Skip to content

update qwen-image text token length error #1042

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
llmc-reviewer merged 1 commit into from
Apr 28, 2026
Merged

update qwen-image text token length error #1042

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
4 changes: 4 additions & 0 deletions lightx2v/models/input_encoders/hf/qwen25/qwen25_vlforconditionalgeneration.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -218,6 +218,10 @@ def infer(self, text, image_list=None):
drop_idx = self.prompt_template_encode_start_idx
txt = [template.format(e) for e in text]

token_lengths = [len(ids) for ids in self.tokenizer(txt, add_special_tokens=True)["input_ids"]]
max_token_len = max(token_lengths)
if max_token_len > self.tokenizer_max_length + drop_idx:
raise ValueError(f"Input text token length ({max_token_len - drop_idx}) exceeds ({self.tokenizer_max_length}). Please shorten the input text.")
Comment on lines +221 to +224
Copy link
Copy Markdown
Contributor

@gemini-code-assist gemini-code-assist Bot Apr 28, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

medium

The tokenizer is called twice for the same input text: once here to calculate lengths and again on the following line to generate the tensors. This is inefficient, especially for large batches or long prompts.\n\nConsider calling the tokenizer once with padding=True and return_tensors="pt", then checking the length of the resulting input_ids tensor before moving it to the device.

model_inputs = self.tokenizer(txt, max_length=self.tokenizer_max_length + drop_idx, padding=True, truncation=True, return_tensors="pt").to(AI_DEVICE)
encoder_hidden_states = self.text_encoder(
input_ids=model_inputs.input_ids,
Expand Down
3 changes: 2 additions & 1 deletion lightx2v/server/api/server.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -145,7 +145,8 @@ async def _process_single_task(self, task_info: Any):

except Exception as e:
logger.exception(f"Task {task_id} processing failed: {str(e)}")
task_manager.fail_task(task_id, str(e))
original_et = getattr(e, "original_error_type", "") or ""
task_manager.fail_task(task_id, str(e), error_type=original_et or None)
finally:
if lock_acquired:
task_manager.release_processing_lock(task_id)
Expand Down
12 changes: 11 additions & 1 deletion lightx2v/server/api/tasks/image.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -34,7 +34,11 @@ async def _wait_task_and_stream_result(task_id: str, timeout_seconds: int, poll_
raise HTTPException(status_code=500, detail=f"Task completed but no in-memory image found: {task_id}")

if status == TaskStatus.FAILED.value:
raise HTTPException(status_code=500, detail=task_status.get("error", "Task failed"))
error_type = task_status.get("error_type", "")
error_detail = task_status.get("error", "Task failed")
if error_type == "ValueError":
raise HTTPException(status_code=413, detail=error_detail)
Comment on lines +39 to +40
Copy link
Copy Markdown
Contributor

@gemini-code-assist gemini-code-assist Bot Apr 28, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

medium

Mapping all ValueError exceptions to HTTP 413 (Payload Too Large) is too broad. A ValueError can be raised for various reasons other than input length (e.g., invalid parameter values or unsupported configurations). This might lead to misleading error codes for the client. It is recommended to use a more specific exception type for token length violations or check the error message to ensure the status code is appropriate.

raise HTTPException(status_code=500, detail=error_detail)

if status == TaskStatus.CANCELLED.value:
raise HTTPException(status_code=409, detail=task_status.get("error", "Task cancelled"))
Expand Down Expand Up @@ -108,6 +112,8 @@ async def create_image_task(message: ImageTaskRequest):
save_result_path=message.save_result_path,
)
except RuntimeError as e:
if getattr(e, "original_error_type", "") == "ValueError":
raise HTTPException(status_code=413, detail=str(e))
Comment on lines +115 to +116
Copy link
Copy Markdown
Contributor

@gemini-code-assist gemini-code-assist Bot Apr 28, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

medium

This check for ValueError in the RuntimeError handler appears to be unreachable for token length errors. Token length validation happens during inference in the worker process, and the resulting error is propagated via the task status, not as a direct exception from task_manager.create_task. This pattern is repeated in other task creation endpoints in this file.

raise HTTPException(status_code=503, detail=str(e))
except Exception as e:
logger.error(f"Failed to create image task: {e}")
Expand Down Expand Up @@ -168,6 +174,8 @@ async def create_image_task_sync(
raise

except RuntimeError as e:
if getattr(e, "original_error_type", "") == "ValueError":
raise HTTPException(status_code=413, detail=str(e))
raise HTTPException(status_code=503, detail=str(e))
except HTTPException:
raise
Expand Down Expand Up @@ -229,6 +237,8 @@ async def save_file_async(file: UploadFile, target_dir: Path) -> str:
save_result_path=message.save_result_path,
)
except RuntimeError as e:
if getattr(e, "original_error_type", "") == "ValueError":
raise HTTPException(status_code=413, detail=str(e))
raise HTTPException(status_code=503, detail=str(e))
except Exception as e:
logger.error(f"Failed to create image form task: {e}")
Expand Down
5 changes: 4 additions & 1 deletion lightx2v/server/services/generation/base.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -190,7 +190,10 @@ async def generate_with_stop_event(self, message: Any, stop_event) -> Optional[A
)
else:
error_msg = result.get("error", "Inference failed")
raise RuntimeError(error_msg)
error_type = result.get("error_type", "")
exc = RuntimeError(error_msg)
exc.original_error_type = error_type
raise exc

except Exception as e:
logger.exception(f"Task {message.task_id} processing failed: {str(e)}")
Expand Down
5 changes: 4 additions & 1 deletion lightx2v/server/services/generation/image.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -78,7 +78,10 @@ async def generate_with_stop_event(self, message: Any, stop_event) -> Optional[A
)
else:
error_msg = result.get("error", "Inference failed")
raise RuntimeError(error_msg)
error_type = result.get("error_type", "")
exc = RuntimeError(error_msg)
exc.original_error_type = error_type
raise exc

except Exception as e:
logger.exception(f"Task {message.task_id} processing failed: {str(e)}")
Expand Down
3 changes: 3 additions & 0 deletions lightx2v/server/services/inference/worker.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -68,6 +68,7 @@ def init(self, args) -> bool:
async def process_request(self, task_data: Dict[str, Any]) -> Dict[str, Any]:
has_error = False
error_msg = ""
error_type = ""
pipeline_return = None

try:
Expand Down Expand Up @@ -103,6 +104,7 @@ async def process_request(self, task_data: Dict[str, Any]) -> Dict[str, Any]:
except Exception as e:
has_error = True
error_msg = str(e)
error_type = type(e).__name__
logger.exception(f"Rank {self.rank} inference failed: {error_msg}")

if self.world_size > 1:
Expand All @@ -114,6 +116,7 @@ async def process_request(self, task_data: Dict[str, Any]) -> Dict[str, Any]:
"task_id": task_data.get("task_id", "unknown"),
"status": "failed",
"error": error_msg,
"error_type": error_type,
"message": f"Inference failed: {error_msg}",
}
else:
Expand Down
14 changes: 12 additions & 2 deletions lightx2v/server/task_manager.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -27,6 +27,7 @@ class TaskInfo:
start_time: datetime = field(default_factory=datetime.now)
end_time: Optional[datetime] = None
error: Optional[str] = None
error_type: Optional[str] = None
save_result_path: Optional[str] = None
result_png: Optional[bytes] = None
stop_event: threading.Event = field(default_factory=threading.Event)
Expand Down Expand Up @@ -97,7 +98,7 @@ def complete_task(self, task_id: str, save_result_path: Optional[str] = None, re
self.completed_tasks += 1
self._emit_queue_metrics_unlocked()

def fail_task(self, task_id: str, error: str):
def fail_task(self, task_id: str, error: str, error_type: Optional[str] = None):
with self._lock:
if task_id not in self._tasks:
logger.warning(f"Task {task_id} not found for failure")
Expand All @@ -107,6 +108,7 @@ def fail_task(self, task_id: str, error: str):
task.status = TaskStatus.FAILED
task.end_time = datetime.now()
task.error = error
task.error_type = error_type

self.failed_tasks += 1
self._emit_queue_metrics_unlocked()
Expand Down Expand Up @@ -154,7 +156,15 @@ def get_task_status(self, task_id: str) -> Optional[Dict[str, Any]]:
if not task:
return None

return {"task_id": task.task_id, "status": task.status.value, "start_time": task.start_time, "end_time": task.end_time, "error": task.error, "save_result_path": task.save_result_path}
return {
"task_id": task.task_id,
"status": task.status.value,
"start_time": task.start_time,
"end_time": task.end_time,
"error": task.error,
"error_type": task.error_type or "",
"save_result_path": task.save_result_path,
}

def get_all_tasks(self):
with self._lock:
Expand Down
Loading