/DeepX

Primary LanguagePythonMIT LicenseMIT

DeepX

Deep learning algorithms implemented with PyTorch and Lightning.

License: MIT

Features

  • Backbone

    • MLP
    • ResNet
    • Vision Transformer
    • Transformer

  • Algorithms/Architectures

    • Vision
      • Classification
        • ResNet
      • Segmentation
        • UNet
      • Object Detection
        • YOLO
        • SSD
      • Generation
        • VAE
        • DCGAN
        • Diffusion Models
    • Language
      • Language Model
      • Text Classification
      • Translation

Documentation

Documentation is available here!!

Installation

  • Docker installation:
cd envs
docker compose up -d
  • PyPI installation:
pip install -e .

Usage

  • Starting a container:
docker exec -it deepx zsh
  • Experiment Tracking:

Access http://localhost:5000 in your browser.

Development

  • Profiling
cd experiments/training
python -m cProfile -o profile.prof <task>.py <args> task.debug=true machine.num_workers=0
snakeviz profile.prof
  • PyTorch Profiler
cd experiments/training
python -m torch.utils.bottleneck <task>.py <args> task.debug=true machine.num_workers=0