File size: 6,015 Bytes
7446b5a d659f17 7446b5a 90040e7 7446b5a |
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 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 |
import streamlit as st
from PIL import Image, ImageOps
import torch
from matplotlib.image import imread
import numpy as np
import tensorflow as tf
import math
class Block(nn.Module):
def __init__(self, in_ch, out_ch, time_emb_dim, up=False):
super().__init__()
self.time_mlp = nn.Linear(time_emb_dim, out_ch)
if up:
self.conv1 = nn.Conv2d(2*in_ch, out_ch, 3, padding=1)
self.transform = nn.ConvTranspose2d(out_ch, out_ch, 4, 2, 1)
self.Upsample = nn.Upsample(scale_factor = 2, mode ='bilinear')
else:
self.conv1 = nn.Conv2d(in_ch, out_ch, 3, padding=1)
self.transform = nn.Conv2d(out_ch, out_ch, 4, 2, 1)
self.maxpool = nn.MaxPool2d(4, 2, 1)
self.conv2 = nn.Conv2d(out_ch, out_ch, 3, padding=1)
self.bnorm1 = nn.BatchNorm2d(out_ch)
self.bnorm2 = nn.BatchNorm2d(out_ch)
self.silu = nn.SiLU()
self.relu = nn.ReLU()
def forward(self, x, t, ):
# First Conv
h = (self.silu(self.bnorm1(self.conv1(x))))
# Time embedding
time_emb = self.relu(self.time_mlp(t))
# Extend last 2 dimensions
time_emb = time_emb[(..., ) + (None, ) * 2]
# Add time channel
h = h + time_emb
# Second Conv
h = (self.silu(self.bnorm2(self.conv2(h))))
# Down or Upsample
return self.transform(h)
class SinusoidalPositionEmbeddings(nn.Module):
def __init__(self, dim):
super().__init__()
self.dim = dim
def forward(self, time):
device = time.device
half_dim = self.dim // 2
embeddings = math.log(10000) / (half_dim - 1)
embeddings = torch.exp(torch.arange(half_dim, device=device) * -embeddings)
embeddings = time[:, None] * embeddings[None, :]
embeddings = torch.cat((embeddings.sin(), embeddings.cos()), dim=-1)
# TODO: Double check the ordering here
return embeddings
class SimpleUnet(nn.Module):
"""
A simplified variant of the Unet architecture.
"""
def __init__(self):
super().__init__()
image_channels = 3
down_channels = (32, 64, 128, 256, 512)
up_channels = (512, 256, 128, 64, 32)
out_dim = 3
time_emb_dim = 32
# Time embedding
self.time_mlp = nn.Sequential(
SinusoidalPositionEmbeddings(time_emb_dim),
nn.Linear(time_emb_dim, time_emb_dim),
nn.ReLU()
)
# Initial projection
self.conv0 = nn.Conv2d(image_channels, down_channels[0], 3, padding=1)
# Downsample
self.downs = nn.ModuleList([Block(down_channels[i], down_channels[i+1], \
time_emb_dim) \
for i in range(len(down_channels)-1)])
# Upsample
self.ups = nn.ModuleList([Block(up_channels[i], up_channels[i+1], \
time_emb_dim, up=True) \
for i in range(len(up_channels)-1)])
# Edit: Corrected a bug found by Jakub C (see YouTube comment)
self.output = nn.Conv2d(up_channels[-1], out_dim, 1)
def forward(self, x, timestep):
# Embedd time
t = self.time_mlp(timestep)
# Initial conv
x = self.conv0(x)
# Unet
residual_inputs = []
for down in self.downs:
x = down(x, t)
residual_inputs.append(x)
for up in self.ups:
residual_x = residual_inputs.pop()
# Add residual x as additional channels
x = torch.cat((x, residual_x), dim=1)
x = up(x, t)
return self.output(x)
def extract(a, t, x_shape):
batch_size = t.shape[0]
out = a.gather(-1, t.cpu())
return out.reshape(batch_size, *((1,) * (len(x_shape) - 1))).to(t.device)
@torch.no_grad()
def p_sample(model, x, t, t_index):
betas_t = extract(betas, t, x.shape)
sqrt_one_minus_alphas_cumprod_t = extract(
sqrt_one_minus_alphas_cumprod, t, x.shape
)
sqrt_recip_alphas_t = extract(sqrt_recip_alphas, t, x.shape)
# Equation 11 in the paper
# Use our model (noise predictor) to predict the mean
model_mean = sqrt_recip_alphas_t * (
x - betas_t * model(x, t) / sqrt_one_minus_alphas_cumprod_t
)
if t_index == 0:
return model_mean
else:
posterior_variance_t = extract(posterior_variance, t, x.shape)
noise = torch.randn_like(x)
# Algorithm 2 line 4:
return model_mean + torch.sqrt(posterior_variance_t) * noise
# Algorithm 2 but save all images:
@torch.no_grad()
def p_sample_loop(model, shape):
device = next(model.parameters()).device
b = shape[0]
# start from pure noise (for each example in the batch)
img = torch.randn(shape, device=device)
imgs = []
for i in tqdm(reversed(range(0, timesteps)), desc='sampling loop time step', total=timesteps):
img = p_sample(model, img, torch.full((b,), i, device=device, dtype=torch.long), 3)
imgs.append(img.cpu().numpy())
return imgs
@torch.no_grad()
def sample(model, image_size, batch_size=16, channels=3):
return p_sample_loop(model, shape=(batch_size, channels, image_size, image_size))
samples = sample(model, image_size=img_size, batch_size=64, channels=3)
reverse_transforms = transforms.Compose([
transforms.Lambda(lambda t: (t + 1) / 2),
transforms.Lambda(lambda t: t.permute(1, 2, 0)), # CHW to HWC
transforms.Lambda(lambda t: t * 255.),
transforms.Lambda(lambda t: t.numpy().astype(np.uint8)),
transforms.ToPILImage(),
])
for i in range(10):
img = reverse_transforms(torch.Tensor((samples[-1][i].reshape(3, img_size, img_size))))
plt.imshow(img)
model = SimpleUnet()
st.title("Generatig images using a diffusion model")
model.load_state_dict(torch.load("new_linear_model_1090.pt"))
result = st.button("Click to generate image")
if(result):
model()
|