feat: Hyperparameter validation #2856

evakravi · 2022-01-19T17:16:20Z

Issue #, if available:

Description of changes:
This PR adds utilities to perform hyperparameter validation. At a high level, a validate function was added to the hyperparameters module. This gives users the ability to validate hyperparameters for their models.

Testing done:
Unit testing has been done on the validate function

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.

sagemaker-bot · 2022-01-19T17:24:36Z

AWS CodeBuild CI Report

CodeBuild project: sagemaker-python-sdk-unit-tests
Commit ID: f4b0536
Result: FAILED
Build Logs (available for 30 days)

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

sagemaker-bot · 2022-01-19T18:18:10Z

AWS CodeBuild CI Report

CodeBuild project: sagemaker-python-sdk-local-mode-tests
Commit ID: f4b0536
Result: SUCCEEDED
Build Logs (available for 30 days)

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

sagemaker-bot · 2022-01-19T18:34:21Z

AWS CodeBuild CI Report

CodeBuild project: sagemaker-python-sdk-slow-tests
Commit ID: f4b0536
Result: SUCCEEDED
Build Logs (available for 30 days)

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

JGuinegagne

Mostly docstring changes and suggestions.

JGuinegagne · 2022-01-20T15:45:30Z

src/sagemaker/jumpstart/enums.py

+
+    TEXT = "text"
+    INT = "int"
+    FLOAT = "float"


shall we add a BOOL member while we are at it?

JGuinegagne · 2022-01-20T15:47:36Z

src/sagemaker/jumpstart/exceptions.py

+
+
+class JumpStartHyperparametersError(Exception):
+    """Exception raised for errors with hyperparameters for JumpStart models."""


nit:

"""Exception raised when validation of the parameter of a JumpStart model fails."""

JGuinegagne · 2022-01-20T15:47:59Z

src/sagemaker/hyperparameters.py

+          If set to``VALIDATE_ALGORITHM``, all algorithm hyperparameters will be validated.
+          If set to ``VALIDATE_ALL``, all hyperparameters for the model will be validated.
+          (Default: None)
+


add a Raises: section in the docstring please.

nit: also add ValueError in Raises docsting.

JGuinegagne · 2022-01-20T15:50:01Z

src/sagemaker/jumpstart/validators.py

+    if len(hyperparameter_spec) == 0:
+        raise JumpStartHyperparametersError(
+            f"Unable to perform validation -- cannot find hyperparameter '{hyperparameter_name}' in model specs."
+        )


could you add another if to assert that hyperparameter_spec length is 1. If not, we should raise another exception as the name attribute should be unique in the hp set.

JGuinegagne · 2022-01-20T15:51:50Z

src/sagemaker/jumpstart/validators.py

+                raise JumpStartHyperparametersError(
+                    f"Hyperparameter '{hyperparameter_name}' must have one of the following values: "
+                    ", ".join(hyperparameter_spec.options)
+                )


could you add support for min and max as well? For a TEXT parameter, that refers to the length of the string.

JGuinegagne · 2022-01-20T15:56:56Z

src/sagemaker/jumpstart/validators.py

+def validate_hyperparameters(
+    model_id: str,
+    model_version: str,
+    hyperparameters: dict,


refine type please: Dict[str, Any]

JGuinegagne · 2022-01-20T15:57:42Z

src/sagemaker/jumpstart/validators.py

+    hyperparameters: dict,
+    validation_mode: HyperparameterValidationMode = HyperparameterValidationMode.VALIDATE_PROVIDED,
+    region: Optional[str] = JUMPSTART_DEFAULT_REGION_NAME,
+):


return type please: -> None

JGuinegagne · 2022-01-20T15:58:02Z

src/sagemaker/jumpstart/validators.py

+    hyperparameter_name: str,
+    hyperparameter_value: Any,
+    hyperparameter_specs: List[JumpStartHyperparameter],
+):


return type please -> None in your implementation

JGuinegagne · 2022-01-20T16:03:03Z

src/sagemaker/jumpstart/validators.py

+
+    if validation_mode == HyperparameterValidationMode.VALIDATE_PROVIDED:
+        for hyperparam_name, hyperparam_value in hyperparameters.items():
+            _validate_hyperparameter(hyperparam_name, hyperparam_value, hyperparameters_specs)


question: this implementation fails fast, ie it raises the first validation exception it encounters. For example, if a model takes hp [a, b, c, d], and a, c and d are wrong, it will take the users 3 attempts to fix it. Do you think that is a concern?

I think this is fine. Many validators fail at the first error, and force the user to correct upstream errors before tackling issues that occur downstream.

JGuinegagne · 2022-01-20T16:13:07Z

src/sagemaker/jumpstart/validators.py

+            if not hyperparameter_value_str[start_index:].isdigit():
+                raise JumpStartHyperparametersError(
+                    f"Hyperparameter '{hyperparameter_name}' must be integer type ('{hyperparameter_value}')."
+                )


would this be simpler:

try: int_value = int(hyperparameter_value) except ValueError: raise JumpStartHyperparametersError(f"...")

That won't work, because floats can be serialized into ints.

Wouldn't the training script arg parser take care of the conversion in such a case?

JGuinegagne · 2022-01-21T16:32:55Z

src/sagemaker/hyperparameters.py

+    model_version: Optional[str] = None,
+    hyperparameters: Optional[dict] = None,
+    validation_mode: Optional[HyperparameterValidationMode] = None,
+):


add return type please: -> None

sagemaker-bot · 2022-01-22T01:36:06Z

AWS CodeBuild CI Report

CodeBuild project: sagemaker-python-sdk-unit-tests
Commit ID: 8f4aecf
Result: SUCCEEDED
Build Logs (available for 30 days)

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

sagemaker-bot · 2022-01-22T01:36:07Z

AWS CodeBuild CI Report

CodeBuild project: sagemaker-python-sdk-unit-tests
Commit ID: c917ad8
Result: SUCCEEDED
Build Logs (available for 30 days)

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

sagemaker-bot · 2022-01-22T01:37:53Z

AWS CodeBuild CI Report

CodeBuild project: sagemaker-python-sdk-unit-tests
Commit ID: dd26117
Result: SUCCEEDED
Build Logs (available for 30 days)

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

sagemaker-bot · 2022-01-22T02:09:56Z

AWS CodeBuild CI Report

CodeBuild project: sagemaker-python-sdk-local-mode-tests
Commit ID: 8f4aecf
Result: SUCCEEDED
Build Logs (available for 30 days)

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

codecov-commenter · 2022-01-22T02:12:13Z

Codecov Report

Merging #2856 (8a35894) into master-jumpstart (c9aa29b) will increase coverage by 0.03%.
The diff coverage is 91.48%.

@@                 Coverage Diff                  @@
##           master-jumpstart    #2856      +/-   ##
====================================================
+ Coverage             89.27%   89.31%   +0.03%     
====================================================
  Files                   186      188       +2     
  Lines                 16063    16200     +137     
====================================================
+ Hits                  14341    14469     +128     
- Misses                 1722     1731       +9

Impacted Files	Coverage Δ
src/sagemaker/environment_variables.py	`90.90% <0.00%> (ø)`
src/sagemaker/model_uris.py	`90.90% <0.00%> (ø)`
src/sagemaker/hyperparameters.py	`84.21% <70.00%> (-6.70%)`	⬇️
src/sagemaker/jumpstart/validators.py	`91.86% <91.86%> (ø)`
src/sagemaker/jumpstart/artifacts.py	`92.95% <100.00%> (+0.10%)`	⬆️
src/sagemaker/jumpstart/enums.py	`100.00% <100.00%> (ø)`
src/sagemaker/jumpstart/exceptions.py	`82.14% <100.00%> (+2.97%)`	⬆️
src/sagemaker/jumpstart/types.py	`94.33% <100.00%> (+0.22%)`	⬆️

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update c9aa29b...8a35894. Read the comment docs.

sagemaker-bot · 2022-01-22T02:12:51Z

AWS CodeBuild CI Report

CodeBuild project: sagemaker-python-sdk-local-mode-tests
Commit ID: c917ad8
Result: SUCCEEDED
Build Logs (available for 30 days)

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

sagemaker-bot · 2022-01-22T02:20:23Z

AWS CodeBuild CI Report

CodeBuild project: sagemaker-python-sdk-local-mode-tests
Commit ID: dd26117
Result: SUCCEEDED
Build Logs (available for 30 days)

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

sagemaker-bot · 2022-01-22T02:27:51Z

AWS CodeBuild CI Report

CodeBuild project: sagemaker-python-sdk-notebook-tests
Commit ID: 8f4aecf
Result: SUCCEEDED
Build Logs (available for 30 days)

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

sagemaker-bot · 2022-01-22T02:28:04Z

AWS CodeBuild CI Report

CodeBuild project: sagemaker-python-sdk-notebook-tests
Commit ID: dd26117
Result: SUCCEEDED
Build Logs (available for 30 days)

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

sagemaker-bot · 2022-01-22T02:36:55Z

AWS CodeBuild CI Report

CodeBuild project: sagemaker-python-sdk-slow-tests
Commit ID: dd26117
Result: SUCCEEDED
Build Logs (available for 30 days)

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

sagemaker-bot · 2022-01-22T02:36:55Z

AWS CodeBuild CI Report

CodeBuild project: sagemaker-python-sdk-slow-tests
Commit ID: 8f4aecf
Result: SUCCEEDED
Build Logs (available for 30 days)

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

sagemaker-bot · 2022-01-22T02:36:56Z

AWS CodeBuild CI Report

CodeBuild project: sagemaker-python-sdk-slow-tests
Commit ID: c917ad8
Result: SUCCEEDED
Build Logs (available for 30 days)

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

sagemaker-bot · 2022-01-22T02:40:31Z

AWS CodeBuild CI Report

CodeBuild project: sagemaker-python-sdk-notebook-tests
Commit ID: c917ad8
Result: FAILED
Build Logs (available for 30 days)

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

…siveMin/exclusiveMax text params

sagemaker-bot · 2022-01-22T16:07:32Z

AWS CodeBuild CI Report

CodeBuild project: sagemaker-python-sdk-unit-tests
Commit ID: 063cdc7
Result: SUCCEEDED
Build Logs (available for 30 days)

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

sagemaker-bot · 2022-01-22T16:46:29Z

AWS CodeBuild CI Report

CodeBuild project: sagemaker-python-sdk-local-mode-tests
Commit ID: 063cdc7
Result: SUCCEEDED
Build Logs (available for 30 days)

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

JGuinegagne

only non-blocking comments

JGuinegagne · 2022-01-22T16:44:33Z

src/sagemaker/jumpstart/validators.py

+                raise JumpStartHyperparametersError(
+                    f"Hyperparameter '{hyperparameter_name}' must have length greater than "
+                    f"{hyperparameter_spec.exclusive_min}"
+                )


non-blocking: there is no real use-case for exclusive min/max with discreet values like integer (in this case text size), it's only useful for continuous values like float or double. That said, it doesn't really hurt to keep it either, it just makes the code a bit longer.

src/sagemaker/jumpstart/validators.py

sagemaker-bot · 2022-01-22T17:30:07Z

AWS CodeBuild CI Report

CodeBuild project: sagemaker-python-sdk-slow-tests
Commit ID: 063cdc7
Result: FAILED
Build Logs (available for 30 days)

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

sagemaker-bot · 2022-01-22T18:40:03Z

AWS CodeBuild CI Report

CodeBuild project: sagemaker-python-sdk-unit-tests
Commit ID: c6eb14d
Result: SUCCEEDED
Build Logs (available for 30 days)

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

sagemaker-bot · 2022-01-22T19:33:30Z

AWS CodeBuild CI Report

CodeBuild project: sagemaker-python-sdk-local-mode-tests
Commit ID: c6eb14d
Result: SUCCEEDED
Build Logs (available for 30 days)

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

sagemaker-bot · 2022-01-22T19:47:15Z

AWS CodeBuild CI Report

CodeBuild project: sagemaker-python-sdk-slow-tests
Commit ID: c6eb14d
Result: SUCCEEDED
Build Logs (available for 30 days)

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

…perparameter-validation

sagemaker-bot · 2022-01-23T17:07:18Z

AWS CodeBuild CI Report

CodeBuild project: sagemaker-python-sdk-unit-tests
Commit ID: 76271ea
Result: SUCCEEDED
Build Logs (available for 30 days)

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

sagemaker-bot · 2022-01-23T17:43:15Z

AWS CodeBuild CI Report

CodeBuild project: sagemaker-python-sdk-local-mode-tests
Commit ID: 76271ea
Result: SUCCEEDED
Build Logs (available for 30 days)

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

sagemaker-bot · 2022-01-23T18:11:01Z

AWS CodeBuild CI Report

CodeBuild project: sagemaker-python-sdk-notebook-tests
Commit ID: 76271ea
Result: SUCCEEDED
Build Logs (available for 30 days)

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

sagemaker-bot · 2022-01-23T18:12:32Z

AWS CodeBuild CI Report

CodeBuild project: sagemaker-python-sdk-slow-tests
Commit ID: 76271ea
Result: FAILED
Build Logs (available for 30 days)

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

sagemaker-bot · 2022-01-23T22:48:51Z

AWS CodeBuild CI Report

CodeBuild project: sagemaker-python-sdk-slow-tests
Commit ID: 76271ea
Result: SUCCEEDED
Build Logs (available for 30 days)

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

…perparameter-validation

sagemaker-bot · 2022-01-24T19:37:53Z

AWS CodeBuild CI Report

CodeBuild project: sagemaker-python-sdk-unit-tests
Commit ID: 8a95942
Result: SUCCEEDED
Build Logs (available for 30 days)

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

sagemaker-bot · 2022-01-24T20:21:36Z

AWS CodeBuild CI Report

CodeBuild project: sagemaker-python-sdk-local-mode-tests
Commit ID: 8a95942
Result: SUCCEEDED
Build Logs (available for 30 days)

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

sagemaker-bot · 2022-01-24T20:33:31Z

AWS CodeBuild CI Report

CodeBuild project: sagemaker-python-sdk-slow-tests
Commit ID: 8a95942
Result: SUCCEEDED
Build Logs (available for 30 days)

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

mufaddal-rohawala · 2022-01-25T17:43:51Z

src/sagemaker/hyperparameters.py

+          If set to``VALIDATE_ALGORITHM``, all algorithm hyperparameters will be validated.
+          If set to ``VALIDATE_ALL``, all hyperparameters for the model will be validated.
+          (Default: None)
+


nit: also add ValueError in Raises docsting.

sagemaker-bot · 2022-01-25T19:11:31Z

AWS CodeBuild CI Report

CodeBuild project: sagemaker-python-sdk-unit-tests
Commit ID: 617adc1
Result: SUCCEEDED
Build Logs (available for 30 days)

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

sagemaker-bot · 2022-01-25T20:06:16Z

AWS CodeBuild CI Report

CodeBuild project: sagemaker-python-sdk-local-mode-tests
Commit ID: 617adc1
Result: SUCCEEDED
Build Logs (available for 30 days)

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

sagemaker-bot · 2022-01-25T20:39:42Z

AWS CodeBuild CI Report

CodeBuild project: sagemaker-python-sdk-local-mode-tests
Commit ID: aa13bcf
Result: FAILED
Build Logs (available for 30 days)

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

sagemaker-bot · 2022-01-25T20:45:36Z

AWS CodeBuild CI Report

CodeBuild project: sagemaker-python-sdk-unit-tests
Commit ID: aa13bcf
Result: SUCCEEDED
Build Logs (available for 30 days)

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

sagemaker-bot · 2022-01-25T20:48:23Z

AWS CodeBuild CI Report

CodeBuild project: sagemaker-python-sdk-slow-tests
Commit ID: 617adc1
Result: SUCCEEDED
Build Logs (available for 30 days)

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

sagemaker-bot · 2022-01-25T21:07:03Z

AWS CodeBuild CI Report

CodeBuild project: sagemaker-python-sdk-unit-tests
Commit ID: 8a35894
Result: SUCCEEDED
Build Logs (available for 30 days)

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

sagemaker-bot · 2022-01-25T21:35:09Z

AWS CodeBuild CI Report

CodeBuild project: sagemaker-python-sdk-slow-tests
Commit ID: aa13bcf
Result: FAILED
Build Logs (available for 30 days)

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

sagemaker-bot · 2022-01-25T21:48:05Z

AWS CodeBuild CI Report

CodeBuild project: sagemaker-python-sdk-local-mode-tests
Commit ID: 8a35894
Result: SUCCEEDED
Build Logs (available for 30 days)

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

sagemaker-bot · 2022-01-25T22:34:05Z

AWS CodeBuild CI Report

CodeBuild project: sagemaker-python-sdk-slow-tests
Commit ID: 8a35894
Result: FAILED
Build Logs (available for 30 days)

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

sagemaker-bot · 2022-01-25T23:31:37Z

AWS CodeBuild CI Report

CodeBuild project: sagemaker-python-sdk-slow-tests
Commit ID: 8a35894
Result: SUCCEEDED
Build Logs (available for 30 days)

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

feat: hyperparameter validation

f4b0536

JGuinegagne requested changes Jan 20, 2022

View reviewed changes

JGuinegagne reviewed Jan 21, 2022

View reviewed changes

evakravi and others added 3 commits January 22, 2022 01:18

change: improve jumpstart hyperparam validation logic

8f4aecf

Merge branch 'master-jumpstart' into feat/hyperparameter-validation

c917ad8

change: add return typing

dd26117

change: add js unit tests for hyperparam validation for min/max/exclu…

063cdc7

…siveMin/exclusiveMax text params

JGuinegagne approved these changes Jan 22, 2022

View reviewed changes

fix: jumpstart docstring

c6eb14d

Merge remote-tracking branch 'upstream/master-jumpstart' into feat/hy…

76271ea

…perparameter-validation

Merge remote-tracking branch 'upstream/master-jumpstart' into feat/hy…

8a95942

…perparameter-validation

mufaddal-rohawala approved these changes Jan 25, 2022

View reviewed changes

fix: update jumpstart docstring

617adc1

shreyapandit approved these changes Jan 25, 2022

View reviewed changes

Merge branch 'master-jumpstart' into feat/hyperparameter-validation

aa13bcf

fix: entry point variable name

8a35894

shreyapandit changed the title ~~feat: hyperparameter validation~~ feat: Hyperparameter validation Jan 25, 2022

shreyapandit merged commit 423f389 into aws:master-jumpstart Jan 25, 2022



		class JumpStartHyperparametersError(Exception):
		"""Exception raised for errors with hyperparameters for JumpStart models."""

feat: Hyperparameter validation #2856

feat: Hyperparameter validation #2856

Uh oh!

Conversation

evakravi commented Jan 19, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Merge Checklist

General

Tests

Uh oh!

sagemaker-bot commented Jan 19, 2022

AWS CodeBuild CI Report

Uh oh!

sagemaker-bot commented Jan 19, 2022

AWS CodeBuild CI Report

Uh oh!

sagemaker-bot commented Jan 19, 2022

AWS CodeBuild CI Report

Uh oh!

JGuinegagne left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

sagemaker-bot commented Jan 22, 2022

AWS CodeBuild CI Report

Uh oh!

sagemaker-bot commented Jan 22, 2022

AWS CodeBuild CI Report

Uh oh!

sagemaker-bot commented Jan 22, 2022

AWS CodeBuild CI Report

Uh oh!

sagemaker-bot commented Jan 22, 2022

AWS CodeBuild CI Report

Uh oh!

codecov-commenter commented Jan 22, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

sagemaker-bot commented Jan 22, 2022

AWS CodeBuild CI Report

Uh oh!

sagemaker-bot commented Jan 22, 2022

AWS CodeBuild CI Report

Uh oh!

sagemaker-bot commented Jan 22, 2022

AWS CodeBuild CI Report

Uh oh!

sagemaker-bot commented Jan 22, 2022

AWS CodeBuild CI Report

Uh oh!

evakravi commented Jan 19, 2022 •

edited

Loading

codecov-commenter commented Jan 22, 2022 •

edited

Loading