aws / aws/sagemaker-distribution
Fix keras tests for GPU based docker images
- Dominant language
- Shell
- Stars
- 154
- Forks
- 120
- Avg merge
- 5d 4h
- Merged PRs (30d)
- 27
Description
Issue: keras tests are failing when running it for GPU .
More background:
We hard code keras to a legacy commit: https://github.com/aws/sagemaker-distribution/blob/main/test/test_artifacts/v1/keras.test.Dockerfile#L19 due to TensorFlow version conflict issue, but with upgrading to new minor version.
After removing this line and rerun the test, test failed again complaining missing dependency like `jax`.
Acceptance Criteria: keras unit tests should pass.
### Image Tags
1.8.0-gpu
Contributor guide
Research direction
Start with test/test_artifacts/v1/keras.test.Dockerfile, especially the legacy Keras commit pin at line 19, and rerun the Keras tests for image tag 1.8.0-gpu. Investigate the TensorFlow version conflict and the reported missing jax dependency. Done means the Keras unit tests pass for the GPU image.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- docker, python, tensorflow
- Domain
- build-system, machine-learning, testing-qa
- Issue type
- Bug
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Stale
- Clarity
- Mostly clear
- Newbie friendliness
- 35/100