[SDK] Allow customising base trainer and storage images in Train API #2261
Conversation
Allow customizing base storage_initializer and trainer images through Env vars. Signed-off-by: Varsha Prasad Narsing <[email protected]>
varshaprasad96
force-pushed
the
sdk/fetch-base-image
branch
from
9cda074
to
40c5c10
Compare
Pull Request Test Coverage Report for Build 10927951593Details
💛 - Coveralls |
There was a problem hiding this comment.
Thank you for creating this PR!
/approve
@deepanker13 Do you have any other comments?
If not, you can just say /lgtm, and then this will be merged into the master branch.
|
[APPROVALNOTIFIER] This PR is APPROVED This pull-request has been approved by: tenzen-y The full list of commands accepted by this bot can be found here. The pull request process is described here
Needs approval from an approver in each of these files:
Approvers can indicate their approval by writing |
|
/assign @deepanker13 |
|
Thanks @varshaprasad96 |
There was a problem hiding this comment.
Thank you for doing this @varshaprasad96!
Please can you submit PR to address my comment ?
| @@ -82,14 +82,14 @@ | |||
|
|
|||
|
|
|||
| # TODO (andreyvelich): We should add image tag for Storage Initializer and Trainer. | |||
| STORAGE_INITIALIZER_IMAGE = "docker.io/kubeflow/storage-initializer" | |||
| STORAGE_INITIALIZER_IMAGE_DEFAULT = "docker.io/kubeflow/storage-initializer" | |||
There was a problem hiding this comment.
@varshaprasad96 Please can you submit PR to add the following change in the constants.py:
STORAGE_INITIALIZER_IMAGE = os.getenv("STORAGE_INITIAILIZER_IMAGE", "docker.io/kubeflow/storage-initializer")
TRAINER_TRANSFORMER_IMAGE = os.getenv("TRAINER_TRANSFORMER_IMAGE", "docker.io/kubeflow/trainer-huggingface")That will allow users to quickly see the env they can modify, instead of searching in the training_client.py.
There was a problem hiding this comment.
I see!.. here we go: #2268
Follow up from kubeflow/training-operator#2261 as this is a user facing change. Signed-off-by: Varsha Prasad Narsing <[email protected]>
Follow up from kubeflow/training-operator#2261 as this is a user facing change. Signed-off-by: Varsha Prasad Narsing <[email protected]>
Follow up from kubeflow/training-operator#2261 as this is a user facing change. Signed-off-by: Varsha Prasad Narsing <[email protected]>
Follow up from kubeflow/training-operator#2261 as this is a user facing change. Signed-off-by: Varsha Prasad Narsing <[email protected]>
Follow up from kubeflow/training-operator#2261 as this is a user facing change. Signed-off-by: Varsha Prasad Narsing <[email protected]>
What this PR does / why we need it:
Allow customising base storage_initializer and trainer images through Env vars.
Example use case: Train API could be expanded to use ROCm libs in addition to CUDA.
Which issue(s) this PR fixes (optional, in
Fixes #<issue number>, #<issue number>, ...format, will close the issue(s) when PR gets merged):Fixes #2247
TODO: Docs to be updated in https://github.com/kubeflow/website.
Checklist: