arnaldovitor/pytorch-gfx803

Installing PyTorch 2.1.2 with ROCm Support on Linux Mint 21.2

This tutorial guides you through installing PyTorch 2.1.2 with ROCm support on Linux Mint 21.2 using a Radeon RX 480/580 8GB GPU.

Prerequisites

Ensure your system meets the following requirements:

Linux Kernel: Version 5.19.0-50-generic is required for compatibility with ROCm 5.5.
ROCm: Version 5.5 is installed.
Python: Version 3.10.6.

Step 1: Downgrade Kernel (if necessary)

If your current kernel is newer than 5.19.0-50-generic, downgrade to this version:

KERN="5.19.0-50-generic"
sudo apt install "linux-image-$KERN" "linux-headers-$KERN" "linux-modules-$KERN" "linux-modules-extra-$KERN"

Step 2: Install Dependencies

Install necessary dependencies for ROCm and PyTorch:

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

Step 3: Install ROCm

Configure ROCm environment variables and install ROCm 5.5:

echo -e "ROC_ENABLE_PRE_VEGA=1\nHSA_OVERRIDE_GFX_VERSION=8.0.3" | sudo tee -a /etc/environment
sudo reboot

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 ROCm installation:

rocminfo
rocm-smi
clinfo

Step 4: Install PyTorch

Download and install torch and torchvision using the wheels built at viebrix's releases or from our Google Drive for guaranteed availability:

Google Drive link

pip install torch-2.1.2-cp310-cp310-linux_x86_64.whl
pip install torchvision-0.16.2-cp310-cp310-linux_x86_64.whl
pip install numpy==1.21.5 --force-reinstall  # Resolve NumPy initialization issue

Step 5: Test Installation

Verify PyTorch installation by running a test script:

python3.10 test_torch.py

This guide should help you install PyTorch 2.1.2 with ROCm support smoothly on your system. If you encounter any issues, feel free to refer to viebrix's repository or raise an issue for assistance.

arnaldovitor / pytorch-gfx803