extern/easydiffusion
1
0
Fork 1
mirror of https://github.com/easydiffusion/easydiffusion.git synced 2025-05-29 22:29:30 +02:00
Code Issues Packages Projects Releases Wiki Activity

Img2img sd2 attempt 2

Browse Source
This commit is contained in:
cmdr2 2022-11-25 23:28:31 +05:30 committed by GitHub
parent 617a8b2814
commit 7fa983b971
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 15 additions and 8 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

23
ui/sd_internal/runtime.py
View File

@ -808,13 +808,6 @@ def _txt2img(opt_W, opt_H, opt_n_samples, opt_ddim_steps, opt_scale, start_code,
def _img2img(init_latent, t_enc, batch_size, opt_scale, c, uc, opt_ddim_steps, opt_ddim_eta, opt_seed, img_callback, mask): def _img2img(init_latent, t_enc, batch_size, opt_scale, c, uc, opt_ddim_steps, opt_ddim_eta, opt_seed, img_callback, mask):
# encode (scaled latent) # encode (scaled latent)
z_enc = thread_data.model.stochastic_encode(
init_latent,
torch.tensor([t_enc] * batch_size).to(thread_data.device),
opt_seed,
opt_ddim_eta,
opt_ddim_steps,
)
x_T = None if mask is None else init_latent x_T = None if mask is None else init_latent
if thread_data.test_sd2: if thread_data.test_sd2:
@ -822,7 +815,13 @@ def _img2img(init_latent, t_enc, batch_size, opt_scale, c, uc, opt_ddim_steps, o
sampler = DDIMSampler(thread_data.model) sampler = DDIMSampler(thread_data.model)
samples_ddim = thread_data.model.sample( sampler.make_schedule(ddim_num_steps=opt_ddim_steps, ddim_eta=opt_ddim_eta, verbose=False)
z_enc = sampler.stochastic_encode(init_latent, torch.tensor([t_enc] * batch_size).to(thread_data.device))
#samples = sampler.decode(z_enc, c, t_enc, unconditional_guidance_scale=opt.scale,
unconditional_conditioning=uc, )
samples_ddim = sampler.sample(
S=t_enc, S=t_enc,
batch_size=opt_n_samples, batch_size=opt_n_samples,
conditioning=c, conditioning=c,
@ -834,6 +833,14 @@ def _img2img(init_latent, t_enc, batch_size, opt_scale, c, uc, opt_ddim_steps, o
x_T=x_T x_T=x_T
) )
else: else:
z_enc = thread_data.model.stochastic_encode(
init_latent,
torch.tensor([t_enc] * batch_size).to(thread_data.device),
opt_seed,
opt_ddim_eta,
opt_ddim_steps,
)
# decode it # decode it
samples_ddim = thread_data.model.sample( samples_ddim = thread_data.model.sample(
t_enc, t_enc,