Why do I run so slow in training？

[Rzlf-47]

Thank you for reading this question. I tried to run the training of SimSwap by myself, 

but its training speed was very slow. I want to know what I should do to improve the training speed. And when I was running training, the GPU utilization rate was very low, and the host fan was turning and stopping. My GPU is NVIDIA 2070s, AMD Ryzen73700x8core. I downloaded Vggface2 myself and changed the address of '--dataset' to the path of this dataset in the code. Since the model.act1 of _make_efficientnet generates errors in the projector.py file, I changed it to nn.SiLU(implace=False). dataloader_worker=8, pin_memory=False in data_loader_butter.py. If you can help me with problems or ideas, I would be most grateful.