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

Adafactor Jax suspiciously slow on ViT and Conformer compared to Adam #392

Closed
priyakasimbeg opened this issue May 23, 2023 · 1 comment
Closed

Comments

@priyakasimbeg
Copy link
Contributor

Adafactor seems to be about 20% slower on ViT and Conformer compared to Adam, while internal runs show comparable runtimes.

Steps to Reproduce

python3 submission_runner.py --framework=jax --workload=librispeech_conformer --submission_path=baselines/adafactor/jax/submission.py --tuning_search_space=baselines/adafactor/tuning_search_space.json --data_dir=/data/librispeech --num_tuning_trials=1 --experiment_dir=/experiment_runs --experiment_name=timing_fancy_adafactor/timing_adafactor --overwrite=True --save_checkpoints=False --max_global_steps=20000 --librispeech_tokenizer_vocab_path=/data/librispeech/spm_model.vocab 
@priyakasimbeg
Copy link
Contributor Author

Won't fix

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

No branches or pull requests

1 participant