Skip to content

Fix beta and exponential sigmas + add tests #9954

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 4 commits into from
Nov 20, 2024
Merged

Fix beta and exponential sigmas + add tests #9954

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
c635bc2
Fix beta and exponential sigmas + add tests
hlky Nov 18, 2024
f60e096
Fix ordering
hlky Nov 19, 2024
01c6f31
Merge branch 'main' into fix-beta-exponential
sayakpaul Nov 19, 2024
d304b1f
Fix unipc
hlky Nov 19, 2024
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
14 changes: 9 additions & 5 deletions src/diffusers/schedulers/scheduling_deis_multistep.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -266,18 +266,22 @@ def set_timesteps(self, num_inference_steps: int, device: Union[str, torch.devic
)

sigmas = np.array(((1 - self.alphas_cumprod) / self.alphas_cumprod) ** 0.5)
log_sigmas = np.log(sigmas)
if self.config.use_karras_sigmas:
log_sigmas = np.log(sigmas)
sigmas = np.flip(sigmas).copy()
sigmas = self._convert_to_karras(in_sigmas=sigmas, num_inference_steps=num_inference_steps)
timesteps = np.array([self._sigma_to_t(sigma, log_sigmas) for sigma in sigmas]).round()
sigmas = np.concatenate([sigmas, sigmas[-1:]]).astype(np.float32)
elif self.config.use_exponential_sigmas:
sigmas = self._convert_to_exponential(in_sigmas=sigmas, num_inference_steps=self.num_inference_steps)
sigmas = np.flip(sigmas).copy()
sigmas = self._convert_to_exponential(in_sigmas=sigmas, num_inference_steps=num_inference_steps)
timesteps = np.array([self._sigma_to_t(sigma, log_sigmas) for sigma in sigmas])
sigmas = np.concatenate([sigmas, sigmas[-1:]]).astype(np.float32)
elif self.config.use_beta_sigmas:
sigmas = self._convert_to_beta(in_sigmas=sigmas, num_inference_steps=self.num_inference_steps)
sigmas = np.flip(sigmas).copy()
sigmas = self._convert_to_beta(in_sigmas=sigmas, num_inference_steps=num_inference_steps)
timesteps = np.array([self._sigma_to_t(sigma, log_sigmas) for sigma in sigmas])
sigmas = np.concatenate([sigmas, sigmas[-1:]]).astype(np.float32)
else:
sigmas = np.interp(timesteps, np.arange(0, len(sigmas)), sigmas)
sigma_last = ((1 - self.alphas_cumprod[0]) / self.alphas_cumprod[0]) ** 0.5
Expand Down Expand Up @@ -408,7 +412,7 @@ def _convert_to_exponential(self, in_sigmas: torch.Tensor, num_inference_steps:
sigma_min = sigma_min if sigma_min is not None else in_sigmas[-1].item()
sigma_max = sigma_max if sigma_max is not None else in_sigmas[0].item()

sigmas = torch.linspace(math.log(sigma_max), math.log(sigma_min), num_inference_steps).exp()
sigmas = np.exp(np.linspace(math.log(sigma_max), math.log(sigma_min), num_inference_steps))
return sigmas

# Copied from diffusers.schedulers.scheduling_euler_discrete.EulerDiscreteScheduler._convert_to_beta
Expand All @@ -432,7 +436,7 @@ def _convert_to_beta(
sigma_min = sigma_min if sigma_min is not None else in_sigmas[-1].item()
sigma_max = sigma_max if sigma_max is not None else in_sigmas[0].item()

sigmas = torch.Tensor(
sigmas = np.array(
[
sigma_min + (ppf * (sigma_max - sigma_min))
for ppf in [
Expand Down
10 changes: 6 additions & 4 deletions src/diffusers/schedulers/scheduling_dpmsolver_multistep.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -400,10 +400,12 @@ def set_timesteps(
sigmas = np.exp(lambdas)
timesteps = np.array([self._sigma_to_t(sigma, log_sigmas) for sigma in sigmas]).round()
elif self.config.use_exponential_sigmas:
sigmas = self._convert_to_exponential(in_sigmas=sigmas, num_inference_steps=self.num_inference_steps)
sigmas = np.flip(sigmas).copy()
sigmas = self._convert_to_exponential(in_sigmas=sigmas, num_inference_steps=num_inference_steps)
timesteps = np.array([self._sigma_to_t(sigma, log_sigmas) for sigma in sigmas])
elif self.config.use_beta_sigmas:
sigmas = self._convert_to_beta(in_sigmas=sigmas, num_inference_steps=self.num_inference_steps)
sigmas = np.flip(sigmas).copy()
sigmas = self._convert_to_beta(in_sigmas=sigmas, num_inference_steps=num_inference_steps)
timesteps = np.array([self._sigma_to_t(sigma, log_sigmas) for sigma in sigmas])
else:
sigmas = np.interp(timesteps, np.arange(0, len(sigmas)), sigmas)
Expand Down Expand Up @@ -556,7 +558,7 @@ def _convert_to_exponential(self, in_sigmas: torch.Tensor, num_inference_steps:
sigma_min = sigma_min if sigma_min is not None else in_sigmas[-1].item()
sigma_max = sigma_max if sigma_max is not None else in_sigmas[0].item()

sigmas = torch.linspace(math.log(sigma_max), math.log(sigma_min), num_inference_steps).exp()
sigmas = np.exp(np.linspace(math.log(sigma_max), math.log(sigma_min), num_inference_steps))
return sigmas

# Copied from diffusers.schedulers.scheduling_euler_discrete.EulerDiscreteScheduler._convert_to_beta
Expand All @@ -580,7 +582,7 @@ def _convert_to_beta(
sigma_min = sigma_min if sigma_min is not None else in_sigmas[-1].item()
sigma_max = sigma_max if sigma_max is not None else in_sigmas[0].item()

sigmas = torch.Tensor(
sigmas = np.array(
[
sigma_min + (ppf * (sigma_max - sigma_min))
for ppf in [
Expand Down
8 changes: 4 additions & 4 deletions src/diffusers/schedulers/scheduling_dpmsolver_multistep_inverse.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -287,10 +287,10 @@ def set_timesteps(self, num_inference_steps: int = None, device: Union[str, torc
timesteps = timesteps.copy().astype(np.int64)
sigmas = np.concatenate([sigmas, sigmas[-1:]]).astype(np.float32)
elif self.config.use_exponential_sigmas:
sigmas = self._convert_to_exponential(in_sigmas=sigmas, num_inference_steps=self.num_inference_steps)
sigmas = self._convert_to_exponential(in_sigmas=sigmas, num_inference_steps=num_inference_steps)
timesteps = np.array([self._sigma_to_t(sigma, log_sigmas) for sigma in sigmas])
elif self.config.use_beta_sigmas:
sigmas = self._convert_to_beta(in_sigmas=sigmas, num_inference_steps=self.num_inference_steps)
sigmas = self._convert_to_beta(in_sigmas=sigmas, num_inference_steps=num_inference_steps)
timesteps = np.array([self._sigma_to_t(sigma, log_sigmas) for sigma in sigmas])
else:
sigmas = np.interp(timesteps, np.arange(0, len(sigmas)), sigmas)
Expand Down Expand Up @@ -429,7 +429,7 @@ def _convert_to_exponential(self, in_sigmas: torch.Tensor, num_inference_steps:
sigma_min = sigma_min if sigma_min is not None else in_sigmas[-1].item()
sigma_max = sigma_max if sigma_max is not None else in_sigmas[0].item()

sigmas = torch.linspace(math.log(sigma_max), math.log(sigma_min), num_inference_steps).exp()
sigmas = np.exp(np.linspace(math.log(sigma_max), math.log(sigma_min), num_inference_steps))
return sigmas

# Copied from diffusers.schedulers.scheduling_euler_discrete.EulerDiscreteScheduler._convert_to_beta
Expand All @@ -453,7 +453,7 @@ def _convert_to_beta(
sigma_min = sigma_min if sigma_min is not None else in_sigmas[-1].item()
sigma_max = sigma_max if sigma_max is not None else in_sigmas[0].item()

sigmas = torch.Tensor(
sigmas = np.array(
[
sigma_min + (ppf * (sigma_max - sigma_min))
for ppf in [
Expand Down
8 changes: 4 additions & 4 deletions src/diffusers/schedulers/scheduling_dpmsolver_sde.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -380,10 +380,10 @@ def set_timesteps(
sigmas = self._convert_to_karras(in_sigmas=sigmas)
timesteps = np.array([self._sigma_to_t(sigma, log_sigmas) for sigma in sigmas])
elif self.config.use_exponential_sigmas:
sigmas = self._convert_to_exponential(in_sigmas=sigmas, num_inference_steps=self.num_inference_steps)
sigmas = self._convert_to_exponential(in_sigmas=sigmas, num_inference_steps=num_inference_steps)
timesteps = np.array([self._sigma_to_t(sigma, log_sigmas) for sigma in sigmas])
elif self.config.use_beta_sigmas:
sigmas = self._convert_to_beta(in_sigmas=sigmas, num_inference_steps=self.num_inference_steps)
sigmas = self._convert_to_beta(in_sigmas=sigmas, num_inference_steps=num_inference_steps)
timesteps = np.array([self._sigma_to_t(sigma, log_sigmas) for sigma in sigmas])

second_order_timesteps = self._second_order_timesteps(sigmas, log_sigmas)
Expand Down Expand Up @@ -484,7 +484,7 @@ def _convert_to_exponential(self, in_sigmas: torch.Tensor, num_inference_steps:
sigma_min = sigma_min if sigma_min is not None else in_sigmas[-1].item()
sigma_max = sigma_max if sigma_max is not None else in_sigmas[0].item()

sigmas = torch.linspace(math.log(sigma_max), math.log(sigma_min), num_inference_steps).exp()
sigmas = np.exp(np.linspace(math.log(sigma_max), math.log(sigma_min), num_inference_steps))
return sigmas

# Copied from diffusers.schedulers.scheduling_euler_discrete.EulerDiscreteScheduler._convert_to_beta
Expand All @@ -508,7 +508,7 @@ def _convert_to_beta(
sigma_min = sigma_min if sigma_min is not None else in_sigmas[-1].item()
sigma_max = sigma_max if sigma_max is not None else in_sigmas[0].item()

sigmas = torch.Tensor(
sigmas = np.array(
[
sigma_min + (ppf * (sigma_max - sigma_min))
for ppf in [
Expand Down
12 changes: 7 additions & 5 deletions src/diffusers/schedulers/scheduling_dpmsolver_singlestep.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -339,16 +339,18 @@ def set_timesteps(
)

sigmas = np.array(((1 - self.alphas_cumprod) / self.alphas_cumprod) ** 0.5)
log_sigmas = np.log(sigmas)
if self.config.use_karras_sigmas:
log_sigmas = np.log(sigmas)
sigmas = np.flip(sigmas).copy()
sigmas = self._convert_to_karras(in_sigmas=sigmas, num_inference_steps=num_inference_steps)
timesteps = np.array([self._sigma_to_t(sigma, log_sigmas) for sigma in sigmas]).round()
elif self.config.use_exponential_sigmas:
sigmas = self._convert_to_exponential(in_sigmas=sigmas, num_inference_steps=self.num_inference_steps)
sigmas = np.flip(sigmas).copy()
sigmas = self._convert_to_exponential(in_sigmas=sigmas, num_inference_steps=num_inference_steps)
timesteps = np.array([self._sigma_to_t(sigma, log_sigmas) for sigma in sigmas])
elif self.config.use_beta_sigmas:
sigmas = self._convert_to_beta(in_sigmas=sigmas, num_inference_steps=self.num_inference_steps)
sigmas = np.flip(sigmas).copy()
sigmas = self._convert_to_beta(in_sigmas=sigmas, num_inference_steps=num_inference_steps)
timesteps = np.array([self._sigma_to_t(sigma, log_sigmas) for sigma in sigmas])
else:
sigmas = np.interp(timesteps, np.arange(0, len(sigmas)), sigmas)
Expand Down Expand Up @@ -498,7 +500,7 @@ def _convert_to_exponential(self, in_sigmas: torch.Tensor, num_inference_steps:
sigma_min = sigma_min if sigma_min is not None else in_sigmas[-1].item()
sigma_max = sigma_max if sigma_max is not None else in_sigmas[0].item()

sigmas = torch.linspace(math.log(sigma_max), math.log(sigma_min), num_inference_steps).exp()
sigmas = np.exp(np.linspace(math.log(sigma_max), math.log(sigma_min), num_inference_steps))
return sigmas

# Copied from diffusers.schedulers.scheduling_euler_discrete.EulerDiscreteScheduler._convert_to_beta
Expand All @@ -522,7 +524,7 @@ def _convert_to_beta(
sigma_min = sigma_min if sigma_min is not None else in_sigmas[-1].item()
sigma_max = sigma_max if sigma_max is not None else in_sigmas[0].item()

sigmas = torch.Tensor(
sigmas = np.array(
[
sigma_min + (ppf * (sigma_max - sigma_min))
for ppf in [
Expand Down
8 changes: 4 additions & 4 deletions src/diffusers/schedulers/scheduling_euler_discrete.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -419,11 +419,11 @@ def set_timesteps(
timesteps = np.array([self._sigma_to_t(sigma, log_sigmas) for sigma in sigmas])

elif self.config.use_exponential_sigmas:
sigmas = self._convert_to_exponential(in_sigmas=sigmas, num_inference_steps=self.num_inference_steps)
sigmas = self._convert_to_exponential(in_sigmas=sigmas, num_inference_steps=num_inference_steps)
timesteps = np.array([self._sigma_to_t(sigma, log_sigmas) for sigma in sigmas])

elif self.config.use_beta_sigmas:
sigmas = self._convert_to_beta(in_sigmas=sigmas, num_inference_steps=self.num_inference_steps)
sigmas = self._convert_to_beta(in_sigmas=sigmas, num_inference_steps=num_inference_steps)
timesteps = np.array([self._sigma_to_t(sigma, log_sigmas) for sigma in sigmas])

if self.config.final_sigmas_type == "sigma_min":
Expand Down Expand Up @@ -517,7 +517,7 @@ def _convert_to_exponential(self, in_sigmas: torch.Tensor, num_inference_steps:
sigma_min = sigma_min if sigma_min is not None else in_sigmas[-1].item()
sigma_max = sigma_max if sigma_max is not None else in_sigmas[0].item()

sigmas = torch.linspace(math.log(sigma_max), math.log(sigma_min), num_inference_steps).exp()
sigmas = np.exp(np.linspace(math.log(sigma_max), math.log(sigma_min), num_inference_steps))
return sigmas

def _convert_to_beta(
Expand All @@ -540,7 +540,7 @@ def _convert_to_beta(
sigma_min = sigma_min if sigma_min is not None else in_sigmas[-1].item()
sigma_max = sigma_max if sigma_max is not None else in_sigmas[0].item()

sigmas = torch.Tensor(
sigmas = np.array(
[
sigma_min + (ppf * (sigma_max - sigma_min))
for ppf in [
Expand Down
8 changes: 4 additions & 4 deletions src/diffusers/schedulers/scheduling_heun_discrete.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -329,10 +329,10 @@ def set_timesteps(
sigmas = self._convert_to_karras(in_sigmas=sigmas, num_inference_steps=self.num_inference_steps)
timesteps = np.array([self._sigma_to_t(sigma, log_sigmas) for sigma in sigmas])
elif self.config.use_exponential_sigmas:
sigmas = self._convert_to_exponential(in_sigmas=sigmas, num_inference_steps=self.num_inference_steps)
sigmas = self._convert_to_exponential(in_sigmas=sigmas, num_inference_steps=num_inference_steps)
timesteps = np.array([self._sigma_to_t(sigma, log_sigmas) for sigma in sigmas])
elif self.config.use_beta_sigmas:
sigmas = self._convert_to_beta(in_sigmas=sigmas, num_inference_steps=self.num_inference_steps)
sigmas = self._convert_to_beta(in_sigmas=sigmas, num_inference_steps=num_inference_steps)
timesteps = np.array([self._sigma_to_t(sigma, log_sigmas) for sigma in sigmas])

sigmas = np.concatenate([sigmas, [0.0]]).astype(np.float32)
Expand Down Expand Up @@ -421,7 +421,7 @@ def _convert_to_exponential(self, in_sigmas: torch.Tensor, num_inference_steps:
sigma_min = sigma_min if sigma_min is not None else in_sigmas[-1].item()
sigma_max = sigma_max if sigma_max is not None else in_sigmas[0].item()

sigmas = torch.linspace(math.log(sigma_max), math.log(sigma_min), num_inference_steps).exp()
sigmas = np.exp(np.linspace(math.log(sigma_max), math.log(sigma_min), num_inference_steps))
return sigmas

# Copied from diffusers.schedulers.scheduling_euler_discrete.EulerDiscreteScheduler._convert_to_beta
Expand All @@ -445,7 +445,7 @@ def _convert_to_beta(
sigma_min = sigma_min if sigma_min is not None else in_sigmas[-1].item()
sigma_max = sigma_max if sigma_max is not None else in_sigmas[0].item()

sigmas = torch.Tensor(
sigmas = np.array(
[
sigma_min + (ppf * (sigma_max - sigma_min))
for ppf in [
Expand Down
8 changes: 4 additions & 4 deletions src/diffusers/schedulers/scheduling_k_dpm_2_ancestral_discrete.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -289,10 +289,10 @@ def set_timesteps(
sigmas = self._convert_to_karras(in_sigmas=sigmas, num_inference_steps=num_inference_steps)
timesteps = np.array([self._sigma_to_t(sigma, log_sigmas) for sigma in sigmas]).round()
elif self.config.use_exponential_sigmas:
sigmas = self._convert_to_exponential(in_sigmas=sigmas, num_inference_steps=self.num_inference_steps)
sigmas = self._convert_to_exponential(in_sigmas=sigmas, num_inference_steps=num_inference_steps)
timesteps = np.array([self._sigma_to_t(sigma, log_sigmas) for sigma in sigmas])
elif self.config.use_beta_sigmas:
sigmas = self._convert_to_beta(in_sigmas=sigmas, num_inference_steps=self.num_inference_steps)
sigmas = self._convert_to_beta(in_sigmas=sigmas, num_inference_steps=num_inference_steps)
timesteps = np.array([self._sigma_to_t(sigma, log_sigmas) for sigma in sigmas])

self.log_sigmas = torch.from_numpy(log_sigmas).to(device)
Expand Down Expand Up @@ -409,7 +409,7 @@ def _convert_to_exponential(self, in_sigmas: torch.Tensor, num_inference_steps:
sigma_min = sigma_min if sigma_min is not None else in_sigmas[-1].item()
sigma_max = sigma_max if sigma_max is not None else in_sigmas[0].item()

sigmas = torch.linspace(math.log(sigma_max), math.log(sigma_min), num_inference_steps).exp()
sigmas = np.exp(np.linspace(math.log(sigma_max), math.log(sigma_min), num_inference_steps))
return sigmas

# Copied from diffusers.schedulers.scheduling_euler_discrete.EulerDiscreteScheduler._convert_to_beta
Expand All @@ -433,7 +433,7 @@ def _convert_to_beta(
sigma_min = sigma_min if sigma_min is not None else in_sigmas[-1].item()
sigma_max = sigma_max if sigma_max is not None else in_sigmas[0].item()

sigmas = torch.Tensor(
sigmas = np.array(
[
sigma_min + (ppf * (sigma_max - sigma_min))
for ppf in [
Expand Down
8 changes: 4 additions & 4 deletions src/diffusers/schedulers/scheduling_k_dpm_2_discrete.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -288,10 +288,10 @@ def set_timesteps(
sigmas = self._convert_to_karras(in_sigmas=sigmas, num_inference_steps=num_inference_steps)
timesteps = np.array([self._sigma_to_t(sigma, log_sigmas) for sigma in sigmas]).round()
elif self.config.use_exponential_sigmas:
sigmas = self._convert_to_exponential(in_sigmas=sigmas, num_inference_steps=self.num_inference_steps)
sigmas = self._convert_to_exponential(in_sigmas=sigmas, num_inference_steps=num_inference_steps)
timesteps = np.array([self._sigma_to_t(sigma, log_sigmas) for sigma in sigmas])
elif self.config.use_beta_sigmas:
sigmas = self._convert_to_beta(in_sigmas=sigmas, num_inference_steps=self.num_inference_steps)
sigmas = self._convert_to_beta(in_sigmas=sigmas, num_inference_steps=num_inference_steps)
timesteps = np.array([self._sigma_to_t(sigma, log_sigmas) for sigma in sigmas])

self.log_sigmas = torch.from_numpy(log_sigmas).to(device=device)
Expand Down Expand Up @@ -422,7 +422,7 @@ def _convert_to_exponential(self, in_sigmas: torch.Tensor, num_inference_steps:
sigma_min = sigma_min if sigma_min is not None else in_sigmas[-1].item()
sigma_max = sigma_max if sigma_max is not None else in_sigmas[0].item()

sigmas = torch.linspace(math.log(sigma_max), math.log(sigma_min), num_inference_steps).exp()
sigmas = np.exp(np.linspace(math.log(sigma_max), math.log(sigma_min), num_inference_steps))
return sigmas

# Copied from diffusers.schedulers.scheduling_euler_discrete.EulerDiscreteScheduler._convert_to_beta
Expand All @@ -446,7 +446,7 @@ def _convert_to_beta(
sigma_min = sigma_min if sigma_min is not None else in_sigmas[-1].item()
sigma_max = sigma_max if sigma_max is not None else in_sigmas[0].item()

sigmas = torch.Tensor(
sigmas = np.array(
[
sigma_min + (ppf * (sigma_max - sigma_min))
for ppf in [
Expand Down
Loading
Loading