NVIDIA Anti-Trust Enforcement#12
Closed
openSourcerer9000 wants to merge 2 commits intoaudiohacking:mainfrom
Closed
NVIDIA Anti-Trust Enforcement#12openSourcerer9000 wants to merge 2 commits intoaudiohacking:mainfrom
openSourcerer9000 wants to merge 2 commits intoaudiohacking:mainfrom
Conversation
Author
|
I got the same issue and set Qwen coder next loose on it for about an hour, and it seems to have fixed it. Someone should really take a look at the diffs though, it managed to break and subsequently fix the server along the way... |
There was a problem hiding this comment.
Pull request overview
Improves backend runtime compatibility on non-CUDA environments (MPS/CPU) and enhances failure observability via traceback logging.
Changes:
- Add a CUDA fallback monkey-patch for non-CUDA devices and disable SageAttention on MPS.
- Improve device selection logging when falling back to CPU.
- Switch several failure logs to include tracebacks (
logger.exception).
Reviewed changes
Copilot reviewed 4 out of 4 changed files in this pull request and generated 3 comments.
| File | Description |
|---|---|
| backend/ltx2_server.py | Adds CUDA fallback patching and disables SageAttention on MPS; logs CPU fallback. |
| backend/handlers/video_generation_handler.py | Adds traceback logging for i2v generation failures. |
| backend/handlers/queue_worker.py | Logs job failures with tracebacks. |
| backend/handlers/generation_handler.py | Logs generation failures with tracebacks during state transitions. |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
You can also share your feedback on Copilot code review. Take the survey.
Comment on lines
+49
to
+115
| def _setup_cuda_fallback() -> None: | ||
| """ | ||
| Monkey-patch torch.cuda functions to handle cases where PyTorch is not | ||
| compiled with CUDA support (e.g., running on MPS or CPU). | ||
|
|
||
| The ltx-pipelines library calls torch.cuda.synchronize() unconditionally, | ||
| which fails with "Torch not compiled with CUDA enabled" on non-CUDA builds. | ||
| """ | ||
| # Check if we're on a device that doesn't have full CUDA support | ||
| device_type = DEVICE.type | ||
|
|
||
| if device_type == "cuda": | ||
| # True CUDA - no fallback needed | ||
| return | ||
|
|
||
| logger.info(f"Setup CUDA fallback for device type: {device_type}") | ||
|
|
||
| # Create safe no-op implementations for CUDA functions | ||
| def safe_cuda_synchronize() -> None: | ||
| """No-op synchronize for non-CUDA devices.""" | ||
| if device_type == "mps": | ||
| try: | ||
| torch.mps.synchronize() | ||
| except Exception: | ||
| pass | ||
|
|
||
| def safe_cuda_empty_cache() -> None: | ||
| """No-op empty_cache for non-CUDA devices.""" | ||
| if device_type == "mps": | ||
| try: | ||
| torch.mps.empty_cache() | ||
| except Exception: | ||
| pass | ||
|
|
||
| def safe_cuda_memory_reserved() -> int: | ||
| """Return 0 for memory reserved on non-CUDA devices.""" | ||
| return 0 | ||
|
|
||
| def safe_cuda_memory_allocated() -> int: | ||
| """Return 0 for memory allocated on non-CUDA devices.""" | ||
| return 0 | ||
|
|
||
| def safe_cuda_get_device_name(device: object = None) -> str: | ||
| """Return device name for non-CUDA devices.""" | ||
| if device_type == "mps" and hasattr(torch, 'mps'): | ||
| return "Apple Silicon MPS" | ||
| return "CPU" | ||
|
|
||
| def safe_cuda_get_device_capability(device: object = None) -> tuple[int, int]: | ||
| """Return (0, 0) for non-CUDA devices.""" | ||
| return (0, 0) | ||
|
|
||
| # Patch torch.cuda module | ||
| if not hasattr(torch.cuda, "_ltx_original_synchronize"): | ||
| # Store original functions if they exist | ||
| try: | ||
| torch.cuda._ltx_original_synchronize = torch.cuda.synchronize # type: ignore[attr-defined] | ||
| except AttributeError: | ||
| pass | ||
|
|
||
| # Replace with safe implementations | ||
| torch.cuda.synchronize = safe_cuda_synchronize # type: ignore[assignment] | ||
| torch.cuda.empty_cache = safe_cuda_empty_cache # type: ignore[assignment] | ||
| torch.cuda.memory_reserved = safe_cuda_memory_reserved # type: ignore[assignment] | ||
| torch.cuda.memory_allocated = safe_cuda_memory_allocated # type: ignore[assignment] | ||
| torch.cuda.get_device_name = safe_cuda_get_device_name # type: ignore[assignment] | ||
| torch.cuda.get_device_capability = safe_cuda_get_device_capability # type: ignore[assignment] |
| logger.info("Generation cancelled by user") | ||
| return GenerateVideoResponse(status="cancelled") | ||
|
|
||
| logger.exception("[i2v] Generation failed with exception: %s", e) |
Comment on lines
+45
to
+48
| # ============================================================ | ||
| # CUDA Fallback Handling (for non-CUDA PyTorch builds on MPS/CPU) | ||
| # ============================================================ | ||
|
|
|
@openSourcerer9000 the _setup_cuda_fallback patch will definitely help but all the changes to logger.exception are causing test errors and are unnecessary. Could you kindly remove those from the PR and limit changes to the CUDA related patches? |
|
@openSourcerer9000 closed and recreated in #14 |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
No description provided.