-
-
Notifications
You must be signed in to change notification settings - Fork 134
1. Installation
Aphrodite is primarily a python library, but also contains CUDA C/C++ code that needs to be compiled. We provide pre-compiled binaries in a wheel, so users don't have to deal with the compile times.
- Linux, Windows (WSL)
- Python 3.8 - 3.11
- NVIDIA (compute capability > 6.1) and AMD (MI200+, NAVI)
Any NVIDIA GPU with a compute capability of 6.1 or higher. Refer to this page for a full list of CUDA GPUs:
https://developer.nvidia.com/cuda-gpus.
Or, you can manually find out your GPU's Compute Capability by opening a Python interpreter and running:
>>> import torch # if you don't have `torch` installed, run `pip install torch` first
>>> print(torch.cuda.get_device_capability())
This should print something like this: (7, 5)
, which would indicate a CC of 7.5
If you do not meet the minimum CC, you will not be able to run Aphrodite. At the moment, compute capability of 7.5 or higher is required for AWQ quantization scheme; you can use GPTQ if your GPU does not support it
For the pip package, all you need is NVIDIA drivers.
$ pip install -U aphrodite-engine --extra-index-url https://downloads.pygmalion.chat/whl
Caution
Ninja during build uses 8GB of RAM per thread. If you're running out of memory, export MAX_JOBS=1
, or the highest number you can do before you OOM. e.g., if you have 32 threads but 64GB of RAM, limit MAX_JOBS to 8.
The embedded runtime sets up micromamba and downloads CUDA 12.2 along with all the required dependencies.
$ git clone https://github.com/PygmalionAI/aphrodite-engine.git && cd aphrodite-engine
$ ./update-runtime.sh # this can take a while
You can then use the runtime.sh
script to enter the Aphrodite Engine environment. You may also simply prefix all your Aphrodite-related commands with ./runtime.sh
to execute them within the created env.
Alternatively, if you already have an environment with CUDA 12, you can simply run the following to build aphrodite-engine:
$ pip install -e . # this can take a while
Credits to @pyroserenus for the initial draft of the guide.
Windows is currently not supported, but you can use WSL2 to install Aphrodite Engine.
- Open Windows PowerShell as administrator.
- Type
wsl --install
and hit enter. This may take a while. - It will ask for a username and password. These are arbitrary.
- Reboot your PC.
- Run this command:
sudo apt update && sudo apt upgrade -y && sudo apt install python3 python3-pip git wget curl bzip2 tar
The above command will update all installed packages on WSL, then install python.
- Install Aphrodite Engine. You can now install Aphrodite Engine via pip. If that fails, try building from source.