wip

src/projected_grpo/train.py

@@ -161,7 +161,7 @@ class Config:
     # preset doesn't burn its first 10 steps at 1e-3-of-peak LR. 0.1 = ariahw
     # canonical 10/100 = 10% at the 100-step regime they used.
     warmup_frac: float = 0.1
-    grad_clip: float = 1.0            # global L2 clip on delta_S grads; set high (e.g. 500) to effectively disable
+    grad_clip: float = 10.0           # global L2 clip on delta_S grads (sane new-env default; was 1.0/500-disabled)
     seed: int = 41
     preserve_magnitude: bool = True
     gate_mode: Literal["one_sided", "no_gate", "reverse"] = "one_sided"
@@ -264,7 +264,7 @@ class FastConfig(Config):
     n_problems=200 keeps teacher_pool coverage (only ~40 prompts touched
     at pp=4 x 20 steps)."""
     model: str = "Qwen/Qwen3-4B"
-    steps: int = 20
+    steps: int = 60                   # sane new-env default (was 20; 60 lets the gap open at convergence)
     group: int = 8                    # G=8 so the locked-in mix_ratio=0.125 gives 1 teacher / 7 student
     max_new: int = 512
     n_problems: int = 200