optuna-lightning-cli 0.1.0

A Typer CLI application for tuning Lightning modules with Optuna.

optuna-lightning-cli

A small Typer CLI for tuning lightning.pytorch modules with Optuna.

Install the MNIST example extra and run the sample HPO flow:

pip install -e ".[examples]"

optuna-lightning tune \
  --training-config examples/mnist-training.yaml \
  --optuna-config examples/mnist-optuna.yaml

MnistClassifier owns the training loop (training_step, validation_step, and configure_optimizers). MnistDataModule owns the MNIST download and data loaders. The CLI validates the config pair, patches Optuna samples into the flat training config, and hands everything to Trainer.fit().

print-config renders the normalized Lightning config under a Lightning Base Config header and the Optuna config under an Optuna Config header.

The CLI also includes config printing, validation, and persisted study inspection:

optuna-lightning print-config \
  --training-config examples/mnist-training.yaml \
  --optuna-config examples/mnist-optuna.yaml

optuna-lightning validate \
  --training-config examples/mnist-training.yaml \
  --optuna-config examples/mnist-optuna.yaml

optuna-lightning studies list --storage sqlite:///optuna.db
optuna-lightning studies show \
  --storage sqlite:///optuna.db \
  --study-name mnist
optuna-lightning studies trials \
  --optuna-config examples/mnist-optuna.yaml

The example Optuna config stores studies in ./optuna.db relative to the directory where the command is run, so studies trials --optuna-config examples/mnist-optuna.yaml can inspect persisted trials later. The trials table labels the value column as Objective [val_acc, maximize] for the MNIST example, and storage-only inspection still shows the study direction in the column header.