fix: Upgrade to SageMaker v2 SDK #76

brightsparc · 2020-08-10T11:22:25Z

Issue #, if available: #69 Feature request: support SageMaker SDK 2.0

Related to #73

Description of changes:

Manually upgrade the step functions SDK to support the breaking changes relevant to this library specifically.

image -> image_uri
train_instance_count -> instance_count
train_instance_type -> instance_type
train_max_run -> max_run
train_max_run_wait -> max_run_wait
train_volume_size -> volume_size
sagemaker.session.s3_input -> sagemaker.inputs.TrainingInput

I upgrade dependencies to:

sagemaker>=2.1.0`
boto3>=1.14.38

I upgrade package version to: 2.0.0

I have an integration tests that validates this code works with - but have a few issues running the complete test suite.

TrainingStep
ModelStep
ProcessingStep

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.

* image -> image_uri * train_instance_count -> instance_count * train_instance_type -> instance_type * train_max_run -> max_run * train_max_run_wait -> max_run_wait * train_volume_size -> volume_size * sagemaker.session.s3_input -> sagemaker.inputs.TrainingInput

StepFunctions-Bot · 2020-08-10T11:25:14Z

AWS CodeBuild CI Report

CodeBuild project: StepFunctionsPythonSDK-integtests
Commit ID: 15f37b0
Result: FAILED
Build Logs (available for 30 days)

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

StepFunctions-Bot · 2020-08-10T11:27:09Z

AWS CodeBuild CI Report

Result: FAILED
Build Logs (available for 30 days)

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

get_image_uri -> sagemaker.image_uris.retrieve()

StepFunctions-Bot · 2020-08-10T11:36:33Z

AWS CodeBuild CI Report

Result: FAILED
Build Logs (available for 30 days)

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

StepFunctions-Bot · 2020-08-10T11:45:43Z

AWS CodeBuild CI Report

Result: FAILED
Build Logs (available for 30 days)

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

StepFunctions-Bot · 2020-08-10T13:26:04Z

AWS CodeBuild CI Report

Result: FAILED
Build Logs (available for 30 days)

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

StepFunctions-Bot · 2020-08-10T15:39:27Z

AWS CodeBuild CI Report

CodeBuild project: StepFunctionsPythonSDK-integtests
Commit ID: 9a88915
Result: FAILED
Build Logs (available for 30 days)

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

* Removed sagemaker_session for SKLearn * Moved checkpoint_path into hyper parameters (https://sagemaker.readthedocs.io/en/v2.0.0.rc0/frameworks/tensorflow/upgrade_from_legacy.html) * Added framework_version and py_version * Update entry_point and renamed image_name to image_uri for TensorFlow

StepFunctions-Bot · 2020-08-11T07:46:31Z

AWS CodeBuild CI Report

Result: FAILED
Build Logs (available for 30 days)

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

brightsparc · 2020-08-11T07:57:30Z

I have validated that there are no more upgrades required by the tool after performing upgrade-v2 to all files and using black to format consistency given the upgrade tool messes with formatting. (although I haven't re-formatted the code using black as part of this branch).

upgrade_dir () {    
    shopt -s nullglob dotglob

    for pathname in "$1"/*; do
        if [ -d "$pathname" ]; then
            upgrade_dir "$pathname"
        else
            case "$pathname" in
                *.py)
                    sagemaker-upgrade-v2 --in-file "$pathname" --out-file "$pathname"
            esac
        fi
    done
}

# install black for formatting
pip install black

# apply upgrades
upgrade_dir ./src
upgrade_dir ./tests

# re-apply black to src and test
black -S ./src
black -S ./tests

git diff -w

StepFunctions-Bot · 2020-08-11T11:04:45Z

AWS CodeBuild CI Report

CodeBuild project: StepFunctionsPythonSDK-integtests
Commit ID: 8b97898
Result: SUCCEEDED
Build Logs (available for 30 days)

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

src/stepfunctions/steps/sagemaker.py

vaib-amz · 2020-09-04T21:15:44Z

Thanks a lot for this PR @brightsparc ! I don't have any feedback besides the docstring. I will ask another team member to look at this PR, and will also have to get a confirmation regarding the version bump.

pouyanhoss · 2020-09-11T01:19:38Z

@brightsparc
Hi Julian,

I am using DeepAR in SageMaker and recently upgraded it to SDK v2. I was able to modify everything except the following part:

predictor = DeepARPredictor(endpoint=best_tuning_job_name, sagemaker_session=sess, content_type='application/json')

Based on the documentation, I modified it as following:

predictor = DeepARPredictor(endpoint_name=best_tuning_job_name, sagemaker_session=sess,serializer=sagemaker.serializers.JSONSerializer())

However, I get a TypeError when I call the predictor in the next cell:

predictor.set_prediction_parameters(freq, prediction_length)
list_of_df = predictor.predict(violation_list_training[:2])
TypeError: Object of type bytes is not JSON serializable

Any help would be much appreciated.

yoodan93

Looks good to me

wong-a · 2020-09-17T18:43:44Z

Thanks for the contribution @brightsparc

A few things to take care of before merging:

Resolve merge conflicts in VERSION, requiremenets.txt, and setup.py
Make sure acceptance and unit tests pass

For the Step Functions team, we need to figure out the plan for maintaining v1 of the stepfunctions SDK and user guidance for migration from v1 to v2.

brightsparc · 2020-09-20T05:51:13Z

For the Step Functions team, we need to figure out the plan for maintaining v1 of the stepfunctions SDK and user guidance for migration from v1 to v2.

I have resolved the conflict to stick with the v2 target version. @wong-a I can revise this as required.

StepFunctions-Bot · 2020-09-20T05:57:14Z

AWS CodeBuild CI Report

Result: FAILED
Build Logs (available for 30 days)

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

StepFunctions-Bot · 2020-09-20T08:49:06Z

AWS CodeBuild CI Report

CodeBuild project: StepFunctionsPythonSDK-integtests
Commit ID: 24f1b61
Result: FAILED
Build Logs (available for 30 days)

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

metrizable

looks good! is there interest in using an auto-formatter like black?

metrizable · 2020-09-22T23:52:41Z

src/stepfunctions/steps/sagemaker.py

@@ -36,12 +36,12 @@ def __init__(self, state_id, estimator, job_name, data=None, hyperparameters=Non
            data: Information about the training data. Please refer to the ``fit()`` method of the associated estimator, as this can take any of the following forms:

                * (str) - The S3 location where training data is saved.
-                * (dict[str, str] or dict[str, sagemaker.session.s3_input]) - If using multiple
+                * (dict[str, str] or dict[str, sagemaker.inputs.TrainingInput]) - If using multiple


suggested edit: is it worth considering for the future is to specify the generic Dict from typing (or even Mapping)? I know these are just docstring comments, but in the future, you may want to include type hinting and the change will be minor in the future at that point:

Suggested change

* (dict[str, str] or dict[str, sagemaker.inputs.TrainingInput]) - If using multiple

* (Dict[str, str] or Dict[str, sagemaker.inputs.TrainingInput]) - If using multiple

I think an auto-formatter is a good idea. I used black when I was ensuring I had applied the correct v2 SM upgrades, but the diff made it hard to compare changes so left it as is.

Those doc comments could be updated with a seperate PR as I suspect there are a few.

…teps, fix Rule evaluator image, and instance count/type

StepFunctions-Bot · 2020-09-23T05:11:37Z

AWS CodeBuild CI Report

Result: SUCCEEDED
Build Logs (available for 30 days)

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

StepFunctions-Bot · 2020-09-23T08:19:34Z

AWS CodeBuild CI Report

CodeBuild project: StepFunctionsPythonSDK-integtests
Commit ID: 50fada4
Result: SUCCEEDED
Build Logs (available for 30 days)

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

vaib-amz · 2020-09-23T20:17:16Z

Updating version to a pre-release candidate. The tests were all okay, only the version has changed. Since the changed were approved already, merging in.

StepFunctions-Bot · 2020-09-23T20:25:10Z

AWS CodeBuild CI Report

Result: SUCCEEDED
Build Logs (available for 30 days)

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

StepFunctions-Bot · 2020-09-23T23:35:42Z

AWS CodeBuild CI Report

CodeBuild project: StepFunctionsPythonSDK-integtests
Commit ID: c5cabdc
Result: FAILED
Build Logs (available for 30 days)

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

brightsparc added 2 commits August 10, 2020 21:03

Update verison to 2.0.0

15f37b0

Add addition mapping

e868bee

get_image_uri -> sagemaker.image_uris.retrieve()

Update RuleEvaluatorImage to account: 199566480951

2273ffa

vaib-amz self-requested a review September 4, 2020 19:36

vaib-amz reviewed Sep 4, 2020

View reviewed changes

src/stepfunctions/steps/sagemaker.py Show resolved Hide resolved

brightsparc mentioned this pull request Sep 15, 2020

Feature request: support SageMaker SDK 2.0 #69

Closed

vaib-amz requested a review from yoodan93 September 16, 2020 21:46

yoodan93 previously approved these changes Sep 17, 2020

View reviewed changes

brightsparc added 2 commits September 20, 2020 15:41

Remove doc string as per comments

a8053db

Merging changes from upstream master. Targeting v2+

24f1b61

brightsparc dismissed yoodan93’s stale review via 24f1b61 September 20, 2020 05:48

metrizable previously approved these changes Sep 23, 2020

View reviewed changes

Addition unit test fixups to remove cloudwtch metrics, add training_s…

50fada4

…teps, fix Rule evaluator image, and instance count/type

brightsparc dismissed metrizable’s stale review via 50fada4 September 23, 2020 05:02

Merge branch 'master' into sagemaker-v2

c5cabdc

vaib-amz merged commit dbcf358 into aws:master Sep 23, 2020

brightsparc mentioned this pull request Sep 25, 2020

Upgrade Data Science SDK notebooks to support SageMaker v2 SDK aws/amazon-sagemaker-examples#1568

Open

yoodan93 mentioned this pull request Dec 19, 2020

v2 Release plans and migration instructions #108

Open

	* (dict[str, str] or dict[str, sagemaker.inputs.TrainingInput]) - If using multiple
	* (Dict[str, str] or Dict[str, sagemaker.inputs.TrainingInput]) - If using multiple

fix: Upgrade to SageMaker v2 SDK #76

fix: Upgrade to SageMaker v2 SDK #76

Uh oh!

Conversation

brightsparc commented Aug 10, 2020 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

StepFunctions-Bot commented Aug 10, 2020

AWS CodeBuild CI Report

Uh oh!

StepFunctions-Bot commented Aug 10, 2020

AWS CodeBuild CI Report

Uh oh!

StepFunctions-Bot commented Aug 10, 2020

AWS CodeBuild CI Report

Uh oh!

StepFunctions-Bot commented Aug 10, 2020

AWS CodeBuild CI Report

Uh oh!

StepFunctions-Bot commented Aug 10, 2020

AWS CodeBuild CI Report

Uh oh!

StepFunctions-Bot commented Aug 10, 2020

AWS CodeBuild CI Report

Uh oh!

StepFunctions-Bot commented Aug 11, 2020

AWS CodeBuild CI Report

Uh oh!

brightsparc commented Aug 11, 2020 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

StepFunctions-Bot commented Aug 11, 2020

AWS CodeBuild CI Report

Uh oh!

Uh oh!

vaib-amz commented Sep 4, 2020

Uh oh!

pouyanhoss commented Sep 11, 2020 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

yoodan93 left a comment

Choose a reason for hiding this comment

Uh oh!

wong-a commented Sep 17, 2020

Uh oh!

brightsparc commented Sep 20, 2020

Uh oh!

StepFunctions-Bot commented Sep 20, 2020

AWS CodeBuild CI Report

Uh oh!

StepFunctions-Bot commented Sep 20, 2020

AWS CodeBuild CI Report

Uh oh!

metrizable left a comment

Choose a reason for hiding this comment

Uh oh!

metrizable Sep 22, 2020

Choose a reason for hiding this comment

Uh oh!

brightsparc Sep 23, 2020

Choose a reason for hiding this comment

Uh oh!

StepFunctions-Bot commented Sep 23, 2020

AWS CodeBuild CI Report

Uh oh!

StepFunctions-Bot commented Sep 23, 2020

AWS CodeBuild CI Report

Uh oh!

vaib-amz commented Sep 23, 2020

Uh oh!

StepFunctions-Bot commented Sep 23, 2020

AWS CodeBuild CI Report

Uh oh!

StepFunctions-Bot commented Sep 23, 2020

AWS CodeBuild CI Report

Uh oh!

Uh oh!

brightsparc commented Aug 10, 2020 •

edited

Loading

brightsparc commented Aug 11, 2020 •

edited

Loading

pouyanhoss commented Sep 11, 2020 •

edited

Loading