File size: 5,134 Bytes
079c32c
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
import time
import torch
from hpc_rll.origin.td import q_nstep_td_error_with_rescale, q_nstep_td_data
from hpc_rll.rl_utils.td import QNStepTDRescale
from testbase import mean_relative_error, times

assert torch.cuda.is_available()
use_cuda = True

T = 1024
B = 64
N = 64
gamma = 0.95


def qntd_rescale_val():
    ori_q = torch.randn(B, N)
    ori_next_n_q = torch.randn(B, N)
    ori_action = torch.randint(0, N, size=(B, ))
    ori_next_n_action = torch.randint(0, N, size=(B, ))
    ori_reward = torch.randn(T, B)
    ori_done = torch.randn(B)
    ori_weight = torch.randn(B)

    hpc_q = ori_q.clone().detach()
    hpc_next_n_q = ori_next_n_q.clone().detach()
    hpc_action = ori_action.clone().detach()
    hpc_next_n_action = ori_next_n_action.clone().detach()
    hpc_reward = ori_reward.clone().detach()
    hpc_done = ori_done.clone().detach()
    hpc_weight = ori_weight.clone().detach()
    hpc_qntd_rescale = QNStepTDRescale(T, B, N)

    if use_cuda:
        ori_q = ori_q.cuda()
        ori_next_n_q = ori_next_n_q.cuda()
        ori_action = ori_action.cuda()
        ori_next_n_action = ori_next_n_action.cuda()
        ori_reward = ori_reward.cuda()
        ori_done = ori_done.cuda()
        ori_weight = ori_weight.cuda()

        hpc_q = hpc_q.cuda()
        hpc_next_n_q = hpc_next_n_q.cuda()
        hpc_action = hpc_action.cuda()
        hpc_next_n_action = hpc_next_n_action.cuda()
        hpc_reward = hpc_reward.cuda()
        hpc_done = hpc_done.cuda()
        hpc_weight = hpc_weight.cuda()
        hpc_qntd_rescale = hpc_qntd_rescale.cuda()

    ori_q.requires_grad_(True)
    ori_loss, _ = q_nstep_td_error_with_rescale(
        q_nstep_td_data(ori_q, ori_next_n_q, ori_action, ori_next_n_action, ori_reward, ori_done, ori_weight), gamma, T
    )
    ori_loss = ori_loss.mean()
    ori_loss.backward()
    if use_cuda:
        torch.cuda.synchronize()

    hpc_q.requires_grad_(True)
    hpc_loss, _ = hpc_qntd_rescale(
        hpc_q, hpc_next_n_q, hpc_action, hpc_next_n_action, hpc_reward, hpc_done, hpc_weight, gamma
    )
    hpc_loss = hpc_loss.mean()
    hpc_loss.backward()
    if use_cuda:
        torch.cuda.synchronize()

    mre = mean_relative_error(
        torch.flatten(ori_loss).cpu().detach().numpy(),
        torch.flatten(hpc_loss).cpu().detach().numpy()
    )
    print("qntd rescale fp mean_relative_error: " + str(mre))
    mre = mean_relative_error(
        torch.flatten(ori_q.grad).cpu().detach().numpy(),
        torch.flatten(hpc_q.grad).cpu().detach().numpy()
    )
    print("qntd rescale bp mean_relative_error: " + str(mre))


def qntd_rescale_perf():
    ori_q = torch.randn(B, N)
    ori_next_n_q = torch.randn(B, N)
    ori_action = torch.randint(0, N, size=(B, ))
    ori_next_n_action = torch.randint(0, N, size=(B, ))
    ori_reward = torch.randn(T, B)
    ori_done = torch.randn(B)
    ori_weight = torch.randn(B)

    hpc_q = ori_q.clone().detach()
    hpc_next_n_q = ori_next_n_q.clone().detach()
    hpc_action = ori_action.clone().detach()
    hpc_next_n_action = ori_next_n_action.clone().detach()
    hpc_reward = ori_reward.clone().detach()
    hpc_done = ori_done.clone().detach()
    hpc_weight = ori_weight.clone().detach()
    hpc_qntd_rescale = QNStepTDRescale(T, B, N)

    if use_cuda:
        ori_q = ori_q.cuda()
        ori_next_n_q = ori_next_n_q.cuda()
        ori_action = ori_action.cuda()
        ori_next_n_action = ori_next_n_action.cuda()
        ori_reward = ori_reward.cuda()
        ori_done = ori_done.cuda()
        ori_weight = ori_weight.cuda()

        hpc_q = hpc_q.cuda()
        hpc_next_n_q = hpc_next_n_q.cuda()
        hpc_action = hpc_action.cuda()
        hpc_next_n_action = hpc_next_n_action.cuda()
        hpc_reward = hpc_reward.cuda()
        hpc_done = hpc_done.cuda()
        hpc_weight = hpc_weight.cuda()
        hpc_qntd_rescale = hpc_qntd_rescale.cuda()

    ori_q.requires_grad_(True)
    for i in range(times):
        t = time.time()
        ori_loss, _ = q_nstep_td_error_with_rescale(
            q_nstep_td_data(ori_q, ori_next_n_q, ori_action, ori_next_n_action, ori_reward, ori_done, ori_weight),
            gamma, T
        )
        ori_loss = ori_loss.mean()
        ori_loss.backward()
        if use_cuda:
            torch.cuda.synchronize()
        print('epoch: {}, original qntd rescale cost time: {}'.format(i, time.time() - t))

    hpc_q.requires_grad_(True)
    for i in range(times):
        t = time.time()
        hpc_loss, _ = hpc_qntd_rescale(
            hpc_q, hpc_next_n_q, hpc_action, hpc_next_n_action, hpc_reward, hpc_done, hpc_weight, gamma
        )
        hpc_loss = hpc_loss.mean()
        hpc_loss.backward()
        if use_cuda:
            torch.cuda.synchronize()
        print('epoch: {}, hpc qntd rescale cost time: {}'.format(i, time.time() - t))


if __name__ == '__main__':
    print("target problem: T = {}, B = {}, N = {}, gamma = {}".format(T, B, N, gamma))
    print("================run qntd rescale validation test================")
    qntd_rescale_val()
    print("================run qntd rescale performance test================")
    qntd_rescale_perf()