aws / aws/amazon-sagemaker-examples
Horovod script mode - incorrect weblink, unclear usage of --model_dir
- Dominant language
- Jupyter Notebook
- Stars
- 11k
- Forks
- 7k
- Avg merge
- 8h 29m
- Merged PRs (30d)
- 8
Description
1. The link https://github.com/aws/sagemaker-containers/blob/master/README.md in "More details can be found here" returns '404'
2. It is unclear from the notebook that SageMaker container invokes mnist_hvd.py with command-line agruments and that --model_dir is one of them. Need a detailed explanation what comman-line agrs are used and how to modify/use them should customer were to use this example as a foundation for his/her code.
3. the model is saved on a local machine, which is fine for 'local' mode, but does not help when using a SageMaker training instance. Could you, please, add code that would save the final model to the session's default s3 bucket?
Contributor guide
Research direction
Start with the Horovod script-mode notebook and the referenced mnist_hvd.py entry point; inspect how the SageMaker container passes command-line arguments, especially --model_dir. Verify the README link, document the arguments and how users can modify them, and trace model output in local versus training-instance mode so completion includes saving the final model to the session's default S3 bucket.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- aws, jupyter-notebook, python
- Domain
- cloud, documentation, machine-learning
- Issue type
- Documentation
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Stale
- Clarity
- Mostly clear
- Newbie friendliness
- 25/100