ROCm v6.2.4

[GZGavinZhao]

• rocm-llvm: Update to v6.2.4
• hsakmt: Update to v6.2.4
• rocm-cmake: Update to v6.2.4
• rocm-device-libs: Update to v6.2.4
• rocm-compilersupport: Update to v6.2.4
• fixup! rocm-compilersupport: Update to v6.2.4
• rocm-runtime: Update to v6.2.4
• hsa-amd-aqlprofile: update to v6.2.0
• rocm-core: Update to v6.2.4
• rocprofiler-register: Add at v6.2.4

Summary

Updated the ROCm stack to v6.2.4 and PyTorch to v2.5, bringing performance and compatibility improvements to ML workflows on Solus.

Test Plan

• Stable Diffusion works on gfx1032
• Stable Diffusion works on gfx900
• Anime4KCPP works on gfx1032 (OpenCL)
• Anime4KCPP works on gfx90c (OpenCL)
• Anime4KCPP works on gfx900 (OpenCL)
• llama.cpp works on gfx1032
• llama.cpp works on gfx90c
• llama.cpp works on gfx900
• gfx1030 code runs on gfx1032
• gfx900 code runs on gfx90c
• pytorch/examples passed on gfx1032 (upstream issue [ROCm] PyTorch multiprocess fails to create memory with IPC pytorch/pytorch#143728)
• pytorch/examples passed on gfx90c (upstream issue [ROCm] PyTorch multiprocess fails to create memory with IPC pytorch/pytorch#143728)
• pytorch/examples passed on gfx900 (upstream issue [ROCm] PyTorch multiprocess fails to create memory with IPC pytorch/pytorch#143728)

Checklist

• Package was built and tested against unstable
• This change could gainfully be listed in the weekly sync notes once merged (Write an appropriate message in the Summary section)

[GZGavinZhao]

Everything has been packaged. Just tests pending (and fixing broken tests)

[GZGavinZhao]

Last rebuild of the entire stack to make sure everything builds in the expected order. Will merge after this finishes.