Transformers: BertForMaskedLM "labels" is an unexpected keyword

[guoxuxu]

🐛 Bug

Information

The official BertForMaskedLM example: https://huggingface.co/transformers/model_doc/bert.html#bertformaskedlm
has a bug. When running: outputs = model(input_ids, labels=input_ids), it alerts: TypeError: forward() got an unexpected keyword argument 'labels'

Model I am using (Bert, XLNet ...):
BERT

Language I am using the model on (English, Chinese ...):
English

The problem arises when using:

• [ ] the official example scripts: (give details below)
  from transformers import BertTokenizer, BertForMaskedLM
  import torch

tokenizer = BertTokenizer.from_pretrained('bert-base-uncased')
model = BertForMaskedLM.from_pretrained('bert-base-uncased')

input_ids = torch.tensor(tokenizer.encode("Hello, my dog is cute", add_special_tokens=True)).unsqueeze(0) # Batch size 1
outputs = model(input_ids, labels=input_ids)

• [ ] my own modified scripts: (give details below)
  The tasks I am working on is:
• [ ] an official GLUE/SQUaD task: (give the name)
• [ ] my own task or dataset: (give details below)

To reproduce

Steps to reproduce the behavior:

1. https://huggingface.co/transformers/model_doc/bert.html#bertformaskedlm
   2.
   3.

Expected behavior

should work with labels??

Environment info

• transformers version: 2.11.0
• Platform: Linux
• Python version: 3.7
• PyTorch version (GPU?): 1.5.1
• Tensorflow version (GPU?): 2.2.0
• Using GPU in script?: yes
• Using distributed or parallel set-up in script?: no

[patil-suraj]

Hi, @guoxuxu , lm_labels is changed to labels in a recent commit on master. If you are using master then use labels otherwise use lm_labels

@sgugger This change is causing a lot of confusion. Would it be a good idea to keep master and release docs separate ?

[sgugger]

This has just been done. The documentation now shows the latest stable release (v2.11.0) and you have to opt-in to see the master documentation.

I'll work on a version selector next.

[patil-suraj]

Thanks @sgugger !

[sgugger]

I think we can close the issue as a result. Please reopen if any problem persists.