task.py

# SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
# SPDX-License-Identifier: Apache-2.0
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

from dataclasses import dataclass
from typing import List

import nemo_run as run


@dataclass
class Model:
    """Dummy model config"""

    hidden_size: int
    num_layers: int
    activation: str


@dataclass
class Optimizer:
    """Dummy optimizer config"""

    learning_rate: float
    weight_decay: float
    betas: List[float]


@run.cli.factory
@run.autoconvert
def my_model(hidden_size: int = 256, num_layers: int = 3, activation: str = "relu") -> Model:
    """
    Create a model configuration.
    """
    return Model(hidden_size=hidden_size, num_layers=num_layers, activation=activation)


@run.cli.factory
def my_optimizer(
    learning_rate: float = 0.001, weight_decay: float = 1e-5, betas: List[float] = [0.9, 0.999]
) -> run.Config[Optimizer]:
    """Create an optimizer configuration."""
    return run.Config(
        Optimizer, learning_rate=learning_rate, weight_decay=weight_decay, betas=betas
    )


def train_model(
    model: Model,
    optimizer: Optimizer,
    epochs: int = 10,
    batch_size: int = 32,
):
    """
    Train a model using the specified configuration.

    Args:
        model (Model): Configuration for the model.
        optimizer (Optimizer): Configuration for the optimizer.
        epochs (int, optional): Number of training epochs. Defaults to 10.
        batch_size (int, optional): Batch size for training. Defaults to 32.
    """
    print("Training model with the following configuration:")
    print(f"Model: {model}")
    print(f"Optimizer: {optimizer}")
    print(f"Epochs: {epochs}")
    print(f"Batch size: {batch_size}")

    # Simulating model training
    for epoch in range(epochs):
        print(f"Epoch {epoch + 1}/{epochs}")

    print("Training completed!")


@run.cli.factory(target=train_model)
def train_recipe() -> run.Partial[train_model]:
    return run.Partial(
        train_model,
        model=my_model(hidden_size=512),
        optimizer=my_optimizer(learning_rate=0.0005),
        epochs=50,
        batch_size=2048,
    )


if __name__ == "__main__":
    run.cli.main(train_model, cmd_defaults={"skip_confirmation": True})