r/stable-diffusion-finetune
1
0
Fork 0
Code Issues Pull requests Projects Releases Wiki Activity

experimental support for guided upscale sampling

Browse source
This commit is contained in:
Patrick Esser 2022-06-13 10:44:10 +00:00 committed by root
parent 255d479088
commit 3c77204c0a
3 changed files with 27 additions and 3 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

2
configs/stable-diffusion/txt2img-upscale-clip-encoder-f16-1024.yaml
View file

@ -24,7 +24,7 @@ model:
linear_start: 0.00085
linear_end: 0.0120
timesteps: 1000
max_noise_level: 250
max_noise_level: 100
output_size: 64
model_config:
target: ldm.models.autoencoder.AutoencoderKL

15
ldm/models/diffusion/ddim.py
View file

@ -179,7 +179,20 @@ class DDIMSampler(object):
else:
x_in = torch.cat([x] * 2)
t_in = torch.cat([t] * 2)
c_in = torch.cat([unconditional_conditioning, c])
if isinstance(c, dict):
assert isinstance(unconditional_conditioning, dict)
c_in = dict()
for k in c:
if isinstance(c[k], list):
c_in[k] = [torch.cat([
unconditional_conditioning[k][i],
c[k][i]]) for i in range(len(c[k]))]
else:
c_in[k] = torch.cat([
unconditional_conditioning[k],
c[k]])
else:
c_in = torch.cat([unconditional_conditioning, c])
e_t_uncond, e_t = self.model.apply_model(x_in, t_in, c_in).chunk(2)
e_t = e_t_uncond + unconditional_guidance_scale * (e_t - e_t_uncond)

13
ldm/models/diffusion/ddpm.py
View file

@ -1543,7 +1543,18 @@ class LatentUpscaleDiffusion(LatentDiffusion):
log["denoise_row"] = denoise_grid
if unconditional_guidance_scale > 1.0:
uc = self.get_unconditional_conditioning(N, unconditional_guidance_label)
uc_tmp = self.get_unconditional_conditioning(N, unconditional_guidance_label)
# TODO explore better "unconditional" choices for the other keys
uc = dict()
for k in c:
if k == "c_crossattn":
assert isinstance(c[k], list) and len(c[k]) == 1
uc[k] = [uc_tmp]
elif isinstance(c[k], list):
uc[k] = [torch.zeros_like(c[k][i]) for i in range(len(c[k]))]
else:
uc[k] = torch.zeros_like(c[k])
with ema_scope("Sampling with classifier-free guidance"):
samples_cfg, _ = self.sample_log(cond=c, batch_size=N, ddim=use_ddim,
ddim_steps=ddim_steps, eta=ddim_eta,