forked from leggedrobotics/MPC-Net
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathPolicyNet.py
103 lines (77 loc) · 3.73 KB
/
PolicyNet.py
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
import torch
import numpy as np
class LinearPolicy(torch.nn.Module):
def __init__(self, d_in, d_out):
super(LinearPolicy, self).__init__()
self.d_out = d_out
self.linear = torch.nn.Linear(d_in, d_out)
def forward(self, tx):
u = self.linear(tx).reshape((1, self.d_out))
return torch.ones((1, 1)), u
def logParameters(self, writer, it):
for param in list(self.named_parameters(prefix='LinearPolicy', recurse=True)):
for scalar_it in range(len(param[1].data.view(-1))):
writer.add_scalar(param[0] + "/" + str(scalar_it), param[1].data.view(-1)[scalar_it].item(), it)
class NonlinearPolicy(torch.nn.Module):
def __init__(self, d_in, d_out):
super(NonlinearPolicy, self).__init__()
self.d_out = d_out
self.n_hidden = d_in * 2 * 2
self.linear1 = torch.nn.Linear(d_in, self.n_hidden)
self.activation1 = torch.tanh
self.linear2 = torch.nn.Linear(self.n_hidden, self.n_hidden)
self.activation2 = torch.tanh
self.linear3 = torch.nn.Linear(self.n_hidden, self.d_out)
def forward(self, tx):
z_h1 = self.activation1(self.linear1(tx))
u = self.linear3(z_h1).reshape((1, self.d_out))
return torch.ones((1, 1)), u
def logParameters(self, writer, it):
for param in list(self.named_parameters(prefix='NonlinearPolicy', recurse=True)):
for scalar_it in range(len(param[1].data.view(-1))):
writer.add_scalar(param[0] + "/" + str(scalar_it), param[1].data.view(-1)[scalar_it].item(), it)
class TwoLayerNLP(torch.nn.Module):
def __init__(self, d_in, d_out):
super(TwoLayerNLP, self).__init__()
self.d_out = d_out
self.n_hidden = 128
self.linear1 = torch.nn.Linear(d_in, self.n_hidden)
self.activation1 = torch.tanh
self.linear2 = torch.nn.Linear(self.n_hidden, self.n_hidden)
self.activation2 = torch.tanh
self.linear3 = torch.nn.Linear(self.n_hidden, self.d_out)
def forward(self, tx):
z_h1 = self.activation1(self.linear1(tx))
z_h2 = self.activation2(self.linear2(z_h1))
u = self.linear3(z_h2).reshape((1, self.d_out))
return torch.ones((1, 1)), u
def logParameters(self, writer, it):
for param in list(self.named_parameters(prefix='TwoLayerNLP', recurse=True)):
for scalar_it in range(len(param[1].data.view(-1))):
writer.add_scalar(param[0] + "/" + str(scalar_it), param[1].data.view(-1)[scalar_it].item(), it)
class ExpertMixturePolicy(torch.nn.Module):
def __init__(self, d_in, d_out):
super(ExpertMixturePolicy, self).__init__()
self.num_experts = 8
self.n_hidden = d_in * 4
self.d_out = d_out
self.linear1 = torch.nn.Linear(d_in, self.n_hidden)
self.activation1 = torch.tanh
self.selector_net = torch.nn.Sequential(
torch.nn.Linear(self.n_hidden, self.num_experts),
# torch.nn.Softmax(dim=-1)
torch.nn.Sigmoid()
)
self.expert_net = torch.nn.Sequential(
torch.nn.Linear(self.n_hidden, d_out*self.num_experts)
)
def forward(self, tx):
z_h = self.activation1(self.linear1(tx))
pi_nonNormalized = self.selector_net(z_h)
pi = pi_nonNormalized / pi_nonNormalized.sum()
u_experts = self.expert_net(z_h).reshape((self.num_experts, self.d_out))
return pi, u_experts
def logParameters(self, writer, it):
for param in list(self.named_parameters(prefix='ExpertMixPolicy', recurse=True)):
for scalar_it in range(len(param[1].data.view(-1))):
writer.add_scalar(param[0]+"/"+str(scalar_it), param[1].data.view(-1)[scalar_it].item(), it)