Skip to content

[Type hint] PNDM schedulers #335

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
Sep 4, 2022
Merged

[Type hint] PNDM schedulers #335

Changes from 3 commits
Commits
Show all changes
6 commits
Select commit Hold shift + click to select a range
7d1f48f
[Type hint] PNDM Schedulers
daspartho Sep 2, 2022
58f6b95
ran make style
daspartho Sep 2, 2022
71a373c
updated timesteps type hint
daspartho Sep 3, 2022
de51dcb
apply suggestions from code review
daspartho Sep 4, 2022
e9c24d5
ran make style
daspartho Sep 4, 2022
50a6d74
removed unused import
daspartho Sep 4, 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
23 changes: 14 additions & 9 deletions src/diffusers/schedulers/scheduling_pndm.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -15,7 +15,7 @@
# DISCLAIMER: This file is strongly influenced by https://github.com/ermongroup/ddim

import math
from typing import Union
from typing import Optional, Union

import numpy as np
import torch
Expand Down Expand Up @@ -51,12 +51,12 @@ class PNDMScheduler(SchedulerMixin, ConfigMixin):
@register_to_config
def __init__(
self,
num_train_timesteps=1000,
beta_start=0.0001,
beta_end=0.02,
beta_schedule="linear",
tensor_format="pt",
skip_prk_steps=False,
num_train_timesteps: int = 1000,
beta_start: float = 0.0001,
beta_end: float = 0.02,
beta_schedule: Optional[str] = "linear",
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Suggested change
beta_schedule: Optional[str] = "linear",
beta_schedule: str = "linear",

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

beta_schedule has to be string I think :-)

tensor_format: Optional[str] = "pt",
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Suggested change
tensor_format: Optional[str] = "pt",
tensor_format: str = "pt",

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Think the tensor_format cannot be None

skip_prk_steps: bool = False,
):

if beta_schedule == "linear":
Expand Down Expand Up @@ -97,7 +97,7 @@ def __init__(
self.tensor_format = tensor_format
self.set_format(tensor_format=tensor_format)

def set_timesteps(self, num_inference_steps, offset=0):
def set_timesteps(self, num_inference_steps: int, offset=0):
Copy link
Contributor

@patrickvonplaten patrickvonplaten Sep 4, 2022 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Suggested change
def set_timesteps(self, num_inference_steps: int, offset=0):
def set_timesteps(self, num_inference_steps: int, offset: int = 0) -> torch.FloatTensor:

self.num_inference_steps = num_inference_steps
self._timesteps = list(
range(0, self.config.num_train_timesteps, self.config.num_train_timesteps // num_inference_steps)
Expand Down Expand Up @@ -264,7 +264,12 @@ def _get_prev_sample(self, sample, timestep, timestep_prev, model_output):

return prev_sample

def add_noise(self, original_samples, noise, timesteps):
def add_noise(
self,
original_samples: Union[torch.FloatTensor, np.ndarray],
noise: Union[torch.FloatTensor, np.ndarray],
timesteps: Union[torch.IntTensor, np.ndarray],
):
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Suggested change
):
) -> torch.Tensor:

sqrt_alpha_prod = self.alphas_cumprod[timesteps] ** 0.5
sqrt_alpha_prod = self.match_shape(sqrt_alpha_prod, original_samples)
sqrt_one_minus_alpha_prod = (1 - self.alphas_cumprod[timesteps]) ** 0.5
Expand Down