Skip to content

[DDIMScheduler] fix noise device in ddim step #1189

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 6 commits into from
Nov 8, 2022
Merged

[DDIMScheduler] fix noise device in ddim step #1189

Changes from all commits
Commits
Show all changes
6 commits
Select commit Hold shift + click to select a range
7d90f1c
fix noise device in ddim sched
patil-suraj Nov 8, 2022
1944914
fix typo
patil-suraj Nov 8, 2022
4bc7d6d
self.device -> device
patil-suraj Nov 8, 2022
8644fd0
remove duplicated if
patil-suraj Nov 8, 2022
93b2cee
use str device
patil-suraj Nov 8, 2022
e7d9fc6
don't use str for device
patil-suraj Nov 8, 2022
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
13 changes: 9 additions & 4 deletions src/diffusers/schedulers/scheduling_ddim.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -288,17 +288,22 @@ def step(

if eta > 0:
# randn_like does not support generator https://github.com/pytorch/pytorch/issues/27072
device = model_output.device if torch.is_tensor(model_output) else "cpu"
device = model_output.device if torch.is_tensor(model_output) else torch.device("cpu")
if variance_noise is not None and generator is not None:
raise ValueError(
"Cannot pass both generator and variance_noise. Please make sure that either `generator` or"
" `variance_noise` stays `None`."
)

if variance_noise is None:
variance_noise = torch.randn(model_output.shape, dtype=model_output.dtype, generator=generator).to(
device
)
if device.type == "mps":
# randn does not work reproducibly on mps
variance_noise = torch.randn(model_output.shape, dtype=model_output.dtype, generator=generator)
variance_noise = variance_noise.to(device)
else:
variance_noise = torch.randn(
model_output.shape, generator=generator, device=device, dtype=model_output.dtype
)
variance = self._get_variance(timestep, prev_timestep) ** (0.5) * eta * variance_noise

prev_sample = prev_sample + variance
Expand Down