Skip to content

fix: hyperparameter tuning with spot instances and checkpoints #1015

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Merged
merged 5 commits into from
Sep 4, 2019
Merged

fix: hyperparameter tuning with spot instances and checkpoints #1015

merged 5 commits into from
Sep 4, 2019
+74 −1

Conversation

speg03
Copy link
Contributor

@speg03 speg03 commented Sep 1, 2019

Issue #, if available:

#1011

Description of changes:

Add parameters (train_use_spot_instances, checkpoint_s3_uri, and checkpoint_local_path) to the request of hyperparameter tuning jobs as it did for training jobs in #990. These parameters are from an estimator.

Merge Checklist

Put an x in the boxes that apply. You can also fill these out after creating the PR. If you're unsure about any of them, don't hesitate to ask. We're here to help! This is simply a reminder of what we are going to look for before merging your pull request.

  • I have read the CONTRIBUTING doc
  • I used the commit message format described in CONTRIBUTING
  • I have added tests that prove my fix is effective or that my feature works (if appropriate)
  • I have updated any necessary documentation (if appropriate)

By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license.

@speg03
Copy link
Contributor Author

speg03 commented Sep 1, 2019 •
edited
Loading

I notice a flake8 error that there are too many arguments for the session.tune method. What should I do?

@sagemaker-bot
Copy link
Collaborator

AWS CodeBuild CI Report

  • Result: FAILED
  • Build Logs (available for 30 days)

Powered by github-codebuild-logs, available on the AWS Serverless Application Repository

@sagemaker-bot
Copy link
Collaborator

AWS CodeBuild CI Report

  • Result: SUCCEEDED
  • Build Logs (available for 30 days)

Powered by github-codebuild-logs, available on the AWS Serverless Application Repository

@sagemaker-bot
Copy link
Collaborator

AWS CodeBuild CI Report

  • Result: SUCCEEDED
  • Build Logs (available for 30 days)

Powered by github-codebuild-logs, available on the AWS Serverless Application Repository

@sagemaker-bot
Copy link
Collaborator

AWS CodeBuild CI Report

  • Result: SUCCEEDED
  • Build Logs (available for 30 days)

Powered by github-codebuild-logs, available on the AWS Serverless Application Repository

@sagemaker-bot
Copy link
Collaborator

AWS CodeBuild CI Report

  • Result: FAILED
  • Build Logs (available for 30 days)

Powered by github-codebuild-logs, available on the AWS Serverless Application Repository

@sagemaker-bot
Copy link
Collaborator

AWS CodeBuild CI Report

  • Result: SUCCEEDED
  • Build Logs (available for 30 days)

Powered by github-codebuild-logs, available on the AWS Serverless Application Repository

@speg03
Copy link
Contributor Author

speg03 commented Sep 3, 2019

I notice a flake8 error that there are too many arguments for the session.tune method. What should I do?

I misunderstood. The cause of the error is not too many arguments, but too complex. I fixed to suppress the error in the same way as session.train method does.

@sagemaker-bot
Copy link
Collaborator

AWS CodeBuild CI Report

  • Result: SUCCEEDED
  • Build Logs (available for 30 days)

Powered by github-codebuild-logs, available on the AWS Serverless Application Repository

@sagemaker-bot
Copy link
Collaborator

AWS CodeBuild CI Report

  • Result: SUCCEEDED
  • Build Logs (available for 30 days)

Powered by github-codebuild-logs, available on the AWS Serverless Application Repository

@sagemaker-bot
Copy link
Collaborator

AWS CodeBuild CI Report

  • Result: SUCCEEDED
  • Build Logs (available for 30 days)

Powered by github-codebuild-logs, available on the AWS Serverless Application Repository

@sagemaker-bot
Copy link
Collaborator

AWS CodeBuild CI Report

  • Result: SUCCEEDED
  • Build Logs (available for 30 days)

Powered by github-codebuild-logs, available on the AWS Serverless Application Repository

@sagemaker-bot
Copy link
Collaborator

AWS CodeBuild CI Report

  • Result: SUCCEEDED
  • Build Logs (available for 30 days)

Powered by github-codebuild-logs, available on the AWS Serverless Application Repository

ChoiByungWook
ChoiByungWook reviewed Sep 3, 2019
View reviewed changes
Copy link
Contributor

@ChoiByungWook ChoiByungWook left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The lack of integration test proving this change works worries me, however the precedent didn't seem to do it either, as shown here: #990

Can you comment confirming that your change works as intended?

src/sagemaker/session.py
@@ -450,6 +450,9 @@ def tune(
early_stopping_type="Off",
encrypt_inter_container_traffic=False,
vpc_config=None,
train_use_spot_instances=False,
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Should we allow the user to provide a train_max_wait similar to the Estimator call?

sagemaker-python-sdk/src/sagemaker/estimator.py

Line 86 in b33d1a2

train_max_wait=None,

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

We don't need to pass train_max_wait to HyperparameterTuner. It has already been passed as stop_condition from the estimator of a HyperparameterTuner instance.

sagemaker-python-sdk/src/sagemaker/job.py

Line 85 in b33d1a2

estimator.train_max_run, estimator.train_max_wait

This is called from here:

sagemaker-python-sdk/src/sagemaker/tuner.py

Line 853 in b33d1a2

config = _Job._load_config(inputs, tuner.estimator)

@speg03
Copy link
Contributor Author

speg03 commented Sep 3, 2019

Hi @ChoiByungWook, thanks for your comment.

Can you comment confirming that your change works as intended?

In my several tuning jobs, I have confirmed that this change is working correctly.

@sagemaker-bot
Copy link
Collaborator

AWS CodeBuild CI Report

  • Result: SUCCEEDED
  • Build Logs (available for 30 days)

Powered by github-codebuild-logs, available on the AWS Serverless Application Repository

@sagemaker-bot
Copy link
Collaborator

AWS CodeBuild CI Report

  • Result: SUCCEEDED
  • Build Logs (available for 30 days)

Powered by github-codebuild-logs, available on the AWS Serverless Application Repository

@sagemaker-bot
Copy link
Collaborator

AWS CodeBuild CI Report

  • Result: SUCCEEDED
  • Build Logs (available for 30 days)

Powered by github-codebuild-logs, available on the AWS Serverless Application Repository

@sagemaker-bot
Copy link
Collaborator

AWS CodeBuild CI Report

  • Result: SUCCEEDED
  • Build Logs (available for 30 days)

Powered by github-codebuild-logs, available on the AWS Serverless Application Repository

@sagemaker-bot
Copy link
Collaborator

AWS CodeBuild CI Report

  • Result: SUCCEEDED
  • Build Logs (available for 30 days)

Powered by github-codebuild-logs, available on the AWS Serverless Application Repository

@sagemaker-bot
Copy link
Collaborator

AWS CodeBuild CI Report

  • Result: SUCCEEDED
  • Build Logs (available for 30 days)

Powered by github-codebuild-logs, available on the AWS Serverless Application Repository

@sagemaker-bot
Copy link
Collaborator

AWS CodeBuild CI Report

  • Result: SUCCEEDED
  • Build Logs (available for 30 days)

Powered by github-codebuild-logs, available on the AWS Serverless Application Repository

@laurenyu laurenyu merged commit c701100 into aws:master Sep 4, 2019
@laurenyu
Copy link
Contributor

laurenyu commented Sep 4, 2019

thanks for your contribution!

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@ChoiByungWook ChoiByungWook ChoiByungWook left review comments

@laurenyu laurenyu laurenyu approved these changes

Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
4 participants
@speg03 @sagemaker-bot @laurenyu @ChoiByungWook