RAVE training requires CUDA verified via nvidia-smi and a dedicated conda environment

Before training RAVE, the system needs a CUDA-enabled NVIDIA GPU with drivers installed. Verification: run nvidia-smi in a terminal — if a message appears with the name of your graphics card and its available memory, CUDA is working. RAVE can be trained on Linux or Windows, but Linux is advised. Installation via a conda environment is recommended to isolate dependencies and keep Python versions separate. Steps: create a conda env (conda create -n rave python=3.9), activate it, then install torch/torchaudio followed by pip install acids-rave. The conda env must be activated each session (conda activate rave) before using any rave CLI commands.

Examples

nvidia-smi → shows GPU stats → good to proceed. conda create -n rave python=3.9 && conda activate rave && pip install acids-rave.

Assessment

A user runs rave train and gets a CUDA error. Walk through the two verification steps they should perform before filing a bug report.