-
Notifications
You must be signed in to change notification settings - Fork 126
/
ppo.py
executable file
·178 lines (145 loc) · 8.23 KB
/
ppo.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
# License: see [LICENSE, LICENSES/rsl_rl/LICENSE]
import torch
import torch.nn as nn
import torch.optim as optim
import torch.nn.functional as F
import numpy as np
from params_proto import PrefixProto
from go1_gym_learn.ppo import ActorCritic
from go1_gym_learn.ppo import RolloutStorage
from go1_gym_learn.ppo import caches
class PPO_Args(PrefixProto):
# algorithm
value_loss_coef = 1.0
use_clipped_value_loss = True
clip_param = 0.2
entropy_coef = 0.01
num_learning_epochs = 5
num_mini_batches = 4 # mini batch size = num_envs*nsteps / nminibatches
learning_rate = 1.e-3 # 5.e-4
adaptation_module_learning_rate = 1.e-3
num_adaptation_module_substeps = 1
schedule = 'adaptive' # could be adaptive, fixed
gamma = 0.99
lam = 0.95
desired_kl = 0.01
max_grad_norm = 1.
class PPO:
actor_critic: ActorCritic
def __init__(self, actor_critic, device='cpu'):
self.device = device
# PPO components
self.actor_critic = actor_critic
self.actor_critic.to(device)
self.storage = None # initialized later
self.optimizer = optim.Adam(self.actor_critic.parameters(), lr=PPO_Args.learning_rate)
self.adaptation_module_optimizer = optim.Adam(self.actor_critic.parameters(),
lr=PPO_Args.adaptation_module_learning_rate)
self.transition = RolloutStorage.Transition()
self.learning_rate = PPO_Args.learning_rate
def init_storage(self, num_envs, num_transitions_per_env, actor_obs_shape, privileged_obs_shape, obs_history_shape,
action_shape):
self.storage = RolloutStorage(num_envs, num_transitions_per_env, actor_obs_shape, privileged_obs_shape,
obs_history_shape, action_shape, self.device)
def test_mode(self):
self.actor_critic.test()
def train_mode(self):
self.actor_critic.train()
def act(self, obs, privileged_obs, obs_history):
# Compute the actions and values
self.transition.actions = self.actor_critic.act(obs, privileged_obs).detach()
self.transition.values = self.actor_critic.evaluate(obs, privileged_obs).detach()
self.transition.actions_log_prob = self.actor_critic.get_actions_log_prob(self.transition.actions).detach()
self.transition.action_mean = self.actor_critic.action_mean.detach()
self.transition.action_sigma = self.actor_critic.action_std.detach()
# need to record obs and critic_obs before env.step()
self.transition.observations = obs
self.transition.critic_observations = obs
self.transition.privileged_observations = privileged_obs
self.transition.observation_histories = obs_history
return self.transition.actions
def process_env_step(self, rewards, dones, infos):
self.transition.rewards = rewards.clone()
self.transition.dones = dones
self.transition.env_bins = infos["env_bins"]
# Bootstrapping on time outs
if 'time_outs' in infos:
self.transition.rewards += PPO_Args.gamma * torch.squeeze(
self.transition.values * infos['time_outs'].unsqueeze(1).to(self.device), 1)
# Record the transition
self.storage.add_transitions(self.transition)
self.transition.clear()
self.actor_critic.reset(dones)
def compute_returns(self, last_critic_obs, last_critic_privileged_obs):
last_values = self.actor_critic.evaluate(last_critic_obs, last_critic_privileged_obs).detach()
self.storage.compute_returns(last_values, PPO_Args.gamma, PPO_Args.lam)
def update(self):
mean_value_loss = 0
mean_surrogate_loss = 0
mean_adaptation_module_loss = 0
generator = self.storage.mini_batch_generator(PPO_Args.num_mini_batches, PPO_Args.num_learning_epochs)
for obs_batch, critic_obs_batch, privileged_obs_batch, obs_history_batch, actions_batch, target_values_batch, advantages_batch, returns_batch, old_actions_log_prob_batch, \
old_mu_batch, old_sigma_batch, masks_batch, env_bins_batch in generator:
self.actor_critic.act(obs_batch, privileged_obs_batch, masks=masks_batch)
actions_log_prob_batch = self.actor_critic.get_actions_log_prob(actions_batch)
value_batch = self.actor_critic.evaluate(critic_obs_batch, privileged_obs_batch, masks=masks_batch)
mu_batch = self.actor_critic.action_mean
sigma_batch = self.actor_critic.action_std
entropy_batch = self.actor_critic.entropy
# KL
if PPO_Args.desired_kl != None and PPO_Args.schedule == 'adaptive':
with torch.inference_mode():
kl = torch.sum(
torch.log(sigma_batch / old_sigma_batch + 1.e-5) + (
torch.square(old_sigma_batch) + torch.square(old_mu_batch - mu_batch)) / (
2.0 * torch.square(sigma_batch)) - 0.5, axis=-1)
kl_mean = torch.mean(kl)
if kl_mean > PPO_Args.desired_kl * 2.0:
self.learning_rate = max(1e-5, self.learning_rate / 1.5)
elif kl_mean < PPO_Args.desired_kl / 2.0 and kl_mean > 0.0:
self.learning_rate = min(1e-2, self.learning_rate * 1.5)
for param_group in self.optimizer.param_groups:
param_group['lr'] = self.learning_rate
# Surrogate loss
ratio = torch.exp(actions_log_prob_batch - torch.squeeze(old_actions_log_prob_batch))
surrogate = -torch.squeeze(advantages_batch) * ratio
surrogate_clipped = -torch.squeeze(advantages_batch) * torch.clamp(ratio, 1.0 - PPO_Args.clip_param,
1.0 + PPO_Args.clip_param)
surrogate_loss = torch.max(surrogate, surrogate_clipped).mean()
# Value function loss
if PPO_Args.use_clipped_value_loss:
value_clipped = target_values_batch + \
(value_batch - target_values_batch).clamp(-PPO_Args.clip_param,
PPO_Args.clip_param)
value_losses = (value_batch - returns_batch).pow(2)
value_losses_clipped = (value_clipped - returns_batch).pow(2)
value_loss = torch.max(value_losses, value_losses_clipped).mean()
else:
value_loss = (returns_batch - value_batch).pow(2).mean()
loss = surrogate_loss + PPO_Args.value_loss_coef * value_loss - PPO_Args.entropy_coef * entropy_batch.mean()
# Gradient step
self.optimizer.zero_grad()
loss.backward()
nn.utils.clip_grad_norm_(self.actor_critic.parameters(), PPO_Args.max_grad_norm)
self.optimizer.step()
mean_value_loss += value_loss.item()
mean_surrogate_loss += surrogate_loss.item()
# Adaptation module gradient step
for epoch in range(PPO_Args.num_adaptation_module_substeps):
adaptation_pred = self.actor_critic.adaptation_module(obs_history_batch)
with torch.no_grad():
adaptation_target = self.actor_critic.env_factor_encoder(privileged_obs_batch)
residual = (adaptation_target - adaptation_pred).norm(dim=1)
caches.slot_cache.log(env_bins_batch[:, 0].cpu().numpy().astype(np.uint8),
sysid_residual=residual.cpu().numpy())
adaptation_loss = F.mse_loss(adaptation_pred, adaptation_target)
self.adaptation_module_optimizer.zero_grad()
adaptation_loss.backward()
self.adaptation_module_optimizer.step()
mean_adaptation_module_loss += adaptation_loss.item()
num_updates = PPO_Args.num_learning_epochs * PPO_Args.num_mini_batches
mean_value_loss /= num_updates
mean_surrogate_loss /= num_updates
mean_adaptation_module_loss /= (num_updates * PPO_Args.num_adaptation_module_substeps)
self.storage.clear()
return mean_value_loss, mean_surrogate_loss, mean_adaptation_module_loss