card: thresholded-ternary best 2.1469 (per-group scale, zero state)

README.md

@@ -39,25 +39,32 @@ Rust.
 | Matrices | 168 ayeOS per-layer files (`m000.json` … `m167.json`) |
 | Runtime assets | `embeddings.f16` `[151936, 896]`, `norms.f32` `[49, 896]` |
 | Training | continued-train, masked CE, deployed-forward QAT |
-| Validation (masked, n=90) | **0.5597** |
-| Checkpoint sha256 | `69af3001...0409` (final, epoch 21) |
+| Validation (masked, n=90) | **2.1469** |
+| Checkpoint sha256 | `21294c68...8285` (best, epoch 2 / early stop) |
 
 ## Training
 
-- **Data**: 14,330 text samples (konstellation corpus: kompress + domain +
+- **Data**: 20,007 text samples (konstellation corpus: kompress + domain +
   c3), stratified 90-sample held-out val.
 - **Loss**: masked cross-entropy — pad tokens (id 0) weighted out, honest mean
   over valid tokens. Dynamic per-batch padding bucketed to multiples of 64.
-- **Optimizer**: AdamW, weight decay 0.1, grad clip off, lr 5e-5 → cosine →
-  5e-6 (2% warmup), early stop patience 5 / min-delta 0.02, 40-epoch cap.
-- **Hardware**: vast.ai H100 (80 GB) — the SOTA run, B8/256, 14,330 samples.
+- **Optimizer**: AdamW, weight decay 0.1, grad clip off, lr 3e-4 → cosine →
+  3e-5 (2% warmup), early stop patience 5 / min-delta 0.02, 40-epoch cap.
+- **Hardware**: vast.ai RTX PRO 6000 (96 GB) — the nightly run, B12/256,
+  20,007 samples. `MLX_CUDA_GRAPH_CACHE_SIZE=2000`.
+- **Quantizer**: true thresholded ternary — `weight_quant` uses a per-group
+  (64) scale and a real zero state (`|w| < 0.5·scale → 0`), so ~30% of the
+  weights land on zero. The old sign-based collapse wasted the third state;
+  this one earns it. Training forward ≡ export ≡ Rust (per-group scale, same
+  dequant `(code−1)·scale`).
 - **Forward**: deployed-forward QAT — the exact forward the Rust runner uses
   (weight-quant-only BitLinear, per-projection RMSNorm + activation quant
   skipped), so training ≡ inference. The Rust runner reproduces the forward
-  to **1e-5** (golden-logits gate, both prompts, identical argmax).
-- **Val trajectory** (SOTA run): 2.90 → 1.68 → 1.27 → 0.77 → 0.67 → 0.63 →
-  0.60 → **0.5597** (epoch 21, early stop). Prior runs: 1.6998 (7,000
-  samples, L40), 3.2862 (2,785, 3090); the old artifact measured 11.34.
+  to **1.3e-5** (golden-logits gate, both prompts, identical argmax 71703).
+- **Val trajectory** (nightly run): 2.38 → **2.1469** (epoch 2, best) → 2.25
+  → 2.49 → 2.72 → 2.76 → 3.07 (overfit after epoch 2; early stop). Prior
+  runs: 1.6998 (7,000 samples, L40), 3.2862 (2,785, 3090); the old artifact
+  measured 11.34.
 
 ## Layout