update doc

hubconfs/gpt_hubconf.py

@@ -146,6 +146,7 @@ def openAIGPTLMHeadModel(*args, **kwargs):
 		# Get the predicted last token
 		>>> predicted_index = torch.argmax(predictions[0, -1, :]).item()
 		>>> predicted_token = tokenizer.convert_ids_to_tokens([predicted_index])[0]
+        '.</w>'
     """
     model = OpenAIGPTLMHeadModel.from_pretrained(*args, **kwargs)
     return model