WIP: Shared-transformer + warmdown-aligned training (not final submis…#420
Open
leofeasby wants to merge 1 commit intoopenai:mainfrom
Open
WIP: Shared-transformer + warmdown-aligned training (not final submis…#420leofeasby wants to merge 1 commit intoopenai:mainfrom
leofeasby wants to merge 1 commit intoopenai:mainfrom
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Shared-transformer architecture with MLP×5 expansion, bigram hash embedding
(4096),
weight decay tuning, SWA, and warmdown-aligned training schedule.
Key finding: performance is dominated by a late-stage phase transition driven
by LR
warmdown. Aligning WARMDOWN_ITERS with the wallclock budget moves this
transition
earlier and converts wasted steps into usable convergence.
Results on 8×H100 (within 10-min budget):
Work in progress — not a final submission.