-
Notifications
You must be signed in to change notification settings - Fork 1.2k
fix: jumpstart async inference config predictor support #3970
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.
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
fix: jumpstart async inference config predictor support #3970
Conversation
AWS CodeBuild CI Report
Powered by github-codebuild-logs, available on the AWS Serverless Application Repository |
AWS CodeBuild CI Report
Powered by github-codebuild-logs, available on the AWS Serverless Application Repository |
AWS CodeBuild CI Report
Powered by github-codebuild-logs, available on the AWS Serverless Application Repository |
Refers to #3969 |
AWS CodeBuild CI Report
Powered by github-codebuild-logs, available on the AWS Serverless Application Repository |
Codecov Report
@@ Coverage Diff @@
## master #3970 +/- ##
==========================================
- Coverage 90.32% 89.59% -0.73%
==========================================
Files 1292 305 -987
Lines 113972 28188 -85784
==========================================
- Hits 102940 25256 -77684
+ Misses 11032 2932 -8100
|
AWS CodeBuild CI Report
Powered by github-codebuild-logs, available on the AWS Serverless Application Repository |
@@ -973,7 +973,7 @@ def deploy( | |||
) |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Please remove the link in the description to our ticket corp ( link this bug instead: #3969)
@@ -432,7 +432,7 @@ def deploy( | |||
predictor = super(JumpStartModel, self).deploy(**deploy_kwargs.to_kwargs_dict()) | |||
|
|||
# If no predictor class was passed, add defaults to predictor | |||
if self.orig_predictor_cls is None: | |||
if self.orig_predictor_cls is None and async_inference_config is None: |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Don't have enough context here, but can you confirm that we will not see similar issues for other types of configs as well - serverless_inference_config, data_capture_config ? i.e. if that is None, do we need to return a defaultPredictor as well?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
To be honest, I am really not sure. I wanted to resolve the customer ticket ASAP, but I will do a deep dive into other kinds of issues with predictor being set under the hood.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
thanks for the callout @akrishna1995.
IIUC, Data capture wouldn't impact the predictor, but could you please let us know if you disagree.
As of today, JumpStart does not claim to support serverless inference for any model. It might work for some of our smaller models, but for foundation models it wouldn't: they all run on GPU. We will look into it however for our older models, but can this be in a separate PRs?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Sure - Thanks for the confirmation, rest is LGTM, will approve and merge the PR
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Changes are LGTM overall, minor comment for confirmation and request to remove amazon specific ticketing link from issue description
AWS CodeBuild CI Report
Powered by github-codebuild-logs, available on the AWS Serverless Application Repository |
AWS CodeBuild CI Report
Powered by github-codebuild-logs, available on the AWS Serverless Application Repository |
AWS CodeBuild CI Report
Powered by github-codebuild-logs, available on the AWS Serverless Application Repository |
AWS CodeBuild CI Report
Powered by github-codebuild-logs, available on the AWS Serverless Application Repository |
AWS CodeBuild CI Report
Powered by github-codebuild-logs, available on the AWS Serverless Application Repository |
@@ -432,7 +432,7 @@ def deploy( | |||
predictor = super(JumpStartModel, self).deploy(**deploy_kwargs.to_kwargs_dict()) | |||
|
|||
# If no predictor class was passed, add defaults to predictor | |||
if self.orig_predictor_cls is None: | |||
if self.orig_predictor_cls is None and async_inference_config is None: |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
thanks for the callout @akrishna1995.
IIUC, Data capture wouldn't impact the predictor, but could you please let us know if you disagree.
As of today, JumpStart does not claim to support serverless inference for any model. It might work for some of our smaller models, but for foundation models it wouldn't: they all run on GPU. We will look into it however for our older models, but can this be in a separate PRs?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
LGTM
Issue #, if available:
#3969
Description of changes:
If
async_inference_config
field is used withJumpStartModel
orJumpStartEstimator
, we use the predictor associated with async inference, not try to create a predictor with JumpStart-defaulted values.Testing done:
Wrote unit tests.
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
Tests
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.