Skip to content

submission: Int6 MLP3x + QAT + SlidingWindow (val_bpb: 1.1702) #117

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.

Sign up for GitHub

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

Open
trovatochris wants to merge 2 commits into
base: main
Choose a base branch
from
Open

submission: Int6 MLP3x + QAT + SlidingWindow (val_bpb: 1.1702) #117

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
b9909ba
Create README.md
trovatochris Mar 19, 2026
ba47f6e
Create train_gpt.py
trovatochris Mar 19, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
22 changes: 22 additions & 0 deletions submissions/trovatochris/README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,22 @@
# Submission: Int6 MLP3x + QAT + SlidingWindow (val_bpb: 1.1702)

## Summary
Stacked int6 per-row quantization + zstd22 compression, 3x MLP expansion, QAT weight-snapping at 70% training, tuned Muon optimizer (momentum=0.99 with warmup), extended warmdown (3000 iters), and stride-64 sliding window evaluation.

## Configuration
MLP_MULT=3 VAL_LOSS_EVERY=0 MATRIX_LR=0.02 SCALAR_LR=0.02 TIED_EMBED_LR=0.03 MUON_MOMENTUM=0.99 MUON_MOMENTUM_WARMUP_STEPS=1500 MUON_MOMENTUM_WARMUP_START=0.92 WARMDOWN_ITERS=3000 ENABLE_QAT=1 QAT_START_FRAC=0.7 USE_SLIDING_EVAL=1 EVAL_STRIDE=64

## Key Metrics
- Training steps: 9,540/20,000 (600s wallclock cap)
- Average step time: 62.86ms
- Model params: 21,778,504
- Standard val_bpb: 1.2040
- Sliding val_bpb: 1.1702
- Total artifact: 15,306,777 bytes (int6+zstd22)
- Under 16MB: YES (693,223 bytes headroom)

## Command
torchrun --standalone --nproc_per_node=8 train_gpt.py

## Seed
SEED=1337
Loading