JaehaerysNL/StableDiffusion
1
0
Fork 0
mirror of https://github.com/Stability-AI/stablediffusion.git synced 2024-12-22 15:44:58 +00:00
Code Issues Projects Releases Packages Wiki Activity

Reformatting

Browse source
This commit is contained in:
Andres Caicedo 2023-05-08 14:42:53 +02:00
parent a825f77092
commit 462a9d3298
No known key found for this signature in database
GPG key ID: 6E797C4F5A327624
15 changed files with 47 additions and 72 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

3
ldm/models/diffusion/ddim.py
View file

@ -360,7 +360,7 @@ class DDIMSampler(object):
raise NotImplementedError() raise NotImplementedError()
# direction pointing to x_t # direction pointing to x_t
dir_xt = (1.0 - a_prev - sigma_t ** 2).sqrt() * e_t dir_xt = (1.0 - a_prev - sigma_t**2).sqrt() * e_t
noise = sigma_t * noise_like(x.shape, device, repeat_noise) * temperature noise = sigma_t * noise_like(x.shape, device, repeat_noise) * temperature
if noise_dropout > 0.0: if noise_dropout > 0.0:
noise = torch.nn.functional.dropout(noise, p=noise_dropout) noise = torch.nn.functional.dropout(noise, p=noise_dropout)
@ -472,7 +472,6 @@ class DDIMSampler(object):
use_original_steps=False, use_original_steps=False,
callback=None, callback=None,
): ):
timesteps = ( timesteps = (
np.arange(self.ddpm_num_timesteps) np.arange(self.ddpm_num_timesteps)
if use_original_steps if use_original_steps

5
ldm/models/diffusion/ddpm.py
View file

@ -242,7 +242,7 @@ class DDPM(pl.LightningModule):
) )
if self.parameterization == "eps": if self.parameterization == "eps":
lvlb_weights = self.betas ** 2 / ( lvlb_weights = self.betas**2 / (
2 2
* self.posterior_variance * self.posterior_variance
* to_torch(alphas) * to_torch(alphas)
@ -256,7 +256,7 @@ class DDPM(pl.LightningModule):
) )
elif self.parameterization == "v": elif self.parameterization == "v":
lvlb_weights = torch.ones_like( lvlb_weights = torch.ones_like(
self.betas ** 2 self.betas**2
/ ( / (
2 2
* self.posterior_variance * self.posterior_variance
@ -1358,7 +1358,6 @@ class LatentDiffusion(DDPM):
start_T=None, start_T=None,
log_every_t=None, log_every_t=None,
): ):
if not log_every_t: if not log_every_t:
log_every_t = self.log_every_t log_every_t = self.log_every_t
device = self.betas.device device = self.betas.device

2
ldm/models/diffusion/plms.py
View file

@ -339,7 +339,7 @@ class PLMSSampler(object):
if dynamic_threshold is not None: if dynamic_threshold is not None:
pred_x0 = norm_thresholding(pred_x0, dynamic_threshold) pred_x0 = norm_thresholding(pred_x0, dynamic_threshold)
# direction pointing to x_t # direction pointing to x_t
dir_xt = (1.0 - a_prev - sigma_t ** 2).sqrt() * e_t dir_xt = (1.0 - a_prev - sigma_t**2).sqrt() * e_t
noise = sigma_t * noise_like(x.shape, device, repeat_noise) * temperature noise = sigma_t * noise_like(x.shape, device, repeat_noise) * temperature
if noise_dropout > 0.0: if noise_dropout > 0.0:
noise = torch.nn.functional.dropout(noise, p=noise_dropout) noise = torch.nn.functional.dropout(noise, p=noise_dropout)

2
ldm/modules/attention.py
View file

@ -144,7 +144,7 @@ class CrossAttention(nn.Module):
inner_dim = dim_head * heads inner_dim = dim_head * heads
context_dim = default(context_dim, query_dim) context_dim = default(context_dim, query_dim)
self.scale = dim_head ** -0.5 self.scale = dim_head**-0.5
self.heads = heads self.heads = heads
self.to_q = nn.Linear(query_dim, inner_dim, bias=False) self.to_q = nn.Linear(query_dim, inner_dim, bias=False)

4
ldm/modules/diffusionmodules/openaimodel.py
View file

@ -43,7 +43,7 @@ class AttentionPool2d(nn.Module):
): ):
super().__init__() super().__init__()
self.positional_embedding = nn.Parameter( self.positional_embedding = nn.Parameter(
th.randn(embed_dim, spacial_dim ** 2 + 1) / embed_dim ** 0.5 th.randn(embed_dim, spacial_dim**2 + 1) / embed_dim**0.5
) )
self.qkv_proj = conv_nd(1, embed_dim, 3 * embed_dim, 1) self.qkv_proj = conv_nd(1, embed_dim, 3 * embed_dim, 1)
self.c_proj = conv_nd(1, embed_dim, output_dim or embed_dim, 1) self.c_proj = conv_nd(1, embed_dim, output_dim or embed_dim, 1)
@ -354,7 +354,7 @@ def count_flops_attn(model, _x, y):
# We perform two matmuls with the same number of ops. # We perform two matmuls with the same number of ops.
# The first computes the weight matrix, the second computes # The first computes the weight matrix, the second computes
# the combination of the value vectors. # the combination of the value vectors.
matmul_ops = 2 * b * (num_spatial ** 2) * c matmul_ops = 2 * b * (num_spatial**2) * c
model.total_ops += th.DoubleTensor([matmul_ops]) model.total_ops += th.DoubleTensor([matmul_ops])

2
ldm/modules/diffusionmodules/util.py
View file

@ -25,7 +25,7 @@ def make_beta_schedule(
if schedule == "linear": if schedule == "linear":
betas = ( betas = (
torch.linspace( torch.linspace(
linear_start ** 0.5, linear_end ** 0.5, n_timestep, dtype=torch.float64 linear_start**0.5, linear_end**0.5, n_timestep, dtype=torch.float64
) )
** 2 ** 2
) )

6
ldm/modules/image_degradation/bsrgan.py
View file

@ -403,7 +403,7 @@ def add_Gaussian_noise(img, noise_level1=2, noise_level2=25):
U = orth(np.random.rand(3, 3)) U = orth(np.random.rand(3, 3))
conv = np.dot(np.dot(np.transpose(U), D), U) conv = np.dot(np.dot(np.transpose(U), D), U)
img = img + np.random.multivariate_normal( img = img + np.random.multivariate_normal(
[0, 0, 0], np.abs(L ** 2 * conv), img.shape[:2] [0, 0, 0], np.abs(L**2 * conv), img.shape[:2]
).astype(np.float32) ).astype(np.float32)
img = np.clip(img, 0.0, 1.0) img = np.clip(img, 0.0, 1.0)
return img return img
@ -427,7 +427,7 @@ def add_speckle_noise(img, noise_level1=2, noise_level2=25):
U = orth(np.random.rand(3, 3)) U = orth(np.random.rand(3, 3))
conv = np.dot(np.dot(np.transpose(U), D), U) conv = np.dot(np.dot(np.transpose(U), D), U)
img += img * np.random.multivariate_normal( img += img * np.random.multivariate_normal(
[0, 0, 0], np.abs(L ** 2 * conv), img.shape[:2] [0, 0, 0], np.abs(L**2 * conv), img.shape[:2]
).astype(np.float32) ).astype(np.float32)
img = np.clip(img, 0.0, 1.0) img = np.clip(img, 0.0, 1.0)
return img return img
@ -519,7 +519,6 @@ def degradation_bsrgan(img, sf=4, lq_patchsize=72, isp_model=None):
) )
for i in shuffle_order: for i in shuffle_order:
if i == 0: if i == 0:
img = add_blur(img, sf=sf) img = add_blur(img, sf=sf)
@ -623,7 +622,6 @@ def degradation_bsrgan_variant(image, sf=4, isp_model=None):
) )
for i in shuffle_order: for i in shuffle_order:
if i == 0: if i == 0:
image = add_blur(image, sf=sf) image = add_blur(image, sf=sf)

6
ldm/modules/image_degradation/bsrgan_light.py
View file

@ -404,7 +404,7 @@ def add_Gaussian_noise(img, noise_level1=2, noise_level2=25):
U = orth(np.random.rand(3, 3)) U = orth(np.random.rand(3, 3))
conv = np.dot(np.dot(np.transpose(U), D), U) conv = np.dot(np.dot(np.transpose(U), D), U)
img = img + np.random.multivariate_normal( img = img + np.random.multivariate_normal(
[0, 0, 0], np.abs(L ** 2 * conv), img.shape[:2] [0, 0, 0], np.abs(L**2 * conv), img.shape[:2]
).astype(np.float32) ).astype(np.float32)
img = np.clip(img, 0.0, 1.0) img = np.clip(img, 0.0, 1.0)
return img return img
@ -428,7 +428,7 @@ def add_speckle_noise(img, noise_level1=2, noise_level2=25):
U = orth(np.random.rand(3, 3)) U = orth(np.random.rand(3, 3))
conv = np.dot(np.dot(np.transpose(U), D), U) conv = np.dot(np.dot(np.transpose(U), D), U)
img += img * np.random.multivariate_normal( img += img * np.random.multivariate_normal(
[0, 0, 0], np.abs(L ** 2 * conv), img.shape[:2] [0, 0, 0], np.abs(L**2 * conv), img.shape[:2]
).astype(np.float32) ).astype(np.float32)
img = np.clip(img, 0.0, 1.0) img = np.clip(img, 0.0, 1.0)
return img return img
@ -520,7 +520,6 @@ def degradation_bsrgan(img, sf=4, lq_patchsize=72, isp_model=None):
) )
for i in shuffle_order: for i in shuffle_order:
if i == 0: if i == 0:
img = add_blur(img, sf=sf) img = add_blur(img, sf=sf)
@ -624,7 +623,6 @@ def degradation_bsrgan_variant(image, sf=4, isp_model=None, up=False):
) )
for i in shuffle_order: for i in shuffle_order:
if i == 0: if i == 0:
image = add_blur(image, sf=sf) image = add_blur(image, sf=sf)

76
ldm/modules/image_degradation/utils_image.py
View file

@ -271,22 +271,18 @@ def read_img(path):
def uint2single(img): def uint2single(img):
return np.float32(img / 255.0) return np.float32(img / 255.0)
def single2uint(img): def single2uint(img):
return np.uint8((img.clip(0, 1) * 255.0).round()) return np.uint8((img.clip(0, 1) * 255.0).round())
def uint162single(img): def uint162single(img):
return np.float32(img / 65535.0) return np.float32(img / 65535.0)
def single2uint16(img): def single2uint16(img):
return np.uint16((img.clip(0, 1) * 65535.0).round()) return np.uint16((img.clip(0, 1) * 65535.0).round())
@ -586,18 +582,14 @@ def rgb2ycbcr(img, only_y=True):
if only_y: if only_y:
rlt = np.dot(img, [65.481, 128.553, 24.966]) / 255.0 + 16.0 rlt = np.dot(img, [65.481, 128.553, 24.966]) / 255.0 + 16.0
else: else:
rlt = ( rlt = np.matmul(
np.matmul( img,
img, [
[ [65.481, -37.797, 112.0],
[65.481, -37.797, 112.0], [128.553, -74.203, -93.786],
[128.553, -74.203, -93.786], [24.966, 112.0, -18.214],
[24.966, 112.0, -18.214], ],
], ) / 255.0 + [16, 128, 128]
)
/ 255.0
+ [16, 128, 128]
)
if in_img_type == np.uint8: if in_img_type == np.uint8:
rlt = rlt.round() rlt = rlt.round()
else: else:
@ -616,18 +608,14 @@ def ycbcr2rgb(img):
if in_img_type != np.uint8: if in_img_type != np.uint8:
img *= 255.0 img *= 255.0
# convert # convert
rlt = ( rlt = np.matmul(
np.matmul( img,
img, [
[ [0.00456621, 0.00456621, 0.00456621],
[0.00456621, 0.00456621, 0.00456621], [0, -0.00153632, 0.00791071],
[0, -0.00153632, 0.00791071], [0.00625893, -0.00318811, 0],
[0.00625893, -0.00318811, 0], ],
], ) * 255.0 + [-222.921, 135.576, -276.836]
)
* 255.0
+ [-222.921, 135.576, -276.836]
)
if in_img_type == np.uint8: if in_img_type == np.uint8:
rlt = rlt.round() rlt = rlt.round()
else: else:
@ -650,18 +638,14 @@ def bgr2ycbcr(img, only_y=True):
if only_y: if only_y:
rlt = np.dot(img, [24.966, 128.553, 65.481]) / 255.0 + 16.0 rlt = np.dot(img, [24.966, 128.553, 65.481]) / 255.0 + 16.0
else: else:
rlt = ( rlt = np.matmul(
np.matmul( img,
img, [
[ [24.966, 112.0, -18.214],
[24.966, 112.0, -18.214], [128.553, -74.203, -93.786],
[128.553, -74.203, -93.786], [65.481, -37.797, 112.0],
[65.481, -37.797, 112.0], ],
], ) / 255.0 + [16, 128, 128]
)
/ 255.0
+ [16, 128, 128]
)
if in_img_type == np.uint8: if in_img_type == np.uint8:
rlt = rlt.round() rlt = rlt.round()
else: else:
@ -752,11 +736,11 @@ def ssim(img1, img2):
mu1 = cv2.filter2D(img1, -1, window)[5:-5, 5:-5] # valid mu1 = cv2.filter2D(img1, -1, window)[5:-5, 5:-5] # valid
mu2 = cv2.filter2D(img2, -1, window)[5:-5, 5:-5] mu2 = cv2.filter2D(img2, -1, window)[5:-5, 5:-5]
mu1_sq = mu1 ** 2 mu1_sq = mu1**2
mu2_sq = mu2 ** 2 mu2_sq = mu2**2
mu1_mu2 = mu1 * mu2 mu1_mu2 = mu1 * mu2
sigma1_sq = cv2.filter2D(img1 ** 2, -1, window)[5:-5, 5:-5] - mu1_sq sigma1_sq = cv2.filter2D(img1**2, -1, window)[5:-5, 5:-5] - mu1_sq
sigma2_sq = cv2.filter2D(img2 ** 2, -1, window)[5:-5, 5:-5] - mu2_sq sigma2_sq = cv2.filter2D(img2**2, -1, window)[5:-5, 5:-5] - mu2_sq
sigma12 = cv2.filter2D(img1 * img2, -1, window)[5:-5, 5:-5] - mu1_mu2 sigma12 = cv2.filter2D(img1 * img2, -1, window)[5:-5, 5:-5] - mu1_mu2
ssim_map = ((2 * mu1_mu2 + C1) * (2 * sigma12 + C2)) / ( ssim_map = ((2 * mu1_mu2 + C1) * (2 * sigma12 + C2)) / (
@ -775,8 +759,8 @@ def ssim(img1, img2):
# matlab 'imresize' function, now only support 'bicubic' # matlab 'imresize' function, now only support 'bicubic'
def cubic(x): def cubic(x):
absx = torch.abs(x) absx = torch.abs(x)
absx2 = absx ** 2 absx2 = absx**2
absx3 = absx ** 3 absx3 = absx**3
return (1.5 * absx3 - 2.5 * absx2 + 1) * ((absx <= 1).type_as(absx)) + ( return (1.5 * absx3 - 2.5 * absx2 + 1) * ((absx <= 1).type_as(absx)) + (
-0.5 * absx3 + 2.5 * absx2 - 4 * absx + 2 -0.5 * absx3 + 2.5 * absx2 - 4 * absx + 2
) * (((absx > 1) * (absx <= 2)).type_as(absx)) ) * (((absx > 1) * (absx <= 2)).type_as(absx))

2
ldm/modules/karlo/kakao/models/clip.py
View file

@ -106,7 +106,7 @@ class CustomizedCLIP(CLIP):
) )
vision_patch_size = None vision_patch_size = None
assert ( assert (
output_width ** 2 + 1 output_width**2 + 1
== state_dict["visual.attnpool.positional_embedding"].shape[0] == state_dict["visual.attnpool.positional_embedding"].shape[0]
) )
image_resolution = output_width * 32 image_resolution = output_width * 32

6
ldm/modules/karlo/kakao/modules/diffusion/gaussian_diffusion.py
View file

@ -26,8 +26,8 @@ def get_beta_schedule(beta_schedule, *, beta_start, beta_end, num_diffusion_time
if beta_schedule == "quad": if beta_schedule == "quad":
betas = ( betas = (
np.linspace( np.linspace(
beta_start ** 0.5, beta_start**0.5,
beta_end ** 0.5, beta_end**0.5,
num_diffusion_timesteps, num_diffusion_timesteps,
dtype=np.float64, dtype=np.float64,
) )
@ -681,7 +681,7 @@ class GaussianDiffusion(th.nn.Module):
noise = th.randn_like(x) noise = th.randn_like(x)
mean_pred = ( mean_pred = (
out["pred_xstart"] * th.sqrt(alpha_bar_prev) out["pred_xstart"] * th.sqrt(alpha_bar_prev)
+ th.sqrt(1 - alpha_bar_prev - sigma ** 2) * eps + th.sqrt(1 - alpha_bar_prev - sigma**2) * eps
) )
nonzero_mask = ( nonzero_mask = (
(t != 0).float().view(-1, *([1] * (len(x.shape) - 1))) (t != 0).float().view(-1, *([1] * (len(x.shape) - 1)))

1
ldm/modules/karlo/kakao/sampler.py
View file

@ -36,7 +36,6 @@ class T2ISampler(BaseSampler):
clip_stat_path: str, clip_stat_path: str,
sampling_type: str = "default", sampling_type: str = "default",
): ):
model = cls( model = cls(
root_dir=root_dir, root_dir=root_dir,
sampling_type=sampling_type, sampling_type=sampling_type,

1
ldm/modules/midas/midas/dpt_depth.py
View file

@ -33,7 +33,6 @@ class DPT(BaseModel):
channels_last=False, channels_last=False,
use_bn=False, use_bn=False,
): ):
super(DPT, self).__init__() super(DPT, self).__init__()
self.channels_last = channels_last self.channels_last = channels_last

1
ldm/modules/midas/utils.py
View file

@ -17,7 +17,6 @@ def read_pfm(path):
tuple: (data, scale) tuple: (data, scale)
""" """
with open(path, "rb") as file: with open(path, "rb") as file:
color = None color = None
width = None width = None
height = None height = None

2
requirements.txt
View file

@ -16,7 +16,7 @@ gradio==3.13.2
kornia==0.6 kornia==0.6
invisible-watermark>=0.1.5 invisible-watermark>=0.1.5
streamlit-drawable-canvas==0.8.0 streamlit-drawable-canvas==0.8.0
black==21.9b0 black==23.3.0
isort==5.9.3 isort==5.9.3
flake8==4.0.1 flake8==4.0.1
click==8.0.3 click==8.0.3