Check for batch_fn before calling close #8786
Conversation
When performing GEPA optimization using a pipeline that involves an embedder, the following error tends to occur:
```
Exception ignored in: <function Unbatchify.__del__ at 0x1263e0180>
Traceback (most recent call last):
File "/Users/zachbam/.local/share/mise/installs/python/3.12.11/lib/python3.12/site-packages/dspy/utils/unbatchify.py", line 112, in __del__
self.close()
File "/Users/zachbam/.local/share/mise/installs/python/3.12.11/lib/python3.12/site-packages/dspy/utils/unbatchify.py", line 92, in close
if not self.stop_event.is_set():
^^^^^^^^^^^^^^^
AttributeError: 'Unbatchify' object has no attribute 'stop_event'
```
The issue seems to stem from instances where deletion is called, but the object is already gone. My suggested fix resolves this issue.
There was a problem hiding this comment.
Pull Request Overview
This PR fixes an AttributeError that occurs during garbage collection of Unbatchify objects in GEPA optimization pipelines with embedders. The error happens when the destructor tries to access attributes that may not exist if the object was not fully initialized.
- Adds a safety check in the
__del__method to verify object initialization before cleanup - Prevents AttributeError exceptions during garbage collection of partially initialized objects
Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.
| Ensures the worker thread is terminated when the object is garbage collected. | ||
| """ | ||
| self.close() | ||
| if hasattr(self, "batch_fn"): |
There was a problem hiding this comment.
The fix checks for batch_fn but the error traceback shows the issue is with missing stop_event attribute. The check should verify the existence of stop_event instead, or check for multiple required attributes to ensure the object is fully initialized.
| if hasattr(self, "batch_fn"): | |
| if hasattr(self, "stop_event") and hasattr(self, "worker_thread"): |
Copilot uses AI. Check for mistakes.
When performing GEPA optimization using a pipeline that involves an embedder, the following error tends to occur:
The issue seems to stem from instances where deletion is called, but the object is already gone. My suggested fix resolves this issue.