feat(sagemaker): add support uncompressed model #30949
Open
Conversation
This file contains 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
github-actions
bot
added
the
admired-contributor
WinterYukky
commented
Jul 25, 2024
Comment on lines
-73
to
-75
| if (!path.toLowerCase().endsWith(ARTIFACT_EXTENSION)) { | ||
| throw new Error(`Asset must be a gzipped tar file with extension ${ARTIFACT_EXTENSION} (${this.path})`); | ||
| } |
There was a problem hiding this comment.
The asset's availability determination was moved after binding it as an asset.
Comment on lines
+172
to
+174
| if (!this.asset.isFile) { | ||
| throw new Error(`Asset must be a file, if you want to use directory you can use 'ModelData.fromBucket()' with the 's3DataType' option to 'S3DataType.S3_PREFIX' and 'compressionType' option to 'CompressionType.NONE' (${this.path})`); | ||
| } |
There was a problem hiding this comment.
The bundled assets are now allowed if they are not directories.
Comment on lines
+179
to
+182
| compressionType: this.asset.assetPath.toLowerCase().endsWith(COMPRESSED_ARTIFACT_EXTENSION) | ||
| ? CompressionType.GZIP | ||
| : CompressionType.NONE, | ||
| s3DataType: S3DataType.S3_OBJECT, |
There was a problem hiding this comment.
Uncompressed single files are also supported
…ithub.com/WinterYukky/aws-cdk into feat/sagemaker/support-uncompressed-model
AWS CodeBuild CI Report
Powered by github-codebuild-logs, available on the AWS Serverless Application Repository |
aws-cdk-automation
added
the
pr/needs-community-review
This was referenced Jul 29, 2024
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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Outdated suggestions cannot be applied.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.
Issue # (if applicable)
None.
Reason for this change
Currently,
sagemaker.Modelonly supportModelDataUrl.ModelDataUrlmust be gzip the model artifact.https://docs.aws.amazon.com/AWSCloudFormation/latest/UserGuide/aws-properties-sagemaker-model-containerdefinition.html#cfn-sagemaker-model-containerdefinition-modeldataurl
LLM, which is popular these days, is large and takes longer to compress. For example, the 70B parameter model could take 2 hours or more to compress. SageMaker has a way to use the uncompressed model, and this PR is needed to support the uncompressed model.
The uncompressed model is also recommended in the official documentation.
Description of changes
Users can now specify the use of the uncompressed model by taking the model's compression type and S3 data type as options. When using an uncompressed model, use
ModelDataSourceinstead ofModelDataUrl. SinceModelDataSourcecannot be used in some workloads, the compression model usesModelDataUrlas before.Note
Currently you cannot use ModelDataSource in conjunction with SageMaker batch transform, SageMaker serverless endpoints, SageMaker multi-model endpoints, and SageMaker Marketplace.
https://docs.aws.amazon.com/sagemaker/latest/APIReference/API_ContainerDefinition.html#sagemaker-Type-ContainerDefinition-ModelDataSource
Also, asset bundling can now be used as a by-product of being able to use something other than tar.gz due to support for uncompressed models. Until now, no model artifacts with extensions other than tar.gz were permitted, but bundled assets are now allowed if they are not directories.
Since it is not possible to upload a directory or extract a zip with
s3_asset.assetalone, the directory cannot continue to be used as an asset. (If there is a way to support directories, please let me know)Description of how you validated changes
I've added an integ test with assertions that invoke the model endpoint and verify the response.
Checklist
Other
sagemaker-alpha module document is not aligned contributing guide recommendation, so also I fix the it.
https://github.com/aws/aws-cdk/blob/main/CONTRIBUTING.md#recommendations
By submitting this pull request, I confirm that my contribution is made under the terms of the Apache-2.0 license