My ramblings from training several types of models on AMD hardware w/PyTorch ROCm
Great post! Just some clarification - /dev/dri is actually the GPU. In Linux, everything is a file, even hardware. DRI = “direct render interface”. —network=host is what configured the container to use the host network namespace.
Yeah, that's right. I fixed it, thanks!
Great post! Just some clarification - /dev/dri is actually the GPU. In Linux, everything is a file, even hardware. DRI = “direct render interface”. —network=host is what configured the container to use the host network namespace.
Yeah, that's right. I fixed it, thanks!