Create scorer.py

scorer.py

@@ -0,0 +1,58 @@
+__version__ = "1.0.0"
+__scorer_id__ = "clarus-f1-thermal-load-v1"
+
+import csv, hashlib, json, sys
+from datetime import datetime, timezone
+
+def _find_label_column(fields):
+    for f in fields:
+        if f.startswith("label_"):
+            return f
+    raise ValueError("No label column")
+
+def _norm(v):
+    v = str(v).strip().lower()
+    return 1 if v in {"1","true","yes"} else 0
+
+def _safe(n,d): return n/d if d else 0.0
+
+def _read(p):
+    with open(p,"r") as f:
+        return list(csv.DictReader(f))
+
+def _hash(p):
+    h=hashlib.sha256()
+    with open(p,"rb") as f:
+        h.update(f.read())
+    return h.hexdigest()
+
+def score(ref, pred):
+    r = _read(ref)
+    p = _read(pred)
+
+    label = _find_label_column(r[0].keys())
+    y_true = [_norm(x[label]) for x in r]
+    y_pred = [_norm(x["prediction"]) for x in p]
+
+    tp = sum(1 for a,b in zip(y_true,y_pred) if a==1 and b==1)
+    tn = sum(1 for a,b in zip(y_true,y_pred) if a==0 and b==0)
+    fp = sum(1 for a,b in zip(y_true,y_pred) if a==0 and b==1)
+    fn = sum(1 for a,b in zip(y_true,y_pred) if a==1 and b==0)
+
+    precision = _safe(tp,tp+fp)
+    recall = _safe(tp,tp+fn)
+    f1 = _safe(2*precision*recall, precision+recall)
+
+    return {
+        "accuracy": _safe(tp+tn,len(y_true)),
+        "precision": precision,
+        "recall": recall,
+        "f1": f1,
+        "false_activation_rate": _safe(fp,fp+tn),
+        "missed_latent_activation_rate": _safe(fn,fn+tp),
+        "hash_ref": _hash(ref),
+        "hash_pred": _hash(pred)
+    }
+
+if __name__ == "__main__":
+    print(json.dumps(score(sys.argv[1], sys.argv[2]), indent=2))