| { | |
| "metadata": { | |
| "total_size": 18811170816.0 | |
| }, | |
| "weight_map": { | |
| "embed_out.bias": "pytorch_model-00010-of-00010.bin", | |
| "embed_out.weight": "pytorch_model-00010-of-00010.bin", | |
| "geov.embed_in.weight": "pytorch_model-00001-of-00010.bin", | |
| "geov.final_layer_norm.bias": "pytorch_model-00010-of-00010.bin", | |
| "geov.final_layer_norm.weight": "pytorch_model-00010-of-00010.bin", | |
| "geov.layers.0.attention.causal_mask": "pytorch_model-00001-of-00010.bin", | |
| "geov.layers.0.attention.dense.weight": "pytorch_model-00001-of-00010.bin", | |
| "geov.layers.0.attention.qkv.bias": "pytorch_model-00001-of-00010.bin", | |
| "geov.layers.0.attention.qkv.weight": "pytorch_model-00001-of-00010.bin", | |
| "geov.layers.0.attention.rotary_emb.inv_freq": "pytorch_model-00001-of-00010.bin", | |
| "geov.layers.0.input_layernorm.bias": "pytorch_model-00001-of-00010.bin", | |
| "geov.layers.0.input_layernorm.weight": "pytorch_model-00001-of-00010.bin", | |
| "geov.layers.0.mlp.dense_2h_to_h.weight": "pytorch_model-00001-of-00010.bin", | |
| "geov.layers.0.mlp.dense_h_to_4h.bias": "pytorch_model-00001-of-00010.bin", | |
| "geov.layers.0.mlp.dense_h_to_4h.weight": "pytorch_model-00001-of-00010.bin", | |
| "geov.layers.0.post_attention_layernorm.bias": "pytorch_model-00001-of-00010.bin", | |
| "geov.layers.0.post_attention_layernorm.weight": "pytorch_model-00001-of-00010.bin", | |
| "geov.layers.1.attention.causal_mask": "pytorch_model-00001-of-00010.bin", | |
| "geov.layers.1.attention.dense.weight": "pytorch_model-00001-of-00010.bin", | |
| "geov.layers.1.attention.qkv.bias": "pytorch_model-00001-of-00010.bin", | |
| "geov.layers.1.attention.qkv.weight": "pytorch_model-00001-of-00010.bin", | |
| "geov.layers.1.attention.rotary_emb.inv_freq": "pytorch_model-00001-of-00010.bin", | |
| "geov.layers.1.input_layernorm.bias": "pytorch_model-00001-of-00010.bin", | |
| "geov.layers.1.input_layernorm.weight": "pytorch_model-00001-of-00010.bin", | |
| "geov.layers.1.mlp.dense_2h_to_h.weight": "pytorch_model-00001-of-00010.bin", | |
| "geov.layers.1.mlp.dense_h_to_4h.bias": "pytorch_model-00001-of-00010.bin", | |
| "geov.layers.1.mlp.dense_h_to_4h.weight": "pytorch_model-00001-of-00010.bin", | |
| "geov.layers.1.post_attention_layernorm.bias": "pytorch_model-00001-of-00010.bin", | |
| "geov.layers.1.post_attention_layernorm.weight": "pytorch_model-00001-of-00010.bin", | |
| "geov.layers.10.attention.causal_mask": "pytorch_model-00004-of-00010.bin", | |
| "geov.layers.10.attention.dense.weight": "pytorch_model-00004-of-00010.bin", | |
| "geov.layers.10.attention.qkv.bias": "pytorch_model-00004-of-00010.bin", | |
| "geov.layers.10.attention.qkv.weight": "pytorch_model-00004-of-00010.bin", | |
| "geov.layers.10.attention.rotary_emb.inv_freq": "pytorch_model-00004-of-00010.bin", | |
| "geov.layers.10.input_layernorm.bias": "pytorch_model-00004-of-00010.bin", | |
| "geov.layers.10.input_layernorm.weight": "pytorch_model-00004-of-00010.bin", | |
| "geov.layers.10.mlp.dense_2h_to_h.weight": "pytorch_model-00004-of-00010.bin", | |
| "geov.layers.10.mlp.dense_h_to_4h.bias": "pytorch_model-00004-of-00010.bin", | |
| "geov.layers.10.mlp.dense_h_to_4h.weight": "pytorch_model-00004-of-00010.bin", | |
| "geov.layers.10.post_attention_layernorm.bias": "pytorch_model-00004-of-00010.bin", | |
| "geov.layers.10.post_attention_layernorm.weight": "pytorch_model-00004-of-00010.bin", | |
| "geov.layers.11.attention.causal_mask": "pytorch_model-00004-of-00010.bin", | |
| "geov.layers.11.attention.dense.weight": "pytorch_model-00004-of-00010.bin", | |
| "geov.layers.11.attention.qkv.bias": "pytorch_model-00004-of-00010.bin", | |
| "geov.layers.11.attention.qkv.weight": "pytorch_model-00004-of-00010.bin", | |
| "geov.layers.11.attention.rotary_emb.inv_freq": "pytorch_model-00004-of-00010.bin", | |
| "geov.layers.11.input_layernorm.bias": "pytorch_model-00004-of-00010.bin", | |
| "geov.layers.11.input_layernorm.weight": "pytorch_model-00004-of-00010.bin", | |
| "geov.layers.11.mlp.dense_2h_to_h.weight": "pytorch_model-00004-of-00010.bin", | |
| "geov.layers.11.mlp.dense_h_to_4h.bias": "pytorch_model-00004-of-00010.bin", | |
| "geov.layers.11.mlp.dense_h_to_4h.weight": "pytorch_model-00004-of-00010.bin", | |
| "geov.layers.11.post_attention_layernorm.bias": "pytorch_model-00004-of-00010.bin", | |
| "geov.layers.11.post_attention_layernorm.weight": "pytorch_model-00004-of-00010.bin", | |
| "geov.layers.12.attention.causal_mask": "pytorch_model-00004-of-00010.bin", | |
| "geov.layers.12.attention.dense.weight": "pytorch_model-00004-of-00010.bin", | |
| "geov.layers.12.attention.qkv.bias": "pytorch_model-00004-of-00010.bin", | |
| "geov.layers.12.attention.qkv.weight": "pytorch_model-00004-of-00010.bin", | |
| "geov.layers.12.attention.rotary_emb.inv_freq": "pytorch_model-00004-of-00010.bin", | |
| "geov.layers.12.input_layernorm.bias": "pytorch_model-00004-of-00010.bin", | |
| "geov.layers.12.input_layernorm.weight": "pytorch_model-00004-of-00010.bin", | |
| "geov.layers.12.mlp.dense_2h_to_h.weight": "pytorch_model-00004-of-00010.bin", | |
| "geov.layers.12.mlp.dense_h_to_4h.bias": "pytorch_model-00004-of-00010.bin", | |
| "geov.layers.12.mlp.dense_h_to_4h.weight": "pytorch_model-00004-of-00010.bin", | |
| "geov.layers.12.post_attention_layernorm.bias": "pytorch_model-00004-of-00010.bin", | |
| "geov.layers.12.post_attention_layernorm.weight": "pytorch_model-00004-of-00010.bin", | |
| "geov.layers.13.attention.causal_mask": "pytorch_model-00004-of-00010.bin", | |
| "geov.layers.13.attention.dense.weight": "pytorch_model-00004-of-00010.bin", | |
| "geov.layers.13.attention.qkv.bias": "pytorch_model-00004-of-00010.bin", | |
| "geov.layers.13.attention.qkv.weight": "pytorch_model-00004-of-00010.bin", | |
| "geov.layers.13.attention.rotary_emb.inv_freq": "pytorch_model-00004-of-00010.bin", | |
| "geov.layers.13.input_layernorm.bias": "pytorch_model-00004-of-00010.bin", | |
| "geov.layers.13.input_layernorm.weight": "pytorch_model-00004-of-00010.bin", | |
| "geov.layers.13.mlp.dense_2h_to_h.weight": "pytorch_model-00005-of-00010.bin", | |
| "geov.layers.13.mlp.dense_h_to_4h.bias": "pytorch_model-00005-of-00010.bin", | |
| "geov.layers.13.mlp.dense_h_to_4h.weight": "pytorch_model-00005-of-00010.bin", | |
| "geov.layers.13.post_attention_layernorm.bias": "pytorch_model-00004-of-00010.bin", | |
| "geov.layers.13.post_attention_layernorm.weight": "pytorch_model-00004-of-00010.bin", | |
| "geov.layers.14.attention.causal_mask": "pytorch_model-00005-of-00010.bin", | |
| "geov.layers.14.attention.dense.weight": "pytorch_model-00005-of-00010.bin", | |
| "geov.layers.14.attention.qkv.bias": "pytorch_model-00005-of-00010.bin", | |
| "geov.layers.14.attention.qkv.weight": "pytorch_model-00005-of-00010.bin", | |
| "geov.layers.14.attention.rotary_emb.inv_freq": "pytorch_model-00005-of-00010.bin", | |
| "geov.layers.14.input_layernorm.bias": "pytorch_model-00005-of-00010.bin", | |
| "geov.layers.14.input_layernorm.weight": "pytorch_model-00005-of-00010.bin", | |
| "geov.layers.14.mlp.dense_2h_to_h.weight": "pytorch_model-00005-of-00010.bin", | |
| "geov.layers.14.mlp.dense_h_to_4h.bias": "pytorch_model-00005-of-00010.bin", | |
| "geov.layers.14.mlp.dense_h_to_4h.weight": "pytorch_model-00005-of-00010.bin", | |
| "geov.layers.14.post_attention_layernorm.bias": "pytorch_model-00005-of-00010.bin", | |
| "geov.layers.14.post_attention_layernorm.weight": "pytorch_model-00005-of-00010.bin", | |
| "geov.layers.15.attention.causal_mask": "pytorch_model-00005-of-00010.bin", | |
| "geov.layers.15.attention.dense.weight": "pytorch_model-00005-of-00010.bin", | |
| "geov.layers.15.attention.qkv.bias": "pytorch_model-00005-of-00010.bin", | |
| "geov.layers.15.attention.qkv.weight": "pytorch_model-00005-of-00010.bin", | |
| "geov.layers.15.attention.rotary_emb.inv_freq": "pytorch_model-00005-of-00010.bin", | |
| "geov.layers.15.input_layernorm.bias": "pytorch_model-00005-of-00010.bin", | |
| "geov.layers.15.input_layernorm.weight": "pytorch_model-00005-of-00010.bin", | |
| "geov.layers.15.mlp.dense_2h_to_h.weight": "pytorch_model-00005-of-00010.bin", | |
| "geov.layers.15.mlp.dense_h_to_4h.bias": "pytorch_model-00005-of-00010.bin", | |
| "geov.layers.15.mlp.dense_h_to_4h.weight": "pytorch_model-00005-of-00010.bin", | |
| "geov.layers.15.post_attention_layernorm.bias": "pytorch_model-00005-of-00010.bin", | |
| "geov.layers.15.post_attention_layernorm.weight": "pytorch_model-00005-of-00010.bin", | |
| "geov.layers.16.attention.causal_mask": "pytorch_model-00005-of-00010.bin", | |
| "geov.layers.16.attention.dense.weight": "pytorch_model-00005-of-00010.bin", | |
| "geov.layers.16.attention.qkv.bias": "pytorch_model-00005-of-00010.bin", | |
| "geov.layers.16.attention.qkv.weight": "pytorch_model-00005-of-00010.bin", | |
| "geov.layers.16.attention.rotary_emb.inv_freq": "pytorch_model-00005-of-00010.bin", | |
| "geov.layers.16.input_layernorm.bias": "pytorch_model-00005-of-00010.bin", | |
| "geov.layers.16.input_layernorm.weight": "pytorch_model-00005-of-00010.bin", | |
| "geov.layers.16.mlp.dense_2h_to_h.weight": "pytorch_model-00005-of-00010.bin", | |
| "geov.layers.16.mlp.dense_h_to_4h.bias": "pytorch_model-00005-of-00010.bin", | |
| "geov.layers.16.mlp.dense_h_to_4h.weight": "pytorch_model-00005-of-00010.bin", | |
| "geov.layers.16.post_attention_layernorm.bias": "pytorch_model-00005-of-00010.bin", | |
| "geov.layers.16.post_attention_layernorm.weight": "pytorch_model-00005-of-00010.bin", | |
| "geov.layers.17.attention.causal_mask": "pytorch_model-00005-of-00010.bin", | |
| "geov.layers.17.attention.dense.weight": "pytorch_model-00006-of-00010.bin", | |
| "geov.layers.17.attention.qkv.bias": "pytorch_model-00006-of-00010.bin", | |
| "geov.layers.17.attention.qkv.weight": "pytorch_model-00006-of-00010.bin", | |
| "geov.layers.17.attention.rotary_emb.inv_freq": "pytorch_model-00005-of-00010.bin", | |
| "geov.layers.17.input_layernorm.bias": "pytorch_model-00005-of-00010.bin", | |
| "geov.layers.17.input_layernorm.weight": "pytorch_model-00005-of-00010.bin", | |
| "geov.layers.17.mlp.dense_2h_to_h.weight": "pytorch_model-00006-of-00010.bin", | |
| "geov.layers.17.mlp.dense_h_to_4h.bias": "pytorch_model-00006-of-00010.bin", | |
| "geov.layers.17.mlp.dense_h_to_4h.weight": "pytorch_model-00006-of-00010.bin", | |
| "geov.layers.17.post_attention_layernorm.bias": "pytorch_model-00005-of-00010.bin", | |
| "geov.layers.17.post_attention_layernorm.weight": "pytorch_model-00005-of-00010.bin", | |
| "geov.layers.18.attention.causal_mask": "pytorch_model-00006-of-00010.bin", | |
| "geov.layers.18.attention.dense.weight": "pytorch_model-00006-of-00010.bin", | |
| "geov.layers.18.attention.qkv.bias": "pytorch_model-00006-of-00010.bin", | |
| "geov.layers.18.attention.qkv.weight": "pytorch_model-00006-of-00010.bin", | |
| "geov.layers.18.attention.rotary_emb.inv_freq": "pytorch_model-00006-of-00010.bin", | |
| "geov.layers.18.input_layernorm.bias": "pytorch_model-00006-of-00010.bin", | |
| "geov.layers.18.input_layernorm.weight": "pytorch_model-00006-of-00010.bin", | |
| "geov.layers.18.mlp.dense_2h_to_h.weight": "pytorch_model-00006-of-00010.bin", | |
| "geov.layers.18.mlp.dense_h_to_4h.bias": "pytorch_model-00006-of-00010.bin", | |
| "geov.layers.18.mlp.dense_h_to_4h.weight": "pytorch_model-00006-of-00010.bin", | |
| "geov.layers.18.post_attention_layernorm.bias": "pytorch_model-00006-of-00010.bin", | |
| "geov.layers.18.post_attention_layernorm.weight": "pytorch_model-00006-of-00010.bin", | |
| "geov.layers.19.attention.causal_mask": "pytorch_model-00006-of-00010.bin", | |
| "geov.layers.19.attention.dense.weight": "pytorch_model-00006-of-00010.bin", | |
| "geov.layers.19.attention.qkv.bias": "pytorch_model-00006-of-00010.bin", | |
| "geov.layers.19.attention.qkv.weight": "pytorch_model-00006-of-00010.bin", | |
| "geov.layers.19.attention.rotary_emb.inv_freq": "pytorch_model-00006-of-00010.bin", | |
| "geov.layers.19.input_layernorm.bias": "pytorch_model-00006-of-00010.bin", | |
| "geov.layers.19.input_layernorm.weight": "pytorch_model-00006-of-00010.bin", | |
| "geov.layers.19.mlp.dense_2h_to_h.weight": "pytorch_model-00006-of-00010.bin", | |
| "geov.layers.19.mlp.dense_h_to_4h.bias": "pytorch_model-00006-of-00010.bin", | |
| "geov.layers.19.mlp.dense_h_to_4h.weight": "pytorch_model-00006-of-00010.bin", | |
| "geov.layers.19.post_attention_layernorm.bias": "pytorch_model-00006-of-00010.bin", | |
| "geov.layers.19.post_attention_layernorm.weight": "pytorch_model-00006-of-00010.bin", | |
| "geov.layers.2.attention.causal_mask": "pytorch_model-00001-of-00010.bin", | |
| "geov.layers.2.attention.dense.weight": "pytorch_model-00001-of-00010.bin", | |
| "geov.layers.2.attention.qkv.bias": "pytorch_model-00001-of-00010.bin", | |
| "geov.layers.2.attention.qkv.weight": "pytorch_model-00001-of-00010.bin", | |
| "geov.layers.2.attention.rotary_emb.inv_freq": "pytorch_model-00001-of-00010.bin", | |
| "geov.layers.2.input_layernorm.bias": "pytorch_model-00001-of-00010.bin", | |
| "geov.layers.2.input_layernorm.weight": "pytorch_model-00001-of-00010.bin", | |
| "geov.layers.2.mlp.dense_2h_to_h.weight": "pytorch_model-00002-of-00010.bin", | |
| "geov.layers.2.mlp.dense_h_to_4h.bias": "pytorch_model-00002-of-00010.bin", | |
| "geov.layers.2.mlp.dense_h_to_4h.weight": "pytorch_model-00002-of-00010.bin", | |
| "geov.layers.2.post_attention_layernorm.bias": "pytorch_model-00001-of-00010.bin", | |
| "geov.layers.2.post_attention_layernorm.weight": "pytorch_model-00001-of-00010.bin", | |
| "geov.layers.20.attention.causal_mask": "pytorch_model-00006-of-00010.bin", | |
| "geov.layers.20.attention.dense.weight": "pytorch_model-00006-of-00010.bin", | |
| "geov.layers.20.attention.qkv.bias": "pytorch_model-00006-of-00010.bin", | |
| "geov.layers.20.attention.qkv.weight": "pytorch_model-00006-of-00010.bin", | |
| "geov.layers.20.attention.rotary_emb.inv_freq": "pytorch_model-00006-of-00010.bin", | |
| "geov.layers.20.input_layernorm.bias": "pytorch_model-00006-of-00010.bin", | |
| "geov.layers.20.input_layernorm.weight": "pytorch_model-00006-of-00010.bin", | |
| "geov.layers.20.mlp.dense_2h_to_h.weight": "pytorch_model-00007-of-00010.bin", | |
| "geov.layers.20.mlp.dense_h_to_4h.bias": "pytorch_model-00006-of-00010.bin", | |
| "geov.layers.20.mlp.dense_h_to_4h.weight": "pytorch_model-00006-of-00010.bin", | |
| "geov.layers.20.post_attention_layernorm.bias": "pytorch_model-00006-of-00010.bin", | |
| "geov.layers.20.post_attention_layernorm.weight": "pytorch_model-00006-of-00010.bin", | |
| "geov.layers.21.attention.causal_mask": "pytorch_model-00007-of-00010.bin", | |
| "geov.layers.21.attention.dense.weight": "pytorch_model-00007-of-00010.bin", | |
| "geov.layers.21.attention.qkv.bias": "pytorch_model-00007-of-00010.bin", | |
| "geov.layers.21.attention.qkv.weight": "pytorch_model-00007-of-00010.bin", | |
| "geov.layers.21.attention.rotary_emb.inv_freq": "pytorch_model-00007-of-00010.bin", | |
| "geov.layers.21.input_layernorm.bias": "pytorch_model-00007-of-00010.bin", | |
| "geov.layers.21.input_layernorm.weight": "pytorch_model-00007-of-00010.bin", | |
| "geov.layers.21.mlp.dense_2h_to_h.weight": "pytorch_model-00007-of-00010.bin", | |
| "geov.layers.21.mlp.dense_h_to_4h.bias": "pytorch_model-00007-of-00010.bin", | |
| "geov.layers.21.mlp.dense_h_to_4h.weight": "pytorch_model-00007-of-00010.bin", | |
| "geov.layers.21.post_attention_layernorm.bias": "pytorch_model-00007-of-00010.bin", | |
| "geov.layers.21.post_attention_layernorm.weight": "pytorch_model-00007-of-00010.bin", | |
| "geov.layers.22.attention.causal_mask": "pytorch_model-00007-of-00010.bin", | |
| "geov.layers.22.attention.dense.weight": "pytorch_model-00007-of-00010.bin", | |
| "geov.layers.22.attention.qkv.bias": "pytorch_model-00007-of-00010.bin", | |
| "geov.layers.22.attention.qkv.weight": "pytorch_model-00007-of-00010.bin", | |
| "geov.layers.22.attention.rotary_emb.inv_freq": "pytorch_model-00007-of-00010.bin", | |
| "geov.layers.22.input_layernorm.bias": "pytorch_model-00007-of-00010.bin", | |
| "geov.layers.22.input_layernorm.weight": "pytorch_model-00007-of-00010.bin", | |
| "geov.layers.22.mlp.dense_2h_to_h.weight": "pytorch_model-00007-of-00010.bin", | |
| "geov.layers.22.mlp.dense_h_to_4h.bias": "pytorch_model-00007-of-00010.bin", | |
| "geov.layers.22.mlp.dense_h_to_4h.weight": "pytorch_model-00007-of-00010.bin", | |
| "geov.layers.22.post_attention_layernorm.bias": "pytorch_model-00007-of-00010.bin", | |
| "geov.layers.22.post_attention_layernorm.weight": "pytorch_model-00007-of-00010.bin", | |
| "geov.layers.23.attention.causal_mask": "pytorch_model-00007-of-00010.bin", | |
| "geov.layers.23.attention.dense.weight": "pytorch_model-00007-of-00010.bin", | |
| "geov.layers.23.attention.qkv.bias": "pytorch_model-00007-of-00010.bin", | |
| "geov.layers.23.attention.qkv.weight": "pytorch_model-00007-of-00010.bin", | |
| "geov.layers.23.attention.rotary_emb.inv_freq": "pytorch_model-00007-of-00010.bin", | |
| "geov.layers.23.input_layernorm.bias": "pytorch_model-00007-of-00010.bin", | |
| "geov.layers.23.input_layernorm.weight": "pytorch_model-00007-of-00010.bin", | |
| "geov.layers.23.mlp.dense_2h_to_h.weight": "pytorch_model-00007-of-00010.bin", | |
| "geov.layers.23.mlp.dense_h_to_4h.bias": "pytorch_model-00007-of-00010.bin", | |
| "geov.layers.23.mlp.dense_h_to_4h.weight": "pytorch_model-00007-of-00010.bin", | |
| "geov.layers.23.post_attention_layernorm.bias": "pytorch_model-00007-of-00010.bin", | |
| "geov.layers.23.post_attention_layernorm.weight": "pytorch_model-00007-of-00010.bin", | |
| "geov.layers.24.attention.causal_mask": "pytorch_model-00007-of-00010.bin", | |
| "geov.layers.24.attention.dense.weight": "pytorch_model-00007-of-00010.bin", | |
| "geov.layers.24.attention.qkv.bias": "pytorch_model-00007-of-00010.bin", | |
| "geov.layers.24.attention.qkv.weight": "pytorch_model-00007-of-00010.bin", | |
| "geov.layers.24.attention.rotary_emb.inv_freq": "pytorch_model-00007-of-00010.bin", | |
| "geov.layers.24.input_layernorm.bias": "pytorch_model-00007-of-00010.bin", | |
| "geov.layers.24.input_layernorm.weight": "pytorch_model-00007-of-00010.bin", | |
| "geov.layers.24.mlp.dense_2h_to_h.weight": "pytorch_model-00008-of-00010.bin", | |
| "geov.layers.24.mlp.dense_h_to_4h.bias": "pytorch_model-00008-of-00010.bin", | |
| "geov.layers.24.mlp.dense_h_to_4h.weight": "pytorch_model-00008-of-00010.bin", | |
| "geov.layers.24.post_attention_layernorm.bias": "pytorch_model-00007-of-00010.bin", | |
| "geov.layers.24.post_attention_layernorm.weight": "pytorch_model-00007-of-00010.bin", | |
| "geov.layers.25.attention.causal_mask": "pytorch_model-00008-of-00010.bin", | |
| "geov.layers.25.attention.dense.weight": "pytorch_model-00008-of-00010.bin", | |
| "geov.layers.25.attention.qkv.bias": "pytorch_model-00008-of-00010.bin", | |
| "geov.layers.25.attention.qkv.weight": "pytorch_model-00008-of-00010.bin", | |
| "geov.layers.25.attention.rotary_emb.inv_freq": "pytorch_model-00008-of-00010.bin", | |
| "geov.layers.25.input_layernorm.bias": "pytorch_model-00008-of-00010.bin", | |
| "geov.layers.25.input_layernorm.weight": "pytorch_model-00008-of-00010.bin", | |
| "geov.layers.25.mlp.dense_2h_to_h.weight": "pytorch_model-00008-of-00010.bin", | |
| "geov.layers.25.mlp.dense_h_to_4h.bias": "pytorch_model-00008-of-00010.bin", | |
| "geov.layers.25.mlp.dense_h_to_4h.weight": "pytorch_model-00008-of-00010.bin", | |
| "geov.layers.25.post_attention_layernorm.bias": "pytorch_model-00008-of-00010.bin", | |
| "geov.layers.25.post_attention_layernorm.weight": "pytorch_model-00008-of-00010.bin", | |
| "geov.layers.26.attention.causal_mask": "pytorch_model-00008-of-00010.bin", | |
| "geov.layers.26.attention.dense.weight": "pytorch_model-00008-of-00010.bin", | |
| "geov.layers.26.attention.qkv.bias": "pytorch_model-00008-of-00010.bin", | |
| "geov.layers.26.attention.qkv.weight": "pytorch_model-00008-of-00010.bin", | |
| "geov.layers.26.attention.rotary_emb.inv_freq": "pytorch_model-00008-of-00010.bin", | |
| "geov.layers.26.input_layernorm.bias": "pytorch_model-00008-of-00010.bin", | |
| "geov.layers.26.input_layernorm.weight": "pytorch_model-00008-of-00010.bin", | |
| "geov.layers.26.mlp.dense_2h_to_h.weight": "pytorch_model-00008-of-00010.bin", | |
| "geov.layers.26.mlp.dense_h_to_4h.bias": "pytorch_model-00008-of-00010.bin", | |
| "geov.layers.26.mlp.dense_h_to_4h.weight": "pytorch_model-00008-of-00010.bin", | |
| "geov.layers.26.post_attention_layernorm.bias": "pytorch_model-00008-of-00010.bin", | |
| "geov.layers.26.post_attention_layernorm.weight": "pytorch_model-00008-of-00010.bin", | |
| "geov.layers.27.attention.causal_mask": "pytorch_model-00008-of-00010.bin", | |
| "geov.layers.27.attention.dense.weight": "pytorch_model-00008-of-00010.bin", | |
| "geov.layers.27.attention.qkv.bias": "pytorch_model-00008-of-00010.bin", | |
| "geov.layers.27.attention.qkv.weight": "pytorch_model-00008-of-00010.bin", | |
| "geov.layers.27.attention.rotary_emb.inv_freq": "pytorch_model-00008-of-00010.bin", | |
| "geov.layers.27.input_layernorm.bias": "pytorch_model-00008-of-00010.bin", | |
| "geov.layers.27.input_layernorm.weight": "pytorch_model-00008-of-00010.bin", | |
| "geov.layers.27.mlp.dense_2h_to_h.weight": "pytorch_model-00008-of-00010.bin", | |
| "geov.layers.27.mlp.dense_h_to_4h.bias": "pytorch_model-00008-of-00010.bin", | |
| "geov.layers.27.mlp.dense_h_to_4h.weight": "pytorch_model-00008-of-00010.bin", | |
| "geov.layers.27.post_attention_layernorm.bias": "pytorch_model-00008-of-00010.bin", | |
| "geov.layers.27.post_attention_layernorm.weight": "pytorch_model-00008-of-00010.bin", | |
| "geov.layers.28.attention.causal_mask": "pytorch_model-00008-of-00010.bin", | |
| "geov.layers.28.attention.dense.weight": "pytorch_model-00009-of-00010.bin", | |
| "geov.layers.28.attention.qkv.bias": "pytorch_model-00009-of-00010.bin", | |
| "geov.layers.28.attention.qkv.weight": "pytorch_model-00009-of-00010.bin", | |
| "geov.layers.28.attention.rotary_emb.inv_freq": "pytorch_model-00008-of-00010.bin", | |
| "geov.layers.28.input_layernorm.bias": "pytorch_model-00008-of-00010.bin", | |
| "geov.layers.28.input_layernorm.weight": "pytorch_model-00008-of-00010.bin", | |
| "geov.layers.28.mlp.dense_2h_to_h.weight": "pytorch_model-00009-of-00010.bin", | |
| "geov.layers.28.mlp.dense_h_to_4h.bias": "pytorch_model-00009-of-00010.bin", | |
| "geov.layers.28.mlp.dense_h_to_4h.weight": "pytorch_model-00009-of-00010.bin", | |
| "geov.layers.28.post_attention_layernorm.bias": "pytorch_model-00008-of-00010.bin", | |
| "geov.layers.28.post_attention_layernorm.weight": "pytorch_model-00008-of-00010.bin", | |
| "geov.layers.29.attention.causal_mask": "pytorch_model-00009-of-00010.bin", | |
| "geov.layers.29.attention.dense.weight": "pytorch_model-00009-of-00010.bin", | |
| "geov.layers.29.attention.qkv.bias": "pytorch_model-00009-of-00010.bin", | |
| "geov.layers.29.attention.qkv.weight": "pytorch_model-00009-of-00010.bin", | |
| "geov.layers.29.attention.rotary_emb.inv_freq": "pytorch_model-00009-of-00010.bin", | |
| "geov.layers.29.input_layernorm.bias": "pytorch_model-00009-of-00010.bin", | |
| "geov.layers.29.input_layernorm.weight": "pytorch_model-00009-of-00010.bin", | |
| "geov.layers.29.mlp.dense_2h_to_h.weight": "pytorch_model-00009-of-00010.bin", | |
| "geov.layers.29.mlp.dense_h_to_4h.bias": "pytorch_model-00009-of-00010.bin", | |
| "geov.layers.29.mlp.dense_h_to_4h.weight": "pytorch_model-00009-of-00010.bin", | |
| "geov.layers.29.post_attention_layernorm.bias": "pytorch_model-00009-of-00010.bin", | |
| "geov.layers.29.post_attention_layernorm.weight": "pytorch_model-00009-of-00010.bin", | |
| "geov.layers.3.attention.causal_mask": "pytorch_model-00002-of-00010.bin", | |
| "geov.layers.3.attention.dense.weight": "pytorch_model-00002-of-00010.bin", | |
| "geov.layers.3.attention.qkv.bias": "pytorch_model-00002-of-00010.bin", | |
| "geov.layers.3.attention.qkv.weight": "pytorch_model-00002-of-00010.bin", | |
| "geov.layers.3.attention.rotary_emb.inv_freq": "pytorch_model-00002-of-00010.bin", | |
| "geov.layers.3.input_layernorm.bias": "pytorch_model-00002-of-00010.bin", | |
| "geov.layers.3.input_layernorm.weight": "pytorch_model-00002-of-00010.bin", | |
| "geov.layers.3.mlp.dense_2h_to_h.weight": "pytorch_model-00002-of-00010.bin", | |
| "geov.layers.3.mlp.dense_h_to_4h.bias": "pytorch_model-00002-of-00010.bin", | |
| "geov.layers.3.mlp.dense_h_to_4h.weight": "pytorch_model-00002-of-00010.bin", | |
| "geov.layers.3.post_attention_layernorm.bias": "pytorch_model-00002-of-00010.bin", | |
| "geov.layers.3.post_attention_layernorm.weight": "pytorch_model-00002-of-00010.bin", | |
| "geov.layers.30.attention.causal_mask": "pytorch_model-00009-of-00010.bin", | |
| "geov.layers.30.attention.dense.weight": "pytorch_model-00009-of-00010.bin", | |
| "geov.layers.30.attention.qkv.bias": "pytorch_model-00009-of-00010.bin", | |
| "geov.layers.30.attention.qkv.weight": "pytorch_model-00009-of-00010.bin", | |
| "geov.layers.30.attention.rotary_emb.inv_freq": "pytorch_model-00009-of-00010.bin", | |
| "geov.layers.30.input_layernorm.bias": "pytorch_model-00009-of-00010.bin", | |
| "geov.layers.30.input_layernorm.weight": "pytorch_model-00009-of-00010.bin", | |
| "geov.layers.30.mlp.dense_2h_to_h.weight": "pytorch_model-00009-of-00010.bin", | |
| "geov.layers.30.mlp.dense_h_to_4h.bias": "pytorch_model-00009-of-00010.bin", | |
| "geov.layers.30.mlp.dense_h_to_4h.weight": "pytorch_model-00009-of-00010.bin", | |
| "geov.layers.30.post_attention_layernorm.bias": "pytorch_model-00009-of-00010.bin", | |
| "geov.layers.30.post_attention_layernorm.weight": "pytorch_model-00009-of-00010.bin", | |
| "geov.layers.31.attention.causal_mask": "pytorch_model-00009-of-00010.bin", | |
| "geov.layers.31.attention.dense.weight": "pytorch_model-00009-of-00010.bin", | |
| "geov.layers.31.attention.qkv.bias": "pytorch_model-00009-of-00010.bin", | |
| "geov.layers.31.attention.qkv.weight": "pytorch_model-00009-of-00010.bin", | |
| "geov.layers.31.attention.rotary_emb.inv_freq": "pytorch_model-00009-of-00010.bin", | |
| "geov.layers.31.input_layernorm.bias": "pytorch_model-00009-of-00010.bin", | |
| "geov.layers.31.input_layernorm.weight": "pytorch_model-00009-of-00010.bin", | |
| "geov.layers.31.mlp.dense_2h_to_h.weight": "pytorch_model-00010-of-00010.bin", | |
| "geov.layers.31.mlp.dense_h_to_4h.bias": "pytorch_model-00009-of-00010.bin", | |
| "geov.layers.31.mlp.dense_h_to_4h.weight": "pytorch_model-00009-of-00010.bin", | |
| "geov.layers.31.post_attention_layernorm.bias": "pytorch_model-00009-of-00010.bin", | |
| "geov.layers.31.post_attention_layernorm.weight": "pytorch_model-00009-of-00010.bin", | |
| "geov.layers.4.attention.causal_mask": "pytorch_model-00002-of-00010.bin", | |
| "geov.layers.4.attention.dense.weight": "pytorch_model-00002-of-00010.bin", | |
| "geov.layers.4.attention.qkv.bias": "pytorch_model-00002-of-00010.bin", | |
| "geov.layers.4.attention.qkv.weight": "pytorch_model-00002-of-00010.bin", | |
| "geov.layers.4.attention.rotary_emb.inv_freq": "pytorch_model-00002-of-00010.bin", | |
| "geov.layers.4.input_layernorm.bias": "pytorch_model-00002-of-00010.bin", | |
| "geov.layers.4.input_layernorm.weight": "pytorch_model-00002-of-00010.bin", | |
| "geov.layers.4.mlp.dense_2h_to_h.weight": "pytorch_model-00002-of-00010.bin", | |
| "geov.layers.4.mlp.dense_h_to_4h.bias": "pytorch_model-00002-of-00010.bin", | |
| "geov.layers.4.mlp.dense_h_to_4h.weight": "pytorch_model-00002-of-00010.bin", | |
| "geov.layers.4.post_attention_layernorm.bias": "pytorch_model-00002-of-00010.bin", | |
| "geov.layers.4.post_attention_layernorm.weight": "pytorch_model-00002-of-00010.bin", | |
| "geov.layers.5.attention.causal_mask": "pytorch_model-00002-of-00010.bin", | |
| "geov.layers.5.attention.dense.weight": "pytorch_model-00002-of-00010.bin", | |
| "geov.layers.5.attention.qkv.bias": "pytorch_model-00002-of-00010.bin", | |
| "geov.layers.5.attention.qkv.weight": "pytorch_model-00002-of-00010.bin", | |
| "geov.layers.5.attention.rotary_emb.inv_freq": "pytorch_model-00002-of-00010.bin", | |
| "geov.layers.5.input_layernorm.bias": "pytorch_model-00002-of-00010.bin", | |
| "geov.layers.5.input_layernorm.weight": "pytorch_model-00002-of-00010.bin", | |
| "geov.layers.5.mlp.dense_2h_to_h.weight": "pytorch_model-00002-of-00010.bin", | |
| "geov.layers.5.mlp.dense_h_to_4h.bias": "pytorch_model-00002-of-00010.bin", | |
| "geov.layers.5.mlp.dense_h_to_4h.weight": "pytorch_model-00002-of-00010.bin", | |
| "geov.layers.5.post_attention_layernorm.bias": "pytorch_model-00002-of-00010.bin", | |
| "geov.layers.5.post_attention_layernorm.weight": "pytorch_model-00002-of-00010.bin", | |
| "geov.layers.6.attention.causal_mask": "pytorch_model-00002-of-00010.bin", | |
| "geov.layers.6.attention.dense.weight": "pytorch_model-00003-of-00010.bin", | |
| "geov.layers.6.attention.qkv.bias": "pytorch_model-00003-of-00010.bin", | |
| "geov.layers.6.attention.qkv.weight": "pytorch_model-00003-of-00010.bin", | |
| "geov.layers.6.attention.rotary_emb.inv_freq": "pytorch_model-00002-of-00010.bin", | |
| "geov.layers.6.input_layernorm.bias": "pytorch_model-00002-of-00010.bin", | |
| "geov.layers.6.input_layernorm.weight": "pytorch_model-00002-of-00010.bin", | |
| "geov.layers.6.mlp.dense_2h_to_h.weight": "pytorch_model-00003-of-00010.bin", | |
| "geov.layers.6.mlp.dense_h_to_4h.bias": "pytorch_model-00003-of-00010.bin", | |
| "geov.layers.6.mlp.dense_h_to_4h.weight": "pytorch_model-00003-of-00010.bin", | |
| "geov.layers.6.post_attention_layernorm.bias": "pytorch_model-00002-of-00010.bin", | |
| "geov.layers.6.post_attention_layernorm.weight": "pytorch_model-00002-of-00010.bin", | |
| "geov.layers.7.attention.causal_mask": "pytorch_model-00003-of-00010.bin", | |
| "geov.layers.7.attention.dense.weight": "pytorch_model-00003-of-00010.bin", | |
| "geov.layers.7.attention.qkv.bias": "pytorch_model-00003-of-00010.bin", | |
| "geov.layers.7.attention.qkv.weight": "pytorch_model-00003-of-00010.bin", | |
| "geov.layers.7.attention.rotary_emb.inv_freq": "pytorch_model-00003-of-00010.bin", | |
| "geov.layers.7.input_layernorm.bias": "pytorch_model-00003-of-00010.bin", | |
| "geov.layers.7.input_layernorm.weight": "pytorch_model-00003-of-00010.bin", | |
| "geov.layers.7.mlp.dense_2h_to_h.weight": "pytorch_model-00003-of-00010.bin", | |
| "geov.layers.7.mlp.dense_h_to_4h.bias": "pytorch_model-00003-of-00010.bin", | |
| "geov.layers.7.mlp.dense_h_to_4h.weight": "pytorch_model-00003-of-00010.bin", | |
| "geov.layers.7.post_attention_layernorm.bias": "pytorch_model-00003-of-00010.bin", | |
| "geov.layers.7.post_attention_layernorm.weight": "pytorch_model-00003-of-00010.bin", | |
| "geov.layers.8.attention.causal_mask": "pytorch_model-00003-of-00010.bin", | |
| "geov.layers.8.attention.dense.weight": "pytorch_model-00003-of-00010.bin", | |
| "geov.layers.8.attention.qkv.bias": "pytorch_model-00003-of-00010.bin", | |
| "geov.layers.8.attention.qkv.weight": "pytorch_model-00003-of-00010.bin", | |
| "geov.layers.8.attention.rotary_emb.inv_freq": "pytorch_model-00003-of-00010.bin", | |
| "geov.layers.8.input_layernorm.bias": "pytorch_model-00003-of-00010.bin", | |
| "geov.layers.8.input_layernorm.weight": "pytorch_model-00003-of-00010.bin", | |
| "geov.layers.8.mlp.dense_2h_to_h.weight": "pytorch_model-00003-of-00010.bin", | |
| "geov.layers.8.mlp.dense_h_to_4h.bias": "pytorch_model-00003-of-00010.bin", | |
| "geov.layers.8.mlp.dense_h_to_4h.weight": "pytorch_model-00003-of-00010.bin", | |
| "geov.layers.8.post_attention_layernorm.bias": "pytorch_model-00003-of-00010.bin", | |
| "geov.layers.8.post_attention_layernorm.weight": "pytorch_model-00003-of-00010.bin", | |
| "geov.layers.9.attention.causal_mask": "pytorch_model-00003-of-00010.bin", | |
| "geov.layers.9.attention.dense.weight": "pytorch_model-00003-of-00010.bin", | |
| "geov.layers.9.attention.qkv.bias": "pytorch_model-00003-of-00010.bin", | |
| "geov.layers.9.attention.qkv.weight": "pytorch_model-00003-of-00010.bin", | |
| "geov.layers.9.attention.rotary_emb.inv_freq": "pytorch_model-00003-of-00010.bin", | |
| "geov.layers.9.input_layernorm.bias": "pytorch_model-00003-of-00010.bin", | |
| "geov.layers.9.input_layernorm.weight": "pytorch_model-00003-of-00010.bin", | |
| "geov.layers.9.mlp.dense_2h_to_h.weight": "pytorch_model-00004-of-00010.bin", | |
| "geov.layers.9.mlp.dense_h_to_4h.bias": "pytorch_model-00003-of-00010.bin", | |
| "geov.layers.9.mlp.dense_h_to_4h.weight": "pytorch_model-00003-of-00010.bin", | |
| "geov.layers.9.post_attention_layernorm.bias": "pytorch_model-00003-of-00010.bin", | |
| "geov.layers.9.post_attention_layernorm.weight": "pytorch_model-00003-of-00010.bin" | |
| } | |
| } | |