chore: deduplicate CuPy installation in Dockerfiles

[yxf0314]

No description provided.

[gemini-code-assist]

Code Review

This pull request introduces a step to deduplicate CuPy in both pack/cuda/Dockerfile and pack/cuda/Dockerfile.vllm by purging conflicting cupy-cuda12x and cupy-cuda13x packages and reinstalling the correct variant matching the CUDA major version. The review feedback correctly identifies a potential build failure: because the shell runs with pipefail and errexit, if uv pip show fails when the package is not already installed, the command substitution will abort the Docker build. The reviewer suggests appending || true inside the command substitution to prevent this issue.

Important

The consumer version of Gemini Code Assist on GitHub is being sunset. Starting June 18, 2026, new organization installations will be blocked, and all code review activity will officially cease on July 17, 2026.
For more details on the timeline and next steps, please review the Help Documentation.

[Copilot]

Pull request overview

This PR adds a CuPy “deduplication” step to the CUDA Docker images to avoid conflicts where multiple cupy-cuda* distributions (e.g., cupy-cuda12x and cupy-cuda13x) are installed simultaneously and overwrite the same cupy/ module tree.

Changes:

• Add a new Docker build step that uninstalls cupy-cuda12x/cupy-cuda13x, removes the shared cupy/ directories, and reinstalls the CUDA-major-appropriate CuPy variant.
• Apply the same CuPy dedupe logic to both the main CUDA image Dockerfile and the vLLM-based Dockerfile.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 2 comments.

File	Description
pack/cuda/Dockerfile.vllm	Adds a CuPy deduplication RUN block before postprocessing.
pack/cuda/Dockerfile	Adds a CuPy deduplication RUN block before postprocessing.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[aiwantaozi]

The pack/cuda/Dockerfile (non-.vllm) copy is dead code, prefer no update it.

[aiwantaozi]

LGTM