Overview
As a follow-up to #7854, extend the cuml.health_checks module to collect and present standardized environment information. This would allow users to simply run python -m cuml.health_checks and include the output in bug reports, replacing the verbose manual environment checklist.
Motivation
Currently, the bug report template asks contributors to manually collect:
- Environment location (bare-metal, Docker, Cloud)
- Linux distro/architecture
- GPU model and driver version
- CUDA version
- Method of cuML install (conda, Docker, source) with version details
This is a barrier to filing good bug reports. A single command that gathers this information in a standardized, well-formatted way would lower that barrier significantly.
Proposed Work
- Add an environment information collector to
cuml.health_checks that gathers:
- CUDA version and compatibility with installed cuML
- GPU model and driver version
- Linux distro and architecture
- cuML version and installation method
- Other relevant RAPIDS component versions
- Surface known incompatible or broken version combinations as actionable warnings
- Update the bug report template to reference
python -m cuml.health_checks instead of the manual checklist
References
Requested by @csadorf in #7854
Overview
As a follow-up to #7854, extend the
cuml.health_checksmodule to collect and present standardized environment information. This would allow users to simply runpython -m cuml.health_checksand include the output in bug reports, replacing the verbose manual environment checklist.Motivation
Currently, the bug report template asks contributors to manually collect:
This is a barrier to filing good bug reports. A single command that gathers this information in a standardized, well-formatted way would lower that barrier significantly.
Proposed Work
cuml.health_checksthat gathers:python -m cuml.health_checksinstead of the manual checklistReferences
rapids doctor#7851Requested by @csadorf in #7854