catalyst_simple.py

"""
Optuna example that optimizes multi-layer perceptrons using Catalyst.

In this example, we optimize the validation accuracy of fashion product recognition using
Catalyst, and FashionMNIST. We optimize the neural network architecture.

You can run this example as follows, pruning can be turned on and off with the `--pruning`
argument.
    $ python catalyst_simple.py [--pruning]

[Why OptunaPruningCallback]
In this example, we use Catalyst's `OptunaPruningCallback` instead of `CatalystPruningCallback`
since `CatalystPruningCallback` is deprecated and a thin wrapper of `OptunaPruningCallback`.

See also: https://catalyst-team.github.io/catalyst/api/callbacks.html?highlight=optuna#catalyst.callbacks.optuna.OptunaPruningCallback
"""  # NOQA

import argparse
import os

from catalyst.dl import AccuracyCallback
from catalyst.dl import OptunaPruningCallback
from catalyst.dl import SupervisedRunner
import optuna
import torch
from torch import nn
from torch.utils.data import DataLoader
from torchvision import datasets
from torchvision import transforms


CLASSES = 10


def define_model(trial: optuna.trial.Trial) -> nn.Sequential:
    n_layers = trial.suggest_int("n_layers", 1, 3)
    dropout = trial.suggest_float("dropout", 0.2, 0.5)
    input_dim = 28 * 28
    layers = [nn.Flatten()]
    for i in range(n_layers):
        output_dim = trial.suggest_int("n_units_l{}".format(i), 4, 128, log=True)
        layers.append(nn.Linear(input_dim, output_dim))
        layers.append(nn.ReLU())
        layers.append(nn.Dropout(dropout))

        input_dim = output_dim
    layers.append(nn.Linear(input_dim, CLASSES))

    return nn.Sequential(*layers)


loaders = {
    "train": DataLoader(
        datasets.FashionMNIST(
            os.getcwd(), train=True, download=True, transform=transforms.ToTensor()
        ),
        batch_size=100,
        shuffle=True,
    ),
    "valid": DataLoader(
        datasets.FashionMNIST(
            os.getcwd(), train=False, download=True, transform=transforms.ToTensor()
        ),
        batch_size=100,
    ),
}


def objective(trial):
    logdir = "./logdir"
    num_epochs = 10

    model = define_model(trial)
    optimizer = torch.optim.Adam(model.parameters(), lr=0.02)
    criterion = torch.nn.CrossEntropyLoss()

    # model training
    runner = SupervisedRunner()
    runner.train(
        model=model,
        criterion=criterion,
        optimizer=optimizer,
        loaders=loaders,
        logdir=logdir,
        num_epochs=num_epochs,
        verbose=True,
        callbacks={
            # NOTE(crcrpar): Consult above [Why OptunaPruningCallback] for the use of
            # Catalys's callback for Optuna, not Optuna's one for Catalyst.
            # top-1 accuracy as metric for pruning
            "optuna": OptunaPruningCallback(
                loader_key="valid",
                metric_key="accuracy01",
                minimize=False,
                trial=trial,
            ),
            "accuracy": AccuracyCallback(
                input_key="logits",
                target_key="targets",
                num_classes=10,
            ),
        },
    )

    return runner.callbacks["optuna"].best_score


if __name__ == "__main__":
    parser = argparse.ArgumentParser(description="Catalyst example.")
    parser.add_argument(
        "--pruning",
        "-p",
        action="store_true",
        help="Activate the pruning feature. `MedianPruner` stops unpromising "
        "trials at the early stages of training.",
    )
    args = parser.parse_args()

    pruner = optuna.pruners.MedianPruner() if args.pruning else optuna.pruners.NopPruner()

    study = optuna.create_study(direction="maximize", pruner=pruner)
    study.optimize(objective, n_trials=10, timeout=600)

    print("Number of finished trials: {}".format(len(study.trials)))

    print("Best trial:")
    trial = study.best_trial

    print("  Value: {}".format(trial.value))

    print("  Params: ")
    for key, value in trial.params.items():
        print("    {}: {}".format(key, value))