-
Notifications
You must be signed in to change notification settings - Fork 0
/
agent.py
258 lines (209 loc) · 11.2 KB
/
agent.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
254
255
256
257
258
# -*- coding: utf-8 -*-
# MIT License
#
# Copyright (c) 2017 Kai Arulkumaran
#
# Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:
#
# The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.
# ==============================================================================
from __future__ import division
import os
import numpy as np
import torch
from torch import optim
from torch.nn.utils import clip_grad_norm_
# import kornia.augmentation as aug
import torchvision.transforms.transforms as transforms
import torch
import torch.nn as nn
from model import DQN, AuxDQN
from torch.nn import functional as F
# random_shift = nn.Sequential(aug.RandomCrop((80, 80)), nn.ReplicationPad2d(4), aug.RandomCrop((84, 84)))
# aug = random_shift
augs = torch.nn.Sequential(
transforms.RandomCrop(80),
nn.ReplicationPad2d(4),
transforms.RandomCrop(84),
)
aug = torch.jit.script(augs)
class Agent():
def __init__(self, args, env, weights):
self.args = args
self.weights = weights
self.action_space = env.action_space()
self.atoms = args.atoms
self.Vmin = args.V_min
self.Vmax = args.V_max
self.support = torch.linspace(args.V_min, args.V_max, self.atoms).to(device=args.device) # Support (range) of z
self.delta_z = (args.V_max - args.V_min) / (self.atoms - 1)
self.batch_size = args.batch_size
self.n = args.multi_step
self.discount = args.discount
self.norm_clip = args.norm_clip
self.coeff = 0.01 if args.game in ['pong', 'boxing', 'private_eye', 'freeway'] else 1.
self.online_net = AuxDQN(args, self.action_space).to(device=args.device)
self.momentum_net = AuxDQN(args, self.action_space).to(device=args.device)
if args.model: # Load pretrained model if provided
if os.path.isfile(args.model):
state_dict = torch.load(args.model, map_location='cpu') # Always load tensors onto CPU by default, will shift to GPU if necessary
if 'conv1.weight' in state_dict.keys():
for old_key, new_key in (('conv1.weight', 'convs.0.weight'), ('conv1.bias', 'convs.0.bias'), ('conv2.weight', 'convs.2.weight'), ('conv2.bias', 'convs.2.bias'), ('conv3.weight', 'convs.4.weight'), ('conv3.bias', 'convs.4.bias')):
state_dict[new_key] = state_dict[old_key] # Re-map state dict for old pretrained models
del state_dict[old_key] # Delete old keys for strict load_state_dict
self.online_net.load_state_dict(state_dict)
print("Loading pretrained model: " + args.model)
else: # Raise error if incorrect model path provided
raise FileNotFoundError(args.model)
self.online_net.train()
self.initialize_momentum_net()
self.momentum_net.train()
self.target_net = AuxDQN(args, self.action_space).to(device=args.device)
self.update_target_net()
self.target_net.train()
for param in self.target_net.parameters():
param.requires_grad = False
for param in self.momentum_net.parameters():
param.requires_grad = False
self.optimiser = optim.Adam(self.online_net.parameters(), lr=args.learning_rate, eps=args.adam_eps)
self.mse_loss = nn.MSELoss()
self.ce_loss = nn.CrossEntropyLoss()
# Resets noisy weights in all linear layers (of online net only)
def reset_noise(self):
self.online_net.reset_noise()
# Acts based on single state (no batch)
def act(self, state):
with torch.no_grad():
a = self.online_net(state.unsqueeze(0))
return (a * self.support).sum(2).argmax(1).item()
# Acts with an ε-greedy policy (used for evaluation only)
def act_e_greedy(self, state, epsilon=0.001): # High ε can reduce evaluation scores drastically
return np.random.randint(0, self.action_space) if np.random.random() < epsilon else self.act(state)
def axu_loss_forward(self, obs, next_obs, action, reward, obs_aug, weights):
assert obs.shape == next_obs.shape
assert len(weights) == 5
obs_h = self.online_net.convs(obs)
h_shape = obs_h.shape
obs_h = obs_h.view(-1, self.online_net.conv_output_size)
next_obs_h = self.online_net.convs(next_obs)
next_obs_h = next_obs_h.view(-1, self.online_net.conv_output_size)
loss = 0
if weights[0] > 0:
# predict future
action = F.one_hot(action, num_classes=self.action_space).float()
act_h = self.online_net.act(action)
pre_t = torch.cat([obs_h, act_h], dim=1)
# print(self.online_net.pred_future(pre_t).shape, next_obs_h.detach().shape)
pred_future_loss = self.mse_loss(self.online_net.pred_future(pre_t), next_obs_h.detach())
# print(pred_future_loss)
loss += weights[0] * pred_future_loss
if weights[1] > 0:
# extract reward
pre_t = torch.cat([obs_h, next_obs_h], dim=1)
extract_reward_loss = self.mse_loss(self.online_net.extract_reward(pre_t), reward.reshape(-1, 1))
loss += weights[1] * extract_reward_loss
# print(extract_reward_loss)
if weights[2] > 0:
# BYOL
obs_h_2 = self.online_net.convs(obs_aug)
obs_h_2 = obs_h_2.view(-1, self.online_net.conv_output_size)
p1, p2 = self.online_net.predict(obs_h), self.online_net.predict(obs_h_2)
t1, t2 = self.momentum_net.project(obs_h), self.momentum_net.project(obs_h_2)
def loss_fn(x, y):
x = F.normalize(x, dim=-1, p=2)
y = F.normalize(y, dim=-1, p=2)
return 2 - 2 * (x * y).sum(dim=-1)
byol_loss = (0.5 * loss_fn(p1, t2) + 0.5 * loss_fn(p2, t1))
byol_loss = byol_loss.mean()
# print(byol_loss)
loss += weights[2] * byol_loss
if weights[3] > 0:
# AE
h = self.online_net.hnet(obs_h)
r = h.reshape(h_shape)
rec = self.online_net.decoder(r)
r_shape = rec.shape
rec_loss = self.mse_loss(rec, obs[:, :, :r_shape[-2], :r_shape[-1]])
latent_loss = (0.5 * h.pow(2).sum(1)).mean()
ae_loss = rec_loss + latent_loss * (1e-6) # latent lambda
# print(ae_loss)
loss += weights[3] * ae_loss
if weights[4] > 0:
# Rotation
b = obs.size(0)
labels = torch.arange(4, dtype=torch.long, device=obs.device).repeat_interleave(b)
obs_cat = obs.repeat(4, 1, 1, 1)
for i in range(4):
obs_cat[i * b: (i + 1) * b] = torch.rot90(obs_cat[i * b:(i + 1) * b], i, [-2, -1])
obs_cat = self.online_net.convs(obs_cat)
rot_loss = self.ce_loss(self.online_net.rot_cls(obs_cat), labels)
# print(rot_loss)
loss += weights[4] * rot_loss
return loss
def learn(self, mem):
# Sample transitions
idxs, states, actions, returns, next_states, nonterminals, weights = mem.sample(self.batch_size)
# print(states.shape)
states = states.to(device=self.args.device)
next_states = next_states.to(device=self.args.device)
# augmentation for extract ar
aug_states = aug(states).to(device=self.args.device)
aug_next_states = aug(next_states).to(device=self.args.device)
aug_states_2 = aug(states).to(device=self.args.device)
aux_loss = self.axu_loss_forward(aug_states, aug_next_states, actions, returns, aug_states_2,self.weights)
# print(logits.shape, actions.shape, returns.shape)
# Calculate current state probabilities (online network noise already sampled)
log_ps = self.online_net(states, log=True) # Log probabilities log p(s_t, ·; θonline)
log_ps_a = log_ps[range(self.batch_size), actions] # log p(s_t, a_t; θonline)
with torch.no_grad():
# Calculate nth next state probabilities
pns = self.online_net(next_states) # Probabilities p(s_t+n, ·; θonline)
dns = self.support.expand_as(pns) * pns # Distribution d_t+n = (z, p(s_t+n, ·; θonline))
argmax_indices_ns = dns.sum(2).argmax(1) # Perform argmax action selection using online network: argmax_a[(z, p(s_t+n, a; θonline))]
self.target_net.reset_noise() # Sample new target net noise
pns = self.target_net(next_states) # Probabilities p(s_t+n, ·; θtarget)
pns_a = pns[range(self.batch_size), argmax_indices_ns] # Double-Q probabilities p(s_t+n, argmax_a[(z, p(s_t+n, a; θonline))]; θtarget)
# Compute Tz (Bellman operator T applied to z)
Tz = returns.unsqueeze(1) + nonterminals * (self.discount ** self.n) * self.support.unsqueeze(0) # Tz = R^n + (γ^n)z (accounting for terminal states)
Tz = Tz.clamp(min=self.Vmin, max=self.Vmax) # Clamp between supported values
# Compute L2 projection of Tz onto fixed support z
b = (Tz - self.Vmin) / self.delta_z # b = (Tz - Vmin) / Δz
l, u = b.floor().to(torch.int64), b.ceil().to(torch.int64)
# Fix disappearing probability mass when l = b = u (b is int)
l[(u > 0) * (l == u)] -= 1
u[(l < (self.atoms - 1)) * (l == u)] += 1
# Distribute probability of Tz
m = states.new_zeros(self.batch_size, self.atoms)
offset = torch.linspace(0, ((self.batch_size - 1) * self.atoms), self.batch_size).unsqueeze(1).expand(self.batch_size, self.atoms).to(actions)
m.view(-1).index_add_(0, (l + offset).view(-1), (pns_a * (u.float() - b)).view(-1)) # m_l = m_l + p(s_t+n, a*)(u - b)
m.view(-1).index_add_(0, (u + offset).view(-1), (pns_a * (b - l.float())).view(-1)) # m_u = m_u + p(s_t+n, a*)(b - l)
loss = -torch.sum(m * log_ps_a, 1) # Cross-entropy loss (minimises DKL(m||p(s_t, a_t)))
aux_loss = loss + (aux_loss * self.coeff)
self.online_net.zero_grad()
(weights * aux_loss).mean().backward() # Backpropagate importance-weighted minibatch loss
clip_grad_norm_(self.online_net.parameters(), self.norm_clip) # Clip gradients by L2 norm
self.optimiser.step()
mem.update_priorities(idxs, loss.detach().cpu().numpy()) # Update priorities of sampled transitions
def update_target_net(self):
self.target_net.load_state_dict(self.online_net.state_dict())
def initialize_momentum_net(self):
for param_q, param_k in zip(self.online_net.parameters(), self.momentum_net.parameters()):
param_k.data.copy_(param_q.data) # update
param_k.requires_grad = False # not update by gradient
# Code for this function from https://github.com/facebookresearch/moco
@torch.no_grad()
def update_momentum_net(self, momentum=0.999):
for param_q, param_k in zip(self.online_net.parameters(), self.momentum_net.parameters()):
param_k.data.copy_(momentum * param_k.data + (1.- momentum) * param_q.data) # update
# Save model parameters on current device (don't move model between devices)
def save(self, path, name='model.pth'):
torch.save(self.online_net.state_dict(), os.path.join(path, name))
# Evaluates Q-value based on single state (no batch)
def evaluate_q(self, state):
with torch.no_grad():
a = self.online_net(state.unsqueeze(0))
return (a * self.support).sum(2).max(1)[0].item()
def train(self):
self.online_net.train()
def eval(self):
self.online_net.eval()