File size: 3,143 Bytes
4450790 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 |
#!/usr/bin/env python3
"""
Tiny AutoEncoder for Stable Diffusion
(DNN for encoding / decoding SD's latent space)
"""
import torch
import torch.nn as nn
import comfy.utils
import comfy.ops
def conv(n_in, n_out, **kwargs):
return comfy.ops.disable_weight_init.Conv2d(n_in, n_out, 3, padding=1, **kwargs)
class Clamp(nn.Module):
def forward(self, x):
return torch.tanh(x / 3) * 3
class Block(nn.Module):
def __init__(self, n_in, n_out):
super().__init__()
self.conv = nn.Sequential(conv(n_in, n_out), nn.ReLU(), conv(n_out, n_out), nn.ReLU(), conv(n_out, n_out))
self.skip = comfy.ops.disable_weight_init.Conv2d(n_in, n_out, 1, bias=False) if n_in != n_out else nn.Identity()
self.fuse = nn.ReLU()
def forward(self, x):
return self.fuse(self.conv(x) + self.skip(x))
def Encoder(latent_channels=4):
return nn.Sequential(
conv(3, 64), Block(64, 64),
conv(64, 64, stride=2, bias=False), Block(64, 64), Block(64, 64), Block(64, 64),
conv(64, 64, stride=2, bias=False), Block(64, 64), Block(64, 64), Block(64, 64),
conv(64, 64, stride=2, bias=False), Block(64, 64), Block(64, 64), Block(64, 64),
conv(64, latent_channels),
)
def Decoder(latent_channels=4):
return nn.Sequential(
Clamp(), conv(latent_channels, 64), nn.ReLU(),
Block(64, 64), Block(64, 64), Block(64, 64), nn.Upsample(scale_factor=2), conv(64, 64, bias=False),
Block(64, 64), Block(64, 64), Block(64, 64), nn.Upsample(scale_factor=2), conv(64, 64, bias=False),
Block(64, 64), Block(64, 64), Block(64, 64), nn.Upsample(scale_factor=2), conv(64, 64, bias=False),
Block(64, 64), conv(64, 3),
)
class TAESD(nn.Module):
latent_magnitude = 3
latent_shift = 0.5
def __init__(self, encoder_path=None, decoder_path=None, latent_channels=4):
"""Initialize pretrained TAESD on the given device from the given checkpoints."""
super().__init__()
self.taesd_encoder = Encoder(latent_channels=latent_channels)
self.taesd_decoder = Decoder(latent_channels=latent_channels)
self.vae_scale = torch.nn.Parameter(torch.tensor(1.0))
self.vae_shift = torch.nn.Parameter(torch.tensor(0.0))
if encoder_path is not None:
self.taesd_encoder.load_state_dict(comfy.utils.load_torch_file(encoder_path, safe_load=True))
if decoder_path is not None:
self.taesd_decoder.load_state_dict(comfy.utils.load_torch_file(decoder_path, safe_load=True))
@staticmethod
def scale_latents(x):
"""raw latents -> [0, 1]"""
return x.div(2 * TAESD.latent_magnitude).add(TAESD.latent_shift).clamp(0, 1)
@staticmethod
def unscale_latents(x):
"""[0, 1] -> raw latents"""
return x.sub(TAESD.latent_shift).mul(2 * TAESD.latent_magnitude)
def decode(self, x):
x_sample = self.taesd_decoder((x - self.vae_shift) * self.vae_scale)
x_sample = x_sample.sub(0.5).mul(2)
return x_sample
def encode(self, x):
return (self.taesd_encoder(x * 0.5 + 0.5) / self.vae_scale) + self.vae_shift
|