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

[BART] FP16 testing fixes #3266

Merged
merged 6 commits into from
Mar 13, 2020

Conversation

sshleifer
Copy link
Contributor

@sshleifer sshleifer commented Mar 13, 2020

closes #3249: fp16 forward pass failing when no decoder_attention_mask provided. Adds test coverage.

closes #3265: test_generate_fp16 was failing since #3140 (by sending proper kwargs to BartForConditionalGenerate.generate)

@sshleifer sshleifer requested review from julien-c and thomwolf March 13, 2020 16:48
@sshleifer
Copy link
Contributor Author

I have a suspicion that this will also fix the GPU test runner.

@sshleifer sshleifer requested a review from LysandreJik March 13, 2020 20:17
Copy link
Member

@thomwolf thomwolf left a comment

Choose a reason for hiding this comment

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

Ok, usually people add a dtype argument to methods like _prepare_bart_decoder_inputs for your info.

@sshleifer sshleifer changed the title [bart.fix] FP16 testing fixes [BART] FP16 testing fixes Mar 13, 2020
@sshleifer sshleifer merged commit 2bd79e2 into huggingface:master Mar 13, 2020
@sshleifer sshleifer deleted the fp16-base-model-support branch March 13, 2020 23:48
jplu pushed a commit to jplu/transformers that referenced this pull request Mar 25, 2020
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.

[BART] test_generate_fp16 fails after PR#3140 Using FP16 on BartModel
3 participants