All of these improvements were developed by Claude running autonomously over ~2 days using autoresearch. I didn't touch anything - incredible. All tuning was done on d12 but generalized easily to larger models (e.g. d24 in particular). This means we will also get a new "Time to GPT-2" Leaderboard entry, which I will push separately.

Optimizer & schedule changes:
- Increase unembedding LR 0.004 -> 0.008, weight decay 0.2 -> 0.28
- Per-group Adam betas and weight decay (instead of shared global betas)
- Muon beta2 0.95 -> 0.9, momentum warmup target 0.95 -> 0.97 over 400 steps
- Warmup: ratio-based -> absolute steps (default 40)
- Warmdown ratio 0.5 -> 0.65, final LR fraction 0.0 -> 0.05
- Weight decay schedule: linear -> cosine decay
- Polar express norm factor 1.02 -> 1.01

Architecture & init changes:
- VE gate: channels 32 -> 12, scale range 2x -> 3x, init small positive
- Add post-QK-norm scaling (q,k *= 1.15) for sharper attention
- Embedding init std 1.0 -> 0.8, MLP c_fc init 0.5x smaller
- RoPE base theta 10K -> 100K
- Short attention window: seq_len/2 -> ~seq_len/3 (ceil to 128 tile)
- Logit softcap 20 -> 15

nanochat/optim.py

@@ -113,7 +113,7 @@ def muon_step_fused(
 
     # Polar express
     X = g.bfloat16()
-    X = X / (X.norm(dim=(-2, -1), keepdim=True) * 1.02 + 1e-6)
+    X = X / (X.norm(dim=(-2, -1), keepdim=True) * 1.01 + 1e-6)
     if g.size(-2) > g.size(-1): # Tall matrix
         for a, b, c in polar_express_coeffs[:ns_steps]:
             A = X.mT @ X