MCPcopy Create free account
hub / github.com/dome272/Diffusion-Models-pytorch / sample

Method sample

ddpm.py:39–58  ·  view source on GitHub ↗
(self, model, n)

Source from the content-addressed store, hash-verified

37 return torch.randint(low=1, high=self.noise_steps, size=(n,))
38
39 def sample(self, model, n):
40 logging.info(f"Sampling {n} new images....")
41 model.eval()
42 with torch.no_grad():
43 x = torch.randn((n, 3, self.img_size, self.img_size)).to(self.device)
44 for i in tqdm(reversed(range(1, self.noise_steps)), position=0):
45 t = (torch.ones(n) * i).long().to(self.device)
46 predicted_noise = model(x, t)
47 alpha = self.alpha[t][:, None, None, None]
48 alpha_hat = self.alpha_hat[t][:, None, None, None]
49 beta = self.beta[t][:, None, None, None]
50 if i > 1:
51 noise = torch.randn_like(x)
52 else:
53 noise = torch.zeros_like(x)
54 x = 1 / torch.sqrt(alpha) * (x - ((1 - alpha) / (torch.sqrt(1 - alpha_hat))) * predicted_noise) + torch.sqrt(beta) * noise
55 model.train()
56 x = (x.clamp(-1, 1) + 1) / 2
57 x = (x * 255).type(torch.uint8)
58 return x
59
60
61def train(args):

Callers 1

trainFunction · 0.95

Calls

no outgoing calls

Tested by

no test coverage detected