Hi All, I’m wondering if anybody has any experiences to share on training from scratch? I’m finding I can get fairly decent results, in a reasonable amount of time, if my dateset is small, but as I increase the size of the dataset things get worse… or maybe they just require much more training (eno…

Training from scratch

sayakpaul April 6, 2023, 11:15pm 6

What do you mean prompt accuracy? Something like CLIP score?

Topic		Replies	Views
Loss drops normally but stops improving quickly 🧨 Diffusers	3	5655	March 9, 2023
Tiny fine tune messed up the pretrained sd v1-4 model 🧨 Diffusers	2	514	April 8, 2023
Loss doesn't converge for latent diffusion model 🧨 Diffusers	0	1199	September 25, 2023
Training loss does not go down during fine-tuning Beginners	0	1810	July 3, 2023
A question about finetuning 🧨 Diffusers	0	274	April 11, 2023