add loggin

Author: pawalt

slime/code_golf_harbor_modal/configs/base.py

@@ -27,6 +27,7 @@ class RLConfig:
     harbor_rm_timeout_sec: int = 120
     harbor_length_bonus_weight: float = 0.2
     harbor_rm_profile: bool = False
+    harbor_rm_log_samples: bool = False
 
     @property
     def train_script(self) -> str:

slime/code_golf_harbor_modal/configs/qwen_8b_multi.py

@@ -20,6 +20,7 @@ def get_config() -> RLConfig:
         wandb_project="slime-code-golf",
         wandb_run_name_prefix="qwen8b-mbpp-harbor",
         harbor_rm_profile=True,
+        harbor_rm_log_samples=True,
         slime_args=f"""
             # Model
             {QWEN3_8B_MODEL_ARGS}

slime/code_golf_harbor_modal/custom_rm.py

@@ -112,6 +112,7 @@ def _compose_reward(
 
 async def _score_sample(sample: Any) -> float:
     profile_enabled = _get_env_bool("HARBOR_RM_PROFILE", default=False)
+    log_samples = _get_env_bool("HARBOR_RM_LOG_SAMPLES", default=False)
     t0 = time.perf_counter()
     timings_s: dict[str, float] = {}
 
@@ -175,6 +176,15 @@ def _mark(name: str) -> None:
             "candidate_bytes": candidate_size,
             "reward": reward,
         }
+        if log_samples:
+            print(
+                "harbor_rm_sample: "
+                f"task_id={label_payload.get('task_id')} "
+                f"pass_rate={pass_rate:.6f} "
+                f"reward={reward:.6f} "
+                f"candidate_bytes={candidate_size} "
+                f"reference_bytes={reference_size}"
+            )
         if profile_enabled:
             timestamps = {**timings_s, "done": time.perf_counter()}
             timing_ms = {
@@ -219,6 +229,14 @@ def _mark(name: str) -> None:
             sample.metadata = {}
         sample.metadata["harbor_rm_error"] = repr(exc)
         print(f"harbor_rm_error: {exc!r}")
+        if log_samples:
+            print(
+                "harbor_rm_sample_error: "
+                f"task_id={label_payload.get('task_id')} "
+                f"error={exc!r} "
+                f"candidate_bytes={candidate_size} "
+                f"reference_bytes={reference_size}"
+            )
         if profile_enabled:
             timing_ms = {
                 "total_until_error": round((time.perf_counter() - t0) * 1000, 2)

slime/code_golf_harbor_modal/modal_train.py

@@ -115,6 +115,7 @@ def _build_runtime_env(config: RLConfig, master_addr: str) -> dict:
             "HARBOR_RM_TIMEOUT_SEC": str(config.harbor_rm_timeout_sec),
             "HARBOR_LENGTH_BONUS_WEIGHT": str(config.harbor_length_bonus_weight),
             "HARBOR_RM_PROFILE": "1" if config.harbor_rm_profile else "0",
+            "HARBOR_RM_LOG_SAMPLES": "1" if config.harbor_rm_log_samples else "0",
         }
     }