-
Notifications
You must be signed in to change notification settings - Fork 0
/
discriminator_pytorch.py
253 lines (203 loc) · 9.14 KB
/
discriminator_pytorch.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
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
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
# import d4rl
# import gym
import numpy as np
import pickle
import torch
import torch.nn as nn
import torch.nn.functional as F
import torch.utils.data
from torch import autograd, device
from tqdm import tqdm
# https://github.com/ikostrikov/pytorch-a2c-ppo-acktr-gail/blob/master/a2c_ppo_acktr/algo/gail.py
class Discriminator(nn.Module):
def __init__(self, input_dim, hidden_dim=256, device='cuda:0'):
super(Discriminator, self).__init__()
self.device = device
self.trunk = nn.Sequential(
nn.Linear(input_dim, hidden_dim), nn.Tanh(),
nn.Linear(hidden_dim, hidden_dim), nn.Tanh(),
nn.Linear(hidden_dim, 1)).to(device)
self.trunk.train()
self.optimizer = torch.optim.Adam(self.trunk.parameters(), lr=3e-4)
def compute_grad_pen(self,
expert_state,
offline_state,
lambda_=10):
alpha = torch.rand(expert_state.size(0), 1)
expert_data = expert_state
offline_data = offline_state
alpha = alpha.expand_as(expert_data).to(expert_data.device)
mixup_data = alpha * expert_data + (1 - alpha) * offline_data
mixup_data.requires_grad = True
disc = self.trunk(mixup_data)
ones = torch.ones(disc.size()).to(disc.device)
grad = autograd.grad(
outputs=disc,
inputs=mixup_data,
grad_outputs=ones,
create_graph=True,
retain_graph=True,
only_inputs=True)[0]
grad_pen = lambda_ * (grad.norm(2, dim=1) - 1).pow(2).mean()
return grad_pen
def update(self, expert_loader, offline_loader):
self.train()
loss = 0
n = 0
for expert_state, offline_state in zip(expert_loader, offline_loader):
expert_state = expert_state[0].to(self.device)
offline_state = offline_state[0][:expert_state.shape[0]].to(self.device)
policy_d = self.trunk(offline_state)
expert_d = self.trunk(expert_state)
expert_loss = F.binary_cross_entropy_with_logits(
expert_d,
torch.ones(expert_d.size()).to(self.device))
policy_loss = F.binary_cross_entropy_with_logits(
policy_d,
torch.zeros(policy_d.size()).to(self.device))
gail_loss = expert_loss + policy_loss
grad_pen = self.compute_grad_pen(expert_state, offline_state)
loss += (gail_loss + grad_pen).item()
n += 1
self.optimizer.zero_grad()
(gail_loss + grad_pen).backward()
self.optimizer.step()
return loss / n
def predict_reward(self, state):
with torch.no_grad():
self.eval()
d = self.trunk(state)
s = torch.sigmoid(d)
# log(d^E/d^O)
# reward = - (1/s-1).log()
reward = s.log() - (1 - s).log()
return reward
class Discriminator_SAS(nn.Module):
def __init__(self, state_dim, action_dim, hidden_dim=256, device='cuda:0'):
super(Discriminator_SAS, self).__init__()
self.device = device
self.state_dim = state_dim
self.action_dim = action_dim
state_hidden_dim = hidden_dim if action_dim == 0 else int(hidden_dim//2) #Changing to hidden_dim/3 since we have SAS
self.state_trunk = nn.Sequential(
nn.Linear(state_dim, state_hidden_dim), nn.Tanh()).to(device)
action_trunk_input_dim = 1 if action_dim == 0 else action_dim
self.action_trunk = nn.Sequential(
nn.Linear(action_trunk_input_dim, int(hidden_dim//2)), nn.Tanh()).to(device) #Changing to hidden_dim/3 since we have SAS
self.next_state_trunk = nn.Sequential(
nn.Linear(state_dim, state_hidden_dim), nn.Tanh()).to(device) #Adding trunk for next state
self.trunk = nn.Sequential(
nn.Linear(3*(hidden_dim//2), hidden_dim), nn.Tanh(), #Taking care of floats div/3
nn.Linear(hidden_dim, 1)).to(device)
self.state_trunk.train()
self.action_trunk.train()
self.trunk.train()
self.optimizer = torch.optim.Adam(self.trunk.parameters(), lr=3e-4)
def forward(self, input):
if input.shape[1] == self.state_dim:
h = self.state_trunk(input)
h = self.trunk(h)
else:
#Separate state, action, next_state
state = input[:, :self.state_dim]
action = input[:, self.state_dim:self.state_dim+self.action_dim] #Since next state is also included
next_state = input[:, self.state_dim+self.action_dim:]
h_state = self.state_trunk(state)
h_action = self.action_trunk(action)
h_next_state = self.next_state_trunk(next_state)
h = torch.cat([h_state, h_action,h_next_state], axis=1)
h = self.trunk(h)
return h
def compute_grad_pen(self,
expert_state,
offline_state,
lambda_=10):
alpha = torch.rand(expert_state.size(0), 1)
expert_data = expert_state
offline_data = offline_state
alpha = alpha.expand_as(expert_data).to(expert_data.device)
mixup_data = alpha * expert_data + (1 - alpha) * offline_data
mixup_data.requires_grad = True
disc = self(mixup_data)
ones = torch.ones(disc.size()).to(disc.device)
grad = autograd.grad(
outputs=disc,
inputs=mixup_data,
grad_outputs=ones,
create_graph=True,
retain_graph=True,
only_inputs=True)[0]
grad_pen = lambda_ * (grad.norm(2, dim=1) - 1).pow(2).mean()
return grad_pen
def update(self, expert_loader, offline_loader):
self.train()
loss = 0
n = 0
for expert_state, offline_state in zip(expert_loader, offline_loader):
expert_state = expert_state[0].to(self.device)
offline_state = offline_state[0][:expert_state.shape[0]].to(self.device)
# '''
# Adding noise like MNM
# '''
# noise_expert = torch.normal(mean=torch.zeros_like(expert_state),std=torch.tensor(0.1))
# noise_offline = torch.normal(mean=torch.zeros_like(offline_state),std=torch.tensor(0.1))
# expert_state += expert_state + noise_expert.cuda()
# offline_state += offline_state + noise_offline.cuda()
# del noise_expert,noise_offline
policy_d = self(offline_state)
expert_d = self(expert_state)
expert_loss = F.binary_cross_entropy_with_logits(
expert_d,
torch.ones(expert_d.size()).to(self.device))
policy_loss = F.binary_cross_entropy_with_logits(
policy_d,
0.1*torch.ones(policy_d.size()).to(self.device))
# del expert_d,policy_d
gail_loss = expert_loss + policy_loss
# del expert_loss,policy_loss
grad_pen = self.compute_grad_pen(expert_state, offline_state)
loss += (gail_loss + grad_pen).item()
n += 1
self.optimizer.zero_grad()
(gail_loss + grad_pen).backward()
self.optimizer.step()
del policy_d,expert_d,expert_state,offline_state
return loss / n
def update2(self, expert_loader, offline_loader):
self.train()
loss = 0
n = 0
for offline_state in offline_loader:
for expert_state in expert_loader:
batch_size = min(offline_state[0].shape[0], expert_state[0].shape[0])
offline_state = offline_state[0][:batch_size].to(self.device)
expert_state = expert_state[0][:batch_size].to(self.device)
policy_d = self(offline_state)
expert_d = self(expert_state)
expert_loss = F.binary_cross_entropy_with_logits(
expert_d,
torch.ones(expert_d.size()).to(self.device))
policy_loss = F.binary_cross_entropy_with_logits(
policy_d,
0.1*torch.ones(policy_d.size()).to(self.device))
gail_loss = expert_loss + policy_loss
grad_pen = self.compute_grad_pen(expert_state, offline_state)
loss += (gail_loss + grad_pen).item()
n += 1
self.optimizer.zero_grad()
(gail_loss + grad_pen).backward()
self.optimizer.step()
# force just once in the inner-loop
break
return loss / n
def predict_reward(self, state):
with torch.no_grad():
self.eval()
state = state.to(self.device)
d = self(state)
s = torch.sigmoid(d)
# log(d^E/d^O)
# reward = - (1/s-1).log()
reward = s.log() - (1 - s).log()
# reward = -(1 - s).log()*10
return reward