pipeline.py

import torch
import torch.nn as nn
from torchvision import transforms

class Generator(nn.Module):
    def __init__(self, nz=128, ngf=64, nc=3):
        super(Generator, self).__init__()
        self.main = nn.Sequential(
            nn.ConvTranspose2d(nz, ngf * 8, 4, 1, 0, bias=False),
            nn.BatchNorm2d(ngf * 8),
            nn.LeakyReLU(0.2, inplace=True),
            nn.Dropout(0.2),

            nn.ConvTranspose2d(ngf * 8, ngf * 4, 4, 2, 1, bias=False),
            nn.BatchNorm2d(ngf * 4),
            nn.LeakyReLU(0.2, inplace=True),

            nn.ConvTranspose2d(ngf * 4, ngf * 2, 4, 2, 1, bias=False),
            nn.BatchNorm2d(ngf * 2),
            nn.LeakyReLU(0.2, inplace=True),

            nn.ConvTranspose2d(ngf * 2, ngf, 4, 2, 1, bias=False),
            nn.BatchNorm2d(ngf),
            nn.LeakyReLU(0.2, inplace=True),

            nn.ConvTranspose2d(ngf, nc, 4, 2, 1, bias=False),
            nn.Tanh()
        )

    def forward(self, input):
        output = self.main(input)
        return output

class PreTrainedPipeline():
    def __init__(self, path=""):
        self.device = "cuda" if torch.cuda.is_available() else "cpu"
        self.model = Generator().to(self.device)
        self.model.load_state_dict(torch.load("pytorch_model.bin", map_location=self.device))
 
    def __call__(self, inputs: str):
        """
        Args:
            inputs (:obj:`str`):
                a string containing some text
        Return:
            A :obj:`PIL.Image` with the raw image representation as PIL.
        """
        noise = torch.randn(1, 128, 1, 1, device=self.device)
        with torch.no_grad():
            output = self.model(noise).cpu()

        img = output[0]
        img = (img + 1) / 2
        img = transforms.ToPILImage()(img)
        return img