File size: 6,880 Bytes
ed01507
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
180
181
182
183
184
185
186
187
188
189
190
191
192
193
import numpy as np
import torch
from loguru import logger
from tqdm import tqdm

from lama_cleaner.model.utils import make_ddim_timesteps, make_ddim_sampling_parameters, noise_like


class DDIMSampler(object):
    def __init__(self, model, schedule="linear"):
        super().__init__()
        self.model = model
        self.ddpm_num_timesteps = model.num_timesteps
        self.schedule = schedule

    def register_buffer(self, name, attr):
        setattr(self, name, attr)

    def make_schedule(
        self, ddim_num_steps, ddim_discretize="uniform", ddim_eta=0.0, verbose=True
    ):
        self.ddim_timesteps = make_ddim_timesteps(
            ddim_discr_method=ddim_discretize,
            num_ddim_timesteps=ddim_num_steps,
            # array([1])
            num_ddpm_timesteps=self.ddpm_num_timesteps,
            verbose=verbose,
        )
        alphas_cumprod = self.model.alphas_cumprod  # torch.Size([1000])
        assert (
                alphas_cumprod.shape[0] == self.ddpm_num_timesteps
        ), "alphas have to be defined for each timestep"
        to_torch = lambda x: x.clone().detach().to(torch.float32).to(self.model.device)

        self.register_buffer("betas", to_torch(self.model.betas))
        self.register_buffer("alphas_cumprod", to_torch(alphas_cumprod))
        self.register_buffer(
            "alphas_cumprod_prev", to_torch(self.model.alphas_cumprod_prev)
        )

        # calculations for diffusion q(x_t | x_{t-1}) and others
        self.register_buffer(
            "sqrt_alphas_cumprod", to_torch(np.sqrt(alphas_cumprod.cpu()))
        )
        self.register_buffer(
            "sqrt_one_minus_alphas_cumprod",
            to_torch(np.sqrt(1.0 - alphas_cumprod.cpu())),
        )
        self.register_buffer(
            "log_one_minus_alphas_cumprod", to_torch(np.log(1.0 - alphas_cumprod.cpu()))
        )
        self.register_buffer(
            "sqrt_recip_alphas_cumprod", to_torch(np.sqrt(1.0 / alphas_cumprod.cpu()))
        )
        self.register_buffer(
            "sqrt_recipm1_alphas_cumprod",
            to_torch(np.sqrt(1.0 / alphas_cumprod.cpu() - 1)),
        )

        # ddim sampling parameters
        ddim_sigmas, ddim_alphas, ddim_alphas_prev = make_ddim_sampling_parameters(
            alphacums=alphas_cumprod.cpu(),
            ddim_timesteps=self.ddim_timesteps,
            eta=ddim_eta,
            verbose=verbose,
        )
        self.register_buffer("ddim_sigmas", ddim_sigmas)
        self.register_buffer("ddim_alphas", ddim_alphas)
        self.register_buffer("ddim_alphas_prev", ddim_alphas_prev)
        self.register_buffer("ddim_sqrt_one_minus_alphas", np.sqrt(1.0 - ddim_alphas))
        sigmas_for_original_sampling_steps = ddim_eta * torch.sqrt(
            (1 - self.alphas_cumprod_prev)
            / (1 - self.alphas_cumprod)
            * (1 - self.alphas_cumprod / self.alphas_cumprod_prev)
        )
        self.register_buffer(
            "ddim_sigmas_for_original_num_steps", sigmas_for_original_sampling_steps
        )

    @torch.no_grad()
    def sample(self, steps, conditioning, batch_size, shape):
        self.make_schedule(ddim_num_steps=steps, ddim_eta=0, verbose=False)
        # sampling
        C, H, W = shape
        size = (batch_size, C, H, W)

        # samples: 1,3,128,128
        return self.ddim_sampling(
            conditioning,
            size,
            quantize_denoised=False,
            ddim_use_original_steps=False,
            noise_dropout=0,
            temperature=1.0,
        )

    @torch.no_grad()
    def ddim_sampling(
        self,
        cond,
        shape,
        ddim_use_original_steps=False,
        quantize_denoised=False,
        temperature=1.0,
        noise_dropout=0.0,
    ):
        device = self.model.betas.device
        b = shape[0]
        img = torch.randn(shape, device=device, dtype=cond.dtype)
        timesteps = (
            self.ddpm_num_timesteps if ddim_use_original_steps else self.ddim_timesteps
        )

        time_range = (
            reversed(range(0, timesteps))
            if ddim_use_original_steps
            else np.flip(timesteps)
        )
        total_steps = timesteps if ddim_use_original_steps else timesteps.shape[0]
        logger.info(f"Running DDIM Sampling with {total_steps} timesteps")

        iterator = tqdm(time_range, desc="DDIM Sampler", total=total_steps)

        for i, step in enumerate(iterator):
            index = total_steps - i - 1
            ts = torch.full((b,), step, device=device, dtype=torch.long)

            outs = self.p_sample_ddim(
                img,
                cond,
                ts,
                index=index,
                use_original_steps=ddim_use_original_steps,
                quantize_denoised=quantize_denoised,
                temperature=temperature,
                noise_dropout=noise_dropout,
            )
            img, _ = outs

        return img

    @torch.no_grad()
    def p_sample_ddim(
        self,
        x,
        c,
        t,
        index,
        repeat_noise=False,
        use_original_steps=False,
        quantize_denoised=False,
        temperature=1.0,
        noise_dropout=0.0,
    ):
        b, *_, device = *x.shape, x.device
        e_t = self.model.apply_model(x, t, c)

        alphas = self.model.alphas_cumprod if use_original_steps else self.ddim_alphas
        alphas_prev = (
            self.model.alphas_cumprod_prev
            if use_original_steps
            else self.ddim_alphas_prev
        )
        sqrt_one_minus_alphas = (
            self.model.sqrt_one_minus_alphas_cumprod
            if use_original_steps
            else self.ddim_sqrt_one_minus_alphas
        )
        sigmas = (
            self.model.ddim_sigmas_for_original_num_steps
            if use_original_steps
            else self.ddim_sigmas
        )
        # select parameters corresponding to the currently considered timestep
        a_t = torch.full((b, 1, 1, 1), alphas[index], device=device)
        a_prev = torch.full((b, 1, 1, 1), alphas_prev[index], device=device)
        sigma_t = torch.full((b, 1, 1, 1), sigmas[index], device=device)
        sqrt_one_minus_at = torch.full(
            (b, 1, 1, 1), sqrt_one_minus_alphas[index], device=device
        )

        # current prediction for x_0
        pred_x0 = (x - sqrt_one_minus_at * e_t) / a_t.sqrt()
        if quantize_denoised:  # 没用
            pred_x0, _, *_ = self.model.first_stage_model.quantize(pred_x0)
        # direction pointing to x_t
        dir_xt = (1.0 - a_prev - sigma_t ** 2).sqrt() * e_t
        noise = sigma_t * noise_like(x.shape, device, repeat_noise) * temperature
        if noise_dropout > 0.0:  # 没用
            noise = torch.nn.functional.dropout(noise, p=noise_dropout)
        x_prev = a_prev.sqrt() * pred_x0 + dir_xt + noise
        return x_prev, pred_x0