Arthur comments

Results 795 comments of


                                            Arthur

error when using from indobenchmark

Hey! The `tokenizer_class` that was set in the configuration.json is wrong as the `IndoNLGTokenizer` does not exist in transformers. You should try to ask the other of the model on...

Code crashes without errors when importing Trainer in TPU context

@muellerzr is there a PR or Issue we can track and link here?

Gradient checkpointing throws use_reentrant warning on PyTorch 2.1

Thanks for raising! given that we had #27020, this should be fairly easy to fix! cc @younesbelkada

Gradient checkpointing throws use_reentrant warning on PyTorch 2.1

Will merge the PR today

[OPT] Adds `GPT2TokenizerFast` to the list of tokenizer to use for OPT.

The previous tests were passing but the tokenizer was `slow` were it should have been `fast` 😅

[OPT] Adds `GPT2TokenizerFast` to the list of tokenizer to use for OPT.

Merging ASAP.

[OPT] Adds `GPT2TokenizerFast` to the list of tokenizer to use for OPT.

cc @ydshieh the failing tests are related to the length of the dictionary of the tokenizer. Spaces are encoded to `222`, which is then passed to the model, while the...

[OPT] Adds `GPT2TokenizerFast` to the list of tokenizer to use for OPT.

The problem is not from the Fast tokenizer (it is a GPT2 tokenizer) but the tiny config test. GPT2TokenizerFast is pretty much full proof at this points. I was just...

[OPT] Adds `GPT2TokenizerFast` to the list of tokenizer to use for OPT.

As @ydshieh said it might have been fixed, the problem is that the CI is still kind of stuck/not working ... not really sure if it is only this PR,...

[OPT] Adds `GPT2TokenizerFast` to the list of tokenizer to use for OPT.

I will run them locally 😉