chore: improve trainer docstrings and contributing guide formatting

CLAUDE.md

@@ -1 +1 @@
-AGENTS.md
+AGENTS.md

CONTRIBUTING.md

@@ -4,11 +4,11 @@ This guide explains how to contribute to the Kubeflow SDK project.
 For the Kubeflow SDK documentation, please check [the official Kubeflow documentation](https://www.kubeflow.org/docs/components/).
 
 ## Requirements
+
 - [Supported Python version](./pyproject.toml#L4)
 - [pre-commit](https://pre-commit.com/)
 - [uv](https://docs.astral.sh/uv/getting-started/installation/)
 
-
 ## Development
 
 The Kubeflow SDK project includes a Makefile with several helpful commands to streamline your development workflow.
@@ -20,6 +20,7 @@ make install-dev
 ```
 
 ### Coding Style
+
 Make sure to install [pre-commit](https://pre-commit.com/) (`uv pip install pre-commit`) and run `pre-commit install` from the root of the repository at least once before creating git commits.
 
 The pre-commit hooks ensure code quality and consistency. They are executed in CI. PRs that fail to comply with the hooks will not be able to pass the corresponding CI gate. The hooks are only executed against staged files unless you run `pre-commit run --all`, in which case, they'll be executed against every file in the repository.
@@ -37,18 +38,19 @@ make verify
 The Kubeflow SDK project includes several types of tests to ensure code quality and functionality.
 
 ### Unit Testing
+
 To run unit tests locally use the following make command:
 
 ```shell
 make test-python
 ```
 
 ### E2E Tests
+
 E2E test run in CI on a kind cluster using [Kubeflow Trainer E2E Scripts](https://github.com/kubeflow/trainer/blob/master/CONTRIBUTING.md#e2e-tests).
 Clone the `Kubeflow Trainer` repo and run the provided commands against `Trainer` Makefile.
 For more details check [the Kubeflow Trainer Contributing Guide](https://github.com/kubeflow/trainer/blob/master/CONTRIBUTING.md#e2e-tests).
 
-
 ## Best Practices
 
 ### Pull Request Title Conventions

kubeflow/trainer/api/trainer_client.py

@@ -75,7 +75,7 @@ def list_runtimes(self) -> list[types.Runtime]:
         return self.backend.list_runtimes()
 
     def get_runtime(self, name: str) -> types.Runtime:
-        """Get the runtime object
+        """Get the runtime object.
 
         Args:
             name: Name of the runtime.
@@ -165,7 +165,7 @@ def list_jobs(self, runtime: types.Runtime | None = None) -> list[types.TrainJob
         return self.backend.list_jobs(runtime=runtime)
 
     def get_job(self, name: str) -> types.TrainJob:
-        """Get the TrainJob object
+        """Get the TrainJob object.
 
         Args:
             name: Name of the TrainJob.
@@ -204,7 +204,6 @@ def get_job_logs(
         Returns:
             Iterator of log lines.
 
-
         Raises:
             TimeoutError: Timeout to get a TrainJob.
             RuntimeError: Failed to get a TrainJob.

kubeflow/trainer/types/types.py

@@ -152,7 +152,7 @@ class TorchTuneInstructDataset:
 @dataclass
 class LoraConfig:
     """Configuration for the LoRA/QLoRA/DoRA.
-    REF: https://meta-pytorch.org/torchtune/main/tutorials/memory_optimizations.html
+    REF: https://pytorch.org/torchtune/main/tutorials/lora_finetune.html
 
     Args:
         apply_lora_to_mlp (`Optional[bool]`):
@@ -200,7 +200,7 @@ class TorchTuneConfig:
         batch_size (`Optional[int]`):
             The number of samples processed before updating model weights.
         epochs (`Optional[int]`):
-            The number of samples processed before updating model weights.
+            The number of times the full training dataset is passed through the model.
         loss (`Optional[Loss]`): The loss algorithm we use to fine-tune the LLM,
             e.g. `torchtune.modules.loss.CEWithChunkedOutputLoss`.
         num_nodes (`Optional[int]`): The number of nodes to use for training.