diff --git "a/model-info.txt" "b/model-info.txt" new file mode 100644--- /dev/null +++ "b/model-info.txt" @@ -0,0 +1,953 @@ +Variable decoder/decoder_norm/scale size 384 shape (embed=384) partition spec (None,) +Variable decoder/layers_0/encoder_decoder_attention/key/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable decoder/layers_0/encoder_decoder_attention/out/kernel size 196608 shape (joined_kv=512, embed=384) partition spec ('model', None) +Variable decoder/layers_0/encoder_decoder_attention/query/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable decoder/layers_0/encoder_decoder_attention/value/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable decoder/layers_0/mlp/wi_0/kernel size 589824 shape (embed=384, mlp=1536) partition spec (None, 'model') +Variable decoder/layers_0/mlp/wi_1/kernel size 589824 shape (embed=384, mlp=1536) partition spec (None, 'model') +Variable decoder/layers_0/mlp/wo/kernel size 589824 shape (mlp=1536, embed=384) partition spec ('model', None) +Variable decoder/layers_0/pre_cross_attention_layer_norm/scale size 384 shape (embed=384) partition spec (None,) +Variable decoder/layers_0/pre_mlp_layer_norm/scale size 384 shape (embed=384) partition spec (None,) +Variable decoder/layers_0/pre_self_attention_layer_norm/scale size 384 shape (embed=384) partition spec (None,) +Variable decoder/layers_0/self_attention/key/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable decoder/layers_0/self_attention/out/kernel size 196608 shape (joined_kv=512, embed=384) partition spec ('model', None) +Variable decoder/layers_0/self_attention/query/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable decoder/layers_0/self_attention/value/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable decoder/layers_1/encoder_decoder_attention/key/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable decoder/layers_1/encoder_decoder_attention/out/kernel size 196608 shape (joined_kv=512, embed=384) partition spec ('model', None) +Variable decoder/layers_1/encoder_decoder_attention/query/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable decoder/layers_1/encoder_decoder_attention/value/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable decoder/layers_1/mlp/wi_0/kernel size 589824 shape (embed=384, mlp=1536) partition spec (None, 'model') +Variable decoder/layers_1/mlp/wi_1/kernel size 589824 shape (embed=384, mlp=1536) partition spec (None, 'model') +Variable decoder/layers_1/mlp/wo/kernel size 589824 shape (mlp=1536, embed=384) partition spec ('model', None) +Variable decoder/layers_1/pre_cross_attention_layer_norm/scale size 384 shape (embed=384) partition spec (None,) +Variable decoder/layers_1/pre_mlp_layer_norm/scale size 384 shape (embed=384) partition spec (None,) +Variable decoder/layers_1/pre_self_attention_layer_norm/scale size 384 shape (embed=384) partition spec (None,) +Variable decoder/layers_1/self_attention/key/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable decoder/layers_1/self_attention/out/kernel size 196608 shape (joined_kv=512, embed=384) partition spec ('model', None) +Variable decoder/layers_1/self_attention/query/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable decoder/layers_1/self_attention/value/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable decoder/layers_2/encoder_decoder_attention/key/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable decoder/layers_2/encoder_decoder_attention/out/kernel size 196608 shape (joined_kv=512, embed=384) partition spec ('model', None) +Variable decoder/layers_2/encoder_decoder_attention/query/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable decoder/layers_2/encoder_decoder_attention/value/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable decoder/layers_2/mlp/wi_0/kernel size 589824 shape (embed=384, mlp=1536) partition spec (None, 'model') +Variable decoder/layers_2/mlp/wi_1/kernel size 589824 shape (embed=384, mlp=1536) partition spec (None, 'model') +Variable decoder/layers_2/mlp/wo/kernel size 589824 shape (mlp=1536, embed=384) partition spec ('model', None) +Variable decoder/layers_2/pre_cross_attention_layer_norm/scale size 384 shape (embed=384) partition spec (None,) +Variable decoder/layers_2/pre_mlp_layer_norm/scale size 384 shape (embed=384) partition spec (None,) +Variable decoder/layers_2/pre_self_attention_layer_norm/scale size 384 shape (embed=384) partition spec (None,) +Variable decoder/layers_2/self_attention/key/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable decoder/layers_2/self_attention/out/kernel size 196608 shape (joined_kv=512, embed=384) partition spec ('model', None) +Variable decoder/layers_2/self_attention/query/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable decoder/layers_2/self_attention/value/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable decoder/layers_3/encoder_decoder_attention/key/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable decoder/layers_3/encoder_decoder_attention/out/kernel size 196608 shape (joined_kv=512, embed=384) partition spec ('model', None) +Variable decoder/layers_3/encoder_decoder_attention/query/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable decoder/layers_3/encoder_decoder_attention/value/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable decoder/layers_3/mlp/wi_0/kernel size 589824 shape (embed=384, mlp=1536) partition spec (None, 'model') +Variable decoder/layers_3/mlp/wi_1/kernel size 589824 shape (embed=384, mlp=1536) partition spec (None, 'model') +Variable decoder/layers_3/mlp/wo/kernel size 589824 shape (mlp=1536, embed=384) partition spec ('model', None) +Variable decoder/layers_3/pre_cross_attention_layer_norm/scale size 384 shape (embed=384) partition spec (None,) +Variable decoder/layers_3/pre_mlp_layer_norm/scale size 384 shape (embed=384) partition spec (None,) +Variable decoder/layers_3/pre_self_attention_layer_norm/scale size 384 shape (embed=384) partition spec (None,) +Variable decoder/layers_3/self_attention/key/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable decoder/layers_3/self_attention/out/kernel size 196608 shape (joined_kv=512, embed=384) partition spec ('model', None) +Variable decoder/layers_3/self_attention/query/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable decoder/layers_3/self_attention/value/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable decoder/layers_4/encoder_decoder_attention/key/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable decoder/layers_4/encoder_decoder_attention/out/kernel size 196608 shape (joined_kv=512, embed=384) partition spec ('model', None) +Variable decoder/layers_4/encoder_decoder_attention/query/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable decoder/layers_4/encoder_decoder_attention/value/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable decoder/layers_4/mlp/wi_0/kernel size 589824 shape (embed=384, mlp=1536) partition spec (None, 'model') +Variable decoder/layers_4/mlp/wi_1/kernel size 589824 shape (embed=384, mlp=1536) partition spec (None, 'model') +Variable decoder/layers_4/mlp/wo/kernel size 589824 shape (mlp=1536, embed=384) partition spec ('model', None) +Variable decoder/layers_4/pre_cross_attention_layer_norm/scale size 384 shape (embed=384) partition spec (None,) +Variable decoder/layers_4/pre_mlp_layer_norm/scale size 384 shape (embed=384) partition spec (None,) +Variable decoder/layers_4/pre_self_attention_layer_norm/scale size 384 shape (embed=384) partition spec (None,) +Variable decoder/layers_4/self_attention/key/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable decoder/layers_4/self_attention/out/kernel size 196608 shape (joined_kv=512, embed=384) partition spec ('model', None) +Variable decoder/layers_4/self_attention/query/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable decoder/layers_4/self_attention/value/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable decoder/layers_5/encoder_decoder_attention/key/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable decoder/layers_5/encoder_decoder_attention/out/kernel size 196608 shape (joined_kv=512, embed=384) partition spec ('model', None) +Variable decoder/layers_5/encoder_decoder_attention/query/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable decoder/layers_5/encoder_decoder_attention/value/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable decoder/layers_5/mlp/wi_0/kernel size 589824 shape (embed=384, mlp=1536) partition spec (None, 'model') +Variable decoder/layers_5/mlp/wi_1/kernel size 589824 shape (embed=384, mlp=1536) partition spec (None, 'model') +Variable decoder/layers_5/mlp/wo/kernel size 589824 shape (mlp=1536, embed=384) partition spec ('model', None) +Variable decoder/layers_5/pre_cross_attention_layer_norm/scale size 384 shape (embed=384) partition spec (None,) +Variable decoder/layers_5/pre_mlp_layer_norm/scale size 384 shape (embed=384) partition spec (None,) +Variable decoder/layers_5/pre_self_attention_layer_norm/scale size 384 shape (embed=384) partition spec (None,) +Variable decoder/layers_5/self_attention/key/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable decoder/layers_5/self_attention/out/kernel size 196608 shape (joined_kv=512, embed=384) partition spec ('model', None) +Variable decoder/layers_5/self_attention/query/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable decoder/layers_5/self_attention/value/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable decoder/layers_6/encoder_decoder_attention/key/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable decoder/layers_6/encoder_decoder_attention/out/kernel size 196608 shape (joined_kv=512, embed=384) partition spec ('model', None) +Variable decoder/layers_6/encoder_decoder_attention/query/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable decoder/layers_6/encoder_decoder_attention/value/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable decoder/layers_6/mlp/wi_0/kernel size 589824 shape (embed=384, mlp=1536) partition spec (None, 'model') +Variable decoder/layers_6/mlp/wi_1/kernel size 589824 shape (embed=384, mlp=1536) partition spec (None, 'model') +Variable decoder/layers_6/mlp/wo/kernel size 589824 shape (mlp=1536, embed=384) partition spec ('model', None) +Variable decoder/layers_6/pre_cross_attention_layer_norm/scale size 384 shape (embed=384) partition spec (None,) +Variable decoder/layers_6/pre_mlp_layer_norm/scale size 384 shape (embed=384) partition spec (None,) +Variable decoder/layers_6/pre_self_attention_layer_norm/scale size 384 shape (embed=384) partition spec (None,) +Variable decoder/layers_6/self_attention/key/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable decoder/layers_6/self_attention/out/kernel size 196608 shape (joined_kv=512, embed=384) partition spec ('model', None) +Variable decoder/layers_6/self_attention/query/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable decoder/layers_6/self_attention/value/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable decoder/layers_7/encoder_decoder_attention/key/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable decoder/layers_7/encoder_decoder_attention/out/kernel size 196608 shape (joined_kv=512, embed=384) partition spec ('model', None) +Variable decoder/layers_7/encoder_decoder_attention/query/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable decoder/layers_7/encoder_decoder_attention/value/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable decoder/layers_7/mlp/wi_0/kernel size 589824 shape (embed=384, mlp=1536) partition spec (None, 'model') +Variable decoder/layers_7/mlp/wi_1/kernel size 589824 shape (embed=384, mlp=1536) partition spec (None, 'model') +Variable decoder/layers_7/mlp/wo/kernel size 589824 shape (mlp=1536, embed=384) partition spec ('model', None) +Variable decoder/layers_7/pre_cross_attention_layer_norm/scale size 384 shape (embed=384) partition spec (None,) +Variable decoder/layers_7/pre_mlp_layer_norm/scale size 384 shape (embed=384) partition spec (None,) +Variable decoder/layers_7/pre_self_attention_layer_norm/scale size 384 shape (embed=384) partition spec (None,) +Variable decoder/layers_7/self_attention/key/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable decoder/layers_7/self_attention/out/kernel size 196608 shape (joined_kv=512, embed=384) partition spec ('model', None) +Variable decoder/layers_7/self_attention/query/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable decoder/layers_7/self_attention/value/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable decoder/logits_dense/kernel size 12337152 shape (embed=384, vocab=32128) partition spec (None, 'model') +Variable decoder/relpos_bias/rel_embedding size 256 shape (heads=8, relpos_buckets=32) partition spec ('model', None) +Variable encoder/encoder_norm/scale size 384 shape (embed=384) partition spec (None,) +Variable encoder/layers_0/attention/key/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable encoder/layers_0/attention/out/kernel size 196608 shape (joined_kv=512, embed=384) partition spec ('model', None) +Variable encoder/layers_0/attention/query/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable encoder/layers_0/attention/value/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable encoder/layers_0/mlp/wi_0/kernel size 589824 shape (embed=384, mlp=1536) partition spec (None, 'model') +Variable encoder/layers_0/mlp/wi_1/kernel size 589824 shape (embed=384, mlp=1536) partition spec (None, 'model') +Variable encoder/layers_0/mlp/wo/kernel size 589824 shape (mlp=1536, embed=384) partition spec ('model', None) +Variable encoder/layers_0/pre_attention_layer_norm/scale size 384 shape (embed=384) partition spec (None,) +Variable encoder/layers_0/pre_mlp_layer_norm/scale size 384 shape (embed=384) partition spec (None,) +Variable encoder/layers_1/attention/key/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable encoder/layers_1/attention/out/kernel size 196608 shape (joined_kv=512, embed=384) partition spec ('model', None) +Variable encoder/layers_1/attention/query/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable encoder/layers_1/attention/value/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable encoder/layers_1/mlp/wi_0/kernel size 589824 shape (embed=384, mlp=1536) partition spec (None, 'model') +Variable encoder/layers_1/mlp/wi_1/kernel size 589824 shape (embed=384, mlp=1536) partition spec (None, 'model') +Variable encoder/layers_1/mlp/wo/kernel size 589824 shape (mlp=1536, embed=384) partition spec ('model', None) +Variable encoder/layers_1/pre_attention_layer_norm/scale size 384 shape (embed=384) partition spec (None,) +Variable encoder/layers_1/pre_mlp_layer_norm/scale size 384 shape (embed=384) partition spec (None,) +Variable encoder/layers_2/attention/key/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable encoder/layers_2/attention/out/kernel size 196608 shape (joined_kv=512, embed=384) partition spec ('model', None) +Variable encoder/layers_2/attention/query/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable encoder/layers_2/attention/value/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable encoder/layers_2/mlp/wi_0/kernel size 589824 shape (embed=384, mlp=1536) partition spec (None, 'model') +Variable encoder/layers_2/mlp/wi_1/kernel size 589824 shape (embed=384, mlp=1536) partition spec (None, 'model') +Variable encoder/layers_2/mlp/wo/kernel size 589824 shape (mlp=1536, embed=384) partition spec ('model', None) +Variable encoder/layers_2/pre_attention_layer_norm/scale size 384 shape (embed=384) partition spec (None,) +Variable encoder/layers_2/pre_mlp_layer_norm/scale size 384 shape (embed=384) partition spec (None,) +Variable encoder/layers_3/attention/key/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable encoder/layers_3/attention/out/kernel size 196608 shape (joined_kv=512, embed=384) partition spec ('model', None) +Variable encoder/layers_3/attention/query/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable encoder/layers_3/attention/value/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable encoder/layers_3/mlp/wi_0/kernel size 589824 shape (embed=384, mlp=1536) partition spec (None, 'model') +Variable encoder/layers_3/mlp/wi_1/kernel size 589824 shape (embed=384, mlp=1536) partition spec (None, 'model') +Variable encoder/layers_3/mlp/wo/kernel size 589824 shape (mlp=1536, embed=384) partition spec ('model', None) +Variable encoder/layers_3/pre_attention_layer_norm/scale size 384 shape (embed=384) partition spec (None,) +Variable encoder/layers_3/pre_mlp_layer_norm/scale size 384 shape (embed=384) partition spec (None,) +Variable encoder/layers_4/attention/key/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable encoder/layers_4/attention/out/kernel size 196608 shape (joined_kv=512, embed=384) partition spec ('model', None) +Variable encoder/layers_4/attention/query/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable encoder/layers_4/attention/value/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable encoder/layers_4/mlp/wi_0/kernel size 589824 shape (embed=384, mlp=1536) partition spec (None, 'model') +Variable encoder/layers_4/mlp/wi_1/kernel size 589824 shape (embed=384, mlp=1536) partition spec (None, 'model') +Variable encoder/layers_4/mlp/wo/kernel size 589824 shape (mlp=1536, embed=384) partition spec ('model', None) +Variable encoder/layers_4/pre_attention_layer_norm/scale size 384 shape (embed=384) partition spec (None,) +Variable encoder/layers_4/pre_mlp_layer_norm/scale size 384 shape (embed=384) partition spec (None,) +Variable encoder/layers_5/attention/key/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable encoder/layers_5/attention/out/kernel size 196608 shape (joined_kv=512, embed=384) partition spec ('model', None) +Variable encoder/layers_5/attention/query/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable encoder/layers_5/attention/value/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable encoder/layers_5/mlp/wi_0/kernel size 589824 shape (embed=384, mlp=1536) partition spec (None, 'model') +Variable encoder/layers_5/mlp/wi_1/kernel size 589824 shape (embed=384, mlp=1536) partition spec (None, 'model') +Variable encoder/layers_5/mlp/wo/kernel size 589824 shape (mlp=1536, embed=384) partition spec ('model', None) +Variable encoder/layers_5/pre_attention_layer_norm/scale size 384 shape (embed=384) partition spec (None,) +Variable encoder/layers_5/pre_mlp_layer_norm/scale size 384 shape (embed=384) partition spec (None,) +Variable encoder/layers_6/attention/key/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable encoder/layers_6/attention/out/kernel size 196608 shape (joined_kv=512, embed=384) partition spec ('model', None) +Variable encoder/layers_6/attention/query/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable encoder/layers_6/attention/value/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable encoder/layers_6/mlp/wi_0/kernel size 589824 shape (embed=384, mlp=1536) partition spec (None, 'model') +Variable encoder/layers_6/mlp/wi_1/kernel size 589824 shape (embed=384, mlp=1536) partition spec (None, 'model') +Variable encoder/layers_6/mlp/wo/kernel size 589824 shape (mlp=1536, embed=384) partition spec ('model', None) +Variable encoder/layers_6/pre_attention_layer_norm/scale size 384 shape (embed=384) partition spec (None,) +Variable encoder/layers_6/pre_mlp_layer_norm/scale size 384 shape (embed=384) partition spec (None,) +Variable encoder/layers_7/attention/key/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable encoder/layers_7/attention/out/kernel size 196608 shape (joined_kv=512, embed=384) partition spec ('model', None) +Variable encoder/layers_7/attention/query/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable encoder/layers_7/attention/value/kernel size 196608 shape (embed=384, joined_kv=512) partition spec (None, 'model') +Variable encoder/layers_7/mlp/wi_0/kernel size 589824 shape (embed=384, mlp=1536) partition spec (None, 'model') +Variable encoder/layers_7/mlp/wi_1/kernel size 589824 shape (embed=384, mlp=1536) partition spec (None, 'model') +Variable encoder/layers_7/mlp/wo/kernel size 589824 shape (mlp=1536, embed=384) partition spec ('model', None) +Variable encoder/layers_7/pre_attention_layer_norm/scale size 384 shape (embed=384) partition spec (None,) +Variable encoder/layers_7/pre_mlp_layer_norm/scale size 384 shape (embed=384) partition spec (None,) +Variable encoder/relpos_bias/rel_embedding size 256 shape (heads=8, relpos_buckets=32) partition spec ('model', None) +Variable token_embedder/embedding size 12337152 shape (vocab=32128, embed=384) partition spec ('model', None) +Total number of parameters: 71876864 + +Variable param_states/decoder/decoder_norm/scale/m size 1 shape (1,) partition spec None +Variable param_states/decoder/decoder_norm/scale/v size 384 shape (embed=384) partition spec (None,) +Variable param_states/decoder/decoder_norm/scale/v_col size 1 shape (1,) partition spec None +Variable param_states/decoder/decoder_norm/scale/v_row size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_0/encoder_decoder_attention/key/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_0/encoder_decoder_attention/key/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_0/encoder_decoder_attention/key/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/decoder/layers_0/encoder_decoder_attention/key/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_0/encoder_decoder_attention/out/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_0/encoder_decoder_attention/out/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_0/encoder_decoder_attention/out/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/decoder/layers_0/encoder_decoder_attention/out/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_0/encoder_decoder_attention/query/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_0/encoder_decoder_attention/query/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_0/encoder_decoder_attention/query/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/decoder/layers_0/encoder_decoder_attention/query/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_0/encoder_decoder_attention/value/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_0/encoder_decoder_attention/value/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_0/encoder_decoder_attention/value/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/decoder/layers_0/encoder_decoder_attention/value/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_0/mlp/wi_0/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_0/mlp/wi_0/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_0/mlp/wi_0/kernel/v_col size 1536 shape (1536,) partition spec None +Variable param_states/decoder/layers_0/mlp/wi_0/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_0/mlp/wi_1/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_0/mlp/wi_1/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_0/mlp/wi_1/kernel/v_col size 1536 shape (1536,) partition spec None +Variable param_states/decoder/layers_0/mlp/wi_1/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_0/mlp/wo/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_0/mlp/wo/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_0/mlp/wo/kernel/v_col size 1536 shape (1536,) partition spec None +Variable param_states/decoder/layers_0/mlp/wo/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_0/pre_cross_attention_layer_norm/scale/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_0/pre_cross_attention_layer_norm/scale/v size 384 shape (embed=384) partition spec (None,) +Variable param_states/decoder/layers_0/pre_cross_attention_layer_norm/scale/v_col size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_0/pre_cross_attention_layer_norm/scale/v_row size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_0/pre_mlp_layer_norm/scale/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_0/pre_mlp_layer_norm/scale/v size 384 shape (embed=384) partition spec (None,) +Variable param_states/decoder/layers_0/pre_mlp_layer_norm/scale/v_col size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_0/pre_mlp_layer_norm/scale/v_row size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_0/pre_self_attention_layer_norm/scale/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_0/pre_self_attention_layer_norm/scale/v size 384 shape (embed=384) partition spec (None,) +Variable param_states/decoder/layers_0/pre_self_attention_layer_norm/scale/v_col size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_0/pre_self_attention_layer_norm/scale/v_row size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_0/self_attention/key/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_0/self_attention/key/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_0/self_attention/key/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/decoder/layers_0/self_attention/key/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_0/self_attention/out/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_0/self_attention/out/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_0/self_attention/out/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/decoder/layers_0/self_attention/out/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_0/self_attention/query/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_0/self_attention/query/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_0/self_attention/query/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/decoder/layers_0/self_attention/query/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_0/self_attention/value/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_0/self_attention/value/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_0/self_attention/value/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/decoder/layers_0/self_attention/value/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_1/encoder_decoder_attention/key/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_1/encoder_decoder_attention/key/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_1/encoder_decoder_attention/key/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/decoder/layers_1/encoder_decoder_attention/key/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_1/encoder_decoder_attention/out/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_1/encoder_decoder_attention/out/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_1/encoder_decoder_attention/out/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/decoder/layers_1/encoder_decoder_attention/out/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_1/encoder_decoder_attention/query/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_1/encoder_decoder_attention/query/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_1/encoder_decoder_attention/query/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/decoder/layers_1/encoder_decoder_attention/query/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_1/encoder_decoder_attention/value/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_1/encoder_decoder_attention/value/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_1/encoder_decoder_attention/value/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/decoder/layers_1/encoder_decoder_attention/value/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_1/mlp/wi_0/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_1/mlp/wi_0/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_1/mlp/wi_0/kernel/v_col size 1536 shape (1536,) partition spec None +Variable param_states/decoder/layers_1/mlp/wi_0/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_1/mlp/wi_1/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_1/mlp/wi_1/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_1/mlp/wi_1/kernel/v_col size 1536 shape (1536,) partition spec None +Variable param_states/decoder/layers_1/mlp/wi_1/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_1/mlp/wo/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_1/mlp/wo/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_1/mlp/wo/kernel/v_col size 1536 shape (1536,) partition spec None +Variable param_states/decoder/layers_1/mlp/wo/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_1/pre_cross_attention_layer_norm/scale/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_1/pre_cross_attention_layer_norm/scale/v size 384 shape (embed=384) partition spec (None,) +Variable param_states/decoder/layers_1/pre_cross_attention_layer_norm/scale/v_col size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_1/pre_cross_attention_layer_norm/scale/v_row size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_1/pre_mlp_layer_norm/scale/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_1/pre_mlp_layer_norm/scale/v size 384 shape (embed=384) partition spec (None,) +Variable param_states/decoder/layers_1/pre_mlp_layer_norm/scale/v_col size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_1/pre_mlp_layer_norm/scale/v_row size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_1/pre_self_attention_layer_norm/scale/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_1/pre_self_attention_layer_norm/scale/v size 384 shape (embed=384) partition spec (None,) +Variable param_states/decoder/layers_1/pre_self_attention_layer_norm/scale/v_col size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_1/pre_self_attention_layer_norm/scale/v_row size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_1/self_attention/key/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_1/self_attention/key/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_1/self_attention/key/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/decoder/layers_1/self_attention/key/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_1/self_attention/out/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_1/self_attention/out/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_1/self_attention/out/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/decoder/layers_1/self_attention/out/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_1/self_attention/query/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_1/self_attention/query/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_1/self_attention/query/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/decoder/layers_1/self_attention/query/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_1/self_attention/value/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_1/self_attention/value/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_1/self_attention/value/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/decoder/layers_1/self_attention/value/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_2/encoder_decoder_attention/key/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_2/encoder_decoder_attention/key/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_2/encoder_decoder_attention/key/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/decoder/layers_2/encoder_decoder_attention/key/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_2/encoder_decoder_attention/out/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_2/encoder_decoder_attention/out/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_2/encoder_decoder_attention/out/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/decoder/layers_2/encoder_decoder_attention/out/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_2/encoder_decoder_attention/query/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_2/encoder_decoder_attention/query/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_2/encoder_decoder_attention/query/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/decoder/layers_2/encoder_decoder_attention/query/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_2/encoder_decoder_attention/value/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_2/encoder_decoder_attention/value/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_2/encoder_decoder_attention/value/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/decoder/layers_2/encoder_decoder_attention/value/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_2/mlp/wi_0/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_2/mlp/wi_0/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_2/mlp/wi_0/kernel/v_col size 1536 shape (1536,) partition spec None +Variable param_states/decoder/layers_2/mlp/wi_0/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_2/mlp/wi_1/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_2/mlp/wi_1/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_2/mlp/wi_1/kernel/v_col size 1536 shape (1536,) partition spec None +Variable param_states/decoder/layers_2/mlp/wi_1/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_2/mlp/wo/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_2/mlp/wo/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_2/mlp/wo/kernel/v_col size 1536 shape (1536,) partition spec None +Variable param_states/decoder/layers_2/mlp/wo/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_2/pre_cross_attention_layer_norm/scale/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_2/pre_cross_attention_layer_norm/scale/v size 384 shape (embed=384) partition spec (None,) +Variable param_states/decoder/layers_2/pre_cross_attention_layer_norm/scale/v_col size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_2/pre_cross_attention_layer_norm/scale/v_row size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_2/pre_mlp_layer_norm/scale/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_2/pre_mlp_layer_norm/scale/v size 384 shape (embed=384) partition spec (None,) +Variable param_states/decoder/layers_2/pre_mlp_layer_norm/scale/v_col size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_2/pre_mlp_layer_norm/scale/v_row size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_2/pre_self_attention_layer_norm/scale/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_2/pre_self_attention_layer_norm/scale/v size 384 shape (embed=384) partition spec (None,) +Variable param_states/decoder/layers_2/pre_self_attention_layer_norm/scale/v_col size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_2/pre_self_attention_layer_norm/scale/v_row size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_2/self_attention/key/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_2/self_attention/key/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_2/self_attention/key/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/decoder/layers_2/self_attention/key/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_2/self_attention/out/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_2/self_attention/out/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_2/self_attention/out/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/decoder/layers_2/self_attention/out/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_2/self_attention/query/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_2/self_attention/query/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_2/self_attention/query/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/decoder/layers_2/self_attention/query/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_2/self_attention/value/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_2/self_attention/value/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_2/self_attention/value/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/decoder/layers_2/self_attention/value/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_3/encoder_decoder_attention/key/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_3/encoder_decoder_attention/key/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_3/encoder_decoder_attention/key/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/decoder/layers_3/encoder_decoder_attention/key/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_3/encoder_decoder_attention/out/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_3/encoder_decoder_attention/out/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_3/encoder_decoder_attention/out/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/decoder/layers_3/encoder_decoder_attention/out/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_3/encoder_decoder_attention/query/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_3/encoder_decoder_attention/query/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_3/encoder_decoder_attention/query/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/decoder/layers_3/encoder_decoder_attention/query/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_3/encoder_decoder_attention/value/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_3/encoder_decoder_attention/value/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_3/encoder_decoder_attention/value/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/decoder/layers_3/encoder_decoder_attention/value/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_3/mlp/wi_0/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_3/mlp/wi_0/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_3/mlp/wi_0/kernel/v_col size 1536 shape (1536,) partition spec None +Variable param_states/decoder/layers_3/mlp/wi_0/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_3/mlp/wi_1/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_3/mlp/wi_1/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_3/mlp/wi_1/kernel/v_col size 1536 shape (1536,) partition spec None +Variable param_states/decoder/layers_3/mlp/wi_1/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_3/mlp/wo/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_3/mlp/wo/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_3/mlp/wo/kernel/v_col size 1536 shape (1536,) partition spec None +Variable param_states/decoder/layers_3/mlp/wo/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_3/pre_cross_attention_layer_norm/scale/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_3/pre_cross_attention_layer_norm/scale/v size 384 shape (embed=384) partition spec (None,) +Variable param_states/decoder/layers_3/pre_cross_attention_layer_norm/scale/v_col size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_3/pre_cross_attention_layer_norm/scale/v_row size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_3/pre_mlp_layer_norm/scale/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_3/pre_mlp_layer_norm/scale/v size 384 shape (embed=384) partition spec (None,) +Variable param_states/decoder/layers_3/pre_mlp_layer_norm/scale/v_col size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_3/pre_mlp_layer_norm/scale/v_row size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_3/pre_self_attention_layer_norm/scale/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_3/pre_self_attention_layer_norm/scale/v size 384 shape (embed=384) partition spec (None,) +Variable param_states/decoder/layers_3/pre_self_attention_layer_norm/scale/v_col size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_3/pre_self_attention_layer_norm/scale/v_row size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_3/self_attention/key/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_3/self_attention/key/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_3/self_attention/key/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/decoder/layers_3/self_attention/key/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_3/self_attention/out/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_3/self_attention/out/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_3/self_attention/out/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/decoder/layers_3/self_attention/out/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_3/self_attention/query/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_3/self_attention/query/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_3/self_attention/query/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/decoder/layers_3/self_attention/query/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_3/self_attention/value/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_3/self_attention/value/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_3/self_attention/value/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/decoder/layers_3/self_attention/value/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_4/encoder_decoder_attention/key/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_4/encoder_decoder_attention/key/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_4/encoder_decoder_attention/key/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/decoder/layers_4/encoder_decoder_attention/key/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_4/encoder_decoder_attention/out/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_4/encoder_decoder_attention/out/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_4/encoder_decoder_attention/out/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/decoder/layers_4/encoder_decoder_attention/out/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_4/encoder_decoder_attention/query/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_4/encoder_decoder_attention/query/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_4/encoder_decoder_attention/query/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/decoder/layers_4/encoder_decoder_attention/query/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_4/encoder_decoder_attention/value/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_4/encoder_decoder_attention/value/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_4/encoder_decoder_attention/value/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/decoder/layers_4/encoder_decoder_attention/value/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_4/mlp/wi_0/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_4/mlp/wi_0/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_4/mlp/wi_0/kernel/v_col size 1536 shape (1536,) partition spec None +Variable param_states/decoder/layers_4/mlp/wi_0/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_4/mlp/wi_1/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_4/mlp/wi_1/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_4/mlp/wi_1/kernel/v_col size 1536 shape (1536,) partition spec None +Variable param_states/decoder/layers_4/mlp/wi_1/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_4/mlp/wo/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_4/mlp/wo/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_4/mlp/wo/kernel/v_col size 1536 shape (1536,) partition spec None +Variable param_states/decoder/layers_4/mlp/wo/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_4/pre_cross_attention_layer_norm/scale/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_4/pre_cross_attention_layer_norm/scale/v size 384 shape (embed=384) partition spec (None,) +Variable param_states/decoder/layers_4/pre_cross_attention_layer_norm/scale/v_col size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_4/pre_cross_attention_layer_norm/scale/v_row size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_4/pre_mlp_layer_norm/scale/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_4/pre_mlp_layer_norm/scale/v size 384 shape (embed=384) partition spec (None,) +Variable param_states/decoder/layers_4/pre_mlp_layer_norm/scale/v_col size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_4/pre_mlp_layer_norm/scale/v_row size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_4/pre_self_attention_layer_norm/scale/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_4/pre_self_attention_layer_norm/scale/v size 384 shape (embed=384) partition spec (None,) +Variable param_states/decoder/layers_4/pre_self_attention_layer_norm/scale/v_col size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_4/pre_self_attention_layer_norm/scale/v_row size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_4/self_attention/key/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_4/self_attention/key/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_4/self_attention/key/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/decoder/layers_4/self_attention/key/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_4/self_attention/out/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_4/self_attention/out/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_4/self_attention/out/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/decoder/layers_4/self_attention/out/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_4/self_attention/query/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_4/self_attention/query/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_4/self_attention/query/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/decoder/layers_4/self_attention/query/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_4/self_attention/value/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_4/self_attention/value/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_4/self_attention/value/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/decoder/layers_4/self_attention/value/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_5/encoder_decoder_attention/key/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_5/encoder_decoder_attention/key/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_5/encoder_decoder_attention/key/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/decoder/layers_5/encoder_decoder_attention/key/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_5/encoder_decoder_attention/out/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_5/encoder_decoder_attention/out/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_5/encoder_decoder_attention/out/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/decoder/layers_5/encoder_decoder_attention/out/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_5/encoder_decoder_attention/query/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_5/encoder_decoder_attention/query/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_5/encoder_decoder_attention/query/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/decoder/layers_5/encoder_decoder_attention/query/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_5/encoder_decoder_attention/value/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_5/encoder_decoder_attention/value/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_5/encoder_decoder_attention/value/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/decoder/layers_5/encoder_decoder_attention/value/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_5/mlp/wi_0/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_5/mlp/wi_0/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_5/mlp/wi_0/kernel/v_col size 1536 shape (1536,) partition spec None +Variable param_states/decoder/layers_5/mlp/wi_0/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_5/mlp/wi_1/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_5/mlp/wi_1/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_5/mlp/wi_1/kernel/v_col size 1536 shape (1536,) partition spec None +Variable param_states/decoder/layers_5/mlp/wi_1/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_5/mlp/wo/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_5/mlp/wo/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_5/mlp/wo/kernel/v_col size 1536 shape (1536,) partition spec None +Variable param_states/decoder/layers_5/mlp/wo/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_5/pre_cross_attention_layer_norm/scale/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_5/pre_cross_attention_layer_norm/scale/v size 384 shape (embed=384) partition spec (None,) +Variable param_states/decoder/layers_5/pre_cross_attention_layer_norm/scale/v_col size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_5/pre_cross_attention_layer_norm/scale/v_row size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_5/pre_mlp_layer_norm/scale/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_5/pre_mlp_layer_norm/scale/v size 384 shape (embed=384) partition spec (None,) +Variable param_states/decoder/layers_5/pre_mlp_layer_norm/scale/v_col size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_5/pre_mlp_layer_norm/scale/v_row size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_5/pre_self_attention_layer_norm/scale/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_5/pre_self_attention_layer_norm/scale/v size 384 shape (embed=384) partition spec (None,) +Variable param_states/decoder/layers_5/pre_self_attention_layer_norm/scale/v_col size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_5/pre_self_attention_layer_norm/scale/v_row size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_5/self_attention/key/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_5/self_attention/key/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_5/self_attention/key/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/decoder/layers_5/self_attention/key/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_5/self_attention/out/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_5/self_attention/out/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_5/self_attention/out/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/decoder/layers_5/self_attention/out/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_5/self_attention/query/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_5/self_attention/query/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_5/self_attention/query/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/decoder/layers_5/self_attention/query/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_5/self_attention/value/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_5/self_attention/value/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_5/self_attention/value/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/decoder/layers_5/self_attention/value/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_6/encoder_decoder_attention/key/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_6/encoder_decoder_attention/key/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_6/encoder_decoder_attention/key/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/decoder/layers_6/encoder_decoder_attention/key/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_6/encoder_decoder_attention/out/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_6/encoder_decoder_attention/out/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_6/encoder_decoder_attention/out/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/decoder/layers_6/encoder_decoder_attention/out/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_6/encoder_decoder_attention/query/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_6/encoder_decoder_attention/query/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_6/encoder_decoder_attention/query/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/decoder/layers_6/encoder_decoder_attention/query/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_6/encoder_decoder_attention/value/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_6/encoder_decoder_attention/value/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_6/encoder_decoder_attention/value/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/decoder/layers_6/encoder_decoder_attention/value/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_6/mlp/wi_0/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_6/mlp/wi_0/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_6/mlp/wi_0/kernel/v_col size 1536 shape (1536,) partition spec None +Variable param_states/decoder/layers_6/mlp/wi_0/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_6/mlp/wi_1/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_6/mlp/wi_1/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_6/mlp/wi_1/kernel/v_col size 1536 shape (1536,) partition spec None +Variable param_states/decoder/layers_6/mlp/wi_1/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_6/mlp/wo/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_6/mlp/wo/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_6/mlp/wo/kernel/v_col size 1536 shape (1536,) partition spec None +Variable param_states/decoder/layers_6/mlp/wo/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_6/pre_cross_attention_layer_norm/scale/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_6/pre_cross_attention_layer_norm/scale/v size 384 shape (embed=384) partition spec (None,) +Variable param_states/decoder/layers_6/pre_cross_attention_layer_norm/scale/v_col size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_6/pre_cross_attention_layer_norm/scale/v_row size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_6/pre_mlp_layer_norm/scale/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_6/pre_mlp_layer_norm/scale/v size 384 shape (embed=384) partition spec (None,) +Variable param_states/decoder/layers_6/pre_mlp_layer_norm/scale/v_col size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_6/pre_mlp_layer_norm/scale/v_row size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_6/pre_self_attention_layer_norm/scale/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_6/pre_self_attention_layer_norm/scale/v size 384 shape (embed=384) partition spec (None,) +Variable param_states/decoder/layers_6/pre_self_attention_layer_norm/scale/v_col size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_6/pre_self_attention_layer_norm/scale/v_row size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_6/self_attention/key/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_6/self_attention/key/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_6/self_attention/key/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/decoder/layers_6/self_attention/key/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_6/self_attention/out/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_6/self_attention/out/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_6/self_attention/out/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/decoder/layers_6/self_attention/out/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_6/self_attention/query/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_6/self_attention/query/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_6/self_attention/query/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/decoder/layers_6/self_attention/query/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_6/self_attention/value/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_6/self_attention/value/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_6/self_attention/value/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/decoder/layers_6/self_attention/value/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_7/encoder_decoder_attention/key/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_7/encoder_decoder_attention/key/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_7/encoder_decoder_attention/key/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/decoder/layers_7/encoder_decoder_attention/key/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_7/encoder_decoder_attention/out/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_7/encoder_decoder_attention/out/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_7/encoder_decoder_attention/out/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/decoder/layers_7/encoder_decoder_attention/out/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_7/encoder_decoder_attention/query/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_7/encoder_decoder_attention/query/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_7/encoder_decoder_attention/query/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/decoder/layers_7/encoder_decoder_attention/query/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_7/encoder_decoder_attention/value/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_7/encoder_decoder_attention/value/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_7/encoder_decoder_attention/value/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/decoder/layers_7/encoder_decoder_attention/value/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_7/mlp/wi_0/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_7/mlp/wi_0/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_7/mlp/wi_0/kernel/v_col size 1536 shape (1536,) partition spec None +Variable param_states/decoder/layers_7/mlp/wi_0/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_7/mlp/wi_1/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_7/mlp/wi_1/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_7/mlp/wi_1/kernel/v_col size 1536 shape (1536,) partition spec None +Variable param_states/decoder/layers_7/mlp/wi_1/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_7/mlp/wo/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_7/mlp/wo/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_7/mlp/wo/kernel/v_col size 1536 shape (1536,) partition spec None +Variable param_states/decoder/layers_7/mlp/wo/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_7/pre_cross_attention_layer_norm/scale/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_7/pre_cross_attention_layer_norm/scale/v size 384 shape (embed=384) partition spec (None,) +Variable param_states/decoder/layers_7/pre_cross_attention_layer_norm/scale/v_col size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_7/pre_cross_attention_layer_norm/scale/v_row size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_7/pre_mlp_layer_norm/scale/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_7/pre_mlp_layer_norm/scale/v size 384 shape (embed=384) partition spec (None,) +Variable param_states/decoder/layers_7/pre_mlp_layer_norm/scale/v_col size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_7/pre_mlp_layer_norm/scale/v_row size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_7/pre_self_attention_layer_norm/scale/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_7/pre_self_attention_layer_norm/scale/v size 384 shape (embed=384) partition spec (None,) +Variable param_states/decoder/layers_7/pre_self_attention_layer_norm/scale/v_col size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_7/pre_self_attention_layer_norm/scale/v_row size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_7/self_attention/key/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_7/self_attention/key/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_7/self_attention/key/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/decoder/layers_7/self_attention/key/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_7/self_attention/out/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_7/self_attention/out/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_7/self_attention/out/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/decoder/layers_7/self_attention/out/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_7/self_attention/query/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_7/self_attention/query/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_7/self_attention/query/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/decoder/layers_7/self_attention/query/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/layers_7/self_attention/value/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_7/self_attention/value/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/layers_7/self_attention/value/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/decoder/layers_7/self_attention/value/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/logits_dense/kernel/m size 1 shape (1,) partition spec None +Variable param_states/decoder/logits_dense/kernel/v size 1 shape (1,) partition spec None +Variable param_states/decoder/logits_dense/kernel/v_col size 32128 shape (32128,) partition spec None +Variable param_states/decoder/logits_dense/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/decoder/relpos_bias/rel_embedding/m size 1 shape (1,) partition spec None +Variable param_states/decoder/relpos_bias/rel_embedding/v size 256 shape (heads=8, relpos_buckets=32) partition spec ('model', None) +Variable param_states/decoder/relpos_bias/rel_embedding/v_col size 1 shape (1,) partition spec None +Variable param_states/decoder/relpos_bias/rel_embedding/v_row size 1 shape (1,) partition spec None +Variable param_states/encoder/encoder_norm/scale/m size 1 shape (1,) partition spec None +Variable param_states/encoder/encoder_norm/scale/v size 384 shape (embed=384) partition spec (None,) +Variable param_states/encoder/encoder_norm/scale/v_col size 1 shape (1,) partition spec None +Variable param_states/encoder/encoder_norm/scale/v_row size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_0/attention/key/kernel/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_0/attention/key/kernel/v size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_0/attention/key/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/encoder/layers_0/attention/key/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/encoder/layers_0/attention/out/kernel/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_0/attention/out/kernel/v size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_0/attention/out/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/encoder/layers_0/attention/out/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/encoder/layers_0/attention/query/kernel/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_0/attention/query/kernel/v size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_0/attention/query/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/encoder/layers_0/attention/query/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/encoder/layers_0/attention/value/kernel/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_0/attention/value/kernel/v size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_0/attention/value/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/encoder/layers_0/attention/value/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/encoder/layers_0/mlp/wi_0/kernel/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_0/mlp/wi_0/kernel/v size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_0/mlp/wi_0/kernel/v_col size 1536 shape (1536,) partition spec None +Variable param_states/encoder/layers_0/mlp/wi_0/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/encoder/layers_0/mlp/wi_1/kernel/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_0/mlp/wi_1/kernel/v size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_0/mlp/wi_1/kernel/v_col size 1536 shape (1536,) partition spec None +Variable param_states/encoder/layers_0/mlp/wi_1/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/encoder/layers_0/mlp/wo/kernel/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_0/mlp/wo/kernel/v size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_0/mlp/wo/kernel/v_col size 1536 shape (1536,) partition spec None +Variable param_states/encoder/layers_0/mlp/wo/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/encoder/layers_0/pre_attention_layer_norm/scale/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_0/pre_attention_layer_norm/scale/v size 384 shape (embed=384) partition spec (None,) +Variable param_states/encoder/layers_0/pre_attention_layer_norm/scale/v_col size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_0/pre_attention_layer_norm/scale/v_row size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_0/pre_mlp_layer_norm/scale/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_0/pre_mlp_layer_norm/scale/v size 384 shape (embed=384) partition spec (None,) +Variable param_states/encoder/layers_0/pre_mlp_layer_norm/scale/v_col size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_0/pre_mlp_layer_norm/scale/v_row size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_1/attention/key/kernel/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_1/attention/key/kernel/v size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_1/attention/key/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/encoder/layers_1/attention/key/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/encoder/layers_1/attention/out/kernel/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_1/attention/out/kernel/v size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_1/attention/out/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/encoder/layers_1/attention/out/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/encoder/layers_1/attention/query/kernel/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_1/attention/query/kernel/v size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_1/attention/query/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/encoder/layers_1/attention/query/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/encoder/layers_1/attention/value/kernel/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_1/attention/value/kernel/v size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_1/attention/value/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/encoder/layers_1/attention/value/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/encoder/layers_1/mlp/wi_0/kernel/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_1/mlp/wi_0/kernel/v size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_1/mlp/wi_0/kernel/v_col size 1536 shape (1536,) partition spec None +Variable param_states/encoder/layers_1/mlp/wi_0/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/encoder/layers_1/mlp/wi_1/kernel/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_1/mlp/wi_1/kernel/v size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_1/mlp/wi_1/kernel/v_col size 1536 shape (1536,) partition spec None +Variable param_states/encoder/layers_1/mlp/wi_1/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/encoder/layers_1/mlp/wo/kernel/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_1/mlp/wo/kernel/v size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_1/mlp/wo/kernel/v_col size 1536 shape (1536,) partition spec None +Variable param_states/encoder/layers_1/mlp/wo/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/encoder/layers_1/pre_attention_layer_norm/scale/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_1/pre_attention_layer_norm/scale/v size 384 shape (embed=384) partition spec (None,) +Variable param_states/encoder/layers_1/pre_attention_layer_norm/scale/v_col size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_1/pre_attention_layer_norm/scale/v_row size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_1/pre_mlp_layer_norm/scale/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_1/pre_mlp_layer_norm/scale/v size 384 shape (embed=384) partition spec (None,) +Variable param_states/encoder/layers_1/pre_mlp_layer_norm/scale/v_col size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_1/pre_mlp_layer_norm/scale/v_row size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_2/attention/key/kernel/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_2/attention/key/kernel/v size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_2/attention/key/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/encoder/layers_2/attention/key/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/encoder/layers_2/attention/out/kernel/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_2/attention/out/kernel/v size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_2/attention/out/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/encoder/layers_2/attention/out/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/encoder/layers_2/attention/query/kernel/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_2/attention/query/kernel/v size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_2/attention/query/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/encoder/layers_2/attention/query/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/encoder/layers_2/attention/value/kernel/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_2/attention/value/kernel/v size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_2/attention/value/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/encoder/layers_2/attention/value/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/encoder/layers_2/mlp/wi_0/kernel/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_2/mlp/wi_0/kernel/v size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_2/mlp/wi_0/kernel/v_col size 1536 shape (1536,) partition spec None +Variable param_states/encoder/layers_2/mlp/wi_0/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/encoder/layers_2/mlp/wi_1/kernel/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_2/mlp/wi_1/kernel/v size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_2/mlp/wi_1/kernel/v_col size 1536 shape (1536,) partition spec None +Variable param_states/encoder/layers_2/mlp/wi_1/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/encoder/layers_2/mlp/wo/kernel/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_2/mlp/wo/kernel/v size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_2/mlp/wo/kernel/v_col size 1536 shape (1536,) partition spec None +Variable param_states/encoder/layers_2/mlp/wo/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/encoder/layers_2/pre_attention_layer_norm/scale/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_2/pre_attention_layer_norm/scale/v size 384 shape (embed=384) partition spec (None,) +Variable param_states/encoder/layers_2/pre_attention_layer_norm/scale/v_col size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_2/pre_attention_layer_norm/scale/v_row size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_2/pre_mlp_layer_norm/scale/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_2/pre_mlp_layer_norm/scale/v size 384 shape (embed=384) partition spec (None,) +Variable param_states/encoder/layers_2/pre_mlp_layer_norm/scale/v_col size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_2/pre_mlp_layer_norm/scale/v_row size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_3/attention/key/kernel/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_3/attention/key/kernel/v size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_3/attention/key/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/encoder/layers_3/attention/key/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/encoder/layers_3/attention/out/kernel/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_3/attention/out/kernel/v size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_3/attention/out/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/encoder/layers_3/attention/out/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/encoder/layers_3/attention/query/kernel/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_3/attention/query/kernel/v size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_3/attention/query/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/encoder/layers_3/attention/query/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/encoder/layers_3/attention/value/kernel/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_3/attention/value/kernel/v size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_3/attention/value/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/encoder/layers_3/attention/value/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/encoder/layers_3/mlp/wi_0/kernel/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_3/mlp/wi_0/kernel/v size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_3/mlp/wi_0/kernel/v_col size 1536 shape (1536,) partition spec None +Variable param_states/encoder/layers_3/mlp/wi_0/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/encoder/layers_3/mlp/wi_1/kernel/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_3/mlp/wi_1/kernel/v size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_3/mlp/wi_1/kernel/v_col size 1536 shape (1536,) partition spec None +Variable param_states/encoder/layers_3/mlp/wi_1/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/encoder/layers_3/mlp/wo/kernel/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_3/mlp/wo/kernel/v size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_3/mlp/wo/kernel/v_col size 1536 shape (1536,) partition spec None +Variable param_states/encoder/layers_3/mlp/wo/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/encoder/layers_3/pre_attention_layer_norm/scale/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_3/pre_attention_layer_norm/scale/v size 384 shape (embed=384) partition spec (None,) +Variable param_states/encoder/layers_3/pre_attention_layer_norm/scale/v_col size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_3/pre_attention_layer_norm/scale/v_row size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_3/pre_mlp_layer_norm/scale/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_3/pre_mlp_layer_norm/scale/v size 384 shape (embed=384) partition spec (None,) +Variable param_states/encoder/layers_3/pre_mlp_layer_norm/scale/v_col size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_3/pre_mlp_layer_norm/scale/v_row size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_4/attention/key/kernel/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_4/attention/key/kernel/v size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_4/attention/key/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/encoder/layers_4/attention/key/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/encoder/layers_4/attention/out/kernel/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_4/attention/out/kernel/v size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_4/attention/out/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/encoder/layers_4/attention/out/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/encoder/layers_4/attention/query/kernel/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_4/attention/query/kernel/v size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_4/attention/query/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/encoder/layers_4/attention/query/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/encoder/layers_4/attention/value/kernel/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_4/attention/value/kernel/v size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_4/attention/value/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/encoder/layers_4/attention/value/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/encoder/layers_4/mlp/wi_0/kernel/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_4/mlp/wi_0/kernel/v size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_4/mlp/wi_0/kernel/v_col size 1536 shape (1536,) partition spec None +Variable param_states/encoder/layers_4/mlp/wi_0/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/encoder/layers_4/mlp/wi_1/kernel/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_4/mlp/wi_1/kernel/v size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_4/mlp/wi_1/kernel/v_col size 1536 shape (1536,) partition spec None +Variable param_states/encoder/layers_4/mlp/wi_1/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/encoder/layers_4/mlp/wo/kernel/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_4/mlp/wo/kernel/v size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_4/mlp/wo/kernel/v_col size 1536 shape (1536,) partition spec None +Variable param_states/encoder/layers_4/mlp/wo/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/encoder/layers_4/pre_attention_layer_norm/scale/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_4/pre_attention_layer_norm/scale/v size 384 shape (embed=384) partition spec (None,) +Variable param_states/encoder/layers_4/pre_attention_layer_norm/scale/v_col size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_4/pre_attention_layer_norm/scale/v_row size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_4/pre_mlp_layer_norm/scale/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_4/pre_mlp_layer_norm/scale/v size 384 shape (embed=384) partition spec (None,) +Variable param_states/encoder/layers_4/pre_mlp_layer_norm/scale/v_col size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_4/pre_mlp_layer_norm/scale/v_row size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_5/attention/key/kernel/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_5/attention/key/kernel/v size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_5/attention/key/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/encoder/layers_5/attention/key/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/encoder/layers_5/attention/out/kernel/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_5/attention/out/kernel/v size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_5/attention/out/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/encoder/layers_5/attention/out/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/encoder/layers_5/attention/query/kernel/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_5/attention/query/kernel/v size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_5/attention/query/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/encoder/layers_5/attention/query/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/encoder/layers_5/attention/value/kernel/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_5/attention/value/kernel/v size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_5/attention/value/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/encoder/layers_5/attention/value/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/encoder/layers_5/mlp/wi_0/kernel/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_5/mlp/wi_0/kernel/v size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_5/mlp/wi_0/kernel/v_col size 1536 shape (1536,) partition spec None +Variable param_states/encoder/layers_5/mlp/wi_0/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/encoder/layers_5/mlp/wi_1/kernel/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_5/mlp/wi_1/kernel/v size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_5/mlp/wi_1/kernel/v_col size 1536 shape (1536,) partition spec None +Variable param_states/encoder/layers_5/mlp/wi_1/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/encoder/layers_5/mlp/wo/kernel/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_5/mlp/wo/kernel/v size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_5/mlp/wo/kernel/v_col size 1536 shape (1536,) partition spec None +Variable param_states/encoder/layers_5/mlp/wo/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/encoder/layers_5/pre_attention_layer_norm/scale/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_5/pre_attention_layer_norm/scale/v size 384 shape (embed=384) partition spec (None,) +Variable param_states/encoder/layers_5/pre_attention_layer_norm/scale/v_col size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_5/pre_attention_layer_norm/scale/v_row size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_5/pre_mlp_layer_norm/scale/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_5/pre_mlp_layer_norm/scale/v size 384 shape (embed=384) partition spec (None,) +Variable param_states/encoder/layers_5/pre_mlp_layer_norm/scale/v_col size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_5/pre_mlp_layer_norm/scale/v_row size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_6/attention/key/kernel/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_6/attention/key/kernel/v size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_6/attention/key/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/encoder/layers_6/attention/key/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/encoder/layers_6/attention/out/kernel/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_6/attention/out/kernel/v size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_6/attention/out/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/encoder/layers_6/attention/out/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/encoder/layers_6/attention/query/kernel/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_6/attention/query/kernel/v size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_6/attention/query/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/encoder/layers_6/attention/query/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/encoder/layers_6/attention/value/kernel/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_6/attention/value/kernel/v size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_6/attention/value/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/encoder/layers_6/attention/value/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/encoder/layers_6/mlp/wi_0/kernel/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_6/mlp/wi_0/kernel/v size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_6/mlp/wi_0/kernel/v_col size 1536 shape (1536,) partition spec None +Variable param_states/encoder/layers_6/mlp/wi_0/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/encoder/layers_6/mlp/wi_1/kernel/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_6/mlp/wi_1/kernel/v size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_6/mlp/wi_1/kernel/v_col size 1536 shape (1536,) partition spec None +Variable param_states/encoder/layers_6/mlp/wi_1/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/encoder/layers_6/mlp/wo/kernel/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_6/mlp/wo/kernel/v size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_6/mlp/wo/kernel/v_col size 1536 shape (1536,) partition spec None +Variable param_states/encoder/layers_6/mlp/wo/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/encoder/layers_6/pre_attention_layer_norm/scale/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_6/pre_attention_layer_norm/scale/v size 384 shape (embed=384) partition spec (None,) +Variable param_states/encoder/layers_6/pre_attention_layer_norm/scale/v_col size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_6/pre_attention_layer_norm/scale/v_row size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_6/pre_mlp_layer_norm/scale/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_6/pre_mlp_layer_norm/scale/v size 384 shape (embed=384) partition spec (None,) +Variable param_states/encoder/layers_6/pre_mlp_layer_norm/scale/v_col size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_6/pre_mlp_layer_norm/scale/v_row size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_7/attention/key/kernel/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_7/attention/key/kernel/v size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_7/attention/key/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/encoder/layers_7/attention/key/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/encoder/layers_7/attention/out/kernel/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_7/attention/out/kernel/v size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_7/attention/out/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/encoder/layers_7/attention/out/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/encoder/layers_7/attention/query/kernel/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_7/attention/query/kernel/v size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_7/attention/query/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/encoder/layers_7/attention/query/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/encoder/layers_7/attention/value/kernel/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_7/attention/value/kernel/v size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_7/attention/value/kernel/v_col size 512 shape (512,) partition spec None +Variable param_states/encoder/layers_7/attention/value/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/encoder/layers_7/mlp/wi_0/kernel/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_7/mlp/wi_0/kernel/v size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_7/mlp/wi_0/kernel/v_col size 1536 shape (1536,) partition spec None +Variable param_states/encoder/layers_7/mlp/wi_0/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/encoder/layers_7/mlp/wi_1/kernel/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_7/mlp/wi_1/kernel/v size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_7/mlp/wi_1/kernel/v_col size 1536 shape (1536,) partition spec None +Variable param_states/encoder/layers_7/mlp/wi_1/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/encoder/layers_7/mlp/wo/kernel/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_7/mlp/wo/kernel/v size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_7/mlp/wo/kernel/v_col size 1536 shape (1536,) partition spec None +Variable param_states/encoder/layers_7/mlp/wo/kernel/v_row size 384 shape (384,) partition spec None +Variable param_states/encoder/layers_7/pre_attention_layer_norm/scale/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_7/pre_attention_layer_norm/scale/v size 384 shape (embed=384) partition spec (None,) +Variable param_states/encoder/layers_7/pre_attention_layer_norm/scale/v_col size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_7/pre_attention_layer_norm/scale/v_row size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_7/pre_mlp_layer_norm/scale/m size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_7/pre_mlp_layer_norm/scale/v size 384 shape (embed=384) partition spec (None,) +Variable param_states/encoder/layers_7/pre_mlp_layer_norm/scale/v_col size 1 shape (1,) partition spec None +Variable param_states/encoder/layers_7/pre_mlp_layer_norm/scale/v_row size 1 shape (1,) partition spec None +Variable param_states/encoder/relpos_bias/rel_embedding/m size 1 shape (1,) partition spec None +Variable param_states/encoder/relpos_bias/rel_embedding/v size 256 shape (heads=8, relpos_buckets=32) partition spec ('model', None) +Variable param_states/encoder/relpos_bias/rel_embedding/v_col size 1 shape (1,) partition spec None +Variable param_states/encoder/relpos_bias/rel_embedding/v_row size 1 shape (1,) partition spec None +Variable param_states/token_embedder/embedding/m size 1 shape (1,) partition spec None +Variable param_states/token_embedder/embedding/v size 1 shape (1,) partition spec None +Variable param_states/token_embedder/embedding/v_col size 32128 shape (32128,) partition spec None +Variable param_states/token_embedder/embedding/v_row size 384 shape (384,) partition spec None +Variable step size 1 shape () partition spec None