[assistant-librarian]

Bumps ROCm/rocm-libraries from ROCm/rocm-libraries@320d658 to ROCm/rocm-libraries@4606f83.

See full comparison here: ROCm/rocm-libraries@320d658...4606f83

[therock-pr-bot]

🤖 Bump PR detected (author @assistant-librarian[bot]). All policy checks are auto-approved for automated dependency bumps.

[dysu-amd]

gfx125X and gfx90a tests are non-blocking, they were just recently enabled in #8169

Linux gfx942 rocJPEG has 24/25 failing tests, many of which are segfaults. Looking into this to see if this could be the same underlying issue as #8265 (tldr: non-CSP pool may be bad, thus is an infra issue and not a bump issue)

[dysu-amd]

gfx125x and gfx90a are nonblocking: #8309

Followed same steps as in previous bump to rerun rocJPEG tests on a CSP gfx942 runner, which passed: https://github.com/ROCm/TheRock/actions/runs/35258423260/job/105329183157

This test was a manual dispatch of multi_arch_ci.yml with following inputs:

branch: bump-rocm-libraries-4606f83
linux gpu families: gfx94X
linux test labels: test:rocjpeg

note: test_filter:standard should have also been added as a test label (otherwise defaults to the full test suite), but in rocjpeg's case the diff between full and standard is a single extra test, so it's more or less the same)

prebuilt stages: all
baseline run ID: 35193881163
build python packages: false
build pytorch: false
build jax: false
build native linux: false
All other params were default/empty.

[dysu-amd]

created rocjpeg issue: #8310