fix: pin ForwardDiff and CUDA versions

[avik-pal]

I am going to pin these till upstream PRs get merged. Can't have CI red unnecessarily.

1. fix: regression in non-fast scalar indexing support JuliaDiff/ForwardDiff.jl#760
2. fix: NCCL import of @checked + support for CUDA 13 JuliaGPU/NCCL.jl#63

[gemini-code-assist]

Summary of Changes

Hello @avik-pal, I'm Gemini Code Assist¹! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed!

This pull request addresses recent continuous integration failures by temporarily pinning specific versions of the ForwardDiff and CUDA dependencies. This proactive measure aims to stabilize the build process and prevent unnecessary CI red states until critical upstream pull requests are merged and integrated, ensuring the project remains functional and reliable.

Highlights

• Dependency Version Pinning: The ForwardDiff package is now explicitly pinned to major version 1 in Project.toml to prevent potential breaking changes from newer versions.
• CUDA Compatibility Range: The CUDA package compatibility in lib/LuxCUDA/Project.toml has been narrowed to 5.8 - 5.8.1 to ensure stability and avoid CI failures related to upstream changes.
• Project Version Bumps: The main Lux project version has been updated to 1.21.2, and the LuxCUDA sub-project version has been updated to 0.3.5.

Using Gemini Code Assist

The full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips.

Invoking Gemini

You can request assistance from Gemini at any point in your pull request via creating an issue comment (i.e. comment on the pull request page) using either /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands.

Feature	Command	Description
Code Review	/gemini review	Performs a code review for the current pull request in its current state.
Pull Request Summary	/gemini summary	Provides a summary of the current pull request in its current state.
Comment	@gemini-code-assist	Responds in comments when explicitly tagged, both in issue comments and review comments.
Help	/gemini help	Displays a list of available commands.

Customization

To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a .gemini/ folder in the base of the repository. Detailed instructions can be found here.

Limitations & Feedback

Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 on @gemini-code-assist comments. If you're interested in giving your feedback about your experience with Gemini Code Assist for Github and other Google products, sign up here.

You can also get AI-powered code generation, chat, as well as code reviews directly in the IDE at no cost with the Gemini Code Assist IDE Extension.

Footnotes

1. Review the Privacy Notices, Generative AI Prohibited Use Policy, Terms of Service, and learn how to configure Gemini Code Assist in GitHub here. Gemini can make mistakes, so double check it and use code with caution. ↩

[gemini-code-assist]

Code Review

This pull request pins the versions of ForwardDiff and CUDA to work around issues in upstream packages and keep CI green. The version numbers for Lux and LuxCUDA are also bumped.

The change to pin the CUDA version in lib/LuxCUDA/Project.toml is correct and effectively constrains the version as intended. However, the change to the ForwardDiff compatibility in the root Project.toml uses a syntax that results in an empty set of allowed versions, which will likely break package installation for users. I've provided a critical comment with a suggested fix for this issue. It's worth noting that the same syntactical problem might exist for other dependencies in that file.

[gemini-code-assist]

The current syntax for the [compat] entry specifies an empty set of allowed versions for ForwardDiff, which will likely cause package resolution to fail.

According to Julia's Pkg documentation, comma-separated version specifiers are treated as an intersection.

• "0.10.36" resolves to the range [0.10.36, 0.11.0).
• "=1" resolves to the range [1.0.0, 1.1.0).

The intersection of these two ranges is empty. The previous value ("0.10.36, 1") had the same problem.

To pin the version to avoid issues with v1.0 (as suggested by the linked PR), you should probably restrict compatibility to the 0.10 series. The suggested change does this.

Please note that other dependencies in this file (e.g., Zygote, LossFunctions) seem to have the same syntactical issue.

Suggested change

	ForwardDiff = "0.10.36, =1"
	ForwardDiff = "0.10.36"

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 69.85%. Comparing base (c28178f) to head (fb0a3c2).
⚠️ Report is 2 commits behind head on main.

❗ There is a different number of reports uploaded between BASE (c28178f) and HEAD (fb0a3c2). Click for more details.

HEAD has 31 uploads less than BASE

Flag	BASE (c28178f)	HEAD (fb0a3c2)
	52	21

Additional details and impacted files

@@             Coverage Diff             @@
##             main    #1477       +/-   ##
===========================================
- Coverage   84.13%   69.85%   -14.28%     
===========================================
  Files         166      166               
  Lines        6724     6711       -13     
===========================================
- Hits         5657     4688      -969     
- Misses       1067     2023      +956     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[github-actions]

Benchmark Results (ASV)

	main	fb0a3c2...	main / fb0a3c2...
basics/overhead	0.137 ± 0.003 μs	0.148 ± 0.007 μs	0.924 ± 0.048
time_to_load	0.945 ± 0.014 s	0.95 ± 0.003 s	0.995 ± 0.015

Benchmark Plots

A plot of the benchmark results have been uploaded as an artifact to the workflow run for this PR.
Go to "Actions"->"Benchmark a pull request"->[the most recent run]->"Artifacts" (at the bottom).