本项目的“diffusionmodules”目录包含了与扩散模型相关的核心组件和工具,包括模型定义、上采样机制、调度策略和实用函数。这些文件共同支持扩散过程的实现,如生成和处理图像的噪声调度、张量操作、网络层构建和参数管理,提供了灵活的基础架构,以便在图像生成任务中实现高效的扩散和上采样操作。
(1)文件model.py定义了一系列用于构建深度学习模型的神经网络模块,包括编码器、解码器、重缩放模块和上采样模块。主要功能是实现一个自编码器结构,用于图像生成和重建任务。通过层叠的残差块和注意力机制,这些模块能够高效地处理输入图像,提取潜在特征,并将其转换为输出图像。文件中的结构设计灵活,支持多种配置参数,以适应不同的应用需求。
import math
import torch
import torch.nn as nn
import numpy as np
from einops import rearrange
from typing import Optional, Any
from ldm.modules.attention import MemoryEfficientCrossAttention
try:
import xformers
import xformers.ops
XFORMERS_IS_AVAILBLE = True
except:
XFORMERS_IS_AVAILBLE = False
print("No module 'xformers'. Proceeding without it.")
def get_timestep_embedding(timesteps, embedding_dim):
"""
This matches the implementation in Denoising Diffusion Probabilistic Models:
From Fairseq.
Build sinusoidal embeddings.
This matches the implementation in tensor2tensor, but differs slightly
from the description in Section 3.5 of "Attention Is All You Need".
"""
assert len(timesteps.shape) == 1
half_dim = embedding_dim // 2
emb = math.log(10000) / (half_dim - 1)
emb = torch.exp(torch.arange(half_dim, dtype=torch.float32) * -emb)
emb = emb.to(device=timesteps.device)
emb = timesteps.float()[:, None] * emb[None, :]
emb = torch.cat([torch.sin(emb), torch.cos(emb)], dim=1)
if embedding_dim % 2 == 1: # zero pad
emb = torch.nn.functional.pad(emb, (0,1,0,0))
return emb
def nonlinearity(x):
# swish
return x*torch.sigmoid(x)
def Normalize(in_channels, num_groups=32):
return torch.nn.GroupNorm(num_groups=num_groups, num_channels=in_channels, eps=1e-6, affine=True)
class Upsample(nn.Module):
def __init__(self, in_channels, with_conv):
super().__init__()
self.with_conv = with_conv
if self.with_conv:
self.conv = torch.nn.Conv2d(in_channels,
in_channels,
kernel_size=3,
stride=1,
padding=1)
def forward(self, x):
x = torch.nn.functional.interpolate(x, scale_factor=2.0, mode="nearest")
if self.with_conv:
x = self.conv(x)
return x
class Downsample(nn.Module):
def __init__(self, in_channels, with_conv):
super().__init__()
self.with_conv = with_conv
if self.with_conv:
# no asymmetric padding in torch conv, must do it ourselves
self.conv = torch.nn.Conv2d(in_channels,
in_channels,
kernel_size=3,
stride=2,
padding=0)
def forward(self, x):
if self.with_conv:
pad = (0,1,0,1)
x = torch.nn.functional.pad(x, pad, mode="constant", value=0)
x = self.conv(x)
else:
x = torch.nn.functional.avg_pool2d(x, kernel_size=2, stride=2)
return x
class ResnetBlock(nn.Module):
def __init__(self, *, in_channels, out_channels=None, conv_shortcut=False,
dropout, temb_channels=512):
super().__init__()
self.in_channels = in_channels
out_channels = in_channels if out_channels is None else out_channels
self.out_channels = out_channels
self.use_conv_shortcut = conv_shortcut
self.norm1 = Normalize(in_channels)
self.conv1 = torch.nn.Conv2d(in_channels,
out_channels,
kernel_size=3,
stride=1,
padding=1)
if temb_channels > 0:
self.temb_proj = torch.nn.Linear(temb_channels,
out_channels)
self.norm2 = Normalize(out_channels)
self.dropout = torch.nn.Dropout(dropout)
self.conv2 = torch.nn.Conv2d(out_channels,
out_channels,
kernel_size=3,
stride=1,
padding=1)
if self.in_channels != self.out_channels:
if self.use_conv_shortcut:
self.conv_shortcut = torch.nn.Conv2d(in_channels,
out_channels,
kernel_size=3,
stride=1,
padding=1)
else:
self.nin_shortcut = torch.nn.Conv2d(in_channels,
out_channels,
kernel_size=1,
stride=1,
padding=0)
def forward(self, x, temb):
h = x
h = self.norm1(h)
h = nonlinearity(h)
h = self.conv1(h)
if temb is not None:
h = h + self.temb_proj(nonlinearity(temb))[:,:,None,None]
h = self.norm2(h)
h = nonlinearity(h)
h = self.dropout(h)
h = self.conv2(h)
if self.in_channels != self.out_channels:
if self.use_conv_shortcut:
x = self.conv_shortcut(x)
else:
x = self.nin_shortcut(x)
return x+h
class AttnBlock(nn.Module):
def __init__(self, in_channels):
super().__init__()
self.in_channels = in_channels
self.norm = Normalize(in_channels)
self.q = torch.nn.Conv2d(in_channels,
in_channels,
kernel_size=1,
stride=1,
padding=0)
self.k = torch.nn.Conv2d(in_channels,
in_channels,
kernel_size=1,
stride=1,
padding=0)
self.v = torch.nn.Conv2d(in_channels,
in_channels,
kernel_size=1,
stride=1,
padding=0)
self.proj_out = torch.nn.Conv2d(in_channels,
in_channels,
kernel_size=1,
stride=1,
padding=0)
def forward(self, x):
h_ = x
h_ = self.norm(h_)
q = self.q(h_)
k = self.k(h_)
v = self.v(h_)
# compute attention
b,c,h,w = q.shape
q = q.reshape(b,c,h*w)
q = q.permute(0,2,1) # b,hw,c
k = k.reshape(b,c,h*w) # b,c,hw
w_ = torch.bmm(q,k) # b,hw,hw w[b,i,j]=sum_c q[b,i,c]k[b,c,j]
w_ = w_ * (int(c)**(-0.5))
w_ = torch.nn.functional.softmax(w_, dim=2)
# attend to values
v = v.reshape(b,c,h*w)
w_ = w_.permute(0,2,1) # b,hw,hw (first hw of k, second of q)
h_ = torch.bmm(v,w_) # b, c,hw (hw of q) h_[b,c,j] = sum_i v[b,c,i] w_[b,i,j]
h_ = h_.reshape(b,c,h,w)
h_ = self.proj_out(h_)
return x+h_
class MemoryEfficientAttnBlock(nn.Module):
"""
Uses xformers efficient implementation,
see https://github.com/MatthieuTPHR/diffusers/blob/d80b531ff8060ec1ea982b65a1b8df70f73aa67c/src/diffusers/models/attention.py#L223
Note: this is a single-head self-attention operation
"""
#
def __init__(self, in_channels):
super().__init__()
self.in_channels = in_channels
self.norm = Normalize(in_channels)
self.q = torch.nn.Conv2d(in_channels,
in_channels,
kernel_size=1,
stride=1,
padding=0)
self.k = torch.nn.Conv2d(in_channels,
in_channels,
kernel_size=1,
stride=1,
padding=0)
self.v = torch.nn.Conv2d(in_channels,
in_channels,
kernel_size=1,
stride=1,
padding=0)
self.proj_out = torch.nn.Conv2d(in_channels,
in_channels,
kernel_size=1,
stride=1,
padding=0)
self.attention_op: Optional[Any] = None
def forward(self, x):
h_ = x
h_ = self.norm(h_)
q = self.q(h_)
k = self.k(h_)
v = self.v(h_)
# compute attention
B, C, H, W = q.shape
q, k, v = map(lambda x: rearrange(x, 'b c h w -> b (h w) c'), (q, k, v))
q, k, v = map(
lambda t: t.unsqueeze(3)
.reshape(B, t.shape[1], 1, C)
.permute(0, 2, 1, 3)
.reshape(B * 1, t.shape[1], C)
.contiguous(),
(q, k, v),
)
out = xformers.ops.memory_efficient_attention(q, k, v, attn_bias=None, op=self.attention_op)
out = (
out.unsqueeze(0)
.reshape(B, 1, out.shape[1], C)
.permute(0, 2, 1, 3)
.reshape(B, out.shape[1], C)
)
out = rearrange(out, 'b (h w) c -> b c h w', b=B, h=H, w=W, c=C)
out = self.proj_out(out)
return x+out
class MemoryEfficientCrossAttentionWrapper(MemoryEfficientCrossAttention):
def forward(self, x, context=None, mask=None):
b, c, h, w = x.shape
x = rearrange(x, 'b c h w -> b (h w) c')
out = super().forward(x, context=context, mask=mask)
out = rearrange(out, 'b (h w) c -> b c h w', h=h, w=w, c=c)
return x + out
def make_attn(in_channels, attn_type="vanilla", attn_kwargs=None):
assert attn_type in ["vanilla", "vanilla-xformers", "memory-efficient-cross-attn", "linear", "none"], f'attn_type {attn_type} unknown'
if XFORMERS_IS_AVAILBLE and attn_type == "vanilla":
attn_type = "vanilla-xformers"
print(f"making attention of type '{attn_type}' with {in_channels} in_channels")
if attn_type == "vanilla":
assert attn_kwargs is None
return AttnBlock(in_channels)
elif attn_type == "vanilla-xformers":
print(f"building MemoryEfficientAttnBlock with {in_channels} in_channels...")
return MemoryEfficientAttnBlock(in_channels)
elif type == "memory-efficient-cross-attn":
attn_kwargs["query_dim"] = in_channels
return MemoryEfficientCrossAttentionWrapper(**attn_kwargs)
elif attn_type == "none":
return nn.Identity(in_channels)
else:
raise NotImplementedError()
class Model(nn.Module):
def __init__(self, *, ch, out_ch, ch_mult=(1,2,4,8), num_res_blocks,
attn_resolutions, dropout=0.0, resamp_with_conv=True, in_channels,
resolution, use_timestep=True, use_linear_attn=False, attn_type="vanilla"):
super().__init__()
if use_linear_attn: attn_type = "linear"
self.ch = ch
self.temb_ch = self.ch*4
self.num_resolutions = len(ch_mult)
self.num_res_blocks = num_res_blocks
self.resolution = resolution
self.in_channels = in_channels
self.use_timestep = use_timestep
if self.use_timestep:
# timestep embedding
self.temb = nn.Module()
self.temb.dense = nn.ModuleList([
torch.nn.Linear(self.ch,
self.temb_ch),
torch.nn.Linear(self.temb_ch,
self.temb_ch),
])
# downsampling
self.conv_in = torch.nn.Conv2d(in_channels,
self.ch,
kernel_size=3,
stride=1,
padding=1)
curr_res = resolution
in_ch_mult = (1,)+tuple(ch_mult)
self.down = nn.ModuleList()
for i_level in range(self.num_resolutions):
block = nn.ModuleList()
attn = nn.ModuleList()
block_in = ch*in_ch_mult[i_level]
block_out = ch*ch_mult[i_level]
for i_block in range(self.num_res_blocks):
block.append(ResnetBlock(in_channels=block_in,
out_channels=block_out,
temb_channels=self.temb_ch,
dropout=dropout))
block_in = block_out
if curr_res in attn_resolutions:
attn.append(make_attn(block_in, attn_type=attn_type))
down = nn.Module()
down.block = block
down.attn = attn
if i_level != self.num_resolutions-1:
down.downsample = Downsample(block_in, resamp_with_conv)
curr_res = curr_res // 2
self.down.append(down)
# middle
self.mid = nn.Module()
self.mid.block_1 = ResnetBlock(in_channels=block_in,
out_channels=block_in,
temb_channels=self.temb_ch,
dropout=dropout)
self.mid.attn_1 = make_attn(block_in, attn_type=attn_type)
self.mid.block_2 = ResnetBlock(in_channels=block_in,
out_channels=block_in,
temb_channels=self.temb_ch,
dropout=dropout)
# upsampling
self.up = nn.ModuleList()
for i_level in reversed(range(self.num_resolutions)):
block = nn.ModuleList()
attn = nn.ModuleList()
block_out = ch*ch_mult[i_level]
skip_in = ch*ch_mult[i_level]
for i_block in range(self.num_res_blocks+1):
if i_block == self.num_res_blocks:
skip_in = ch*in_ch_mult[i_level]
block.append(ResnetBlock(in_channels=block_in+skip_in,
out_channels=block_out,
temb_channels=self.temb_ch,
dropout=dropout))
block_in = block_out
if curr_res in attn_resolutions:
attn.append(make_attn(block_in, attn_type=attn_type))
up = nn.Module()
up.block = block
up.attn = attn
if i_level != 0:
up.upsample = Upsample(block_in, resamp_with_conv)
curr_res = curr_res * 2
self.up.insert(0, up) # prepend to get consistent order
# end
self.norm_out = Normalize(block_in)
self.conv_out = torch.nn.Conv2d(block_in,
out_ch,
kernel_size=3,
stride=1,
padding=1)
def forward(self, x, t=None, context=None):
#assert x.shape[2] == x.shape[3] == self.resolution
if context is not None:
# assume aligned context, cat along channel axis
x = torch.cat((x, context), dim=1)
if self.use_timestep:
# timestep embedding
assert t is not None
temb = get_timestep_embedding(t, self.ch)
temb = self.temb.dense[0](temb)
temb = nonlinearity(temb)
temb = self.temb.dense[1](temb)
else:
temb = None
# downsampling
hs = [self.conv_in(x)]
for i_level in range(self.num_resolutions):
for i_block in range(self.num_res_blocks):
h = self.down[i_level].block[i_block](hs[-1], temb)
if len(self.down[i_level].attn) > 0:
h = self.down[i_level].attn[i_block](h)
hs.append(h)
if i_level != self.num_resolutions-1:
hs.append(self.down[i_level].downsample(hs[-1]))
# middle
h = hs[-1]
h = self.mid.block_1(h, temb)
h = self.mid.attn_1(h)
h = self.mid.block_2(h, temb)
# upsampling
for i_level in reversed(range(self.num_resolutions)):
for i_block in range(self.num_res_blocks+1):
h = self.up[i_level].block[i_block](
torch.cat([h, hs.pop()], dim=1), temb)
if len(self.up[i_level].attn) > 0:
h = self.up[i_level].attn[i_block](h)
if i_level != 0:
h = self.up[i_level].upsample(h)
# end
h = self.norm_out(h)
h = nonlinearity(h)
h = self.conv_out(h)
return h
def get_last_layer(self):
return self.conv_out.weight
class Encoder(nn.Module):
def __init__(self, *, ch, out_ch, ch_mult=(1,2,4,8), num_res_blocks,
attn_resolutions, dropout=0.0, resamp_with_conv=True, in_channels,
resolution, z_channels, double_z=True, use_linear_attn=False, attn_type="vanilla",
**ignore_kwargs):
super().__init__()
if use_linear_attn: attn_type = "linear"
self.ch = ch
self.temb_ch = 0
self.num_resolutions = len(ch_mult)
self.num_res_blocks = num_res_blocks
self.resolution = resolution
self.in_channels = in_channels
# downsampling
self.conv_in = torch.nn.Conv2d(in_channels,
self.ch,
kernel_size=3,
stride=1,
padding=1)
curr_res = resolution
in_ch_mult = (1,)+tuple(ch_mult)
self.in_ch_mult = in_ch_mult
self.down = nn.ModuleList()
for i_level in range(self.num_resolutions):
block = nn.ModuleList()
attn = nn.ModuleList()
block_in = ch*in_ch_mult[i_level]
block_out = ch*ch_mult[i_level]
for i_block in range(self.num_res_blocks):
block.append(ResnetBlock(in_channels=block_in,
out_channels=block_out,
temb_channels=self.temb_ch,
dropout=dropout))
block_in = block_out
if curr_res in attn_resolutions:
attn.append(make_attn(block_in, attn_type=attn_type))
down = nn.Module()
down.block = block
down.attn = attn
if i_level != self.num_resolutions-1:
down.downsample = Downsample(block_in, resamp_with_conv)
curr_res = curr_res // 2
self.down.append(down)
# middle
self.mid = nn.Module()
self.mid.block_1 = ResnetBlock(in_channels=block_in,
out_channels=block_in,
temb_channels=self.temb_ch,
dropout=dropout)
self.mid.attn_1 = make_attn(block_in, attn_type=attn_type)
self.mid.block_2 = ResnetBlock(in_channels=block_in,
out_channels=block_in,
temb_channels=self.temb_ch,
dropout=dropout)
# end
self.norm_out = Normalize(block_in)
self.conv_out = torch.nn.Conv2d(block_in,
2*z_channels if double_z else z_channels,
kernel_size=3,
stride=1,
padding=1)
def forward(self, x):
# timestep embedding
temb = None
# downsampling
hs = [self.conv_in(x)]
for i_level in range(self.num_resolutions):
for i_block in range(self.num_res_blocks):
h = self.down[i_level].block[i_block](hs[-1], temb)
if len(self.down[i_level].attn) > 0:
h = self.down[i_level].attn[i_block](h)
hs.append(h)
if i_level != self.num_resolutions-1:
hs.append(self.down[i_level].downsample(hs[-1]))
# middle
h = hs[-1]
h = self.mid.block_1(h, temb)
h = self.mid.attn_1(h)
h = self.mid.block_2(h, temb)
# end
h = self.norm_out(h)
h = nonlinearity(h)
h = self.conv_out(h)
return h
class Decoder(nn.Module):
def __init__(self, *, ch, out_ch, ch_mult=(1,2,4,8), num_res_blocks,
attn_resolutions, dropout=0.0, resamp_with_conv=True, in_channels,
resolution, z_channels, give_pre_end=False, tanh_out=False, use_linear_attn=False,
attn_type="vanilla", **ignorekwargs):
super().__init__()
if use_linear_attn: attn_type = "linear"
self.ch = ch
self.temb_ch = 0
self.num_resolutions = len(ch_mult)
self.num_res_blocks = num_res_blocks
self.resolution = resolution
self.in_channels = in_channels
self.give_pre_end = give_pre_end
self.tanh_out = tanh_out
# compute in_ch_mult, block_in and curr_res at lowest res
in_ch_mult = (1,)+tuple(ch_mult)
block_in = ch*ch_mult[self.num_resolutions-1]
curr_res = resolution // 2**(self.num_resolutions-1)
self.z_shape = (1,z_channels,curr_res,curr_res)
print("Working with z of shape {} = {} dimensions.".format(
self.z_shape, np.prod(self.z_shape)))
# z to block_in
self.conv_in = torch.nn.Conv2d(z_channels,
block_in,
kernel_size=3,
stride=1,
padding=1)
# middle
self.mid = nn.Module()
self.mid.block_1 = ResnetBlock(in_channels=block_in,
out_channels=block_in,
temb_channels=self.temb_ch,
dropout=dropout)
self.mid.attn_1 = make_attn(block_in, attn_type=attn_type)
self.mid.block_2 = ResnetBlock(in_channels=block_in,
out_channels=block_in,
temb_channels=self.temb_ch,
dropout=dropout)
# upsampling
self.up = nn.ModuleList()
for i_level in reversed(range(self.num_resolutions)):
block = nn.ModuleList()
attn = nn.ModuleList()
block_out = ch*ch_mult[i_level]
for i_block in range(self.num_res_blocks+1):
block.append(ResnetBlock(in_channels=block_in,
out_channels=block_out,
temb_channels=self.temb_ch,
dropout=dropout))
block_in = block_out
if curr_res in attn_resolutions:
attn.append(make_attn(block_in, attn_type=attn_type))
up = nn.Module()
up.block = block
up.attn = attn
if i_level != 0:
up.upsample = Upsample(block_in, resamp_with_conv)
curr_res = curr_res * 2
self.up.insert(0, up) # prepend to get consistent order
# end
self.norm_out = Normalize(block_in)
self.conv_out = torch.nn.Conv2d(block_in,
out_ch,
kernel_size=3,
stride=1,
padding=1)
def forward(self, z):
#assert z.shape[1:] == self.z_shape[1:]
self.last_z_shape = z.shape
# timestep embedding
temb = None
# z to block_in
h = self.conv_in(z)
# middle
h = self.mid.block_1(h, temb)
h = self.mid.attn_1(h)
h = self.mid.block_2(h, temb)
# upsampling
for i_level in reversed(range(self.num_resolutions)):
for i_block in range(self.num_res_blocks+1):
h = self.up[i_level].block[i_block](h, temb)
if len(self.up[i_level].attn) > 0:
h = self.up[i_level].attn[i_block](h)
if i_level != 0:
h = self.up[i_level].upsample(h)
# end
if self.give_pre_end:
return h
h = self.norm_out(h)
h = nonlinearity(h)
h = self.conv_out(h)
if self.tanh_out:
h = torch.tanh(h)
return h
class SimpleDecoder(nn.Module):
def __init__(self, in_channels, out_channels, *args, **kwargs):
super().__init__()
self.model = nn.ModuleList([nn.Conv2d(in_channels, in_channels, 1),
ResnetBlock(in_channels=in_channels,
out_channels=2 * in_channels,
temb_channels=0, dropout=0.0),
ResnetBlock(in_channels=2 * in_channels,
out_channels=4 * in_channels,
temb_channels=0, dropout=0.0),
ResnetBlock(in_channels=4 * in_channels,
out_channels=2 * in_channels,
temb_channels=0, dropout=0.0),
nn.Conv2d(2*in_channels, in_channels, 1),
Upsample(in_channels, with_conv=True)])
# end
self.norm_out = Normalize(in_channels)
self.conv_out = torch.nn.Conv2d(in_channels,
out_channels,
kernel_size=3,
stride=1,
padding=1)
def forward(self, x):
for i, layer in enumerate(self.model):
if i in [1,2,3]:
x = layer(x, None)
else:
x = layer(x)
h = self.norm_out(x)
h = nonlinearity(h)
x = self.conv_out(h)
return x
class UpsampleDecoder(nn.Module):
def __init__(self, in_channels, out_channels, ch, num_res_blocks, resolution,
ch_mult=(2,2), dropout=0.0):
super().__init__()
# upsampling
self.temb_ch = 0
self.num_resolutions = len(ch_mult)
self.num_res_blocks = num_res_blocks
block_in = in_channels
curr_res = resolution // 2 ** (self.num_resolutions - 1)
self.res_blocks = nn.ModuleList()
self.upsample_blocks = nn.ModuleList()
for i_level in range(self.num_resolutions):
res_block = []
block_out = ch * ch_mult[i_level]
for i_block in range(self.num_res_blocks + 1):
res_block.append(ResnetBlock(in_channels=block_in,
out_channels=block_out,
temb_channels=self.temb_ch,
dropout=dropout))
block_in = block_out
self.res_blocks.append(nn.ModuleList(res_block))
if i_level != self.num_resolutions - 1:
self.upsample_blocks.append(Upsample(block_in, True))
curr_res = curr_res * 2
# end
self.norm_out = Normalize(block_in)
self.conv_out = torch.nn.Conv2d(block_in,
out_channels,
kernel_size=3,
stride=1,
padding=1)
def forward(self, x):
# upsampling
h = x
for k, i_level in enumerate(range(self.num_resolutions)):
for i_block in range(self.num_res_blocks + 1):
h = self.res_blocks[i_level][i_block](h, None)
if i_level != self.num_resolutions - 1:
h = self.upsample_blocks[k](h)
h = self.norm_out(h)
h = nonlinearity(h)
h = self.conv_out(h)
return h
class LatentRescaler(nn.Module):
def __init__(self, factor, in_channels, mid_channels, out_channels, depth=2):
super().__init__()
# residual block, interpolate, residual block
self.factor = factor
self.conv_in = nn.Conv2d(in_channels,
mid_channels,
kernel_size=3,
stride=1,
padding=1)
self.res_block1 = nn.ModuleList([ResnetBlock(in_channels=mid_channels,
out_channels=mid_channels,
temb_channels=0,
dropout=0.0) for _ in range(depth)])
self.attn = AttnBlock(mid_channels)
self.res_block2 = nn.ModuleList([ResnetBlock(in_channels=mid_channels,
out_channels=mid_channels,
temb_channels=0,
dropout=0.0) for _ in range(depth)])
self.conv_out = nn.Conv2d(mid_channels,
out_channels,
kernel_size=1,
)
def forward(self, x):
x = self.conv_in(x)
for block in self.res_block1:
x = block(x, None)
x = torch.nn.functional.interpolate(x, size=(int(round(x.shape[2]*self.factor)), int(round(x.shape[3]*self.factor))))
x = self.attn(x)
for block in self.res_block2:
x = block(x, None)
x = self.conv_out(x)
return x
class MergedRescaleEncoder(nn.Module):
def __init__(self, in_channels, ch, resolution, out_ch, num_res_blocks,
attn_resolutions, dropout=0.0, resamp_with_conv=True,
ch_mult=(1,2,4,8), rescale_factor=1.0, rescale_module_depth=1):
super().__init__()
intermediate_chn = ch * ch_mult[-1]
self.encoder = Encoder(in_channels=in_channels, num_res_blocks=num_res_blocks, ch=ch, ch_mult=ch_mult,
z_channels=intermediate_chn, double_z=False, resolution=resolution,
attn_resolutions=attn_resolutions, dropout=dropout, resamp_with_conv=resamp_with_conv,
out_ch=None)
self.rescaler = LatentRescaler(factor=rescale_factor, in_channels=intermediate_chn,
mid_channels=intermediate_chn, out_channels=out_ch, depth=rescale_module_depth)
def forward(self, x):
x = self.encoder(x)
x = self.rescaler(x)
return x
class MergedRescaleDecoder(nn.Module):
def __init__(self, z_channels, out_ch, resolution, num_res_blocks, attn_resolutions, ch, ch_mult=(1,2,4,8),
dropout=0.0, resamp_with_conv=True, rescale_factor=1.0, rescale_module_depth=1):
super().__init__()
tmp_chn = z_channels*ch_mult[-1]
self.decoder = Decoder(out_ch=out_ch, z_channels=tmp_chn, attn_resolutions=attn_resolutions, dropout=dropout,
resamp_with_conv=resamp_with_conv, in_channels=None, num_res_blocks=num_res_blocks,
ch_mult=ch_mult, resolution=resolution, ch=ch)
self.rescaler = LatentRescaler(factor=rescale_factor, in_channels=z_channels, mid_channels=tmp_chn,
out_channels=tmp_chn, depth=rescale_module_depth)
def forward(self, x):
x = self.rescaler(x)
x = self.decoder(x)
return x
class Upsampler(nn.Module):
def __init__(self, in_size, out_size, in_channels, out_channels, ch_mult=2):
super().__init__()
assert out_size >= in_size
num_blocks = int(np.log2(out_size//in_size))+1
factor_up = 1.+ (out_size % in_size)
print(f"Building {self.__class__.__name__} with in_size: {in_size} --> out_size {out_size} and factor {factor_up}")
self.rescaler = LatentRescaler(factor=factor_up, in_channels=in_channels, mid_channels=2*in_channels,
out_channels=in_channels)
self.decoder = Decoder(out_ch=out_channels, resolution=out_size, z_channels=in_channels, num_res_blocks=2,
attn_resolutions=[], in_channels=None, ch=in_channels,
ch_mult=[ch_mult for _ in range(num_blocks)])
def forward(self, x):
x = self.rescaler(x)
x = self.decoder(x)
return x
class Resize(nn.Module):
def __init__(self, in_channels=None, learned=False, mode="bilinear"):
super().__init__()
self.with_conv = learned
self.mode = mode
if self.with_conv:
print(f"Note: {self.__class__.__name} uses learned downsampling and will ignore the fixed {mode} mode")
raise NotImplementedError()
assert in_channels is not None
# no asymmetric padding in torch conv, must do it ourselves
self.conv = torch.nn.Conv2d(in_channels,
in_channels,
kernel_size=4,
stride=2,
padding=1)
def forward(self, x, scale_factor=1.0):
if scale_factor==1.0:
return x
else:
x = torch.nn.functional.interpolate(x, mode=self.mode, align_corners=False, scale_factor=scale_factor)
return x
(2)文件openaimodel.py定义了一个完整的 UNet 模型,结合了注意力机制和时间步嵌入。该模型通过多个输入块、残差块和输出块构建,允许灵活的通道设置、下采样和上采样,支持条件生成(例如类别条件生成)和自注意力机制,以增强模型对输入特征的表示能力。模型还包括在前向传播过程中处理输入数据的逻辑,能够根据需要转换数据类型以优化性能。
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 ldm.modules.diffusionmodules.util import (
checkpoint,
conv_nd,
linear,
avg_pool_nd,
zero_module,
normalization,
timestep_embedding,
)
from ldm.modules.attention import SpatialTransformer
from ldm.util import exists
# dummy replace
def convert_module_to_f16(x):
pass
def convert_module_to_f32(x):
pass
## go
class AttentionPool2d(nn.Module):
"""
Adapted from CLIP: https://github.com/openai/CLIP/blob/main/clip/model.py
"""
def __init__(
self,
spacial_dim: int,
embed_dim: int,
num_heads_channels: int,
output_dim: int = None,
):
super().__init__()
self.positional_embedding = nn.Parameter(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.c_proj = conv_nd(1, embed_dim, output_dim or embed_dim, 1)
self.num_heads = embed_dim // num_heads_channels
self.attention = QKVAttention(self.num_heads)
def forward(self, x):
b, c, *_spatial = x.shape
x = x.reshape(b, c, -1) # NC(HW)
x = th.cat([x.mean(dim=-1, keepdim=True), x], dim=-1) # NC(HW+1)
x = x + self.positional_embedding[None, :, :].to(x.dtype) # NC(HW+1)
x = self.qkv_proj(x)
x = self.attention(x)
x = self.c_proj(x)
return x[:, :, 0]
class TimestepBlock(nn.Module):
"""
Any module where forward() takes timestep embeddings as a second argument.
"""
@abstractmethod
def forward(self, x, emb):
"""
Apply the module to `x` given `emb` timestep embeddings.
"""
class TimestepEmbedSequential(nn.Sequential, TimestepBlock):
"""
A sequential module that passes timestep embeddings to the children that
support it as an extra input.
"""
def forward(self, x, emb, context=None):
for layer in self:
if isinstance(layer, TimestepBlock):
x = layer(x, emb)
elif isinstance(layer, SpatialTransformer):
x = layer(x, context)
else:
x = layer(x)
return x
class Upsample(nn.Module):
"""
An upsampling layer with an optional convolution.
:param channels: channels in the inputs and outputs.
:param use_conv: a bool determining if a convolution is applied.
:param dims: determines if the signal is 1D, 2D, or 3D. If 3D, then
upsampling occurs in the inner-two dimensions.
"""
def __init__(self, channels, use_conv, dims=2, out_channels=None, padding=1):
super().__init__()
self.channels = channels
self.out_channels = out_channels or channels
self.use_conv = use_conv
self.dims = dims
if use_conv:
self.conv = conv_nd(dims, self.channels, self.out_channels, 3, padding=padding)
def forward(self, x):
assert x.shape[1] == self.channels
if self.dims == 3:
x = F.interpolate(
x, (x.shape[2], x.shape[3] * 2, x.shape[4] * 2), mode="nearest"
)
else:
x = F.interpolate(x, scale_factor=2, mode="nearest")
if self.use_conv:
x = self.conv(x)
return x
class TransposedUpsample(nn.Module):
'Learned 2x upsampling without padding'
def __init__(self, channels, out_channels=None, ks=5):
super().__init__()
self.channels = channels
self.out_channels = out_channels or channels
self.up = nn.ConvTranspose2d(self.channels,self.out_channels,kernel_size=ks,stride=2)
def forward(self,x):
return self.up(x)
class Downsample(nn.Module):
"""
A downsampling layer with an optional convolution.
:param channels: channels in the inputs and outputs.
:param use_conv: a bool determining if a convolution is applied.
:param dims: determines if the signal is 1D, 2D, or 3D. If 3D, then
downsampling occurs in the inner-two dimensions.
"""
def __init__(self, channels, use_conv, dims=2, out_channels=None,padding=1):
super().__init__()
self.channels = channels
self.out_channels = out_channels or channels
self.use_conv = use_conv
self.dims = dims
stride = 2 if dims != 3 else (1, 2, 2)
if use_conv:
self.op = conv_nd(
dims, self.channels, self.out_channels, 3, stride=stride, padding=padding
)
else:
assert self.channels == self.out_channels
self.op = avg_pool_nd(dims, kernel_size=stride, stride=stride)
def forward(self, x):
assert x.shape[1] == self.channels
return self.op(x)
class ResBlock(TimestepBlock):
"""
A residual block that can optionally change the number of channels.
:param channels: the number of input channels.
:param emb_channels: the number of timestep embedding channels.
:param dropout: the rate of dropout.
:param out_channels: if specified, the number of out channels.
:param use_conv: if True and out_channels is specified, use a spatial
convolution instead of a smaller 1x1 convolution to change the
channels in the skip connection.
:param dims: determines if the signal is 1D, 2D, or 3D.
:param use_checkpoint: if True, use gradient checkpointing on this module.
:param up: if True, use this block for upsampling.
:param down: if True, use this block for downsampling.
"""
def __init__(
self,
channels,
emb_channels,
dropout,
out_channels=None,
use_conv=False,
use_scale_shift_norm=False,
dims=2,
use_checkpoint=False,
up=False,
down=False,
):
super().__init__()
self.channels = channels
self.emb_channels = emb_channels
self.dropout = dropout
self.out_channels = out_channels or channels
self.use_conv = use_conv
self.use_checkpoint = use_checkpoint
self.use_scale_shift_norm = use_scale_shift_norm
self.in_layers = nn.Sequential(
normalization(channels),
nn.SiLU(),
conv_nd(dims, channels, self.out_channels, 3, padding=1),
)
self.updown = up or down
if up:
self.h_upd = Upsample(channels, False, dims)
self.x_upd = Upsample(channels, False, dims)
elif down:
self.h_upd = Downsample(channels, False, dims)
self.x_upd = Downsample(channels, False, dims)
else:
self.h_upd = self.x_upd = nn.Identity()
self.emb_layers = nn.Sequential(
nn.SiLU(),
linear(
emb_channels,
2 * self.out_channels if use_scale_shift_norm else self.out_channels,
),
)
self.out_layers = nn.Sequential(
normalization(self.out_channels),
nn.SiLU(),
nn.Dropout(p=dropout),
zero_module(
conv_nd(dims, self.out_channels, self.out_channels, 3, padding=1)
),
)
if self.out_channels == channels:
self.skip_connection = nn.Identity()
elif use_conv:
self.skip_connection = conv_nd(
dims, channels, self.out_channels, 3, padding=1
)
else:
self.skip_connection = conv_nd(dims, channels, self.out_channels, 1)
def forward(self, x, emb):
"""
Apply the block to a Tensor, conditioned on a timestep embedding.
:param x: an [N x C x ...] Tensor of features.
:param emb: an [N x emb_channels] Tensor of timestep embeddings.
:return: an [N x C x ...] Tensor of outputs.
"""
return checkpoint(
self._forward, (x, emb), self.parameters(), self.use_checkpoint
)
def _forward(self, x, emb):
if self.updown:
in_rest, in_conv = self.in_layers[:-1], self.in_layers[-1]
h = in_rest(x)
h = self.h_upd(h)
x = self.x_upd(x)
h = in_conv(h)
else:
h = self.in_layers(x)
emb_out = self.emb_layers(emb).type(h.dtype)
while len(emb_out.shape) < len(h.shape):
emb_out = emb_out[..., None]
if self.use_scale_shift_norm:
out_norm, out_rest = self.out_layers[0], self.out_layers[1:]
scale, shift = th.chunk(emb_out, 2, dim=1)
h = out_norm(h) * (1 + scale) + shift
h = out_rest(h)
else:
h = h + emb_out
h = self.out_layers(h)
return self.skip_connection(x) + h
class AttentionBlock(nn.Module):
"""
An attention block that allows spatial positions to attend to each other.
Originally ported from here, but adapted to the N-d case.
https://github.com/hojonathanho/diffusion/blob/1e0dceb3b3495bbe19116a5e1b3596cd0706c543/diffusion_tf/models/unet.py#L66.
"""
def __init__(
self,
channels,
num_heads=1,
num_head_channels=-1,
use_checkpoint=False,
use_new_attention_order=False,
):
super().__init__()
self.channels = channels
if num_head_channels == -1:
self.num_heads = num_heads
else:
assert (
channels % num_head_channels == 0
), f"q,k,v channels {channels} is not divisible by num_head_channels {num_head_channels}"
self.num_heads = channels // num_head_channels
self.use_checkpoint = use_checkpoint
self.norm = normalization(channels)
self.qkv = conv_nd(1, channels, channels * 3, 1)
if use_new_attention_order:
# split qkv before split heads
self.attention = QKVAttention(self.num_heads)
else:
# split heads before split qkv
self.attention = QKVAttentionLegacy(self.num_heads)
self.proj_out = zero_module(conv_nd(1, channels, channels, 1))
def forward(self, x):
return checkpoint(self._forward, (x,), self.parameters(), True) # TODO: check checkpoint usage, is True # TODO: fix the .half call!!!
#return pt_checkpoint(self._forward, x) # pytorch
def _forward(self, x):
b, c, *spatial = x.shape
x = x.reshape(b, c, -1)
qkv = self.qkv(self.norm(x))
h = self.attention(qkv)
h = self.proj_out(h)
return (x + h).reshape(b, c, *spatial)
def count_flops_attn(model, _x, y):
"""
A counter for the `thop` package to count the operations in an
attention operation.
Meant to be used like:
macs, params = thop.profile(
model,
inputs=(inputs, timestamps),
custom_ops={QKVAttention: QKVAttention.count_flops},
)
"""
b, c, *spatial = y[0].shape
num_spatial = int(np.prod(spatial))
# We perform two matmuls with the same number of ops.
# The first computes the weight matrix, the second computes
# the combination of the value vectors.
matmul_ops = 2 * b * (num_spatial ** 2) * c
model.total_ops += th.DoubleTensor([matmul_ops])
class QKVAttentionLegacy(nn.Module):
"""
A module which performs QKV attention. Matches legacy QKVAttention + input/ouput heads shaping
"""
def __init__(self, n_heads):
super().__init__()
self.n_heads = n_heads
def forward(self, qkv):
"""
Apply QKV attention.
:param qkv: an [N x (H * 3 * C) x T] tensor of Qs, Ks, and Vs.
:return: an [N x (H * C) x T] tensor after attention.
"""
bs, width, length = qkv.shape
assert width % (3 * self.n_heads) == 0
ch = width // (3 * self.n_heads)
q, k, v = qkv.reshape(bs * self.n_heads, ch * 3, length).split(ch, dim=1)
scale = 1 / math.sqrt(math.sqrt(ch))
weight = th.einsum(
"bct,bcs->bts", q * scale, k * scale
) # More stable with f16 than dividing afterwards
weight = th.softmax(weight.float(), dim=-1).type(weight.dtype)
a = th.einsum("bts,bcs->bct", weight, v)
return a.reshape(bs, -1, length)
@staticmethod
def count_flops(model, _x, y):
return count_flops_attn(model, _x, y)
class QKVAttention(nn.Module):
"""
A module which performs QKV attention and splits in a different order.
"""
def __init__(self, n_heads):
super().__init__()
self.n_heads = n_heads
def forward(self, qkv):
"""
Apply QKV attention.
:param qkv: an [N x (3 * H * C) x T] tensor of Qs, Ks, and Vs.
:return: an [N x (H * C) x T] tensor after attention.
"""
bs, width, length = qkv.shape
assert width % (3 * self.n_heads) == 0
ch = width // (3 * self.n_heads)
q, k, v = qkv.chunk(3, dim=1)
scale = 1 / math.sqrt(math.sqrt(ch))
weight = th.einsum(
"bct,bcs->bts",
(q * scale).view(bs * self.n_heads, ch, length),
(k * scale).view(bs * self.n_heads, ch, length),
) # More stable with f16 than dividing afterwards
weight = th.softmax(weight.float(), dim=-1).type(weight.dtype)
a = th.einsum("bts,bcs->bct", weight, v.reshape(bs * self.n_heads, ch, length))
return a.reshape(bs, -1, length)
@staticmethod
def count_flops(model, _x, y):
return count_flops_attn(model, _x, y)
class Timestep(nn.Module):
def __init__(self, dim):
super().__init__()
self.dim = dim
def forward(self, t):
return timestep_embedding(t, self.dim)
class UNetModel(nn.Module):
"""
The full UNet model with attention and timestep embedding.
:param in_channels: channels in the input Tensor.
:param model_channels: base channel count for the model.
:param out_channels: channels in the output Tensor.
:param num_res_blocks: number of residual blocks per downsample.
:param attention_resolutions: a collection of downsample rates at which
attention will take place. May be a set, list, or tuple.
For example, if this contains 4, then at 4x downsampling, attention
will be used.
:param dropout: the dropout probability.
:param channel_mult: channel multiplier for each level of the UNet.
:param conv_resample: if True, use learned convolutions for upsampling and
downsampling.
:param dims: determines if the signal is 1D, 2D, or 3D.
:param num_classes: if specified (as an int), then this model will be
class-conditional with `num_classes` classes.
:param use_checkpoint: use gradient checkpointing to reduce memory usage.
:param num_heads: the number of attention heads in each attention layer.
:param num_heads_channels: if specified, ignore num_heads and instead use
a fixed channel width per attention head.
:param num_heads_upsample: works with num_heads to set a different number
of heads for upsampling. Deprecated.
:param use_scale_shift_norm: use a FiLM-like conditioning mechanism.
:param resblock_updown: use residual blocks for up/downsampling.
:param use_new_attention_order: use a different attention pattern for potentially
increased efficiency.
"""
def __init__(
self,
image_size,
in_channels,
model_channels,
out_channels,
num_res_blocks,
attention_resolutions,
dropout=0,
channel_mult=(1, 2, 4, 8),
conv_resample=True,
dims=2,
num_classes=None,
use_checkpoint=False,
use_fp16=False,
use_bf16=False,
num_heads=-1,
num_head_channels=-1,
num_heads_upsample=-1,
use_scale_shift_norm=False,
resblock_updown=False,
use_new_attention_order=False,
use_spatial_transformer=False, # custom transformer support
transformer_depth=1, # custom transformer support
context_dim=None, # custom transformer support
n_embed=None, # custom support for prediction of discrete ids into codebook of first stage vq model
legacy=True,
disable_self_attentions=None,
num_attention_blocks=None,
disable_middle_self_attn=False,
use_linear_in_transformer=False,
adm_in_channels=None,
):
super().__init__()
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...'
from omegaconf.listconfig import ListConfig
if type(context_dim) == ListConfig:
context_dim = list(context_dim)
if num_heads_upsample == -1:
num_heads_upsample = num_heads
if num_heads == -1:
assert num_head_channels != -1, 'Either num_heads or num_head_channels has to be set'
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
if isinstance(num_res_blocks, int):
self.num_res_blocks = len(channel_mult) * [num_res_blocks]
else:
if len(num_res_blocks) != len(channel_mult):
raise ValueError("provide num_res_blocks either as an int (globally constant) or "
"as a list/tuple (per-level) with the same length as channel_mult")
self.num_res_blocks = num_res_blocks
if disable_self_attentions is not None:
# should be a list of booleans, indicating whether to disable self-attention in TransformerBlocks or not
assert len(disable_self_attentions) == len(channel_mult)
if num_attention_blocks is not None:
assert len(num_attention_blocks) == len(self.num_res_blocks)
assert all(map(lambda i: self.num_res_blocks[i] >= num_attention_blocks[i], range(len(num_attention_blocks))))
print(f"Constructor of UNetModel received num_attention_blocks={num_attention_blocks}. "
f"This option has LESS priority than attention_resolutions {attention_resolutions}, "
f"i.e., in cases where num_attention_blocks[i] > 0 but 2**i not in attention_resolutions, "
f"attention will still not be set.")
self.attention_resolutions = attention_resolutions
self.dropout = dropout
self.channel_mult = channel_mult
self.conv_resample = conv_resample
self.num_classes = num_classes
self.use_checkpoint = use_checkpoint
self.dtype = th.float16 if use_fp16 else th.float32
self.dtype = th.bfloat16 if use_bf16 else self.dtype
self.num_heads = num_heads
self.num_head_channels = num_head_channels
self.num_heads_upsample = num_heads_upsample
self.predict_codebook_ids = n_embed is not None
time_embed_dim = model_channels * 4
self.time_embed = nn.Sequential(
linear(model_channels, time_embed_dim),
nn.SiLU(),
linear(time_embed_dim, time_embed_dim),
)
if self.num_classes is not None:
if isinstance(self.num_classes, int):
self.label_emb = nn.Embedding(num_classes, time_embed_dim)
elif self.num_classes == "continuous":
print("setting up linear c_adm embedding layer")
self.label_emb = nn.Linear(1, time_embed_dim)
elif self.num_classes == "sequential":
assert adm_in_channels is not None
self.label_emb = nn.Sequential(
nn.Sequential(
linear(adm_in_channels, time_embed_dim),
nn.SiLU(),
linear(time_embed_dim, time_embed_dim),
)
)
else:
raise ValueError()
self.input_blocks = nn.ModuleList(
[
TimestepEmbedSequential(
conv_nd(dims, in_channels, model_channels, 3, padding=1)
)
]
)
self._feature_size = model_channels
input_block_chans = [model_channels]
ch = model_channels
ds = 1
for level, mult in enumerate(channel_mult):
for nr in range(self.num_res_blocks[level]):
layers = [
ResBlock(
ch,
time_embed_dim,
dropout,
out_channels=mult * model_channels,
dims=dims,
use_checkpoint=use_checkpoint,
use_scale_shift_norm=use_scale_shift_norm,
)
]
ch = mult * model_channels
if ds in attention_resolutions:
if num_head_channels == -1:
dim_head = ch // num_heads
else:
num_heads = ch // num_head_channels
dim_head = num_head_channels
if legacy:
#num_heads = 1
dim_head = ch // num_heads if use_spatial_transformer else num_head_channels
if exists(disable_self_attentions):
disabled_sa = disable_self_attentions[level]
else:
disabled_sa = False
if not exists(num_attention_blocks) or nr < num_attention_blocks[level]:
layers.append(
AttentionBlock(
ch,
use_checkpoint=use_checkpoint,
num_heads=num_heads,
num_head_channels=dim_head,
use_new_attention_order=use_new_attention_order,
) if not use_spatial_transformer else SpatialTransformer(
ch, num_heads, dim_head, depth=transformer_depth, context_dim=context_dim,
disable_self_attn=disabled_sa, use_linear=use_linear_in_transformer,
use_checkpoint=use_checkpoint
)
)
self.input_blocks.append(TimestepEmbedSequential(*layers))
self._feature_size += ch
input_block_chans.append(ch)
if level != len(channel_mult) - 1:
out_ch = ch
self.input_blocks.append(
TimestepEmbedSequential(
ResBlock(
ch,
time_embed_dim,
dropout,
out_channels=out_ch,
dims=dims,
use_checkpoint=use_checkpoint,
use_scale_shift_norm=use_scale_shift_norm,
down=True,
)
if resblock_updown
else Downsample(
ch, conv_resample, dims=dims, out_channels=out_ch
)
)
)
ch = out_ch
input_block_chans.append(ch)
ds *= 2
self._feature_size += ch
if num_head_channels == -1:
dim_head = ch // num_heads
else:
num_heads = ch // num_head_channels
dim_head = num_head_channels
if legacy:
#num_heads = 1
dim_head = ch // num_heads if use_spatial_transformer else num_head_channels
self.middle_block = TimestepEmbedSequential(
ResBlock(
ch,
time_embed_dim,
dropout,
dims=dims,
use_checkpoint=use_checkpoint,
use_scale_shift_norm=use_scale_shift_norm,
),
AttentionBlock(
ch,
use_checkpoint=use_checkpoint,
num_heads=num_heads,
num_head_channels=dim_head,
use_new_attention_order=use_new_attention_order,
) if not use_spatial_transformer else SpatialTransformer( # always uses a self-attn
ch, num_heads, dim_head, depth=transformer_depth, context_dim=context_dim,
disable_self_attn=disable_middle_self_attn, use_linear=use_linear_in_transformer,
use_checkpoint=use_checkpoint
),
ResBlock(
ch,
time_embed_dim,
dropout,
dims=dims,
use_checkpoint=use_checkpoint,
use_scale_shift_norm=use_scale_shift_norm,
),
)
self._feature_size += ch
self.output_blocks = nn.ModuleList([])
for level, mult in list(enumerate(channel_mult))[::-1]:
for i in range(self.num_res_blocks[level] + 1):
ich = input_block_chans.pop()
layers = [
ResBlock(
ch + ich,
time_embed_dim,
dropout,
out_channels=model_channels * mult,
dims=dims,
use_checkpoint=use_checkpoint,
use_scale_shift_norm=use_scale_shift_norm,
)
]
ch = model_channels * mult
if ds in attention_resolutions:
if num_head_channels == -1:
dim_head = ch // num_heads
else:
num_heads = ch // num_head_channels
dim_head = num_head_channels
if legacy:
#num_heads = 1
dim_head = ch // num_heads if use_spatial_transformer else num_head_channels
if exists(disable_self_attentions):
disabled_sa = disable_self_attentions[level]
else:
disabled_sa = False
if not exists(num_attention_blocks) or i < num_attention_blocks[level]:
layers.append(
AttentionBlock(
ch,
use_checkpoint=use_checkpoint,
num_heads=num_heads_upsample,
num_head_channels=dim_head,
use_new_attention_order=use_new_attention_order,
) if not use_spatial_transformer else SpatialTransformer(
ch, num_heads, dim_head, depth=transformer_depth, context_dim=context_dim,
disable_self_attn=disabled_sa, use_linear=use_linear_in_transformer,
use_checkpoint=use_checkpoint
)
)
if level and i == self.num_res_blocks[level]:
out_ch = ch
layers.append(
ResBlock(
ch,
time_embed_dim,
dropout,
out_channels=out_ch,
dims=dims,
use_checkpoint=use_checkpoint,
use_scale_shift_norm=use_scale_shift_norm,
up=True,
)
if resblock_updown
else Upsample(ch, conv_resample, dims=dims, out_channels=out_ch)
)
ds //= 2
self.output_blocks.append(TimestepEmbedSequential(*layers))
self._feature_size += ch
self.out = nn.Sequential(
normalization(ch),
nn.SiLU(),
zero_module(conv_nd(dims, model_channels, out_channels, 3, padding=1)),
)
if self.predict_codebook_ids:
self.id_predictor = nn.Sequential(
normalization(ch),
conv_nd(dims, model_channels, n_embed, 1),
#nn.LogSoftmax(dim=1) # change to cross_entropy and produce non-normalized logits
)
def convert_to_fp16(self):
"""
Convert the torso of the model to float16.
"""
self.input_blocks.apply(convert_module_to_f16)
self.middle_block.apply(convert_module_to_f16)
self.output_blocks.apply(convert_module_to_f16)
def convert_to_fp32(self):
"""
Convert the torso of the model to float32.
"""
self.input_blocks.apply(convert_module_to_f32)
self.middle_block.apply(convert_module_to_f32)
self.output_blocks.apply(convert_module_to_f32)
def forward(self, x, timesteps=None, context=None, y=None,**kwargs):
"""
Apply the model to an input batch.
:param x: an [N x C x ...] Tensor of inputs.
:param timesteps: a 1-D batch of timesteps.
:param context: conditioning plugged in via crossattn
:param y: an [N] Tensor of labels, if class-conditional.
:return: an [N x C x ...] Tensor of outputs.
"""
assert (y is not None) == (
self.num_classes is not None
), "must specify y if and only if the model is class-conditional"
hs = []
t_emb = timestep_embedding(timesteps, self.model_channels, repeat_only=False)
emb = self.time_embed(t_emb)
if self.num_classes is not None:
assert y.shape[0] == x.shape[0]
emb = emb + self.label_emb(y)
h = x.type(self.dtype)
for module in self.input_blocks:
h = module(h, emb, context)
hs.append(h)
h = self.middle_block(h, emb, context)
for module in self.output_blocks:
h = th.cat([h, hs.pop()], dim=1)
h = module(h, emb, context)
h = h.type(x.dtype)
if self.predict_codebook_ids:
return self.id_predictor(h)
else:
return self.out(h)
(3)文件upscaling.py定义了一些用于处理图像上采样的类,主要包括一个抽象基类 AbstractLowScaleModel,提供了一些基础功能,例如注册噪声调度和生成样本的函数 q_sample。类AbstractLowScaleModel的子类 SimpleImageConcat 实现了不带噪声水平条件的图像连接,返回输入图像及固定噪声水平。另一个子类 ImageConcatWithNoiseAugmentation 则允许根据随机噪声水平对图像进行增强处理,生成带噪声的样本,并返回生成的样本及相应的噪声水平。这些类主要用于在图像上采样过程中对低分辨率图像进行处理和增强。
import torch
import torch.nn as nn
import numpy as np
from functools import partial
from ldm.modules.diffusionmodules.util import extract_into_tensor, make_beta_schedule
from ldm.util import default
class AbstractLowScaleModel(nn.Module):
# for concatenating a downsampled image to the latent representation
def __init__(self, noise_schedule_config=None):
super(AbstractLowScaleModel, self).__init__()
if noise_schedule_config is not None:
self.register_schedule(**noise_schedule_config)
def register_schedule(self, beta_schedule="linear", timesteps=1000,
linear_start=1e-4, linear_end=2e-2, cosine_s=8e-3):
betas = make_beta_schedule(beta_schedule, timesteps, linear_start=linear_start, linear_end=linear_end,
cosine_s=cosine_s)
alphas = 1. - betas
alphas_cumprod = np.cumprod(alphas, axis=0)
alphas_cumprod_prev = np.append(1., alphas_cumprod[:-1])
timesteps, = betas.shape
self.num_timesteps = int(timesteps)
self.linear_start = linear_start
self.linear_end = linear_end
assert alphas_cumprod.shape[0] == self.num_timesteps, 'alphas have to be defined for each timestep'
to_torch = partial(torch.tensor, dtype=torch.float32)
self.register_buffer('betas', to_torch(betas))
self.register_buffer('alphas_cumprod', to_torch(alphas_cumprod))
self.register_buffer('alphas_cumprod_prev', to_torch(alphas_cumprod_prev))
# calculations for diffusion q(x_t | x_{t-1}) and others
self.register_buffer('sqrt_alphas_cumprod', to_torch(np.sqrt(alphas_cumprod)))
self.register_buffer('sqrt_one_minus_alphas_cumprod', to_torch(np.sqrt(1. - alphas_cumprod)))
self.register_buffer('log_one_minus_alphas_cumprod', to_torch(np.log(1. - alphas_cumprod)))
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))
return (extract_into_tensor(self.sqrt_alphas_cumprod, t, x_start.shape) * x_start +
extract_into_tensor(self.sqrt_one_minus_alphas_cumprod, t, x_start.shape) * noise)
def forward(self, x):
return x, None
def decode(self, x):
return x
class SimpleImageConcat(AbstractLowScaleModel):
# no noise level conditioning
def __init__(self):
super(SimpleImageConcat, self).__init__(noise_schedule_config=None)
self.max_noise_level = 0
def forward(self, x):
# fix to constant noise level
return x, torch.zeros(x.shape[0], device=x.device).long()
class ImageConcatWithNoiseAugmentation(AbstractLowScaleModel):
def __init__(self, noise_schedule_config, max_noise_level=1000, to_cuda=False):
super().__init__(noise_schedule_config=noise_schedule_config)
self.max_noise_level = max_noise_level
def forward(self, x, noise_level=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)
return z, noise_level
(4)文件util.py提供了一系列实用工具和函数,用于支持图像生成和处理任务,包括调度函数(如生成 beta 值和 DDIM 时间步)、张量操作(如噪声生成和提取函数)、模型参数处理(如归零和缩放参数)、正则化和嵌入生成、网络层创建、混合条件生成器,以及检查点功能以优化内存使用。这些功能为扩展和优化图像生成模型提供了灵活的基础设施。
import os
import math
import torch
import torch.nn as nn
import numpy as np
from einops import repeat
from ldm.util import instantiate_from_config
def make_beta_schedule(schedule, n_timestep, linear_start=1e-4, linear_end=2e-2, cosine_s=8e-3):
if schedule == "linear":
betas = (
torch.linspace(linear_start ** 0.5, linear_end ** 0.5, n_timestep, dtype=torch.float64) ** 2
)
elif schedule == "cosine":
timesteps = (
torch.arange(n_timestep + 1, dtype=torch.float64) / n_timestep + cosine_s
)
alphas = timesteps / (1 + cosine_s) * np.pi / 2
alphas = torch.cos(alphas).pow(2)
alphas = alphas / alphas[0]
betas = 1 - alphas[1:] / alphas[:-1]
betas = np.clip(betas, a_min=0, a_max=0.999)
elif schedule == "squaredcos_cap_v2": # used for karlo prior
# return early
return betas_for_alpha_bar(
n_timestep,
lambda t: math.cos((t + 0.008) / 1.008 * math.pi / 2) ** 2,
)
elif schedule == "sqrt_linear":
betas = torch.linspace(linear_start, linear_end, n_timestep, dtype=torch.float64)
elif schedule == "sqrt":
betas = torch.linspace(linear_start, linear_end, n_timestep, dtype=torch.float64) ** 0.5
else:
raise ValueError(f"schedule '{schedule}' unknown.")
return betas.numpy()
def make_ddim_timesteps(ddim_discr_method, num_ddim_timesteps, num_ddpm_timesteps, verbose=True):
if ddim_discr_method == 'uniform':
c = num_ddpm_timesteps // num_ddim_timesteps
ddim_timesteps = np.asarray(list(range(0, num_ddpm_timesteps, c)))
elif ddim_discr_method == 'quad':
ddim_timesteps = ((np.linspace(0, np.sqrt(num_ddpm_timesteps * .8), num_ddim_timesteps)) ** 2).astype(int)
else:
raise NotImplementedError(f'There is no ddim discretization method called "{ddim_discr_method}"')
# assert ddim_timesteps.shape[0] == num_ddim_timesteps
# add one to get the final alpha values right (the ones from first scale to data during sampling)
steps_out = ddim_timesteps + 1
if verbose:
print(f'Selected timesteps for ddim sampler: {steps_out}')
return steps_out
def make_ddim_sampling_parameters(alphacums, ddim_timesteps, eta, verbose=True):
# select alphas for computing the variance schedule
alphas = alphacums[ddim_timesteps]
alphas_prev = np.asarray([alphacums[0]] + alphacums[ddim_timesteps[:-1]].tolist())
# according the the formula provided in https://arxiv.org/abs/2010.02502
sigmas = eta * np.sqrt((1 - alphas_prev) / (1 - alphas) * (1 - alphas / alphas_prev))
if verbose:
print(f'Selected alphas for ddim sampler: a_t: {alphas}; a_(t-1): {alphas_prev}')
print(f'For the chosen value of eta, which is {eta}, '
f'this results in the following sigma_t schedule for ddim sampler {sigmas}')
return sigmas, alphas, alphas_prev
def betas_for_alpha_bar(num_diffusion_timesteps, alpha_bar, max_beta=0.999):
"""
Create a beta schedule that discretizes the given alpha_t_bar function,
which defines the cumulative product of (1-beta) over time from t = [0,1].
:param num_diffusion_timesteps: the number of betas to produce.
:param alpha_bar: a lambda that takes an argument t from 0 to 1 and
produces the cumulative product of (1-beta) up to that
part of the diffusion process.
:param max_beta: the maximum beta to use; use values lower than 1 to
prevent singularities.
"""
betas = []
for i in range(num_diffusion_timesteps):
t1 = i / num_diffusion_timesteps
t2 = (i + 1) / num_diffusion_timesteps
betas.append(min(1 - alpha_bar(t2) / alpha_bar(t1), max_beta))
return np.array(betas)
def extract_into_tensor(a, t, x_shape):
b, *_ = t.shape
out = a.gather(-1, t)
return out.reshape(b, *((1,) * (len(x_shape) - 1)))
def checkpoint(func, inputs, params, flag):
"""
Evaluate a function without caching intermediate activations, allowing for
reduced memory at the expense of extra compute in the backward pass.
:param func: the function to evaluate.
:param inputs: the argument sequence to pass to `func`.
:param params: a sequence of parameters `func` depends on but does not
explicitly take as arguments.
:param flag: if False, disable gradient checkpointing.
"""
if flag:
args = tuple(inputs) + tuple(params)
return CheckpointFunction.apply(func, len(inputs), *args)
else:
return func(*inputs)
class CheckpointFunction(torch.autograd.Function):
@staticmethod
def forward(ctx, run_function, length, *args):
ctx.run_function = run_function
ctx.input_tensors = list(args[:length])
ctx.input_params = list(args[length:])
ctx.gpu_autocast_kwargs = {"enabled": torch.is_autocast_enabled(),
"dtype": torch.get_autocast_gpu_dtype(),
"cache_enabled": torch.is_autocast_cache_enabled()}
with torch.no_grad():
output_tensors = ctx.run_function(*ctx.input_tensors)
return output_tensors
@staticmethod
def backward(ctx, *output_grads):
ctx.input_tensors = [x.detach().requires_grad_(True) for x in ctx.input_tensors]
with torch.enable_grad(), \
torch.cuda.amp.autocast(**ctx.gpu_autocast_kwargs):
# Fixes a bug where the first op in run_function modifies the
# Tensor storage in place, which is not allowed for detach()'d
# Tensors.
shallow_copies = [x.view_as(x) for x in ctx.input_tensors]
output_tensors = ctx.run_function(*shallow_copies)
input_grads = torch.autograd.grad(
output_tensors,
ctx.input_tensors + ctx.input_params,
output_grads,
allow_unused=True,
)
del ctx.input_tensors
del ctx.input_params
del output_tensors
return (None, None) + input_grads
def timestep_embedding(timesteps, dim, max_period=10000, repeat_only=False):
"""
Create sinusoidal timestep embeddings.
:param timesteps: a 1-D Tensor of N indices, one per batch element.
These may be fractional.
:param dim: the dimension of the output.
:param max_period: controls the minimum frequency of the embeddings.
:return: an [N x dim] Tensor of positional embeddings.
"""
if not repeat_only:
half = dim // 2
freqs = torch.exp(
-math.log(max_period) * torch.arange(start=0, end=half, dtype=torch.float32) / half
).to(device=timesteps.device)
args = timesteps[:, None].float() * freqs[None]
embedding = torch.cat([torch.cos(args), torch.sin(args)], dim=-1)
if dim % 2:
embedding = torch.cat([embedding, torch.zeros_like(embedding[:, :1])], dim=-1)
else:
embedding = repeat(timesteps, 'b -> b d', d=dim)
return embedding
def zero_module(module):
"""
Zero out the parameters of a module and return it.
"""
for p in module.parameters():
p.detach().zero_()
return module
def scale_module(module, scale):
"""
Scale the parameters of a module and return it.
"""
for p in module.parameters():
p.detach().mul_(scale)
return module
def mean_flat(tensor):
"""
Take the mean over all non-batch dimensions.
"""
return tensor.mean(dim=list(range(1, len(tensor.shape))))
def normalization(channels):
"""
Make a standard normalization layer.
:param channels: number of input channels.
:return: an nn.Module for normalization.
"""
return GroupNorm32(32, channels)
# PyTorch 1.7 has SiLU, but we support PyTorch 1.5.
class SiLU(nn.Module):
def forward(self, x):
return x * torch.sigmoid(x)
class GroupNorm32(nn.GroupNorm):
def forward(self, x):
return super().forward(x.float()).type(x.dtype)
def conv_nd(dims, *args, **kwargs):
"""
Create a 1D, 2D, or 3D convolution module.
"""
if dims == 1:
return nn.Conv1d(*args, **kwargs)
elif dims == 2:
return nn.Conv2d(*args, **kwargs)
elif dims == 3:
return nn.Conv3d(*args, **kwargs)
raise ValueError(f"unsupported dimensions: {dims}")
def linear(*args, **kwargs):
"""
Create a linear module.
"""
return nn.Linear(*args, **kwargs)
def avg_pool_nd(dims, *args, **kwargs):
"""
Create a 1D, 2D, or 3D average pooling module.
"""
if dims == 1:
return nn.AvgPool1d(*args, **kwargs)
elif dims == 2:
return nn.AvgPool2d(*args, **kwargs)
elif dims == 3:
return nn.AvgPool3d(*args, **kwargs)
raise ValueError(f"unsupported dimensions: {dims}")
class HybridConditioner(nn.Module):
def __init__(self, c_concat_config, c_crossattn_config):
super().__init__()
self.concat_conditioner = instantiate_from_config(c_concat_config)
self.crossattn_conditioner = instantiate_from_config(c_crossattn_config)
def forward(self, c_concat, c_crossattn):
c_concat = self.concat_conditioner(c_concat)
c_crossattn = self.crossattn_conditioner(c_crossattn)
return {'c_concat': [c_concat], 'c_crossattn': [c_crossattn]}
def noise_like(shape, device, repeat=False):
repeat_noise = lambda: torch.randn((1, *shape[1:]), device=device).repeat(shape[0], *((1,) * (len(shape) - 1)))
noise = lambda: torch.randn(shape, device=device)
return repeat_noise() if repeat else noise()