re-triage test_nn #653
Comments
github-merge-queue bot
pushed a commit
that referenced
this issue
Jul 31, 2024
…ompatible for XPU (#665) This fixes issues from #653 The case is verifying an expected assertion log raised in kernel. We have different log in XPU backend, different keyword (we have no 'CUDA'), different index (we are using work item). Original test case uses CUDA specific calls, and expected assert message returned from CUDA. The test cases are now rewritten in `test_nn_xpu.py` to adapt to XPU specific counterparts. - "test_cross_entropy_loss_2d_out_of_bounds_class_index_xpu_float16", - "test_cross_entropy_loss_2d_out_of_bounds_class_index_xpu_float32",
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
🐛 Describe the bug
#258
these cases are target 2.5
LossNLL2d has no correct assert
"test_cross_entropy_loss_2d_out_of_bounds_class_index_xpu_float16",
"test_cross_entropy_loss_2d_out_of_bounds_class_index_xpu_float32", - Jianghang
native_group_norm : RuntimeError: Expected X.is_contiguous(memory_format) to be true, but got false. (Could this error message be improved? If so, please report an enhancement request to PyTorch.)
"test_GroupNorm_memory_format_xpu", - Resolve the memory format issue of GroupNorm #677
upsamplingNearest2d: Failed: Unexpected success
"test_upsamplingNearest2d_launch_fail_xpu", - grid size check
do not aline with cuda fix cuda bias code in test_nn #656 fix it
"test_upsamplingBiMode2d_consistency",
"test_upsamplingBiLinear2d_consistency_interp_size_bug",
cause by cuda hard code
"test_device_mask_xpu", fix cuda bias code in test_nn #656 fix it
"test_overwrite_module_params_on_conversion_cpu_device_xpu",
(https://github.com/pytorch/pytorch/blob/1fb498d6e34e0e9b43b2c26dc0a18a4fc3a52605/aten/src/ATen/native/cuda/UpSampleNearest2d.cu#L303) is specially for cuda, keep in skip list
Versions
main
The text was updated successfully, but these errors were encountered: