Skip to content

Description and whl files for using Whisper with gfx803 RX580

Notifications You must be signed in to change notification settings

viebrix/pytorch-gfx803-for-Whisper

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

2 Commits
 
 

Repository files navigation

pytorch-gfx803-for-Whisper (WEBUI)

Description and whl files for using Whisper with gfx803 RX580

PyTorch ROCm gfx803

build pytorch 1.13.1 with ROCm support for Whisper

Linux Mint 21.2
Radeon RX 580 8GB
RoCm 5.5

Python 3.10.6
- pytorch 1.13.1
- torchvision 0.14.1

Install dependencies

sudo apt autoremove rocm-core amdgpu-dkms
sudo apt install libopenmpi3 libstdc++-12-dev libdnnl-dev ninja-build libopenblas-dev libpng-dev libjpeg-dev

Install ROCm

sudo -i
sudo echo ROC_ENABLE_PRE_VEGA=1 >> /etc/environment
sudo echo HSA_OVERRIDE_GFX_VERSION=8.0.3 >> /etc/environment
# Reboot after this

wget https://repo.radeon.com/amdgpu-install/5.5/ubuntu/jammy/amdgpu-install_5.5.50500-1_all.deb
sudo apt install ./amdgpu-install_5.5.50500-1_all.deb
sudo amdgpu-install -y --usecase=rocm,hiplibsdk,mlsdk

sudo usermod -aG video $LOGNAME
sudo usermod -aG render $LOGNAME

# verify
rocminfo
clinfo

Build

You may need to install addional dependencies, and the build will take a long time.

#in home directory create directory pytorch1.13.1 #Build Torch

cd pytorch1.13.1
git clone --recursive https://github.com/pytorch/pytorch.git -b v1.13.1
cd pytorch
pip install cmake mkl mkl-include
pip install -r requirements.txt
sudo ln -s /usr/lib/x86_64-linux-gnu/librt.so.1 /usr/lib/x86_64-linux-gnu/librt.so
export PATH=/opt/rocm/bin:$PATH ROCM_PATH=/opt/rocm HIP_PATH=/opt/rocm/hip
export PYTORCH_ROCM_ARCH=gfx803
export PYTORCH_BUILD_VERSION=1.13.1 PYTORCH_BUILD_NUMBER=1
export USE_CUDA=0 USE_ROCM=1 USE_NINJA=1
python3 tools/amd_build/build_amd.py
python3 setup.py bdist_wheel
#pip install dist/torch-1.13.1-cp310-cp310-linux_x86_64.whl --force-reinstall

Build torchvision

cd ..
git clone https://github.com/pytorch/vision.git -b v0.14.1
cd vision
export BUILD_VERSION=0.14.1
FORCE_CUDA=1 ROCM_HOME=/opt/rocm/ python3 setup.py bdist_wheel
#pip install dist/torchvision-0.14.1-cp310-cp310-linux_x86_64.whl --force-reinstall

If you are using WHISPER-WEBUI you have to turn of FASTER-WHISPER because it does not work with ROCm, therefore you have to call it with argument --disable_faster_whisper

install in WHISPER-WEBUI

cd Whisper-WebUI
python3 -m venv venv
source venv/bin/activate
python -m pip install --upgrade pip wheel
pip uninstall torch torchvision
pip3 install /home/*******/pytorch1.13.1/pytorch/dist/torch-1.13.1-cp310-cp310-linux_x86_64.whl
pip3 install /home/*******/pytorch1.13.1/vision/dist/torchvision-0.14.1-cp310-cp310-linux_x86_64.whl
pip list | grep 'torch'
python3 app.py --disable_faster_whisper

******* is your home directory username

I have tested a 45 minutes mp4 viedeo in language german to transcibe on my local computer (language setting was also preset to german)

  1. with faster-whisper and CPU mode
  2. with whisper and RX580

Results (which was not really a clean test - because other programs (firefox and so on where running during the test) 1: fasterwhisper with cpu 27.0 minutes 44 seconds! 2: whisper with gpu rx580 8GB (gfx803) 22.0 minutes 54 seconds!

Reference

About

Description and whl files for using Whisper with gfx803 RX580

Resources

Stars

Watchers

Forks

Packages

No packages published