Repository navigation
Fix packaged cuDNN loading and prepare qualified v0.1.5 release - #11
Merged
Merged
Conversation
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Transformer Engine could not find the cuDNN shared libraries installed by the NVIDIA wheel in a fresh GPU environment, so v0.1.4 qualification correctly blocked publication. Preload the matching CUDA-major distribution's cuDNN9 libraries before importing Transformer Engine, retain their handles, and handle dependency ordering with bounded retries. Doctor, model loading, ops and Userbuffers use the same loader.
The release now checks the installed portable package's native runtime import on CPU before allocating GPUs. Prepare v0.1.5 and document v0.1.4 as unpublished without modifying its tag.
Validation: patched production loader imports Transformer Engine and its ops in the isolated Modal release image with no PYTHONPATH or library-path workaround; four regression tests cover dependency ordering, repeat loading, failure/retry, CPU and missing-distribution behavior. Repository/configuration checks, lock validation, Ruff, actionlint and checkpoint resume pass. GPU doctor, training smoke and correctness qualification remain mandatory before publication.