small fixes

src/bespokelabs/curator/request_processor/base_online_request_processor.py

@@ -222,7 +222,6 @@ def run(
                 self.process_requests_from_file(
                     generic_request_filepath=request_file,
                     save_filepath=response_file,
-                    max_attempts=self.max_retries,
                     resume=True,
                 )
             )
@@ -233,7 +232,6 @@ async def process_requests_from_file(
         self,
         generic_request_filepath: str,
         save_filepath: str,
-        max_attempts: int,
         resume: bool,
         resume_no_retry: bool = False,
     ) -> None:
@@ -353,7 +351,7 @@ async def process_requests_from_file(
                         task_id=status_tracker.num_tasks_started,
                         generic_request=generic_request,
                         api_specific_request=self.create_api_specific_request(generic_request),
-                        attempts_left=max_attempts,
+                        attempts_left=self.max_retries,
                         prompt_formatter=self.prompt_formatter,
                     )
 
@@ -406,7 +404,7 @@ async def process_requests_from_file(
                     token_estimate = self.estimate_total_tokens(
                         retry_request.generic_request.messages
                     )
-                    attempt_number = 1 + self.max_retries - retry_request.attempts_left
+                    attempt_number = self.max_retries - retry_request.attempts_left
                     logger.info(
                         f"Processing retry for request {retry_request.task_id} "
                         f"(attempt #{attempt_number} of {self.max_retries}). "
@@ -483,7 +481,7 @@ async def handle_single_request_with_retries(
 
         except Exception as e:
             logger.warning(
-                f"Request {request.task_id} failed with Exception {e}, attempts left {request.attempts_left}"
+                f"Request {request.task_id} failed with Exception: {e}, attempts left: {request.attempts_left}"
             )
             status_tracker.num_other_errors += 1
             request.result.append(e)

src/bespokelabs/curator/request_processor/base_request_processor.py

@@ -77,15 +77,17 @@ def run(
         """
         pass
 
-    def _verify_existing_request_files(self, working_dir: str, dataset: Optional[Dataset]) -> List[int]:
+    def _verify_existing_request_files(
+        self, working_dir: str, dataset: Optional[Dataset]
+    ) -> List[int]:
         """
         Verify integrity of the cache (each request file has associated metadata, and the number of rows is correct),
         and return the indices of request files that need to be regenerated (so that no work is repeated).
-        
+
         Args:
             working_dir (str): Working directory where cache files are expected to be (requests.jsonl, metadata.json)
             dataset (Optional[Dataset]): The dataset that we want to create requests from
-            
+
         Returns:
             List[int]: Indices of missing files
         """
@@ -114,11 +116,15 @@ def _verify_existing_request_files(self, working_dir: str, dataset: Optional[Dat
                     if num_jobs != expected_num_jobs:
                         incomplete_files.append(i)
 
-            logger.info(f"Cache missing {len(incomplete_files)} complete request files - regenerating missing ones.")
+            logger.info(
+                f"Cache missing {len(incomplete_files)} complete request files - regenerating missing ones."
+            )
             return incomplete_files
-            
+
         except:
-            logger.info("Cache verification failed for unexpected reasons - regenerating all request files.")
+            logger.info(
+                "Cache verification failed for unexpected reasons - regenerating all request files."
+            )
             incomplete_files = list(range(expected_num_files))
             return incomplete_files
 
@@ -195,13 +201,16 @@ async def create_all_request_files():
                         metadata_files[i],
                         start_idx=i * self.batch_size,
                     )
-                    for i in range(num_batches) if i in incomplete_files
+                    for i in range(num_batches)
+                    if i in incomplete_files
                 ]
                 await asyncio.gather(*tasks)
 
             run_in_event_loop(create_all_request_files())
         else:
-            run_in_event_loop(self.acreate_request_file(dataset, prompt_formatter, request_file, metadata_file))
+            run_in_event_loop(
+                self.acreate_request_file(dataset, prompt_formatter, request_file, metadata_file)
+            )
 
         return request_files
 

src/bespokelabs/curator/request_processor/litellm_online_request_processor.py

@@ -301,9 +301,14 @@ async def call_single_request(
         except litellm.NotFoundError as e:
             cost = 0
 
+        if completion_obj.choices[0].finish_reason == "content_filter":
+            raise ValueError(
+                f"finish_reason was content_filter with raw response {completion_obj.model_dump()} for request {request.generic_request.messages}"
+            )
+
         if response_message is None:
             raise ValueError(
-                f"Request {request.task_id} returned no response message with raw response {completion_obj.model_dump()}"
+                f"response_message was None with raw response {completion_obj.model_dump()}"
             )
 
         # Create and return response