This package contains Metal kernels for fast on-device SSM inference on Apple silicon.
To install this package, first install Xcode, which can be downloaded from https://developer.apple.com/xcode/. Accept the license agreement with:
sudo xcodebuild -license
We recommend using an environment management tool such as conda
or virtualenv
.
Note: This package has been tested on macOS Sonoma 14.1 with the M3 chip.
Create and activate a new conda environment. Ensure that it is an arm64
environment. Example:
CONDA_SUBDIR=osx-arm64 conda create -n cartesia_mlx python=3.11
conda activate cartesia_mlx
In your conda environment, install the nanobind
package (exact version required), followed by cartesia-metal
:
pip install nanobind@git+/~https://github.com/wjakob/nanobind.git@2f04eac452a6d9142dedb957701bdb20125561e4
pip install git+/~https://github.com/cartesia-ai/edge.git#subdirectory=cartesia-metal
Create and activate a new virtualenv:
python -m venv cartesia_mlx
source cartesia_mlx/bin/activate
In your virtual environment, install the wheel
package, and the nanobind
package (exact version required), followed by cartesia-metal
:
pip install wheel
pip install nanobind@git+/~https://github.com/wjakob/nanobind.git@2f04eac452a6d9142dedb957701bdb20125561e4
pip install git+/~https://github.com/cartesia-ai/edge.git#subdirectory=cartesia-metal --no-build-isolation