report pipeline: build_report.py (self-contained HTML), eval --runs-dir, plot fix, halting per-input steps

1 files changed, 5 insertions, 2 deletions

diff --git a/src/eval.py b/src/eval.py
index 0e769ac..7dcd289 100644
--- a/src/eval.py
+++ b/src/eval.py
@@ -132,7 +132,8 @@ def halting_report(model, cfg: Config) -> dict:
     else:
         code = "H4"          # intermediate but no positive gap correlation — noisy/unused
     return {"code": code, "mean_steps": mean, "corr_gap": rho,
-            "min_steps": cfg.min_steps, "max_steps": cfg.max_steps}
+            "min_steps": cfg.min_steps, "max_steps": cfg.max_steps,
+            "gaps": gaps, "steps_by_n": steps}
 
 
 def _load(out_dir: str, ckpt: str, cfg: Config):
@@ -146,8 +147,10 @@ def main() -> None:
     ap = argparse.ArgumentParser(description="prime-grokking eval")
     ap.add_argument("model")
     ap.add_argument("seed")
+    ap.add_argument("--ckpt", default="last.pt")
+    ap.add_argument("--runs-dir", default="runs")
     a = ap.parse_args()
-    out_dir = os.path.join("runs", a.model, f"seed{a.seed}")
+    out_dir = os.path.join(a.runs_dir, a.model, f"seed{a.seed}")
     cfg = Config.load(os.path.join(out_dir, "config.json"))
 
     _, val_in = get_splits(cfg)