FEAT: add job runner for vertex ai custom training job (#18)

cmd/job-runner/.env

@@ -0,0 +1 @@
+GCS_PATH="/gcs/bucket/path"

cmd/job-runner/.gitignore

@@ -0,0 +1 @@
+!.env

cmd/job-runner/.python-version

@@ -0,0 +1 @@
+3.12.8

cmd/job-runner/Makefile

@@ -0,0 +1,26 @@
+.PHONY: lint fmt help test lock setup train
+.DEFAULT_GOAL := help
+
+lint: ## Run Linter
+	uv run ruff check .
+	uv run mypy .
+
+fmt: ## Run formatter
+	uv run ruff check --fix .
+	uv run ruff format .
+
+test: ## Run tests
+	uv run pytest .
+
+lock: ## Lock dependencies
+	uv lock
+
+install: ## Setup the project
+	uv sync --all-groups
+
+train: ## Train the model
+	uv run python train.py
+
+help: ## Show options
+	@grep -E '^[a-zA-Z_-]+:.*?## .*$$' $(MAKEFILE_LIST) | \
+		awk 'BEGIN {FS = ":.*?## "}; {printf "\033[36m%-20s\033[0m %s\n", $$1, $$2}'

cmd/job-runner/README.md

@@ -0,0 +1,9 @@
+# Vertex AI Custom Training Job Runner
+
+A command line utility for running machine learning training jobs on Google Cloud Vertex AI.
+
+## Usage
+
+```sh
+uv run main.py --args=...
+```

cmd/job-runner/main.py

@@ -0,0 +1,31 @@
+import pathlib
+from typing import Literal
+
+from loguru import logger
+from pydantic import Field
+from pydantic_settings import BaseSettings, CliApp, SettingsConfigDict
+
+
+class Settings(BaseSettings):
+    """Settings for the job-runner"""
+
+    model_config = SettingsConfigDict(
+        cli_parse_args=True, cli_prog_name="job-runner", env_file=".env", env_file_encoding="utf-8"
+    )
+
+    # from environmental variables
+    # FIXME: これも必須parameterになり、CLIから指定する必要があるように見える
+    gcs_path: pathlib.Path = Field(description="[ENV] Google Cloud Storage path to the model")
+
+    # from args
+    machine_type: Literal["g2-instance=4", "g2-instance=12"] = Field(
+        description="Machine type to use for the job"
+    )
+
+    def cli_cmd(self) -> None:
+        logger.info("Running the job-runner cli command")
+        logger.info(self.model_dump())
+
+
+if __name__ == "__main__":
+    s = CliApp.run(Settings)

cmd/job-runner/pyproject.toml

@@ -0,0 +1,35 @@
+[project]
+name = "job-runner"
+version = "0.1.0"
+description = "Command line utilities to run ml training job"
+readme = "README.md"
+requires-python = ">=3.12"
+dependencies = ["loguru~=0.7.3", "pydantic~=2.10.4", "pydantic-settings~=2.7.1"]
+
+[dependency-groups]
+dev = ["pytest~=8.3.4", "mypy~=1.14.1"]
+lint = ["ruff~=0.8.4"]
+
+[tool.ruff]
+target-version = "py312"
+line-length = 100
+
+[tool.ruff.lint]
+extend-select = ["I"]
+
+[tool.mypy]
+python_version = "3.12"
+exclude = [".venv"]
+plugins = ["pydantic.mypy"]
+
+follow_imports = "silent"
+warn_redundant_casts = true
+warn_unused_ignores = true
+disallow_any_generics = true
+no_implicit_reexport = true
+disallow_untyped_defs = true
+
+[tool.pydantic-mypy]
+init_forbid_extra = true
+init_typed = true
+warn_required_dynamic_aliases = true