v5b: add --submit-only SFT mode

training/train_sft.py

@@ -92,6 +92,7 @@ class SftConfig:
     eval_holdout_canonical: int = 3
     eval_holdout_hard: int = 3
     eval_every_epoch: bool = True
+    submit_only: bool = False
 
     lr: float = 5e-5
     grad_clip: float = 1.0
@@ -309,6 +310,9 @@ def main() -> None:
     env = InvoiceGuardEnvironment()
     train_tasks, eval_tasks = split_tasks(cfg)
     examples = build_sft_examples(tokenizer, env, train_tasks, cfg.max_prompt_tokens)
+    if cfg.submit_only:
+        examples = [ex for ex in examples if ex["action_type"] == "submit_final_resolution"]
+        print(f"[setup] submit-only mode: filtered to {len(examples)} submit examples", flush=True)
     print(f"[setup] train_tasks={len(train_tasks)} eval_tasks={len(eval_tasks)} examples={len(examples)}", flush=True)
 
     tracker = None
@@ -434,6 +438,7 @@ def _parse_args() -> SftConfig:
     p.add_argument("--max-prompt-tokens", type=int, default=None)
     p.add_argument("--no-push", action="store_true")
     p.add_argument("--no-4bit", action="store_true")
+    p.add_argument("--submit-only", action="store_true")
     args = p.parse_args()
 
     cfg = SftConfig()
@@ -459,6 +464,8 @@ def _parse_args() -> SftConfig:
         cfg.push_to_hub = False
     if args.no_4bit:
         cfg.use_4bit = False
+    if args.submit_only:
+        cfg.submit_only = True
     return cfg