Open crackedpotato007 opened 3 years ago
Most ML libraries/frameworks use CUDA which is an Nvidia technology so it's generally unsupported for AMD GPU's to my knowledge, am I sure there could be a workaround somewhere but I am unsure. The standard for machine learning right now is Nvidia. Sorry to disappoint, however Google Colaboratory lets you run the code with GPU acceleration
I know this is pretty old, but I figured I still answer for someone finds their way here. To train AI-Models with AMD GPUs you first need ROCm: https://docs.amd.com/bundle/ROCm-Installation-Guide-v5.5/page/Introduction_to_ROCm_Installation_Guide_for_Linux.html As the URL suggest, you need a supported Linux-Distribution and a fitting pytorch version for ROCm. ROCm supports the cuda nomenclature and therefore doesn't need any changes in the code as pytorch/ROCm just "translate" it. After all this is installed properly, you can start training on your AMD-GPU.
With version 5.6 there is supposedly Windows support coming.
Hey, i own a amd gpu and look forward to make models but i cant as it needs nvidia drivers it is there any workarounds