Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Replace unsafe_convert with pointer #564

Merged
merged 1 commit into from
Feb 20, 2024

Conversation

mcabbott
Copy link
Member

Closes #563, perhaps

@ToucheSir
Copy link
Member

AMDGPU.jl compat issues on nightly are blocking us from seeing what CI thinks about this. What happens if you trim AMDGPU from Project.toml and re-run here or locally?

@mcabbott
Copy link
Member Author

Locally it passes, without AMDGPU, provided I relax isapprox tolerance on two batched_mul tests.

@pxl-th
Copy link
Member

pxl-th commented Feb 20, 2024

AMDGPU.jl compat issues on nightly are blocking us

Should no longer happen with AMDGPU 0.8.8, but I think nightly currently requires JuliaGPU/GPUCompiler.jl#546 to work.

@ToucheSir ToucheSir closed this Feb 20, 2024
@ToucheSir ToucheSir reopened this Feb 20, 2024
Copy link
Member

@ToucheSir ToucheSir left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

CI is complaining about CUDA stuff now instead, but we should just land this and revise if necessary. The main follow-up I see is whether the accuracy changes on CI or not, because it seems to still be fine on stable Julia versions.

Side note, we should look into removing the GPU packages from the test environment when working with GHA CI. All they do right now is cause precompile latency and errors on nightly.

@mcabbott mcabbott merged commit 9c31d9d into FluxML:master Feb 20, 2024
21 of 26 checks passed
@mcabbott mcabbott deleted the convert_pointer branch February 20, 2024 18:26
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

batched_mul broken on Julia nightly
3 participants