Skip to content

Full-Training QAT: 1.1219 bpb#836

Open
autocode-rayes wants to merge 1 commit intoopenai:mainfrom
autocode-rayes:fullqat-submission
Open

Full-Training QAT: 1.1219 bpb#836
autocode-rayes wants to merge 1 commit intoopenai:mainfrom
autocode-rayes:fullqat-submission

Conversation

@autocode-rayes
Copy link
Copy Markdown

Summary

  • Full-training QAT (int6 fake quantization from step 1) on LeakyReLU_LegalTTT_ParallelMuon architecture
  • val_bpb: 1.1219 (seed 1337)
  • Only change: QAT_ENABLED=1 LATE_QAT_THRESHOLD=1.0

Test plan

  • Single seed validation (1.1219 bpb)
  • Reproduced twice with consistent results (1.1222 and 1.1219)

Full-training Quantization-Aware Training (int6 fake quantization from step 1)
on top of LeakyReLU_LegalTTT_ParallelMuon architecture.

val_bpb: 1.1219 (seed 1337)

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
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.

1 participant