Examples

Complete examples for various use cases.

Image Classification

CIFAR-10 with ResNet

See Tutorials for a complete CIFAR-10 classification example.

Transfer Learning

Fine-tune a pre-trained ResNet50 on a custom dataset:

cd scripts
python torch_trainer_mnist_example.py

Semantic Segmentation

Binary Segmentation

Road segmentation example:

cd notebooks
jupyter notebook Road_Segmentation_Example.ipynb

Multiclass Segmentation

Scene segmentation with multiple classes:

cd notebooks
jupyter notebook Multiclass_Scene_Segmentation.ipynb

Or run the script:

cd scripts
python fabric_trainer_multiclass_segmentation_example.py

Image Regression

Age/Depth Estimation

cd notebooks
jupyter notebook Image_Regression_Example.ipynb

Distributed Training

FabricTrainer Example

cd scripts
python fabric_trainer_mnist_example.py

AcceleratorTrainer Example

cd scripts
python accelerator_trainer_mnist_example.py

Multi-GPU Training

# Automatically uses all available GPUs
python torch_trainer_binary_segmentation_example.py

Experiment Tracking

MLflow Integration

Check the Experiment Tracking documentation for MLflow examples.

Weights & Biases

Examples with wandb integration are available in the scripts directory.

Custom Metrics

Implementing Custom Metrics

See Metrics for examples of custom metric implementations.

Project Structure

Recommended Structure

my_project/
├── data/
│   ├── train/
│   ├── val/
│   └── test/
├── checkpoints/
├── runs/  # TensorBoard logs
├── mlruns/  # MLflow logs
├── notebooks/
│   └── exploration.ipynb
├── src/
│   ├── models/
│   │   ├── __init__.py
│   │   └── custom_model.py
│   ├── datasets/
│   │   ├── __init__.py
│   │   └── custom_dataset.py
│   └── utils/
│       ├── __init__.py
│       └── helpers.py
├── train.py
├── evaluate.py
├── predict.py
└── requirements.txt

Example train.py

import argparse
import torch
from torch.utils.data import DataLoader
from deepml.tasks import ImageClassification
from deepml.fabric_trainer import FabricTrainer

def main(args):
    # Data
    train_dataset = create_dataset(args.data_dir, split='train')
    val_dataset = create_dataset(args.data_dir, split='val')

    train_loader = DataLoader(
        train_dataset,
        batch_size=args.batch_size,
        shuffle=True,
        num_workers=args.num_workers
    )

    val_loader = DataLoader(
        val_dataset,
        batch_size=args.batch_size,
        num_workers=args.num_workers
    )

    # Model
    model = create_model(
        arch=args.arch,
        num_classes=args.num_classes,
        pretrained=args.pretrained
    )

    # Task
    task = ImageClassification(
        model=model,
        model_dir=args.checkpoint_dir
    )

    # Training setup
    optimizer = torch.optim.Adam(
        model.parameters(),
        lr=args.lr,
        weight_decay=args.weight_decay
    )

    criterion = torch.nn.CrossEntropyLoss()

    # Trainer
    trainer = FabricTrainer(
        task=task,
        optimizer=optimizer,
        criterion=criterion,
        accelerator=args.accelerator,
        devices=args.devices,
        precision=args.precision
    )

    # Train
    trainer.fit(
        train_loader=train_loader,
        val_loader=val_loader,
        epochs=args.epochs,
        save_model_after_every_epoch=args.save_freq
    )

if __name__ == '__main__':
    parser = argparse.ArgumentParser()
    parser.add_argument('--data_dir', type=str, required=True)
    parser.add_argument('--checkpoint_dir', type=str, default='./checkpoints')
    parser.add_argument('--arch', type=str, default='resnet18')
    parser.add_argument('--num_classes', type=int, required=True)
    parser.add_argument('--pretrained', action='store_true')
    parser.add_argument('--batch_size', type=int, default=32)
    parser.add_argument('--epochs', type=int, default=50)
    parser.add_argument('--lr', type=float, default=1e-3)
    parser.add_argument('--weight_decay', type=float, default=1e-4)
    parser.add_argument('--num_workers', type=int, default=4)
    parser.add_argument('--accelerator', type=str, default='auto')
    parser.add_argument('--devices', type=str, default='auto')
    parser.add_argument('--precision', type=str, default='16-mixed')
    parser.add_argument('--save_freq', type=int, default=10)

    args = parser.parse_args()
    main(args)

Run Training

python train.py \
    --data_dir ./data \
    --checkpoint_dir ./checkpoints \
    --arch resnet50 \
    --num_classes 10 \
    --pretrained \
    --batch_size 64 \
    --epochs 100 \
    --lr 1e-3

More Examples

Check the following for more examples:

scripts/ directory in the repository
notebooks/ directory for Jupyter notebooks
GitHub repository: https://github.com/sagar100rathod/deep-ml

Community Examples

Share your examples and projects using deep-ml!

Contributing examples is welcome. See Contributing for guidelines.