debug loading

app.py

@@ -3,23 +3,22 @@ import transformers
 # import tokenizers
 import torch
 from transformers import pipeline, set_seed
-from transformers import GPT2Model, GPT2Config, GPT2LMHeadModel
-from transformers import BertTokenizerFast
+from transformers import GPT2Model, GPT2Config, GPT2LMHeadModel, AutoModel
+from transformers import BertTokenizerFast, BertTokenizer
 # https://huggingface.co/docs/hub/spaces-sdks-gradio
 
-tokenizer_bert = BertTokenizerFast.from_pretrained('bert-base-chinese',
-    additional_special_tokens=["<s>","<pad>","</s>","<unk>","<mask>"],
-    pad_token='<pad>' ,max_len=512)
+# tokenizer_bert = BertTokenizer.from_pretrained('bert-base-chinese',
+#     additional_special_tokens=["<s>","<pad>","</s>","<unk>","<mask>"],
+#     pad_token='<pad>', max_len=512)
 
-configuration = GPT2Config(vocab_size=25000, n_layer=8)
-model = GPT2LMHeadModel(config=configuration)
-#%%
-# path2pytorch_model = "pytorch_model.bin"
+# configuration = GPT2Config(vocab_size=25000, n_layer=8)
+# model = GPT2LMHeadModel(config=configuration)
 # path2pytorch_model = "/home/binxuwang/Datasets/ancChn_L8_LB_cont_output/checkpoint-100000/pytorch_model.bin"
 # model.load_state_dict(torch.load(path2pytorch_model))
-model.from_pretrained(("binxu/Ziyue-GPT2"))
-
-generator = pipeline('text-generation', model=model, tokenizer=tokenizer_bert)
+# model.from_pretrained("binxu/Ziyue-GPT2")
+#%%
+model = GPT2LMHeadModel.from_pretrained("binxu/Ziyue-GPT2")
+generator = pipeline('text-generation', model=model, tokenizer='bert-base-chinese')
 
 def generate(prompt):
     outputs = generator(prompt, max_length=50, num_return_sequences=5, num_beams=10, repetition_penalty=1.5)