FAVDBench
FAVDBench copied to clipboard
Hardware requirements
Hi, can I know what was the GPUs used to train the model?
Hi, our model has been trained on either 16*A100-40G or 8*A100-80G configurations. The code is designed to function effectively on a single A100, but the prior settings are advantageous for larger batch sizes and thus better results.
Additionally, we've validated the code on v100-32G and 3090-24G.