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

Remove gas (beano) #1144

Merged
merged 3 commits into from
Feb 5, 2024
Merged

Remove gas (beano) #1144

merged 3 commits into from
Feb 5, 2024

Conversation

segyges
Copy link
Contributor

@segyges segyges commented Feb 4, 2024

per code being removed:

    """gradient_accumulation_steps"""  # TODO this is a duplicate, remove?

This is set in many configs and never assigned from. Its continued existence in neox only encourages people to set gas, which does apparently nothing, instead of setting gradient_accumulation_steps, which is probably what they intended to do.

Pythia configs left alone because the source of truth for those configs is the pythia repo.

@segyges segyges changed the title Remove gas Remove gas (beano) Feb 4, 2024
@segyges
Copy link
Contributor Author

segyges commented Feb 4, 2024

Oh: And any model run with "gas" set at anything other than "1" will have had batch size smaller by 1/their_gas_size than they intended. People might want to check their previous work for that.

@Quentin-Anthony
Copy link
Member

Good catch!

@Quentin-Anthony Quentin-Anthony merged commit 91c44bc into EleutherAI:main Feb 5, 2024
3 of 5 checks passed
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.

None yet

2 participants