gemini-code-assist[bot] commented on code in PR #19777:
URL: https://github.com/apache/tvm/pull/19777#discussion_r3411453924
##########
tests/scripts/task_python_integration_gpuonly.sh:
##########
@@ -19,7 +19,9 @@
set -exo pipefail
export
TVM_TEST_TARGETS='cuda;opencl;metal;rocm;nvptx;{"kind":"opencl","device":"mali,adreno"}'
-export PYTEST_ADDOPTS="-m gpu $PYTEST_ADDOPTS"
+# Select GPU tests by any GPU-family marker (the umbrella "gpu" marker is no
+# longer auto-applied via parent inheritance).
+export PYTEST_ADDOPTS="-m 'gpu or cuda or rocm or metal or vulkan or opencl or
nvptx or tensorcore or cudnn or cublas or nccl or cudagraph or nvshmem or
matrixcore or hipblas or adreno_opencl_vulkan or adreno_vulkan or adreno_opencl
or adreno_clml' $PYTEST_ADDOPTS"
Review Comment:
The markers `multi_gpu`, `OpenCLML`, and `adreno_opencl_real` are missing
from the `-m` filter. Since the umbrella `gpu` marker is no longer auto-applied
via parent inheritance, any tests marked solely with these markers will be
skipped in the GPU-only integration tests. They should be added to the filter
list.
```suggestion
export PYTEST_ADDOPTS="-m 'gpu or cuda or rocm or metal or vulkan or opencl
or nvptx or tensorcore or cudnn or cublas or nccl or cudagraph or nvshmem or
matrixcore or hipblas or adreno_opencl_vulkan or adreno_vulkan or adreno_opencl
or adreno_clml or adreno_opencl_real or OpenCLML or multi_gpu' $PYTEST_ADDOPTS"
```
##########
tests/scripts/task_python_unittest_gpuonly.sh:
##########
@@ -18,7 +18,10 @@
set -euxo pipefail
-export PYTEST_ADDOPTS="-m gpu ${PYTEST_ADDOPTS:-}"
+# Select GPU tests by any GPU-family marker. The umbrella "gpu" marker is no
+# longer auto-applied via parent inheritance, so enumerate every marker that
+# previously inherited it (cuda/rocm backends and their libraries).
+export PYTEST_ADDOPTS="-m 'gpu or cuda or rocm or metal or vulkan or opencl or
nvptx or tensorcore or cudnn or cublas or nccl or cudagraph or nvshmem or
matrixcore or hipblas or adreno_opencl_vulkan or adreno_vulkan or adreno_opencl
or adreno_clml' ${PYTEST_ADDOPTS:-}"
Review Comment:
The markers `multi_gpu`, `OpenCLML`, and `adreno_opencl_real` are missing
from the `-m` filter. Since the umbrella `gpu` marker is no longer auto-applied
via parent inheritance, any tests marked solely with these markers will be
skipped in the GPU-only unit tests. They should be added to the filter list.
```suggestion
export PYTEST_ADDOPTS="-m 'gpu or cuda or rocm or metal or vulkan or opencl
or nvptx or tensorcore or cudnn or cublas or nccl or cudagraph or nvshmem or
matrixcore or hipblas or adreno_opencl_vulkan or adreno_vulkan or adreno_opencl
or adreno_clml or adreno_opencl_real or OpenCLML or multi_gpu'
${PYTEST_ADDOPTS:-}"
```
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
