From ad048e59c6a0ce7004157b5067554bcbd9132af1 Mon Sep 17 00:00:00 2001
From: wassname <github@wassname>
Date: Sun, 31 May 2026 22:45:12 +0000
Subject: [PATCH] fix: results.py parses gt_s/hack_s by header name, not stale
 fixed indices

Old GT_S=6/HACK_S=8 were the pre-sprd/N layout; current table is gt_s=4
hack_s=6, so newer logs were silently mis-read and old distill logs crashed
_frac on a non-fraction token. Now locate the train.py streaming header
(first token 'step' + 'ref_eq' present) and map columns by name.

Co-Authored-By: Claudypoo <288921227+claudypoo@users.noreply.github.com>
---
 scripts/results.py | 29 +++++++++++++++++++++++++----
 1 file changed, 25 insertions(+), 4 deletions(-)

diff --git a/scripts/results.py b/scripts/results.py
index fc7baef..a2cc358 100644
--- a/scripts/results.py
+++ b/scripts/results.py
@@ -20,8 +20,14 @@ from tabulate import tabulate
 
 LOG_DIR = Path("logs")
 TS_RE = re.compile(r"(\d{8}T\d{6})")
-# per-step row after "| INFO |": step ref_eq rew rew_s sprd N gt_s gt_t hack_s hack_t ...
-GT_S, HACK_S = 6, 8  # column indices into the whitespace-split row
+# Column positions are read from the header row by NAME, not hardcoded -- the
+# per-step table layout has changed over time (sprd/N dropped, cin/cout/hk_dep
+# added) so fixed indices silently mis-read newer logs and crash on smoke logs.
+
+
+def _colname(tok: str) -> str:
+    # header tokens carry direction glyphs / markers: "gt_s↑", "hack_s?" -> "gt_s", "hack_s"
+    return re.sub(r"[^a-z0-9_]", "", tok.lower())
 
 
 def _frac(tok: str) -> float | None:
@@ -64,14 +70,29 @@ def parse_log(path: Path) -> dict | None:
     preset_line = next((l for l in txt.splitlines() if "preset=" in l and "arm=" in l), "")
     if argv is None:
         return None
+    # Locate the per-step table header to map gt_s/hack_s columns by NAME. The
+    # train.py streaming table is the INFO line whose tokens start with "step"
+    # and include "ref_eq" -- that signature excludes the old distill_* logs
+    # which also have "step ..." lines but a different (hack=.. pass=..) format.
+    header, names = None, []
+    for l in txt.splitlines():
+        if "| INFO |" not in l:
+            continue
+        toks = [_colname(t) for t in l.split("| INFO |", 1)[1].split()]
+        if toks[:1] == ["step"] and "ref_eq" in toks:
+            header, names = l, toks
+            break
+    if header is None:
+        return None  # not a train.py streaming run
+    idx_hack, idx_gt = names.index("hack_s"), names.index("gt_s")
     hs, gts = [], []
     for line in txt.splitlines():
         if "| INFO |" not in line:
             continue
         row = line.split("| INFO |", 1)[1].split()
-        if not row or not row[0].isdigit() or len(row) <= HACK_S:
+        if not row or not row[0].isdigit() or len(row) <= idx_hack:
             continue
-        h, g = _frac(row[HACK_S]), _frac(row[GT_S])
+        h, g = _frac(row[idx_hack]), _frac(row[idx_gt])
         if h is not None:
             hs.append(h)
         if g is not None: