Can I train a new stable diffusion model from scratch?
#141
by
zefyrus
- opened
I'm concluding my master thesis on generative models in machine learning and last chapter is a comparison between GANs and diffusion models. Can we train a stable diffusion model from zero in the style of deep learning models?
for epoch in range(epochs):
for data, description in tqdm(dataloader):
#generate, minimize losses, calculate gradients, etc.