imagand/diffusion.py at main · GenerativeDrugDiscovery/imagand · GitHub

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
# Disclaimer: This code is strongly influenced by
# https://github.com/huggingface/diffusers/blob/main/src/diffusers/schedulers/scheduling_ddim.py

import math
from typing import Union

import numpy as np
import torch
from tqdm import tqdm

from utils import match_shape, clip


def cosine_beta_schedule(timesteps, beta_start=0.0, beta_end=0.999, s=0.008):
    steps = timesteps + 1
    x = torch.linspace(0, timesteps, steps, dtype=torch.float32)
    alphas_cumprod = torch.cos(
        ((x / timesteps) + s) / (1 + s) * math.pi * 0.5)**2
    alphas_cumprod = alphas_cumprod / alphas_cumprod[0]
    betas = 1 - (alphas_cumprod[1:] / alphas_cumprod[:-1])
    return torch.clip(betas, beta_start, beta_end)


class DDIMScheduler:

    def __init__(self,
                 num_train_timesteps=500,
                 beta_start=0.0001,
                 beta_end=0.02,
                 beta_schedule="cosine",
                 clip_sample=True,
                 set_alpha_to_one=True):

        if beta_schedule == "linear":
            self.betas = np.linspace(beta_start,
                                     beta_end,
                                     num_train_timesteps,
                                     dtype=np.float32)
            self.betas = torch.tensor(self.betas)
        elif beta_schedule == "cosine":
            self.betas = cosine_beta_schedule(num_train_timesteps,
                                              beta_start=beta_start,
                                              beta_end=beta_end)
        else:
            raise NotImplementedError(
                f"{beta_schedule} does is not implemented for {self.__class__}")

        self.num_train_timesteps = num_train_timesteps
        self.clip_sample = clip_sample
        self.alphas = 1.0 - self.betas
        self.alphas_cumprod = np.cumprod(self.alphas, axis=0)

        self.final_alpha_cumprod = np.array(
            1.0) if set_alpha_to_one else self.alphas_cumprod[0]

        self.num_inference_steps = None
        self.timesteps = np.arange(0, num_train_timesteps)[::-1].copy()

    def _get_variance(self, timestep, prev_timestep):
        alpha_prod_t = self.alphas_cumprod[timestep]
        alpha_prod_t_prev = self.alphas_cumprod[
            prev_timestep] if prev_timestep >= 0 else self.final_alpha_cumprod
        beta_prod_t = 1 - alpha_prod_t
        beta_prod_t_prev = 1 - alpha_prod_t_prev

        variance = (beta_prod_t_prev /
                    beta_prod_t) * (1 - alpha_prod_t / alpha_prod_t_prev)

        return variance

    def set_timesteps(self, num_inference_steps, offset=0):
        self.num_inference_steps = num_inference_steps
        self.timesteps = np.arange(0, self.num_train_timesteps,
                                   self.num_train_timesteps // num_inference_steps)[::-1].copy()
        self.timesteps += offset

    def step(
        self,
        model_output: Union[torch.FloatTensor, np.ndarray],
        timestep: int,
        sample: Union[torch.FloatTensor, np.ndarray],
        eta: float = 1.0,
        use_clipped_model_output: bool = True,
        generator=None,
    ):
        # 1. get previous step value (=t-1)
        prev_timestep = timestep - self.num_train_timesteps // self.num_inference_steps

        # 2. compute alphas, betas
        alpha_prod_t = self.alphas_cumprod[timestep]
        alpha_prod_t_prev = self.alphas_cumprod[
            prev_timestep] if prev_timestep >= 0 else self.final_alpha_cumprod
        beta_prod_t = 1 - alpha_prod_t

        # 3. compute predicted original sample from predicted noise also called
        # "predicted x_0" of formula (12) from https://arxiv.org/pdf/2010.02502.pdf
        pred_original_sample = (sample - beta_prod_t**
                                (0.5) * model_output) / alpha_prod_t**(0.5)

        # 4. Clip "predicted x_0"
        if self.clip_sample:
            pred_original_sample = clip(pred_original_sample, -1, 1)

        # 5. compute variance: "sigma_t(η)" -> see formula (16)
        # σ_t = sqrt((1 − α_t−1)/(1 − α_t)) * sqrt(1 − α_t/α_t−1)
        variance = self._get_variance(timestep, prev_timestep)
        std_dev_t = eta * variance**(0.5)

        if use_clipped_model_output:
            # the model_output is always re-derived from the clipped x_0 in Glide
            model_output = (sample - alpha_prod_t**
                            (0.5) * pred_original_sample) / beta_prod_t**(0.5)
            #print("after",model_output)

        # 6. compute "direction pointing to x_t" of formula (12) from https://arxiv.org/pdf/2010.02502.pdf
        pred_sample_direction = (1 - alpha_prod_t_prev -
                                 std_dev_t**2)**(0.5) * model_output
        #print("direction", pred_sample_direction)

        # 7. compute x_t without "random noise" of formula (12) from https://arxiv.org/pdf/2010.02502.pdf
        prev_sample = alpha_prod_t_prev**(
            0.5) * pred_original_sample + pred_sample_direction
        #print("v",alpha_prod_t_prev**(0.5))
        if eta > 0:
            device = model_output.device if torch.is_tensor(
                model_output) else "cpu"
            noise = torch.randn(model_output.shape,
                                generator=generator).to(device)
            variance = self._get_variance(timestep,
                                          prev_timestep)**(0.5) * eta * noise

            if not torch.is_tensor(model_output):
                variance = variance.numpy()

            prev_sample = prev_sample + variance

        return prev_sample

    def add_noise(self, original_samples, noise, timesteps):
        timesteps = timesteps.cpu()
        sqrt_alpha_prod = self.alphas_cumprod[timesteps]**0.5
        sqrt_alpha_prod = match_shape(sqrt_alpha_prod, original_samples)
        sqrt_one_minus_alpha_prod = (1 - self.alphas_cumprod[timesteps])**0.5
        sqrt_one_minus_alpha_prod = match_shape(sqrt_one_minus_alpha_prod,
                                                original_samples)
        noisy_samples = sqrt_alpha_prod * original_samples + sqrt_one_minus_alpha_prod * noise
        return noisy_samples

    @torch.no_grad()
    def generate(self,
                 model,
                 cond,
                 ys,
                 generator=None,
                 eta=1.0,
                 use_clipped_model_output=False,
                 num_inference_steps=500,
                 device=None):
        if device is None:
            device = "cuda" if torch.cuda.is_available() else "cpu"

        # batch_size = cond.shape[0]
        # ys = torch.randn(
        #     (batch_size, model.out_dim),
        #     generator=generator,
        # )
        # ys = ys.to(device)

        self.set_timesteps(num_inference_steps)

        for t in self.timesteps:
            # 1. predict noise model_output
            model_output = model(cond, ys, t)

            # 2. predict previous mean of ys x_t-1 and add variance depending on eta
            # do x_t -> x_t-1
            ys = self.step(model_output,
                              t,
                              ys,
                              eta,
                              use_clipped_model_output=use_clipped_model_output)

        return ys

    def __len__(self):
        return self.num_train_timesteps