Install PyTorch and ComfyUI on ROCm 7.10.0

This topic guides you through installing PyTorch with ROCm support on AMD hardware. It applies to supported AMD GPUs and platforms. It also includes additional setup steps for ComfyUI, showcasing AI-powered image generation.

AMD device family

Instinct

Radeon PRO

Radeon RX

Ryzen AI

Instinct GPU

Instinct MI355X
Instinct MI350X

Instinct MI325X
Instinct MI300X
Instinct MI300A

Instinct MI250X
Instinct MI250
Instinct MI210

Radeon PRO GPU

Radeon PRO W7900D
Radeon PRO W7900
Radeon PRO W7800 48GB
Radeon PRO W7800

Radeon PRO W7700

Radeon RX GPU

Radeon RX 7900 XTX
Radeon RX 7900 XT
Radeon RX 7900 GRE

Radeon RX 7800 XT
Radeon RX 7700 XT

Ryzen AI APU

Ryzen AI Max+ PRO 395
Ryzen AI Max PRO 390, 385, 380
Ryzen AI Max+ 395
Ryzen AI Max 390, 385

Ryzen AI 9 HX 375
Ryzen AI 9 HX 370
Ryzen AI 9 365

Operating system

Linux

Operating system

Linux

Windows

Prerequisites

To run ComfyUI workloads on Windows, ensure you have Adrenalin Driver version 25.20.01.17. For details and the download link, see AMD Software: PyTorch on Windows Edition 7.1.1. See the related known issue.

Ensure your system has a supported Python version installed and accessible: 3.11, 3.12, or 3.13.

Review the ROCm 7.10.0 compatibility matrix for more details.

Install PyTorch

For prerequisite steps and post-installation recommendations, see the ROCm installation instructions.

1. Set up your Python virtual environment. If you already have a successful ROCm 7.10.0 installation using pip, skip this step.

   For example, run the following command to create a virtual environment:

   python3.12 -m venv .venv
   

2. Activate your Python virtual environment. For example:

   source .venv/bin/activate
   

   .venv\Scripts\activate
   

3. Install the appropriate ROCm-enabled PyTorch build for your operating system and AMD hardware architecture.

   python -m pip install --index-url https://repo.amd.com/rocm/whl/gfx950-dcgpu/ torch torchvision torchaudio
   

   python -m pip install --index-url https://repo.amd.com/rocm/whl/gfx94X-dcgpu/ torch torchvision torchaudio
   

   python -m pip install --index-url https://repo.amd.com/rocm/whl/gfx90X-dcgpu/ torch torchvision torchaudio
   

   python -m pip install --index-url https://repo.amd.com/rocm/whl/gfx1151/ torch torchvision torchaudio
   

   python -m pip install --index-url https://repo.amd.com/rocm/whl/gfx1150/ torch torchvision torchaudio
   

   python -m pip install --index-url https://repo.amd.com/rocm/whl/gfx110X-dgpu/ torch torchvision torchaudio
   

4. Check your PyTorch installation.

   python -c "import torch; print(torch.cuda.is_available())"
   

   This prints True if PyTorch and ROCm are installed properly.

Optionally, for a demonstration of ROCm-powered image generation, try ComfyUI on Windows.

Install and run ComfyUI

1. Ensure your working environment is running ROCm-enabled PyTorch on a supported system. See Install PyTorch for instructions.

2. Clone the ComfyUI repository.

   git clone https://github.com/comfyanonymous/ComfyUI.git
   

3. Install Python dependencies.

   pip install -r ComfyUI/requirements.txt
   

   pip install -r ComfyUI\requirements.txt
   

4. Run ComfyUI.

   1. Start the ComfyUI server from the command line.

      python ComfyUI/main.py
      

      python ComfyUI\main.py
      

      This will start the server and display a prompt like:

      To see the GUI go to: http://127.0.0.1:8188
      

   2. Navigate to http://127.0.0.1:8188 in your web browser. You might need to replace 8188 with the appropriate port number.

      

   3. Search for one of the following templates and download any missing models.

      SD3.5 Simple

      Select Template → Model Filter → SD3.5 → SD3.5 Simple

      

      Download required models, if missing.

      

      Chroma1 Radiance text to image

      Select Template → Model Filter → Chroma → Chroma1 Radiance text to image

      

      Download required models, if missing.

      

   4. Click the Run button.

   The application will use your AMD GPU to convert the prompted text to an image.

Known limitations

ComfyUI might not start if Smart App Control is enabled in your Windows security settings.