proper set to deprecated

Author: mayinghan

eval_protocol/models.py

@@ -812,6 +812,7 @@ class ExecutionMetadata(BaseModel):
     # deprecated: use rollout_duration_seconds and eval_duration_seconds instead
     duration_seconds: Optional[float] = Field(
         default=None,
+        deprecated=True,
         description="[Deprecated] Processing duration in seconds for this evaluation row. Note that if it gets retried, this will be the duration of the last attempt.",
     )
 

eval_protocol/pytest/priority_scheduler.py

@@ -197,16 +197,13 @@ async def _run_eval(rows_to_eval: Union[EvaluationRow, List[EvaluationRow]]):
                 if isinstance(eval_res, list):
                     for row in eval_res:
                         self.results.append(row)
-                    # Update eval progress bar (groupwise: 1 eval for the group)
-                    if self.eval_pbar:
-                        self.eval_pbar.update(1)
                 else:
                     self.results.append(eval_res)
-                    # Update eval progress bar (pointwise: 1 eval per row)
-                    if self.eval_pbar:
-                        self.eval_pbar.update(1)
                 return eval_res
             finally:
+                # Always update progress bar (handles both success and failure cases)
+                if self.eval_pbar:
+                    self.eval_pbar.update(1)
                 # Decrement active eval counter
                 async with self.active_evals_lock:
                     self.active_evals -= 1