[Non-record] H-Net MAMBA Outer-Layer Ablation: OL2 collapses, OL1 converges to 1.5194 INT6 BPB#1757
Open
aiejvn wants to merge 1 commit intoopenai:mainfrom
Open
[Non-record] H-Net MAMBA Outer-Layer Ablation: OL2 collapses, OL1 converges to 1.5194 INT6 BPB#1757aiejvn wants to merge 1 commit intoopenai:mainfrom
aiejvn 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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Investigates the effect of MAMBA outer-layer count on H-Net training stability within a fixed 9-layer budget. Increasing from OL1 to OL2 (2 enc + 5 main + 2 dec) caused catastrophic training collapse at step 2000 — val BPB spiked from 1.81 to 2.53 in a single validation window — while also running 2.2× slower per step (~9.2s vs ~4.3s). Reverting to OL1 (1 enc + 7 main + 1 dec) stabilized training across all 20k steps, converging to 1.4770 float BPB and 1.5194 INT6 sliding-window BPB. Val BPB was still decreasing at the 20k cutoff, suggesting longer runs would improve further.