fix: jumpstart estimator for gated uncompressed training #5175
Conversation
| data_source = train_args["InputDataConfig"][idx]["DataSource"] | ||
| if "S3DataSource" in data_source: | ||
| s3_data_source = data_source["S3DataSource"] | ||
| if "ModelAccessConfig" not in s3_data_source: |
There was a problem hiding this comment.
Who sets this ModelAccessConfig? Is it the default artifacts set by JumpStart or is it something user would explicitly add to their inputs?
There was a problem hiding this comment.
it's set by us when the customer inputs accept_eula=True
| train_args["checkpoint_local_path"] = estimator.checkpoint_local_path | ||
|
|
||
| @classmethod | ||
| def _set_accept_eula_for_model_channel_input_data_config(cls, train_args, accept_eula): |
There was a problem hiding this comment.
qq- I know @Narrohag added some code that does something very similar. I want to make sure we're not doing the same work twice
There was a problem hiding this comment.
Yeah I'm not fully following why this extra accept_eula logic is needed. For reference Evan this is where I added most of the model access config logic: https://github.com/aws/sagemaker-python-sdk/pull/5070/files. I'll take a closer look in a bit and maybe we can chat about it today
4 participants
Issue #, if available:
Description of changes:
JumpStartEstimatornot including the model channel when invokingCreateTrainingJobaccept_eulafield in baseEstimatorclass, and added support for setting low-level field accordingly.Testing done:
Merge Checklist
Put an
xin 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_baseto 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.