gusanagy/holoocean-cuda-dev-4 repository overview

⁠🐳 HoloOcean CUDA Dev Container

A ready-to-use Docker environment for developing and running HoloOcean⁠ simulations with CUDA, Vulkan, and full GUI support.

⁠📦 Features

• ✅ CUDA 12.8.1 support for GPU-based training and simulation
• 🧠 HoloOcean Python API pre-installed
• 🎮 Vulkan/Mesa graphics stack for real-time rendering
• 📸 OpenCV with Python bindings
• 🖥️ X11 GUI support for graphical environments
• 🐧 Based on Ubuntu, with development tools and dependencies

⁠🗂️ Repository Structure

. ├── Dockerfile # Builds the dev image ├── start_container.sh # 🧠 Automates container startup with GUI support ├── /client/tests # Contains test scripts for HoloOcean environments └── README.md

⁠⚙️ Automated Container Startup

The script start_container.sh simplifies the initialization of the development container by:

• 🔐 Setting correct permissions for X11, enabling GUI apps inside Docker
• 📂 Automatically mounting your host directory inside the container
• 🎮 Enabling GPU access via --gpus all and NVIDIA runtime
• 📌 Running the container in interactive mode with a custom working directory

⁠📜 Example Script (start_container.sh)

#!/bin/bash
# Declara que o script será interpretado pelo bash

# ============================
# DEFINIÇÕES INICIAIS
# ============================

# Nome que será dado ao container em execução
CONTAINER_NAME="holoocean_dev_container_cuda_12.2"

# Nome da imagem Docker que você construiu previamente com Dockerfile
IMAGE_NAME="gusanagy/holoocean-cuda-dev-4:latest"

# Diretório atual do host, será montado dentro do container
HOST_DIR=$(pwd)

# ============================
# PERMISSÃO PARA INTERFACE GRÁFICA (X11)
# ============================

# Libera acesso ao servidor X11 para o Docker (permite abrir janelas gráficas dentro do container)
#xhost +local:docker

# ============================
# EXECUÇÃO DO CONTAINER
# ============================

# Inicia o container Docker com todos os parâmetros
docker run -it --rm \
  --name "$CONTAINER_NAME" \
  --gpus all \
  --runtime=nvidia \
  --net=host \
  --ipc=host \
  --privileged \
  -e DISPLAY=$DISPLAY \
  -e NVIDIA_VISIBLE_DEVICES=all \
  -e NVIDIA_DRIVER_CAPABILITIES=compute,graphics,utility \
  -e QT_X11_NO_MITSHM=1 \
  -v /tmp/.X11-unix:/tmp/.X11-unix \
  -v "$HOST_DIR:/tmp/shared" \
  -w /home/developer/HoloOcean \
  -u developer \
  "$IMAGE_NAME"

💡 Why it's important: Starting GPU-based GUI containers manually can be complex. This script automates the setup, avoiding common issues with graphics, permissions, and device visibility.

⁠🚀 Getting Started

⁠1. Pull the Docker image

docker pull your_dockerhub_username/holoocean-cuda-dev-4:latest

1. Make the startup script executable

chmod +x start_container.sh

1. Run the container

./start_container.sh

⁠👨‍🔬 Use Cases

🤖 Reinforcement learning with underwater agents

🌊 Marine robotics simulation using Unreal Engine

🧪 Research and prototyping for AI, control systems, and simulation environments

🤝 Contributing Contributions are welcome! Feel free to open issues, suggest improvements, or submit pull requests to help expand and enhance this development container for other users and use cases.