1
0
mirror of https://github.com/huggingface/diffusers.git synced 2026-01-29 07:22:12 +03:00

[bug fix] dpm multistep solver duplicate timesteps

This commit is contained in:
William Berman
2023-04-04 09:29:15 -07:00
committed by Daniel Gu
parent 3de609dd66
commit b8bfa562dc

View File

@@ -192,14 +192,22 @@ class DPMSolverMultistepScheduler(SchedulerMixin, ConfigMixin):
device (`str` or `torch.device`, optional):
the device to which the timesteps should be moved to. If `None`, the timesteps are not moved.
"""
self.num_inference_steps = num_inference_steps
timesteps = (
np.linspace(0, self.num_train_timesteps - 1, num_inference_steps + 1)
.round()[::-1][:-1]
.copy()
.astype(np.int64)
)
# when num_inference_steps == num_train_timesteps, we can end up with
# duplicates in timesteps.
_, unique_indices = np.unique(timesteps, return_index=True)
timesteps = timesteps[np.sort(unique_indices)]
self.timesteps = torch.from_numpy(timesteps).to(device)
self.num_inference_steps = len(timesteps)
self.model_outputs = [
None,
] * self.config.solver_order