Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[test] add test for --config_overrides #14466

Merged
merged 2 commits into from Nov 22, 2021
Merged

Conversation

stas00
Copy link
Contributor

@stas00 stas00 commented Nov 20, 2021

#14389 suggested that --config_overrides doesn't work.

The feature works just fine, it's just the multiple logging of the config done by the framework is confusing. I already flagged this issue here: #11104 I have no idea why loading a tokenizer triggers dumping of model config - as its contents are mostly irrelevant to the tokenizer and surely doesn't contribute anything useful to the user, other than avoiding looking at the log completely.

So there will be no confusion I added an additional dump with the updated config (terrible!) and a test so that we don't accidentally break this feature.

I'm not sure how else to improve that other than revisiting the design of when the model config is dumped. IMHO:

  1. it's dumped too soon for the model (before it can be updated)
  2. it shouldn't be dumped at all for the tokenizer

Fixes: #14389

@sgugger, @LysandreJik

Copy link
Collaborator

@sgugger sgugger left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks for adding the test and the logs!

examples/pytorch/test_examples.py Outdated Show resolved Hide resolved
Copy link
Member

@LysandreJik LysandreJik left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Great, thanks @stas00!

@LysandreJik LysandreJik merged commit 11f65d4 into master Nov 22, 2021
@LysandreJik LysandreJik deleted the config_overrides_test branch November 22, 2021 16:33
Albertobegue pushed a commit to Albertobegue/transformers that referenced this pull request Jan 27, 2022
* add test for --config_overrides

* remove unneeded parts of the test
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

--config_overrides doesn't appear to work in run_clm.py when trying to specify a larger GPT model
3 participants