Running model on AMD

#4
by Fl1ntSt0n3 - opened

For those wondering how to run on AMD based HW:

Clone the github repository
Go on liquid-audio folder and then remove the existing .venv folder

Add the following onto the pyproject.toml

[tool.uv.sources]
apex = { index = "rocm"}
torch = { index = "rocm" }
triton = { index = "rocm"}
torchaudio = { index = "rocm"}
torchvision = { index = "rocm"}

[[tool.uv.index]]
name = "pypi"
url = "https://pypi.org/simple"
default = true

[[tool.uv.index]]
name = "rocm"
url = "https://download.pytorch.org/whl/nightly/rocm7.1"

Create a new venv:
uv venv

Add the missing dependencies:
uv add gradle
&
uv add fastrtc[vad]

Then sync the project dependencies:
uv sync

Once its done, you can run the demo script:
uv run liquid-audio-demo

Sign up or log in to comment