Foreign-Mirrors/Fooocus
1
0
Fork 0
mirror of https://github.com/lllyasviel/Fooocus.git synced 2026-08-16 13:13:16 +02:00
Code Issues Packages Projects Releases Wiki Activity

backend

Browse Source
This commit is contained in:
lllyasviel
2024-01-27 05:12:34 -08:00
parent 0b00aaf897
commit 1cc40d24d7
44 changed files with 1885 additions and 167 deletions
Download Patch File Download Diff File Expand all files Collapse all files
ldm_patched/ldm/modules/diffusionmodules/openaimodel.py
+1 -8
View File
@@ -1,12 +1,9 @@
from abc import abstractmethod
import math
import numpy as np
import torch as th
import torch.nn as nn
import torch.nn.functional as F
from einops import rearrange
from functools import partial
from .util import (
checkpoint,
@@ -437,9 +434,6 @@ class UNetModel(nn.Module):
operations=ops,
):
super().__init__()
assert use_spatial_transformer == True, "use_spatial_transformer has to be true"
if use_spatial_transformer:
assert context_dim is not None, 'Fool!! You forgot to include the dimension of your cross-attention conditioning...'
if context_dim is not None:
assert use_spatial_transformer, 'Fool!! You forgot to use the spatial transformer for your cross-attention conditioning...'
@@ -456,7 +450,6 @@ class UNetModel(nn.Module):
if num_head_channels == -1:
assert num_heads != -1, 'Either num_heads or num_head_channels has to be set'
self.image_size = image_size
self.in_channels = in_channels
self.model_channels = model_channels
self.out_channels = out_channels
@@ -502,7 +495,7 @@ class UNetModel(nn.Module):
if self.num_classes is not None:
if isinstance(self.num_classes, int):
self.label_emb = nn.Embedding(num_classes, time_embed_dim)
self.label_emb = nn.Embedding(num_classes, time_embed_dim, dtype=self.dtype, device=device)
elif self.num_classes == "continuous":
print("setting up linear c_adm embedding layer")
self.label_emb = nn.Linear(1, time_embed_dim)
ldm_patched/ldm/modules/diffusionmodules/upscaling.py
+8 -4
View File
@@ -41,8 +41,12 @@ class AbstractLowScaleModel(nn.Module):
self.register_buffer('sqrt_recip_alphas_cumprod', to_torch(np.sqrt(1. / alphas_cumprod)))
self.register_buffer('sqrt_recipm1_alphas_cumprod', to_torch(np.sqrt(1. / alphas_cumprod - 1)))
def q_sample(self, x_start, t, noise=None):
noise = default(noise, lambda: torch.randn_like(x_start))
def q_sample(self, x_start, t, noise=None, seed=None):
if noise is None:
if seed is None:
noise = torch.randn_like(x_start)
else:
noise = torch.randn(x_start.size(), dtype=x_start.dtype, layout=x_start.layout, generator=torch.manual_seed(seed)).to(x_start.device)
return (extract_into_tensor(self.sqrt_alphas_cumprod.to(x_start.device), t, x_start.shape) * x_start +
extract_into_tensor(self.sqrt_one_minus_alphas_cumprod.to(x_start.device), t, x_start.shape) * noise)
@@ -69,12 +73,12 @@ class ImageConcatWithNoiseAugmentation(AbstractLowScaleModel):
super().__init__(noise_schedule_config=noise_schedule_config)
self.max_noise_level = max_noise_level
def forward(self, x, noise_level=None):
def forward(self, x, noise_level=None, seed=None):
if noise_level is None:
noise_level = torch.randint(0, self.max_noise_level, (x.shape[0],), device=x.device).long()
else:
assert isinstance(noise_level, torch.Tensor)
z = self.q_sample(x, noise_level)
z = self.q_sample(x, noise_level, seed=seed)
return z, noise_level