[mypy] Enable mypy type checking for vllm/core

[jberkhahn]

Partially address #3680

Hopefully I did this right.

[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 consists a small and essential subset of CI tests to quickly catch errors. You can run other CI tests on top of default ones by unblocking the steps in your fast-check build on Buildkite UI.

Once the PR is approved and ready to go, please make sure to run full CI as it is required to merge (or just use auto-merge).

To run full CI, you can do one of these:

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

🚀

[jberkhahn]

so for the mypy stuff, I think the failure is because it's not picking up the changes in the ci config files? the yapf formatting stuff I have no idea, it looks like a bunch of stuff in the repo doesn't adhere to that. I added the formatting changes just for vllm/core, but that doesn't seem to have satisfied it.

[DarkLight1337]

so for the mypy stuff, I think the failure is because it's not picking up the changes in the ci config files? the yapf formatting stuff I have no idea, it looks like a bunch of stuff in the repo doesn't adhere to that. I added the formatting changes just for vllm/core, but that doesn't seem to have satisfied it.

format.sh has recently been updated to skip all imports when run locally because otherwise it takes too long to run. You can manually run mypy (without --follow-imports=skip) to perform the stricter checks locally.

[jberkhahn]

so for the mypy stuff, I think the failure is because it's not picking up the changes in the ci config files? the yapf formatting stuff I have no idea, it looks like a bunch of stuff in the repo doesn't adhere to that. I added the formatting changes just for vllm/core, but that doesn't seem to have satisfied it.

format.sh has recently been updated to skip all imports when run locally because otherwise it takes too long to run. You can manually run mypy (without --follow-imports=skip) to perform the stricter checks locally.

Right, but when I run them locally with just mypy vllm/core they pass. Also, the errors in CI refer to line numbers that don't match up with my edited files, so I think what's wrong is something with how CI is configured.

I'll take a look at the suggested changes. Thanks for the review!

[jberkhahn]

Think this is good to go now?

[DarkLight1337]

I don't think this is required? Since num_prompt_blocks = 0 if seq is None anyway

[jberkhahn]

The issue is seq is optional and directly referenced on lines 328-330:

                block = self.gpu_allocator.allocate(
                    seq.hash_of_block(logical_idx),
                    seq.num_hashed_tokens_of_block(logical_idx))

It's not technically possible to get there if seq is none but mypy is still complaining because it doesn't care about the check in _get_seq_num_required_blocks(seq). Do you have a more straightforward suggestion?

[DarkLight1337]

I see, thanks for the explanation.

[njhill]

@DarkLight1337 this one good with you to merge now?

[DarkLight1337]

@DarkLight1337 this one good with you to merge now?

Yeah, let's merge this