Fix V100 CUDA compatibility for demeter4 runners#199
Open
ChrisRackauckas-Claude wants to merge 9 commits intoSciML:mainfrom
Open
Fix V100 CUDA compatibility for demeter4 runners#199ChrisRackauckas-Claude wants to merge 9 commits intoSciML:mainfrom
ChrisRackauckas-Claude wants to merge 9 commits intoSciML:mainfrom
Conversation
Add LocalPreferences.toml to pin CUDA runtime 12.6 and disable forward-compat driver. V100 GPUs (compute capability 7.0) require system driver since CUDA_Driver_jll v13+ drops cc7.0 support. Ref: ChrisRackauckas/InternalJunk#19
Move LocalPreferences.toml from test/ to root so Pkg.test() picks up CUDA 12.6 pinning for V100 compatibility. Add JULIA_CUDA_VERSION and JULIA_CUDA_USE_COMPAT env vars in CI as backup. Add warnonly for example_block in docs to handle pre-existing upstream Zygote/ChainRulesCore gradient errors. Co-Authored-By: Chris Rackauckas <accounts@chrisrackauckas.com> Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Co-Authored-By: Chris Rackauckas <accounts@chrisrackauckas.com> Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
CUDA_Runtime_jll and CUDA_Driver_jll need to be direct test dependencies so Pkg.test() properly propagates LocalPreferences.toml to the temp test environment. Remove deprecated JULIA_CUDA_VERSION env vars and unnecessary docs Preferences step. Co-Authored-By: Chris Rackauckas <accounts@chrisrackauckas.com> Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Fix Aqua.test_deps_compat failure by adding compat entries for CUDA_Driver_jll and CUDA_Runtime_jll. Add nvidia-smi step to diagnose GPU memory issues on runners. Co-Authored-By: Chris Rackauckas <accounts@chrisrackauckas.com> Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Contributor
Author
CI Status UpdatePassing (7/8 non-skipped):
Failing:
CUDA GPU Test Failure AnalysisThe nvidia-smi on arctic1 shows: 7029MiB / 15360MiB already used by other processes, leaving only ~8GB free. Two pre-existing issues on the T4 runner:
These failures also occur on What this PR does:
The V100 compat fix will be verifiable once demeter4's driver is repaired and the |
Match DiffEqGPU.jl pattern: CUDA tests on gpu-t4 (arctic1, T4 16GB) and documentation on gpu-v100 (demeter4, V100 32GB). The generic 'gpu' label caused tests to land on congested runners with OOM. Co-Authored-By: Chris Rackauckas <accounts@chrisrackauckas.com> Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
arctic1 T4 (15GB) is shared by 16 runners and consistently has <500MB free from other Julia CI processes. Use gpu-v100 (demeter4, V100 32GB) for both CUDA tests and docs, matching the V100 compat focus of this PR. Co-Authored-By: Chris Rackauckas <accounts@chrisrackauckas.com> Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Contributor
Author
Final CI Status (commit f69d6c4 → 2f5fc8e)All non-GPU checks: PASS ✅
CUDA GPU Tests: PARTIAL ✅/❌
Runner label fixSwitched from generic
V100 CUDA 12.6 pinning: VERIFIED ✓nvidia-smi on demeter4-2 shows
Remaining upstream issues (pre-existing, not caused by this PR)
Both are CUDA-specific gradient bugs; forward passes work. CPU gradient tests all pass. These failures exist in the current package versions independent of this PR. |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Adds
LocalPreferences.tomlto pin CUDA runtime 12.6 and disable forward-compat driver for V100 GPU compatibility on demeter4 self-hosted runners.Changes
docs/LocalPreferences.toml: Pin CUDA_Runtime_jll to 12.6 and set CUDA_Driver_jll compat="false" for documentation buildstest/LocalPreferences.toml: Same configuration for GPU testsdocs/Project.toml: Add CUDA_Driver_jll and CUDA_Runtime_jll depsBackground
V100 GPUs (compute capability 7.0) require the system driver since CUDA_Driver_jll v13+ drops cc7.0 support. This matches the pattern established in OrdinaryDiffEq.jl#3162.
Ref: ChrisRackauckas/InternalJunk#19