A few months ago I discovered "Deepfacelab", a software with which you can create deepfakes. Recently I tried switching from my home pc to a server which can hold multiple graphics-cards to speed up the training process. The problem was that the training process did not start (at least SAEHD and AMP didn't, Quick 96 worked for some reason). So we tried many, many things, updated the GPU dirvers, tried different builds of DFL, installed several extensions, like python, numpy and so on.. We came to the conclusion that the CPU in my home PC has AVX and the one in the server has not, which apparently is a requirement for Deepfacelab. Since replacing the old Cpu with a new one is not an option at this moment, I wondered whether you can prevent DFL from wanting to use AVX. Is this possible at all, and if yes, where one could turn this reliance to AVX off? If you're in and want to take a look, here is the link to the github page: https://github.com/iperov/DeepFaceLab Under "releases" you can find the download links; I used the Mega.nz link for Windows and chose the following build: DeepFaceLab_NVIDIA_up_to_RTX2080Ti_build_11_20_2021.exe Continue reading...
