Mirrored_Repos/InvokeAI
1
0
Fork 0
mirror of https://github.com/invoke-ai/InvokeAI synced 2024-08-30 20:32:17 +00:00
Code Issues Packages Projects Releases Wiki Activity

revert previous change

Browse Source
This commit is contained in:
Lincoln Stein 2022-10-06 14:57:06 -04:00
parent 2c27e759cd
commit 440065f7f8
2 changed files with 17 additions and 10 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

22
ldm/models/diffusion/ksampler.py
View File

@ -123,15 +123,15 @@ class KSampler(Sampler):
if img_callback is not None:
img_callback(k_callback_values['x'],k_callback_values['i'])
# sigmas = self.model.get_sigmas(S)
# sigmas are now set up in make_schedule - we take the last steps items
# sigmas are set up in make_schedule - we take the last steps items
sigmas = self.sigmas[-S-1:]
x = torch.randn([batch_size, *shape], device=self.device) * sigmas[0]
# for GPU draw
if x_T is not None:
x = x_T + x
x = x_T * sigmas[0]
# x = x_T + torch.randn([batch_size, *shape], device=self.device) * sigmas[0]
else:
x = torch.randn([batch_size, *shape], device=self.device) * sigmas[0]
model_wrap_cfg = CFGDenoiser(self.model, threshold=threshold, warmup=max(0.8*S,S-10))
extra_args = {
@ -194,11 +194,17 @@ class KSampler(Sampler):
return img, None, None
# REVIEW THIS METHOD: it has never been tested. In particular,
# we should not be multiplying by self.sigmas[0] if we
# are at an intermediate step in img2img. See similar in
# sample() which does work.
def get_initial_image(self,x_T,shape,steps):
x = (torch.randn(shape, device=self.device) * self.sigmas[0])
if x_T is not None:
return x_T + x_T * self.sigmas[0]
return x_T + x
else:
return (torch.randn(shape, device=self.device) * self.sigmas[0])
return x
def prepare_to_sample(self,t_enc):
self.t_enc = t_enc

5
ldm/models/diffusion/sampler.py
View File

@ -351,10 +351,11 @@ class Sampler(object):
return x_dec
def get_initial_image(self,x_T,shape,timesteps=None):
x = torch.randn(shape, device=self.device)
if x_T is None:
return torch.randn(shape, device=self.device)
return x
else:
return x_T
return x_T + x
def p_sample(
self,