diff --git "a/openvino/openvino_model_qint8_quantized.xml" "b/openvino/openvino_model_qint8_quantized.xml" new file mode 100644--- /dev/null +++ "b/openvino/openvino_model_qint8_quantized.xml" @@ -0,0 +1,20864 @@ +<?xml version="1.0"?> +<net name="Model518" version="11"> + <layers> + <layer id="2" name="input_ids" type="Parameter" version="opset1"> + <data shape="?,?" element_type="i64" /> + <output> + <port id="0" precision="I64" names="input_ids"> + <dim>-1</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="1" name="attention_mask" type="Parameter" version="opset1"> + <data shape="?,?" element_type="i64" /> + <output> + <port id="0" precision="I64" names="attention_mask"> + <dim>-1</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="0" name="token_type_ids" type="Parameter" version="opset1"> + <data shape="?,?" element_type="i64" /> + <output> + <port id="0" precision="I64" names="token_type_ids"> + <dim>-1</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="3" name="Constant_578496" type="Const" version="opset1"> + <data element_type="i8" shape="30522, 384" offset="0" size="11720448" /> + <output> + <port id="0" precision="I8"> + <dim>30522</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="4" name="Convert_578497" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>30522</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>30522</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="5" name="Constant_578498" type="Const" version="opset1"> + <data element_type="f32" shape="30522, 1" offset="11720448" size="122088" /> + <output> + <port id="0" precision="FP32"> + <dim>30522</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="6" name="__module.embeddings.word_embeddings/aten::embedding/Gather/fq_weights_0" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>30522</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>30522</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>30522</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="7" name="__module.embeddings.word_embeddings/aten::embedding/Convert" type="Convert" version="opset1"> + <data destination_type="i32" /> + <input> + <port id="0" precision="I64"> + <dim>-1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="1" precision="I32"> + <dim>-1</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="8" name="__module.embeddings.word_embeddings/aten::embedding/Constant" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="11842536" size="4" /> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="9" name="__module.embeddings.word_embeddings/aten::embedding/Gather" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="FP32"> + <dim>30522</dim> + <dim>384</dim> + </port> + <port id="1" precision="I32"> + <dim>-1</dim> + <dim>-1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="FP32" names="79,inputs_embeds"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="10" name="Constant_578500" type="Const" version="opset1"> + <data element_type="i8" shape="2, 384" offset="11842540" size="768" /> + <output> + <port id="0" precision="I8"> + <dim>2</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="11" name="Convert_578501" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>2</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>2</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="12" name="Constant_578502" type="Const" version="opset1"> + <data element_type="f32" shape="2, 1" offset="11843308" size="8" /> + <output> + <port id="0" precision="FP32"> + <dim>2</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="13" name="__module.embeddings.token_type_embeddings/aten::embedding/Gather/fq_weights_0" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>2</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>2</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>2</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="14" name="__module.embeddings.token_type_embeddings/aten::embedding/Convert" type="Convert" version="opset1"> + <data destination_type="i32" /> + <input> + <port id="0" precision="I64"> + <dim>-1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="1" precision="I32"> + <dim>-1</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="15" name="__module.embeddings.token_type_embeddings/aten::embedding/Constant" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="11842536" size="4" /> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="16" name="__module.embeddings.token_type_embeddings/aten::embedding/Gather" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="FP32"> + <dim>2</dim> + <dim>384</dim> + </port> + <port id="1" precision="I32"> + <dim>-1</dim> + <dim>-1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="FP32" names="81,token_type_embeddings.1"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="17" name="__module.embeddings/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="82_1"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="18" name="Constant_578504" type="Const" version="opset1"> + <data element_type="i8" shape="512, 384" offset="11843316" size="196608" /> + <output> + <port id="0" precision="I8"> + <dim>512</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="19" name="Convert_578505" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>512</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>512</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="20" name="Constant_578506" type="Const" version="opset1"> + <data element_type="f32" shape="512, 1" offset="12039924" size="2048" /> + <output> + <port id="0" precision="FP32"> + <dim>512</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="21" name="__module.embeddings.position_embeddings/aten::embedding/Gather/fq_weights_0" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>512</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>512</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>512</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="22" name="__module.embeddings/aten::slice/Slice" type="Const" version="opset1"> + <data element_type="i64" shape="1, 512" offset="12041972" size="4096" /> + <output> + <port id="0" precision="I64" names="76"> + <dim>1</dim> + <dim>512</dim> + </port> + </output> + </layer> + <layer id="23" name="__module.embeddings/aten::slice/Reshape" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="12046068" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="24" name="ShapeOf_131929" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="I64"> + <dim>-1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="25" name="Constant_132057" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="12046076" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="26" name="Constant_131931" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="12046068" size="8" /> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="27" name="Gather_131932" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64" names="10,17,19,72,74,75,8"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="28" name="__module.embeddings/aten::slice/Reshape_2" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="12046076" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="29" name="__module.embeddings/aten::slice/Reshape_3" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="12046076" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="30" name="__module.embeddings/aten::slice/Slice_1" type="Slice" version="opset8"> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + <dim>512</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>1</dim> + </port> + <port id="4" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="5" precision="I64" names="77"> + <dim>1</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="31" name="__module.embeddings.position_embeddings/aten::embedding/Convert" type="Convert" version="opset1"> + <data destination_type="i32" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="1" precision="I32"> + <dim>1</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="32" name="__module.embeddings.position_embeddings/aten::embedding/Constant" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="11842536" size="4" /> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="33" name="__module.embeddings.position_embeddings/aten::embedding/Gather" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="FP32"> + <dim>512</dim> + <dim>384</dim> + </port> + <port id="1" precision="I32"> + <dim>1</dim> + <dim>-1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="FP32" names="84,position_embeddings.1"> + <dim>1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="34" name="__module.embeddings/aten::add_/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="82,embeddings.1"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="35" name="__module.embeddings.LayerNorm/aten::layer_norm/Multiply" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="12046084" size="4" /> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="36" name="__module.embeddings.LayerNorm/aten::layer_norm/MVN" type="MVN" version="opset6"> + <data eps="9.999999960041972e-13" normalize_variance="true" eps_mode="INSIDE_SQRT" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="37" name="Constant_131702" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="12046088" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="38" name="__module.embeddings.LayerNorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="39" name="Constant_131703" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="12047624" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="40" name="__module.embeddings.LayerNorm/aten::layer_norm/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="89,input.1"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="41" name="__module.embeddings.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/scale" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="12049160" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="42" name="__module.embeddings.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="43" name="__module.embeddings.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="12050696" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="44" name="__module.embeddings.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="12050700" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="45" name="__module.embeddings.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="12050696" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="46" name="__module.embeddings.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="12050700" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="47" name="__module.embeddings.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="48" name="Constant_578508" type="Const" version="opset1"> + <data element_type="i8" shape="384, 384" offset="12050704" size="147456" /> + <output> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="49" name="Convert_578509" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="50" name="Constant_578510" type="Const" version="opset1"> + <data element_type="f32" shape="384, 1" offset="12198160" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="51" name="__module.encoder.layer.0.attention.self.query/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="52" name="__module.encoder.layer.0.attention.self.query/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="53" name="Constant_131704" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="12199696" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="54" name="__module.encoder.layer.0.attention.self.query/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="141,x.1"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="55" name="__module.encoder.layer.0.attention.self.query/aten::linear/Add/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="12201232" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="56" name="__module.encoder.layer.0.attention.self.query/aten::linear/Add/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="12201236" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="57" name="__module.encoder.layer.0.attention.self.query/aten::linear/Add/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="12201232" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="58" name="__module.encoder.layer.0.attention.self.query/aten::linear/Add/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="12201236" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="59" name="__module.encoder.layer.0.attention.self.query/aten::linear/Add/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="60" name="__module.encoder.layer.0.attention.self/prim::ListConstruct/Concat" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201240" size="32" /> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="61" name="__module.encoder.layer.0.attention.self/aten::view/Reshape" type="Reshape" version="opset1"> + <data special_zero="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="145,x.3"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="62" name="Constant_122581" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201272" size="32" /> + <output> + <port id="0" precision="I64" names="146"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="63" name="__module.encoder.layer.0.attention.self/aten::permute/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="147"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="64" name="Constant_578512" type="Const" version="opset1"> + <data element_type="i8" shape="384, 384" offset="12201304" size="147456" /> + <output> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="65" name="Convert_578513" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="66" name="Constant_578514" type="Const" version="opset1"> + <data element_type="f32" shape="384, 1" offset="12348760" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="67" name="__module.encoder.layer.0.attention.self.key/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="68" name="__module.encoder.layer.0.attention.self.key/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="69" name="Constant_131705" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="12350296" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="70" name="__module.encoder.layer.0.attention.self.key/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="150,x.5"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="71" name="__module.encoder.layer.0.attention.self.key/aten::linear/Add/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="12351832" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="72" name="__module.encoder.layer.0.attention.self.key/aten::linear/Add/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="12351836" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="73" name="__module.encoder.layer.0.attention.self.key/aten::linear/Add/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="12351832" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="74" name="__module.encoder.layer.0.attention.self.key/aten::linear/Add/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="12351836" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="75" name="__module.encoder.layer.0.attention.self.key/aten::linear/Add/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="76" name="__module.encoder.layer.0.attention.self/prim::ListConstruct/Concat_1" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201240" size="32" /> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="77" name="__module.encoder.layer.0.attention.self/aten::view/Reshape_1" type="Reshape" version="opset1"> + <data special_zero="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="154,x.7"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="78" name="Constant_122606" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201272" size="32" /> + <output> + <port id="0" precision="I64" names="155"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="79" name="__module.encoder.layer.0.attention.self/aten::permute/Transpose_1" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="156"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="80" name="Constant_578516" type="Const" version="opset1"> + <data element_type="i8" shape="384, 384" offset="12351840" size="147456" /> + <output> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="81" name="Convert_578517" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="82" name="Constant_578518" type="Const" version="opset1"> + <data element_type="f32" shape="384, 1" offset="12499296" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="83" name="__module.encoder.layer.0.attention.self.value/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="84" name="__module.encoder.layer.0.attention.self.value/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="85" name="Constant_131706" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="12500832" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="86" name="__module.encoder.layer.0.attention.self.value/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="159,x.9"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="87" name="__module.encoder.layer.0.attention.self/prim::ListConstruct/Concat_2" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201240" size="32" /> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="88" name="__module.encoder.layer.0.attention.self/aten::view/Reshape_2" type="Reshape" version="opset1"> + <data special_zero="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="163,x.11"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="89" name="Constant_122631" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201272" size="32" /> + <output> + <port id="0" precision="I64" names="164"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="90" name="__module.encoder.layer.0.attention.self/aten::permute/Transpose_2" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="165"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="91" name="Constant_131708" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 1, 1" offset="12502368" size="4" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="92" name="25" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="12046076" size="8" /> + <output> + <port id="0" precision="I64" names="25" /> + </output> + </layer> + <layer id="93" name="aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="I64"> + <dim>-1</dim> + <dim>-1</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" names="26"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="94" name="27" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="12502372" size="8" /> + <output> + <port id="0" precision="I64" names="27" /> + </output> + </layer> + <layer id="95" name="aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="I64"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" names="28,33"> + <dim>-1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="96" name="Constant_132060" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="12046068" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="97" name="Constant_131939" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="12046068" size="8" /> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="98" name="Gather_131940" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64" names="13,15"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="99" name="Constant_130326" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="12046076" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="100" name="Constant_132062" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="12502380" size="16" /> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="101" name="Constant_132063" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="12046068" size="8" /> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="102" name="Gather_132064" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="103" name="prim::ListConstruct/Concat" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="35"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="104" name="aten::expand/Broadcast" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="I64"> + <dim>-1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>-1</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="I64" names="37"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="105" name="aten::to/Convert" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I64"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="42"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="106" name="Constant_131707" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 1, 1" offset="12502368" size="4" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="107" name="aten::rsub/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="108" name="aten::rsub/Subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="45,inverted_mask"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="109" name="aten::to/Convert_1" type="Convert" version="opset1"> + <data destination_type="boolean" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="1" precision="BOOL" names="50"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="110" name="aten::masked_fill/ConvertLike" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="12502396" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="111" name="aten::masked_fill/Select" type="Select" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="BOOL"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="3" precision="FP32" names="52"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="112" name="__module.encoder.layer.0.attention.self/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> + <data causal="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="166,attn_output.1"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="113" name="__module.encoder.layer.0.attention.self/aten::transpose/ScatterElementsUpdate" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="12502400" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="114" name="__module.encoder.layer.0.attention.self/aten::transpose/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="167,attn_output.3"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="115" name="Constant_131948" type="Const" version="opset1"> + <data element_type="i64" shape="3" offset="12502416" size="24" /> + <output> + <port id="0" precision="I64"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="116" name="__module.encoder.layer.0.attention.self/aten::reshape/Reshape" type="Reshape" version="opset1"> + <data special_zero="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="169"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="117" name="__module.encoder.layer.0.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/scale" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="12502440" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="118" name="__module.encoder.layer.0.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="119" name="__module.encoder.layer.0.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="12503976" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="120" name="__module.encoder.layer.0.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="12503980" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="121" name="__module.encoder.layer.0.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="12503976" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="122" name="__module.encoder.layer.0.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="12503980" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="123" name="__module.encoder.layer.0.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="124" name="Constant_578520" type="Const" version="opset1"> + <data element_type="i8" shape="384, 384" offset="12503984" size="147456" /> + <output> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="125" name="Convert_578521" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="126" name="Constant_578522" type="Const" version="opset1"> + <data element_type="f32" shape="384, 1" offset="12651440" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="127" name="__module.encoder.layer.0.attention.output.dense/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="128" name="__module.encoder.layer.0.attention.output.dense/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="129" name="Constant_131709" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="12652976" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="130" name="__module.encoder.layer.0.attention.output.dense/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="175,input.3"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="131" name="__module.encoder.layer.0.attention.output/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="177"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="132" name="__module.encoder.layer.0.attention.output.LayerNorm/aten::layer_norm/Multiply" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="12046084" size="4" /> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="133" name="__module.encoder.layer.0.attention.output.LayerNorm/aten::layer_norm/MVN" type="MVN" version="opset6"> + <data eps="9.999999960041972e-13" normalize_variance="true" eps_mode="INSIDE_SQRT" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="134" name="Constant_131710" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="12654512" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="135" name="__module.encoder.layer.0.attention.output.LayerNorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="136" name="Constant_131711" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="12656048" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="137" name="__module.encoder.layer.0.attention.output.LayerNorm/aten::layer_norm/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="181,input_tensor.1"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="138" name="__module.encoder.layer.0.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/scale" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="12657584" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="139" name="__module.encoder.layer.0.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="140" name="__module.encoder.layer.0.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="12659120" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="141" name="__module.encoder.layer.0.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="12659124" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="142" name="__module.encoder.layer.0.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="12659120" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="143" name="__module.encoder.layer.0.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="12659124" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="144" name="__module.encoder.layer.0.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="145" name="Constant_578524" type="Const" version="opset1"> + <data element_type="i8" shape="1536, 384" offset="12659128" size="589824" /> + <output> + <port id="0" precision="I8"> + <dim>1536</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="146" name="Convert_578525" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>1536</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1536</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="147" name="Constant_578526" type="Const" version="opset1"> + <data element_type="f32" shape="1536, 1" offset="13248952" size="6144" /> + <output> + <port id="0" precision="FP32"> + <dim>1536</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="148" name="__module.encoder.layer.0.intermediate.dense/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>1536</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1536</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>1536</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="149" name="__module.encoder.layer.0.intermediate.dense/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1536</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="150" name="Constant_131712" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 1536" offset="13255096" size="6144" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="151" name="__module.encoder.layer.0.intermediate.dense/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1536</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="186"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="152" name="__module.encoder.layer.0.intermediate.intermediate_act_fn/aten::gelu/Gelu" type="Gelu" version="opset7"> + <data approximation_mode="ERF" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="187"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="153" name="__module.encoder.layer.0.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/scale" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 1536" offset="13261240" size="6144" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="154" name="__module.encoder.layer.0.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1536</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="155" name="__module.encoder.layer.0.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="13267384" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="156" name="__module.encoder.layer.0.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="13267388" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="157" name="__module.encoder.layer.0.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="13267384" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="158" name="__module.encoder.layer.0.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="13267388" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="159" name="__module.encoder.layer.0.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="160" name="Constant_578528" type="Const" version="opset1"> + <data element_type="i8" shape="384, 1536" offset="13267392" size="589824" /> + <output> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="161" name="Convert_578529" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>1536</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="162" name="Constant_578530" type="Const" version="opset1"> + <data element_type="f32" shape="384, 1" offset="13857216" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="163" name="__module.encoder.layer.0.output.dense/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1536</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>384</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="164" name="__module.encoder.layer.0.output.dense/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1536</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="165" name="Constant_131713" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="13858752" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="166" name="__module.encoder.layer.0.output.dense/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="193,input.5"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="167" name="__module.encoder.layer.0.output/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="195"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="168" name="__module.encoder.layer.0.output.LayerNorm/aten::layer_norm/Multiply" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="12046084" size="4" /> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="169" name="__module.encoder.layer.0.output.LayerNorm/aten::layer_norm/MVN" type="MVN" version="opset6"> + <data eps="9.999999960041972e-13" normalize_variance="true" eps_mode="INSIDE_SQRT" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="170" name="Constant_131714" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="13860288" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="171" name="__module.encoder.layer.0.output.LayerNorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="172" name="Constant_131715" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="13861824" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="173" name="__module.encoder.layer.0.output.LayerNorm/aten::layer_norm/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="199,hidden_states.7"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="174" name="__module.encoder.layer.0.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/scale" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="13863360" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="175" name="__module.encoder.layer.0.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="176" name="__module.encoder.layer.0.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="13864896" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="177" name="__module.encoder.layer.0.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="13864900" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="178" name="__module.encoder.layer.0.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="13864896" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="179" name="__module.encoder.layer.0.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="13864900" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="180" name="__module.encoder.layer.0.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="181" name="Constant_578532" type="Const" version="opset1"> + <data element_type="i8" shape="384, 384" offset="13864904" size="147456" /> + <output> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="182" name="Convert_578533" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="183" name="Constant_578534" type="Const" version="opset1"> + <data element_type="f32" shape="384, 1" offset="14012360" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="184" name="__module.encoder.layer.1.attention.self.query/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="185" name="__module.encoder.layer.1.attention.self.query/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="186" name="Constant_131716" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="14013896" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="187" name="__module.encoder.layer.1.attention.self.query/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="212,x.13"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="188" name="__module.encoder.layer.1.attention.self.query/aten::linear/Add/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="14015432" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="189" name="__module.encoder.layer.1.attention.self.query/aten::linear/Add/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="14015436" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="190" name="__module.encoder.layer.1.attention.self.query/aten::linear/Add/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="14015432" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="191" name="__module.encoder.layer.1.attention.self.query/aten::linear/Add/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="14015436" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="192" name="__module.encoder.layer.1.attention.self.query/aten::linear/Add/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="193" name="__module.encoder.layer.1.attention.self/prim::ListConstruct/Concat" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201240" size="32" /> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="194" name="__module.encoder.layer.1.attention.self/aten::view/Reshape" type="Reshape" version="opset1"> + <data special_zero="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="216,x.15"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="195" name="Constant_122813" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201272" size="32" /> + <output> + <port id="0" precision="I64" names="217"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="196" name="__module.encoder.layer.1.attention.self/aten::permute/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="218"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="197" name="Constant_578536" type="Const" version="opset1"> + <data element_type="i8" shape="384, 384" offset="14015440" size="147456" /> + <output> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="198" name="Convert_578537" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="199" name="Constant_578538" type="Const" version="opset1"> + <data element_type="f32" shape="384, 1" offset="14162896" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="200" name="__module.encoder.layer.1.attention.self.key/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="201" name="__module.encoder.layer.1.attention.self.key/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="202" name="Constant_131717" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="14164432" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="203" name="__module.encoder.layer.1.attention.self.key/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="221,x.17"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="204" name="__module.encoder.layer.1.attention.self.key/aten::linear/Add/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="14165968" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="205" name="__module.encoder.layer.1.attention.self.key/aten::linear/Add/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="14165972" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="206" name="__module.encoder.layer.1.attention.self.key/aten::linear/Add/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="14165968" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="207" name="__module.encoder.layer.1.attention.self.key/aten::linear/Add/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="14165972" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="208" name="__module.encoder.layer.1.attention.self.key/aten::linear/Add/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="209" name="__module.encoder.layer.1.attention.self/prim::ListConstruct/Concat_1" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201240" size="32" /> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="210" name="__module.encoder.layer.1.attention.self/aten::view/Reshape_1" type="Reshape" version="opset1"> + <data special_zero="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="225,x.19"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="211" name="Constant_122836" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201272" size="32" /> + <output> + <port id="0" precision="I64" names="226"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="212" name="__module.encoder.layer.1.attention.self/aten::permute/Transpose_1" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="227"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="213" name="Constant_578540" type="Const" version="opset1"> + <data element_type="i8" shape="384, 384" offset="14165976" size="147456" /> + <output> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="214" name="Convert_578541" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="215" name="Constant_578542" type="Const" version="opset1"> + <data element_type="f32" shape="384, 1" offset="14313432" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="216" name="__module.encoder.layer.1.attention.self.value/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="217" name="__module.encoder.layer.1.attention.self.value/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="218" name="Constant_131718" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="14314968" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="219" name="__module.encoder.layer.1.attention.self.value/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="230,x.21"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="220" name="__module.encoder.layer.1.attention.self/prim::ListConstruct/Concat_2" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201240" size="32" /> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="221" name="__module.encoder.layer.1.attention.self/aten::view/Reshape_2" type="Reshape" version="opset1"> + <data special_zero="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="234,x.23"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="222" name="Constant_122859" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201272" size="32" /> + <output> + <port id="0" precision="I64" names="235"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="223" name="__module.encoder.layer.1.attention.self/aten::permute/Transpose_2" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="236"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="224" name="__module.encoder.layer.1.attention.self/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> + <data causal="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="237,attn_output.5"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="225" name="__module.encoder.layer.1.attention.self/aten::transpose/ScatterElementsUpdate" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="12502400" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="226" name="__module.encoder.layer.1.attention.self/aten::transpose/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="238,attn_output.7"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="227" name="Constant_131949" type="Const" version="opset1"> + <data element_type="i64" shape="3" offset="12502416" size="24" /> + <output> + <port id="0" precision="I64"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="228" name="__module.encoder.layer.1.attention.self/aten::reshape/Reshape" type="Reshape" version="opset1"> + <data special_zero="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="240"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="229" name="__module.encoder.layer.1.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/scale" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="14316504" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="230" name="__module.encoder.layer.1.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="231" name="__module.encoder.layer.1.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="14318040" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="232" name="__module.encoder.layer.1.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="14318044" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="233" name="__module.encoder.layer.1.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="14318040" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="234" name="__module.encoder.layer.1.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="14318044" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="235" name="__module.encoder.layer.1.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="236" name="Constant_578544" type="Const" version="opset1"> + <data element_type="i8" shape="384, 384" offset="14318048" size="147456" /> + <output> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="237" name="Convert_578545" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="238" name="Constant_578546" type="Const" version="opset1"> + <data element_type="f32" shape="384, 1" offset="14465504" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="239" name="__module.encoder.layer.1.attention.output.dense/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="240" name="__module.encoder.layer.1.attention.output.dense/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="241" name="Constant_131719" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="14467040" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="242" name="__module.encoder.layer.1.attention.output.dense/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="246,input.7"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="243" name="__module.encoder.layer.1.attention.output/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="248"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="244" name="__module.encoder.layer.1.attention.output.LayerNorm/aten::layer_norm/Multiply" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="12046084" size="4" /> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="245" name="__module.encoder.layer.1.attention.output.LayerNorm/aten::layer_norm/MVN" type="MVN" version="opset6"> + <data eps="9.999999960041972e-13" normalize_variance="true" eps_mode="INSIDE_SQRT" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="246" name="Constant_131720" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="14468576" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="247" name="__module.encoder.layer.1.attention.output.LayerNorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="248" name="Constant_131721" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="14470112" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="249" name="__module.encoder.layer.1.attention.output.LayerNorm/aten::layer_norm/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="252,input_tensor.3"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="250" name="__module.encoder.layer.1.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/scale" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="14471648" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="251" name="__module.encoder.layer.1.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="252" name="__module.encoder.layer.1.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="14473184" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="253" name="__module.encoder.layer.1.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="14473188" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="254" name="__module.encoder.layer.1.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="14473184" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="255" name="__module.encoder.layer.1.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="14473188" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="256" name="__module.encoder.layer.1.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="257" name="Constant_578548" type="Const" version="opset1"> + <data element_type="i8" shape="1536, 384" offset="14473192" size="589824" /> + <output> + <port id="0" precision="I8"> + <dim>1536</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="258" name="Convert_578549" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>1536</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1536</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="259" name="Constant_578550" type="Const" version="opset1"> + <data element_type="f32" shape="1536, 1" offset="15063016" size="6144" /> + <output> + <port id="0" precision="FP32"> + <dim>1536</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="260" name="__module.encoder.layer.1.intermediate.dense/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>1536</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1536</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>1536</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="261" name="__module.encoder.layer.1.intermediate.dense/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1536</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="262" name="Constant_131722" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 1536" offset="15069160" size="6144" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="263" name="__module.encoder.layer.1.intermediate.dense/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1536</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="257"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="264" name="__module.encoder.layer.1.intermediate.intermediate_act_fn/aten::gelu/Gelu" type="Gelu" version="opset7"> + <data approximation_mode="ERF" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="258"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="265" name="__module.encoder.layer.1.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/scale" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 1536" offset="15075304" size="6144" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="266" name="__module.encoder.layer.1.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1536</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="267" name="__module.encoder.layer.1.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="15081448" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="268" name="__module.encoder.layer.1.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="15081452" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="269" name="__module.encoder.layer.1.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="15081448" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="270" name="__module.encoder.layer.1.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="15081452" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="271" name="__module.encoder.layer.1.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="272" name="Constant_578552" type="Const" version="opset1"> + <data element_type="i8" shape="384, 1536" offset="15081456" size="589824" /> + <output> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="273" name="Convert_578553" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>1536</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="274" name="Constant_578554" type="Const" version="opset1"> + <data element_type="f32" shape="384, 1" offset="15671280" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="275" name="__module.encoder.layer.1.output.dense/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1536</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>384</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="276" name="__module.encoder.layer.1.output.dense/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1536</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="277" name="Constant_131723" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="15672816" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="278" name="__module.encoder.layer.1.output.dense/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="264,input.9"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="279" name="__module.encoder.layer.1.output/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="266"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="280" name="__module.encoder.layer.1.output.LayerNorm/aten::layer_norm/Multiply" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="12046084" size="4" /> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="281" name="__module.encoder.layer.1.output.LayerNorm/aten::layer_norm/MVN" type="MVN" version="opset6"> + <data eps="9.999999960041972e-13" normalize_variance="true" eps_mode="INSIDE_SQRT" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="282" name="Constant_131724" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="15674352" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="283" name="__module.encoder.layer.1.output.LayerNorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="284" name="Constant_131725" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="15675888" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="285" name="__module.encoder.layer.1.output.LayerNorm/aten::layer_norm/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="270,hidden_states.13"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="286" name="__module.encoder.layer.1.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/scale" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="15677424" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="287" name="__module.encoder.layer.1.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="288" name="__module.encoder.layer.1.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="15678960" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="289" name="__module.encoder.layer.1.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="15678964" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="290" name="__module.encoder.layer.1.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="15678960" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="291" name="__module.encoder.layer.1.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="15678964" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="292" name="__module.encoder.layer.1.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="293" name="Constant_578556" type="Const" version="opset1"> + <data element_type="i8" shape="384, 384" offset="15678968" size="147456" /> + <output> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="294" name="Convert_578557" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="295" name="Constant_578558" type="Const" version="opset1"> + <data element_type="f32" shape="384, 1" offset="15826424" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="296" name="__module.encoder.layer.2.attention.self.query/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="297" name="__module.encoder.layer.2.attention.self.query/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="298" name="Constant_131726" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="15827960" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="299" name="__module.encoder.layer.2.attention.self.query/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="283,x.25"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="300" name="__module.encoder.layer.2.attention.self.query/aten::linear/Add/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="15829496" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="301" name="__module.encoder.layer.2.attention.self.query/aten::linear/Add/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="15829500" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="302" name="__module.encoder.layer.2.attention.self.query/aten::linear/Add/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="15829496" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="303" name="__module.encoder.layer.2.attention.self.query/aten::linear/Add/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="15829500" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="304" name="__module.encoder.layer.2.attention.self.query/aten::linear/Add/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="305" name="__module.encoder.layer.2.attention.self/prim::ListConstruct/Concat" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201240" size="32" /> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="306" name="__module.encoder.layer.2.attention.self/aten::view/Reshape" type="Reshape" version="opset1"> + <data special_zero="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="287,x.27"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="307" name="Constant_123039" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201272" size="32" /> + <output> + <port id="0" precision="I64" names="288"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="308" name="__module.encoder.layer.2.attention.self/aten::permute/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="289"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="309" name="Constant_578560" type="Const" version="opset1"> + <data element_type="i8" shape="384, 384" offset="15829504" size="147456" /> + <output> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="310" name="Convert_578561" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="311" name="Constant_578562" type="Const" version="opset1"> + <data element_type="f32" shape="384, 1" offset="15976960" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="312" name="__module.encoder.layer.2.attention.self.key/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="313" name="__module.encoder.layer.2.attention.self.key/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="314" name="Constant_131727" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="15978496" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="315" name="__module.encoder.layer.2.attention.self.key/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="292,x.29"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="316" name="__module.encoder.layer.2.attention.self.key/aten::linear/Add/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="15980032" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="317" name="__module.encoder.layer.2.attention.self.key/aten::linear/Add/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="15980036" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="318" name="__module.encoder.layer.2.attention.self.key/aten::linear/Add/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="15980032" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="319" name="__module.encoder.layer.2.attention.self.key/aten::linear/Add/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="15980036" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="320" name="__module.encoder.layer.2.attention.self.key/aten::linear/Add/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="321" name="__module.encoder.layer.2.attention.self/prim::ListConstruct/Concat_1" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201240" size="32" /> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="322" name="__module.encoder.layer.2.attention.self/aten::view/Reshape_1" type="Reshape" version="opset1"> + <data special_zero="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="296,x.31"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="323" name="Constant_123062" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201272" size="32" /> + <output> + <port id="0" precision="I64" names="297"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="324" name="__module.encoder.layer.2.attention.self/aten::permute/Transpose_1" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="298"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="325" name="Constant_578564" type="Const" version="opset1"> + <data element_type="i8" shape="384, 384" offset="15980040" size="147456" /> + <output> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="326" name="Convert_578565" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="327" name="Constant_578566" type="Const" version="opset1"> + <data element_type="f32" shape="384, 1" offset="16127496" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="328" name="__module.encoder.layer.2.attention.self.value/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="329" name="__module.encoder.layer.2.attention.self.value/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="330" name="Constant_131728" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="16129032" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="331" name="__module.encoder.layer.2.attention.self.value/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="301,x.33"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="332" name="__module.encoder.layer.2.attention.self/prim::ListConstruct/Concat_2" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201240" size="32" /> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="333" name="__module.encoder.layer.2.attention.self/aten::view/Reshape_2" type="Reshape" version="opset1"> + <data special_zero="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="305,x.35"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="334" name="Constant_123085" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201272" size="32" /> + <output> + <port id="0" precision="I64" names="306"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="335" name="__module.encoder.layer.2.attention.self/aten::permute/Transpose_2" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="307"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="336" name="__module.encoder.layer.2.attention.self/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> + <data causal="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="308,attn_output.9"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="337" name="__module.encoder.layer.2.attention.self/aten::transpose/ScatterElementsUpdate" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="12502400" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="338" name="__module.encoder.layer.2.attention.self/aten::transpose/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="309,attn_output.11"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="339" name="Constant_131950" type="Const" version="opset1"> + <data element_type="i64" shape="3" offset="12502416" size="24" /> + <output> + <port id="0" precision="I64"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="340" name="__module.encoder.layer.2.attention.self/aten::reshape/Reshape" type="Reshape" version="opset1"> + <data special_zero="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="311"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="341" name="__module.encoder.layer.2.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/scale" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="16130568" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="342" name="__module.encoder.layer.2.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="343" name="__module.encoder.layer.2.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="16132104" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="344" name="__module.encoder.layer.2.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="16132108" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="345" name="__module.encoder.layer.2.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="16132104" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="346" name="__module.encoder.layer.2.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="16132108" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="347" name="__module.encoder.layer.2.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="348" name="Constant_578568" type="Const" version="opset1"> + <data element_type="i8" shape="384, 384" offset="16132112" size="147456" /> + <output> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="349" name="Convert_578569" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="350" name="Constant_578570" type="Const" version="opset1"> + <data element_type="f32" shape="384, 1" offset="16279568" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="351" name="__module.encoder.layer.2.attention.output.dense/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="352" name="__module.encoder.layer.2.attention.output.dense/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="353" name="Constant_131729" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="16281104" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="354" name="__module.encoder.layer.2.attention.output.dense/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="317,input.11"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="355" name="__module.encoder.layer.2.attention.output/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="319"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="356" name="__module.encoder.layer.2.attention.output.LayerNorm/aten::layer_norm/Multiply" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="12046084" size="4" /> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="357" name="__module.encoder.layer.2.attention.output.LayerNorm/aten::layer_norm/MVN" type="MVN" version="opset6"> + <data eps="9.999999960041972e-13" normalize_variance="true" eps_mode="INSIDE_SQRT" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="358" name="Constant_131730" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="16282640" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="359" name="__module.encoder.layer.2.attention.output.LayerNorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="360" name="Constant_131731" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="16284176" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="361" name="__module.encoder.layer.2.attention.output.LayerNorm/aten::layer_norm/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="323,input_tensor.5"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="362" name="__module.encoder.layer.2.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/scale" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="16285712" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="363" name="__module.encoder.layer.2.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="364" name="__module.encoder.layer.2.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="16287248" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="365" name="__module.encoder.layer.2.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="16287252" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="366" name="__module.encoder.layer.2.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="16287248" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="367" name="__module.encoder.layer.2.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="16287252" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="368" name="__module.encoder.layer.2.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="369" name="Constant_578572" type="Const" version="opset1"> + <data element_type="i8" shape="1536, 384" offset="16287256" size="589824" /> + <output> + <port id="0" precision="I8"> + <dim>1536</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="370" name="Convert_578573" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>1536</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1536</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="371" name="Constant_578574" type="Const" version="opset1"> + <data element_type="f32" shape="1536, 1" offset="16877080" size="6144" /> + <output> + <port id="0" precision="FP32"> + <dim>1536</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="372" name="__module.encoder.layer.2.intermediate.dense/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>1536</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1536</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>1536</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="373" name="__module.encoder.layer.2.intermediate.dense/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1536</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="374" name="Constant_131732" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 1536" offset="16883224" size="6144" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="375" name="__module.encoder.layer.2.intermediate.dense/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1536</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="328"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="376" name="__module.encoder.layer.2.intermediate.intermediate_act_fn/aten::gelu/Gelu" type="Gelu" version="opset7"> + <data approximation_mode="ERF" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="329"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="377" name="__module.encoder.layer.2.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/scale" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 1536" offset="16889368" size="6144" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="378" name="__module.encoder.layer.2.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1536</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="379" name="__module.encoder.layer.2.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="16895512" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="380" name="__module.encoder.layer.2.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="16895516" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="381" name="__module.encoder.layer.2.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="16895512" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="382" name="__module.encoder.layer.2.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="16895516" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="383" name="__module.encoder.layer.2.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="384" name="Constant_578576" type="Const" version="opset1"> + <data element_type="i8" shape="384, 1536" offset="16895520" size="589824" /> + <output> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="385" name="Convert_578577" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>1536</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="386" name="Constant_578578" type="Const" version="opset1"> + <data element_type="f32" shape="384, 1" offset="17485344" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="387" name="__module.encoder.layer.2.output.dense/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1536</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>384</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="388" name="__module.encoder.layer.2.output.dense/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1536</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="389" name="Constant_131733" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="17486880" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="390" name="__module.encoder.layer.2.output.dense/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="335,input.13"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="391" name="__module.encoder.layer.2.output/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="337"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="392" name="__module.encoder.layer.2.output.LayerNorm/aten::layer_norm/Multiply" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="12046084" size="4" /> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="393" name="__module.encoder.layer.2.output.LayerNorm/aten::layer_norm/MVN" type="MVN" version="opset6"> + <data eps="9.999999960041972e-13" normalize_variance="true" eps_mode="INSIDE_SQRT" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="394" name="Constant_131734" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="17488416" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="395" name="__module.encoder.layer.2.output.LayerNorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="396" name="Constant_131735" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="17489952" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="397" name="__module.encoder.layer.2.output.LayerNorm/aten::layer_norm/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="341,hidden_states.19"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="398" name="__module.encoder.layer.2.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/scale" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="17491488" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="399" name="__module.encoder.layer.2.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="400" name="__module.encoder.layer.2.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="17493024" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="401" name="__module.encoder.layer.2.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="17493028" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="402" name="__module.encoder.layer.2.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="17493024" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="403" name="__module.encoder.layer.2.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="17493028" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="404" name="__module.encoder.layer.2.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="405" name="Constant_578580" type="Const" version="opset1"> + <data element_type="i8" shape="384, 384" offset="17493032" size="147456" /> + <output> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="406" name="Convert_578581" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="407" name="Constant_578582" type="Const" version="opset1"> + <data element_type="f32" shape="384, 1" offset="17640488" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="408" name="__module.encoder.layer.3.attention.self.query/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="409" name="__module.encoder.layer.3.attention.self.query/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="410" name="Constant_131736" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="17642024" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="411" name="__module.encoder.layer.3.attention.self.query/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="354,x.37"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="412" name="__module.encoder.layer.3.attention.self.query/aten::linear/Add/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="17643560" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="413" name="__module.encoder.layer.3.attention.self.query/aten::linear/Add/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="17643564" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="414" name="__module.encoder.layer.3.attention.self.query/aten::linear/Add/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="17643560" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="415" name="__module.encoder.layer.3.attention.self.query/aten::linear/Add/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="17643564" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="416" name="__module.encoder.layer.3.attention.self.query/aten::linear/Add/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="417" name="__module.encoder.layer.3.attention.self/prim::ListConstruct/Concat" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201240" size="32" /> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="418" name="__module.encoder.layer.3.attention.self/aten::view/Reshape" type="Reshape" version="opset1"> + <data special_zero="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="358,x.39"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="419" name="Constant_123265" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201272" size="32" /> + <output> + <port id="0" precision="I64" names="359"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="420" name="__module.encoder.layer.3.attention.self/aten::permute/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="360"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="421" name="Constant_578584" type="Const" version="opset1"> + <data element_type="i8" shape="384, 384" offset="17643568" size="147456" /> + <output> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="422" name="Convert_578585" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="423" name="Constant_578586" type="Const" version="opset1"> + <data element_type="f32" shape="384, 1" offset="17791024" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="424" name="__module.encoder.layer.3.attention.self.key/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="425" name="__module.encoder.layer.3.attention.self.key/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="426" name="Constant_131737" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="17792560" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="427" name="__module.encoder.layer.3.attention.self.key/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="363,x.41"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="428" name="__module.encoder.layer.3.attention.self.key/aten::linear/Add/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="17794096" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="429" name="__module.encoder.layer.3.attention.self.key/aten::linear/Add/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="17794100" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="430" name="__module.encoder.layer.3.attention.self.key/aten::linear/Add/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="17794096" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="431" name="__module.encoder.layer.3.attention.self.key/aten::linear/Add/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="17794100" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="432" name="__module.encoder.layer.3.attention.self.key/aten::linear/Add/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="433" name="__module.encoder.layer.3.attention.self/prim::ListConstruct/Concat_1" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201240" size="32" /> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="434" name="__module.encoder.layer.3.attention.self/aten::view/Reshape_1" type="Reshape" version="opset1"> + <data special_zero="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="367,x.43"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="435" name="Constant_123288" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201272" size="32" /> + <output> + <port id="0" precision="I64" names="368"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="436" name="__module.encoder.layer.3.attention.self/aten::permute/Transpose_1" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="369"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="437" name="Constant_578588" type="Const" version="opset1"> + <data element_type="i8" shape="384, 384" offset="17794104" size="147456" /> + <output> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="438" name="Convert_578589" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="439" name="Constant_578590" type="Const" version="opset1"> + <data element_type="f32" shape="384, 1" offset="17941560" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="440" name="__module.encoder.layer.3.attention.self.value/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="441" name="__module.encoder.layer.3.attention.self.value/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="442" name="Constant_131738" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="17943096" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="443" name="__module.encoder.layer.3.attention.self.value/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="372,x.45"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="444" name="__module.encoder.layer.3.attention.self/prim::ListConstruct/Concat_2" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201240" size="32" /> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="445" name="__module.encoder.layer.3.attention.self/aten::view/Reshape_2" type="Reshape" version="opset1"> + <data special_zero="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="376,x.47"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="446" name="Constant_123311" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201272" size="32" /> + <output> + <port id="0" precision="I64" names="377"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="447" name="__module.encoder.layer.3.attention.self/aten::permute/Transpose_2" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="378"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="448" name="__module.encoder.layer.3.attention.self/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> + <data causal="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="379,attn_output.13"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="449" name="__module.encoder.layer.3.attention.self/aten::transpose/ScatterElementsUpdate" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="12502400" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="450" name="__module.encoder.layer.3.attention.self/aten::transpose/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="380,attn_output.15"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="451" name="Constant_131951" type="Const" version="opset1"> + <data element_type="i64" shape="3" offset="12502416" size="24" /> + <output> + <port id="0" precision="I64"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="452" name="__module.encoder.layer.3.attention.self/aten::reshape/Reshape" type="Reshape" version="opset1"> + <data special_zero="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="382"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="453" name="__module.encoder.layer.3.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/scale" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="17944632" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="454" name="__module.encoder.layer.3.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="455" name="__module.encoder.layer.3.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="17946168" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="456" name="__module.encoder.layer.3.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="17946172" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="457" name="__module.encoder.layer.3.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="17946168" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="458" name="__module.encoder.layer.3.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="17946172" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="459" name="__module.encoder.layer.3.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="460" name="Constant_578592" type="Const" version="opset1"> + <data element_type="i8" shape="384, 384" offset="17946176" size="147456" /> + <output> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="461" name="Convert_578593" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="462" name="Constant_578594" type="Const" version="opset1"> + <data element_type="f32" shape="384, 1" offset="18093632" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="463" name="__module.encoder.layer.3.attention.output.dense/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="464" name="__module.encoder.layer.3.attention.output.dense/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="465" name="Constant_131739" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="18095168" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="466" name="__module.encoder.layer.3.attention.output.dense/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="388,input.15"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="467" name="__module.encoder.layer.3.attention.output/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="390"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="468" name="__module.encoder.layer.3.attention.output.LayerNorm/aten::layer_norm/Multiply" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="12046084" size="4" /> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="469" name="__module.encoder.layer.3.attention.output.LayerNorm/aten::layer_norm/MVN" type="MVN" version="opset6"> + <data eps="9.999999960041972e-13" normalize_variance="true" eps_mode="INSIDE_SQRT" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="470" name="Constant_131740" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="18096704" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="471" name="__module.encoder.layer.3.attention.output.LayerNorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="472" name="Constant_131741" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="18098240" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="473" name="__module.encoder.layer.3.attention.output.LayerNorm/aten::layer_norm/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="394,input_tensor.7"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="474" name="__module.encoder.layer.3.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/scale" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="18099776" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="475" name="__module.encoder.layer.3.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="476" name="__module.encoder.layer.3.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="18101312" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="477" name="__module.encoder.layer.3.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="18101316" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="478" name="__module.encoder.layer.3.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="18101312" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="479" name="__module.encoder.layer.3.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="18101316" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="480" name="__module.encoder.layer.3.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="481" name="Constant_578596" type="Const" version="opset1"> + <data element_type="i8" shape="1536, 384" offset="18101320" size="589824" /> + <output> + <port id="0" precision="I8"> + <dim>1536</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="482" name="Convert_578597" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>1536</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1536</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="483" name="Constant_578598" type="Const" version="opset1"> + <data element_type="f32" shape="1536, 1" offset="18691144" size="6144" /> + <output> + <port id="0" precision="FP32"> + <dim>1536</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="484" name="__module.encoder.layer.3.intermediate.dense/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>1536</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1536</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>1536</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="485" name="__module.encoder.layer.3.intermediate.dense/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1536</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="486" name="Constant_131742" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 1536" offset="18697288" size="6144" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="487" name="__module.encoder.layer.3.intermediate.dense/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1536</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="399"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="488" name="__module.encoder.layer.3.intermediate.intermediate_act_fn/aten::gelu/Gelu" type="Gelu" version="opset7"> + <data approximation_mode="ERF" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="400"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="489" name="__module.encoder.layer.3.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/scale" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 1536" offset="18703432" size="6144" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="490" name="__module.encoder.layer.3.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1536</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="491" name="__module.encoder.layer.3.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="18709576" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="492" name="__module.encoder.layer.3.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="18709580" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="493" name="__module.encoder.layer.3.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="18709576" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="494" name="__module.encoder.layer.3.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="18709580" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="495" name="__module.encoder.layer.3.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="496" name="Constant_578600" type="Const" version="opset1"> + <data element_type="i8" shape="384, 1536" offset="18709584" size="589824" /> + <output> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="497" name="Convert_578601" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>1536</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="498" name="Constant_578602" type="Const" version="opset1"> + <data element_type="f32" shape="384, 1" offset="19299408" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="499" name="__module.encoder.layer.3.output.dense/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1536</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>384</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="500" name="__module.encoder.layer.3.output.dense/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1536</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="501" name="Constant_131743" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="19300944" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="502" name="__module.encoder.layer.3.output.dense/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="406,input.17"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="503" name="__module.encoder.layer.3.output/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="408"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="504" name="__module.encoder.layer.3.output.LayerNorm/aten::layer_norm/Multiply" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="12046084" size="4" /> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="505" name="__module.encoder.layer.3.output.LayerNorm/aten::layer_norm/MVN" type="MVN" version="opset6"> + <data eps="9.999999960041972e-13" normalize_variance="true" eps_mode="INSIDE_SQRT" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="506" name="Constant_131744" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="19302480" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="507" name="__module.encoder.layer.3.output.LayerNorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="508" name="Constant_131745" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="19304016" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="509" name="__module.encoder.layer.3.output.LayerNorm/aten::layer_norm/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="412,hidden_states.25"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="510" name="__module.encoder.layer.3.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/scale" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="19305552" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="511" name="__module.encoder.layer.3.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="512" name="__module.encoder.layer.3.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="19307088" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="513" name="__module.encoder.layer.3.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="19307092" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="514" name="__module.encoder.layer.3.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="19307088" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="515" name="__module.encoder.layer.3.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="19307092" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="516" name="__module.encoder.layer.3.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="517" name="Constant_578604" type="Const" version="opset1"> + <data element_type="i8" shape="384, 384" offset="19307096" size="147456" /> + <output> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="518" name="Convert_578605" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="519" name="Constant_578606" type="Const" version="opset1"> + <data element_type="f32" shape="384, 1" offset="19454552" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="520" name="__module.encoder.layer.4.attention.self.query/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="521" name="__module.encoder.layer.4.attention.self.query/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="522" name="Constant_131746" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="19456088" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="523" name="__module.encoder.layer.4.attention.self.query/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="425,x.49"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="524" name="__module.encoder.layer.4.attention.self.query/aten::linear/Add/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="19457624" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="525" name="__module.encoder.layer.4.attention.self.query/aten::linear/Add/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="19457628" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="526" name="__module.encoder.layer.4.attention.self.query/aten::linear/Add/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="19457624" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="527" name="__module.encoder.layer.4.attention.self.query/aten::linear/Add/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="19457628" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="528" name="__module.encoder.layer.4.attention.self.query/aten::linear/Add/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="529" name="__module.encoder.layer.4.attention.self/prim::ListConstruct/Concat" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201240" size="32" /> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="530" name="__module.encoder.layer.4.attention.self/aten::view/Reshape" type="Reshape" version="opset1"> + <data special_zero="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="429,x.51"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="531" name="Constant_123491" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201272" size="32" /> + <output> + <port id="0" precision="I64" names="430"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="532" name="__module.encoder.layer.4.attention.self/aten::permute/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="431"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="533" name="Constant_578608" type="Const" version="opset1"> + <data element_type="i8" shape="384, 384" offset="19457632" size="147456" /> + <output> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="534" name="Convert_578609" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="535" name="Constant_578610" type="Const" version="opset1"> + <data element_type="f32" shape="384, 1" offset="19605088" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="536" name="__module.encoder.layer.4.attention.self.key/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="537" name="__module.encoder.layer.4.attention.self.key/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="538" name="Constant_131747" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="19606624" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="539" name="__module.encoder.layer.4.attention.self.key/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="434,x.53"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="540" name="__module.encoder.layer.4.attention.self.key/aten::linear/Add/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="19608160" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="541" name="__module.encoder.layer.4.attention.self.key/aten::linear/Add/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="19608164" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="542" name="__module.encoder.layer.4.attention.self.key/aten::linear/Add/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="19608160" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="543" name="__module.encoder.layer.4.attention.self.key/aten::linear/Add/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="19608164" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="544" name="__module.encoder.layer.4.attention.self.key/aten::linear/Add/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="545" name="__module.encoder.layer.4.attention.self/prim::ListConstruct/Concat_1" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201240" size="32" /> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="546" name="__module.encoder.layer.4.attention.self/aten::view/Reshape_1" type="Reshape" version="opset1"> + <data special_zero="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="438,x.55"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="547" name="Constant_123514" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201272" size="32" /> + <output> + <port id="0" precision="I64" names="439"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="548" name="__module.encoder.layer.4.attention.self/aten::permute/Transpose_1" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="440"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="549" name="Constant_578612" type="Const" version="opset1"> + <data element_type="i8" shape="384, 384" offset="19608168" size="147456" /> + <output> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="550" name="Convert_578613" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="551" name="Constant_578614" type="Const" version="opset1"> + <data element_type="f32" shape="384, 1" offset="19755624" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="552" name="__module.encoder.layer.4.attention.self.value/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="553" name="__module.encoder.layer.4.attention.self.value/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="554" name="Constant_131748" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="19757160" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="555" name="__module.encoder.layer.4.attention.self.value/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="443,x.57"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="556" name="__module.encoder.layer.4.attention.self/prim::ListConstruct/Concat_2" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201240" size="32" /> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="557" name="__module.encoder.layer.4.attention.self/aten::view/Reshape_2" type="Reshape" version="opset1"> + <data special_zero="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="447,x.59"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="558" name="Constant_123537" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201272" size="32" /> + <output> + <port id="0" precision="I64" names="448"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="559" name="__module.encoder.layer.4.attention.self/aten::permute/Transpose_2" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="449"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="560" name="__module.encoder.layer.4.attention.self/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> + <data causal="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="450,attn_output.17"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="561" name="__module.encoder.layer.4.attention.self/aten::transpose/ScatterElementsUpdate" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="12502400" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="562" name="__module.encoder.layer.4.attention.self/aten::transpose/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="451,attn_output.19"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="563" name="Constant_131952" type="Const" version="opset1"> + <data element_type="i64" shape="3" offset="12502416" size="24" /> + <output> + <port id="0" precision="I64"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="564" name="__module.encoder.layer.4.attention.self/aten::reshape/Reshape" type="Reshape" version="opset1"> + <data special_zero="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="453"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="565" name="__module.encoder.layer.4.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/scale" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="19758696" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="566" name="__module.encoder.layer.4.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="567" name="__module.encoder.layer.4.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="19760232" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="568" name="__module.encoder.layer.4.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="19760236" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="569" name="__module.encoder.layer.4.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="19760232" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="570" name="__module.encoder.layer.4.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="19760236" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="571" name="__module.encoder.layer.4.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="572" name="Constant_578616" type="Const" version="opset1"> + <data element_type="i8" shape="384, 384" offset="19760240" size="147456" /> + <output> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="573" name="Convert_578617" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="574" name="Constant_578618" type="Const" version="opset1"> + <data element_type="f32" shape="384, 1" offset="19907696" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="575" name="__module.encoder.layer.4.attention.output.dense/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="576" name="__module.encoder.layer.4.attention.output.dense/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="577" name="Constant_131749" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="19909232" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="578" name="__module.encoder.layer.4.attention.output.dense/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="459,input.19"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="579" name="__module.encoder.layer.4.attention.output/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="461"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="580" name="__module.encoder.layer.4.attention.output.LayerNorm/aten::layer_norm/Multiply" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="12046084" size="4" /> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="581" name="__module.encoder.layer.4.attention.output.LayerNorm/aten::layer_norm/MVN" type="MVN" version="opset6"> + <data eps="9.999999960041972e-13" normalize_variance="true" eps_mode="INSIDE_SQRT" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="582" name="Constant_131750" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="19910768" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="583" name="__module.encoder.layer.4.attention.output.LayerNorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="584" name="Constant_131751" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="19912304" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="585" name="__module.encoder.layer.4.attention.output.LayerNorm/aten::layer_norm/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="465,input_tensor.9"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="586" name="__module.encoder.layer.4.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/scale" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="19913840" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="587" name="__module.encoder.layer.4.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="588" name="__module.encoder.layer.4.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="19915376" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="589" name="__module.encoder.layer.4.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="19915380" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="590" name="__module.encoder.layer.4.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="19915376" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="591" name="__module.encoder.layer.4.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="19915380" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="592" name="__module.encoder.layer.4.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="593" name="Constant_578620" type="Const" version="opset1"> + <data element_type="i8" shape="1536, 384" offset="19915384" size="589824" /> + <output> + <port id="0" precision="I8"> + <dim>1536</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="594" name="Convert_578621" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>1536</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1536</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="595" name="Constant_578622" type="Const" version="opset1"> + <data element_type="f32" shape="1536, 1" offset="20505208" size="6144" /> + <output> + <port id="0" precision="FP32"> + <dim>1536</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="596" name="__module.encoder.layer.4.intermediate.dense/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>1536</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1536</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>1536</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="597" name="__module.encoder.layer.4.intermediate.dense/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1536</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="598" name="Constant_131752" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 1536" offset="20511352" size="6144" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="599" name="__module.encoder.layer.4.intermediate.dense/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1536</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="470"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="600" name="__module.encoder.layer.4.intermediate.intermediate_act_fn/aten::gelu/Gelu" type="Gelu" version="opset7"> + <data approximation_mode="ERF" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="471"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="601" name="__module.encoder.layer.4.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/scale" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 1536" offset="20517496" size="6144" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="602" name="__module.encoder.layer.4.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1536</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="603" name="__module.encoder.layer.4.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="20523640" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="604" name="__module.encoder.layer.4.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="20523644" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="605" name="__module.encoder.layer.4.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="20523640" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="606" name="__module.encoder.layer.4.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="20523644" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="607" name="__module.encoder.layer.4.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="608" name="Constant_578624" type="Const" version="opset1"> + <data element_type="i8" shape="384, 1536" offset="20523648" size="589824" /> + <output> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="609" name="Convert_578625" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>1536</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="610" name="Constant_578626" type="Const" version="opset1"> + <data element_type="f32" shape="384, 1" offset="21113472" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="611" name="__module.encoder.layer.4.output.dense/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1536</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>384</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="612" name="__module.encoder.layer.4.output.dense/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1536</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="613" name="Constant_131753" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="21115008" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="614" name="__module.encoder.layer.4.output.dense/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="477,input.21"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="615" name="__module.encoder.layer.4.output/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="479"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="616" name="__module.encoder.layer.4.output.LayerNorm/aten::layer_norm/Multiply" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="12046084" size="4" /> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="617" name="__module.encoder.layer.4.output.LayerNorm/aten::layer_norm/MVN" type="MVN" version="opset6"> + <data eps="9.999999960041972e-13" normalize_variance="true" eps_mode="INSIDE_SQRT" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="618" name="Constant_131754" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="21116544" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="619" name="__module.encoder.layer.4.output.LayerNorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="620" name="Constant_131755" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="21118080" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="621" name="__module.encoder.layer.4.output.LayerNorm/aten::layer_norm/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="483,hidden_states.31"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="622" name="__module.encoder.layer.4.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/scale" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="21119616" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="623" name="__module.encoder.layer.4.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="624" name="__module.encoder.layer.4.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="21121152" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="625" name="__module.encoder.layer.4.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="21121156" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="626" name="__module.encoder.layer.4.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="21121152" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="627" name="__module.encoder.layer.4.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="21121156" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="628" name="__module.encoder.layer.4.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="629" name="Constant_578628" type="Const" version="opset1"> + <data element_type="i8" shape="384, 384" offset="21121160" size="147456" /> + <output> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="630" name="Convert_578629" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="631" name="Constant_578630" type="Const" version="opset1"> + <data element_type="f32" shape="384, 1" offset="21268616" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="632" name="__module.encoder.layer.5.attention.self.query/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="633" name="__module.encoder.layer.5.attention.self.query/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="634" name="Constant_131756" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="21270152" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="635" name="__module.encoder.layer.5.attention.self.query/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="496,x.61"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="636" name="__module.encoder.layer.5.attention.self.query/aten::linear/Add/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="21271688" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="637" name="__module.encoder.layer.5.attention.self.query/aten::linear/Add/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="21271692" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="638" name="__module.encoder.layer.5.attention.self.query/aten::linear/Add/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="21271688" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="639" name="__module.encoder.layer.5.attention.self.query/aten::linear/Add/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="21271692" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="640" name="__module.encoder.layer.5.attention.self.query/aten::linear/Add/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="641" name="__module.encoder.layer.5.attention.self/prim::ListConstruct/Concat" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201240" size="32" /> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="642" name="__module.encoder.layer.5.attention.self/aten::view/Reshape" type="Reshape" version="opset1"> + <data special_zero="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="500,x.63"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="643" name="Constant_123717" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201272" size="32" /> + <output> + <port id="0" precision="I64" names="501"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="644" name="__module.encoder.layer.5.attention.self/aten::permute/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="502"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="645" name="Constant_578632" type="Const" version="opset1"> + <data element_type="i8" shape="384, 384" offset="21271696" size="147456" /> + <output> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="646" name="Convert_578633" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="647" name="Constant_578634" type="Const" version="opset1"> + <data element_type="f32" shape="384, 1" offset="21419152" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="648" name="__module.encoder.layer.5.attention.self.key/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="649" name="__module.encoder.layer.5.attention.self.key/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="650" name="Constant_131757" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="21420688" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="651" name="__module.encoder.layer.5.attention.self.key/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="505,x.65"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="652" name="__module.encoder.layer.5.attention.self.key/aten::linear/Add/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="21422224" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="653" name="__module.encoder.layer.5.attention.self.key/aten::linear/Add/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="21422228" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="654" name="__module.encoder.layer.5.attention.self.key/aten::linear/Add/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="21422224" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="655" name="__module.encoder.layer.5.attention.self.key/aten::linear/Add/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="21422228" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="656" name="__module.encoder.layer.5.attention.self.key/aten::linear/Add/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="657" name="__module.encoder.layer.5.attention.self/prim::ListConstruct/Concat_1" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201240" size="32" /> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="658" name="__module.encoder.layer.5.attention.self/aten::view/Reshape_1" type="Reshape" version="opset1"> + <data special_zero="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="509,x.67"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="659" name="Constant_123740" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201272" size="32" /> + <output> + <port id="0" precision="I64" names="510"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="660" name="__module.encoder.layer.5.attention.self/aten::permute/Transpose_1" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="511"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="661" name="Constant_578636" type="Const" version="opset1"> + <data element_type="i8" shape="384, 384" offset="21422232" size="147456" /> + <output> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="662" name="Convert_578637" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="663" name="Constant_578638" type="Const" version="opset1"> + <data element_type="f32" shape="384, 1" offset="21569688" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="664" name="__module.encoder.layer.5.attention.self.value/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="665" name="__module.encoder.layer.5.attention.self.value/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="666" name="Constant_131758" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="21571224" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="667" name="__module.encoder.layer.5.attention.self.value/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="514,x.69"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="668" name="__module.encoder.layer.5.attention.self/prim::ListConstruct/Concat_2" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201240" size="32" /> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="669" name="__module.encoder.layer.5.attention.self/aten::view/Reshape_2" type="Reshape" version="opset1"> + <data special_zero="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="518,x.71"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="670" name="Constant_123763" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201272" size="32" /> + <output> + <port id="0" precision="I64" names="519"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="671" name="__module.encoder.layer.5.attention.self/aten::permute/Transpose_2" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="520"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="672" name="__module.encoder.layer.5.attention.self/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> + <data causal="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="521,attn_output.21"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="673" name="__module.encoder.layer.5.attention.self/aten::transpose/ScatterElementsUpdate" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="12502400" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="674" name="__module.encoder.layer.5.attention.self/aten::transpose/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="522,attn_output.23"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="675" name="Constant_131953" type="Const" version="opset1"> + <data element_type="i64" shape="3" offset="12502416" size="24" /> + <output> + <port id="0" precision="I64"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="676" name="__module.encoder.layer.5.attention.self/aten::reshape/Reshape" type="Reshape" version="opset1"> + <data special_zero="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="524"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="677" name="__module.encoder.layer.5.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/scale" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="21572760" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="678" name="__module.encoder.layer.5.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="679" name="__module.encoder.layer.5.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="21574296" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="680" name="__module.encoder.layer.5.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="21574300" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="681" name="__module.encoder.layer.5.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="21574296" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="682" name="__module.encoder.layer.5.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="21574300" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="683" name="__module.encoder.layer.5.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="684" name="Constant_578640" type="Const" version="opset1"> + <data element_type="i8" shape="384, 384" offset="21574304" size="147456" /> + <output> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="685" name="Convert_578641" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="686" name="Constant_578642" type="Const" version="opset1"> + <data element_type="f32" shape="384, 1" offset="21721760" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="687" name="__module.encoder.layer.5.attention.output.dense/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="688" name="__module.encoder.layer.5.attention.output.dense/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="689" name="Constant_131759" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="21723296" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="690" name="__module.encoder.layer.5.attention.output.dense/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="530,input.23"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="691" name="__module.encoder.layer.5.attention.output/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="532"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="692" name="__module.encoder.layer.5.attention.output.LayerNorm/aten::layer_norm/Multiply" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="12046084" size="4" /> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="693" name="__module.encoder.layer.5.attention.output.LayerNorm/aten::layer_norm/MVN" type="MVN" version="opset6"> + <data eps="9.999999960041972e-13" normalize_variance="true" eps_mode="INSIDE_SQRT" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="694" name="Constant_131760" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="21724832" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="695" name="__module.encoder.layer.5.attention.output.LayerNorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="696" name="Constant_131761" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="21726368" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="697" name="__module.encoder.layer.5.attention.output.LayerNorm/aten::layer_norm/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="536,input_tensor.11"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="698" name="__module.encoder.layer.5.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/scale" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="21727904" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="699" name="__module.encoder.layer.5.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="700" name="__module.encoder.layer.5.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="21729440" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="701" name="__module.encoder.layer.5.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="21729444" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="702" name="__module.encoder.layer.5.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="21729440" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="703" name="__module.encoder.layer.5.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="21729444" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="704" name="__module.encoder.layer.5.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="705" name="Constant_578644" type="Const" version="opset1"> + <data element_type="i8" shape="1536, 384" offset="21729448" size="589824" /> + <output> + <port id="0" precision="I8"> + <dim>1536</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="706" name="Convert_578645" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>1536</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1536</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="707" name="Constant_578646" type="Const" version="opset1"> + <data element_type="f32" shape="1536, 1" offset="22319272" size="6144" /> + <output> + <port id="0" precision="FP32"> + <dim>1536</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="708" name="__module.encoder.layer.5.intermediate.dense/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>1536</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1536</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>1536</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="709" name="__module.encoder.layer.5.intermediate.dense/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1536</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="710" name="Constant_131762" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 1536" offset="22325416" size="6144" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="711" name="__module.encoder.layer.5.intermediate.dense/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1536</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="541"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="712" name="__module.encoder.layer.5.intermediate.intermediate_act_fn/aten::gelu/Gelu" type="Gelu" version="opset7"> + <data approximation_mode="ERF" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="542"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="713" name="__module.encoder.layer.5.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/scale" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 1536" offset="22331560" size="6144" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="714" name="__module.encoder.layer.5.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1536</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="715" name="__module.encoder.layer.5.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="22337704" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="716" name="__module.encoder.layer.5.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="22337708" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="717" name="__module.encoder.layer.5.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="22337704" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="718" name="__module.encoder.layer.5.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="22337708" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="719" name="__module.encoder.layer.5.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="720" name="Constant_578648" type="Const" version="opset1"> + <data element_type="i8" shape="384, 1536" offset="22337712" size="589824" /> + <output> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="721" name="Convert_578649" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>1536</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="722" name="Constant_578650" type="Const" version="opset1"> + <data element_type="f32" shape="384, 1" offset="22927536" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="723" name="__module.encoder.layer.5.output.dense/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1536</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>384</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="724" name="__module.encoder.layer.5.output.dense/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1536</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="725" name="Constant_131763" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="22929072" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="726" name="__module.encoder.layer.5.output.dense/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="548,input.25"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="727" name="__module.encoder.layer.5.output/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="550"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="728" name="__module.encoder.layer.5.output.LayerNorm/aten::layer_norm/Multiply" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="12046084" size="4" /> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="729" name="__module.encoder.layer.5.output.LayerNorm/aten::layer_norm/MVN" type="MVN" version="opset6"> + <data eps="9.999999960041972e-13" normalize_variance="true" eps_mode="INSIDE_SQRT" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="730" name="Constant_131764" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="22930608" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="731" name="__module.encoder.layer.5.output.LayerNorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="732" name="Constant_131765" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="22932144" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="733" name="__module.encoder.layer.5.output.LayerNorm/aten::layer_norm/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="554,hidden_states.37"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="734" name="__module.encoder.layer.5.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/scale" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="22933680" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="735" name="__module.encoder.layer.5.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="736" name="__module.encoder.layer.5.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="22935216" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="737" name="__module.encoder.layer.5.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="22935220" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="738" name="__module.encoder.layer.5.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="22935216" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="739" name="__module.encoder.layer.5.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="22935220" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="740" name="__module.encoder.layer.5.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="741" name="Constant_578652" type="Const" version="opset1"> + <data element_type="i8" shape="384, 384" offset="22935224" size="147456" /> + <output> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="742" name="Convert_578653" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="743" name="Constant_578654" type="Const" version="opset1"> + <data element_type="f32" shape="384, 1" offset="23082680" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="744" name="__module.encoder.layer.6.attention.self.query/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="745" name="__module.encoder.layer.6.attention.self.query/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="746" name="Constant_131766" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="23084216" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="747" name="__module.encoder.layer.6.attention.self.query/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="567,x.73"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="748" name="__module.encoder.layer.6.attention.self.query/aten::linear/Add/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="23085752" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="749" name="__module.encoder.layer.6.attention.self.query/aten::linear/Add/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="23085756" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="750" name="__module.encoder.layer.6.attention.self.query/aten::linear/Add/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="23085752" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="751" name="__module.encoder.layer.6.attention.self.query/aten::linear/Add/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="23085756" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="752" name="__module.encoder.layer.6.attention.self.query/aten::linear/Add/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="753" name="__module.encoder.layer.6.attention.self/prim::ListConstruct/Concat" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201240" size="32" /> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="754" name="__module.encoder.layer.6.attention.self/aten::view/Reshape" type="Reshape" version="opset1"> + <data special_zero="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="571,x.75"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="755" name="Constant_123943" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201272" size="32" /> + <output> + <port id="0" precision="I64" names="572"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="756" name="__module.encoder.layer.6.attention.self/aten::permute/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="573"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="757" name="Constant_578656" type="Const" version="opset1"> + <data element_type="i8" shape="384, 384" offset="23085760" size="147456" /> + <output> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="758" name="Convert_578657" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="759" name="Constant_578658" type="Const" version="opset1"> + <data element_type="f32" shape="384, 1" offset="23233216" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="760" name="__module.encoder.layer.6.attention.self.key/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="761" name="__module.encoder.layer.6.attention.self.key/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="762" name="Constant_131767" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="23234752" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="763" name="__module.encoder.layer.6.attention.self.key/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="576,x.77"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="764" name="__module.encoder.layer.6.attention.self.key/aten::linear/Add/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="23236288" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="765" name="__module.encoder.layer.6.attention.self.key/aten::linear/Add/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="23236292" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="766" name="__module.encoder.layer.6.attention.self.key/aten::linear/Add/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="23236288" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="767" name="__module.encoder.layer.6.attention.self.key/aten::linear/Add/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="23236292" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="768" name="__module.encoder.layer.6.attention.self.key/aten::linear/Add/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="769" name="__module.encoder.layer.6.attention.self/prim::ListConstruct/Concat_1" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201240" size="32" /> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="770" name="__module.encoder.layer.6.attention.self/aten::view/Reshape_1" type="Reshape" version="opset1"> + <data special_zero="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="580,x.79"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="771" name="Constant_123966" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201272" size="32" /> + <output> + <port id="0" precision="I64" names="581"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="772" name="__module.encoder.layer.6.attention.self/aten::permute/Transpose_1" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="582"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="773" name="Constant_578660" type="Const" version="opset1"> + <data element_type="i8" shape="384, 384" offset="23236296" size="147456" /> + <output> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="774" name="Convert_578661" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="775" name="Constant_578662" type="Const" version="opset1"> + <data element_type="f32" shape="384, 1" offset="23383752" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="776" name="__module.encoder.layer.6.attention.self.value/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="777" name="__module.encoder.layer.6.attention.self.value/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="778" name="Constant_131768" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="23385288" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="779" name="__module.encoder.layer.6.attention.self.value/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="585,x.81"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="780" name="__module.encoder.layer.6.attention.self/prim::ListConstruct/Concat_2" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201240" size="32" /> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="781" name="__module.encoder.layer.6.attention.self/aten::view/Reshape_2" type="Reshape" version="opset1"> + <data special_zero="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="589,x.83"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="782" name="Constant_123989" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201272" size="32" /> + <output> + <port id="0" precision="I64" names="590"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="783" name="__module.encoder.layer.6.attention.self/aten::permute/Transpose_2" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="591"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="784" name="__module.encoder.layer.6.attention.self/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> + <data causal="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="592,attn_output.25"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="785" name="__module.encoder.layer.6.attention.self/aten::transpose/ScatterElementsUpdate" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="12502400" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="786" name="__module.encoder.layer.6.attention.self/aten::transpose/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="593,attn_output.27"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="787" name="Constant_131954" type="Const" version="opset1"> + <data element_type="i64" shape="3" offset="12502416" size="24" /> + <output> + <port id="0" precision="I64"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="788" name="__module.encoder.layer.6.attention.self/aten::reshape/Reshape" type="Reshape" version="opset1"> + <data special_zero="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="595"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="789" name="__module.encoder.layer.6.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/scale" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="23386824" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="790" name="__module.encoder.layer.6.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="791" name="__module.encoder.layer.6.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="23388360" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="792" name="__module.encoder.layer.6.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="23388364" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="793" name="__module.encoder.layer.6.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="23388360" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="794" name="__module.encoder.layer.6.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="23388364" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="795" name="__module.encoder.layer.6.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="796" name="Constant_578664" type="Const" version="opset1"> + <data element_type="i8" shape="384, 384" offset="23388368" size="147456" /> + <output> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="797" name="Convert_578665" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="798" name="Constant_578666" type="Const" version="opset1"> + <data element_type="f32" shape="384, 1" offset="23535824" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="799" name="__module.encoder.layer.6.attention.output.dense/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="800" name="__module.encoder.layer.6.attention.output.dense/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="801" name="Constant_131769" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="23537360" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="802" name="__module.encoder.layer.6.attention.output.dense/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="601,input.27"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="803" name="__module.encoder.layer.6.attention.output/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="603"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="804" name="__module.encoder.layer.6.attention.output.LayerNorm/aten::layer_norm/Multiply" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="12046084" size="4" /> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="805" name="__module.encoder.layer.6.attention.output.LayerNorm/aten::layer_norm/MVN" type="MVN" version="opset6"> + <data eps="9.999999960041972e-13" normalize_variance="true" eps_mode="INSIDE_SQRT" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="806" name="Constant_131770" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="23538896" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="807" name="__module.encoder.layer.6.attention.output.LayerNorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="808" name="Constant_131771" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="23540432" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="809" name="__module.encoder.layer.6.attention.output.LayerNorm/aten::layer_norm/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="607,input_tensor.13"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="810" name="__module.encoder.layer.6.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/scale" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="23541968" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="811" name="__module.encoder.layer.6.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="812" name="__module.encoder.layer.6.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="23543504" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="813" name="__module.encoder.layer.6.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="23543508" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="814" name="__module.encoder.layer.6.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="23543504" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="815" name="__module.encoder.layer.6.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="23543508" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="816" name="__module.encoder.layer.6.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="817" name="Constant_578668" type="Const" version="opset1"> + <data element_type="i8" shape="1536, 384" offset="23543512" size="589824" /> + <output> + <port id="0" precision="I8"> + <dim>1536</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="818" name="Convert_578669" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>1536</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1536</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="819" name="Constant_578670" type="Const" version="opset1"> + <data element_type="f32" shape="1536, 1" offset="24133336" size="6144" /> + <output> + <port id="0" precision="FP32"> + <dim>1536</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="820" name="__module.encoder.layer.6.intermediate.dense/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>1536</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1536</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>1536</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="821" name="__module.encoder.layer.6.intermediate.dense/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1536</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="822" name="Constant_131772" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 1536" offset="24139480" size="6144" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="823" name="__module.encoder.layer.6.intermediate.dense/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1536</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="612"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="824" name="__module.encoder.layer.6.intermediate.intermediate_act_fn/aten::gelu/Gelu" type="Gelu" version="opset7"> + <data approximation_mode="ERF" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="613"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="825" name="__module.encoder.layer.6.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/scale" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 1536" offset="24145624" size="6144" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="826" name="__module.encoder.layer.6.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1536</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="827" name="__module.encoder.layer.6.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="24151768" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="828" name="__module.encoder.layer.6.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="24151772" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="829" name="__module.encoder.layer.6.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="24151768" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="830" name="__module.encoder.layer.6.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="24151772" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="831" name="__module.encoder.layer.6.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="832" name="Constant_578672" type="Const" version="opset1"> + <data element_type="i8" shape="384, 1536" offset="24151776" size="589824" /> + <output> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="833" name="Convert_578673" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>1536</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="834" name="Constant_578674" type="Const" version="opset1"> + <data element_type="f32" shape="384, 1" offset="24741600" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="835" name="__module.encoder.layer.6.output.dense/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1536</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>384</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="836" name="__module.encoder.layer.6.output.dense/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1536</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="837" name="Constant_131773" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="24743136" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="838" name="__module.encoder.layer.6.output.dense/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="619,input.29"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="839" name="__module.encoder.layer.6.output/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="621"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="840" name="__module.encoder.layer.6.output.LayerNorm/aten::layer_norm/Multiply" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="12046084" size="4" /> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="841" name="__module.encoder.layer.6.output.LayerNorm/aten::layer_norm/MVN" type="MVN" version="opset6"> + <data eps="9.999999960041972e-13" normalize_variance="true" eps_mode="INSIDE_SQRT" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="842" name="Constant_131774" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="24744672" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="843" name="__module.encoder.layer.6.output.LayerNorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="844" name="Constant_131775" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="24746208" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="845" name="__module.encoder.layer.6.output.LayerNorm/aten::layer_norm/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="625,hidden_states.43"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="846" name="__module.encoder.layer.6.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/scale" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="24747744" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="847" name="__module.encoder.layer.6.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="848" name="__module.encoder.layer.6.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="24749280" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="849" name="__module.encoder.layer.6.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="24749284" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="850" name="__module.encoder.layer.6.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="24749280" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="851" name="__module.encoder.layer.6.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="24749284" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="852" name="__module.encoder.layer.6.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="853" name="Constant_578676" type="Const" version="opset1"> + <data element_type="i8" shape="384, 384" offset="24749288" size="147456" /> + <output> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="854" name="Convert_578677" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="855" name="Constant_578678" type="Const" version="opset1"> + <data element_type="f32" shape="384, 1" offset="24896744" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="856" name="__module.encoder.layer.7.attention.self.query/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="857" name="__module.encoder.layer.7.attention.self.query/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="858" name="Constant_131776" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="24898280" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="859" name="__module.encoder.layer.7.attention.self.query/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="638,x.85"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="860" name="__module.encoder.layer.7.attention.self.query/aten::linear/Add/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="24899816" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="861" name="__module.encoder.layer.7.attention.self.query/aten::linear/Add/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="24899820" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="862" name="__module.encoder.layer.7.attention.self.query/aten::linear/Add/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="24899816" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="863" name="__module.encoder.layer.7.attention.self.query/aten::linear/Add/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="24899820" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="864" name="__module.encoder.layer.7.attention.self.query/aten::linear/Add/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="865" name="__module.encoder.layer.7.attention.self/prim::ListConstruct/Concat" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201240" size="32" /> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="866" name="__module.encoder.layer.7.attention.self/aten::view/Reshape" type="Reshape" version="opset1"> + <data special_zero="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="642,x.87"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="867" name="Constant_124169" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201272" size="32" /> + <output> + <port id="0" precision="I64" names="643"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="868" name="__module.encoder.layer.7.attention.self/aten::permute/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="644"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="869" name="Constant_578680" type="Const" version="opset1"> + <data element_type="i8" shape="384, 384" offset="24899824" size="147456" /> + <output> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="870" name="Convert_578681" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="871" name="Constant_578682" type="Const" version="opset1"> + <data element_type="f32" shape="384, 1" offset="25047280" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="872" name="__module.encoder.layer.7.attention.self.key/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="873" name="__module.encoder.layer.7.attention.self.key/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="874" name="Constant_131777" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="25048816" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="875" name="__module.encoder.layer.7.attention.self.key/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="647,x.89"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="876" name="__module.encoder.layer.7.attention.self.key/aten::linear/Add/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="25050352" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="877" name="__module.encoder.layer.7.attention.self.key/aten::linear/Add/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="25050356" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="878" name="__module.encoder.layer.7.attention.self.key/aten::linear/Add/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="25050352" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="879" name="__module.encoder.layer.7.attention.self.key/aten::linear/Add/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="25050356" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="880" name="__module.encoder.layer.7.attention.self.key/aten::linear/Add/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="881" name="__module.encoder.layer.7.attention.self/prim::ListConstruct/Concat_1" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201240" size="32" /> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="882" name="__module.encoder.layer.7.attention.self/aten::view/Reshape_1" type="Reshape" version="opset1"> + <data special_zero="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="651,x.91"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="883" name="Constant_124192" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201272" size="32" /> + <output> + <port id="0" precision="I64" names="652"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="884" name="__module.encoder.layer.7.attention.self/aten::permute/Transpose_1" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="653"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="885" name="Constant_578684" type="Const" version="opset1"> + <data element_type="i8" shape="384, 384" offset="25050360" size="147456" /> + <output> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="886" name="Convert_578685" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="887" name="Constant_578686" type="Const" version="opset1"> + <data element_type="f32" shape="384, 1" offset="25197816" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="888" name="__module.encoder.layer.7.attention.self.value/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="889" name="__module.encoder.layer.7.attention.self.value/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="890" name="Constant_131778" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="25199352" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="891" name="__module.encoder.layer.7.attention.self.value/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="656,x.93"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="892" name="__module.encoder.layer.7.attention.self/prim::ListConstruct/Concat_2" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201240" size="32" /> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="893" name="__module.encoder.layer.7.attention.self/aten::view/Reshape_2" type="Reshape" version="opset1"> + <data special_zero="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="660,x.95"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="894" name="Constant_124215" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201272" size="32" /> + <output> + <port id="0" precision="I64" names="661"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="895" name="__module.encoder.layer.7.attention.self/aten::permute/Transpose_2" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="662"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="896" name="__module.encoder.layer.7.attention.self/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> + <data causal="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="663,attn_output.29"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="897" name="__module.encoder.layer.7.attention.self/aten::transpose/ScatterElementsUpdate" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="12502400" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="898" name="__module.encoder.layer.7.attention.self/aten::transpose/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="664,attn_output.31"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="899" name="Constant_131955" type="Const" version="opset1"> + <data element_type="i64" shape="3" offset="12502416" size="24" /> + <output> + <port id="0" precision="I64"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="900" name="__module.encoder.layer.7.attention.self/aten::reshape/Reshape" type="Reshape" version="opset1"> + <data special_zero="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="666"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="901" name="__module.encoder.layer.7.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/scale" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="25200888" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="902" name="__module.encoder.layer.7.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="903" name="__module.encoder.layer.7.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="25202424" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="904" name="__module.encoder.layer.7.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="25202428" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="905" name="__module.encoder.layer.7.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="25202424" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="906" name="__module.encoder.layer.7.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="25202428" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="907" name="__module.encoder.layer.7.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="908" name="Constant_578688" type="Const" version="opset1"> + <data element_type="i8" shape="384, 384" offset="25202432" size="147456" /> + <output> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="909" name="Convert_578689" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="910" name="Constant_578690" type="Const" version="opset1"> + <data element_type="f32" shape="384, 1" offset="25349888" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="911" name="__module.encoder.layer.7.attention.output.dense/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="912" name="__module.encoder.layer.7.attention.output.dense/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="913" name="Constant_131779" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="25351424" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="914" name="__module.encoder.layer.7.attention.output.dense/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="672,input.31"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="915" name="__module.encoder.layer.7.attention.output/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="674"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="916" name="__module.encoder.layer.7.attention.output.LayerNorm/aten::layer_norm/Multiply" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="12046084" size="4" /> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="917" name="__module.encoder.layer.7.attention.output.LayerNorm/aten::layer_norm/MVN" type="MVN" version="opset6"> + <data eps="9.999999960041972e-13" normalize_variance="true" eps_mode="INSIDE_SQRT" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="918" name="Constant_131780" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="25352960" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="919" name="__module.encoder.layer.7.attention.output.LayerNorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="920" name="Constant_131781" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="25354496" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="921" name="__module.encoder.layer.7.attention.output.LayerNorm/aten::layer_norm/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="678,input_tensor.15"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="922" name="__module.encoder.layer.7.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/scale" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="25356032" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="923" name="__module.encoder.layer.7.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="924" name="__module.encoder.layer.7.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="25357568" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="925" name="__module.encoder.layer.7.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="25357572" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="926" name="__module.encoder.layer.7.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="25357568" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="927" name="__module.encoder.layer.7.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="25357572" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="928" name="__module.encoder.layer.7.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="929" name="Constant_578692" type="Const" version="opset1"> + <data element_type="i8" shape="1536, 384" offset="25357576" size="589824" /> + <output> + <port id="0" precision="I8"> + <dim>1536</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="930" name="Convert_578693" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>1536</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1536</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="931" name="Constant_578694" type="Const" version="opset1"> + <data element_type="f32" shape="1536, 1" offset="25947400" size="6144" /> + <output> + <port id="0" precision="FP32"> + <dim>1536</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="932" name="__module.encoder.layer.7.intermediate.dense/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>1536</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1536</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>1536</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="933" name="__module.encoder.layer.7.intermediate.dense/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1536</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="934" name="Constant_131782" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 1536" offset="25953544" size="6144" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="935" name="__module.encoder.layer.7.intermediate.dense/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1536</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="683"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="936" name="__module.encoder.layer.7.intermediate.intermediate_act_fn/aten::gelu/Gelu" type="Gelu" version="opset7"> + <data approximation_mode="ERF" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="684"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="937" name="__module.encoder.layer.7.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/scale" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 1536" offset="25959688" size="6144" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="938" name="__module.encoder.layer.7.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1536</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="939" name="__module.encoder.layer.7.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="25965832" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="940" name="__module.encoder.layer.7.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="25965836" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="941" name="__module.encoder.layer.7.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="25965832" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="942" name="__module.encoder.layer.7.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="25965836" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="943" name="__module.encoder.layer.7.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="944" name="Constant_578696" type="Const" version="opset1"> + <data element_type="i8" shape="384, 1536" offset="25965840" size="589824" /> + <output> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="945" name="Convert_578697" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>1536</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="946" name="Constant_578698" type="Const" version="opset1"> + <data element_type="f32" shape="384, 1" offset="26555664" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="947" name="__module.encoder.layer.7.output.dense/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1536</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>384</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="948" name="__module.encoder.layer.7.output.dense/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1536</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="949" name="Constant_131783" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="26557200" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="950" name="__module.encoder.layer.7.output.dense/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="690,input.33"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="951" name="__module.encoder.layer.7.output/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="692"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="952" name="__module.encoder.layer.7.output.LayerNorm/aten::layer_norm/Multiply" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="12046084" size="4" /> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="953" name="__module.encoder.layer.7.output.LayerNorm/aten::layer_norm/MVN" type="MVN" version="opset6"> + <data eps="9.999999960041972e-13" normalize_variance="true" eps_mode="INSIDE_SQRT" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="954" name="Constant_131784" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="26558736" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="955" name="__module.encoder.layer.7.output.LayerNorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="956" name="Constant_131785" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="26560272" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="957" name="__module.encoder.layer.7.output.LayerNorm/aten::layer_norm/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="696,hidden_states.49"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="958" name="__module.encoder.layer.7.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/scale" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="26561808" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="959" name="__module.encoder.layer.7.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="960" name="__module.encoder.layer.7.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="26563344" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="961" name="__module.encoder.layer.7.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="26563348" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="962" name="__module.encoder.layer.7.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="26563344" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="963" name="__module.encoder.layer.7.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="26563348" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="964" name="__module.encoder.layer.7.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="965" name="Constant_578700" type="Const" version="opset1"> + <data element_type="i8" shape="384, 384" offset="26563352" size="147456" /> + <output> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="966" name="Convert_578701" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="967" name="Constant_578702" type="Const" version="opset1"> + <data element_type="f32" shape="384, 1" offset="26710808" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="968" name="__module.encoder.layer.8.attention.self.query/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="969" name="__module.encoder.layer.8.attention.self.query/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="970" name="Constant_131786" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="26712344" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="971" name="__module.encoder.layer.8.attention.self.query/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="709,x.97"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="972" name="__module.encoder.layer.8.attention.self.query/aten::linear/Add/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="26713880" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="973" name="__module.encoder.layer.8.attention.self.query/aten::linear/Add/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="26713884" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="974" name="__module.encoder.layer.8.attention.self.query/aten::linear/Add/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="26713880" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="975" name="__module.encoder.layer.8.attention.self.query/aten::linear/Add/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="26713884" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="976" name="__module.encoder.layer.8.attention.self.query/aten::linear/Add/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="977" name="__module.encoder.layer.8.attention.self/prim::ListConstruct/Concat" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201240" size="32" /> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="978" name="__module.encoder.layer.8.attention.self/aten::view/Reshape" type="Reshape" version="opset1"> + <data special_zero="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="713,x.99"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="979" name="Constant_124395" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201272" size="32" /> + <output> + <port id="0" precision="I64" names="714"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="980" name="__module.encoder.layer.8.attention.self/aten::permute/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="715"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="981" name="Constant_578704" type="Const" version="opset1"> + <data element_type="i8" shape="384, 384" offset="26713888" size="147456" /> + <output> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="982" name="Convert_578705" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="983" name="Constant_578706" type="Const" version="opset1"> + <data element_type="f32" shape="384, 1" offset="26861344" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="984" name="__module.encoder.layer.8.attention.self.key/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="985" name="__module.encoder.layer.8.attention.self.key/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="986" name="Constant_131787" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="26862880" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="987" name="__module.encoder.layer.8.attention.self.key/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="718,x.101"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="988" name="__module.encoder.layer.8.attention.self.key/aten::linear/Add/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="26864416" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="989" name="__module.encoder.layer.8.attention.self.key/aten::linear/Add/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="26864420" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="990" name="__module.encoder.layer.8.attention.self.key/aten::linear/Add/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="26864416" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="991" name="__module.encoder.layer.8.attention.self.key/aten::linear/Add/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="26864420" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="992" name="__module.encoder.layer.8.attention.self.key/aten::linear/Add/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="993" name="__module.encoder.layer.8.attention.self/prim::ListConstruct/Concat_1" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201240" size="32" /> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="994" name="__module.encoder.layer.8.attention.self/aten::view/Reshape_1" type="Reshape" version="opset1"> + <data special_zero="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="722,x.103"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="995" name="Constant_124418" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201272" size="32" /> + <output> + <port id="0" precision="I64" names="723"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="996" name="__module.encoder.layer.8.attention.self/aten::permute/Transpose_1" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="724"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="997" name="Constant_578708" type="Const" version="opset1"> + <data element_type="i8" shape="384, 384" offset="26864424" size="147456" /> + <output> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="998" name="Convert_578709" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="999" name="Constant_578710" type="Const" version="opset1"> + <data element_type="f32" shape="384, 1" offset="27011880" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1000" name="__module.encoder.layer.8.attention.self.value/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1001" name="__module.encoder.layer.8.attention.self.value/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1002" name="Constant_131788" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="27013416" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1003" name="__module.encoder.layer.8.attention.self.value/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="727,x.105"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1004" name="__module.encoder.layer.8.attention.self/prim::ListConstruct/Concat_2" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201240" size="32" /> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1005" name="__module.encoder.layer.8.attention.self/aten::view/Reshape_2" type="Reshape" version="opset1"> + <data special_zero="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="731,x.107"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="1006" name="Constant_124441" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201272" size="32" /> + <output> + <port id="0" precision="I64" names="732"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1007" name="__module.encoder.layer.8.attention.self/aten::permute/Transpose_2" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="733"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="1008" name="__module.encoder.layer.8.attention.self/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> + <data causal="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="734,attn_output.33"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="1009" name="__module.encoder.layer.8.attention.self/aten::transpose/ScatterElementsUpdate" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="12502400" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1010" name="__module.encoder.layer.8.attention.self/aten::transpose/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="735,attn_output.35"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="1011" name="Constant_131956" type="Const" version="opset1"> + <data element_type="i64" shape="3" offset="12502416" size="24" /> + <output> + <port id="0" precision="I64"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="1012" name="__module.encoder.layer.8.attention.self/aten::reshape/Reshape" type="Reshape" version="opset1"> + <data special_zero="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="737"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1013" name="__module.encoder.layer.8.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/scale" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="27014952" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1014" name="__module.encoder.layer.8.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1015" name="__module.encoder.layer.8.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="27016488" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1016" name="__module.encoder.layer.8.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="27016492" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1017" name="__module.encoder.layer.8.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="27016488" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1018" name="__module.encoder.layer.8.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="27016492" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1019" name="__module.encoder.layer.8.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1020" name="Constant_578712" type="Const" version="opset1"> + <data element_type="i8" shape="384, 384" offset="27016496" size="147456" /> + <output> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1021" name="Convert_578713" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1022" name="Constant_578714" type="Const" version="opset1"> + <data element_type="f32" shape="384, 1" offset="27163952" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1023" name="__module.encoder.layer.8.attention.output.dense/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1024" name="__module.encoder.layer.8.attention.output.dense/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1025" name="Constant_131789" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="27165488" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1026" name="__module.encoder.layer.8.attention.output.dense/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="743,input.35"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1027" name="__module.encoder.layer.8.attention.output/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="745"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1028" name="__module.encoder.layer.8.attention.output.LayerNorm/aten::layer_norm/Multiply" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="12046084" size="4" /> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1029" name="__module.encoder.layer.8.attention.output.LayerNorm/aten::layer_norm/MVN" type="MVN" version="opset6"> + <data eps="9.999999960041972e-13" normalize_variance="true" eps_mode="INSIDE_SQRT" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1030" name="Constant_131790" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="27167024" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1031" name="__module.encoder.layer.8.attention.output.LayerNorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1032" name="Constant_131791" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="27168560" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1033" name="__module.encoder.layer.8.attention.output.LayerNorm/aten::layer_norm/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="749,input_tensor.17"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1034" name="__module.encoder.layer.8.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/scale" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="27170096" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1035" name="__module.encoder.layer.8.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1036" name="__module.encoder.layer.8.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="27171632" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1037" name="__module.encoder.layer.8.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="27171636" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1038" name="__module.encoder.layer.8.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="27171632" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1039" name="__module.encoder.layer.8.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="27171636" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1040" name="__module.encoder.layer.8.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1041" name="Constant_578716" type="Const" version="opset1"> + <data element_type="i8" shape="1536, 384" offset="27171640" size="589824" /> + <output> + <port id="0" precision="I8"> + <dim>1536</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1042" name="Convert_578717" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>1536</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1536</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1043" name="Constant_578718" type="Const" version="opset1"> + <data element_type="f32" shape="1536, 1" offset="27761464" size="6144" /> + <output> + <port id="0" precision="FP32"> + <dim>1536</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1044" name="__module.encoder.layer.8.intermediate.dense/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>1536</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1536</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>1536</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1045" name="__module.encoder.layer.8.intermediate.dense/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1536</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="1046" name="Constant_131792" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 1536" offset="27767608" size="6144" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="1047" name="__module.encoder.layer.8.intermediate.dense/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1536</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="754"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="1048" name="__module.encoder.layer.8.intermediate.intermediate_act_fn/aten::gelu/Gelu" type="Gelu" version="opset7"> + <data approximation_mode="ERF" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="755"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="1049" name="__module.encoder.layer.8.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/scale" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 1536" offset="27773752" size="6144" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="1050" name="__module.encoder.layer.8.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1536</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="1051" name="__module.encoder.layer.8.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="27779896" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1052" name="__module.encoder.layer.8.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="27779900" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1053" name="__module.encoder.layer.8.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="27779896" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1054" name="__module.encoder.layer.8.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="27779900" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1055" name="__module.encoder.layer.8.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="1056" name="Constant_578720" type="Const" version="opset1"> + <data element_type="i8" shape="384, 1536" offset="27779904" size="589824" /> + <output> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="1057" name="Convert_578721" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>1536</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="1058" name="Constant_578722" type="Const" version="opset1"> + <data element_type="f32" shape="384, 1" offset="28369728" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1059" name="__module.encoder.layer.8.output.dense/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1536</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>384</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="1060" name="__module.encoder.layer.8.output.dense/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1536</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1061" name="Constant_131793" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="28371264" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1062" name="__module.encoder.layer.8.output.dense/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="761,input.37"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1063" name="__module.encoder.layer.8.output/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="763"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1064" name="__module.encoder.layer.8.output.LayerNorm/aten::layer_norm/Multiply" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="12046084" size="4" /> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1065" name="__module.encoder.layer.8.output.LayerNorm/aten::layer_norm/MVN" type="MVN" version="opset6"> + <data eps="9.999999960041972e-13" normalize_variance="true" eps_mode="INSIDE_SQRT" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1066" name="Constant_131794" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="28372800" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1067" name="__module.encoder.layer.8.output.LayerNorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1068" name="Constant_131795" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="28374336" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1069" name="__module.encoder.layer.8.output.LayerNorm/aten::layer_norm/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="767,hidden_states.55"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1070" name="__module.encoder.layer.8.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/scale" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="28375872" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1071" name="__module.encoder.layer.8.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1072" name="__module.encoder.layer.8.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="28377408" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1073" name="__module.encoder.layer.8.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="28377412" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1074" name="__module.encoder.layer.8.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="28377408" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1075" name="__module.encoder.layer.8.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="28377412" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1076" name="__module.encoder.layer.8.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1077" name="Constant_578724" type="Const" version="opset1"> + <data element_type="i8" shape="384, 384" offset="28377416" size="147456" /> + <output> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1078" name="Convert_578725" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1079" name="Constant_578726" type="Const" version="opset1"> + <data element_type="f32" shape="384, 1" offset="28524872" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1080" name="__module.encoder.layer.9.attention.self.query/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1081" name="__module.encoder.layer.9.attention.self.query/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1082" name="Constant_131796" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="28526408" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1083" name="__module.encoder.layer.9.attention.self.query/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="780,x.109"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1084" name="__module.encoder.layer.9.attention.self.query/aten::linear/Add/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="28527944" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1085" name="__module.encoder.layer.9.attention.self.query/aten::linear/Add/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="28527948" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1086" name="__module.encoder.layer.9.attention.self.query/aten::linear/Add/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="28527944" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1087" name="__module.encoder.layer.9.attention.self.query/aten::linear/Add/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="28527948" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1088" name="__module.encoder.layer.9.attention.self.query/aten::linear/Add/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1089" name="__module.encoder.layer.9.attention.self/prim::ListConstruct/Concat" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201240" size="32" /> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1090" name="__module.encoder.layer.9.attention.self/aten::view/Reshape" type="Reshape" version="opset1"> + <data special_zero="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="784,x.111"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="1091" name="Constant_124621" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201272" size="32" /> + <output> + <port id="0" precision="I64" names="785"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1092" name="__module.encoder.layer.9.attention.self/aten::permute/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="786"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="1093" name="Constant_578728" type="Const" version="opset1"> + <data element_type="i8" shape="384, 384" offset="28527952" size="147456" /> + <output> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1094" name="Convert_578729" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1095" name="Constant_578730" type="Const" version="opset1"> + <data element_type="f32" shape="384, 1" offset="28675408" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1096" name="__module.encoder.layer.9.attention.self.key/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1097" name="__module.encoder.layer.9.attention.self.key/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1098" name="Constant_131797" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="28676944" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1099" name="__module.encoder.layer.9.attention.self.key/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="789,x.113"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1100" name="__module.encoder.layer.9.attention.self.key/aten::linear/Add/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="28678480" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1101" name="__module.encoder.layer.9.attention.self.key/aten::linear/Add/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="28678484" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1102" name="__module.encoder.layer.9.attention.self.key/aten::linear/Add/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="28678480" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1103" name="__module.encoder.layer.9.attention.self.key/aten::linear/Add/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="28678484" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1104" name="__module.encoder.layer.9.attention.self.key/aten::linear/Add/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1105" name="__module.encoder.layer.9.attention.self/prim::ListConstruct/Concat_1" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201240" size="32" /> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1106" name="__module.encoder.layer.9.attention.self/aten::view/Reshape_1" type="Reshape" version="opset1"> + <data special_zero="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="793,x.115"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="1107" name="Constant_124644" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201272" size="32" /> + <output> + <port id="0" precision="I64" names="794"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1108" name="__module.encoder.layer.9.attention.self/aten::permute/Transpose_1" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="795"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="1109" name="Constant_578732" type="Const" version="opset1"> + <data element_type="i8" shape="384, 384" offset="28678488" size="147456" /> + <output> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1110" name="Convert_578733" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1111" name="Constant_578734" type="Const" version="opset1"> + <data element_type="f32" shape="384, 1" offset="28825944" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1112" name="__module.encoder.layer.9.attention.self.value/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1113" name="__module.encoder.layer.9.attention.self.value/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1114" name="Constant_131798" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="28827480" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1115" name="__module.encoder.layer.9.attention.self.value/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="798,x.117"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1116" name="__module.encoder.layer.9.attention.self/prim::ListConstruct/Concat_2" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201240" size="32" /> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1117" name="__module.encoder.layer.9.attention.self/aten::view/Reshape_2" type="Reshape" version="opset1"> + <data special_zero="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="802,x.119"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="1118" name="Constant_124667" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201272" size="32" /> + <output> + <port id="0" precision="I64" names="803"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1119" name="__module.encoder.layer.9.attention.self/aten::permute/Transpose_2" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="804"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="1120" name="__module.encoder.layer.9.attention.self/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> + <data causal="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="805,attn_output.37"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="1121" name="__module.encoder.layer.9.attention.self/aten::transpose/ScatterElementsUpdate" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="12502400" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1122" name="__module.encoder.layer.9.attention.self/aten::transpose/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="806,attn_output.39"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="1123" name="Constant_131957" type="Const" version="opset1"> + <data element_type="i64" shape="3" offset="12502416" size="24" /> + <output> + <port id="0" precision="I64"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="1124" name="__module.encoder.layer.9.attention.self/aten::reshape/Reshape" type="Reshape" version="opset1"> + <data special_zero="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="808"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1125" name="__module.encoder.layer.9.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/scale" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="28829016" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1126" name="__module.encoder.layer.9.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1127" name="__module.encoder.layer.9.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="28830552" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1128" name="__module.encoder.layer.9.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="28830556" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1129" name="__module.encoder.layer.9.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="28830552" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1130" name="__module.encoder.layer.9.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="28830556" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1131" name="__module.encoder.layer.9.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1132" name="Constant_578736" type="Const" version="opset1"> + <data element_type="i8" shape="384, 384" offset="28830560" size="147456" /> + <output> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1133" name="Convert_578737" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1134" name="Constant_578738" type="Const" version="opset1"> + <data element_type="f32" shape="384, 1" offset="28978016" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1135" name="__module.encoder.layer.9.attention.output.dense/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1136" name="__module.encoder.layer.9.attention.output.dense/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1137" name="Constant_131799" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="28979552" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1138" name="__module.encoder.layer.9.attention.output.dense/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="814,input.39"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1139" name="__module.encoder.layer.9.attention.output/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="816"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1140" name="__module.encoder.layer.9.attention.output.LayerNorm/aten::layer_norm/Multiply" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="12046084" size="4" /> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1141" name="__module.encoder.layer.9.attention.output.LayerNorm/aten::layer_norm/MVN" type="MVN" version="opset6"> + <data eps="9.999999960041972e-13" normalize_variance="true" eps_mode="INSIDE_SQRT" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1142" name="Constant_131800" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="28981088" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1143" name="__module.encoder.layer.9.attention.output.LayerNorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1144" name="Constant_131801" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="28982624" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1145" name="__module.encoder.layer.9.attention.output.LayerNorm/aten::layer_norm/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="820,input_tensor.19"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1146" name="__module.encoder.layer.9.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/scale" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="28984160" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1147" name="__module.encoder.layer.9.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1148" name="__module.encoder.layer.9.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="28985696" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1149" name="__module.encoder.layer.9.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="28985700" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1150" name="__module.encoder.layer.9.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="28985696" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1151" name="__module.encoder.layer.9.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="28985700" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1152" name="__module.encoder.layer.9.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1153" name="Constant_578740" type="Const" version="opset1"> + <data element_type="i8" shape="1536, 384" offset="28985704" size="589824" /> + <output> + <port id="0" precision="I8"> + <dim>1536</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1154" name="Convert_578741" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>1536</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1536</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1155" name="Constant_578742" type="Const" version="opset1"> + <data element_type="f32" shape="1536, 1" offset="29575528" size="6144" /> + <output> + <port id="0" precision="FP32"> + <dim>1536</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1156" name="__module.encoder.layer.9.intermediate.dense/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>1536</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1536</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>1536</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1157" name="__module.encoder.layer.9.intermediate.dense/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1536</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="1158" name="Constant_131802" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 1536" offset="29581672" size="6144" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="1159" name="__module.encoder.layer.9.intermediate.dense/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1536</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="825"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="1160" name="__module.encoder.layer.9.intermediate.intermediate_act_fn/aten::gelu/Gelu" type="Gelu" version="opset7"> + <data approximation_mode="ERF" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="826"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="1161" name="__module.encoder.layer.9.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/scale" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 1536" offset="29587816" size="6144" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="1162" name="__module.encoder.layer.9.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1536</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="1163" name="__module.encoder.layer.9.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="29593960" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1164" name="__module.encoder.layer.9.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="29593964" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1165" name="__module.encoder.layer.9.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="29593960" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1166" name="__module.encoder.layer.9.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="29593964" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1167" name="__module.encoder.layer.9.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="1168" name="Constant_578744" type="Const" version="opset1"> + <data element_type="i8" shape="384, 1536" offset="29593968" size="589824" /> + <output> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="1169" name="Convert_578745" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>1536</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="1170" name="Constant_578746" type="Const" version="opset1"> + <data element_type="f32" shape="384, 1" offset="30183792" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1171" name="__module.encoder.layer.9.output.dense/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1536</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>384</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="1172" name="__module.encoder.layer.9.output.dense/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1536</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1173" name="Constant_131803" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="30185328" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1174" name="__module.encoder.layer.9.output.dense/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="832,input.41"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1175" name="__module.encoder.layer.9.output/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="834"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1176" name="__module.encoder.layer.9.output.LayerNorm/aten::layer_norm/Multiply" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="12046084" size="4" /> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1177" name="__module.encoder.layer.9.output.LayerNorm/aten::layer_norm/MVN" type="MVN" version="opset6"> + <data eps="9.999999960041972e-13" normalize_variance="true" eps_mode="INSIDE_SQRT" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1178" name="Constant_131804" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="30186864" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1179" name="__module.encoder.layer.9.output.LayerNorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1180" name="Constant_131805" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="30188400" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1181" name="__module.encoder.layer.9.output.LayerNorm/aten::layer_norm/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="838,hidden_states.61"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1182" name="__module.encoder.layer.9.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/scale" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="30189936" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1183" name="__module.encoder.layer.9.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1184" name="__module.encoder.layer.9.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="30191472" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1185" name="__module.encoder.layer.9.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="30191476" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1186" name="__module.encoder.layer.9.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="30191472" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1187" name="__module.encoder.layer.9.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="30191476" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1188" name="__module.encoder.layer.9.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1189" name="Constant_578748" type="Const" version="opset1"> + <data element_type="i8" shape="384, 384" offset="30191480" size="147456" /> + <output> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1190" name="Convert_578749" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1191" name="Constant_578750" type="Const" version="opset1"> + <data element_type="f32" shape="384, 1" offset="30338936" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1192" name="__module.encoder.layer.10.attention.self.query/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1193" name="__module.encoder.layer.10.attention.self.query/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1194" name="Constant_131806" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="30340472" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1195" name="__module.encoder.layer.10.attention.self.query/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="851,x.121"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1196" name="__module.encoder.layer.10.attention.self.query/aten::linear/Add/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="30342008" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1197" name="__module.encoder.layer.10.attention.self.query/aten::linear/Add/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="30342012" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1198" name="__module.encoder.layer.10.attention.self.query/aten::linear/Add/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="30342008" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1199" name="__module.encoder.layer.10.attention.self.query/aten::linear/Add/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="30342012" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1200" name="__module.encoder.layer.10.attention.self.query/aten::linear/Add/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1201" name="__module.encoder.layer.10.attention.self/prim::ListConstruct/Concat" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201240" size="32" /> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1202" name="__module.encoder.layer.10.attention.self/aten::view/Reshape" type="Reshape" version="opset1"> + <data special_zero="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="855,x.123"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="1203" name="Constant_124847" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201272" size="32" /> + <output> + <port id="0" precision="I64" names="856"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1204" name="__module.encoder.layer.10.attention.self/aten::permute/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="857"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="1205" name="Constant_578752" type="Const" version="opset1"> + <data element_type="i8" shape="384, 384" offset="30342016" size="147456" /> + <output> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1206" name="Convert_578753" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1207" name="Constant_578754" type="Const" version="opset1"> + <data element_type="f32" shape="384, 1" offset="30489472" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1208" name="__module.encoder.layer.10.attention.self.key/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1209" name="__module.encoder.layer.10.attention.self.key/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1210" name="Constant_131807" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="30491008" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1211" name="__module.encoder.layer.10.attention.self.key/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="860,x.125"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1212" name="__module.encoder.layer.10.attention.self.key/aten::linear/Add/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="30492544" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1213" name="__module.encoder.layer.10.attention.self.key/aten::linear/Add/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="30492548" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1214" name="__module.encoder.layer.10.attention.self.key/aten::linear/Add/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="30492544" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1215" name="__module.encoder.layer.10.attention.self.key/aten::linear/Add/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="30492548" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1216" name="__module.encoder.layer.10.attention.self.key/aten::linear/Add/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1217" name="__module.encoder.layer.10.attention.self/prim::ListConstruct/Concat_1" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201240" size="32" /> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1218" name="__module.encoder.layer.10.attention.self/aten::view/Reshape_1" type="Reshape" version="opset1"> + <data special_zero="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="864,x.127"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="1219" name="Constant_124870" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201272" size="32" /> + <output> + <port id="0" precision="I64" names="865"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1220" name="__module.encoder.layer.10.attention.self/aten::permute/Transpose_1" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="866"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="1221" name="Constant_578756" type="Const" version="opset1"> + <data element_type="i8" shape="384, 384" offset="30492552" size="147456" /> + <output> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1222" name="Convert_578757" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1223" name="Constant_578758" type="Const" version="opset1"> + <data element_type="f32" shape="384, 1" offset="30640008" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1224" name="__module.encoder.layer.10.attention.self.value/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1225" name="__module.encoder.layer.10.attention.self.value/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1226" name="Constant_131808" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="30641544" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1227" name="__module.encoder.layer.10.attention.self.value/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="869,x.129"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1228" name="__module.encoder.layer.10.attention.self/prim::ListConstruct/Concat_2" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201240" size="32" /> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1229" name="__module.encoder.layer.10.attention.self/aten::view/Reshape_2" type="Reshape" version="opset1"> + <data special_zero="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="873,x.131"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="1230" name="Constant_124893" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201272" size="32" /> + <output> + <port id="0" precision="I64" names="874"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1231" name="__module.encoder.layer.10.attention.self/aten::permute/Transpose_2" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="875"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="1232" name="__module.encoder.layer.10.attention.self/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> + <data causal="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="876,attn_output.41"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="1233" name="__module.encoder.layer.10.attention.self/aten::transpose/ScatterElementsUpdate" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="12502400" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1234" name="__module.encoder.layer.10.attention.self/aten::transpose/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="877,attn_output.43"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="1235" name="Constant_131958" type="Const" version="opset1"> + <data element_type="i64" shape="3" offset="12502416" size="24" /> + <output> + <port id="0" precision="I64"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="1236" name="__module.encoder.layer.10.attention.self/aten::reshape/Reshape" type="Reshape" version="opset1"> + <data special_zero="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="879"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1237" name="__module.encoder.layer.10.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/scale" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="30643080" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1238" name="__module.encoder.layer.10.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1239" name="__module.encoder.layer.10.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="30644616" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1240" name="__module.encoder.layer.10.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="30644620" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1241" name="__module.encoder.layer.10.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="30644616" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1242" name="__module.encoder.layer.10.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="30644620" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1243" name="__module.encoder.layer.10.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1244" name="Constant_578760" type="Const" version="opset1"> + <data element_type="i8" shape="384, 384" offset="30644624" size="147456" /> + <output> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1245" name="Convert_578761" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1246" name="Constant_578762" type="Const" version="opset1"> + <data element_type="f32" shape="384, 1" offset="30792080" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1247" name="__module.encoder.layer.10.attention.output.dense/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1248" name="__module.encoder.layer.10.attention.output.dense/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1249" name="Constant_131809" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="30793616" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1250" name="__module.encoder.layer.10.attention.output.dense/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="885,input.43"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1251" name="__module.encoder.layer.10.attention.output/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="887"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1252" name="__module.encoder.layer.10.attention.output.LayerNorm/aten::layer_norm/Multiply" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="12046084" size="4" /> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1253" name="__module.encoder.layer.10.attention.output.LayerNorm/aten::layer_norm/MVN" type="MVN" version="opset6"> + <data eps="9.999999960041972e-13" normalize_variance="true" eps_mode="INSIDE_SQRT" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1254" name="Constant_131810" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="30795152" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1255" name="__module.encoder.layer.10.attention.output.LayerNorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1256" name="Constant_131811" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="30796688" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1257" name="__module.encoder.layer.10.attention.output.LayerNorm/aten::layer_norm/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="891,input_tensor.21"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1258" name="__module.encoder.layer.10.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/scale" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="30798224" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1259" name="__module.encoder.layer.10.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1260" name="__module.encoder.layer.10.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="30799760" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1261" name="__module.encoder.layer.10.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="30799764" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1262" name="__module.encoder.layer.10.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="30799760" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1263" name="__module.encoder.layer.10.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="30799764" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1264" name="__module.encoder.layer.10.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1265" name="Constant_578764" type="Const" version="opset1"> + <data element_type="i8" shape="1536, 384" offset="30799768" size="589824" /> + <output> + <port id="0" precision="I8"> + <dim>1536</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1266" name="Convert_578765" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>1536</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1536</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1267" name="Constant_578766" type="Const" version="opset1"> + <data element_type="f32" shape="1536, 1" offset="31389592" size="6144" /> + <output> + <port id="0" precision="FP32"> + <dim>1536</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1268" name="__module.encoder.layer.10.intermediate.dense/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>1536</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1536</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>1536</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1269" name="__module.encoder.layer.10.intermediate.dense/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1536</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="1270" name="Constant_131812" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 1536" offset="31395736" size="6144" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="1271" name="__module.encoder.layer.10.intermediate.dense/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1536</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="896"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="1272" name="__module.encoder.layer.10.intermediate.intermediate_act_fn/aten::gelu/Gelu" type="Gelu" version="opset7"> + <data approximation_mode="ERF" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="897"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="1273" name="__module.encoder.layer.10.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/scale" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 1536" offset="31401880" size="6144" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="1274" name="__module.encoder.layer.10.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1536</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="1275" name="__module.encoder.layer.10.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="31408024" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1276" name="__module.encoder.layer.10.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="31408028" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1277" name="__module.encoder.layer.10.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="31408024" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1278" name="__module.encoder.layer.10.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="31408028" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1279" name="__module.encoder.layer.10.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="1280" name="Constant_578768" type="Const" version="opset1"> + <data element_type="i8" shape="384, 1536" offset="31408032" size="589824" /> + <output> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="1281" name="Convert_578769" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>1536</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="1282" name="Constant_578770" type="Const" version="opset1"> + <data element_type="f32" shape="384, 1" offset="31997856" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1283" name="__module.encoder.layer.10.output.dense/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1536</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>384</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="1284" name="__module.encoder.layer.10.output.dense/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1536</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1285" name="Constant_131813" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="31999392" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1286" name="__module.encoder.layer.10.output.dense/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="903,input.45"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1287" name="__module.encoder.layer.10.output/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="905"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1288" name="__module.encoder.layer.10.output.LayerNorm/aten::layer_norm/Multiply" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="12046084" size="4" /> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1289" name="__module.encoder.layer.10.output.LayerNorm/aten::layer_norm/MVN" type="MVN" version="opset6"> + <data eps="9.999999960041972e-13" normalize_variance="true" eps_mode="INSIDE_SQRT" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1290" name="Constant_131814" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="32000928" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1291" name="__module.encoder.layer.10.output.LayerNorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1292" name="Constant_131815" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="32002464" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1293" name="__module.encoder.layer.10.output.LayerNorm/aten::layer_norm/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="909,hidden_states.67"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1294" name="__module.encoder.layer.10.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/scale" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="32004000" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1295" name="__module.encoder.layer.10.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1296" name="__module.encoder.layer.10.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="32005536" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1297" name="__module.encoder.layer.10.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="32005540" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1298" name="__module.encoder.layer.10.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="32005536" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1299" name="__module.encoder.layer.10.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="32005540" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1300" name="__module.encoder.layer.10.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1301" name="Constant_578772" type="Const" version="opset1"> + <data element_type="i8" shape="384, 384" offset="32005544" size="147456" /> + <output> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1302" name="Convert_578773" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1303" name="Constant_578774" type="Const" version="opset1"> + <data element_type="f32" shape="384, 1" offset="32153000" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1304" name="__module.encoder.layer.11.attention.self.query/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1305" name="__module.encoder.layer.11.attention.self.query/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1306" name="Constant_131816" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="32154536" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1307" name="__module.encoder.layer.11.attention.self.query/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="922,x.133"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1308" name="__module.encoder.layer.11.attention.self.query/aten::linear/Add/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="32156072" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1309" name="__module.encoder.layer.11.attention.self.query/aten::linear/Add/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="32156076" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1310" name="__module.encoder.layer.11.attention.self.query/aten::linear/Add/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="32156072" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1311" name="__module.encoder.layer.11.attention.self.query/aten::linear/Add/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="32156076" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1312" name="__module.encoder.layer.11.attention.self.query/aten::linear/Add/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1313" name="__module.encoder.layer.11.attention.self/prim::ListConstruct/Concat" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201240" size="32" /> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1314" name="__module.encoder.layer.11.attention.self/aten::view/Reshape" type="Reshape" version="opset1"> + <data special_zero="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="926,x.135"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="1315" name="Constant_125073" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201272" size="32" /> + <output> + <port id="0" precision="I64" names="927"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1316" name="__module.encoder.layer.11.attention.self/aten::permute/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="928"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="1317" name="Constant_578776" type="Const" version="opset1"> + <data element_type="i8" shape="384, 384" offset="32156080" size="147456" /> + <output> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1318" name="Convert_578777" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1319" name="Constant_578778" type="Const" version="opset1"> + <data element_type="f32" shape="384, 1" offset="32303536" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1320" name="__module.encoder.layer.11.attention.self.key/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1321" name="__module.encoder.layer.11.attention.self.key/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1322" name="Constant_131817" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="32305072" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1323" name="__module.encoder.layer.11.attention.self.key/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="931,x.137"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1324" name="__module.encoder.layer.11.attention.self.key/aten::linear/Add/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="32306608" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1325" name="__module.encoder.layer.11.attention.self.key/aten::linear/Add/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="32306612" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1326" name="__module.encoder.layer.11.attention.self.key/aten::linear/Add/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="32306608" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1327" name="__module.encoder.layer.11.attention.self.key/aten::linear/Add/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="32306612" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1328" name="__module.encoder.layer.11.attention.self.key/aten::linear/Add/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1329" name="__module.encoder.layer.11.attention.self/prim::ListConstruct/Concat_1" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201240" size="32" /> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1330" name="__module.encoder.layer.11.attention.self/aten::view/Reshape_1" type="Reshape" version="opset1"> + <data special_zero="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="935,x.139"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="1331" name="Constant_125096" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201272" size="32" /> + <output> + <port id="0" precision="I64" names="936"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1332" name="__module.encoder.layer.11.attention.self/aten::permute/Transpose_1" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="937"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="1333" name="Constant_578780" type="Const" version="opset1"> + <data element_type="i8" shape="384, 384" offset="32306616" size="147456" /> + <output> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1334" name="Convert_578781" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1335" name="Constant_578782" type="Const" version="opset1"> + <data element_type="f32" shape="384, 1" offset="32454072" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1336" name="__module.encoder.layer.11.attention.self.value/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1337" name="__module.encoder.layer.11.attention.self.value/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1338" name="Constant_131818" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="32455608" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1339" name="__module.encoder.layer.11.attention.self.value/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="940,x.141"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1340" name="__module.encoder.layer.11.attention.self/prim::ListConstruct/Concat_2" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201240" size="32" /> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1341" name="__module.encoder.layer.11.attention.self/aten::view/Reshape_2" type="Reshape" version="opset1"> + <data special_zero="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="944,x"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="1342" name="Constant_125119" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="12201272" size="32" /> + <output> + <port id="0" precision="I64" names="945"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1343" name="__module.encoder.layer.11.attention.self/aten::permute/Transpose_2" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="946"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="1344" name="__module.encoder.layer.11.attention.self/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> + <data causal="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="947,attn_output.45"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="1345" name="__module.encoder.layer.11.attention.self/aten::transpose/ScatterElementsUpdate" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="12502400" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1346" name="__module.encoder.layer.11.attention.self/aten::transpose/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>12</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="948,attn_output"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="1347" name="Constant_131959" type="Const" version="opset1"> + <data element_type="i64" shape="3" offset="12502416" size="24" /> + <output> + <port id="0" precision="I64"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="1348" name="__module.encoder.layer.11.attention.self/aten::reshape/Reshape" type="Reshape" version="opset1"> + <data special_zero="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>12</dim> + <dim>32</dim> + </port> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="950"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1349" name="__module.encoder.layer.11.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/scale" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="32457144" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1350" name="__module.encoder.layer.11.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1351" name="__module.encoder.layer.11.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="32458680" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1352" name="__module.encoder.layer.11.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="32458684" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1353" name="__module.encoder.layer.11.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="32458680" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1354" name="__module.encoder.layer.11.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="32458684" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1355" name="__module.encoder.layer.11.attention.self/aten::reshape/Reshape_0_0/nncf_smooth_quant/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1356" name="Constant_578784" type="Const" version="opset1"> + <data element_type="i8" shape="384, 384" offset="32458688" size="147456" /> + <output> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1357" name="Convert_578785" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1358" name="Constant_578786" type="Const" version="opset1"> + <data element_type="f32" shape="384, 1" offset="32606144" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1359" name="__module.encoder.layer.11.attention.output.dense/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1360" name="__module.encoder.layer.11.attention.output.dense/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1361" name="Constant_131819" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="32607680" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1362" name="__module.encoder.layer.11.attention.output.dense/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="956,input.47"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1363" name="__module.encoder.layer.11.attention.output/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="958"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1364" name="__module.encoder.layer.11.attention.output.LayerNorm/aten::layer_norm/Multiply" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="12046084" size="4" /> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1365" name="__module.encoder.layer.11.attention.output.LayerNorm/aten::layer_norm/MVN" type="MVN" version="opset6"> + <data eps="9.999999960041972e-13" normalize_variance="true" eps_mode="INSIDE_SQRT" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1366" name="Constant_131820" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="32609216" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1367" name="__module.encoder.layer.11.attention.output.LayerNorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1368" name="Constant_131821" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="32610752" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1369" name="__module.encoder.layer.11.attention.output.LayerNorm/aten::layer_norm/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="962,input_tensor"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1370" name="__module.encoder.layer.11.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/scale" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="32612288" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1371" name="__module.encoder.layer.11.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1372" name="__module.encoder.layer.11.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="32613824" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1373" name="__module.encoder.layer.11.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="32613828" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1374" name="__module.encoder.layer.11.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="32613824" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1375" name="__module.encoder.layer.11.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="32613828" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1376" name="__module.encoder.layer.11.attention.output.LayerNorm/aten::layer_norm/Add_0_0/nncf_smooth_quant/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1377" name="Constant_578788" type="Const" version="opset1"> + <data element_type="i8" shape="1536, 384" offset="32613832" size="589824" /> + <output> + <port id="0" precision="I8"> + <dim>1536</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1378" name="Convert_578789" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>1536</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1536</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1379" name="Constant_578790" type="Const" version="opset1"> + <data element_type="f32" shape="1536, 1" offset="33203656" size="6144" /> + <output> + <port id="0" precision="FP32"> + <dim>1536</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1380" name="__module.encoder.layer.11.intermediate.dense/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>1536</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1536</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>1536</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1381" name="__module.encoder.layer.11.intermediate.dense/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1536</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="1382" name="Constant_131822" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 1536" offset="33209800" size="6144" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="1383" name="__module.encoder.layer.11.intermediate.dense/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1536</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="967"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="1384" name="__module.encoder.layer.11.intermediate.intermediate_act_fn/aten::gelu/Gelu" type="Gelu" version="opset7"> + <data approximation_mode="ERF" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="968"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="1385" name="__module.encoder.layer.11.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/scale" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 1536" offset="33215944" size="6144" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="1386" name="__module.encoder.layer.11.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1536</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="1387" name="__module.encoder.layer.11.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/fq_output_0/input_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="33222088" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1388" name="__module.encoder.layer.11.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/fq_output_0/input_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="33222092" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1389" name="__module.encoder.layer.11.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/fq_output_0/output_low" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="33222088" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1390" name="__module.encoder.layer.11.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/fq_output_0/output_high" type="Const" version="opset1"> + <data element_type="f32" shape="" offset="33222092" size="4" /> + <output> + <port id="0" precision="FP32" /> + </output> + </layer> + <layer id="1391" name="__module.encoder.layer.11.intermediate.intermediate_act_fn/aten::gelu/Gelu_0_0/nncf_smooth_quant/fq_output_0" type="FakeQuantize" version="opset1"> + <data levels="256" auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32" /> + <port id="3" precision="FP32" /> + <port id="4" precision="FP32" /> + </input> + <output> + <port id="5" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="1392" name="Constant_578792" type="Const" version="opset1"> + <data element_type="i8" shape="384, 1536" offset="33222096" size="589824" /> + <output> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="1393" name="Convert_578793" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I8"> + <dim>384</dim> + <dim>1536</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="1394" name="Constant_578794" type="Const" version="opset1"> + <data element_type="f32" shape="384, 1" offset="33811920" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1395" name="__module.encoder.layer.11.output.dense/aten::linear/MatMul/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>384</dim> + <dim>1536</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>384</dim> + <dim>1536</dim> + </port> + </output> + </layer> + <layer id="1396" name="__module.encoder.layer.11.output.dense/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1536</dim> + </port> + <port id="1" precision="FP32"> + <dim>384</dim> + <dim>1536</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1397" name="Constant_131823" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="33813456" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1398" name="__module.encoder.layer.11.output.dense/aten::linear/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="974,input"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1399" name="__module.encoder.layer.11.output/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="976"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1400" name="__module.encoder.layer.11.output.LayerNorm/aten::layer_norm/Multiply" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="12046084" size="4" /> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1401" name="__module.encoder.layer.11.output.LayerNorm/aten::layer_norm/MVN" type="MVN" version="opset6"> + <data eps="9.999999960041972e-13" normalize_variance="true" eps_mode="INSIDE_SQRT" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1402" name="Constant_131824" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="33814992" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1403" name="__module.encoder.layer.11.output.LayerNorm/aten::layer_norm/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1404" name="Constant_131825" type="Const" version="opset1"> + <data element_type="f32" shape="1, 1, 384" offset="33816528" size="1536" /> + <output> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1405" name="__module.encoder.layer.11.output.LayerNorm/aten::layer_norm/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>384</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="last_hidden_state"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </output> + </layer> + <layer id="1406" name="Result_127011" type="Result" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>384</dim> + </port> + </input> + </layer> + </layers> + <edges> + <edge from-layer="0" from-port="0" to-layer="14" to-port="0" /> + <edge from-layer="1" from-port="0" to-layer="93" to-port="0" /> + <edge from-layer="2" from-port="0" to-layer="7" to-port="0" /> + <edge from-layer="2" from-port="0" to-layer="24" to-port="0" /> + <edge from-layer="3" from-port="0" to-layer="4" to-port="0" /> + <edge from-layer="4" from-port="1" to-layer="6" to-port="0" /> + <edge from-layer="5" from-port="0" to-layer="6" to-port="1" /> + <edge from-layer="6" from-port="2" to-layer="9" to-port="0" /> + <edge from-layer="7" from-port="1" to-layer="9" to-port="1" /> + <edge from-layer="8" from-port="0" to-layer="9" to-port="2" /> + <edge from-layer="9" from-port="3" to-layer="17" to-port="0" /> + <edge from-layer="10" from-port="0" to-layer="11" to-port="0" /> + <edge from-layer="11" from-port="1" to-layer="13" to-port="0" /> + <edge from-layer="12" from-port="0" to-layer="13" to-port="1" /> + <edge from-layer="13" from-port="2" to-layer="16" to-port="0" /> + <edge from-layer="14" from-port="1" to-layer="16" to-port="1" /> + <edge from-layer="15" from-port="0" to-layer="16" to-port="2" /> + <edge from-layer="16" from-port="3" to-layer="17" to-port="1" /> + <edge from-layer="17" from-port="2" to-layer="34" to-port="0" /> + <edge from-layer="18" from-port="0" to-layer="19" to-port="0" /> + <edge from-layer="19" from-port="1" to-layer="21" to-port="0" /> + <edge from-layer="20" from-port="0" to-layer="21" to-port="1" /> + <edge from-layer="21" from-port="2" to-layer="33" to-port="0" /> + <edge from-layer="22" from-port="0" to-layer="30" to-port="0" /> + <edge from-layer="23" from-port="0" to-layer="30" to-port="1" /> + <edge from-layer="24" from-port="1" to-layer="27" to-port="0" /> + <edge from-layer="24" from-port="1" to-layer="98" to-port="0" /> + <edge from-layer="24" from-port="1" to-layer="102" to-port="0" /> + <edge from-layer="25" from-port="0" to-layer="27" to-port="1" /> + <edge from-layer="26" from-port="0" to-layer="27" to-port="2" /> + <edge from-layer="27" from-port="3" to-layer="30" to-port="2" /> + <edge from-layer="28" from-port="0" to-layer="30" to-port="3" /> + <edge from-layer="29" from-port="0" to-layer="30" to-port="4" /> + <edge from-layer="30" from-port="5" to-layer="31" to-port="0" /> + <edge from-layer="31" from-port="1" to-layer="33" to-port="1" /> + <edge from-layer="32" from-port="0" to-layer="33" to-port="2" /> + <edge from-layer="33" from-port="3" to-layer="34" to-port="1" /> + <edge from-layer="34" from-port="2" to-layer="36" to-port="0" /> + <edge from-layer="35" from-port="0" to-layer="36" to-port="1" /> + <edge from-layer="36" from-port="2" to-layer="38" to-port="0" /> + <edge from-layer="37" from-port="0" to-layer="38" to-port="1" /> + <edge from-layer="38" from-port="2" to-layer="40" to-port="0" /> + <edge from-layer="39" from-port="0" to-layer="40" to-port="1" /> + <edge from-layer="40" from-port="2" to-layer="131" to-port="1" /> + <edge from-layer="40" from-port="2" to-layer="42" to-port="0" /> + <edge from-layer="41" from-port="0" to-layer="42" to-port="1" /> + <edge from-layer="42" from-port="2" to-layer="47" to-port="0" /> + <edge from-layer="43" from-port="0" to-layer="47" to-port="1" /> + <edge from-layer="44" from-port="0" to-layer="47" to-port="2" /> + <edge from-layer="45" from-port="0" to-layer="47" to-port="3" /> + <edge from-layer="46" from-port="0" to-layer="47" to-port="4" /> + <edge from-layer="47" from-port="5" to-layer="84" to-port="0" /> + <edge from-layer="47" from-port="5" to-layer="52" to-port="0" /> + <edge from-layer="47" from-port="5" to-layer="68" to-port="0" /> + <edge from-layer="48" from-port="0" to-layer="49" to-port="0" /> + <edge from-layer="49" from-port="1" to-layer="51" to-port="0" /> + <edge from-layer="50" from-port="0" to-layer="51" to-port="1" /> + <edge from-layer="51" from-port="2" to-layer="52" to-port="1" /> + <edge from-layer="52" from-port="2" to-layer="54" to-port="0" /> + <edge from-layer="53" from-port="0" to-layer="54" to-port="1" /> + <edge from-layer="54" from-port="2" to-layer="59" to-port="0" /> + <edge from-layer="55" from-port="0" to-layer="59" to-port="1" /> + <edge from-layer="56" from-port="0" to-layer="59" to-port="2" /> + <edge from-layer="57" from-port="0" to-layer="59" to-port="3" /> + <edge from-layer="58" from-port="0" to-layer="59" to-port="4" /> + <edge from-layer="59" from-port="5" to-layer="61" to-port="0" /> + <edge from-layer="60" from-port="0" to-layer="61" to-port="1" /> + <edge from-layer="61" from-port="2" to-layer="63" to-port="0" /> + <edge from-layer="62" from-port="0" to-layer="63" to-port="1" /> + <edge from-layer="63" from-port="2" to-layer="112" to-port="0" /> + <edge from-layer="64" from-port="0" to-layer="65" to-port="0" /> + <edge from-layer="65" from-port="1" to-layer="67" to-port="0" /> + <edge from-layer="66" from-port="0" to-layer="67" to-port="1" /> + <edge from-layer="67" from-port="2" to-layer="68" to-port="1" /> + <edge from-layer="68" from-port="2" to-layer="70" to-port="0" /> + <edge from-layer="69" from-port="0" to-layer="70" to-port="1" /> + <edge from-layer="70" from-port="2" to-layer="75" to-port="0" /> + <edge from-layer="71" from-port="0" to-layer="75" to-port="1" /> + <edge from-layer="72" from-port="0" to-layer="75" to-port="2" /> + <edge from-layer="73" from-port="0" to-layer="75" to-port="3" /> + <edge from-layer="74" from-port="0" to-layer="75" to-port="4" /> + <edge from-layer="75" from-port="5" to-layer="77" to-port="0" /> + <edge from-layer="76" from-port="0" to-layer="77" to-port="1" /> + <edge from-layer="77" from-port="2" to-layer="79" to-port="0" /> + <edge from-layer="78" from-port="0" to-layer="79" to-port="1" /> + <edge from-layer="79" from-port="2" to-layer="112" to-port="1" /> + <edge from-layer="80" from-port="0" to-layer="81" to-port="0" /> + <edge from-layer="81" from-port="1" to-layer="83" to-port="0" /> + <edge from-layer="82" from-port="0" to-layer="83" to-port="1" /> + <edge from-layer="83" from-port="2" to-layer="84" to-port="1" /> + <edge from-layer="84" from-port="2" to-layer="86" to-port="0" /> + <edge from-layer="85" from-port="0" to-layer="86" to-port="1" /> + <edge from-layer="86" from-port="2" to-layer="88" to-port="0" /> + <edge from-layer="87" from-port="0" to-layer="88" to-port="1" /> + <edge from-layer="88" from-port="2" to-layer="90" to-port="0" /> + <edge from-layer="89" from-port="0" to-layer="90" to-port="1" /> + <edge from-layer="90" from-port="2" to-layer="112" to-port="2" /> + <edge from-layer="91" from-port="0" to-layer="108" to-port="0" /> + <edge from-layer="92" from-port="0" to-layer="93" to-port="1" /> + <edge from-layer="93" from-port="2" to-layer="95" to-port="0" /> + <edge from-layer="94" from-port="0" to-layer="95" to-port="1" /> + <edge from-layer="95" from-port="2" to-layer="104" to-port="0" /> + <edge from-layer="96" from-port="0" to-layer="98" to-port="1" /> + <edge from-layer="97" from-port="0" to-layer="98" to-port="2" /> + <edge from-layer="98" from-port="3" to-layer="103" to-port="0" /> + <edge from-layer="99" from-port="0" to-layer="103" to-port="1" /> + <edge from-layer="100" from-port="0" to-layer="102" to-port="1" /> + <edge from-layer="101" from-port="0" to-layer="102" to-port="2" /> + <edge from-layer="102" from-port="3" to-layer="103" to-port="2" /> + <edge from-layer="103" from-port="3" to-layer="104" to-port="1" /> + <edge from-layer="104" from-port="2" to-layer="105" to-port="0" /> + <edge from-layer="105" from-port="1" to-layer="107" to-port="0" /> + <edge from-layer="106" from-port="0" to-layer="107" to-port="1" /> + <edge from-layer="107" from-port="2" to-layer="108" to-port="1" /> + <edge from-layer="108" from-port="2" to-layer="109" to-port="0" /> + <edge from-layer="108" from-port="2" to-layer="111" to-port="2" /> + <edge from-layer="109" from-port="1" to-layer="111" to-port="0" /> + <edge from-layer="110" from-port="0" to-layer="111" to-port="1" /> + <edge from-layer="111" from-port="3" to-layer="448" to-port="3" /> + <edge from-layer="111" from-port="3" to-layer="560" to-port="3" /> + <edge from-layer="111" from-port="3" to-layer="672" to-port="3" /> + <edge from-layer="111" from-port="3" to-layer="1344" to-port="3" /> + <edge from-layer="111" from-port="3" to-layer="1232" to-port="3" /> + <edge from-layer="111" from-port="3" to-layer="1120" to-port="3" /> + <edge from-layer="111" from-port="3" to-layer="1008" to-port="3" /> + <edge from-layer="111" from-port="3" to-layer="896" to-port="3" /> + <edge from-layer="111" from-port="3" to-layer="784" to-port="3" /> + <edge from-layer="111" from-port="3" to-layer="224" to-port="3" /> + <edge from-layer="111" from-port="3" to-layer="336" to-port="3" /> + <edge from-layer="111" from-port="3" to-layer="112" to-port="3" /> + <edge from-layer="112" from-port="4" to-layer="114" to-port="0" /> + <edge from-layer="113" from-port="0" to-layer="114" to-port="1" /> + <edge from-layer="114" from-port="2" to-layer="116" to-port="0" /> + <edge from-layer="115" from-port="0" to-layer="116" to-port="1" /> + <edge from-layer="116" from-port="2" to-layer="118" to-port="0" /> + <edge from-layer="117" from-port="0" to-layer="118" to-port="1" /> + <edge from-layer="118" from-port="2" to-layer="123" to-port="0" /> + <edge from-layer="119" from-port="0" to-layer="123" to-port="1" /> + <edge from-layer="120" from-port="0" to-layer="123" to-port="2" /> + <edge from-layer="121" from-port="0" to-layer="123" to-port="3" /> + <edge from-layer="122" from-port="0" to-layer="123" to-port="4" /> + <edge from-layer="123" from-port="5" to-layer="128" to-port="0" /> + <edge from-layer="124" from-port="0" to-layer="125" to-port="0" /> + <edge from-layer="125" from-port="1" to-layer="127" to-port="0" /> + <edge from-layer="126" from-port="0" to-layer="127" to-port="1" /> + <edge from-layer="127" from-port="2" to-layer="128" to-port="1" /> + <edge from-layer="128" from-port="2" to-layer="130" to-port="0" /> + <edge from-layer="129" from-port="0" to-layer="130" to-port="1" /> + <edge from-layer="130" from-port="2" to-layer="131" to-port="0" /> + <edge from-layer="131" from-port="2" to-layer="133" to-port="0" /> + <edge from-layer="132" from-port="0" to-layer="133" to-port="1" /> + <edge from-layer="133" from-port="2" to-layer="135" to-port="0" /> + <edge from-layer="134" from-port="0" to-layer="135" to-port="1" /> + <edge from-layer="135" from-port="2" to-layer="137" to-port="0" /> + <edge from-layer="136" from-port="0" to-layer="137" to-port="1" /> + <edge from-layer="137" from-port="2" to-layer="139" to-port="0" /> + <edge from-layer="137" from-port="2" to-layer="167" to-port="1" /> + <edge from-layer="138" from-port="0" to-layer="139" to-port="1" /> + <edge from-layer="139" from-port="2" to-layer="144" to-port="0" /> + <edge from-layer="140" from-port="0" to-layer="144" to-port="1" /> + <edge from-layer="141" from-port="0" to-layer="144" to-port="2" /> + <edge from-layer="142" from-port="0" to-layer="144" to-port="3" /> + <edge from-layer="143" from-port="0" to-layer="144" to-port="4" /> + <edge from-layer="144" from-port="5" to-layer="149" to-port="0" /> + <edge from-layer="145" from-port="0" to-layer="146" to-port="0" /> + <edge from-layer="146" from-port="1" to-layer="148" to-port="0" /> + <edge from-layer="147" from-port="0" to-layer="148" to-port="1" /> + <edge from-layer="148" from-port="2" to-layer="149" to-port="1" /> + <edge from-layer="149" from-port="2" to-layer="151" to-port="0" /> + <edge from-layer="150" from-port="0" to-layer="151" to-port="1" /> + <edge from-layer="151" from-port="2" to-layer="152" to-port="0" /> + <edge from-layer="152" from-port="1" to-layer="154" to-port="0" /> + <edge from-layer="153" from-port="0" to-layer="154" to-port="1" /> + <edge from-layer="154" from-port="2" to-layer="159" to-port="0" /> + <edge from-layer="155" from-port="0" to-layer="159" to-port="1" /> + <edge from-layer="156" from-port="0" to-layer="159" to-port="2" /> + <edge from-layer="157" from-port="0" to-layer="159" to-port="3" /> + <edge from-layer="158" from-port="0" to-layer="159" to-port="4" /> + <edge from-layer="159" from-port="5" to-layer="164" to-port="0" /> + <edge from-layer="160" from-port="0" to-layer="161" to-port="0" /> + <edge from-layer="161" from-port="1" to-layer="163" to-port="0" /> + <edge from-layer="162" from-port="0" to-layer="163" to-port="1" /> + <edge from-layer="163" from-port="2" to-layer="164" to-port="1" /> + <edge from-layer="164" from-port="2" to-layer="166" to-port="0" /> + <edge from-layer="165" from-port="0" to-layer="166" to-port="1" /> + <edge from-layer="166" from-port="2" to-layer="167" to-port="0" /> + <edge from-layer="167" from-port="2" to-layer="169" to-port="0" /> + <edge from-layer="168" from-port="0" to-layer="169" to-port="1" /> + <edge from-layer="169" from-port="2" to-layer="171" to-port="0" /> + <edge from-layer="170" from-port="0" to-layer="171" to-port="1" /> + <edge from-layer="171" from-port="2" to-layer="173" to-port="0" /> + <edge from-layer="172" from-port="0" to-layer="173" to-port="1" /> + <edge from-layer="173" from-port="2" to-layer="175" to-port="0" /> + <edge from-layer="173" from-port="2" to-layer="243" to-port="1" /> + <edge from-layer="174" from-port="0" to-layer="175" to-port="1" /> + <edge from-layer="175" from-port="2" to-layer="180" to-port="0" /> + <edge from-layer="176" from-port="0" to-layer="180" to-port="1" /> + <edge from-layer="177" from-port="0" to-layer="180" to-port="2" /> + <edge from-layer="178" from-port="0" to-layer="180" to-port="3" /> + <edge from-layer="179" from-port="0" to-layer="180" to-port="4" /> + <edge from-layer="180" from-port="5" to-layer="185" to-port="0" /> + <edge from-layer="180" from-port="5" to-layer="217" to-port="0" /> + <edge from-layer="180" from-port="5" to-layer="201" to-port="0" /> + <edge from-layer="181" from-port="0" to-layer="182" to-port="0" /> + <edge from-layer="182" from-port="1" to-layer="184" to-port="0" /> + <edge from-layer="183" from-port="0" to-layer="184" to-port="1" /> + <edge from-layer="184" from-port="2" to-layer="185" to-port="1" /> + <edge from-layer="185" from-port="2" to-layer="187" to-port="0" /> + <edge from-layer="186" from-port="0" to-layer="187" to-port="1" /> + <edge from-layer="187" from-port="2" to-layer="192" to-port="0" /> + <edge from-layer="188" from-port="0" to-layer="192" to-port="1" /> + <edge from-layer="189" from-port="0" to-layer="192" to-port="2" /> + <edge from-layer="190" from-port="0" to-layer="192" to-port="3" /> + <edge from-layer="191" from-port="0" to-layer="192" to-port="4" /> + <edge from-layer="192" from-port="5" to-layer="194" to-port="0" /> + <edge from-layer="193" from-port="0" to-layer="194" to-port="1" /> + <edge from-layer="194" from-port="2" to-layer="196" to-port="0" /> + <edge from-layer="195" from-port="0" to-layer="196" to-port="1" /> + <edge from-layer="196" from-port="2" to-layer="224" to-port="0" /> + <edge from-layer="197" from-port="0" to-layer="198" to-port="0" /> + <edge from-layer="198" from-port="1" to-layer="200" to-port="0" /> + <edge from-layer="199" from-port="0" to-layer="200" to-port="1" /> + <edge from-layer="200" from-port="2" to-layer="201" to-port="1" /> + <edge from-layer="201" from-port="2" to-layer="203" to-port="0" /> + <edge from-layer="202" from-port="0" to-layer="203" to-port="1" /> + <edge from-layer="203" from-port="2" to-layer="208" to-port="0" /> + <edge from-layer="204" from-port="0" to-layer="208" to-port="1" /> + <edge from-layer="205" from-port="0" to-layer="208" to-port="2" /> + <edge from-layer="206" from-port="0" to-layer="208" to-port="3" /> + <edge from-layer="207" from-port="0" to-layer="208" to-port="4" /> + <edge from-layer="208" from-port="5" to-layer="210" to-port="0" /> + <edge from-layer="209" from-port="0" to-layer="210" to-port="1" /> + <edge from-layer="210" from-port="2" to-layer="212" to-port="0" /> + <edge from-layer="211" from-port="0" to-layer="212" to-port="1" /> + <edge from-layer="212" from-port="2" to-layer="224" to-port="1" /> + <edge from-layer="213" from-port="0" to-layer="214" to-port="0" /> + <edge from-layer="214" from-port="1" to-layer="216" to-port="0" /> + <edge from-layer="215" from-port="0" to-layer="216" to-port="1" /> + <edge from-layer="216" from-port="2" to-layer="217" to-port="1" /> + <edge from-layer="217" from-port="2" to-layer="219" to-port="0" /> + <edge from-layer="218" from-port="0" to-layer="219" to-port="1" /> + <edge from-layer="219" from-port="2" to-layer="221" to-port="0" /> + <edge from-layer="220" from-port="0" to-layer="221" to-port="1" /> + <edge from-layer="221" from-port="2" to-layer="223" to-port="0" /> + <edge from-layer="222" from-port="0" to-layer="223" to-port="1" /> + <edge from-layer="223" from-port="2" to-layer="224" to-port="2" /> + <edge from-layer="224" from-port="4" to-layer="226" to-port="0" /> + <edge from-layer="225" from-port="0" to-layer="226" to-port="1" /> + <edge from-layer="226" from-port="2" to-layer="228" to-port="0" /> + <edge from-layer="227" from-port="0" to-layer="228" to-port="1" /> + <edge from-layer="228" from-port="2" to-layer="230" to-port="0" /> + <edge from-layer="229" from-port="0" to-layer="230" to-port="1" /> + <edge from-layer="230" from-port="2" to-layer="235" to-port="0" /> + <edge from-layer="231" from-port="0" to-layer="235" to-port="1" /> + <edge from-layer="232" from-port="0" to-layer="235" to-port="2" /> + <edge from-layer="233" from-port="0" to-layer="235" to-port="3" /> + <edge from-layer="234" from-port="0" to-layer="235" to-port="4" /> + <edge from-layer="235" from-port="5" to-layer="240" to-port="0" /> + <edge from-layer="236" from-port="0" to-layer="237" to-port="0" /> + <edge from-layer="237" from-port="1" to-layer="239" to-port="0" /> + <edge from-layer="238" from-port="0" to-layer="239" to-port="1" /> + <edge from-layer="239" from-port="2" to-layer="240" to-port="1" /> + <edge from-layer="240" from-port="2" to-layer="242" to-port="0" /> + <edge from-layer="241" from-port="0" to-layer="242" to-port="1" /> + <edge from-layer="242" from-port="2" to-layer="243" to-port="0" /> + <edge from-layer="243" from-port="2" to-layer="245" to-port="0" /> + <edge from-layer="244" from-port="0" to-layer="245" to-port="1" /> + <edge from-layer="245" from-port="2" to-layer="247" to-port="0" /> + <edge from-layer="246" from-port="0" to-layer="247" to-port="1" /> + <edge from-layer="247" from-port="2" to-layer="249" to-port="0" /> + <edge from-layer="248" from-port="0" to-layer="249" to-port="1" /> + <edge from-layer="249" from-port="2" to-layer="251" to-port="0" /> + <edge from-layer="249" from-port="2" to-layer="279" to-port="1" /> + <edge from-layer="250" from-port="0" to-layer="251" to-port="1" /> + <edge from-layer="251" from-port="2" to-layer="256" to-port="0" /> + <edge from-layer="252" from-port="0" to-layer="256" to-port="1" /> + <edge from-layer="253" from-port="0" to-layer="256" to-port="2" /> + <edge from-layer="254" from-port="0" to-layer="256" to-port="3" /> + <edge from-layer="255" from-port="0" to-layer="256" to-port="4" /> + <edge from-layer="256" from-port="5" to-layer="261" to-port="0" /> + <edge from-layer="257" from-port="0" to-layer="258" to-port="0" /> + <edge from-layer="258" from-port="1" to-layer="260" to-port="0" /> + <edge from-layer="259" from-port="0" to-layer="260" to-port="1" /> + <edge from-layer="260" from-port="2" to-layer="261" to-port="1" /> + <edge from-layer="261" from-port="2" to-layer="263" to-port="0" /> + <edge from-layer="262" from-port="0" to-layer="263" to-port="1" /> + <edge from-layer="263" from-port="2" to-layer="264" to-port="0" /> + <edge from-layer="264" from-port="1" to-layer="266" to-port="0" /> + <edge from-layer="265" from-port="0" to-layer="266" to-port="1" /> + <edge from-layer="266" from-port="2" to-layer="271" to-port="0" /> + <edge from-layer="267" from-port="0" to-layer="271" to-port="1" /> + <edge from-layer="268" from-port="0" to-layer="271" to-port="2" /> + <edge from-layer="269" from-port="0" to-layer="271" to-port="3" /> + <edge from-layer="270" from-port="0" to-layer="271" to-port="4" /> + <edge from-layer="271" from-port="5" to-layer="276" to-port="0" /> + <edge from-layer="272" from-port="0" to-layer="273" to-port="0" /> + <edge from-layer="273" from-port="1" to-layer="275" to-port="0" /> + <edge from-layer="274" from-port="0" to-layer="275" to-port="1" /> + <edge from-layer="275" from-port="2" to-layer="276" to-port="1" /> + <edge from-layer="276" from-port="2" to-layer="278" to-port="0" /> + <edge from-layer="277" from-port="0" to-layer="278" to-port="1" /> + <edge from-layer="278" from-port="2" to-layer="279" to-port="0" /> + <edge from-layer="279" from-port="2" to-layer="281" to-port="0" /> + <edge from-layer="280" from-port="0" to-layer="281" to-port="1" /> + <edge from-layer="281" from-port="2" to-layer="283" to-port="0" /> + <edge from-layer="282" from-port="0" to-layer="283" to-port="1" /> + <edge from-layer="283" from-port="2" to-layer="285" to-port="0" /> + <edge from-layer="284" from-port="0" to-layer="285" to-port="1" /> + <edge from-layer="285" from-port="2" to-layer="355" to-port="1" /> + <edge from-layer="285" from-port="2" to-layer="287" to-port="0" /> + <edge from-layer="286" from-port="0" to-layer="287" to-port="1" /> + <edge from-layer="287" from-port="2" to-layer="292" to-port="0" /> + <edge from-layer="288" from-port="0" to-layer="292" to-port="1" /> + <edge from-layer="289" from-port="0" to-layer="292" to-port="2" /> + <edge from-layer="290" from-port="0" to-layer="292" to-port="3" /> + <edge from-layer="291" from-port="0" to-layer="292" to-port="4" /> + <edge from-layer="292" from-port="5" to-layer="297" to-port="0" /> + <edge from-layer="292" from-port="5" to-layer="329" to-port="0" /> + <edge from-layer="292" from-port="5" to-layer="313" to-port="0" /> + <edge from-layer="293" from-port="0" to-layer="294" to-port="0" /> + <edge from-layer="294" from-port="1" to-layer="296" to-port="0" /> + <edge from-layer="295" from-port="0" to-layer="296" to-port="1" /> + <edge from-layer="296" from-port="2" to-layer="297" to-port="1" /> + <edge from-layer="297" from-port="2" to-layer="299" to-port="0" /> + <edge from-layer="298" from-port="0" to-layer="299" to-port="1" /> + <edge from-layer="299" from-port="2" to-layer="304" to-port="0" /> + <edge from-layer="300" from-port="0" to-layer="304" to-port="1" /> + <edge from-layer="301" from-port="0" to-layer="304" to-port="2" /> + <edge from-layer="302" from-port="0" to-layer="304" to-port="3" /> + <edge from-layer="303" from-port="0" to-layer="304" to-port="4" /> + <edge from-layer="304" from-port="5" to-layer="306" to-port="0" /> + <edge from-layer="305" from-port="0" to-layer="306" to-port="1" /> + <edge from-layer="306" from-port="2" to-layer="308" to-port="0" /> + <edge from-layer="307" from-port="0" to-layer="308" to-port="1" /> + <edge from-layer="308" from-port="2" to-layer="336" to-port="0" /> + <edge from-layer="309" from-port="0" to-layer="310" to-port="0" /> + <edge from-layer="310" from-port="1" to-layer="312" to-port="0" /> + <edge from-layer="311" from-port="0" to-layer="312" to-port="1" /> + <edge from-layer="312" from-port="2" to-layer="313" to-port="1" /> + <edge from-layer="313" from-port="2" to-layer="315" to-port="0" /> + <edge from-layer="314" from-port="0" to-layer="315" to-port="1" /> + <edge from-layer="315" from-port="2" to-layer="320" to-port="0" /> + <edge from-layer="316" from-port="0" to-layer="320" to-port="1" /> + <edge from-layer="317" from-port="0" to-layer="320" to-port="2" /> + <edge from-layer="318" from-port="0" to-layer="320" to-port="3" /> + <edge from-layer="319" from-port="0" to-layer="320" to-port="4" /> + <edge from-layer="320" from-port="5" to-layer="322" to-port="0" /> + <edge from-layer="321" from-port="0" to-layer="322" to-port="1" /> + <edge from-layer="322" from-port="2" to-layer="324" to-port="0" /> + <edge from-layer="323" from-port="0" to-layer="324" to-port="1" /> + <edge from-layer="324" from-port="2" to-layer="336" to-port="1" /> + <edge from-layer="325" from-port="0" to-layer="326" to-port="0" /> + <edge from-layer="326" from-port="1" to-layer="328" to-port="0" /> + <edge from-layer="327" from-port="0" to-layer="328" to-port="1" /> + <edge from-layer="328" from-port="2" to-layer="329" to-port="1" /> + <edge from-layer="329" from-port="2" to-layer="331" to-port="0" /> + <edge from-layer="330" from-port="0" to-layer="331" to-port="1" /> + <edge from-layer="331" from-port="2" to-layer="333" to-port="0" /> + <edge from-layer="332" from-port="0" to-layer="333" to-port="1" /> + <edge from-layer="333" from-port="2" to-layer="335" to-port="0" /> + <edge from-layer="334" from-port="0" to-layer="335" to-port="1" /> + <edge from-layer="335" from-port="2" to-layer="336" to-port="2" /> + <edge from-layer="336" from-port="4" to-layer="338" to-port="0" /> + <edge from-layer="337" from-port="0" to-layer="338" to-port="1" /> + <edge from-layer="338" from-port="2" to-layer="340" to-port="0" /> + <edge from-layer="339" from-port="0" to-layer="340" to-port="1" /> + <edge from-layer="340" from-port="2" to-layer="342" to-port="0" /> + <edge from-layer="341" from-port="0" to-layer="342" to-port="1" /> + <edge from-layer="342" from-port="2" to-layer="347" to-port="0" /> + <edge from-layer="343" from-port="0" to-layer="347" to-port="1" /> + <edge from-layer="344" from-port="0" to-layer="347" to-port="2" /> + <edge from-layer="345" from-port="0" to-layer="347" to-port="3" /> + <edge from-layer="346" from-port="0" to-layer="347" to-port="4" /> + <edge from-layer="347" from-port="5" to-layer="352" to-port="0" /> + <edge from-layer="348" from-port="0" to-layer="349" to-port="0" /> + <edge from-layer="349" from-port="1" to-layer="351" to-port="0" /> + <edge from-layer="350" from-port="0" to-layer="351" to-port="1" /> + <edge from-layer="351" from-port="2" to-layer="352" to-port="1" /> + <edge from-layer="352" from-port="2" to-layer="354" to-port="0" /> + <edge from-layer="353" from-port="0" to-layer="354" to-port="1" /> + <edge from-layer="354" from-port="2" to-layer="355" to-port="0" /> + <edge from-layer="355" from-port="2" to-layer="357" to-port="0" /> + <edge from-layer="356" from-port="0" to-layer="357" to-port="1" /> + <edge from-layer="357" from-port="2" to-layer="359" to-port="0" /> + <edge from-layer="358" from-port="0" to-layer="359" to-port="1" /> + <edge from-layer="359" from-port="2" to-layer="361" to-port="0" /> + <edge from-layer="360" from-port="0" to-layer="361" to-port="1" /> + <edge from-layer="361" from-port="2" to-layer="363" to-port="0" /> + <edge from-layer="361" from-port="2" to-layer="391" to-port="1" /> + <edge from-layer="362" from-port="0" to-layer="363" to-port="1" /> + <edge from-layer="363" from-port="2" to-layer="368" to-port="0" /> + <edge from-layer="364" from-port="0" to-layer="368" to-port="1" /> + <edge from-layer="365" from-port="0" to-layer="368" to-port="2" /> + <edge from-layer="366" from-port="0" to-layer="368" to-port="3" /> + <edge from-layer="367" from-port="0" to-layer="368" to-port="4" /> + <edge from-layer="368" from-port="5" to-layer="373" to-port="0" /> + <edge from-layer="369" from-port="0" to-layer="370" to-port="0" /> + <edge from-layer="370" from-port="1" to-layer="372" to-port="0" /> + <edge from-layer="371" from-port="0" to-layer="372" to-port="1" /> + <edge from-layer="372" from-port="2" to-layer="373" to-port="1" /> + <edge from-layer="373" from-port="2" to-layer="375" to-port="0" /> + <edge from-layer="374" from-port="0" to-layer="375" to-port="1" /> + <edge from-layer="375" from-port="2" to-layer="376" to-port="0" /> + <edge from-layer="376" from-port="1" to-layer="378" to-port="0" /> + <edge from-layer="377" from-port="0" to-layer="378" to-port="1" /> + <edge from-layer="378" from-port="2" to-layer="383" to-port="0" /> + <edge from-layer="379" from-port="0" to-layer="383" to-port="1" /> + <edge from-layer="380" from-port="0" to-layer="383" to-port="2" /> + <edge from-layer="381" from-port="0" to-layer="383" to-port="3" /> + <edge from-layer="382" from-port="0" to-layer="383" to-port="4" /> + <edge from-layer="383" from-port="5" to-layer="388" to-port="0" /> + <edge from-layer="384" from-port="0" to-layer="385" to-port="0" /> + <edge from-layer="385" from-port="1" to-layer="387" to-port="0" /> + <edge from-layer="386" from-port="0" to-layer="387" to-port="1" /> + <edge from-layer="387" from-port="2" to-layer="388" to-port="1" /> + <edge from-layer="388" from-port="2" to-layer="390" to-port="0" /> + <edge from-layer="389" from-port="0" to-layer="390" to-port="1" /> + <edge from-layer="390" from-port="2" to-layer="391" to-port="0" /> + <edge from-layer="391" from-port="2" to-layer="393" to-port="0" /> + <edge from-layer="392" from-port="0" to-layer="393" to-port="1" /> + <edge from-layer="393" from-port="2" to-layer="395" to-port="0" /> + <edge from-layer="394" from-port="0" to-layer="395" to-port="1" /> + <edge from-layer="395" from-port="2" to-layer="397" to-port="0" /> + <edge from-layer="396" from-port="0" to-layer="397" to-port="1" /> + <edge from-layer="397" from-port="2" to-layer="467" to-port="1" /> + <edge from-layer="397" from-port="2" to-layer="399" to-port="0" /> + <edge from-layer="398" from-port="0" to-layer="399" to-port="1" /> + <edge from-layer="399" from-port="2" to-layer="404" to-port="0" /> + <edge from-layer="400" from-port="0" to-layer="404" to-port="1" /> + <edge from-layer="401" from-port="0" to-layer="404" to-port="2" /> + <edge from-layer="402" from-port="0" to-layer="404" to-port="3" /> + <edge from-layer="403" from-port="0" to-layer="404" to-port="4" /> + <edge from-layer="404" from-port="5" to-layer="409" to-port="0" /> + <edge from-layer="404" from-port="5" to-layer="441" to-port="0" /> + <edge from-layer="404" from-port="5" to-layer="425" to-port="0" /> + <edge from-layer="405" from-port="0" to-layer="406" to-port="0" /> + <edge from-layer="406" from-port="1" to-layer="408" to-port="0" /> + <edge from-layer="407" from-port="0" to-layer="408" to-port="1" /> + <edge from-layer="408" from-port="2" to-layer="409" to-port="1" /> + <edge from-layer="409" from-port="2" to-layer="411" to-port="0" /> + <edge from-layer="410" from-port="0" to-layer="411" to-port="1" /> + <edge from-layer="411" from-port="2" to-layer="416" to-port="0" /> + <edge from-layer="412" from-port="0" to-layer="416" to-port="1" /> + <edge from-layer="413" from-port="0" to-layer="416" to-port="2" /> + <edge from-layer="414" from-port="0" to-layer="416" to-port="3" /> + <edge from-layer="415" from-port="0" to-layer="416" to-port="4" /> + <edge from-layer="416" from-port="5" to-layer="418" to-port="0" /> + <edge from-layer="417" from-port="0" to-layer="418" to-port="1" /> + <edge from-layer="418" from-port="2" to-layer="420" to-port="0" /> + <edge from-layer="419" from-port="0" to-layer="420" to-port="1" /> + <edge from-layer="420" from-port="2" to-layer="448" to-port="0" /> + <edge from-layer="421" from-port="0" to-layer="422" to-port="0" /> + <edge from-layer="422" from-port="1" to-layer="424" to-port="0" /> + <edge from-layer="423" from-port="0" to-layer="424" to-port="1" /> + <edge from-layer="424" from-port="2" to-layer="425" to-port="1" /> + <edge from-layer="425" from-port="2" to-layer="427" to-port="0" /> + <edge from-layer="426" from-port="0" to-layer="427" to-port="1" /> + <edge from-layer="427" from-port="2" to-layer="432" to-port="0" /> + <edge from-layer="428" from-port="0" to-layer="432" to-port="1" /> + <edge from-layer="429" from-port="0" to-layer="432" to-port="2" /> + <edge from-layer="430" from-port="0" to-layer="432" to-port="3" /> + <edge from-layer="431" from-port="0" to-layer="432" to-port="4" /> + <edge from-layer="432" from-port="5" to-layer="434" to-port="0" /> + <edge from-layer="433" from-port="0" to-layer="434" to-port="1" /> + <edge from-layer="434" from-port="2" to-layer="436" to-port="0" /> + <edge from-layer="435" from-port="0" to-layer="436" to-port="1" /> + <edge from-layer="436" from-port="2" to-layer="448" to-port="1" /> + <edge from-layer="437" from-port="0" to-layer="438" to-port="0" /> + <edge from-layer="438" from-port="1" to-layer="440" to-port="0" /> + <edge from-layer="439" from-port="0" to-layer="440" to-port="1" /> + <edge from-layer="440" from-port="2" to-layer="441" to-port="1" /> + <edge from-layer="441" from-port="2" to-layer="443" to-port="0" /> + <edge from-layer="442" from-port="0" to-layer="443" to-port="1" /> + <edge from-layer="443" from-port="2" to-layer="445" to-port="0" /> + <edge from-layer="444" from-port="0" to-layer="445" to-port="1" /> + <edge from-layer="445" from-port="2" to-layer="447" to-port="0" /> + <edge from-layer="446" from-port="0" to-layer="447" to-port="1" /> + <edge from-layer="447" from-port="2" to-layer="448" to-port="2" /> + <edge from-layer="448" from-port="4" to-layer="450" to-port="0" /> + <edge from-layer="449" from-port="0" to-layer="450" to-port="1" /> + <edge from-layer="450" from-port="2" to-layer="452" to-port="0" /> + <edge from-layer="451" from-port="0" to-layer="452" to-port="1" /> + <edge from-layer="452" from-port="2" to-layer="454" to-port="0" /> + <edge from-layer="453" from-port="0" to-layer="454" to-port="1" /> + <edge from-layer="454" from-port="2" to-layer="459" to-port="0" /> + <edge from-layer="455" from-port="0" to-layer="459" to-port="1" /> + <edge from-layer="456" from-port="0" to-layer="459" to-port="2" /> + <edge from-layer="457" from-port="0" to-layer="459" to-port="3" /> + <edge from-layer="458" from-port="0" to-layer="459" to-port="4" /> + <edge from-layer="459" from-port="5" to-layer="464" to-port="0" /> + <edge from-layer="460" from-port="0" to-layer="461" to-port="0" /> + <edge from-layer="461" from-port="1" to-layer="463" to-port="0" /> + <edge from-layer="462" from-port="0" to-layer="463" to-port="1" /> + <edge from-layer="463" from-port="2" to-layer="464" to-port="1" /> + <edge from-layer="464" from-port="2" to-layer="466" to-port="0" /> + <edge from-layer="465" from-port="0" to-layer="466" to-port="1" /> + <edge from-layer="466" from-port="2" to-layer="467" to-port="0" /> + <edge from-layer="467" from-port="2" to-layer="469" to-port="0" /> + <edge from-layer="468" from-port="0" to-layer="469" to-port="1" /> + <edge from-layer="469" from-port="2" to-layer="471" to-port="0" /> + <edge from-layer="470" from-port="0" to-layer="471" to-port="1" /> + <edge from-layer="471" from-port="2" to-layer="473" to-port="0" /> + <edge from-layer="472" from-port="0" to-layer="473" to-port="1" /> + <edge from-layer="473" from-port="2" to-layer="503" to-port="1" /> + <edge from-layer="473" from-port="2" to-layer="475" to-port="0" /> + <edge from-layer="474" from-port="0" to-layer="475" to-port="1" /> + <edge from-layer="475" from-port="2" to-layer="480" to-port="0" /> + <edge from-layer="476" from-port="0" to-layer="480" to-port="1" /> + <edge from-layer="477" from-port="0" to-layer="480" to-port="2" /> + <edge from-layer="478" from-port="0" to-layer="480" to-port="3" /> + <edge from-layer="479" from-port="0" to-layer="480" to-port="4" /> + <edge from-layer="480" from-port="5" to-layer="485" to-port="0" /> + <edge from-layer="481" from-port="0" to-layer="482" to-port="0" /> + <edge from-layer="482" from-port="1" to-layer="484" to-port="0" /> + <edge from-layer="483" from-port="0" to-layer="484" to-port="1" /> + <edge from-layer="484" from-port="2" to-layer="485" to-port="1" /> + <edge from-layer="485" from-port="2" to-layer="487" to-port="0" /> + <edge from-layer="486" from-port="0" to-layer="487" to-port="1" /> + <edge from-layer="487" from-port="2" to-layer="488" to-port="0" /> + <edge from-layer="488" from-port="1" to-layer="490" to-port="0" /> + <edge from-layer="489" from-port="0" to-layer="490" to-port="1" /> + <edge from-layer="490" from-port="2" to-layer="495" to-port="0" /> + <edge from-layer="491" from-port="0" to-layer="495" to-port="1" /> + <edge from-layer="492" from-port="0" to-layer="495" to-port="2" /> + <edge from-layer="493" from-port="0" to-layer="495" to-port="3" /> + <edge from-layer="494" from-port="0" to-layer="495" to-port="4" /> + <edge from-layer="495" from-port="5" to-layer="500" to-port="0" /> + <edge from-layer="496" from-port="0" to-layer="497" to-port="0" /> + <edge from-layer="497" from-port="1" to-layer="499" to-port="0" /> + <edge from-layer="498" from-port="0" to-layer="499" to-port="1" /> + <edge from-layer="499" from-port="2" to-layer="500" to-port="1" /> + <edge from-layer="500" from-port="2" to-layer="502" to-port="0" /> + <edge from-layer="501" from-port="0" to-layer="502" to-port="1" /> + <edge from-layer="502" from-port="2" to-layer="503" to-port="0" /> + <edge from-layer="503" from-port="2" to-layer="505" to-port="0" /> + <edge from-layer="504" from-port="0" to-layer="505" to-port="1" /> + <edge from-layer="505" from-port="2" to-layer="507" to-port="0" /> + <edge from-layer="506" from-port="0" to-layer="507" to-port="1" /> + <edge from-layer="507" from-port="2" to-layer="509" to-port="0" /> + <edge from-layer="508" from-port="0" to-layer="509" to-port="1" /> + <edge from-layer="509" from-port="2" to-layer="579" to-port="1" /> + <edge from-layer="509" from-port="2" to-layer="511" to-port="0" /> + <edge from-layer="510" from-port="0" to-layer="511" to-port="1" /> + <edge from-layer="511" from-port="2" to-layer="516" to-port="0" /> + <edge from-layer="512" from-port="0" to-layer="516" to-port="1" /> + <edge from-layer="513" from-port="0" to-layer="516" to-port="2" /> + <edge from-layer="514" from-port="0" to-layer="516" to-port="3" /> + <edge from-layer="515" from-port="0" to-layer="516" to-port="4" /> + <edge from-layer="516" from-port="5" to-layer="553" to-port="0" /> + <edge from-layer="516" from-port="5" to-layer="521" to-port="0" /> + <edge from-layer="516" from-port="5" to-layer="537" to-port="0" /> + <edge from-layer="517" from-port="0" to-layer="518" to-port="0" /> + <edge from-layer="518" from-port="1" to-layer="520" to-port="0" /> + <edge from-layer="519" from-port="0" to-layer="520" to-port="1" /> + <edge from-layer="520" from-port="2" to-layer="521" to-port="1" /> + <edge from-layer="521" from-port="2" to-layer="523" to-port="0" /> + <edge from-layer="522" from-port="0" to-layer="523" to-port="1" /> + <edge from-layer="523" from-port="2" to-layer="528" to-port="0" /> + <edge from-layer="524" from-port="0" to-layer="528" to-port="1" /> + <edge from-layer="525" from-port="0" to-layer="528" to-port="2" /> + <edge from-layer="526" from-port="0" to-layer="528" to-port="3" /> + <edge from-layer="527" from-port="0" to-layer="528" to-port="4" /> + <edge from-layer="528" from-port="5" to-layer="530" to-port="0" /> + <edge from-layer="529" from-port="0" to-layer="530" to-port="1" /> + <edge from-layer="530" from-port="2" to-layer="532" to-port="0" /> + <edge from-layer="531" from-port="0" to-layer="532" to-port="1" /> + <edge from-layer="532" from-port="2" to-layer="560" to-port="0" /> + <edge from-layer="533" from-port="0" to-layer="534" to-port="0" /> + <edge from-layer="534" from-port="1" to-layer="536" to-port="0" /> + <edge from-layer="535" from-port="0" to-layer="536" to-port="1" /> + <edge from-layer="536" from-port="2" to-layer="537" to-port="1" /> + <edge from-layer="537" from-port="2" to-layer="539" to-port="0" /> + <edge from-layer="538" from-port="0" to-layer="539" to-port="1" /> + <edge from-layer="539" from-port="2" to-layer="544" to-port="0" /> + <edge from-layer="540" from-port="0" to-layer="544" to-port="1" /> + <edge from-layer="541" from-port="0" to-layer="544" to-port="2" /> + <edge from-layer="542" from-port="0" to-layer="544" to-port="3" /> + <edge from-layer="543" from-port="0" to-layer="544" to-port="4" /> + <edge from-layer="544" from-port="5" to-layer="546" to-port="0" /> + <edge from-layer="545" from-port="0" to-layer="546" to-port="1" /> + <edge from-layer="546" from-port="2" to-layer="548" to-port="0" /> + <edge from-layer="547" from-port="0" to-layer="548" to-port="1" /> + <edge from-layer="548" from-port="2" to-layer="560" to-port="1" /> + <edge from-layer="549" from-port="0" to-layer="550" to-port="0" /> + <edge from-layer="550" from-port="1" to-layer="552" to-port="0" /> + <edge from-layer="551" from-port="0" to-layer="552" to-port="1" /> + <edge from-layer="552" from-port="2" to-layer="553" to-port="1" /> + <edge from-layer="553" from-port="2" to-layer="555" to-port="0" /> + <edge from-layer="554" from-port="0" to-layer="555" to-port="1" /> + <edge from-layer="555" from-port="2" to-layer="557" to-port="0" /> + <edge from-layer="556" from-port="0" to-layer="557" to-port="1" /> + <edge from-layer="557" from-port="2" to-layer="559" to-port="0" /> + <edge from-layer="558" from-port="0" to-layer="559" to-port="1" /> + <edge from-layer="559" from-port="2" to-layer="560" to-port="2" /> + <edge from-layer="560" from-port="4" to-layer="562" to-port="0" /> + <edge from-layer="561" from-port="0" to-layer="562" to-port="1" /> + <edge from-layer="562" from-port="2" to-layer="564" to-port="0" /> + <edge from-layer="563" from-port="0" to-layer="564" to-port="1" /> + <edge from-layer="564" from-port="2" to-layer="566" to-port="0" /> + <edge from-layer="565" from-port="0" to-layer="566" to-port="1" /> + <edge from-layer="566" from-port="2" to-layer="571" to-port="0" /> + <edge from-layer="567" from-port="0" to-layer="571" to-port="1" /> + <edge from-layer="568" from-port="0" to-layer="571" to-port="2" /> + <edge from-layer="569" from-port="0" to-layer="571" to-port="3" /> + <edge from-layer="570" from-port="0" to-layer="571" to-port="4" /> + <edge from-layer="571" from-port="5" to-layer="576" to-port="0" /> + <edge from-layer="572" from-port="0" to-layer="573" to-port="0" /> + <edge from-layer="573" from-port="1" to-layer="575" to-port="0" /> + <edge from-layer="574" from-port="0" to-layer="575" to-port="1" /> + <edge from-layer="575" from-port="2" to-layer="576" to-port="1" /> + <edge from-layer="576" from-port="2" to-layer="578" to-port="0" /> + <edge from-layer="577" from-port="0" to-layer="578" to-port="1" /> + <edge from-layer="578" from-port="2" to-layer="579" to-port="0" /> + <edge from-layer="579" from-port="2" to-layer="581" to-port="0" /> + <edge from-layer="580" from-port="0" to-layer="581" to-port="1" /> + <edge from-layer="581" from-port="2" to-layer="583" to-port="0" /> + <edge from-layer="582" from-port="0" to-layer="583" to-port="1" /> + <edge from-layer="583" from-port="2" to-layer="585" to-port="0" /> + <edge from-layer="584" from-port="0" to-layer="585" to-port="1" /> + <edge from-layer="585" from-port="2" to-layer="615" to-port="1" /> + <edge from-layer="585" from-port="2" to-layer="587" to-port="0" /> + <edge from-layer="586" from-port="0" to-layer="587" to-port="1" /> + <edge from-layer="587" from-port="2" to-layer="592" to-port="0" /> + <edge from-layer="588" from-port="0" to-layer="592" to-port="1" /> + <edge from-layer="589" from-port="0" to-layer="592" to-port="2" /> + <edge from-layer="590" from-port="0" to-layer="592" to-port="3" /> + <edge from-layer="591" from-port="0" to-layer="592" to-port="4" /> + <edge from-layer="592" from-port="5" to-layer="597" to-port="0" /> + <edge from-layer="593" from-port="0" to-layer="594" to-port="0" /> + <edge from-layer="594" from-port="1" to-layer="596" to-port="0" /> + <edge from-layer="595" from-port="0" to-layer="596" to-port="1" /> + <edge from-layer="596" from-port="2" to-layer="597" to-port="1" /> + <edge from-layer="597" from-port="2" to-layer="599" to-port="0" /> + <edge from-layer="598" from-port="0" to-layer="599" to-port="1" /> + <edge from-layer="599" from-port="2" to-layer="600" to-port="0" /> + <edge from-layer="600" from-port="1" to-layer="602" to-port="0" /> + <edge from-layer="601" from-port="0" to-layer="602" to-port="1" /> + <edge from-layer="602" from-port="2" to-layer="607" to-port="0" /> + <edge from-layer="603" from-port="0" to-layer="607" to-port="1" /> + <edge from-layer="604" from-port="0" to-layer="607" to-port="2" /> + <edge from-layer="605" from-port="0" to-layer="607" to-port="3" /> + <edge from-layer="606" from-port="0" to-layer="607" to-port="4" /> + <edge from-layer="607" from-port="5" to-layer="612" to-port="0" /> + <edge from-layer="608" from-port="0" to-layer="609" to-port="0" /> + <edge from-layer="609" from-port="1" to-layer="611" to-port="0" /> + <edge from-layer="610" from-port="0" to-layer="611" to-port="1" /> + <edge from-layer="611" from-port="2" to-layer="612" to-port="1" /> + <edge from-layer="612" from-port="2" to-layer="614" to-port="0" /> + <edge from-layer="613" from-port="0" to-layer="614" to-port="1" /> + <edge from-layer="614" from-port="2" to-layer="615" to-port="0" /> + <edge from-layer="615" from-port="2" to-layer="617" to-port="0" /> + <edge from-layer="616" from-port="0" to-layer="617" to-port="1" /> + <edge from-layer="617" from-port="2" to-layer="619" to-port="0" /> + <edge from-layer="618" from-port="0" to-layer="619" to-port="1" /> + <edge from-layer="619" from-port="2" to-layer="621" to-port="0" /> + <edge from-layer="620" from-port="0" to-layer="621" to-port="1" /> + <edge from-layer="621" from-port="2" to-layer="691" to-port="1" /> + <edge from-layer="621" from-port="2" to-layer="623" to-port="0" /> + <edge from-layer="622" from-port="0" to-layer="623" to-port="1" /> + <edge from-layer="623" from-port="2" to-layer="628" to-port="0" /> + <edge from-layer="624" from-port="0" to-layer="628" to-port="1" /> + <edge from-layer="625" from-port="0" to-layer="628" to-port="2" /> + <edge from-layer="626" from-port="0" to-layer="628" to-port="3" /> + <edge from-layer="627" from-port="0" to-layer="628" to-port="4" /> + <edge from-layer="628" from-port="5" to-layer="633" to-port="0" /> + <edge from-layer="628" from-port="5" to-layer="665" to-port="0" /> + <edge from-layer="628" from-port="5" to-layer="649" to-port="0" /> + <edge from-layer="629" from-port="0" to-layer="630" to-port="0" /> + <edge from-layer="630" from-port="1" to-layer="632" to-port="0" /> + <edge from-layer="631" from-port="0" to-layer="632" to-port="1" /> + <edge from-layer="632" from-port="2" to-layer="633" to-port="1" /> + <edge from-layer="633" from-port="2" to-layer="635" to-port="0" /> + <edge from-layer="634" from-port="0" to-layer="635" to-port="1" /> + <edge from-layer="635" from-port="2" to-layer="640" to-port="0" /> + <edge from-layer="636" from-port="0" to-layer="640" to-port="1" /> + <edge from-layer="637" from-port="0" to-layer="640" to-port="2" /> + <edge from-layer="638" from-port="0" to-layer="640" to-port="3" /> + <edge from-layer="639" from-port="0" to-layer="640" to-port="4" /> + <edge from-layer="640" from-port="5" to-layer="642" to-port="0" /> + <edge from-layer="641" from-port="0" to-layer="642" to-port="1" /> + <edge from-layer="642" from-port="2" to-layer="644" to-port="0" /> + <edge from-layer="643" from-port="0" to-layer="644" to-port="1" /> + <edge from-layer="644" from-port="2" to-layer="672" to-port="0" /> + <edge from-layer="645" from-port="0" to-layer="646" to-port="0" /> + <edge from-layer="646" from-port="1" to-layer="648" to-port="0" /> + <edge from-layer="647" from-port="0" to-layer="648" to-port="1" /> + <edge from-layer="648" from-port="2" to-layer="649" to-port="1" /> + <edge from-layer="649" from-port="2" to-layer="651" to-port="0" /> + <edge from-layer="650" from-port="0" to-layer="651" to-port="1" /> + <edge from-layer="651" from-port="2" to-layer="656" to-port="0" /> + <edge from-layer="652" from-port="0" to-layer="656" to-port="1" /> + <edge from-layer="653" from-port="0" to-layer="656" to-port="2" /> + <edge from-layer="654" from-port="0" to-layer="656" to-port="3" /> + <edge from-layer="655" from-port="0" to-layer="656" to-port="4" /> + <edge from-layer="656" from-port="5" to-layer="658" to-port="0" /> + <edge from-layer="657" from-port="0" to-layer="658" to-port="1" /> + <edge from-layer="658" from-port="2" to-layer="660" to-port="0" /> + <edge from-layer="659" from-port="0" to-layer="660" to-port="1" /> + <edge from-layer="660" from-port="2" to-layer="672" to-port="1" /> + <edge from-layer="661" from-port="0" to-layer="662" to-port="0" /> + <edge from-layer="662" from-port="1" to-layer="664" to-port="0" /> + <edge from-layer="663" from-port="0" to-layer="664" to-port="1" /> + <edge from-layer="664" from-port="2" to-layer="665" to-port="1" /> + <edge from-layer="665" from-port="2" to-layer="667" to-port="0" /> + <edge from-layer="666" from-port="0" to-layer="667" to-port="1" /> + <edge from-layer="667" from-port="2" to-layer="669" to-port="0" /> + <edge from-layer="668" from-port="0" to-layer="669" to-port="1" /> + <edge from-layer="669" from-port="2" to-layer="671" to-port="0" /> + <edge from-layer="670" from-port="0" to-layer="671" to-port="1" /> + <edge from-layer="671" from-port="2" to-layer="672" to-port="2" /> + <edge from-layer="672" from-port="4" to-layer="674" to-port="0" /> + <edge from-layer="673" from-port="0" to-layer="674" to-port="1" /> + <edge from-layer="674" from-port="2" to-layer="676" to-port="0" /> + <edge from-layer="675" from-port="0" to-layer="676" to-port="1" /> + <edge from-layer="676" from-port="2" to-layer="678" to-port="0" /> + <edge from-layer="677" from-port="0" to-layer="678" to-port="1" /> + <edge from-layer="678" from-port="2" to-layer="683" to-port="0" /> + <edge from-layer="679" from-port="0" to-layer="683" to-port="1" /> + <edge from-layer="680" from-port="0" to-layer="683" to-port="2" /> + <edge from-layer="681" from-port="0" to-layer="683" to-port="3" /> + <edge from-layer="682" from-port="0" to-layer="683" to-port="4" /> + <edge from-layer="683" from-port="5" to-layer="688" to-port="0" /> + <edge from-layer="684" from-port="0" to-layer="685" to-port="0" /> + <edge from-layer="685" from-port="1" to-layer="687" to-port="0" /> + <edge from-layer="686" from-port="0" to-layer="687" to-port="1" /> + <edge from-layer="687" from-port="2" to-layer="688" to-port="1" /> + <edge from-layer="688" from-port="2" to-layer="690" to-port="0" /> + <edge from-layer="689" from-port="0" to-layer="690" to-port="1" /> + <edge from-layer="690" from-port="2" to-layer="691" to-port="0" /> + <edge from-layer="691" from-port="2" to-layer="693" to-port="0" /> + <edge from-layer="692" from-port="0" to-layer="693" to-port="1" /> + <edge from-layer="693" from-port="2" to-layer="695" to-port="0" /> + <edge from-layer="694" from-port="0" to-layer="695" to-port="1" /> + <edge from-layer="695" from-port="2" to-layer="697" to-port="0" /> + <edge from-layer="696" from-port="0" to-layer="697" to-port="1" /> + <edge from-layer="697" from-port="2" to-layer="699" to-port="0" /> + <edge from-layer="697" from-port="2" to-layer="727" to-port="1" /> + <edge from-layer="698" from-port="0" to-layer="699" to-port="1" /> + <edge from-layer="699" from-port="2" to-layer="704" to-port="0" /> + <edge from-layer="700" from-port="0" to-layer="704" to-port="1" /> + <edge from-layer="701" from-port="0" to-layer="704" to-port="2" /> + <edge from-layer="702" from-port="0" to-layer="704" to-port="3" /> + <edge from-layer="703" from-port="0" to-layer="704" to-port="4" /> + <edge from-layer="704" from-port="5" to-layer="709" to-port="0" /> + <edge from-layer="705" from-port="0" to-layer="706" to-port="0" /> + <edge from-layer="706" from-port="1" to-layer="708" to-port="0" /> + <edge from-layer="707" from-port="0" to-layer="708" to-port="1" /> + <edge from-layer="708" from-port="2" to-layer="709" to-port="1" /> + <edge from-layer="709" from-port="2" to-layer="711" to-port="0" /> + <edge from-layer="710" from-port="0" to-layer="711" to-port="1" /> + <edge from-layer="711" from-port="2" to-layer="712" to-port="0" /> + <edge from-layer="712" from-port="1" to-layer="714" to-port="0" /> + <edge from-layer="713" from-port="0" to-layer="714" to-port="1" /> + <edge from-layer="714" from-port="2" to-layer="719" to-port="0" /> + <edge from-layer="715" from-port="0" to-layer="719" to-port="1" /> + <edge from-layer="716" from-port="0" to-layer="719" to-port="2" /> + <edge from-layer="717" from-port="0" to-layer="719" to-port="3" /> + <edge from-layer="718" from-port="0" to-layer="719" to-port="4" /> + <edge from-layer="719" from-port="5" to-layer="724" to-port="0" /> + <edge from-layer="720" from-port="0" to-layer="721" to-port="0" /> + <edge from-layer="721" from-port="1" to-layer="723" to-port="0" /> + <edge from-layer="722" from-port="0" to-layer="723" to-port="1" /> + <edge from-layer="723" from-port="2" to-layer="724" to-port="1" /> + <edge from-layer="724" from-port="2" to-layer="726" to-port="0" /> + <edge from-layer="725" from-port="0" to-layer="726" to-port="1" /> + <edge from-layer="726" from-port="2" to-layer="727" to-port="0" /> + <edge from-layer="727" from-port="2" to-layer="729" to-port="0" /> + <edge from-layer="728" from-port="0" to-layer="729" to-port="1" /> + <edge from-layer="729" from-port="2" to-layer="731" to-port="0" /> + <edge from-layer="730" from-port="0" to-layer="731" to-port="1" /> + <edge from-layer="731" from-port="2" to-layer="733" to-port="0" /> + <edge from-layer="732" from-port="0" to-layer="733" to-port="1" /> + <edge from-layer="733" from-port="2" to-layer="803" to-port="1" /> + <edge from-layer="733" from-port="2" to-layer="735" to-port="0" /> + <edge from-layer="734" from-port="0" to-layer="735" to-port="1" /> + <edge from-layer="735" from-port="2" to-layer="740" to-port="0" /> + <edge from-layer="736" from-port="0" to-layer="740" to-port="1" /> + <edge from-layer="737" from-port="0" to-layer="740" to-port="2" /> + <edge from-layer="738" from-port="0" to-layer="740" to-port="3" /> + <edge from-layer="739" from-port="0" to-layer="740" to-port="4" /> + <edge from-layer="740" from-port="5" to-layer="777" to-port="0" /> + <edge from-layer="740" from-port="5" to-layer="761" to-port="0" /> + <edge from-layer="740" from-port="5" to-layer="745" to-port="0" /> + <edge from-layer="741" from-port="0" to-layer="742" to-port="0" /> + <edge from-layer="742" from-port="1" to-layer="744" to-port="0" /> + <edge from-layer="743" from-port="0" to-layer="744" to-port="1" /> + <edge from-layer="744" from-port="2" to-layer="745" to-port="1" /> + <edge from-layer="745" from-port="2" to-layer="747" to-port="0" /> + <edge from-layer="746" from-port="0" to-layer="747" to-port="1" /> + <edge from-layer="747" from-port="2" to-layer="752" to-port="0" /> + <edge from-layer="748" from-port="0" to-layer="752" to-port="1" /> + <edge from-layer="749" from-port="0" to-layer="752" to-port="2" /> + <edge from-layer="750" from-port="0" to-layer="752" to-port="3" /> + <edge from-layer="751" from-port="0" to-layer="752" to-port="4" /> + <edge from-layer="752" from-port="5" to-layer="754" to-port="0" /> + <edge from-layer="753" from-port="0" to-layer="754" to-port="1" /> + <edge from-layer="754" from-port="2" to-layer="756" to-port="0" /> + <edge from-layer="755" from-port="0" to-layer="756" to-port="1" /> + <edge from-layer="756" from-port="2" to-layer="784" to-port="0" /> + <edge from-layer="757" from-port="0" to-layer="758" to-port="0" /> + <edge from-layer="758" from-port="1" to-layer="760" to-port="0" /> + <edge from-layer="759" from-port="0" to-layer="760" to-port="1" /> + <edge from-layer="760" from-port="2" to-layer="761" to-port="1" /> + <edge from-layer="761" from-port="2" to-layer="763" to-port="0" /> + <edge from-layer="762" from-port="0" to-layer="763" to-port="1" /> + <edge from-layer="763" from-port="2" to-layer="768" to-port="0" /> + <edge from-layer="764" from-port="0" to-layer="768" to-port="1" /> + <edge from-layer="765" from-port="0" to-layer="768" to-port="2" /> + <edge from-layer="766" from-port="0" to-layer="768" to-port="3" /> + <edge from-layer="767" from-port="0" to-layer="768" to-port="4" /> + <edge from-layer="768" from-port="5" to-layer="770" to-port="0" /> + <edge from-layer="769" from-port="0" to-layer="770" to-port="1" /> + <edge from-layer="770" from-port="2" to-layer="772" to-port="0" /> + <edge from-layer="771" from-port="0" to-layer="772" to-port="1" /> + <edge from-layer="772" from-port="2" to-layer="784" to-port="1" /> + <edge from-layer="773" from-port="0" to-layer="774" to-port="0" /> + <edge from-layer="774" from-port="1" to-layer="776" to-port="0" /> + <edge from-layer="775" from-port="0" to-layer="776" to-port="1" /> + <edge from-layer="776" from-port="2" to-layer="777" to-port="1" /> + <edge from-layer="777" from-port="2" to-layer="779" to-port="0" /> + <edge from-layer="778" from-port="0" to-layer="779" to-port="1" /> + <edge from-layer="779" from-port="2" to-layer="781" to-port="0" /> + <edge from-layer="780" from-port="0" to-layer="781" to-port="1" /> + <edge from-layer="781" from-port="2" to-layer="783" to-port="0" /> + <edge from-layer="782" from-port="0" to-layer="783" to-port="1" /> + <edge from-layer="783" from-port="2" to-layer="784" to-port="2" /> + <edge from-layer="784" from-port="4" to-layer="786" to-port="0" /> + <edge from-layer="785" from-port="0" to-layer="786" to-port="1" /> + <edge from-layer="786" from-port="2" to-layer="788" to-port="0" /> + <edge from-layer="787" from-port="0" to-layer="788" to-port="1" /> + <edge from-layer="788" from-port="2" to-layer="790" to-port="0" /> + <edge from-layer="789" from-port="0" to-layer="790" to-port="1" /> + <edge from-layer="790" from-port="2" to-layer="795" to-port="0" /> + <edge from-layer="791" from-port="0" to-layer="795" to-port="1" /> + <edge from-layer="792" from-port="0" to-layer="795" to-port="2" /> + <edge from-layer="793" from-port="0" to-layer="795" to-port="3" /> + <edge from-layer="794" from-port="0" to-layer="795" to-port="4" /> + <edge from-layer="795" from-port="5" to-layer="800" to-port="0" /> + <edge from-layer="796" from-port="0" to-layer="797" to-port="0" /> + <edge from-layer="797" from-port="1" to-layer="799" to-port="0" /> + <edge from-layer="798" from-port="0" to-layer="799" to-port="1" /> + <edge from-layer="799" from-port="2" to-layer="800" to-port="1" /> + <edge from-layer="800" from-port="2" to-layer="802" to-port="0" /> + <edge from-layer="801" from-port="0" to-layer="802" to-port="1" /> + <edge from-layer="802" from-port="2" to-layer="803" to-port="0" /> + <edge from-layer="803" from-port="2" to-layer="805" to-port="0" /> + <edge from-layer="804" from-port="0" to-layer="805" to-port="1" /> + <edge from-layer="805" from-port="2" to-layer="807" to-port="0" /> + <edge from-layer="806" from-port="0" to-layer="807" to-port="1" /> + <edge from-layer="807" from-port="2" to-layer="809" to-port="0" /> + <edge from-layer="808" from-port="0" to-layer="809" to-port="1" /> + <edge from-layer="809" from-port="2" to-layer="811" to-port="0" /> + <edge from-layer="809" from-port="2" to-layer="839" to-port="1" /> + <edge from-layer="810" from-port="0" to-layer="811" to-port="1" /> + <edge from-layer="811" from-port="2" to-layer="816" to-port="0" /> + <edge from-layer="812" from-port="0" to-layer="816" to-port="1" /> + <edge from-layer="813" from-port="0" to-layer="816" to-port="2" /> + <edge from-layer="814" from-port="0" to-layer="816" to-port="3" /> + <edge from-layer="815" from-port="0" to-layer="816" to-port="4" /> + <edge from-layer="816" from-port="5" to-layer="821" to-port="0" /> + <edge from-layer="817" from-port="0" to-layer="818" to-port="0" /> + <edge from-layer="818" from-port="1" to-layer="820" to-port="0" /> + <edge from-layer="819" from-port="0" to-layer="820" to-port="1" /> + <edge from-layer="820" from-port="2" to-layer="821" to-port="1" /> + <edge from-layer="821" from-port="2" to-layer="823" to-port="0" /> + <edge from-layer="822" from-port="0" to-layer="823" to-port="1" /> + <edge from-layer="823" from-port="2" to-layer="824" to-port="0" /> + <edge from-layer="824" from-port="1" to-layer="826" to-port="0" /> + <edge from-layer="825" from-port="0" to-layer="826" to-port="1" /> + <edge from-layer="826" from-port="2" to-layer="831" to-port="0" /> + <edge from-layer="827" from-port="0" to-layer="831" to-port="1" /> + <edge from-layer="828" from-port="0" to-layer="831" to-port="2" /> + <edge from-layer="829" from-port="0" to-layer="831" to-port="3" /> + <edge from-layer="830" from-port="0" to-layer="831" to-port="4" /> + <edge from-layer="831" from-port="5" to-layer="836" to-port="0" /> + <edge from-layer="832" from-port="0" to-layer="833" to-port="0" /> + <edge from-layer="833" from-port="1" to-layer="835" to-port="0" /> + <edge from-layer="834" from-port="0" to-layer="835" to-port="1" /> + <edge from-layer="835" from-port="2" to-layer="836" to-port="1" /> + <edge from-layer="836" from-port="2" to-layer="838" to-port="0" /> + <edge from-layer="837" from-port="0" to-layer="838" to-port="1" /> + <edge from-layer="838" from-port="2" to-layer="839" to-port="0" /> + <edge from-layer="839" from-port="2" to-layer="841" to-port="0" /> + <edge from-layer="840" from-port="0" to-layer="841" to-port="1" /> + <edge from-layer="841" from-port="2" to-layer="843" to-port="0" /> + <edge from-layer="842" from-port="0" to-layer="843" to-port="1" /> + <edge from-layer="843" from-port="2" to-layer="845" to-port="0" /> + <edge from-layer="844" from-port="0" to-layer="845" to-port="1" /> + <edge from-layer="845" from-port="2" to-layer="847" to-port="0" /> + <edge from-layer="845" from-port="2" to-layer="915" to-port="1" /> + <edge from-layer="846" from-port="0" to-layer="847" to-port="1" /> + <edge from-layer="847" from-port="2" to-layer="852" to-port="0" /> + <edge from-layer="848" from-port="0" to-layer="852" to-port="1" /> + <edge from-layer="849" from-port="0" to-layer="852" to-port="2" /> + <edge from-layer="850" from-port="0" to-layer="852" to-port="3" /> + <edge from-layer="851" from-port="0" to-layer="852" to-port="4" /> + <edge from-layer="852" from-port="5" to-layer="889" to-port="0" /> + <edge from-layer="852" from-port="5" to-layer="857" to-port="0" /> + <edge from-layer="852" from-port="5" to-layer="873" to-port="0" /> + <edge from-layer="853" from-port="0" to-layer="854" to-port="0" /> + <edge from-layer="854" from-port="1" to-layer="856" to-port="0" /> + <edge from-layer="855" from-port="0" to-layer="856" to-port="1" /> + <edge from-layer="856" from-port="2" to-layer="857" to-port="1" /> + <edge from-layer="857" from-port="2" to-layer="859" to-port="0" /> + <edge from-layer="858" from-port="0" to-layer="859" to-port="1" /> + <edge from-layer="859" from-port="2" to-layer="864" to-port="0" /> + <edge from-layer="860" from-port="0" to-layer="864" to-port="1" /> + <edge from-layer="861" from-port="0" to-layer="864" to-port="2" /> + <edge from-layer="862" from-port="0" to-layer="864" to-port="3" /> + <edge from-layer="863" from-port="0" to-layer="864" to-port="4" /> + <edge from-layer="864" from-port="5" to-layer="866" to-port="0" /> + <edge from-layer="865" from-port="0" to-layer="866" to-port="1" /> + <edge from-layer="866" from-port="2" to-layer="868" to-port="0" /> + <edge from-layer="867" from-port="0" to-layer="868" to-port="1" /> + <edge from-layer="868" from-port="2" to-layer="896" to-port="0" /> + <edge from-layer="869" from-port="0" to-layer="870" to-port="0" /> + <edge from-layer="870" from-port="1" to-layer="872" to-port="0" /> + <edge from-layer="871" from-port="0" to-layer="872" to-port="1" /> + <edge from-layer="872" from-port="2" to-layer="873" to-port="1" /> + <edge from-layer="873" from-port="2" to-layer="875" to-port="0" /> + <edge from-layer="874" from-port="0" to-layer="875" to-port="1" /> + <edge from-layer="875" from-port="2" to-layer="880" to-port="0" /> + <edge from-layer="876" from-port="0" to-layer="880" to-port="1" /> + <edge from-layer="877" from-port="0" to-layer="880" to-port="2" /> + <edge from-layer="878" from-port="0" to-layer="880" to-port="3" /> + <edge from-layer="879" from-port="0" to-layer="880" to-port="4" /> + <edge from-layer="880" from-port="5" to-layer="882" to-port="0" /> + <edge from-layer="881" from-port="0" to-layer="882" to-port="1" /> + <edge from-layer="882" from-port="2" to-layer="884" to-port="0" /> + <edge from-layer="883" from-port="0" to-layer="884" to-port="1" /> + <edge from-layer="884" from-port="2" to-layer="896" to-port="1" /> + <edge from-layer="885" from-port="0" to-layer="886" to-port="0" /> + <edge from-layer="886" from-port="1" to-layer="888" to-port="0" /> + <edge from-layer="887" from-port="0" to-layer="888" to-port="1" /> + <edge from-layer="888" from-port="2" to-layer="889" to-port="1" /> + <edge from-layer="889" from-port="2" to-layer="891" to-port="0" /> + <edge from-layer="890" from-port="0" to-layer="891" to-port="1" /> + <edge from-layer="891" from-port="2" to-layer="893" to-port="0" /> + <edge from-layer="892" from-port="0" to-layer="893" to-port="1" /> + <edge from-layer="893" from-port="2" to-layer="895" to-port="0" /> + <edge from-layer="894" from-port="0" to-layer="895" to-port="1" /> + <edge from-layer="895" from-port="2" to-layer="896" to-port="2" /> + <edge from-layer="896" from-port="4" to-layer="898" to-port="0" /> + <edge from-layer="897" from-port="0" to-layer="898" to-port="1" /> + <edge from-layer="898" from-port="2" to-layer="900" to-port="0" /> + <edge from-layer="899" from-port="0" to-layer="900" to-port="1" /> + <edge from-layer="900" from-port="2" to-layer="902" to-port="0" /> + <edge from-layer="901" from-port="0" to-layer="902" to-port="1" /> + <edge from-layer="902" from-port="2" to-layer="907" to-port="0" /> + <edge from-layer="903" from-port="0" to-layer="907" to-port="1" /> + <edge from-layer="904" from-port="0" to-layer="907" to-port="2" /> + <edge from-layer="905" from-port="0" to-layer="907" to-port="3" /> + <edge from-layer="906" from-port="0" to-layer="907" to-port="4" /> + <edge from-layer="907" from-port="5" to-layer="912" to-port="0" /> + <edge from-layer="908" from-port="0" to-layer="909" to-port="0" /> + <edge from-layer="909" from-port="1" to-layer="911" to-port="0" /> + <edge from-layer="910" from-port="0" to-layer="911" to-port="1" /> + <edge from-layer="911" from-port="2" to-layer="912" to-port="1" /> + <edge from-layer="912" from-port="2" to-layer="914" to-port="0" /> + <edge from-layer="913" from-port="0" to-layer="914" to-port="1" /> + <edge from-layer="914" from-port="2" to-layer="915" to-port="0" /> + <edge from-layer="915" from-port="2" to-layer="917" to-port="0" /> + <edge from-layer="916" from-port="0" to-layer="917" to-port="1" /> + <edge from-layer="917" from-port="2" to-layer="919" to-port="0" /> + <edge from-layer="918" from-port="0" to-layer="919" to-port="1" /> + <edge from-layer="919" from-port="2" to-layer="921" to-port="0" /> + <edge from-layer="920" from-port="0" to-layer="921" to-port="1" /> + <edge from-layer="921" from-port="2" to-layer="951" to-port="1" /> + <edge from-layer="921" from-port="2" to-layer="923" to-port="0" /> + <edge from-layer="922" from-port="0" to-layer="923" to-port="1" /> + <edge from-layer="923" from-port="2" to-layer="928" to-port="0" /> + <edge from-layer="924" from-port="0" to-layer="928" to-port="1" /> + <edge from-layer="925" from-port="0" to-layer="928" to-port="2" /> + <edge from-layer="926" from-port="0" to-layer="928" to-port="3" /> + <edge from-layer="927" from-port="0" to-layer="928" to-port="4" /> + <edge from-layer="928" from-port="5" to-layer="933" to-port="0" /> + <edge from-layer="929" from-port="0" to-layer="930" to-port="0" /> + <edge from-layer="930" from-port="1" to-layer="932" to-port="0" /> + <edge from-layer="931" from-port="0" to-layer="932" to-port="1" /> + <edge from-layer="932" from-port="2" to-layer="933" to-port="1" /> + <edge from-layer="933" from-port="2" to-layer="935" to-port="0" /> + <edge from-layer="934" from-port="0" to-layer="935" to-port="1" /> + <edge from-layer="935" from-port="2" to-layer="936" to-port="0" /> + <edge from-layer="936" from-port="1" to-layer="938" to-port="0" /> + <edge from-layer="937" from-port="0" to-layer="938" to-port="1" /> + <edge from-layer="938" from-port="2" to-layer="943" to-port="0" /> + <edge from-layer="939" from-port="0" to-layer="943" to-port="1" /> + <edge from-layer="940" from-port="0" to-layer="943" to-port="2" /> + <edge from-layer="941" from-port="0" to-layer="943" to-port="3" /> + <edge from-layer="942" from-port="0" to-layer="943" to-port="4" /> + <edge from-layer="943" from-port="5" to-layer="948" to-port="0" /> + <edge from-layer="944" from-port="0" to-layer="945" to-port="0" /> + <edge from-layer="945" from-port="1" to-layer="947" to-port="0" /> + <edge from-layer="946" from-port="0" to-layer="947" to-port="1" /> + <edge from-layer="947" from-port="2" to-layer="948" to-port="1" /> + <edge from-layer="948" from-port="2" to-layer="950" to-port="0" /> + <edge from-layer="949" from-port="0" to-layer="950" to-port="1" /> + <edge from-layer="950" from-port="2" to-layer="951" to-port="0" /> + <edge from-layer="951" from-port="2" to-layer="953" to-port="0" /> + <edge from-layer="952" from-port="0" to-layer="953" to-port="1" /> + <edge from-layer="953" from-port="2" to-layer="955" to-port="0" /> + <edge from-layer="954" from-port="0" to-layer="955" to-port="1" /> + <edge from-layer="955" from-port="2" to-layer="957" to-port="0" /> + <edge from-layer="956" from-port="0" to-layer="957" to-port="1" /> + <edge from-layer="957" from-port="2" to-layer="1027" to-port="1" /> + <edge from-layer="957" from-port="2" to-layer="959" to-port="0" /> + <edge from-layer="958" from-port="0" to-layer="959" to-port="1" /> + <edge from-layer="959" from-port="2" to-layer="964" to-port="0" /> + <edge from-layer="960" from-port="0" to-layer="964" to-port="1" /> + <edge from-layer="961" from-port="0" to-layer="964" to-port="2" /> + <edge from-layer="962" from-port="0" to-layer="964" to-port="3" /> + <edge from-layer="963" from-port="0" to-layer="964" to-port="4" /> + <edge from-layer="964" from-port="5" to-layer="985" to-port="0" /> + <edge from-layer="964" from-port="5" to-layer="1001" to-port="0" /> + <edge from-layer="964" from-port="5" to-layer="969" to-port="0" /> + <edge from-layer="965" from-port="0" to-layer="966" to-port="0" /> + <edge from-layer="966" from-port="1" to-layer="968" to-port="0" /> + <edge from-layer="967" from-port="0" to-layer="968" to-port="1" /> + <edge from-layer="968" from-port="2" to-layer="969" to-port="1" /> + <edge from-layer="969" from-port="2" to-layer="971" to-port="0" /> + <edge from-layer="970" from-port="0" to-layer="971" to-port="1" /> + <edge from-layer="971" from-port="2" to-layer="976" to-port="0" /> + <edge from-layer="972" from-port="0" to-layer="976" to-port="1" /> + <edge from-layer="973" from-port="0" to-layer="976" to-port="2" /> + <edge from-layer="974" from-port="0" to-layer="976" to-port="3" /> + <edge from-layer="975" from-port="0" to-layer="976" to-port="4" /> + <edge from-layer="976" from-port="5" to-layer="978" to-port="0" /> + <edge from-layer="977" from-port="0" to-layer="978" to-port="1" /> + <edge from-layer="978" from-port="2" to-layer="980" to-port="0" /> + <edge from-layer="979" from-port="0" to-layer="980" to-port="1" /> + <edge from-layer="980" from-port="2" to-layer="1008" to-port="0" /> + <edge from-layer="981" from-port="0" to-layer="982" to-port="0" /> + <edge from-layer="982" from-port="1" to-layer="984" to-port="0" /> + <edge from-layer="983" from-port="0" to-layer="984" to-port="1" /> + <edge from-layer="984" from-port="2" to-layer="985" to-port="1" /> + <edge from-layer="985" from-port="2" to-layer="987" to-port="0" /> + <edge from-layer="986" from-port="0" to-layer="987" to-port="1" /> + <edge from-layer="987" from-port="2" to-layer="992" to-port="0" /> + <edge from-layer="988" from-port="0" to-layer="992" to-port="1" /> + <edge from-layer="989" from-port="0" to-layer="992" to-port="2" /> + <edge from-layer="990" from-port="0" to-layer="992" to-port="3" /> + <edge from-layer="991" from-port="0" to-layer="992" to-port="4" /> + <edge from-layer="992" from-port="5" to-layer="994" to-port="0" /> + <edge from-layer="993" from-port="0" to-layer="994" to-port="1" /> + <edge from-layer="994" from-port="2" to-layer="996" to-port="0" /> + <edge from-layer="995" from-port="0" to-layer="996" to-port="1" /> + <edge from-layer="996" from-port="2" to-layer="1008" to-port="1" /> + <edge from-layer="997" from-port="0" to-layer="998" to-port="0" /> + <edge from-layer="998" from-port="1" to-layer="1000" to-port="0" /> + <edge from-layer="999" from-port="0" to-layer="1000" to-port="1" /> + <edge from-layer="1000" from-port="2" to-layer="1001" to-port="1" /> + <edge from-layer="1001" from-port="2" to-layer="1003" to-port="0" /> + <edge from-layer="1002" from-port="0" to-layer="1003" to-port="1" /> + <edge from-layer="1003" from-port="2" to-layer="1005" to-port="0" /> + <edge from-layer="1004" from-port="0" to-layer="1005" to-port="1" /> + <edge from-layer="1005" from-port="2" to-layer="1007" to-port="0" /> + <edge from-layer="1006" from-port="0" to-layer="1007" to-port="1" /> + <edge from-layer="1007" from-port="2" to-layer="1008" to-port="2" /> + <edge from-layer="1008" from-port="4" to-layer="1010" to-port="0" /> + <edge from-layer="1009" from-port="0" to-layer="1010" to-port="1" /> + <edge from-layer="1010" from-port="2" to-layer="1012" to-port="0" /> + <edge from-layer="1011" from-port="0" to-layer="1012" to-port="1" /> + <edge from-layer="1012" from-port="2" to-layer="1014" to-port="0" /> + <edge from-layer="1013" from-port="0" to-layer="1014" to-port="1" /> + <edge from-layer="1014" from-port="2" to-layer="1019" to-port="0" /> + <edge from-layer="1015" from-port="0" to-layer="1019" to-port="1" /> + <edge from-layer="1016" from-port="0" to-layer="1019" to-port="2" /> + <edge from-layer="1017" from-port="0" to-layer="1019" to-port="3" /> + <edge from-layer="1018" from-port="0" to-layer="1019" to-port="4" /> + <edge from-layer="1019" from-port="5" to-layer="1024" to-port="0" /> + <edge from-layer="1020" from-port="0" to-layer="1021" to-port="0" /> + <edge from-layer="1021" from-port="1" to-layer="1023" to-port="0" /> + <edge from-layer="1022" from-port="0" to-layer="1023" to-port="1" /> + <edge from-layer="1023" from-port="2" to-layer="1024" to-port="1" /> + <edge from-layer="1024" from-port="2" to-layer="1026" to-port="0" /> + <edge from-layer="1025" from-port="0" to-layer="1026" to-port="1" /> + <edge from-layer="1026" from-port="2" to-layer="1027" to-port="0" /> + <edge from-layer="1027" from-port="2" to-layer="1029" to-port="0" /> + <edge from-layer="1028" from-port="0" to-layer="1029" to-port="1" /> + <edge from-layer="1029" from-port="2" to-layer="1031" to-port="0" /> + <edge from-layer="1030" from-port="0" to-layer="1031" to-port="1" /> + <edge from-layer="1031" from-port="2" to-layer="1033" to-port="0" /> + <edge from-layer="1032" from-port="0" to-layer="1033" to-port="1" /> + <edge from-layer="1033" from-port="2" to-layer="1063" to-port="1" /> + <edge from-layer="1033" from-port="2" to-layer="1035" to-port="0" /> + <edge from-layer="1034" from-port="0" to-layer="1035" to-port="1" /> + <edge from-layer="1035" from-port="2" to-layer="1040" to-port="0" /> + <edge from-layer="1036" from-port="0" to-layer="1040" to-port="1" /> + <edge from-layer="1037" from-port="0" to-layer="1040" to-port="2" /> + <edge from-layer="1038" from-port="0" to-layer="1040" to-port="3" /> + <edge from-layer="1039" from-port="0" to-layer="1040" to-port="4" /> + <edge from-layer="1040" from-port="5" to-layer="1045" to-port="0" /> + <edge from-layer="1041" from-port="0" to-layer="1042" to-port="0" /> + <edge from-layer="1042" from-port="1" to-layer="1044" to-port="0" /> + <edge from-layer="1043" from-port="0" to-layer="1044" to-port="1" /> + <edge from-layer="1044" from-port="2" to-layer="1045" to-port="1" /> + <edge from-layer="1045" from-port="2" to-layer="1047" to-port="0" /> + <edge from-layer="1046" from-port="0" to-layer="1047" to-port="1" /> + <edge from-layer="1047" from-port="2" to-layer="1048" to-port="0" /> + <edge from-layer="1048" from-port="1" to-layer="1050" to-port="0" /> + <edge from-layer="1049" from-port="0" to-layer="1050" to-port="1" /> + <edge from-layer="1050" from-port="2" to-layer="1055" to-port="0" /> + <edge from-layer="1051" from-port="0" to-layer="1055" to-port="1" /> + <edge from-layer="1052" from-port="0" to-layer="1055" to-port="2" /> + <edge from-layer="1053" from-port="0" to-layer="1055" to-port="3" /> + <edge from-layer="1054" from-port="0" to-layer="1055" to-port="4" /> + <edge from-layer="1055" from-port="5" to-layer="1060" to-port="0" /> + <edge from-layer="1056" from-port="0" to-layer="1057" to-port="0" /> + <edge from-layer="1057" from-port="1" to-layer="1059" to-port="0" /> + <edge from-layer="1058" from-port="0" to-layer="1059" to-port="1" /> + <edge from-layer="1059" from-port="2" to-layer="1060" to-port="1" /> + <edge from-layer="1060" from-port="2" to-layer="1062" to-port="0" /> + <edge from-layer="1061" from-port="0" to-layer="1062" to-port="1" /> + <edge from-layer="1062" from-port="2" to-layer="1063" to-port="0" /> + <edge from-layer="1063" from-port="2" to-layer="1065" to-port="0" /> + <edge from-layer="1064" from-port="0" to-layer="1065" to-port="1" /> + <edge from-layer="1065" from-port="2" to-layer="1067" to-port="0" /> + <edge from-layer="1066" from-port="0" to-layer="1067" to-port="1" /> + <edge from-layer="1067" from-port="2" to-layer="1069" to-port="0" /> + <edge from-layer="1068" from-port="0" to-layer="1069" to-port="1" /> + <edge from-layer="1069" from-port="2" to-layer="1071" to-port="0" /> + <edge from-layer="1069" from-port="2" to-layer="1139" to-port="1" /> + <edge from-layer="1070" from-port="0" to-layer="1071" to-port="1" /> + <edge from-layer="1071" from-port="2" to-layer="1076" to-port="0" /> + <edge from-layer="1072" from-port="0" to-layer="1076" to-port="1" /> + <edge from-layer="1073" from-port="0" to-layer="1076" to-port="2" /> + <edge from-layer="1074" from-port="0" to-layer="1076" to-port="3" /> + <edge from-layer="1075" from-port="0" to-layer="1076" to-port="4" /> + <edge from-layer="1076" from-port="5" to-layer="1113" to-port="0" /> + <edge from-layer="1076" from-port="5" to-layer="1097" to-port="0" /> + <edge from-layer="1076" from-port="5" to-layer="1081" to-port="0" /> + <edge from-layer="1077" from-port="0" to-layer="1078" to-port="0" /> + <edge from-layer="1078" from-port="1" to-layer="1080" to-port="0" /> + <edge from-layer="1079" from-port="0" to-layer="1080" to-port="1" /> + <edge from-layer="1080" from-port="2" to-layer="1081" to-port="1" /> + <edge from-layer="1081" from-port="2" to-layer="1083" to-port="0" /> + <edge from-layer="1082" from-port="0" to-layer="1083" to-port="1" /> + <edge from-layer="1083" from-port="2" to-layer="1088" to-port="0" /> + <edge from-layer="1084" from-port="0" to-layer="1088" to-port="1" /> + <edge from-layer="1085" from-port="0" to-layer="1088" to-port="2" /> + <edge from-layer="1086" from-port="0" to-layer="1088" to-port="3" /> + <edge from-layer="1087" from-port="0" to-layer="1088" to-port="4" /> + <edge from-layer="1088" from-port="5" to-layer="1090" to-port="0" /> + <edge from-layer="1089" from-port="0" to-layer="1090" to-port="1" /> + <edge from-layer="1090" from-port="2" to-layer="1092" to-port="0" /> + <edge from-layer="1091" from-port="0" to-layer="1092" to-port="1" /> + <edge from-layer="1092" from-port="2" to-layer="1120" to-port="0" /> + <edge from-layer="1093" from-port="0" to-layer="1094" to-port="0" /> + <edge from-layer="1094" from-port="1" to-layer="1096" to-port="0" /> + <edge from-layer="1095" from-port="0" to-layer="1096" to-port="1" /> + <edge from-layer="1096" from-port="2" to-layer="1097" to-port="1" /> + <edge from-layer="1097" from-port="2" to-layer="1099" to-port="0" /> + <edge from-layer="1098" from-port="0" to-layer="1099" to-port="1" /> + <edge from-layer="1099" from-port="2" to-layer="1104" to-port="0" /> + <edge from-layer="1100" from-port="0" to-layer="1104" to-port="1" /> + <edge from-layer="1101" from-port="0" to-layer="1104" to-port="2" /> + <edge from-layer="1102" from-port="0" to-layer="1104" to-port="3" /> + <edge from-layer="1103" from-port="0" to-layer="1104" to-port="4" /> + <edge from-layer="1104" from-port="5" to-layer="1106" to-port="0" /> + <edge from-layer="1105" from-port="0" to-layer="1106" to-port="1" /> + <edge from-layer="1106" from-port="2" to-layer="1108" to-port="0" /> + <edge from-layer="1107" from-port="0" to-layer="1108" to-port="1" /> + <edge from-layer="1108" from-port="2" to-layer="1120" to-port="1" /> + <edge from-layer="1109" from-port="0" to-layer="1110" to-port="0" /> + <edge from-layer="1110" from-port="1" to-layer="1112" to-port="0" /> + <edge from-layer="1111" from-port="0" to-layer="1112" to-port="1" /> + <edge from-layer="1112" from-port="2" to-layer="1113" to-port="1" /> + <edge from-layer="1113" from-port="2" to-layer="1115" to-port="0" /> + <edge from-layer="1114" from-port="0" to-layer="1115" to-port="1" /> + <edge from-layer="1115" from-port="2" to-layer="1117" to-port="0" /> + <edge from-layer="1116" from-port="0" to-layer="1117" to-port="1" /> + <edge from-layer="1117" from-port="2" to-layer="1119" to-port="0" /> + <edge from-layer="1118" from-port="0" to-layer="1119" to-port="1" /> + <edge from-layer="1119" from-port="2" to-layer="1120" to-port="2" /> + <edge from-layer="1120" from-port="4" to-layer="1122" to-port="0" /> + <edge from-layer="1121" from-port="0" to-layer="1122" to-port="1" /> + <edge from-layer="1122" from-port="2" to-layer="1124" to-port="0" /> + <edge from-layer="1123" from-port="0" to-layer="1124" to-port="1" /> + <edge from-layer="1124" from-port="2" to-layer="1126" to-port="0" /> + <edge from-layer="1125" from-port="0" to-layer="1126" to-port="1" /> + <edge from-layer="1126" from-port="2" to-layer="1131" to-port="0" /> + <edge from-layer="1127" from-port="0" to-layer="1131" to-port="1" /> + <edge from-layer="1128" from-port="0" to-layer="1131" to-port="2" /> + <edge from-layer="1129" from-port="0" to-layer="1131" to-port="3" /> + <edge from-layer="1130" from-port="0" to-layer="1131" to-port="4" /> + <edge from-layer="1131" from-port="5" to-layer="1136" to-port="0" /> + <edge from-layer="1132" from-port="0" to-layer="1133" to-port="0" /> + <edge from-layer="1133" from-port="1" to-layer="1135" to-port="0" /> + <edge from-layer="1134" from-port="0" to-layer="1135" to-port="1" /> + <edge from-layer="1135" from-port="2" to-layer="1136" to-port="1" /> + <edge from-layer="1136" from-port="2" to-layer="1138" to-port="0" /> + <edge from-layer="1137" from-port="0" to-layer="1138" to-port="1" /> + <edge from-layer="1138" from-port="2" to-layer="1139" to-port="0" /> + <edge from-layer="1139" from-port="2" to-layer="1141" to-port="0" /> + <edge from-layer="1140" from-port="0" to-layer="1141" to-port="1" /> + <edge from-layer="1141" from-port="2" to-layer="1143" to-port="0" /> + <edge from-layer="1142" from-port="0" to-layer="1143" to-port="1" /> + <edge from-layer="1143" from-port="2" to-layer="1145" to-port="0" /> + <edge from-layer="1144" from-port="0" to-layer="1145" to-port="1" /> + <edge from-layer="1145" from-port="2" to-layer="1147" to-port="0" /> + <edge from-layer="1145" from-port="2" to-layer="1175" to-port="1" /> + <edge from-layer="1146" from-port="0" to-layer="1147" to-port="1" /> + <edge from-layer="1147" from-port="2" to-layer="1152" to-port="0" /> + <edge from-layer="1148" from-port="0" to-layer="1152" to-port="1" /> + <edge from-layer="1149" from-port="0" to-layer="1152" to-port="2" /> + <edge from-layer="1150" from-port="0" to-layer="1152" to-port="3" /> + <edge from-layer="1151" from-port="0" to-layer="1152" to-port="4" /> + <edge from-layer="1152" from-port="5" to-layer="1157" to-port="0" /> + <edge from-layer="1153" from-port="0" to-layer="1154" to-port="0" /> + <edge from-layer="1154" from-port="1" to-layer="1156" to-port="0" /> + <edge from-layer="1155" from-port="0" to-layer="1156" to-port="1" /> + <edge from-layer="1156" from-port="2" to-layer="1157" to-port="1" /> + <edge from-layer="1157" from-port="2" to-layer="1159" to-port="0" /> + <edge from-layer="1158" from-port="0" to-layer="1159" to-port="1" /> + <edge from-layer="1159" from-port="2" to-layer="1160" to-port="0" /> + <edge from-layer="1160" from-port="1" to-layer="1162" to-port="0" /> + <edge from-layer="1161" from-port="0" to-layer="1162" to-port="1" /> + <edge from-layer="1162" from-port="2" to-layer="1167" to-port="0" /> + <edge from-layer="1163" from-port="0" to-layer="1167" to-port="1" /> + <edge from-layer="1164" from-port="0" to-layer="1167" to-port="2" /> + <edge from-layer="1165" from-port="0" to-layer="1167" to-port="3" /> + <edge from-layer="1166" from-port="0" to-layer="1167" to-port="4" /> + <edge from-layer="1167" from-port="5" to-layer="1172" to-port="0" /> + <edge from-layer="1168" from-port="0" to-layer="1169" to-port="0" /> + <edge from-layer="1169" from-port="1" to-layer="1171" to-port="0" /> + <edge from-layer="1170" from-port="0" to-layer="1171" to-port="1" /> + <edge from-layer="1171" from-port="2" to-layer="1172" to-port="1" /> + <edge from-layer="1172" from-port="2" to-layer="1174" to-port="0" /> + <edge from-layer="1173" from-port="0" to-layer="1174" to-port="1" /> + <edge from-layer="1174" from-port="2" to-layer="1175" to-port="0" /> + <edge from-layer="1175" from-port="2" to-layer="1177" to-port="0" /> + <edge from-layer="1176" from-port="0" to-layer="1177" to-port="1" /> + <edge from-layer="1177" from-port="2" to-layer="1179" to-port="0" /> + <edge from-layer="1178" from-port="0" to-layer="1179" to-port="1" /> + <edge from-layer="1179" from-port="2" to-layer="1181" to-port="0" /> + <edge from-layer="1180" from-port="0" to-layer="1181" to-port="1" /> + <edge from-layer="1181" from-port="2" to-layer="1251" to-port="1" /> + <edge from-layer="1181" from-port="2" to-layer="1183" to-port="0" /> + <edge from-layer="1182" from-port="0" to-layer="1183" to-port="1" /> + <edge from-layer="1183" from-port="2" to-layer="1188" to-port="0" /> + <edge from-layer="1184" from-port="0" to-layer="1188" to-port="1" /> + <edge from-layer="1185" from-port="0" to-layer="1188" to-port="2" /> + <edge from-layer="1186" from-port="0" to-layer="1188" to-port="3" /> + <edge from-layer="1187" from-port="0" to-layer="1188" to-port="4" /> + <edge from-layer="1188" from-port="5" to-layer="1193" to-port="0" /> + <edge from-layer="1188" from-port="5" to-layer="1225" to-port="0" /> + <edge from-layer="1188" from-port="5" to-layer="1209" to-port="0" /> + <edge from-layer="1189" from-port="0" to-layer="1190" to-port="0" /> + <edge from-layer="1190" from-port="1" to-layer="1192" to-port="0" /> + <edge from-layer="1191" from-port="0" to-layer="1192" to-port="1" /> + <edge from-layer="1192" from-port="2" to-layer="1193" to-port="1" /> + <edge from-layer="1193" from-port="2" to-layer="1195" to-port="0" /> + <edge from-layer="1194" from-port="0" to-layer="1195" to-port="1" /> + <edge from-layer="1195" from-port="2" to-layer="1200" to-port="0" /> + <edge from-layer="1196" from-port="0" to-layer="1200" to-port="1" /> + <edge from-layer="1197" from-port="0" to-layer="1200" to-port="2" /> + <edge from-layer="1198" from-port="0" to-layer="1200" to-port="3" /> + <edge from-layer="1199" from-port="0" to-layer="1200" to-port="4" /> + <edge from-layer="1200" from-port="5" to-layer="1202" to-port="0" /> + <edge from-layer="1201" from-port="0" to-layer="1202" to-port="1" /> + <edge from-layer="1202" from-port="2" to-layer="1204" to-port="0" /> + <edge from-layer="1203" from-port="0" to-layer="1204" to-port="1" /> + <edge from-layer="1204" from-port="2" to-layer="1232" to-port="0" /> + <edge from-layer="1205" from-port="0" to-layer="1206" to-port="0" /> + <edge from-layer="1206" from-port="1" to-layer="1208" to-port="0" /> + <edge from-layer="1207" from-port="0" to-layer="1208" to-port="1" /> + <edge from-layer="1208" from-port="2" to-layer="1209" to-port="1" /> + <edge from-layer="1209" from-port="2" to-layer="1211" to-port="0" /> + <edge from-layer="1210" from-port="0" to-layer="1211" to-port="1" /> + <edge from-layer="1211" from-port="2" to-layer="1216" to-port="0" /> + <edge from-layer="1212" from-port="0" to-layer="1216" to-port="1" /> + <edge from-layer="1213" from-port="0" to-layer="1216" to-port="2" /> + <edge from-layer="1214" from-port="0" to-layer="1216" to-port="3" /> + <edge from-layer="1215" from-port="0" to-layer="1216" to-port="4" /> + <edge from-layer="1216" from-port="5" to-layer="1218" to-port="0" /> + <edge from-layer="1217" from-port="0" to-layer="1218" to-port="1" /> + <edge from-layer="1218" from-port="2" to-layer="1220" to-port="0" /> + <edge from-layer="1219" from-port="0" to-layer="1220" to-port="1" /> + <edge from-layer="1220" from-port="2" to-layer="1232" to-port="1" /> + <edge from-layer="1221" from-port="0" to-layer="1222" to-port="0" /> + <edge from-layer="1222" from-port="1" to-layer="1224" to-port="0" /> + <edge from-layer="1223" from-port="0" to-layer="1224" to-port="1" /> + <edge from-layer="1224" from-port="2" to-layer="1225" to-port="1" /> + <edge from-layer="1225" from-port="2" to-layer="1227" to-port="0" /> + <edge from-layer="1226" from-port="0" to-layer="1227" to-port="1" /> + <edge from-layer="1227" from-port="2" to-layer="1229" to-port="0" /> + <edge from-layer="1228" from-port="0" to-layer="1229" to-port="1" /> + <edge from-layer="1229" from-port="2" to-layer="1231" to-port="0" /> + <edge from-layer="1230" from-port="0" to-layer="1231" to-port="1" /> + <edge from-layer="1231" from-port="2" to-layer="1232" to-port="2" /> + <edge from-layer="1232" from-port="4" to-layer="1234" to-port="0" /> + <edge from-layer="1233" from-port="0" to-layer="1234" to-port="1" /> + <edge from-layer="1234" from-port="2" to-layer="1236" to-port="0" /> + <edge from-layer="1235" from-port="0" to-layer="1236" to-port="1" /> + <edge from-layer="1236" from-port="2" to-layer="1238" to-port="0" /> + <edge from-layer="1237" from-port="0" to-layer="1238" to-port="1" /> + <edge from-layer="1238" from-port="2" to-layer="1243" to-port="0" /> + <edge from-layer="1239" from-port="0" to-layer="1243" to-port="1" /> + <edge from-layer="1240" from-port="0" to-layer="1243" to-port="2" /> + <edge from-layer="1241" from-port="0" to-layer="1243" to-port="3" /> + <edge from-layer="1242" from-port="0" to-layer="1243" to-port="4" /> + <edge from-layer="1243" from-port="5" to-layer="1248" to-port="0" /> + <edge from-layer="1244" from-port="0" to-layer="1245" to-port="0" /> + <edge from-layer="1245" from-port="1" to-layer="1247" to-port="0" /> + <edge from-layer="1246" from-port="0" to-layer="1247" to-port="1" /> + <edge from-layer="1247" from-port="2" to-layer="1248" to-port="1" /> + <edge from-layer="1248" from-port="2" to-layer="1250" to-port="0" /> + <edge from-layer="1249" from-port="0" to-layer="1250" to-port="1" /> + <edge from-layer="1250" from-port="2" to-layer="1251" to-port="0" /> + <edge from-layer="1251" from-port="2" to-layer="1253" to-port="0" /> + <edge from-layer="1252" from-port="0" to-layer="1253" to-port="1" /> + <edge from-layer="1253" from-port="2" to-layer="1255" to-port="0" /> + <edge from-layer="1254" from-port="0" to-layer="1255" to-port="1" /> + <edge from-layer="1255" from-port="2" to-layer="1257" to-port="0" /> + <edge from-layer="1256" from-port="0" to-layer="1257" to-port="1" /> + <edge from-layer="1257" from-port="2" to-layer="1287" to-port="1" /> + <edge from-layer="1257" from-port="2" to-layer="1259" to-port="0" /> + <edge from-layer="1258" from-port="0" to-layer="1259" to-port="1" /> + <edge from-layer="1259" from-port="2" to-layer="1264" to-port="0" /> + <edge from-layer="1260" from-port="0" to-layer="1264" to-port="1" /> + <edge from-layer="1261" from-port="0" to-layer="1264" to-port="2" /> + <edge from-layer="1262" from-port="0" to-layer="1264" to-port="3" /> + <edge from-layer="1263" from-port="0" to-layer="1264" to-port="4" /> + <edge from-layer="1264" from-port="5" to-layer="1269" to-port="0" /> + <edge from-layer="1265" from-port="0" to-layer="1266" to-port="0" /> + <edge from-layer="1266" from-port="1" to-layer="1268" to-port="0" /> + <edge from-layer="1267" from-port="0" to-layer="1268" to-port="1" /> + <edge from-layer="1268" from-port="2" to-layer="1269" to-port="1" /> + <edge from-layer="1269" from-port="2" to-layer="1271" to-port="0" /> + <edge from-layer="1270" from-port="0" to-layer="1271" to-port="1" /> + <edge from-layer="1271" from-port="2" to-layer="1272" to-port="0" /> + <edge from-layer="1272" from-port="1" to-layer="1274" to-port="0" /> + <edge from-layer="1273" from-port="0" to-layer="1274" to-port="1" /> + <edge from-layer="1274" from-port="2" to-layer="1279" to-port="0" /> + <edge from-layer="1275" from-port="0" to-layer="1279" to-port="1" /> + <edge from-layer="1276" from-port="0" to-layer="1279" to-port="2" /> + <edge from-layer="1277" from-port="0" to-layer="1279" to-port="3" /> + <edge from-layer="1278" from-port="0" to-layer="1279" to-port="4" /> + <edge from-layer="1279" from-port="5" to-layer="1284" to-port="0" /> + <edge from-layer="1280" from-port="0" to-layer="1281" to-port="0" /> + <edge from-layer="1281" from-port="1" to-layer="1283" to-port="0" /> + <edge from-layer="1282" from-port="0" to-layer="1283" to-port="1" /> + <edge from-layer="1283" from-port="2" to-layer="1284" to-port="1" /> + <edge from-layer="1284" from-port="2" to-layer="1286" to-port="0" /> + <edge from-layer="1285" from-port="0" to-layer="1286" to-port="1" /> + <edge from-layer="1286" from-port="2" to-layer="1287" to-port="0" /> + <edge from-layer="1287" from-port="2" to-layer="1289" to-port="0" /> + <edge from-layer="1288" from-port="0" to-layer="1289" to-port="1" /> + <edge from-layer="1289" from-port="2" to-layer="1291" to-port="0" /> + <edge from-layer="1290" from-port="0" to-layer="1291" to-port="1" /> + <edge from-layer="1291" from-port="2" to-layer="1293" to-port="0" /> + <edge from-layer="1292" from-port="0" to-layer="1293" to-port="1" /> + <edge from-layer="1293" from-port="2" to-layer="1295" to-port="0" /> + <edge from-layer="1293" from-port="2" to-layer="1363" to-port="1" /> + <edge from-layer="1294" from-port="0" to-layer="1295" to-port="1" /> + <edge from-layer="1295" from-port="2" to-layer="1300" to-port="0" /> + <edge from-layer="1296" from-port="0" to-layer="1300" to-port="1" /> + <edge from-layer="1297" from-port="0" to-layer="1300" to-port="2" /> + <edge from-layer="1298" from-port="0" to-layer="1300" to-port="3" /> + <edge from-layer="1299" from-port="0" to-layer="1300" to-port="4" /> + <edge from-layer="1300" from-port="5" to-layer="1337" to-port="0" /> + <edge from-layer="1300" from-port="5" to-layer="1321" to-port="0" /> + <edge from-layer="1300" from-port="5" to-layer="1305" to-port="0" /> + <edge from-layer="1301" from-port="0" to-layer="1302" to-port="0" /> + <edge from-layer="1302" from-port="1" to-layer="1304" to-port="0" /> + <edge from-layer="1303" from-port="0" to-layer="1304" to-port="1" /> + <edge from-layer="1304" from-port="2" to-layer="1305" to-port="1" /> + <edge from-layer="1305" from-port="2" to-layer="1307" to-port="0" /> + <edge from-layer="1306" from-port="0" to-layer="1307" to-port="1" /> + <edge from-layer="1307" from-port="2" to-layer="1312" to-port="0" /> + <edge from-layer="1308" from-port="0" to-layer="1312" to-port="1" /> + <edge from-layer="1309" from-port="0" to-layer="1312" to-port="2" /> + <edge from-layer="1310" from-port="0" to-layer="1312" to-port="3" /> + <edge from-layer="1311" from-port="0" to-layer="1312" to-port="4" /> + <edge from-layer="1312" from-port="5" to-layer="1314" to-port="0" /> + <edge from-layer="1313" from-port="0" to-layer="1314" to-port="1" /> + <edge from-layer="1314" from-port="2" to-layer="1316" to-port="0" /> + <edge from-layer="1315" from-port="0" to-layer="1316" to-port="1" /> + <edge from-layer="1316" from-port="2" to-layer="1344" to-port="0" /> + <edge from-layer="1317" from-port="0" to-layer="1318" to-port="0" /> + <edge from-layer="1318" from-port="1" to-layer="1320" to-port="0" /> + <edge from-layer="1319" from-port="0" to-layer="1320" to-port="1" /> + <edge from-layer="1320" from-port="2" to-layer="1321" to-port="1" /> + <edge from-layer="1321" from-port="2" to-layer="1323" to-port="0" /> + <edge from-layer="1322" from-port="0" to-layer="1323" to-port="1" /> + <edge from-layer="1323" from-port="2" to-layer="1328" to-port="0" /> + <edge from-layer="1324" from-port="0" to-layer="1328" to-port="1" /> + <edge from-layer="1325" from-port="0" to-layer="1328" to-port="2" /> + <edge from-layer="1326" from-port="0" to-layer="1328" to-port="3" /> + <edge from-layer="1327" from-port="0" to-layer="1328" to-port="4" /> + <edge from-layer="1328" from-port="5" to-layer="1330" to-port="0" /> + <edge from-layer="1329" from-port="0" to-layer="1330" to-port="1" /> + <edge from-layer="1330" from-port="2" to-layer="1332" to-port="0" /> + <edge from-layer="1331" from-port="0" to-layer="1332" to-port="1" /> + <edge from-layer="1332" from-port="2" to-layer="1344" to-port="1" /> + <edge from-layer="1333" from-port="0" to-layer="1334" to-port="0" /> + <edge from-layer="1334" from-port="1" to-layer="1336" to-port="0" /> + <edge from-layer="1335" from-port="0" to-layer="1336" to-port="1" /> + <edge from-layer="1336" from-port="2" to-layer="1337" to-port="1" /> + <edge from-layer="1337" from-port="2" to-layer="1339" to-port="0" /> + <edge from-layer="1338" from-port="0" to-layer="1339" to-port="1" /> + <edge from-layer="1339" from-port="2" to-layer="1341" to-port="0" /> + <edge from-layer="1340" from-port="0" to-layer="1341" to-port="1" /> + <edge from-layer="1341" from-port="2" to-layer="1343" to-port="0" /> + <edge from-layer="1342" from-port="0" to-layer="1343" to-port="1" /> + <edge from-layer="1343" from-port="2" to-layer="1344" to-port="2" /> + <edge from-layer="1344" from-port="4" to-layer="1346" to-port="0" /> + <edge from-layer="1345" from-port="0" to-layer="1346" to-port="1" /> + <edge from-layer="1346" from-port="2" to-layer="1348" to-port="0" /> + <edge from-layer="1347" from-port="0" to-layer="1348" to-port="1" /> + <edge from-layer="1348" from-port="2" to-layer="1350" to-port="0" /> + <edge from-layer="1349" from-port="0" to-layer="1350" to-port="1" /> + <edge from-layer="1350" from-port="2" to-layer="1355" to-port="0" /> + <edge from-layer="1351" from-port="0" to-layer="1355" to-port="1" /> + <edge from-layer="1352" from-port="0" to-layer="1355" to-port="2" /> + <edge from-layer="1353" from-port="0" to-layer="1355" to-port="3" /> + <edge from-layer="1354" from-port="0" to-layer="1355" to-port="4" /> + <edge from-layer="1355" from-port="5" to-layer="1360" to-port="0" /> + <edge from-layer="1356" from-port="0" to-layer="1357" to-port="0" /> + <edge from-layer="1357" from-port="1" to-layer="1359" to-port="0" /> + <edge from-layer="1358" from-port="0" to-layer="1359" to-port="1" /> + <edge from-layer="1359" from-port="2" to-layer="1360" to-port="1" /> + <edge from-layer="1360" from-port="2" to-layer="1362" to-port="0" /> + <edge from-layer="1361" from-port="0" to-layer="1362" to-port="1" /> + <edge from-layer="1362" from-port="2" to-layer="1363" to-port="0" /> + <edge from-layer="1363" from-port="2" to-layer="1365" to-port="0" /> + <edge from-layer="1364" from-port="0" to-layer="1365" to-port="1" /> + <edge from-layer="1365" from-port="2" to-layer="1367" to-port="0" /> + <edge from-layer="1366" from-port="0" to-layer="1367" to-port="1" /> + <edge from-layer="1367" from-port="2" to-layer="1369" to-port="0" /> + <edge from-layer="1368" from-port="0" to-layer="1369" to-port="1" /> + <edge from-layer="1369" from-port="2" to-layer="1399" to-port="1" /> + <edge from-layer="1369" from-port="2" to-layer="1371" to-port="0" /> + <edge from-layer="1370" from-port="0" to-layer="1371" to-port="1" /> + <edge from-layer="1371" from-port="2" to-layer="1376" to-port="0" /> + <edge from-layer="1372" from-port="0" to-layer="1376" to-port="1" /> + <edge from-layer="1373" from-port="0" to-layer="1376" to-port="2" /> + <edge from-layer="1374" from-port="0" to-layer="1376" to-port="3" /> + <edge from-layer="1375" from-port="0" to-layer="1376" to-port="4" /> + <edge from-layer="1376" from-port="5" to-layer="1381" to-port="0" /> + <edge from-layer="1377" from-port="0" to-layer="1378" to-port="0" /> + <edge from-layer="1378" from-port="1" to-layer="1380" to-port="0" /> + <edge from-layer="1379" from-port="0" to-layer="1380" to-port="1" /> + <edge from-layer="1380" from-port="2" to-layer="1381" to-port="1" /> + <edge from-layer="1381" from-port="2" to-layer="1383" to-port="0" /> + <edge from-layer="1382" from-port="0" to-layer="1383" to-port="1" /> + <edge from-layer="1383" from-port="2" to-layer="1384" to-port="0" /> + <edge from-layer="1384" from-port="1" to-layer="1386" to-port="0" /> + <edge from-layer="1385" from-port="0" to-layer="1386" to-port="1" /> + <edge from-layer="1386" from-port="2" to-layer="1391" to-port="0" /> + <edge from-layer="1387" from-port="0" to-layer="1391" to-port="1" /> + <edge from-layer="1388" from-port="0" to-layer="1391" to-port="2" /> + <edge from-layer="1389" from-port="0" to-layer="1391" to-port="3" /> + <edge from-layer="1390" from-port="0" to-layer="1391" to-port="4" /> + <edge from-layer="1391" from-port="5" to-layer="1396" to-port="0" /> + <edge from-layer="1392" from-port="0" to-layer="1393" to-port="0" /> + <edge from-layer="1393" from-port="1" to-layer="1395" to-port="0" /> + <edge from-layer="1394" from-port="0" to-layer="1395" to-port="1" /> + <edge from-layer="1395" from-port="2" to-layer="1396" to-port="1" /> + <edge from-layer="1396" from-port="2" to-layer="1398" to-port="0" /> + <edge from-layer="1397" from-port="0" to-layer="1398" to-port="1" /> + <edge from-layer="1398" from-port="2" to-layer="1399" to-port="0" /> + <edge from-layer="1399" from-port="2" to-layer="1401" to-port="0" /> + <edge from-layer="1400" from-port="0" to-layer="1401" to-port="1" /> + <edge from-layer="1401" from-port="2" to-layer="1403" to-port="0" /> + <edge from-layer="1402" from-port="0" to-layer="1403" to-port="1" /> + <edge from-layer="1403" from-port="2" to-layer="1405" to-port="0" /> + <edge from-layer="1404" from-port="0" to-layer="1405" to-port="1" /> + <edge from-layer="1405" from-port="2" to-layer="1406" to-port="0" /> + </edges> + <rt_info> + <Runtime_version value="2024.4.1-16618-643f23d1318-releases/2024/4" /> + <conversion_parameters> + <framework value="pytorch" /> + <is_python_object value="True" /> + </conversion_parameters> + <nncf> + <friendly_names_were_updated value="True" /> + <quantization> + <advanced_parameters value="{'overflow_fix': 'disable', 'quantize_outputs': False, 'inplace_statistics': True, 'disable_channel_alignment': True, 'disable_bias_correction': False, 'batchwise_statistics': None, 'activations_quantization_params': None, 'weights_quantization_params': None, 'activations_range_estimator_params': {'min': {'statistics_type': None, 'aggregator_type': None, 'clipping_value': None, 'quantile_outlier_prob': 0.0001}, 'max': {'statistics_type': None, 'aggregator_type': None, 'clipping_value': None, 'quantile_outlier_prob': 0.0001}}, 'weights_range_estimator_params': {'min': {'statistics_type': None, 'aggregator_type': None, 'clipping_value': None, 'quantile_outlier_prob': 0.0001}, 'max': {'statistics_type': None, 'aggregator_type': None, 'clipping_value': None, 'quantile_outlier_prob': 0.0001}}, 'bias_correction_params': {'apply_for_all_nodes': False, 'threshold': None}, 'smooth_quant_alphas': {'convolution': -1, 'matmul': 0.95}, 'smooth_quant_alpha': None, 'backend_params': {}}" /> + <fast_bias_correction value="True" /> + <ignored_scope> + <types value="['GroupNormalization']" /> + </ignored_scope> + <model_type value="transformer" /> + <preset value="mixed" /> + <subset_size value="300" /> + <target_device value="ANY" /> + </quantization> + </nncf> + <optimum> + <optimum_intel_version value="1.20.1" /> + <optimum_version value="1.23.3" /> + <pytorch_version value="2.5.1" /> + <transformers_version value="4.46.2" /> + </optimum> + </rt_info> +</net>