update to transformers

train.py

@@ -1,4 +1,4 @@
-import pytorch_transformers
+import transformers
 import torch
 import os
 import json
@@ -78,7 +78,7 @@ def main():
 
     os.environ["CUDA_VISIBLE_DEVICES"] = args.device  # 此处设置程序使用哪些显卡
 
-    model_config = pytorch_transformers.modeling_gpt2.GPT2Config.from_json_file(args.model_config)
+    model_config = transformers.modeling_gpt2.GPT2Config.from_json_file(args.model_config)
     print('config:\n' + model_config.to_json_string())
 
     n_ctx = model_config.n_ctx
@@ -118,9 +118,9 @@ def main():
         print('files built')
 
     if not args.pretrained_model:
-        model = pytorch_transformers.modeling_gpt2.GPT2LMHeadModel(config=model_config)
+        model = transformers.modeling_gpt2.GPT2LMHeadModel(config=model_config)
     else:
-        model = pytorch_transformers.modeling_gpt2.GPT2LMHeadModel.from_pretrained(args.pretrained_model)
+        model = transformers.modeling_gpt2.GPT2LMHeadModel.from_pretrained(args.pretrained_model)
     model.train()
     model.to(device)
 
@@ -139,8 +139,8 @@ def main():
     total_steps = int(full_len / stride * epochs / batch_size / gradient_accumulation)
     print('total steps = {}'.format(total_steps))
 
-    optimizer = pytorch_transformers.AdamW(model.parameters(), lr=lr, correct_bias=True)
-    scheduler = pytorch_transformers.WarmupLinearSchedule(optimizer, warmup_steps=warmup_steps,
+    optimizer = transformers.AdamW(model.parameters(), lr=lr, correct_bias=True)
+    scheduler = transformers.WarmupLinearSchedule(optimizer, warmup_steps=warmup_steps,
                                                           t_total=total_steps)
     if fp16:
         try: