Create pipeline.py
Browse files- pipeline.py +55 -0
pipeline.py
ADDED
@@ -0,0 +1,55 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import torch
|
2 |
+
import torch.nn as nn
|
3 |
+
from torchvision import transforms
|
4 |
+
|
5 |
+
class Generator(nn.Module):
|
6 |
+
def __init__(self, nz=128, ngf=64, nc=3):
|
7 |
+
super(Generator, self).__init__()
|
8 |
+
self.main = nn.Sequential(
|
9 |
+
nn.ConvTranspose2d(nz, ngf * 8, 4, 1, 0, bias=False),
|
10 |
+
nn.BatchNorm2d(ngf * 8),
|
11 |
+
nn.LeakyReLU(0.2, inplace=True),
|
12 |
+
nn.Dropout(0.2),
|
13 |
+
|
14 |
+
nn.ConvTranspose2d(ngf * 8, ngf * 4, 4, 2, 1, bias=False),
|
15 |
+
nn.BatchNorm2d(ngf * 4),
|
16 |
+
nn.LeakyReLU(0.2, inplace=True),
|
17 |
+
|
18 |
+
nn.ConvTranspose2d(ngf * 4, ngf * 2, 4, 2, 1, bias=False),
|
19 |
+
nn.BatchNorm2d(ngf * 2),
|
20 |
+
nn.LeakyReLU(0.2, inplace=True),
|
21 |
+
|
22 |
+
nn.ConvTranspose2d(ngf * 2, ngf, 4, 2, 1, bias=False),
|
23 |
+
nn.BatchNorm2d(ngf),
|
24 |
+
nn.LeakyReLU(0.2, inplace=True),
|
25 |
+
|
26 |
+
nn.ConvTranspose2d(ngf, nc, 4, 2, 1, bias=False),
|
27 |
+
nn.Tanh()
|
28 |
+
)
|
29 |
+
|
30 |
+
def forward(self, input):
|
31 |
+
output = self.main(input)
|
32 |
+
return output
|
33 |
+
|
34 |
+
class PreTrainedPipeline():
|
35 |
+
def __init__(self, path=""):
|
36 |
+
self.device = "cuda" if torch.cuda.is_available() else "cpu"
|
37 |
+
self.model = Generator().to(self.device)
|
38 |
+
self.model.load_state_dict(torch.load("pytorch_model.bin", map_location=device))
|
39 |
+
|
40 |
+
def __call__(self, inputs: str):
|
41 |
+
"""
|
42 |
+
Args:
|
43 |
+
inputs (:obj:`str`):
|
44 |
+
a string containing some text
|
45 |
+
Return:
|
46 |
+
A :obj:`PIL.Image` with the raw image representation as PIL.
|
47 |
+
"""
|
48 |
+
noise = torch.randn(1, 128, 1, 1, device=self.device)
|
49 |
+
with torch.no_grad():
|
50 |
+
output = self.model(noise).cpu()
|
51 |
+
|
52 |
+
img = output[0]
|
53 |
+
img = (img + 1) / 2
|
54 |
+
img = transforms.ToPILImage()(img)
|
55 |
+
return img
|