|
{ |
|
"weight_map": { |
|
"embeddings.position_embedding.weight" : [77,768], |
|
"embeddings.token_embedding.weight" : [49408,768], |
|
"encoder.layers.0.layer_norm1.bias" : [768], |
|
"encoder.layers.0.layer_norm1.weight" : [768], |
|
"encoder.layers.0.layer_norm2.bias" : [768], |
|
"encoder.layers.0.layer_norm2.weight" : [768], |
|
"encoder.layers.0.mlp.fc1.bias" : [3072], |
|
"encoder.layers.0.mlp.fc1.weight" : [3072,768], |
|
"encoder.layers.0.mlp.fc2.bias" : [768], |
|
"encoder.layers.0.mlp.fc2.weight" : [768,3072], |
|
"encoder.layers.0.self_attn.k_proj.bias" : [768], |
|
"encoder.layers.0.self_attn.k_proj.weight" : [768,768], |
|
"encoder.layers.0.self_attn.out_proj.bias" : [768], |
|
"encoder.layers.0.self_attn.out_proj.weight" : [768,768], |
|
"encoder.layers.0.self_attn.q_proj.bias" : [768], |
|
"encoder.layers.0.self_attn.q_proj.weight" : [768,768], |
|
"encoder.layers.0.self_attn.v_proj.bias" : [768], |
|
"encoder.layers.0.self_attn.v_proj.weight" : [768,768], |
|
"encoder.layers.1.layer_norm1.bias" : [768], |
|
"encoder.layers.1.layer_norm1.weight" : [768], |
|
"encoder.layers.1.layer_norm2.bias" : [768], |
|
"encoder.layers.1.layer_norm2.weight" : [768], |
|
"encoder.layers.1.mlp.fc1.bias" : [3072], |
|
"encoder.layers.1.mlp.fc1.weight" : [3072,768], |
|
"encoder.layers.1.mlp.fc2.bias" : [768], |
|
"encoder.layers.1.mlp.fc2.weight" : [768,3072], |
|
"encoder.layers.1.self_attn.k_proj.bias" : [768], |
|
"encoder.layers.1.self_attn.k_proj.weight" : [768,768], |
|
"encoder.layers.1.self_attn.out_proj.bias" : [768], |
|
"encoder.layers.1.self_attn.out_proj.weight" : [768,768], |
|
"encoder.layers.1.self_attn.q_proj.bias" : [768], |
|
"encoder.layers.1.self_attn.q_proj.weight" : [768,768], |
|
"encoder.layers.1.self_attn.v_proj.bias" : [768], |
|
"encoder.layers.1.self_attn.v_proj.weight" : [768,768], |
|
"encoder.layers.10.layer_norm1.bias" : [768], |
|
"encoder.layers.10.layer_norm1.weight" : [768], |
|
"encoder.layers.10.layer_norm2.bias" : [768], |
|
"encoder.layers.10.layer_norm2.weight" : [768], |
|
"encoder.layers.10.mlp.fc1.bias" : [3072], |
|
"encoder.layers.10.mlp.fc1.weight" : [3072,768], |
|
"encoder.layers.10.mlp.fc2.bias" : [768], |
|
"encoder.layers.10.mlp.fc2.weight" : [768,3072], |
|
"encoder.layers.10.self_attn.k_proj.bias" : [768], |
|
"encoder.layers.10.self_attn.k_proj.weight" : [768,768], |
|
"encoder.layers.10.self_attn.out_proj.bias" : [768], |
|
"encoder.layers.10.self_attn.out_proj.weight" : [768,768], |
|
"encoder.layers.10.self_attn.q_proj.bias" : [768], |
|
"encoder.layers.10.self_attn.q_proj.weight" : [768,768], |
|
"encoder.layers.10.self_attn.v_proj.bias" : [768], |
|
"encoder.layers.10.self_attn.v_proj.weight" : [768,768], |
|
"encoder.layers.11.layer_norm1.bias" : [768], |
|
"encoder.layers.11.layer_norm1.weight" : [768], |
|
"encoder.layers.11.layer_norm2.bias" : [768], |
|
"encoder.layers.11.layer_norm2.weight" : [768], |
|
"encoder.layers.11.mlp.fc1.bias" : [3072], |
|
"encoder.layers.11.mlp.fc1.weight" : [3072,768], |
|
"encoder.layers.11.mlp.fc2.bias" : [768], |
|
"encoder.layers.11.mlp.fc2.weight" : [768,3072], |
|
"encoder.layers.11.self_attn.k_proj.bias" : [768], |
|
"encoder.layers.11.self_attn.k_proj.weight" : [768,768], |
|
"encoder.layers.11.self_attn.out_proj.bias" : [768], |
|
"encoder.layers.11.self_attn.out_proj.weight" : [768,768], |
|
"encoder.layers.11.self_attn.q_proj.bias" : [768], |
|
"encoder.layers.11.self_attn.q_proj.weight" : [768,768], |
|
"encoder.layers.11.self_attn.v_proj.bias" : [768], |
|
"encoder.layers.11.self_attn.v_proj.weight" : [768,768], |
|
"encoder.layers.2.layer_norm1.bias" : [768], |
|
"encoder.layers.2.layer_norm1.weight" : [768], |
|
"encoder.layers.2.layer_norm2.bias" : [768], |
|
"encoder.layers.2.layer_norm2.weight" : [768], |
|
"encoder.layers.2.mlp.fc1.bias" : [3072], |
|
"encoder.layers.2.mlp.fc1.weight" : [3072,768], |
|
"encoder.layers.2.mlp.fc2.bias" : [768], |
|
"encoder.layers.2.mlp.fc2.weight" : [768,3072], |
|
"encoder.layers.2.self_attn.k_proj.bias" : [768], |
|
"encoder.layers.2.self_attn.k_proj.weight" : [768,768], |
|
"encoder.layers.2.self_attn.out_proj.bias" : [768], |
|
"encoder.layers.2.self_attn.out_proj.weight" : [768,768], |
|
"encoder.layers.2.self_attn.q_proj.bias" : [768], |
|
"encoder.layers.2.self_attn.q_proj.weight" : [768,768], |
|
"encoder.layers.2.self_attn.v_proj.bias" : [768], |
|
"encoder.layers.2.self_attn.v_proj.weight" : [768,768], |
|
"encoder.layers.3.layer_norm1.bias" : [768], |
|
"encoder.layers.3.layer_norm1.weight" : [768], |
|
"encoder.layers.3.layer_norm2.bias" : [768], |
|
"encoder.layers.3.layer_norm2.weight" : [768], |
|
"encoder.layers.3.mlp.fc1.bias" : [3072], |
|
"encoder.layers.3.mlp.fc1.weight" : [3072,768], |
|
"encoder.layers.3.mlp.fc2.bias" : [768], |
|
"encoder.layers.3.mlp.fc2.weight" : [768,3072], |
|
"encoder.layers.3.self_attn.k_proj.bias" : [768], |
|
"encoder.layers.3.self_attn.k_proj.weight" : [768,768], |
|
"encoder.layers.3.self_attn.out_proj.bias" : [768], |
|
"encoder.layers.3.self_attn.out_proj.weight" : [768,768], |
|
"encoder.layers.3.self_attn.q_proj.bias" : [768], |
|
"encoder.layers.3.self_attn.q_proj.weight" : [768,768], |
|
"encoder.layers.3.self_attn.v_proj.bias" : [768], |
|
"encoder.layers.3.self_attn.v_proj.weight" : [768,768], |
|
"encoder.layers.4.layer_norm1.bias" : [768], |
|
"encoder.layers.4.layer_norm1.weight" : [768], |
|
"encoder.layers.4.layer_norm2.bias" : [768], |
|
"encoder.layers.4.layer_norm2.weight" : [768], |
|
"encoder.layers.4.mlp.fc1.bias" : [3072], |
|
"encoder.layers.4.mlp.fc1.weight" : [3072,768], |
|
"encoder.layers.4.mlp.fc2.bias" : [768], |
|
"encoder.layers.4.mlp.fc2.weight" : [768,3072], |
|
"encoder.layers.4.self_attn.k_proj.bias" : [768], |
|
"encoder.layers.4.self_attn.k_proj.weight" : [768,768], |
|
"encoder.layers.4.self_attn.out_proj.bias" : [768], |
|
"encoder.layers.4.self_attn.out_proj.weight" : [768,768], |
|
"encoder.layers.4.self_attn.q_proj.bias" : [768], |
|
"encoder.layers.4.self_attn.q_proj.weight" : [768,768], |
|
"encoder.layers.4.self_attn.v_proj.bias" : [768], |
|
"encoder.layers.4.self_attn.v_proj.weight" : [768,768], |
|
"encoder.layers.5.layer_norm1.bias" : [768], |
|
"encoder.layers.5.layer_norm1.weight" : [768], |
|
"encoder.layers.5.layer_norm2.bias" : [768], |
|
"encoder.layers.5.layer_norm2.weight" : [768], |
|
"encoder.layers.5.mlp.fc1.bias" : [3072], |
|
"encoder.layers.5.mlp.fc1.weight" : [3072,768], |
|
"encoder.layers.5.mlp.fc2.bias" : [768], |
|
"encoder.layers.5.mlp.fc2.weight" : [768,3072], |
|
"encoder.layers.5.self_attn.k_proj.bias" : [768], |
|
"encoder.layers.5.self_attn.k_proj.weight" : [768,768], |
|
"encoder.layers.5.self_attn.out_proj.bias" : [768], |
|
"encoder.layers.5.self_attn.out_proj.weight" : [768,768], |
|
"encoder.layers.5.self_attn.q_proj.bias" : [768], |
|
"encoder.layers.5.self_attn.q_proj.weight" : [768,768], |
|
"encoder.layers.5.self_attn.v_proj.bias" : [768], |
|
"encoder.layers.5.self_attn.v_proj.weight" : [768,768], |
|
"encoder.layers.6.layer_norm1.bias" : [768], |
|
"encoder.layers.6.layer_norm1.weight" : [768], |
|
"encoder.layers.6.layer_norm2.bias" : [768], |
|
"encoder.layers.6.layer_norm2.weight" : [768], |
|
"encoder.layers.6.mlp.fc1.bias" : [3072], |
|
"encoder.layers.6.mlp.fc1.weight" : [3072,768], |
|
"encoder.layers.6.mlp.fc2.bias" : [768], |
|
"encoder.layers.6.mlp.fc2.weight" : [768,3072], |
|
"encoder.layers.6.self_attn.k_proj.bias" : [768], |
|
"encoder.layers.6.self_attn.k_proj.weight" : [768,768], |
|
"encoder.layers.6.self_attn.out_proj.bias" : [768], |
|
"encoder.layers.6.self_attn.out_proj.weight" : [768,768], |
|
"encoder.layers.6.self_attn.q_proj.bias" : [768], |
|
"encoder.layers.6.self_attn.q_proj.weight" : [768,768], |
|
"encoder.layers.6.self_attn.v_proj.bias" : [768], |
|
"encoder.layers.6.self_attn.v_proj.weight" : [768,768], |
|
"encoder.layers.7.layer_norm1.bias" : [768], |
|
"encoder.layers.7.layer_norm1.weight" : [768], |
|
"encoder.layers.7.layer_norm2.bias" : [768], |
|
"encoder.layers.7.layer_norm2.weight" : [768], |
|
"encoder.layers.7.mlp.fc1.bias" : [3072], |
|
"encoder.layers.7.mlp.fc1.weight" : [3072,768], |
|
"encoder.layers.7.mlp.fc2.bias" : [768], |
|
"encoder.layers.7.mlp.fc2.weight" : [768,3072], |
|
"encoder.layers.7.self_attn.k_proj.bias" : [768], |
|
"encoder.layers.7.self_attn.k_proj.weight" : [768,768], |
|
"encoder.layers.7.self_attn.out_proj.bias" : [768], |
|
"encoder.layers.7.self_attn.out_proj.weight" : [768,768], |
|
"encoder.layers.7.self_attn.q_proj.bias" : [768], |
|
"encoder.layers.7.self_attn.q_proj.weight" : [768,768], |
|
"encoder.layers.7.self_attn.v_proj.bias" : [768], |
|
"encoder.layers.7.self_attn.v_proj.weight" : [768,768], |
|
"encoder.layers.8.layer_norm1.bias" : [768], |
|
"encoder.layers.8.layer_norm1.weight" : [768], |
|
"encoder.layers.8.layer_norm2.bias" : [768], |
|
"encoder.layers.8.layer_norm2.weight" : [768], |
|
"encoder.layers.8.mlp.fc1.bias" : [3072], |
|
"encoder.layers.8.mlp.fc1.weight" : [3072,768], |
|
"encoder.layers.8.mlp.fc2.bias" : [768], |
|
"encoder.layers.8.mlp.fc2.weight" : [768,3072], |
|
"encoder.layers.8.self_attn.k_proj.bias" : [768], |
|
"encoder.layers.8.self_attn.k_proj.weight" : [768,768], |
|
"encoder.layers.8.self_attn.out_proj.bias" : [768], |
|
"encoder.layers.8.self_attn.out_proj.weight" : [768,768], |
|
"encoder.layers.8.self_attn.q_proj.bias" : [768], |
|
"encoder.layers.8.self_attn.q_proj.weight" : [768,768], |
|
"encoder.layers.8.self_attn.v_proj.bias" : [768], |
|
"encoder.layers.8.self_attn.v_proj.weight" : [768,768], |
|
"encoder.layers.9.layer_norm1.bias" : [768], |
|
"encoder.layers.9.layer_norm1.weight" : [768], |
|
"encoder.layers.9.layer_norm2.bias" : [768], |
|
"encoder.layers.9.layer_norm2.weight" : [768], |
|
"encoder.layers.9.mlp.fc1.bias" : [3072], |
|
"encoder.layers.9.mlp.fc1.weight" : [3072,768], |
|
"encoder.layers.9.mlp.fc2.bias" : [768], |
|
"encoder.layers.9.mlp.fc2.weight" : [768,3072], |
|
"encoder.layers.9.self_attn.k_proj.bias" : [768], |
|
"encoder.layers.9.self_attn.k_proj.weight" : [768,768], |
|
"encoder.layers.9.self_attn.out_proj.bias" : [768], |
|
"encoder.layers.9.self_attn.out_proj.weight" : [768,768], |
|
"encoder.layers.9.self_attn.q_proj.bias" : [768], |
|
"encoder.layers.9.self_attn.q_proj.weight" : [768,768], |
|
"encoder.layers.9.self_attn.v_proj.bias" : [768], |
|
"encoder.layers.9.self_attn.v_proj.weight" : [768,768], |
|
"final_layer_norm.bias" : [768], |
|
"final_layer_norm.weight" : [768] |
|
} |
|
} |
|
|