feat(trainer): add dataset and model initializer support to container backend by HKanoje · Pull Request #188 · kubeflow/sdk

HKanoje · 2025-12-04T03:46:58Z

What this PR does / why we need it

This PR implements dataset and model initializer support in the container backend, bringing it to feature parity with the Kubernetes backend. This addresses issue #171 by enabling users to automatically download and prepare datasets and models before training starts.

Solution Overview

This implementation adds full initializer support to the container backend by:

Running initializer containers before training - Initializers execute sequentially (dataset first, then model) before any training containers start
Using shared volumes - All containers (initializers and training nodes) share the same workspace directory on the host
Proper lifecycle management - Initializers must complete successfully before training begins, with automatic cleanup on failures
Comprehensive error handling - Clear error messages and proper resource cleanup when initialization fails

Detailed Changes

1. New Utility Functions (`kubeflow/trainer/backends/container/utils.py`)

`build_initializer_command(initializer, init_type)`

Builds the appropriate container command based on initializer type:

HuggingFace: Uses kubeflow.storage_initializer.hugging_face module
S3: Uses kubeflow.storage_initializer.s3 module
DataCache: Uses kubeflow.storage_initializer.datacache module

`build_initializer_env(initializer, init_type)`

Constructs environment variables from initializer configuration:

Sets STORAGE_URI from the initializer config
Sets OUTPUT_PATH to /workspace/dataset or /workspace/model based on type
Adds optional fields like ACCESS_TOKEN, ENDPOINT, REGION, etc.
Handles DataCache-specific variables like CLUSTER_SIZE, METADATA_LOC

`get_initializer_image()`

Returns the initializer container image (kubeflow/training-operator:latest).
This can be made configurable via backend config in future iterations.

2. Enhanced ContainerBackend (`kubeflow/trainer/backends/container/backend.py`)

`_run_initializers(job_name, initializer, workdir, network_id)`

Orchestrates the initialization phase:

Pulls the initializer image if needed (respects pull_policy)
Runs dataset initializer if configured
Runs model initializer if configured
Ensures proper sequencing and error propagation

`_run_single_initializer(job_name, initializer_config, init_type, image, workdir, network_id)`

Executes a single initializer container:

Creates container with proper labels for tracking
Mounts shared volume to /workspace
Monitors container status with configurable timeout (default 10 minutes)
Waits for successful completion (exit code 0)
Captures and reports logs on failure
Cleans up failed containers automatically

Updated `train()` method

Creates network and working directory first
Runs initializers before generating training script
Only proceeds to training if initialization succeeds
Maintains backward compatibility (initializers are optional)

Updated `__get_trainjob_from_containers()`

Correctly counts only training nodes for num_nodes (excludes initializers)
Includes initializer containers in the steps list
Properly tracks initializer status

Updated `get_job_logs()`

Supports fetching logs from specific initializer steps
When requesting node-0 logs (default), only shows training container logs
Can explicitly request initializer logs with step="dataset-initializer" or step="model-initializer"

3. Comprehensive Test Coverage (`kubeflow/trainer/backends/container/backend_test.py`)

Added 11 new test cases covering:

Initialization Success Scenarios

HuggingFace dataset initializer - Tests storage_uri parsing and access_token handling
S3 dataset initializer - Tests endpoint, region, and credential configuration
HuggingFace model initializer - Tests model downloads with ignore_patterns
S3 model initializer - Tests S3-compatible storage with authentication
Both dataset and model - Tests sequential execution of both initializers
DataCache initializer - Tests distributed cache configuration with metadata_loc and cluster size

Log Retrieval

Tests getting logs from dataset-initializer step
Tests getting logs from model-initializer step
Tests that default node logs exclude initializer logs

Error Handling

Non-zero exit code - Verifies proper error reporting when initializer fails
Timeout handling - Ensures timeout errors are caught and reported
Resource cleanup - Confirms containers and networks are cleaned up on failure

Implementation Details

Initialization Flow

User calls train() with initializer parameter
ContainerBackend creates:

   ├── Working directory: ~/.kubeflow/trainer/containers/<job-name>/
   └── Network: <job-name>-net

If initializer.dataset is set:

   ├── Create dataset-initializer container
   ├── Mount workdir to /workspace
   ├── Run initialization (downloads to /workspace/dataset)
   └── Wait for completion (exit code 0)

If initializer.model is set:

   ├── Create model-initializer container
   ├── Mount workdir to /workspace
   ├── Run initialization (downloads to /workspace/model)
   └── Wait for completion (exit code 0)

Create training containers:

   ├── Mount same workdir to /workspace
   └── Access data at /workspace/dataset and /workspace/model

Volume Layout

Host: ~/.kubeflow/trainer/containers/<job-name>/

├── dataset/ (from dataset-initializer, if configured)
│   └── <downloaded dataset files>
├── model/ (from model-initializer, if configured)
│   └── <downloaded model files>
└── outputs/ (accessible to all training nodes)
    └── <training outputs>

Container Mount: /workspace/

├── dataset/ (read by training code)
├── model/ (read by training code)
└── outputs/ (written by training code)

Testing Results

All tests pass with no regressions:

43/43 container backend tests passed
173/173 trainer module tests passed
make verify (ruff lint + format) passed
pre-commit hooks passed

Usage Example

from kubeflow.trainer import TrainerClient
from kubeflow.trainer.types import (
    CustomTrainer,
    Initializer,
    HuggingFaceDatasetInitializer,
    HuggingFaceModelInitializer,
)

# Define initializers
initializer = Initializer(
    dataset=HuggingFaceDatasetInitializer(
        storage_uri="hf://username/my-dataset",
        access_token="hf_xxxxx",
    ),
    model=HuggingFaceModelInitializer(
        storage_uri="hf://username/my-model",
        access_token="hf_xxxxx",
    ),
)

# Define trainer
trainer = CustomTrainer(
    func=train_func,
    num_nodes=2,
)

# Train with initializers
client = TrainerClient()
job_name = client.train(
    trainer=trainer,
    initializer=initializer,  # <-- Now works with container backend!
)

# Data is automatically available at:
# - /workspace/dataset (in training containers)
# - /workspace/model (in training containers)

github-actions · 2025-12-04T03:47:08Z

🎉 Welcome to the Kubeflow SDK! 🎉

Thanks for opening your first PR! We're happy to have you as part of our community 🚀

Here's what happens next:

If you haven't already, please check out our Contributing Guide for repo-specific guidelines and the Kubeflow Contributor Guide for general community standards
Our team will review your PR soon! cc @kubeflow/kubeflow-sdk-team

Join the community:

Slack: Join our #kubeflow-ml-experience and #kubeflow-trainer Slack channels
Meetings: Attend the Kubeflow SDK and ML Experience bi-weekly meetings

Feel free to ask questions in the comments if you need any help or clarification!
Thanks again for contributing to Kubeflow! 🙏

Fiona-Waters

Thanks @HKanoje for this PR! I've left some comments, ptal! Thanks

Fiona-Waters · 2026-01-05T11:30:48Z

+    """
+    # Use the training-operator image which contains initializer scripts
+    # This can be made configurable via backend config in the future
+    return "kubeflow/training-operator:latest"


Should we make this configurable rather than hardcoding it?

Yes, Definitely. I've made this configurable via ContainerBackendConfig.initializer_image (default: kubeflow/training-operator:latest). Users can now customize it when creating the backend.

Fiona-Waters · 2026-01-05T11:32:08Z

+        try:
+            import time
+
+            timeout = 600  # 10 minutes timeout for initialization


Should this be configurable, or is 10 minutes always going to be enough time?

Added ContainerBackendConfig.initializer_timeout (default: 600 seconds / 10 minutes). This gives users flexibility for large datasets/models that may take longer to download.

Fiona-Waters · 2026-01-05T11:36:52Z

+            # Clean up the failed container
+            from contextlib import suppress
+
+            with suppress(Exception):


As well as cleaning up when a failure occurs, should we clean up the initializer containers when they have been successful also?

Implemented wait_for_container() in the adapter interface and both Docker/Podman adapters. This replaces the polling loop with a single blocking wait call - much more efficient.

Fiona-Waters · 2026-01-05T11:39:34Z

+            logger.debug(f"Created network: {network_id}")
+
+            # Run initializers if configured
+            if initializer:


If the initializer fails should we clean up the network we have created?

Absolutely! Added cleanup for successful initializer containers after completion to prevent accumulation. Also added cleanup for timed-out containers.

Great! Would it make sense to add a helper function for the cleanup logic to reduce duplication?

Done! Added _cleanup_container_resources() helper method in commit 0b7a952 to consolidate the duplicated cleanup logic across exception handlers and delete_job().

Fiona-Waters · 2026-01-05T11:48:15Z

+        if isinstance(
+            initializer, (types.HuggingFaceDatasetInitializer, types.HuggingFaceModelInitializer)
+        )
+        else "python -m kubeflow.storage_initializer.datacache "


Here we are setting datacache as the default/fallback, do we want to do this? In thekubernetes backend we offer 2 options and raise a value error if the type is invalid.

You're right - that was inconsistent with the kubernetes backend. Changed to raise ValueError with a clear message listing all supported types instead of defaulting to datacache.

Fiona-Waters · 2026-01-05T11:58:41Z

+            import time
+
+            timeout = 600  # 10 minutes timeout for initialization
+            polling_interval = 2


Could using wait API be supported?

kramaranya · 2026-01-05T16:14:57Z

/ok-to-test

- Make initializer image configurable via ContainerBackendConfig - Make initializer timeout configurable (default 600 seconds) - Implement wait API in adapters instead of polling - Clean up successful initializer containers after completion - Clean up network on initializer failure - Raise ValueError for unsupported initializer types (no datacache fallback) All tests passing (173/173). Addresses all feedback from PR kubeflow#188.

kramaranya · 2026-01-05T16:42:09Z

Hey @HKanoje, could you please sign your commits?

coveralls · 2026-01-05T16:43:17Z

Pull Request Test Coverage Report for Build 21811391977

Warning: This coverage report may be inaccurate.

This pull request's base commit is no longer the HEAD commit of its target branch. This means it includes changes from outside the original pull request, including, potentially, unrelated coverage changes.

For more information on this, see Tracking coverage changes with pull request builds.
To avoid this issue with future PRs, see these Recommended CI Configurations.
For a quick fix, rebase this PR at GitHub. Your next report should be accurate.

Details

160 of 195 (82.05%) changed or added relevant lines in 7 files are covered.
21 unchanged lines in 1 file lost coverage.
Overall coverage increased (+0.9%) to 68.424%

Changes Missing Coverage	Covered Lines	Changed/Added Lines	%
kubeflow/trainer/backends/container/adapters/base.py	2	3	66.67%
kubeflow/trainer/backends/container/backend_test.py	59	61	96.72%
kubeflow/trainer/backends/container/utils.py	28	30	93.33%
kubeflow/trainer/backends/container/adapters/podman.py	1	9	11.11%
kubeflow/trainer/backends/container/adapters/docker.py	1	11	9.09%
kubeflow/trainer/backends/container/backend.py	66	78	84.62%

Files with Coverage Reduction	New Missed Lines	%
kubeflow/trainer/backends/kubernetes/utils.py	21	77.64%

Totals
Change from base Build 21637340746:	0.9%
Covered Lines:	2921
Relevant Lines:	4269

💛 - Coveralls

- Make initializer image configurable via ContainerBackendConfig - Make initializer timeout configurable (default 600 seconds) - Implement wait API in adapters instead of polling - Clean up successful initializer containers after completion - Clean up network on initializer failure - Raise ValueError for unsupported initializer types (no datacache fallback) All tests passing (173/173). Addresses all feedback from PR kubeflow#188. Signed-off-by: HKanoje <[email protected]>

HKanoje · 2026-01-05T16:50:49Z

@kramaranya Done! All commits are now signed.

HKanoje · 2026-02-01T19:07:43Z

@kramaranya @szaher Please Review the changes whenever you get a chance! Thanks!

Fiona-Waters

I ran this locally using docker/colima, I have left some comments that should be addressed.
I have created this quick PR in the trainer repo moving a kubernetes import to method level as it was causing an error locally looking for kubeconfig. This PR should be merged after that one. Hope that makes sense. Thanks for your work on this.

Fiona-Waters · 2026-02-03T15:24:42Z

        description="Configuration for training runtime sources",
    )
+    initializer_image: str = Field(
+        default="kubeflow/training-operator:latest",


I tested the initializer functionality with the local Docker/Podman container backend and found some issues relating to the use of this image here that need to be addressed in this PR. This image is a controller image and does not container initializer code. Instead we should use these: kubeflow/dataset-initializer:latest, kubeflow/model-initializer:latest. Functionality should be updated to select whichever one is required for each initialiser.

Those images come from here FYI https://github.com/kubeflow/trainer/tree/master/cmd/initializers

Thanks for catching this! You're right - I've updated the implementation to use the correct images.

Fiona-Waters · 2026-02-03T15:26:27Z

+    elif isinstance(
+        initializer, (types.HuggingFaceDatasetInitializer, types.HuggingFaceModelInitializer)
+    ):
+        python_cmd = "python -m kubeflow.storage_initializer.hugging_face "


Suggested change

python_cmd = "python -m kubeflow.storage_initializer.hugging_face "

python_cmd = "python -m pkg.initializers.dataset"

Fiona-Waters · 2026-02-03T15:29:19Z

+            environment=env,
+            labels=labels,
+            volumes=volumes,
+            working_dir=constants.WORKSPACE_PATH,


This should be /app - https://github.com/kubeflow/trainer/blob/master/cmd/initializers/dataset/Dockerfile#L3

Suggested change

working_dir=constants.WORKSPACE_PATH,

working_dir=/app,

Fixed! I've changed the working_dir from /workspace to /app to match the Dockerfile convention. Added a comment referencing the Dockerfile for future maintainability.

… backend Add support for dataset and model initializers in the container backend to bring it to feature parity with the Kubernetes backend. Changes: - Add utility functions for building initializer commands and environment variables - Implement _run_initializers() and _run_single_initializer() methods in ContainerBackend - Run initializers sequentially before training containers start - Download datasets to /workspace/dataset and models to /workspace/model - Track initializer containers as separate steps in TrainJob - Support all initializer types: HuggingFace, S3, and DataCache - Add comprehensive unit tests for all initializer configurations - Handle initializer failures with proper cleanup and error messages Fixes kubeflow#171 Signed-off-by: HKanoje <[email protected]>

- Make initializer image configurable via ContainerBackendConfig - Make initializer timeout configurable (default 600 seconds) - Implement wait API in adapters instead of polling - Clean up successful initializer containers after completion - Clean up network on initializer failure - Raise ValueError for unsupported initializer types (no datacache fallback) All tests passing (173/173). Addresses all feedback from PR kubeflow#188. Signed-off-by: HKanoje <[email protected]>

Add _cleanup_container_resources() helper method to consolidate duplicated cleanup logic for stopping/removing containers and deleting networks. Refactor 5 locations across train(), initializer handlers, and delete_job() to use this helper. Signed-off-by: HKanoje <[email protected]>

Address feedback for initializer support in container backend: - Use separate images for dataset/model initializers: - kubeflow/dataset-initializer:latest for datasets - kubeflow/model-initializer:latest for models (instead of kubeflow/training-operator:latest) - Update python commands to use pkg.initializers module: - python -m pkg.initializers.dataset (for dataset) - python -m pkg.initializers.model (for model) - Change initializer working_dir from /workspace to /app per Dockerfile convention Refs: https://github.com/kubeflow/trainer/tree/master/cmd/initializers Signed-off-by: HKanoje <[email protected]>

Fiona-Waters · 2026-02-04T17:14:38Z

Thanks for this @HKanoje I've re-run with your latest changes (and the trainer changes) and it works as expected.
/lgtm

andreyvelich

This looks great @HKanoje!
I left a few thoughts.

andreyvelich · 2026-02-05T01:44:31Z

+        # Stop and remove containers
+        if container_ids:
+            for container_id in container_ids:
+                with suppress(Exception):


Why do we use suppress here? We don't do that in other part of SDK.

andreyvelich · 2026-02-05T01:47:47Z

+# Tests for Initializer Support
+@pytest.mark.parametrize(


Can you add these test cases to the test_train() API:

sdk/kubeflow/trainer/backends/container/backend_test.py

Line 470 in 6241fae

def test_train(container_backend, test_case):

andreyvelich · 2026-02-05T01:48:26Z

+        ),
+    ],
+)
+def test_get_logs_with_initializers(container_backend, test_case):


Same question, add it to test_get_job_logs()

andreyvelich · 2026-02-05T01:48:54Z

+        ),
+    ],
+)
+def test_initializer_failures(container_backend, test_case):


This can be added to test_train()

andreyvelich · 2026-02-05T01:49:34Z

+        default="kubeflow/dataset-initializer:latest",
+        description="Container image for dataset initializers",
+    )
+    model_initializer_image: str = Field(
+        default="kubeflow/model-initializer:latest",


Can you use GHCR image by default: https://github.com/kubeflow/trainer/pkgs/container/trainer%2Fdataset-initializer

andreyvelich · 2026-02-05T01:54:07Z

+    elif isinstance(initializer, (types.S3DatasetInitializer, types.S3ModelInitializer)):
+        if initializer.endpoint:
+            env["ENDPOINT"] = initializer.endpoint
+        if initializer.access_key_id:
+            env["ACCESS_KEY_ID"] = initializer.access_key_id
+        if initializer.secret_access_key:
+            env["SECRET_ACCESS_KEY"] = initializer.secret_access_key
+        if initializer.region:
+            env["REGION"] = initializer.region
+        if initializer.role_arn:
+            env["ROLE_ARN"] = initializer.role_arn
+        if hasattr(initializer, "ignore_patterns") and initializer.ignore_patterns:
+            env["IGNORE_PATTERNS"] = ",".join(initializer.ignore_patterns)
+
+    elif isinstance(initializer, types.DataCacheInitializer):
+        env["CLUSTER_SIZE"] = str(initializer.num_data_nodes + 1)
+        env["METADATA_LOC"] = initializer.metadata_loc
+        if initializer.head_cpu:
+            env["HEAD_CPU"] = initializer.head_cpu
+        if initializer.head_mem:
+            env["HEAD_MEM"] = initializer.head_mem
+        if initializer.worker_cpu:
+            env["WORKER_CPU"] = initializer.worker_cpu
+        if initializer.worker_mem:
+            env["WORKER_MEM"] = initializer.worker_mem
+        if initializer.iam_role:
+            env["IAM_ROLE"] = initializer.iam_role


Can we simplify this to something like we do here:

sdk/kubeflow/trainer/backends/kubernetes/utils.py

Lines 574 to 587 in 6241fae

def get_optional_initializer_envs(

initializer: types.BaseInitializer, required_fields: set

) -> list[models.IoK8sApiCoreV1EnvVar]:

"""Get the optional envs from the initializer config"""

envs = []

for f in fields(initializer):

if f.name not in required_fields:

value = getattr(initializer, f.name)

if value is not None:

# Convert list values (like ignore_patterns) to comma-separated strings

if isinstance(value, list):

value = ",".join(str(item) for item in value)

envs.append(models.IoK8sApiCoreV1EnvVar(name=f.name.upper(), value=value))

return envs

Env variables are always have the same name as fields, but just upper cased.

andreyvelich · 2026-02-05T01:55:54Z

    return aggregate_status_from_containers(statuses)
+
+
+def build_initializer_command(initializer: types.BaseInitializer, init_type: str) -> list[str]:


you don't need the init_type variable since you can understand whether it is dataset or model initializer by checking type of initializer field.

Suggested change

def build_initializer_command(initializer: types.BaseInitializer, init_type: str) -> list[str]:

def build_initializer_command(initializer: types.BaseInitializer) -> list[str]:

andreyvelich · 2026-02-05T01:56:54Z

+        types.S3ModelInitializer,
+        types.HuggingFaceDatasetInitializer,
+        types.HuggingFaceModelInitializer,
+        types.DataCacheInitializer,


Let's remove it from now, I am not sure how data cache can be supported in Container backend at the moment.
cc @akshaychitneni

Suggested change

types.DataCacheInitializer,

andreyvelich · 2026-02-05T01:57:23Z

+    return ["bash", "-c", python_cmd]
+
+
+def build_initializer_env(initializer: types.BaseInitializer, init_type: str) -> dict[str, str]:


Same point

Suggested change

def build_initializer_env(initializer: types.BaseInitializer, init_type: str) -> dict[str, str]:

def build_initializer_env(initializer: types.BaseInitializer) -> dict[str, str]:

andreyvelich · 2026-02-05T02:05:28Z

+        # Run dataset initializer if configured
+        if initializer.dataset:
+            # Get and pull dataset initializer image
+            dataset_image = container_utils.get_initializer_image(self.cfg, "dataset")


Can you simplify the logic here to be consistent with what we do in Kubernetes backend:

sdk/kubeflow/trainer/backends/kubernetes/backend.py

Line 704 in 6241fae

dataset=utils.get_dataset_initializer(initializer.dataset)

Just simple define two utils function in container utils:

container_utils.get_dataset_initializer() container_utils.get_model_initializer()

Which returns internal type that you can use in the _adapter.create_and_start_container() API:

@dataclass class ContainerInitializer: image: str command: str env: dict

WDYT @HKanoje @Fiona-Waters ?

- Use GHCR images as default for dataset/model initializers - Replace suppress with try-except blocks - Refactor initializer utils with ContainerInitializer dataclass - Add get_dataset_initializer and get_model_initializer functions - Remove DataCache support (unsupported in container backend) - Merge initializer tests into test_train() and test_get_job_logs() - Remove duplicate test functions Signed-off-by: HKanoje <[email protected]>

HKanoje · 2026-02-05T08:08:43Z

Hi @andreyvelich @Fiona-Waters,

I've addressed all the review comments in the latest commit:

Updated default images to GHCR (ghcr.io/kubeflow/trainer/dataset-initializer:latest and ghcr.io/kubeflow/trainer/model-initializer:latest)
Replaced suppress with try-except blocks (added # noqa: SIM105 to satisfy linter)
Refactored initializer utils with ContainerInitializer dataclass
Added get_dataset_initializer() and get_model_initializer() functions following kubernetes backend pattern
Simplified env building using get_optional_initializer_envs() helper
Removed init_type parameter (now determined from initializer type)
Removed DataCache support (not supported in container backend)
Merged initializer tests into test_train() and test_get_job_logs()
Removed duplicate test functions

andreyvelich · 2026-02-05T14:57:32Z

+        self,
+        job_name: str,
+        container_init: container_utils.ContainerInitializer,
+        init_type: str,


Instead of passing init_type separately, you can simple add name to the ContainerInitializer type which can be:

name = dataset-initializer name = model-initializer

Then, just use this name in the f"{self.label_prefix}/step"

cc @HKanoje

Done! @andreyvelich

Changes made:

Added name field to ContainerInitializer dataclass
Set name="dataset-initializer" in get_dataset_initializer()
Set name="model-initializer" in get_model_initializer()
Removed init_type parameter from _run_single_initializer()
Now using container_init.name for labels and log messages

…t_type - Add name field to ContainerInitializer dataclass - Set name='dataset-initializer' and name='model-initializer' in utils - Remove init_type parameter from _run_single_initializer() - Use container_init.name for labels and log messages Signed-off-by: HKanoje <[email protected]>

andreyvelich

Thanks for this @HKanoje!
Please can you create tracking issue for this: #188 (comment)
/lgtm
/assign @Fiona-Waters @kramaranya

Fiona-Waters

/lgtm
Thanks @HKanoje !

HKanoje · 2026-02-12T05:45:13Z

@andreyvelich Done, I have created the issue #290

andreyvelich · 2026-02-12T10:00:00Z

Thanks! /approve

andreyvelich · 2026-02-12T10:00:45Z

/approve

google-oss-prow · 2026-02-12T10:00:57Z

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: andreyvelich

The full list of commands accepted by this bot can be found here.

The pull request process is described here

Details

Needs approval from an approver in each of these files:

~~OWNERS~~ [andreyvelich]

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

* chore!: upgrade to Python 3.10 (kubeflow#282) This upgrades the minimum Python version for the project from 3.9 to 3.10. Python 3.9 is past end-of-life and dependencies will likely require a supported version soon. Signed-off-by: Jon Burdo <[email protected]> * chore: Confirm that a public ConfigMap exists to check version (kubeflow#250) * Confirm that a public ConfigMap exists to check version Signed-off-by: Surya Sameer Datta Vaddadi <[email protected]> * python 3.9 fix Co-authored-by: Copilot <[email protected]> Signed-off-by: Surya Sameer Datta Vaddadi <[email protected]> * Exceptiom handling better Co-authored-by: Copilot <[email protected]> Signed-off-by: Surya Sameer Datta Vaddadi <[email protected]> * Addressing comments Signed-off-by: Surya Sameer Datta Vaddadi <[email protected]> * Update kubeflow/trainer/backends/kubernetes/backend.py Co-authored-by: Andrey Velichkevich <[email protected]> Signed-off-by: Surya Sameer Datta Vaddadi <[email protected]> * Refactored tests into a single function and followed agents.md Signed-off-by: Surya Sameer Datta Vaddadi <[email protected]> * CI friendly edit Signed-off-by: Surya Sameer Datta Vaddadi <[email protected]> * pre-commit format checked Signed-off-by: Surya Sameer Datta Vaddadi <[email protected]> * Modified according to new updates Signed-off-by: Surya Sameer Datta Vaddadi <[email protected]> * Ran pre-commit locally to fix formatting Signed-off-by: Surya Sameer Datta Vaddadi <[email protected]> * unix2dos CLAUDE.md Signed-off-by: Surya Sameer Datta Vaddadi <[email protected]> * Revert CLAUDE.md Signed-off-by: Surya Sameer Datta Vaddadi <[email protected]> --------- Signed-off-by: Surya Sameer Datta Vaddadi <[email protected]> Signed-off-by: Surya Sameer Datta Vaddadi <[email protected]> Co-authored-by: Copilot <[email protected]> Co-authored-by: Andrey Velichkevich <[email protected]> * chore: added sdk docs website to readme (kubeflow#284) * docs: added sdk docs website to readme Signed-off-by: Akash Jaiswal <[email protected]> * format: order of sdk docs Signed-off-by: Akash Jaiswal <[email protected]> --------- Signed-off-by: Akash Jaiswal <[email protected]> * feat(trainer): add dataset and model initializer support to container backend (kubeflow#188) * feat(trainer): add dataset and model initializer support to container backend Add support for dataset and model initializers in the container backend to bring it to feature parity with the Kubernetes backend. Changes: - Add utility functions for building initializer commands and environment variables - Implement _run_initializers() and _run_single_initializer() methods in ContainerBackend - Run initializers sequentially before training containers start - Download datasets to /workspace/dataset and models to /workspace/model - Track initializer containers as separate steps in TrainJob - Support all initializer types: HuggingFace, S3, and DataCache - Add comprehensive unit tests for all initializer configurations - Handle initializer failures with proper cleanup and error messages Fixes kubeflow#171 Signed-off-by: HKanoje <[email protected]> * feat(trainer): address reviewer feedback for initializer support - Make initializer image configurable via ContainerBackendConfig - Make initializer timeout configurable (default 600 seconds) - Implement wait API in adapters instead of polling - Clean up successful initializer containers after completion - Clean up network on initializer failure - Raise ValueError for unsupported initializer types (no datacache fallback) All tests passing (173/173). Addresses all feedback from PR kubeflow#188. Signed-off-by: HKanoje <[email protected]> * chore(trainer): add cleanup helper to reduce duplication Add _cleanup_container_resources() helper method to consolidate duplicated cleanup logic for stopping/removing containers and deleting networks. Refactor 5 locations across train(), initializer handlers, and delete_job() to use this helper. Signed-off-by: HKanoje <[email protected]> * fix(trainer): use correct initializer images and working directory Address feedback for initializer support in container backend: - Use separate images for dataset/model initializers: - kubeflow/dataset-initializer:latest for datasets - kubeflow/model-initializer:latest for models (instead of kubeflow/training-operator:latest) - Update python commands to use pkg.initializers module: - python -m pkg.initializers.dataset (for dataset) - python -m pkg.initializers.model (for model) - Change initializer working_dir from /workspace to /app per Dockerfile convention Refs: https://github.com/kubeflow/trainer/tree/master/cmd/initializers Signed-off-by: HKanoje <[email protected]> * fix(container): address PR review comments for initializer support - Use GHCR images as default for dataset/model initializers - Replace suppress with try-except blocks - Refactor initializer utils with ContainerInitializer dataclass - Add get_dataset_initializer and get_model_initializer functions - Remove DataCache support (unsupported in container backend) - Merge initializer tests into test_train() and test_get_job_logs() - Remove duplicate test functions Signed-off-by: HKanoje <[email protected]> * fix(container): add name field to ContainerInitializer and remove init_type - Add name field to ContainerInitializer dataclass - Set name='dataset-initializer' and name='model-initializer' in utils - Remove init_type parameter from _run_single_initializer() - Use container_init.name for labels and log messages Signed-off-by: HKanoje <[email protected]> --------- Signed-off-by: HKanoje <[email protected]> * feat: add SparkClient API for SparkConnect session management (kubeflow#225) * feat(spark): add core types, dataclasses, and constants - Add SparkConnectInfo, SparkConnectState, Driver, Executor types - Add type tests for validation - Add Kubernetes backend constants (CRD group, version, defaults) Signed-off-by: Shekhar Rajak <[email protected]> * feat(spark): add backend base class and options pattern - Add RuntimeBackend abstract base class with session lifecycle methods - Add options pattern (Name, Image, Timeout, etc.) aligned with trainer SDK - Add validation utilities for connect parameters - Add comprehensive option tests Signed-off-by: Shekhar Rajak <[email protected]> * feat(spark): add KubernetesBackend for SparkConnect CRD operations - Implement KubernetesBackend with create/get/list/delete session methods - Add port-forward support for out-of-cluster connections - Add CRD builder utilities and URL validation - Add comprehensive backend and utils tests with parametrized patterns Signed-off-by: Shekhar Rajak <[email protected]> * feat(spark): add SparkClient API with KEP-107 compliant connect method - Implement SparkClient as main user interface for SparkConnect sessions - Support connect to existing server (base_url) or auto-create new session - Add public exports for SparkClient, Driver, Executor, options - Add SparkClient unit tests Signed-off-by: Shekhar Rajak <[email protected]> * chore(spark): add test infrastructure and package init files - Add test common utilities and fixtures - Add package __init__ files for test directories - Setup test/e2e/spark structure Signed-off-by: Shekhar Rajak <[email protected]> * feat(spark): add example scripts demonstrating SparkClient usage - Add spark_connect_simple.py with 3 usage levels (minimal, simple, advanced) - Add spark_advanced_options.py with full configuration examples - Add connect_existing_session.py for connecting to existing servers - Add demo and test scripts for local development Signed-off-by: Shekhar Rajak <[email protected]> * docs(spark): add documentation for SparkClient and E2E testing - Add examples/spark/README.md with usage guide - Add local Spark Connect testing documentation - Add E2E test README with CI/CD integration guide - Update KEP-107 proposal documentation Signed-off-by: Shekhar Rajak <[email protected]> * test(spark): add E2E test framework with cluster watcher - Add test_spark_examples.py with example validation tests - Add cluster_watcher.py for monitoring SparkConnect and pods during tests - Add run_in_cluster.py for executing examples as K8s Jobs Signed-off-by: Shekhar Rajak <[email protected]> * ci(spark): add GitHub Actions workflow and E2E cluster setup - Add test-spark-examples.yaml workflow for E2E validation - Add e2e-setup-cluster.sh for Kind cluster with Spark Operator - Add SparkConnect CRD, Kind config, and E2E runner Dockerfile - Update Makefile with E2E setup target - Update PR title check for spark prefix Signed-off-by: Shekhar Rajak <[email protected]> * chore(spark): add pyspark[connect] dependency and update lock file - Add spark extra with pyspark[connect]==3.4.1 for grpcio, pandas, pyarrow - Update uv.lock with resolved dependencies - Update .gitignore for spark-related files Signed-off-by: Shekhar Rajak <[email protected]> * Update kubeflow/spark/backends/base.py Co-authored-by: Andrey Velichkevich <[email protected]> Signed-off-by: Shekhar Prasad Rajak <[email protected]> * refactor(spark): rename backend.connect_session() to connect() Signed-off-by: Shekhar Rajak <[email protected]> * refactor: move session creation flow from SparkClient to backend.create_and_connect() Signed-off-by: Shekhar Rajak <[email protected]> --------- Signed-off-by: Shekhar Rajak <[email protected]> Signed-off-by: Shekhar Prasad Rajak <[email protected]> Co-authored-by: Andrey Velichkevich <[email protected]> * chore: bump minimum model-registry version to 0.3.6 (kubeflow#289) Signed-off-by: Jon Burdo <[email protected]> * fix: Improve CVE workflow (kubeflow#267) * fix: Improve CVE workflow Signed-off-by: Fiona-Waters <[email protected]> * fix: fix issue with bash compare Signed-off-by: Fiona-Waters <[email protected]> * feat: Add workflow to cleanup overrides in pyproject.toml Signed-off-by: Fiona-Waters <[email protected]> * fix: address review comments Signed-off-by: Fiona-Waters <[email protected]> * chore: refactor to reduce size of cve related workflows Signed-off-by: Fiona-Waters <[email protected]> --------- Signed-off-by: Fiona-Waters <[email protected]> * chore: upgrade code style for python3.10 (kubeflow#288) * chore: update code style for Python 3.10 This disables a couple ruff rules in pyproject.toml: ``` "UP007", # Use X | Y instead of Union[X, Y] (requires Python 3.10+) "UP045", # Use X | None instead of Optional[X] (requires Python 3.10+) ``` Then the code changes are made with: ``` uv run ruff check --fix uv run ruff format ``` Signed-off-by: Jon Burdo <[email protected]> * fix: handle unions, bools in convert_value The convert_value function didn't seems to be handling union types properly and also needs to handle `T | None` similarly to `Optional[None]` after the upgrade to Python 3.10. This fixes union types, an issue with bool conversion, and adds tests for this function. Signed-off-by: Jon Burdo <[email protected]> --------- Signed-off-by: Jon Burdo <[email protected]> * chore(ci): bump astral-sh/setup-uv from 5 to 7 (kubeflow#276) Bumps [astral-sh/setup-uv](https://github.com/astral-sh/setup-uv) from 5 to 7. - [Release notes](https://github.com/astral-sh/setup-uv/releases) - [Commits](astral-sh/setup-uv@v5...v7) --- updated-dependencies: - dependency-name: astral-sh/setup-uv dependency-version: '7' dependency-type: direct:production update-type: version-update:semver-major ... Signed-off-by: dependabot[bot] <[email protected]> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com> * chore(deps): bump the python-minor group across 1 directory with 4 updates (kubeflow#291) Bumps the python-minor group with 4 updates in the / directory: [coverage](https://github.com/coveragepy/coveragepy), [ruff](https://github.com/astral-sh/ruff), [pre-commit](https://github.com/pre-commit/pre-commit) and [ty](https://github.com/astral-sh/ty). Updates `coverage` from 7.10.7 to 7.13.4 - [Release notes](https://github.com/coveragepy/coveragepy/releases) - [Changelog](https://github.com/coveragepy/coveragepy/blob/main/CHANGES.rst) - [Commits](coveragepy/coveragepy@7.10.7...7.13.4) Updates `ruff` from 0.14.14 to 0.15.0 - [Release notes](https://github.com/astral-sh/ruff/releases) - [Changelog](https://github.com/astral-sh/ruff/blob/main/CHANGELOG.md) - [Commits](astral-sh/ruff@0.14.14...0.15.0) Updates `pre-commit` from 4.3.0 to 4.5.1 - [Release notes](https://github.com/pre-commit/pre-commit/releases) - [Changelog](https://github.com/pre-commit/pre-commit/blob/main/CHANGELOG.md) - [Commits](pre-commit/pre-commit@v4.3.0...v4.5.1) Updates `ty` from 0.0.14 to 0.0.16 - [Release notes](https://github.com/astral-sh/ty/releases) - [Changelog](https://github.com/astral-sh/ty/blob/main/CHANGELOG.md) - [Commits](astral-sh/ty@0.0.14...0.0.16) --- updated-dependencies: - dependency-name: coverage dependency-version: 7.13.4 dependency-type: direct:development update-type: version-update:semver-minor dependency-group: python-minor - dependency-name: ruff dependency-version: 0.15.0 dependency-type: direct:development update-type: version-update:semver-minor dependency-group: python-minor - dependency-name: pre-commit dependency-version: 4.5.1 dependency-type: direct:development update-type: version-update:semver-minor dependency-group: python-minor - dependency-name: ty dependency-version: 0.0.16 dependency-type: direct:development update-type: version-update:semver-patch dependency-group: python-minor ... Signed-off-by: dependabot[bot] <[email protected]> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com> * feat: Added examples to the documentation demonstrating different ways to handle ports (kubeflow#243) * update docs and add test cases. Signed-off-by: osamaahmed17 <[email protected]> * pre-commit error solved Signed-off-by: osamaahmed17 <[email protected]> * Update kubeflow/hub/api/model_registry_client.py Co-authored-by: Jon Burdo <[email protected]> Signed-off-by: Osama Tahir <[email protected]> * readme updated Signed-off-by: Osama Tahir <[email protected]> * Refactor model registry client test cases for clarity Signed-off-by: Osama Tahir <[email protected]> --------- Signed-off-by: osamaahmed17 <[email protected]> Signed-off-by: Osama Tahir <[email protected]> Co-authored-by: Jon Burdo <[email protected]> * chore(ci): bump peter-evans/create-pull-request from 6 to 8 (kubeflow#277) Bumps [peter-evans/create-pull-request](https://github.com/peter-evans/create-pull-request) from 6 to 8. - [Release notes](https://github.com/peter-evans/create-pull-request/releases) - [Commits](peter-evans/create-pull-request@v6...v8) --- updated-dependencies: - dependency-name: peter-evans/create-pull-request dependency-version: '8' dependency-type: direct:production update-type: version-update:semver-major ... Signed-off-by: dependabot[bot] <[email protected]> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com> * chore(ci): bump actions/checkout from 4 to 6 (kubeflow#278) Bumps [actions/checkout](https://github.com/actions/checkout) from 4 to 6. - [Release notes](https://github.com/actions/checkout/releases) - [Changelog](https://github.com/actions/checkout/blob/main/CHANGELOG.md) - [Commits](actions/checkout@v4...v6) --- updated-dependencies: - dependency-name: actions/checkout dependency-version: '6' dependency-type: direct:production update-type: version-update:semver-major ... Signed-off-by: dependabot[bot] <[email protected]> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com> * feat: Adds a GitHub Actions workflow to check kubeflow/hub/OWNERS. (kubeflow#280) * Add OWNERS validation Signed-off-by: muhammadjunaid8047 <[email protected]> * Update .github/workflows/check-owners.yaml Co-authored-by: Jon Burdo <[email protected]> Signed-off-by: Muhammad Junaid <[email protected]> * Update OWNERS file check in workflow Signed-off-by: Muhammad Junaid <[email protected]> * Update paths in check-owners workflow Signed-off-by: Muhammad Junaid <[email protected]> --------- Signed-off-by: muhammadjunaid8047 <[email protected]> Signed-off-by: Muhammad Junaid <[email protected]> Co-authored-by: Jon Burdo <[email protected]> * fix: nightly security dependency updates (kubeflow#296) Co-authored-by: google-oss-prow <[email protected]> * chore(ci): bump aquasecurity/trivy-action from 0.33.1 to 0.34.0 in the actions group (kubeflow#297) Bumps the actions group with 1 update: [aquasecurity/trivy-action](https://github.com/aquasecurity/trivy-action). Updates `aquasecurity/trivy-action` from 0.33.1 to 0.34.0 - [Release notes](https://github.com/aquasecurity/trivy-action/releases) - [Commits](aquasecurity/trivy-action@0.33.1...0.34.0) --- updated-dependencies: - dependency-name: aquasecurity/trivy-action dependency-version: 0.34.0 dependency-type: direct:production update-type: version-update:semver-minor dependency-group: actions ... Signed-off-by: dependabot[bot] <[email protected]> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com> * chore(deps): bump pytest from 8.4.2 to 9.0.2 (kubeflow#301) Bumps [pytest](https://github.com/pytest-dev/pytest) from 8.4.2 to 9.0.2. - [Release notes](https://github.com/pytest-dev/pytest/releases) - [Changelog](https://github.com/pytest-dev/pytest/blob/main/CHANGELOG.rst) - [Commits](pytest-dev/pytest@8.4.2...9.0.2) --- updated-dependencies: - dependency-name: pytest dependency-version: 9.0.2 dependency-type: direct:development update-type: version-update:semver-major ... Signed-off-by: dependabot[bot] <[email protected]> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com> * feat(trainer): Support namespaced TrainingRuntime in the SDK (kubeflow#130) * feat(backend): Support namespaced TrainingRuntime in the SDK Signed-off-by: Moeed Shaik <[email protected]> * Fixed bugs and validated current test cases Signed-off-by: Moeed Shaik <[email protected]> * Fixed pre-commit test failure Signed-off-by: Moeed Shaik <[email protected]> * Addressed comments Signed-off-by: Moeed Shaik <[email protected]> * Fixed no attribute 'DEFAULT_TIMEOUT' error Signed-off-by: Moeed Shaik <[email protected]> * Added namespace-scoped runtime to test cases Signed-off-by: Moeed Shaik <[email protected]> * Addressed fallback logic bugs Signed-off-by: Moeed Shaik <[email protected]> * Added scope field to Runtime Signed-off-by: Moeed Shaik <[email protected]> * Improved code Signed-off-by: Moeed Shaik <[email protected]> * Fixed copilot's comments Signed-off-by: Moeed Shaik <[email protected]> * Shadow duplicate runtimes, priority to ns Signed-off-by: Moeed Shaik <[email protected]> * Fixed bug Signed-off-by: Moeed Shaik <[email protected]> * Fixed copilot comments Signed-off-by: Moeed Shaik <[email protected]> * Improved test cases to validate all possible cases Signed-off-by: Moeed Shaik <[email protected]> * small fix Signed-off-by: Moeed Shaik <[email protected]> * lint fix Signed-off-by: Moeed Shaik <[email protected]> * improved error message Co-authored-by: Copilot <[email protected]> Signed-off-by: Moeed <[email protected]> * refactored code Signed-off-by: Moeed Shaik <[email protected]> * improve code Signed-off-by: Moeed Shaik <[email protected]> * Removed RuntimeScope Signed-off-by: Moeed Shaik <[email protected]> * removed scope references and improved error handling as per kubeflow standards Signed-off-by: Moeed Shaik <[email protected]> --------- Signed-off-by: Moeed Shaik <[email protected]> Signed-off-by: Moeed <[email protected]> Co-authored-by: Copilot <[email protected]> * fix: Fix runtime lookup fallback and test local SDK in E2E (kubeflow#307) * fix: Install SDK locally in E2E workflow and improve error handling for runtime fetching in Kubernetes backend. Signed-off-by: XploY04 <[email protected]> * refactor: Explicitly return errors from and refine exception handling in . Signed-off-by: XploY04 <[email protected]> * docs: update comment to clarify Kubeflow SDK installation from source in e2e workflow. Signed-off-by: XploY04 <[email protected]> * feat: Enhance runtime retrieval tests to cover Kubernetes API 404/403 errors and partial success for list operations on timeout. Signed-off-by: XploY04 <[email protected]> * refactor: Update runtime listing to immediately raise exceptions on failure instead of collecting partial results. Signed-off-by: XploY04 <[email protected]> --------- Signed-off-by: XploY04 <[email protected]> * chore(ci): bump actions/setup-python from 5 to 6 (kubeflow#298) Bumps [actions/setup-python](https://github.com/actions/setup-python) from 5 to 6. - [Release notes](https://github.com/actions/setup-python/releases) - [Commits](actions/setup-python@v5...v6) --- updated-dependencies: - dependency-name: actions/setup-python dependency-version: '6' dependency-type: direct:production update-type: version-update:semver-major ... Signed-off-by: dependabot[bot] <[email protected]> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com> * chore(deps): bump the python-minor group with 2 updates (kubeflow#299) Bumps the python-minor group with 2 updates: [ruff](https://github.com/astral-sh/ruff) and [ty](https://github.com/astral-sh/ty). Updates `ruff` from 0.15.0 to 0.15.1 - [Release notes](https://github.com/astral-sh/ruff/releases) - [Changelog](https://github.com/astral-sh/ruff/blob/main/CHANGELOG.md) - [Commits](astral-sh/ruff@0.15.0...0.15.1) Updates `ty` from 0.0.16 to 0.0.17 - [Release notes](https://github.com/astral-sh/ty/releases) - [Changelog](https://github.com/astral-sh/ty/blob/main/CHANGELOG.md) - [Commits](astral-sh/ty@0.0.16...0.0.17) --- updated-dependencies: - dependency-name: ruff dependency-version: 0.15.1 dependency-type: direct:development update-type: version-update:semver-patch dependency-group: python-minor - dependency-name: ty dependency-version: 0.0.17 dependency-type: direct:development update-type: version-update:semver-patch dependency-group: python-minor ... Signed-off-by: dependabot[bot] <[email protected]> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com> * fix: improve logging around packages_to_install (kubeflow#269) * improve logging around packages_to_install Signed-off-by: Brian Gallagher <[email protected]> * exit when pip install fails, append errors from both attempts Signed-off-by: Brian Gallagher <[email protected]> * Add shlex to address command injection vulnerabilities. Write pip install logfile to cwd Signed-off-by: Brian Gallagher <[email protected]> --------- Signed-off-by: Brian Gallagher <[email protected]> * feat: Add validate lockfile workflow to complement CVE scanning (kubeflow#306) * feat: Add validate lockfile workflow to complement CVE scanning Signed-off-by: Fiona-Waters <[email protected]> * fix: make cve fix pr branch static Signed-off-by: Fiona-Waters <[email protected]> --------- Signed-off-by: Fiona-Waters <[email protected]> * fix(trainer): handle falsy values in get_args_from_peft_config (kubeflow#328) * fix(trainer): handle falsy values in get_args_from_peft_config Signed-off-by: krishdef7 <[email protected]> * fix: apply pre-commit formatting Signed-off-by: krishdef7 <[email protected]> * fix: also handle falsy train_on_input in dataset_preprocess_config Signed-off-by: krishdef7 <[email protected]> * fix: add missing newline at end of utils_test.py Signed-off-by: krishdef7 <[email protected]> * fix: pre-commit formatting Signed-off-by: krishdef7 <[email protected]> --------- Signed-off-by: krishdef7 <[email protected]> * fix(optimizer): prevent input mutation in optimize() (kubeflow#322) * fix(optimizer): prevent input mutation in optimize() Signed-off-by: ruskaruma <[email protected]> * remove unnecessary things Signed-off-by: ruskaruma <[email protected]> * rename test Signed-off-by: ruskaruma <[email protected]> --------- Signed-off-by: ruskaruma <[email protected]> * feat: add TrainerClient examples for local PyTorch distributed training (kubeflow#312) * docs: add TrainerClient examples for local PyTorch distributed training - Add examples/trainer/pytorch_distributed_simple.py - Add examples/trainer/README.md - Demonstrates LocalProcessBackend usage without Kubernetes - Fixes kubeflow#218 Signed-off-by: Mansi Singh <[email protected]> * docs: add training examples table to SDK website Signed-off-by: Mansi Singh <[email protected]> * docs: expand examples table with PyTorch, MLX, DeepSpeed, and TorchTune examples grouped by framework Signed-off-by: Mansi Singh <[email protected]> --------- Signed-off-by: Mansi Singh <[email protected]> * chore: fix docstrings in TrainerClient (kubeflow#333) Signed-off-by: Transcendental-Programmer <[email protected]> * feat(spark): Refactor unit tests to sdk coding standards (kubeflow#293) * Refactored unit test Signed-off-by: digvijay-y <[email protected]> * Changes made Signed-off-by: digvijay-y <[email protected]> * Version Signed-off-by: digvijay-y <[email protected]> * Restructured clien_test Signed-off-by: digvijay-y <[email protected]> * reformated backend_test.py Signed-off-by: digvijay-y <[email protected]> * revert pyproject.toml and uv.lock changes Signed-off-by: digvijay-y <[email protected]> * Standarized spark backend tests Signed-off-by: digvijay-y <[email protected]> * backend_tests Signed-off-by: digvijay-y <[email protected]> --------- Signed-off-by: digvijay-y <[email protected]> * fix(optimizer): add missing get_job_events() to RuntimeBackend base c… (kubeflow#325) * fix(optimizer): add missing get_job_events() to RuntimeBackend base class Signed-off-by: ruskaruma <[email protected]> * Update kubeflow/optimizer/backends/base.py Co-authored-by: Copilot <[email protected]> Signed-off-by: Ruskaruma <[email protected]> * Update kubeflow/optimizer/backends/base.py Co-authored-by: Andrey Velichkevich <[email protected]> Signed-off-by: Ruskaruma <[email protected]> * fix: add abstractmethod, remove docstrings Signed-off-by: ruskaruma <[email protected]> * make get_job_events abstract in RuntimeBackend Signed-off-by: ruskaruma <[email protected]> * Update kubeflow/trainer/backends/localprocess/backend.py Co-authored-by: Andrey Velichkevich <[email protected]> Signed-off-by: Ruskaruma <[email protected]> * fix Signed-off-by: ruskaruma <[email protected]> --------- Signed-off-by: ruskaruma <[email protected]> Signed-off-by: Ruskaruma <[email protected]> Co-authored-by: Copilot <[email protected]> Co-authored-by: Andrey Velichkevich <[email protected]> * chore(spark): migrate SDK to kubeflow_spark_api Pydantic models (kubeflow#295) * chore(spark): add kubeflow-spark-api dependency Signed-off-by: tariq-hasan <[email protected]> * chore(spark): migrate options to typed Pydantic models Signed-off-by: tariq-hasan <[email protected]> * chore(spark): migrate utils to typed Pydantic models Signed-off-by: tariq-hasan <[email protected]> * chore(spark): migrate backend to typed Pydantic models Signed-off-by: tariq-hasan <[email protected]> * chore(spark): refactor tests to use typed models and cleanup Signed-off-by: tariq-hasan <[email protected]> * chore(spark): rename build_spark_connect_crd to build_spark_connect_cr Signed-off-by: tariq-hasan <[email protected]> * fix(spark): use typed model helpers in mock handlers Signed-off-by: tariq-hasan <[email protected]> * chore(spark): bump kubeflow-spark-api to 2.4.0 Signed-off-by: tariq-hasan <[email protected]> --------- Signed-off-by: tariq-hasan <[email protected]> * feat(docs): Update README with Spark Support (kubeflow#349) Signed-off-by: Andrey Velichkevich <[email protected]> * fix(trainer): return TRAINJOB_COMPLETE when all steps are done (kubeflow#340) * fix(local): return TRAINJOB_COMPLETE when all steps are done (kubeflow#338) Signed-off-by: priyank <[email protected]> * test(trainer): add test case for __get_job_status Signed-off-by: priyank <[email protected]> * fix(trainer): early return TRAINJOB_CREATED when job has no steps Signed-off-by: priyank <[email protected]> * test(trainer): refactor test_get_job_status with TestCase fixture Signed-off-by: priyank <[email protected]> --------- Signed-off-by: priyank <[email protected]> * fix(trainer): adapt SDK to removal of numProcPerNode from TorchMLPolicySource (kubeflow#360) Signed-off-by: tariq-hasan <[email protected]> * fix: Make validate-lockfile action non-blocking (kubeflow#361) Signed-off-by: Fiona-Waters <[email protected]> * chore(spark): change pyspark[connect] dependency (kubeflow#357) Change pyspark[connect] 3.4.1 dependency to pyspark-connect 4.0.1. This matches the version of Spark in the spark-operator container image (https://github.com/kubeflow/spark-operator/blob/master/Dockerfile#L17). Signed-off-by: Ali Maredia <[email protected]> * chore(spark): remove SDK-side validation from SparkClient (kubeflow#345) Remove all SDK-side input validation from the spark module. Validation will be handled server-side by the Spark Operator admission webhooks (spark-operator#2862). - Remove validation.py and validation_test.py - Remove isinstance checks from _create_session() - Remove ValidationError from public API Closes: kubeflow#272 Signed-off-by: Yassin Nouh <[email protected]> Signed-off-by: yassinnouh21 <[email protected]> * chore: Merge upstream/main (preserving downstream config) Signed-off-by: Brian Gallagher <[email protected]> * update workflow to skip requirements generation on merge conflict Signed-off-by: Brian Gallagher <[email protected]> * remove compatibility with python 3.9 and udpated tests Signed-off-by: Brian Gallagher <[email protected]> * fix tests Signed-off-by: Brian Gallagher <[email protected]> --------- Signed-off-by: Jon Burdo <[email protected]> Signed-off-by: Surya Sameer Datta Vaddadi <[email protected]> Signed-off-by: Surya Sameer Datta Vaddadi <[email protected]> Signed-off-by: Akash Jaiswal <[email protected]> Signed-off-by: HKanoje <[email protected]> Signed-off-by: Shekhar Rajak <[email protected]> Signed-off-by: Shekhar Prasad Rajak <[email protected]> Signed-off-by: Fiona-Waters <[email protected]> Signed-off-by: dependabot[bot] <[email protected]> Signed-off-by: osamaahmed17 <[email protected]> Signed-off-by: Osama Tahir <[email protected]> Signed-off-by: muhammadjunaid8047 <[email protected]> Signed-off-by: Muhammad Junaid <[email protected]> Signed-off-by: Moeed Shaik <[email protected]> Signed-off-by: Moeed <[email protected]> Signed-off-by: XploY04 <[email protected]> Signed-off-by: Brian Gallagher <[email protected]> Signed-off-by: krishdef7 <[email protected]> Signed-off-by: ruskaruma <[email protected]> Signed-off-by: Mansi Singh <[email protected]> Signed-off-by: Transcendental-Programmer <[email protected]> Signed-off-by: digvijay-y <[email protected]> Signed-off-by: Ruskaruma <[email protected]> Signed-off-by: tariq-hasan <[email protected]> Signed-off-by: Andrey Velichkevich <[email protected]> Signed-off-by: priyank <[email protected]> Signed-off-by: Ali Maredia <[email protected]> Signed-off-by: Yassin Nouh <[email protected]> Signed-off-by: yassinnouh21 <[email protected]> Co-authored-by: Jon Burdo <[email protected]> Co-authored-by: Surya Sameer Datta Vaddadi <[email protected]> Co-authored-by: Copilot <[email protected]> Co-authored-by: Andrey Velichkevich <[email protected]> Co-authored-by: Akash Jaiswal <[email protected]> Co-authored-by: Hrithik Kanoje <[email protected]> Co-authored-by: Shekhar Prasad Rajak <[email protected]> Co-authored-by: Fiona Waters <[email protected]> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com> Co-authored-by: Osama Tahir <[email protected]> Co-authored-by: Muhammad Junaid <[email protected]> Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> Co-authored-by: google-oss-prow <[email protected]> Co-authored-by: Moeed <[email protected]> Co-authored-by: Yash Agarwal <[email protected]> Co-authored-by: krishdef7 <[email protected]> Co-authored-by: Ruskaruma <[email protected]> Co-authored-by: Mansi Singh <[email protected]> Co-authored-by: Priyansh Saxena <[email protected]> Co-authored-by: DIGVIJAY <[email protected]> Co-authored-by: Tariq Hasan <[email protected]> Co-authored-by: Priyank Patel <[email protected]> Co-authored-by: Ali Maredia <[email protected]> Co-authored-by: Yassin Nouh <[email protected]>

Add docs/source/train/initializers.rst documenting the dataset and model initializer types (HuggingFaceDatasetInitializer, S3DatasetInitializer, DataCacheInitializer, HuggingFaceModelInitializer, S3ModelInitializer) that were added to the container backend in PRs kubeflow#188 and kubeflow#313. The guide covers: concept overview, per-initializer code examples, combined usage, ContainerBackendConfig options (images, timeout), log-based debugging, and backend limitations. Co-Authored-By: Claude Sonnet 4.6 <[email protected]>

Add docs/source/train/initializers.rst covering dataset and model initializers for the container backend (added in kubeflow#188, parallelised in kubeflow#313). Includes per-type code examples, combined usage, ContainerBackendConfig options, and debugging via get_job_logs().

Add docs/source/train/initializers.rst covering dataset and model initializers for the container backend (added in kubeflow#188, parallelised in kubeflow#313). Includes per-type code examples, combined usage, ContainerBackendConfig options, and debugging via get_job_logs(). Signed-off-by: Ayush Petwal <[email protected]>

* chore(trainer): add data and model initializers guide Add docs/source/train/initializers.rst covering dataset and model initializers for the container backend (added in #188, parallelised in #313). Includes per-type code examples, combined usage, ContainerBackendConfig options, and debugging via get_job_logs(). Signed-off-by: Ayush Petwal <[email protected]> * docs(trainer): address PR review comments on initializers guide - Fix model output path from /workspace/model-weights to /workspace/model to match the MODEL_PATH constant in constants.py - Clarify DataCacheInitializer is Kubernetes-only in the backend note and annotate the Available Initializers table row accordingly - Add DataCacheInitializer usage example with required fields (storage_uri, metadata_loc, num_data_nodes) and backend constraint note Signed-off-by: 1Ayush-Petwal <[email protected]> * docs(trainer): use latest tag for initializer images in guide Signed-off-by: 1Ayush-Petwal <[email protected]> --------- Signed-off-by: Ayush Petwal <[email protected]> Signed-off-by: 1Ayush-Petwal <[email protected]>

google-oss-prow Bot requested review from kramaranya and szaher December 4, 2025 03:47

google-oss-prow Bot added the size/XL label Dec 4, 2025

kramaranya mentioned this pull request Dec 4, 2025

Add welcome workflow for new contributors kubeflow/trainer#3016

Closed

Fiona-Waters reviewed Jan 5, 2026

View reviewed changes

google-oss-prow Bot added the ok-to-test label Jan 5, 2026

HKanoje force-pushed the feat/171-add-initializer-support-container-backend branch from 768a6a9 to 0dbb6b6 Compare January 5, 2026 16:49

Fiona-Waters mentioned this pull request Feb 3, 2026

fix: Defer kubernetes imports to method level for use with local mode kubeflow/trainer#3167

Merged

1 task

Fiona-Waters reviewed Feb 3, 2026

View reviewed changes

HKanoje added 4 commits February 3, 2026 23:45

HKanoje force-pushed the feat/171-add-initializer-support-container-backend branch from 2bf8662 to 6241fae Compare February 4, 2026 07:45

google-oss-prow Bot assigned Fiona-Waters Feb 4, 2026

google-oss-prow Bot added the lgtm label Feb 4, 2026

andreyvelich reviewed Feb 5, 2026

View reviewed changes

google-oss-prow Bot removed the lgtm label Feb 5, 2026

HKanoje force-pushed the feat/171-add-initializer-support-container-backend branch from e214292 to a1c9ce0 Compare February 5, 2026 08:06

andreyvelich reviewed Feb 5, 2026

View reviewed changes

andreyvelich reviewed Feb 11, 2026

View reviewed changes

google-oss-prow Bot assigned kramaranya and andreyvelich Feb 11, 2026

google-oss-prow Bot added the lgtm label Feb 11, 2026

Fiona-Waters reviewed Feb 11, 2026

View reviewed changes

HKanoje mentioned this pull request Feb 12, 2026

feat(container): Run dataset and model initializers in parallel #290

Closed

google-oss-prow Bot added the approved label Feb 12, 2026

google-oss-prow Bot merged commit deed1ce into kubeflow:main Feb 12, 2026
17 checks passed

google-oss-prow Bot added this to the v0.4 milestone Feb 12, 2026

HKanoje deleted the feat/171-add-initializer-support-container-backend branch February 13, 2026 05:26

Fiona-Waters mentioned this pull request Mar 21, 2026

Document data and model initializers #413

Closed

	python_cmd = "python -m kubeflow.storage_initializer.hugging_face "
	python_cmd = "python -m pkg.initializers.dataset"

	def get_optional_initializer_envs(
	initializer: types.BaseInitializer, required_fields: set
	) -> list[models.IoK8sApiCoreV1EnvVar]:
	"""Get the optional envs from the initializer config"""
	envs = []
	for f in fields(initializer):
	if f.name not in required_fields:
	value = getattr(initializer, f.name)
	if value is not None:
	# Convert list values (like ignore_patterns) to comma-separated strings
	if isinstance(value, list):
	value = ",".join(str(item) for item in value)
	envs.append(models.IoK8sApiCoreV1EnvVar(name=f.name.upper(), value=value))
	return envs

		return aggregate_status_from_containers(statuses)


		def build_initializer_command(initializer: types.BaseInitializer, init_type: str) -> list[str]:

		return ["bash", "-c", python_cmd]


		def build_initializer_env(initializer: types.BaseInitializer, init_type: str) -> dict[str, str]:

Uh oh!

Conversation

HKanoje commented Dec 4, 2025

What this PR does / why we need it

Solution Overview

Detailed Changes

1. New Utility Functions (kubeflow/trainer/backends/container/utils.py)

build_initializer_command(initializer, init_type)

build_initializer_env(initializer, init_type)

get_initializer_image()

2. Enhanced ContainerBackend (kubeflow/trainer/backends/container/backend.py)

_run_initializers(job_name, initializer, workdir, network_id)

_run_single_initializer(job_name, initializer_config, init_type, image, workdir, network_id)

Updated train() method

Updated __get_trainjob_from_containers()

Updated get_job_logs()

3. Comprehensive Test Coverage (kubeflow/trainer/backends/container/backend_test.py)

Initialization Success Scenarios

Log Retrieval

Error Handling

Implementation Details

Initialization Flow

Volume Layout

Testing Results

Usage Example

Uh oh!

github-actions Bot commented Dec 4, 2025

Uh oh!

Fiona-Waters 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!

kramaranya commented Jan 5, 2026

Uh oh!

kramaranya commented Jan 5, 2026

Uh oh!

coveralls commented Jan 5, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Pull Request Test Coverage Report for Build 21811391977

Warning: This coverage report may be inaccurate.

Details

💛 - Coveralls

Uh oh!

HKanoje commented Jan 5, 2026

Uh oh!

HKanoje commented Feb 1, 2026

Uh oh!

Fiona-Waters 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

1. New Utility Functions (`kubeflow/trainer/backends/container/utils.py`)

`build_initializer_command(initializer, init_type)`

`build_initializer_env(initializer, init_type)`

`get_initializer_image()`

2. Enhanced ContainerBackend (`kubeflow/trainer/backends/container/backend.py`)

`_run_initializers(job_name, initializer, workdir, network_id)`

`_run_single_initializer(job_name, initializer_config, init_type, image, workdir, network_id)`

Updated `train()` method

Updated `__get_trainjob_from_containers()`

Updated `get_job_logs()`

3. Comprehensive Test Coverage (`kubeflow/trainer/backends/container/backend_test.py`)

coveralls commented Jan 5, 2026 •

edited

Loading

andreyvelich Feb 5, 2026 •

edited

Loading