Update amd-installation.md

[johnnynunez]

rocm 6.3

[github-actions]

👋 Hi! Thank you for contributing to the vLLM project.
Just a reminder: PRs would not trigger full CI run by default. Instead, it would only run fastcheck CI which starts running only a small and essential subset of CI tests to quickly catch errors. You can run other CI tests on top of those by going to your fastcheck build on Buildkite UI (linked in the PR checks section) and unblock them. If you do not have permission to unblock, ping simon-mo or khluu to add you in our Buildkite org.

Once the PR is approved and ready to go, your PR reviewer(s) can run CI to test the changes comprehensively before merging.

To run CI, PR reviewers can do one of these:

• Add ready label to the PR
• Enable auto-merge.

🚀

[DarkLight1337]

We still use ROCm 6.2 in the Dockerfile.rocm, could you elaborate on why the docs need to be changed?

[johnnynunez]

We still use ROCm 6.2 in the Dockerfile.rocm, could you elaborate on why the docs need to be changed?

I'm updating (still not docker). It is for new hardware device and improve a lot in flash attention
https://www.amd.com/es/products/accelerators/instinct/mi300/mi325x.html

flash attention performance:
https://community.amd.com/t5/ai/unlocking-new-horizons-in-ai-and-hpc-with-the-release-of-amd/ba-p/726434
FlashAttention-2 optimized for ROCm 6.3 is talking up to 3x speed-ups and other benefits.

Pytorch:
6.3.0 ROCM is on CI nightly from this week

[johnnynunez]

rocm6.3.1 and pytorch 2.5 is coming also

[mergify]

This pull request has merge conflicts that must be resolved before it can be
merged. Please rebase the PR, @johnnynunez.

https://docs.github.com/en/pull-requests/collaborating-with-pull-requests/working-with-forks/syncing-a-fork

[DarkLight1337]

Is this superseded by #11777?