[Frontend] Remove custom_cache_manager

[fulvius31]

Remove custom_cache_manager since vllm uses torch 2.5.1 which uses triton 3.1.0 that includes triton-lang/triton#4295

cc @tdoublep @jeejeelee

[github-actions]

👋 Hi! Thank you for contributing to the vLLM project.

💬 Join our developer Slack at https://slack.vllm.ai to discuss your PR in #pr-reviews, coordinate on features in #feat- channels, or join special interest groups in #sig- channels.

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 either: Add ready label to the PR or enable auto-merge.

🚀

[jeejeelee]

I am aware of this PR, but I haven't found this code in the triton/runtime/cache.py corresponding to 3.1.0, so I understand that the 3.1.0 triton might not include this PR yet.

[jeejeelee]

torch 2.6.0 will install triton 3.2.0, which looks include triton-lang/triton#4295. Therefore, we can consider this PR after #12721 is merged.

[fulvius31]

I am aware of this PR, but I haven't found this code in the triton/runtime/cache.py corresponding to 3.1.0, so I understand that the 3.1.0 triton might not include this PR yet.

You are totally right!
I faced the cache issue on 3.0.0 and then when I updated to 3.1.0 the issue was gone. Then I checked quickly (too quick) the cache.py code and I thought it was merged already.

Let's wait #12721 !