File size: 4,689 Bytes
f1069cc
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
import torch

from diffusers import KDPM2DiscreteScheduler
from diffusers.utils import torch_device

from .test_schedulers import SchedulerCommonTest


class KDPM2DiscreteSchedulerTest(SchedulerCommonTest):
    scheduler_classes = (KDPM2DiscreteScheduler,)
    num_inference_steps = 10

    def get_scheduler_config(self, **kwargs):
        config = {
            "num_train_timesteps": 1100,
            "beta_start": 0.0001,
            "beta_end": 0.02,
            "beta_schedule": "linear",
        }

        config.update(**kwargs)
        return config

    def test_timesteps(self):
        for timesteps in [10, 50, 100, 1000]:
            self.check_over_configs(num_train_timesteps=timesteps)

    def test_betas(self):
        for beta_start, beta_end in zip([0.00001, 0.0001, 0.001], [0.0002, 0.002, 0.02]):
            self.check_over_configs(beta_start=beta_start, beta_end=beta_end)

    def test_schedules(self):
        for schedule in ["linear", "scaled_linear"]:
            self.check_over_configs(beta_schedule=schedule)

    def test_prediction_type(self):
        for prediction_type in ["epsilon", "v_prediction"]:
            self.check_over_configs(prediction_type=prediction_type)

    def test_full_loop_with_v_prediction(self):
        scheduler_class = self.scheduler_classes[0]
        scheduler_config = self.get_scheduler_config(prediction_type="v_prediction")
        scheduler = scheduler_class(**scheduler_config)

        scheduler.set_timesteps(self.num_inference_steps)

        model = self.dummy_model()
        sample = self.dummy_sample_deter * scheduler.init_noise_sigma
        sample = sample.to(torch_device)

        for i, t in enumerate(scheduler.timesteps):
            sample = scheduler.scale_model_input(sample, t)

            model_output = model(sample, t)

            output = scheduler.step(model_output, t, sample)
            sample = output.prev_sample

        result_sum = torch.sum(torch.abs(sample))
        result_mean = torch.mean(torch.abs(sample))

        if torch_device in ["cpu", "mps"]:
            assert abs(result_sum.item() - 4.6934e-07) < 1e-2
            assert abs(result_mean.item() - 6.1112e-10) < 1e-3
        else:
            # CUDA
            assert abs(result_sum.item() - 4.693428650170972e-07) < 1e-2
            assert abs(result_mean.item() - 0.0002) < 1e-3

    def test_full_loop_no_noise(self):
        if torch_device == "mps":
            return
        scheduler_class = self.scheduler_classes[0]
        scheduler_config = self.get_scheduler_config()
        scheduler = scheduler_class(**scheduler_config)

        scheduler.set_timesteps(self.num_inference_steps)

        model = self.dummy_model()
        sample = self.dummy_sample_deter * scheduler.init_noise_sigma
        sample = sample.to(torch_device)

        for i, t in enumerate(scheduler.timesteps):
            sample = scheduler.scale_model_input(sample, t)

            model_output = model(sample, t)

            output = scheduler.step(model_output, t, sample)
            sample = output.prev_sample

        result_sum = torch.sum(torch.abs(sample))
        result_mean = torch.mean(torch.abs(sample))

        if torch_device in ["cpu", "mps"]:
            assert abs(result_sum.item() - 20.4125) < 1e-2
            assert abs(result_mean.item() - 0.0266) < 1e-3
        else:
            # CUDA
            assert abs(result_sum.item() - 20.4125) < 1e-2
            assert abs(result_mean.item() - 0.0266) < 1e-3

    def test_full_loop_device(self):
        if torch_device == "mps":
            return
        scheduler_class = self.scheduler_classes[0]
        scheduler_config = self.get_scheduler_config()
        scheduler = scheduler_class(**scheduler_config)

        scheduler.set_timesteps(self.num_inference_steps, device=torch_device)

        model = self.dummy_model()
        sample = self.dummy_sample_deter.to(torch_device) * scheduler.init_noise_sigma

        for t in scheduler.timesteps:
            sample = scheduler.scale_model_input(sample, t)

            model_output = model(sample, t)

            output = scheduler.step(model_output, t, sample)
            sample = output.prev_sample

        result_sum = torch.sum(torch.abs(sample))
        result_mean = torch.mean(torch.abs(sample))

        if str(torch_device).startswith("cpu"):
            # The following sum varies between 148 and 156 on mps. Why?
            assert abs(result_sum.item() - 20.4125) < 1e-2
            assert abs(result_mean.item() - 0.0266) < 1e-3
        else:
            # CUDA
            assert abs(result_sum.item() - 20.4125) < 1e-2
            assert abs(result_mean.item() - 0.0266) < 1e-3