Spaces:
Runtime error
Runtime error
File size: 4,857 Bytes
7900c16 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 |
import argparse
import collections
import torch
parser = argparse.ArgumentParser(formatter_class=argparse.ArgumentDefaultsHelpFormatter)
parser.add_argument("--input_model_path", type=str, default="models/input_model.bin",
help=".")
parser.add_argument("--output_model_path", type=str, default="models/output_model.bin",
help=".")
args = parser.parse_args()
input_model = torch.load(args.input_model_path, map_location="cpu")
output_model = collections.OrderedDict()
output_model["embedding.word.embedding.weight"] = \
input_model["albert.embeddings.word_embeddings.weight"]
output_model["embedding.pos.embedding.weight"] = \
input_model["albert.embeddings.position_embeddings.weight"]
output_model["embedding.seg.embedding.weight"] = \
torch.cat((torch.Tensor([[0]*input_model["albert.embeddings.token_type_embeddings.weight"].size()[1]]),
input_model["albert.embeddings.token_type_embeddings.weight"]), dim=0)
output_model["embedding.layer_norm.gamma"] = \
input_model["albert.embeddings.LayerNorm.weight"]
output_model["embedding.layer_norm.beta"] = \
input_model["albert.embeddings.LayerNorm.bias"]
output_model["encoder.linear.weight"] = \
input_model["albert.encoder.embedding_hidden_mapping_in.weight"]
output_model["encoder.linear.bias"] = \
input_model["albert.encoder.embedding_hidden_mapping_in.bias"]
output_model["encoder.transformer.layer_norm_2.gamma"] = \
input_model["albert.encoder.albert_layer_groups.0.albert_layers.0.full_layer_layer_norm.weight"]
output_model["encoder.transformer.layer_norm_2.beta"] = \
input_model["albert.encoder.albert_layer_groups.0.albert_layers.0.full_layer_layer_norm.bias"]
output_model["encoder.transformer.self_attn.linear_layers.0.weight"] = \
input_model["albert.encoder.albert_layer_groups.0.albert_layers.0.attention.query.weight"]
output_model["encoder.transformer.self_attn.linear_layers.0.bias"] = \
input_model["albert.encoder.albert_layer_groups.0.albert_layers.0.attention.query.bias"]
output_model["encoder.transformer.self_attn.linear_layers.1.weight"] = \
input_model["albert.encoder.albert_layer_groups.0.albert_layers.0.attention.key.weight"]
output_model["encoder.transformer.self_attn.linear_layers.1.bias"] = \
input_model["albert.encoder.albert_layer_groups.0.albert_layers.0.attention.key.bias"]
output_model["encoder.transformer.self_attn.linear_layers.2.weight"] = \
input_model["albert.encoder.albert_layer_groups.0.albert_layers.0.attention.value.weight"]
output_model["encoder.transformer.self_attn.linear_layers.2.bias"] = \
input_model["albert.encoder.albert_layer_groups.0.albert_layers.0.attention.value.bias"]
output_model["encoder.transformer.self_attn.final_linear.weight"] = \
input_model["albert.encoder.albert_layer_groups.0.albert_layers.0.attention.dense.weight"]
output_model["encoder.transformer.self_attn.final_linear.bias"] = \
input_model["albert.encoder.albert_layer_groups.0.albert_layers.0.attention.dense.bias"]
output_model["encoder.transformer.layer_norm_1.gamma"] = \
input_model["albert.encoder.albert_layer_groups.0.albert_layers.0.attention.LayerNorm.weight"]
output_model["encoder.transformer.layer_norm_1.beta"] = \
input_model["albert.encoder.albert_layer_groups.0.albert_layers.0.attention.LayerNorm.bias"]
output_model["encoder.transformer.feed_forward.linear_1.weight"] = \
input_model["albert.encoder.albert_layer_groups.0.albert_layers.0.ffn.weight"]
output_model["encoder.transformer.feed_forward.linear_1.bias"] = \
input_model["albert.encoder.albert_layer_groups.0.albert_layers.0.ffn.bias"]
output_model["encoder.transformer.feed_forward.linear_2.weight"] = \
input_model["albert.encoder.albert_layer_groups.0.albert_layers.0.ffn_output.weight"]
output_model["encoder.transformer.feed_forward.linear_2.bias"] = \
input_model["albert.encoder.albert_layer_groups.0.albert_layers.0.ffn_output.bias"]
output_model["target.sp.linear_1.weight"] = input_model["albert.pooler.weight"]
output_model["target.sp.linear_1.bias"] = input_model["albert.pooler.bias"]
output_model["target.sp.linear_2.weight"] = input_model["sop_classifier.classifier.weight"]
output_model["target.sp.linear_2.bias"] = input_model["sop_classifier.classifier.bias"]
output_model["target.mlm.linear_1.weight"] = input_model["predictions.dense.weight"]
output_model["target.mlm.linear_1.bias"] = input_model["predictions.dense.bias"]
output_model["target.mlm.linear_2.weight"] = input_model["predictions.decoder.weight"]
output_model["target.mlm.linear_2.bias"] = input_model["predictions.bias"]
output_model["target.mlm.layer_norm.gamma"] = input_model["predictions.LayerNorm.weight"]
output_model["target.mlm.layer_norm.beta"] = input_model["predictions.LayerNorm.bias"]
torch.save(output_model, args.output_model_path)
|