zgcharaf / deep-learning-ultra

Open source Deep Learning Containers (DLCs) are a set of Docker images for training and serving models in PyTorch, OpenCV (compiled for GPU), TensorFlow 2 for GPU, PyG and NVIDIA RAPIDS

Deep Learning Ultra

Open source Deep Learning Containers (DLCs) are a set of Docker images for training and serving models with MLFlow in PyTorch, OpenCV (compiled for GPU), TensorFlow 2 for GPU, PyG and NVIDIA RAPIDS, running on CUDA 12.1. Tensorboard included for visualizations into model explainability and fine-tuning/understanding how your model learns.

Used and adopted by many!

Ultra easy deployment

Edit line 18 in docker-compose.yaml for however many GPUs you have: count: 3 # num of gpus, then run:

docker compose up --build -d

• Get security token to log into the notebook:

  token=$(docker exec -it ultra /bin/bash -c "jupyter notebook list") \
  echo ${token::-8}
  

Requirements

• Linux configured with nvidia-container-toolkit found here, CUDA 12, NVIDIA Drivers v.525+
• NVIDIA GPU -ARCH_7.5+

Deep Learning Container Services

1. Deep Learning Notebook, or you can mount VSCode to /app | http://localhost:8888
2. MLFlow is an open source platform for managing the end-to-end machine learning lifecycle, see more here | http://localhost:5000
3. Tensorboard povides the visualization and tooling needed for machine learning experimentation | http://localhost:6006

• Deep learning solution - all python bindings specifically compiled for c++/CUDA:

  • Pytorch 2
  • PyG (Graph Neural Networks)
  • NVIDIA RAPIDS
  • OpenCV v4.8
  • TensorFlow 2

• CuPy, Anaconda Python v3.11.5, Captum, MLFlow and more!

• Supports LLMs, HuggingFace, Computer Vision, Navigation, Physics Informed ML, and Graph Neural Networks

---