Merge pull request #973 from FeiWang96/bert_config

 Fix examples of loading pretrained models in docstring

pytorch_transformers/modeling_bert.py

@@ -643,9 +643,8 @@ class BertModel(BertPreTrainedModel):
 
     Examples::
 
-        config = BertConfig.from_pretrained('bert-base-uncased')
         tokenizer = BertTokenizer.from_pretrained('bert-base-uncased')
-        model = BertModel(config)
+        model = BertModel.from_pretrained('bert-base-uncased')
         input_ids = torch.tensor(tokenizer.encode("Hello, my dog is cute")).unsqueeze(0)  # Batch size 1
         outputs = model(input_ids)
         last_hidden_states = outputs[0]  # The last hidden-state is the first element of the output tuple
@@ -754,10 +753,8 @@ class BertForPreTraining(BertPreTrainedModel):
 
     Examples::
 
-        config = BertConfig.from_pretrained('bert-base-uncased')
         tokenizer = BertTokenizer.from_pretrained('bert-base-uncased')
-        
-        model = BertForPreTraining(config)
+        model = BertForPreTraining.from_pretrained('bert-base-uncased')
         input_ids = torch.tensor(tokenizer.encode("Hello, my dog is cute")).unsqueeze(0)  # Batch size 1
         outputs = model(input_ids)
         prediction_scores, seq_relationship_scores = outputs[:2]
@@ -824,10 +821,8 @@ class BertForMaskedLM(BertPreTrainedModel):
 
     Examples::
 
-        config = BertConfig.from_pretrained('bert-base-uncased')
         tokenizer = BertTokenizer.from_pretrained('bert-base-uncased')
-        
-        model = BertForMaskedLM(config)
+        model = BertForMaskedLM.from_pretrained('bert-base-uncased')
         input_ids = torch.tensor(tokenizer.encode("Hello, my dog is cute")).unsqueeze(0)  # Batch size 1
         outputs = model(input_ids, masked_lm_labels=input_ids)
         loss, prediction_scores = outputs[:2]
@@ -891,10 +886,8 @@ class BertForNextSentencePrediction(BertPreTrainedModel):
 
     Examples::
 
-        config = BertConfig.from_pretrained('bert-base-uncased')
         tokenizer = BertTokenizer.from_pretrained('bert-base-uncased')
-        
-        model = BertForNextSentencePrediction(config)
+        model = BertForNextSentencePrediction.from_pretrained('bert-base-uncased')
         input_ids = torch.tensor(tokenizer.encode("Hello, my dog is cute")).unsqueeze(0)  # Batch size 1
         outputs = model(input_ids)
         seq_relationship_scores = outputs[0]
@@ -951,10 +944,8 @@ class BertForSequenceClassification(BertPreTrainedModel):
 
     Examples::
 
-        config = BertConfig.from_pretrained('bert-base-uncased')
         tokenizer = BertTokenizer.from_pretrained('bert-base-uncased')
-        
-        model = BertForSequenceClassification(config)
+        model = BertForSequenceClassification.from_pretrained('bert-base-uncased')
         input_ids = torch.tensor(tokenizer.encode("Hello, my dog is cute")).unsqueeze(0)  # Batch size 1
         labels = torch.tensor([1]).unsqueeze(0)  # Batch size 1
         outputs = model(input_ids, labels=labels)
@@ -1057,10 +1048,8 @@ class BertForMultipleChoice(BertPreTrainedModel):
 
     Examples::
 
-        config = BertConfig.from_pretrained('bert-base-uncased')
         tokenizer = BertTokenizer.from_pretrained('bert-base-uncased')
-        
-        model = BertForMultipleChoice(config)
+        model = BertForMultipleChoice.from_pretrained('bert-base-uncased')
         choices = ["Hello, my dog is cute", "Hello, my cat is amazing"]
         input_ids = torch.tensor([tokenizer.encode(s) for s in choices]).unsqueeze(0)  # Batch size 1, 2 choices
         labels = torch.tensor(1).unsqueeze(0)  # Batch size 1
@@ -1127,10 +1116,8 @@ class BertForTokenClassification(BertPreTrainedModel):
 
     Examples::
 
-        config = BertConfig.from_pretrained('bert-base-uncased')
         tokenizer = BertTokenizer.from_pretrained('bert-base-uncased')
-        
-        model = BertForTokenClassification(config)
+        model = BertForTokenClassification.from_pretrained('bert-base-uncased')
         input_ids = torch.tensor(tokenizer.encode("Hello, my dog is cute")).unsqueeze(0)  # Batch size 1
         labels = torch.tensor([1] * input_ids.size(1)).unsqueeze(0)  # Batch size 1
         outputs = model(input_ids, labels=labels)
@@ -1203,10 +1190,8 @@ class BertForQuestionAnswering(BertPreTrainedModel):
 
     Examples::
 
-        config = BertConfig.from_pretrained('bert-base-uncased')
         tokenizer = BertTokenizer.from_pretrained('bert-base-uncased')
-        
-        model = BertForQuestionAnswering(config)
+        model = BertForQuestionAnswering.from_pretrained('bert-base-uncased')
         input_ids = torch.tensor(tokenizer.encode("Hello, my dog is cute")).unsqueeze(0)  # Batch size 1
         start_positions = torch.tensor([1])
         end_positions = torch.tensor([3])

pytorch_transformers/modeling_gpt2.py

@@ -433,9 +433,8 @@ class GPT2Model(GPT2PreTrainedModel):
 
     Examples::
 
-        config = GPT2Config.from_pretrained('gpt2')
         tokenizer = GPT2Tokenizer.from_pretrained('gpt2')
-        model = GPT2Model(config)
+        model = GPT2Model.from_pretrained('gpt2')
         input_ids = torch.tensor(tokenizer.encode("Hello, my dog is cute")).unsqueeze(0)  # Batch size 1
         outputs = model(input_ids)
         last_hidden_states = outputs[0]  # The last hidden-state is the first element of the output tuple
@@ -567,9 +566,8 @@ class GPT2LMHeadModel(GPT2PreTrainedModel):
 
     Examples::
 
-        config = GPT2Config.from_pretrained('gpt2')
         tokenizer = GPT2Tokenizer.from_pretrained('gpt2')
-        model = GPT2LMHeadModel(config)
+        model = GPT2LMHeadModel.from_pretrained('gpt2')
         input_ids = torch.tensor(tokenizer.encode("Hello, my dog is cute")).unsqueeze(0)  # Batch size 1
         outputs = model(input_ids, labels=input_ids)
         loss, logits = outputs[:2]
@@ -683,9 +681,8 @@ class GPT2DoubleHeadsModel(GPT2PreTrainedModel):
 
     Examples::
 
-        config = GPT2Config.from_pretrained('gpt2')
         tokenizer = GPT2Tokenizer.from_pretrained('gpt2')
-        model = GPT2DoubleHeadsModel(config)
+        model = GPT2DoubleHeadsModel.from_pretrained('gpt2')
         choices = ["Hello, my dog is cute [CLS]", "Hello, my cat is cute [CLS]"]  # Assume you've added [CLS] to the vocabulary
         input_ids = torch.tensor([tokenizer.encode(s) for s in choices]).unsqueeze(0)  # Batch size 1, 2 choices
         mc_token_ids = torch.tensor([-1, -1]).unsqueeze(0)  # Batch size 1

pytorch_transformers/modeling_openai.py

@@ -439,9 +439,8 @@ class OpenAIGPTModel(OpenAIGPTPreTrainedModel):
 
     Examples::
 
-        config = OpenAIGPTConfig.from_pretrained('openai-gpt')
         tokenizer = OpenAIGPTTokenizer.from_pretrained('openai-gpt')
-        model = OpenAIGPTModel(config)
+        model = OpenAIGPTModel.from_pretrained('openai-gpt')
         input_ids = torch.tensor(tokenizer.encode("Hello, my dog is cute")).unsqueeze(0)  # Batch size 1
         outputs = model(input_ids)
         last_hidden_states = outputs[0]  # The last hidden-state is the first element of the output tuple
@@ -558,9 +557,8 @@ class OpenAIGPTLMHeadModel(OpenAIGPTPreTrainedModel):
 
     Examples::
 
-        config = OpenAIGPTConfig.from_pretrained('openai-gpt')
         tokenizer = OpenAIGPTTokenizer.from_pretrained('openai-gpt')
-        model = OpenAIGPTLMHeadModel(config)
+        model = OpenAIGPTLMHeadModel.from_pretrained('openai-gpt')
         input_ids = torch.tensor(tokenizer.encode("Hello, my dog is cute")).unsqueeze(0)  # Batch size 1
         outputs = model(input_ids, labels=input_ids)
         loss, logits = outputs[:2]
@@ -665,9 +663,8 @@ class OpenAIGPTDoubleHeadsModel(OpenAIGPTPreTrainedModel):
 
     Examples::
 
-        config = OpenAIGPTConfig.from_pretrained('openai-gpt')
         tokenizer = OpenAIGPTTokenizer.from_pretrained('openai-gpt')
-        model = OpenAIGPTDoubleHeadsModel(config)
+        model = OpenAIGPTDoubleHeadsModel.from_pretrained('openai-gpt')
         choices = ["Hello, my dog is cute [CLS]", "Hello, my cat is cute [CLS]"]  # Assume you've added [CLS] to the vocabulary
         input_ids = torch.tensor([tokenizer.encode(s) for s in choices]).unsqueeze(0)  # Batch size 1, 2 choices
         mc_token_ids = torch.tensor([-1, -1]).unsqueeze(0)  # Batch size 1

pytorch_transformers/modeling_transfo_xl.py

@@ -968,9 +968,8 @@ class TransfoXLModel(TransfoXLPreTrainedModel):
 
     Examples::
 
-        config = TransfoXLConfig.from_pretrained('transfo-xl-wt103')
         tokenizer = TransfoXLTokenizer.from_pretrained('transfo-xl-wt103')
-        model = TransfoXLModel(config)
+        model = TransfoXLModel.from_pretrained('transfo-xl-wt103')
         input_ids = torch.tensor(tokenizer.encode("Hello, my dog is cute")).unsqueeze(0)  # Batch size 1
         outputs = model(input_ids)
         last_hidden_states, mems = outputs[:2]
@@ -1284,9 +1283,8 @@ class TransfoXLLMHeadModel(TransfoXLPreTrainedModel):
 
     Examples::
 
-        config = TransfoXLConfig.from_pretrained('transfo-xl-wt103')
         tokenizer = TransfoXLTokenizer.from_pretrained('transfo-xl-wt103')
-        model = TransfoXLLMHeadModel(config)
+        model = TransfoXLLMHeadModel.from_pretrained('transfo-xl-wt103')
         input_ids = torch.tensor(tokenizer.encode("Hello, my dog is cute")).unsqueeze(0)  # Batch size 1
         outputs = model(input_ids)
         prediction_scores, mems = outputs[:2]

pytorch_transformers/modeling_xlm.py

@@ -472,9 +472,8 @@ class XLMModel(XLMPreTrainedModel):
 
     Examples::
 
-        config = XLMConfig.from_pretrained('xlm-mlm-en-2048')
         tokenizer = XLMTokenizer.from_pretrained('xlm-mlm-en-2048')
-        model = XLMModel(config)
+        model = XLMModel.from_pretrained('xlm-mlm-en-2048')
         input_ids = torch.tensor(tokenizer.encode("Hello, my dog is cute")).unsqueeze(0)  # Batch size 1
         outputs = model(input_ids)
         last_hidden_states = outputs[0]  # The last hidden-state is the first element of the output tuple
@@ -745,9 +744,8 @@ class XLMWithLMHeadModel(XLMPreTrainedModel):
 
     Examples::
 
-        config = XLMConfig.from_pretrained('xlm-mlm-en-2048')
         tokenizer = XLMTokenizer.from_pretrained('xlm-mlm-en-2048')
-        model = XLMWithLMHeadModel(config)
+        model = XLMWithLMHeadModel.from_pretrained('xlm-mlm-en-2048')
         input_ids = torch.tensor(tokenizer.encode("Hello, my dog is cute")).unsqueeze(0)  # Batch size 1
         outputs = model(input_ids)
         last_hidden_states = outputs[0]  # The last hidden-state is the first element of the output tuple
@@ -805,10 +803,8 @@ class XLMForSequenceClassification(XLMPreTrainedModel):
 
     Examples::
 
-        config = XLMConfig.from_pretrained('xlm-mlm-en-2048')
         tokenizer = XLMTokenizer.from_pretrained('xlm-mlm-en-2048')
-        
-        model = XLMForSequenceClassification(config)
+        model = XLMForSequenceClassification.from_pretrained('xlm-mlm-en-2048')
         input_ids = torch.tensor(tokenizer.encode("Hello, my dog is cute")).unsqueeze(0)  # Batch size 1
         labels = torch.tensor([1]).unsqueeze(0)  # Batch size 1
         outputs = model(input_ids, labels=labels)
@@ -885,10 +881,8 @@ class XLMForQuestionAnswering(XLMPreTrainedModel):
 
     Examples::
 
-        config = XLMConfig.from_pretrained('xlm-mlm-en-2048')
         tokenizer = XLMTokenizer.from_pretrained('xlm-mlm-en-2048')
-        
-        model = XLMForQuestionAnswering(config)
+        model = XLMForQuestionAnswering.from_pretrained('xlm-mlm-en-2048')
         input_ids = torch.tensor(tokenizer.encode("Hello, my dog is cute")).unsqueeze(0)  # Batch size 1
         start_positions = torch.tensor([1])
         end_positions = torch.tensor([3])

pytorch_transformers/modeling_xlnet.py

@@ -712,9 +712,8 @@ class XLNetModel(XLNetPreTrainedModel):
 
     Examples::
 
-        config = XLNetConfig.from_pretrained('xlnet-large-cased')
         tokenizer = XLNetTokenizer.from_pretrained('xlnet-large-cased')
-        model = XLNetModel(config)
+        model = XLNetModel.from_pretrained('xlnet-large-cased')
         input_ids = torch.tensor(tokenizer.encode("Hello, my dog is cute")).unsqueeze(0)  # Batch size 1
         outputs = model(input_ids)
         last_hidden_states = outputs[0]  # The last hidden-state is the first element of the output tuple
@@ -1019,9 +1018,8 @@ class XLNetLMHeadModel(XLNetPreTrainedModel):
 
     Examples::
 
-        config = XLNetConfig.from_pretrained('xlnet-large-cased')
         tokenizer = XLNetTokenizer.from_pretrained('xlnet-large-cased')
-        model = XLNetLMHeadModel(config)
+        model = XLNetLMHeadModel.from_pretrained('xlnet-large-cased')
         # We show how to setup inputs to predict a next token using a bi-directional context.
         input_ids = torch.tensor(tokenizer.encode("Hello, my dog is very <mask>")).unsqueeze(0)  # We will predict the masked token
         perm_mask = torch.zeros((1, input_ids.shape[1], input_ids.shape[1]), dtype=torch.float)
@@ -1100,10 +1098,8 @@ class XLNetForSequenceClassification(XLNetPreTrainedModel):
 
     Examples::
 
-        config = XLNetConfig.from_pretrained('xlnet-large-cased')
         tokenizer = XLNetTokenizer.from_pretrained('xlnet-large-cased')
-        
-        model = XLNetForSequenceClassification(config)
+        model = XLNetForSequenceClassification.from_pretrained('xlnet-large-cased')
         input_ids = torch.tensor(tokenizer.encode("Hello, my dog is cute")).unsqueeze(0)  # Batch size 1
         labels = torch.tensor([1]).unsqueeze(0)  # Batch size 1
         outputs = model(input_ids, labels=labels)
@@ -1200,10 +1196,8 @@ class XLNetForQuestionAnswering(XLNetPreTrainedModel):
 
     Examples::
 
-        config = XLMConfig.from_pretrained('xlm-mlm-en-2048')
         tokenizer = XLMTokenizer.from_pretrained('xlm-mlm-en-2048')
-        
-        model = XLMForQuestionAnswering(config)
+        model = XLMForQuestionAnswering.from_pretrained('xlnet-large-cased')
         input_ids = torch.tensor(tokenizer.encode("Hello, my dog is cute")).unsqueeze(0)  # Batch size 1
         start_positions = torch.tensor([1])
         end_positions = torch.tensor([3])