Non-record: MDLM Masked Diffusion + Depth Recurrence — val_bpb 1.3428 (8×H100, seed=1337)#1582
Open
He-Wenhao wants to merge 1 commit intoopenai:mainfrom
Open
Non-record: MDLM Masked Diffusion + Depth Recurrence — val_bpb 1.3428 (8×H100, seed=1337)#1582He-Wenhao wants to merge 1 commit intoopenai:mainfrom
He-Wenhao wants to merge 1 commit intoopenai:mainfrom
Conversation
Extends PR openai#1403 MDLM baseline with depth recurrence (L1-L3 looped 1x extra = 12 effective layers), QAT/STE, EMA decay=0.997, GPTQ-lite clip search, linear LR->0, relu^2 MLP, Muon WD=0.01. val_bpb: 1.3428 | quant penalty: 0.0049 | artifact: 14.73 MB 8xH100 SXM, 600s, seed=1337 Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Summary
val_bpb: 1.3428 (int8+zlib roundtrip) | 14.73 MB | 8×H100 SXM, 600s | Beats #1403 by 0.0057 BPB
Extends the MDLM baseline (#1403) with depth recurrence and quantization improvements.
Stack
Results (8×H100 SXM, seed=1337, 600s)
EMA + GPTQ-lite cuts quant penalty from 0.0076 → 0.0049. Depth recurrence improves pre-quant quality (1.3379 vs 1.3409) even with fewer steps, because ~12 effective layers of compute per forward pass.