-
Notifications
You must be signed in to change notification settings - Fork 3
/
train.py
executable file
·215 lines (173 loc) · 9.15 KB
/
train.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
import argparse
import os
import random
import gym
from gym.wrappers import TimeLimit
import numpy as np
import pandas as pd
import yaml
from env_aug import AntEnvAug, HalfCheetahEnvAug, HopperEnvAug, fixedSwimmerEnv
from model import EnsembleGymEnv
from ppo import PPO, Memory
from train_funcs import (collect_data, test_agent, train_agent,
train_agent_model_free, train_agent_model_free_debug)
from online_learning import ExpWeights
from utils import MeanStdevFilter, reward_func, weights_init
os.environ['KMP_DUPLICATE_LIB_OK']='True'
### Main PPO Loop
def train_ppo(params):
## random rollouts
params['zeros'] = False
b = ExpWeights()
if params['env_name'] == 'HalfCheetah-v2':
env = HalfCheetahEnvAug()
elif params['env_name'] == 'Ant-v2':
env = AntEnvAug()
elif params['env_name'] == 'Swimmer-v2':
env = fixedSwimmerEnv()
elif params['env_name'] == 'Hopper-v2':
env = HopperEnvAug()
else:
raise Exception('Environment not supported')
env = TimeLimit(env, params['steps'])
params['ob_dim'] = env.observation_space.shape[0]
params['ac_dim'] = env.action_space.shape[0]
params['is_done_func'] = None
if hasattr(env, 'is_done_func'):
params['is_done_func'] = env.is_done_func
env = EnsembleGymEnv(params, env)
# TODO: put these into argparse/separate yaml files
############## Hyperparameters ##############
log_interval = 100 # print avg reward in the interval
policy_iters = params['policy_iters'] # max training episodes
ep_steps = params['steps'] # max timesteps in one episode
update_timestep = params['update_timestep'] # update policy every n timesteps
action_std = 0.5 # constant std for action distribution (Multivariate Normal)
K_epochs = 10 # update policy for K epochs
eps_clip = 0.2 # clip parameter for PPO
gamma = 0.99 # discount factor
lr = 0.0003 # parameters for Adam optimizer
betas = (0.9, 0.999)
#############################################
state_dim = env.observation_space.shape[0]
action_dim = env.action_space.shape[0]
n_parallel = int(update_timestep / ep_steps)
env_resets = []
env_resets_real = []
env.real_env.seed(params['seed'])
np.random.seed(params['seed'])
random.seed(params['seed'])
for _ in range(n_parallel):
env_resets.append(env.real_env.reset())
env_resets_real.append(env.real_env.unwrapped.state_vector())
env_resets = np.array(env_resets)
env_resets_real = np.array(env_resets_real)
memory = Memory()
ppo = PPO(params['seed'], state_dim, action_dim, action_std, lr, betas, gamma, K_epochs, eps_clip)
if params['var_max']:
params['filename'] += '-var-max'
ppo_rew = PPO(params['seed'], state_dim, action_dim, action_std, lr, betas, gamma, K_epochs, eps_clip)
if params['model_free']:
train_agent_model_free(ppo, env, memory, update_timestep, params['seed'], 500, ep_steps, env_resets, env_resets_real)
return
iterations = 0
print("\nCollecting random rollouts...")
total_timesteps = 0
timesteps, error = collect_data(params, ppo, memory, env)
total_timesteps += timesteps
samples = [total_timesteps]
rewards, rewards_m, lambdas, errors = [], [], [], []
b.error_buffer.append(error) # initial baseline
if params['adapt']:
params['lam'] = 0
while total_timesteps < params['max_timesteps']:
## train model
print("\nTraining Model...")
env.train_model(max_epochs=params['model_epochs'])
## train policy in model
print("\nTraining PPO Policy in Model...")
ppo.change_policy_std(action_std)
if params['var_max']:
train_agent(ppo, env, policy_iters, ep_steps, memory, update_timestep, env_resets, 500, 1, n_parallel, params['var_type'])
train_agent(ppo_rew, env, policy_iters, ep_steps, memory, update_timestep, env_resets, 500, 0, n_parallel, params['var_type'])
else:
train_agent(ppo, env, policy_iters, ep_steps, memory, update_timestep, env_resets, 500, params['lam'], n_parallel, params['var_type'])
## test policy in the env
subset_resets_idx = np.random.randint(0, n_parallel, 10)
subset_resets = env_resets[subset_resets_idx]
subset_resets_real = env_resets_real[subset_resets_idx]
if params['var_max']:
reward_model = test_agent(ppo_rew, env, memory, ep_steps, subset_resets, subset_resets_real, use_model=True)
reward_actual = test_agent(ppo_rew, env, memory, ep_steps, subset_resets, subset_resets_real, use_model=False)
else:
reward_model = test_agent(ppo, env, memory, ep_steps, subset_resets, subset_resets_real, use_model=True)
reward_actual = test_agent(ppo, env, memory, ep_steps, subset_resets, subset_resets_real, use_model=False)
print("\nSamples: %s, Reward in WM: %s, True Reward: %s" %(total_timesteps, np.round(reward_model,4), np.round(reward_actual, 4)))
## log progress to file
rewards.append(reward_actual)
rewards_m.append(reward_model)
errors.append(error)
lambdas.append(params['lam'])
df = pd.DataFrame({'Samples': samples, 'Reward': rewards, 'Reward_WM': rewards_m, 'Lambdas': lambdas, 'MSEs': errors})
lam = ['Adaptive' if params['adapt']==1 else 'fixed{}'.format(str(params['lam']))][0]
save_name = "{}_{}_resid{}_{}_{}".format(params['env_name'], lam, str(params['pca']), params['filename'], str(params['seed']))
if params['comment']:
save_name = save_name + '_' + params['comment']
save_name += '.csv'
df.to_csv(save_name)
## collect more data with the new policy
print("\nCollecting more data with the new policy...")
timesteps, error = collect_data(params, ppo, memory, env)
total_timesteps += timesteps
samples.append(total_timesteps)
b.update_dists(error, env.model.valid_loss)
if params['adapt']:
params['lam'] = b.sample()
print("\n Using Lambda = {}".format(str(params['lam'])))
iterations += 1
def main():
parser = argparse.ArgumentParser()
parser.add_argument('--env_name', type=str, default='HalfCheetah-v2') ## only works properly for HalfCheetah and Ant
parser.add_argument('--seed', '-se', type=int, default=0)
parser.add_argument('--num_models', '-nm', type=int, default=3)
parser.add_argument('--adapt', '-ad', type=int, default=0) ## set to 1 for adaptive
parser.add_argument('--steps', '-s', type=int, default=100) ## maximum time we step through an env per episode
parser.add_argument('--outer_steps', '-in', type=int, default=3000) ## how many time steps/samples we collect each outer loop (including initially)
parser.add_argument('--max_timesteps', '-maxt', type=int, default=1e8) ## total number of timesteps
parser.add_argument('--model_epochs', '-me', type=int, default=2000) ## max number of times we improve model
parser.add_argument('--update_timestep', '-ut', type=int, default=50000) ## for PPO only; how many steps to accumulate before training on them
parser.add_argument('--policy_iters', '-it', type=int, default=2000) ## max number of times we improve policy
parser.add_argument('--learning_rate', '-lr', type=float, default=0.1)
parser.add_argument('--lam', '-la', type=float, default=0)
parser.add_argument('--pca', '-pc', type=float, default=0) ## threshold for residual to stop, try [1e-4,2-e4]
parser.add_argument('--sigma', '-si', type=float, default=0.01)
parser.add_argument('--filename', '-f', type=str, default='ModelBased')
parser.add_argument('--dir', '-d', type=str, default='data')
parser.add_argument('--yaml_file', '-yml', type=str, default=None)
parser.add_argument('--uuid', '-id', type=str, default=None)
parser.add_argument('--fix_std', dest='fix_std', action='store_true')
parser.add_argument('--var_type', type=str, default='reward', choices=('reward', 'state'))
parser.add_argument('--model_free', dest='model_free', action='store_true')
parser.add_argument('--var_max', dest='var_max', action='store_true')
parser.add_argument('--comment', '-c', type=str, default=None)
parser.set_defaults(fix_std=False)
parser.set_defaults(model_free=False)
parser.set_defaults(var_max=False)
args = parser.parse_args()
params = vars(args)
if params['yaml_file']:
with open(args.yaml_file, 'r') as f:
yaml_config = yaml.load(f, Loader=yaml.FullLoader)
for config in yaml_config['args']:
if config in params:
params[config] = yaml_config['args'][config]
if not(os.path.exists(params['dir'])):
os.makedirs(params['dir'])
os.chdir(params['dir'])
if params['uuid']:
if not(os.path.exists(params['uuid'])):
os.makedirs(params['uuid'])
os.chdir(params['uuid'])
train_ppo(params)
if __name__ == '__main__':
main()