ci: pass -Dcuda-arch on the CUDA cache-hit path - #2
Merged
Merged
Conversation
The arch list is part of the module fingerprint. The "Restore prebuilt CUDA module" step computes the asset name with -Dcuda-arch="$CUDA_ARCH", so the download matched, but the build step then invoked -Dcuda-prebuilt without it — recomputing the fingerprint with arch=ggml-default and rejecting the module it had just fetched: stored: arch=75-real;80-real;86-real;89-real;90-real;120-real;120-virtual current: arch=ggml-default Both Linux and Windows failed this way ~90s in, so the first run that could have used the cache never got to compile. Same in all three cache-hit sites. Claude-Session: https://claude.ai/code/session_01N2CCEhM8ewtAuHJceuWYAB
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
The arch list is part of the module fingerprint. The restore step computed the
asset name with
-Dcuda-arch="$CUDA_ARCH"so the download matched, but thebuild step then invoked
-Dcuda-prebuiltwithout it — recomputing thefingerprint with
arch=ggml-defaultand rejecting the module it had justfetched:
Both CUDA jobs died ~90s in on run 31346796779, so the first run that could
have used the cache never compiled anything.
Verified on run 31346997904 (all three green,
Install CUDA Toolkitskippedon both CUDA jobs):
https://claude.ai/code/session_01N2CCEhM8ewtAuHJceuWYAB