| name | mlx-fix-test-py |
| description | Build MLX, reproduce one failing Python test, and implement the smallest Vulkan-only fix. Use when asked to repair a Python test failure in this repository while preserving test code and avoiding unrelated fixes. |
Fix One Vulkan Python Test
Fix exactly one Python failure per invocation.
- Run
./dev.sh build.
- After the build completes, run
./dev.sh test-py -x.
- Identify the first failing test and reproduce only that failure as needed.
- Do not modify test code. Modify implementation only under
mlx/mlx/backend/vulkan/.
- Before editing, inspect corresponding CUDA and Metal implementations under
mlx/mlx/backend/cuda/ and mlx/mlx/backend/metal/.
- Make the smallest correct fix consistent with existing backend patterns and the async Vulkan pipeline policy.
- Run
./dev.sh build, then ./dev.sh test-py -x sequentially to verify the addressed failure. Continue only on the same failure until it passes or a clear blocker remains.
- After implementation changes, run
./dev.sh generate to validate inference output coherence.
Report the failure addressed, Vulkan files changed, reference backend files consulted, and final test and generation results. Do not fix later or unrelated failures.
