I'm trying to use Stable Diffusion locally on my MacBook Pro Intel and I have many problems.
My question for this post is: which checkpoint should I use (and why) ?
Bellow is the details of my experiment.
On Stable Diffusion GitHub's page, it is written :
First, download the weights for SD2.1-v and SD2.1-base.
From the SD2.1-v, I've downloaded the v2-1_768-ema-pruned.ckpt checkpoint.
From the SD2.1-base, I've downloaded the v2-1_512-ema-pruned.ckpt checkpoint.
When I used SD2.1-v, I was able to generate an image:
python scripts/txt2img.py \
--prompt "a sunset behind a mountain range, vector image" \
--ckpt ckpt/v2-1_768-ema-pruned.ckpt \
--H 512 --W 512 \
--n_samples 1 --n_iter 1 \
--config configs/stable-diffusion/intel/v2-inference-v-fp32.yaml \
--precision full
But when I used SD2.1-base, the output image was not correct.
For both tries, I use the following configuration:
configs/stable-diffusion/intel/v2-inference-v-fp32.yaml
Thanks for your help.
