OmniOpt2 helps you effortlessly optimize complex hyperparameter configurations — even when gradients are unavailable or too complicated. Built on Ax and BoTorch, it also supports any method of generating new hyperparameter points using External Generators.

You can run OmniOpt2 on any Linux system with python3. Missing dependencies will be installed automatically.

Have Slurm? Jobs will be automatically parallelized. No Slurm? Jobs run sequentially, hassle-free.

All Python dependencies are installed automatically in a dedicated virtual environment on the first run. This may take a few minutes, but ensures your system stays clean and you can focus on optimization. For complete isolation, use Docker.

In short, OmniOpt2 tests different hyperparameter combinations and uses the BoTorch-Modular (among others) algorithm to find settings that minimize your target metric.

• Works on any Linux + Python3
• Automatic dependency installation
• Supports Slurm for parallel jobs
• Docker-ready for full isolation