將tensorflow版本的.ckpt模型轉成pytorch的.bin模型

用google-research官方的bert源碼（tensorflow版本）對新的法律語料進行微調，迭代次數為100000次，每隔1000次保存一下模型，得到的結果如下：

將最后三個文件取出，改名為bert_model.ckpt.data-00000-of-00001、bert_model.ckpt.index、bert_model.ckpt.meta

加上之前微調使用過的config.json以及vocab.txt文件，運行如下文件后生成pytorch.bin，之后就可以被pytorch得代碼調用了。

from __future__ import absolute_import
from __future__ import division
from __future__ import print_function

import argparse
import torch

from transformers import BertConfig, BertForPreTraining, load_tf_weights_in_bert

import logging
logging.basicConfig(level=logging.INFO)

def convert_tf_checkpoint_to_pytorch(tf_checkpoint_path, bert_config_file, pytorch_dump_path):
    # Initialise PyTorch model
    config = BertConfig.from_json_file(bert_config_file)
    print("Building PyTorch model from configuration: {}".format(str(config)))
    model = BertForPreTraining(config)

    # Load weights from tf checkpoint
    load_tf_weights_in_bert(model, config, tf_checkpoint_path)

    # Save pytorch-model
    print("Save PyTorch model to {}".format(pytorch_dump_path))
    torch.save(model.state_dict(), pytorch_dump_path)

#
if __name__ == "__main__":
    parser = argparse.ArgumentParser()
    ## Required parameters
    parser.add_argument("--tf_checkpoint_path",
                        default = './chinese_L-12_H-768_A-12_improve1/bert_model.ckpt',
                        type = str,
                        help = "Path to the TensorFlow checkpoint path.")
    parser.add_argument("--bert_config_file",
                        default = './chinese_L-12_H-768_A-12_improve1/config.json',
                        type = str,
                        help = "The config json file corresponding to the pre-trained BERT model. \n"
                            "This specifies the model architecture.")
    parser.add_argument("--pytorch_dump_path",
                        default = './chinese_L-12_H-768_A-12_improve1/pytorch_model.bin',
                        type = str,
                        help = "Path to the output PyTorch model.")
    args = parser.parse_args()
    convert_tf_checkpoint_to_pytorch(args.tf_checkpoint_path,
                                     args.bert_config_file,
                                     args.pytorch_dump_path)

Tip：如果不是BERT模型，是BERT模型的變種，比如MobileBERT，DistilBERT等，數據形式可能不匹配，報錯AttributeError: 'BertForPreTraining' object has no attribute 'bias'

此時需要根據transformers庫里的源碼修改convert_tf_checkpoint_to_pytorch函數，以MobileBERT為例

#參考transformers庫里的transformers/models/mobilebert/convert_mobilebert_original_tf_checkpoint_to_pytorch.py
from transformers import MobileBertConfig, MobileBertForPreTraining, load_tf_weights_in_mobilebert


def convert_tf_checkpoint_to_pytorch(tf_checkpoint_path, mobilebert_config_file, pytorch_dump_path):
    # Initialise PyTorch model
    config = MobileBertConfig.from_json_file(mobilebert_config_file)
    print(f"Building PyTorch model from configuration: {config}")
    model = MobileBertForPreTraining(config)
    # Load weights from tf checkpoint
    model = load_tf_weights_in_mobilebert(model, config, tf_checkpoint_path)
    # Save pytorch-model
    print(f"Save PyTorch model to {pytorch_dump_path}")
    torch.save(model.state_dict(), pytorch_dump_path)