Skip to content

feature: Adding support in HuggingFace estimator for Training Compiler enhanced PyTorch 1.11 #3307

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 26 commits into from
Aug 30, 2022
Merged

feature: Adding support in HuggingFace estimator for Training Compiler enhanced PyTorch 1.11 #3307

merged 26 commits into from
Aug 30, 2022

Conversation

Lokiiiiii
Copy link
Contributor

Description of changes:

  1. Adding new DLC for trcomp-hf-pt 1.11
  2. Adding support for new distribution mechanism 'pytorch_xla' to be used only with trcomp-hf-pt 1.11
  3. Invalidating heterogeneous clusters with Training Compiler
  4. Adding support for G5 instances to be used with Training Compiler

Testing done:

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.

General

  • I have read the CONTRIBUTING doc
  • I certify that the changes I am introducing will be backword compatible, and I have discussed concerns about this, if any, with the Python SDK team
  • I used the commit message format described in CONTRIBUTING
  • I have passed the region in to all S3 and STS clients that I've initialized as part of this change.
  • I have updated any necessary documentation, including READMEs and API docs (if appropriate)

Tests

  • I have added tests that prove my fix is effective or that my feature works (if appropriate)
  • I have added unit and/or integration tests as appropriate to ensure backward compatibility of the changes
  • I have checked that my tests are not configured for a specific region or account (if appropriate)
  • I have used unique_name_from_base to create resource names in integ tests (if appropriate)

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

pinaraws
pinaraws reviewed Aug 18, 2022
View reviewed changes
src/sagemaker/huggingface/training_compiler/config.py
@@ -24,7 +26,7 @@
class TrainingCompilerConfig(BaseConfig):
"""The SageMaker Training Compiler configuration class."""

SUPPORTED_INSTANCE_CLASS_PREFIXES = ["p3", "g4dn", "p4"]
SUPPORTED_INSTANCE_CLASS_PREFIXES = ["p3", "g4dn", "p4d", "g5"]
Copy link
Member

Choose a reason for hiding this comment

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

did you mean to remove p4? is it intentional?

Copy link
Contributor Author

Choose a reason for hiding this comment

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

p4d is the only available flavor of p4 AFAIK. Just making the check more specific.

src/sagemaker/huggingface/training_compiler/config.py Outdated
if estimator.pytorch_version:
if Version(estimator.pytorch_version) in SpecifierSet(">= 1.11"):
error_helper_string = (
"'pytorch_xla' is the only distribution mechanism currently supported "
Copy link
Member

Choose a reason for hiding this comment

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

you can still run without specifying distribution, correct?

Copy link
Contributor Author

@Lokiiiiii Lokiiiiii Aug 18, 2022 •
edited
Loading

Choose a reason for hiding this comment

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

Yes, this is just blocking customers from using other built-in distribution mechanisms like mpi or pytorch ddp.

mchoi8739
mchoi8739 reviewed Aug 18, 2022
View reviewed changes
mchoi8739
mchoi8739 reviewed Aug 18, 2022
View reviewed changes
mchoi8739
mchoi8739 reviewed Aug 18, 2022
View reviewed changes
mchoi8739
mchoi8739 reviewed Aug 18, 2022
View reviewed changes
mchoi8739
mchoi8739 reviewed Aug 18, 2022
View reviewed changes
pinaraws
pinaraws reviewed Aug 19, 2022
View reviewed changes
mchoi8739
mchoi8739 suggested changes Aug 22, 2022
View reviewed changes
mufaddal-rohawala
mufaddal-rohawala reviewed Aug 25, 2022
View reviewed changes
Copy link
Member

@mufaddal-rohawala mufaddal-rohawala left a comment •
edited
Loading

Choose a reason for hiding this comment

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

/bot run all

@sagemaker-bot
Copy link
Collaborator

AWS CodeBuild CI Report

  • CodeBuild project: sagemaker-python-sdk-unit-tests
  • Commit ID: 10b7c4e
  • 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

  • CodeBuild project: sagemaker-python-sdk-local-mode-tests
  • Commit ID: 10b7c4e
  • 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

  • CodeBuild project: sagemaker-python-sdk-pr
  • Commit ID: 10b7c4e
  • 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

  • CodeBuild project: sagemaker-python-sdk-notebook-tests
  • Commit ID: 10b7c4e
  • 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

  • CodeBuild project: sagemaker-python-sdk-slow-tests
  • Commit ID: 10b7c4e
  • 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

  • CodeBuild project: sagemaker-python-sdk-slow-tests
  • Commit ID: 6688892
  • 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

  • CodeBuild project: sagemaker-python-sdk-notebook-tests
  • Commit ID: 6688892
  • Result: SUCCEEDED
  • Build Logs (available for 30 days)

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

@Lokiiiiii Lokiiiiii marked this pull request as ready for review August 30, 2022 16:20
harryzorus
harryzorus reviewed Aug 30, 2022
View reviewed changes
src/sagemaker/training_compiler/config.py
for i in cls.SUPPORTED_INSTANCE_CLASS_PREFIXES
]
):
error_helper_string = (
Copy link

@harryzorus harryzorus Aug 30, 2022 •
edited
Loading

Choose a reason for hiding this comment

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

can we add info logging to inform customers that p3dn.24xlarge & p4d.24xlarge instances offer the best performance for distributed training?

Copy link
Contributor Author

Choose a reason for hiding this comment

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

Do we just want to generalize to EFA enabled instances or what is special about these 2 instance types ?

Copy link

@harryzorus harryzorus Aug 30, 2022 •
edited
Loading

Choose a reason for hiding this comment

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

That's a good idea, we eventually would be to recommend all EFA instances that improve performance. We should start with recommending EFA instances we benchmarked thoroughly, we can add other instances to the list once we benchmark them internally.

@sagemaker-bot
Copy link
Collaborator

AWS CodeBuild CI Report

  • CodeBuild project: sagemaker-python-sdk-unit-tests
  • Commit ID: e0580b5
  • 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

  • CodeBuild project: sagemaker-python-sdk-local-mode-tests
  • Commit ID: e0580b5
  • 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

  • CodeBuild project: sagemaker-python-sdk-pr
  • Commit ID: e0580b5
  • 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

  • CodeBuild project: sagemaker-python-sdk-slow-tests
  • Commit ID: e0580b5
  • 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

  • CodeBuild project: sagemaker-python-sdk-notebook-tests
  • Commit ID: e0580b5
  • 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

  • CodeBuild project: sagemaker-python-sdk-unit-tests
  • Commit ID: 65086e5
  • 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

  • CodeBuild project: sagemaker-python-sdk-pr
  • Commit ID: 65086e5
  • 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

  • CodeBuild project: sagemaker-python-sdk-local-mode-tests
  • Commit ID: 65086e5
  • 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

  • CodeBuild project: sagemaker-python-sdk-slow-tests
  • Commit ID: 65086e5
  • 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

  • CodeBuild project: sagemaker-python-sdk-unit-tests
  • Commit ID: 65086e5
  • 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

  • CodeBuild project: sagemaker-python-sdk-pr
  • Commit ID: 65086e5
  • Result: SUCCEEDED
  • Build Logs (available for 30 days)

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

mchoi8739
mchoi8739 approved these changes Aug 30, 2022
View reviewed changes
@sagemaker-bot
Copy link
Collaborator

AWS CodeBuild CI Report

  • CodeBuild project: sagemaker-python-sdk-notebook-tests
  • Commit ID: 65086e5
  • 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

  • CodeBuild project: sagemaker-python-sdk-unit-tests
  • Commit ID: 280f625
  • 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

  • CodeBuild project: sagemaker-python-sdk-pr
  • Commit ID: 280f625
  • 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

  • CodeBuild project: sagemaker-python-sdk-local-mode-tests
  • Commit ID: 280f625
  • 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

  • CodeBuild project: sagemaker-python-sdk-slow-tests
  • Commit ID: 280f625
  • 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

  • CodeBuild project: sagemaker-python-sdk-notebook-tests
  • Commit ID: 280f625
  • Result: SUCCEEDED
  • Build Logs (available for 30 days)

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

@BasilBeirouti BasilBeirouti merged commit 01fd66a into aws:master Aug 30, 2022
JoseJuan98 pushed a commit to JoseJuan98/sagemaker-python-sdk that referenced this pull request Mar 4, 2023
@Lokiiiiii @mchoi8739 @JoseJuan98
feature: Adding support in HuggingFace estimator for Training Compile…
7289b89 
…r enhanced PyTorch 1.11 (aws#3307)

* feature: Adding support in HuggingFace estimator for Training Compiler enhanced PyTorch 1.11

* Update src/sagemaker/huggingface/training_compiler/config.py

Co-authored-by: Miyoung <[email protected]>

* Update src/sagemaker/huggingface/training_compiler/config.py

Co-authored-by: Miyoung <[email protected]>

* Update src/sagemaker/huggingface/training_compiler/config.py

Co-authored-by: Miyoung <[email protected]>

* Update src/sagemaker/training_compiler/config.py

Co-authored-by: Miyoung <[email protected]>

* fix: renaming distribution parameters pytorch_xla -> pytorchxla

* Update src/sagemaker/huggingface/training_compiler/config.py

Co-authored-by: Miyoung <[email protected]>

* Update src/sagemaker/huggingface/estimator.py

Co-authored-by: Miyoung <[email protected]>

* Update src/sagemaker/huggingface/training_compiler/config.py

Co-authored-by: Miyoung <[email protected]>

* Update src/sagemaker/huggingface/training_compiler/config.py

Co-authored-by: Miyoung <[email protected]>

* Update src/sagemaker/huggingface/estimator.py

Co-authored-by: Miyoung <[email protected]>

* Update src/sagemaker/huggingface/estimator.py

Co-authored-by: Miyoung <[email protected]>

* Fix: syntax error in trcomp tests

* fix: linting

* fix: linting to break up long lines

* fix: fixture scoping issue in integ test

* fix: broken unit tests for trcomp

* fix: broken skip logic in version fixtures

* fix: update test and version compatibility

* feature: added warning recommending EFA instances with training compiler

* Update src/sagemaker/huggingface/estimator.py

Co-authored-by: Miyoung <[email protected]>

* Update src/sagemaker/training_compiler/config.py

Co-authored-by: Miyoung <[email protected]>

Co-authored-by: Miyoung <[email protected]>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@harryzorus harryzorus harryzorus approved these changes

@mchoi8739 mchoi8739 mchoi8739 approved these changes

@pinaraws pinaraws pinaraws approved these changes

@mufaddal-rohawala mufaddal-rohawala mufaddal-rohawala 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.
8 participants
@Lokiiiiii @sagemaker-bot @codecov-commenter @harryzorus @mchoi8739 @pinaraws @mufaddal-rohawala @BasilBeirouti