director: 160 training scripts + fixed block lookup

kaggle/director_sft.py

@@ -35,11 +35,13 @@ def find_blocks(model):
     """Longest ModuleList of decoder blocks, whatever it is nested under."""
     import torch.nn as nn
 
-    best = (0, None)
+    best = (0, "", None)
     for name, mod in model.named_modules():
         if isinstance(mod, nn.ModuleList) and len(mod) > best[0]:
-            best = (len(mod), (name, mod))
-    return best[1]
+            best = (len(mod), name, mod)
+    if best[2] is None:
+        raise RuntimeError("no decoder block list found")
+    return best
 
 
 def build_example(tok, topic: str, lines: list[dict]) -> dict:
@@ -67,7 +69,7 @@ def train(rows: list[dict]) -> str:
     tok = AutoTokenizer.from_pretrained(MODEL_ID)
     model = AutoModelForCausalLM.from_pretrained(MODEL_ID, dtype=torch.float32,
                                                  device_map="cuda:0")
-    n_layers, (lname, blocks) = find_blocks(model)
+    n_layers, lname, blocks = find_blocks(model)
     log(f"blocks: {lname} x{n_layers}; training the last {TRAIN_LAYERS}")
 
     for p in model.parameters():
@@ -115,6 +117,8 @@ def train(rows: list[dict]) -> str:
 def main() -> None:
     OUT.mkdir(parents=True, exist_ok=True)
     path = OUT / "sft.jsonl"
+    if not path.exists():  # reuse a set collected by an earlier run
+        path = ROOT / "webfm-eval" / "director" / "sft.jsonl"
     rows = [json.loads(l) for l in path.read_text().splitlines() if l.strip()]
     log(f"{len(rows)} training scripts from {path}")
     if not rows: