Matthew Furseman

I have a server with three RTX 3090 24GB GPUs. When trying to run inference with more than one GPU this fails with CUDA invalid value. ## Runtime output ```...