235 lines
8.9 KiB
Python
235 lines
8.9 KiB
Python
import math
|
|
import torch
|
|
import torch.nn as nn
|
|
import torch.nn.functional as F
|
|
from model.BrownianBridge.base.modules.diffusionmodules.util import GroupNorm32
|
|
|
|
def nonlinearity(x):
|
|
return x*torch.sigmoid(x)
|
|
|
|
def Normalize(in_channels):
|
|
return GroupNorm32(32, 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, 3, 1, 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:
|
|
self.conv = torch.nn.Conv2d(in_channels, in_channels, 3, 2, 0)
|
|
|
|
def forward(self, x):
|
|
if self.with_conv:
|
|
x = torch.nn.functional.pad(x, (0,1,0,1), 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, 3, 1, 1)
|
|
|
|
self.norm2 = Normalize(out_channels)
|
|
self.dropout = torch.nn.Dropout(dropout)
|
|
self.conv2 = torch.nn.Conv2d(out_channels, out_channels, 3, 1, 1)
|
|
|
|
if self.in_channels != self.out_channels:
|
|
if self.use_conv_shortcut:
|
|
self.conv_shortcut = torch.nn.Conv2d(in_channels, out_channels, 3, 1, 1)
|
|
else:
|
|
self.nin_shortcut = torch.nn.Conv2d(in_channels, out_channels, 1, 1, 0)
|
|
|
|
def forward(self, x, temb):
|
|
h = self.norm1(x)
|
|
h = nonlinearity(h)
|
|
h = self.conv1(h)
|
|
|
|
if temb is not None:
|
|
# VQGAN usually doesn't use temb in this node's inference path
|
|
pass
|
|
|
|
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, 1)
|
|
self.k = torch.nn.Conv2d(in_channels, in_channels, 1)
|
|
self.v = torch.nn.Conv2d(in_channels, in_channels, 1)
|
|
self.proj_out = torch.nn.Conv2d(in_channels, in_channels, 1)
|
|
|
|
def forward(self, x):
|
|
h_ = self.norm(x)
|
|
q = self.q(h_)
|
|
k = self.k(h_)
|
|
v = self.v(h_)
|
|
|
|
b,c,h,w = q.shape
|
|
q = q.reshape(b,c,h*w).permute(0,2,1)
|
|
k = k.reshape(b,c,h*w)
|
|
w_ = torch.bmm(q.float(), k.float()) * (int(c)**(-0.5))
|
|
w_ = torch.nn.functional.softmax(w_, dim=2).type(x.dtype)
|
|
|
|
v = v.reshape(b,c,h*w)
|
|
h_ = torch.bmm(v, w_.permute(0,2,1)).reshape(b,c,h,w)
|
|
h_ = self.proj_out(h_)
|
|
return x+h_
|
|
|
|
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, **ignore_kwargs):
|
|
super().__init__()
|
|
self.ch = ch
|
|
self.num_resolutions = len(ch_mult)
|
|
self.num_res_blocks = num_res_blocks
|
|
self.conv_in = torch.nn.Conv2d(in_channels, self.ch, 3, 1, 1)
|
|
|
|
in_ch_mult = (1,)+tuple(ch_mult)
|
|
self.down = nn.ModuleList()
|
|
curr_res = resolution
|
|
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=0, dropout=dropout))
|
|
block_in = block_out
|
|
if curr_res in attn_resolutions:
|
|
attn.append(AttnBlock(block_in))
|
|
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)
|
|
|
|
self.mid = nn.Module()
|
|
self.mid.block_1 = ResnetBlock(in_channels=block_in, out_channels=block_in, temb_channels=0, dropout=dropout)
|
|
self.mid.attn_1 = AttnBlock(block_in)
|
|
self.mid.block_2 = ResnetBlock(in_channels=block_in, out_channels=block_in, temb_channels=0, dropout=dropout)
|
|
|
|
self.norm_out = Normalize(block_in)
|
|
self.conv_out = torch.nn.Conv2d(block_in, 2*z_channels if double_z else z_channels, 3, 1, 1)
|
|
|
|
def forward(self, x):
|
|
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], None)
|
|
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]))
|
|
|
|
h = hs[-1]
|
|
h = self.mid.block_1(h, None)
|
|
h = self.mid.attn_1(h)
|
|
h = self.mid.block_2(h, None)
|
|
|
|
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, **ignorekwargs):
|
|
super().__init__()
|
|
self.ch = ch
|
|
self.num_resolutions = len(ch_mult)
|
|
self.num_res_blocks = num_res_blocks
|
|
self.give_pre_end = give_pre_end
|
|
|
|
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.conv_in = torch.nn.Conv2d(z_channels, block_in, 3, 1, 1)
|
|
|
|
self.mid = nn.Module()
|
|
self.mid.block_1 = ResnetBlock(in_channels=block_in, out_channels=block_in, temb_channels=0, dropout=dropout)
|
|
self.mid.attn_1 = AttnBlock(block_in)
|
|
self.mid.block_2 = ResnetBlock(in_channels=block_in, out_channels=block_in, temb_channels=0, dropout=dropout)
|
|
|
|
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=0, dropout=dropout))
|
|
block_in = block_out
|
|
if curr_res in attn_resolutions:
|
|
attn.append(AttnBlock(block_in))
|
|
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)
|
|
|
|
self.norm_out = Normalize(block_in)
|
|
self.conv_out = torch.nn.Conv2d(block_in, out_ch, 3, 1, 1)
|
|
|
|
def forward(self, z):
|
|
h = self.conv_in(z)
|
|
h = self.mid.block_1(h, None)
|
|
h = self.mid.attn_1(h)
|
|
h = self.mid.block_2(h, None)
|
|
|
|
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, None)
|
|
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)
|
|
|
|
if self.give_pre_end:
|
|
return h
|
|
|
|
h = self.norm_out(h)
|
|
h = nonlinearity(h)
|
|
h = self.conv_out(h)
|
|
return h
|