-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathtrain_dilo.py
188 lines (147 loc) · 7.72 KB
/
train_dilo.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
import os
os.environ['MKL_SERVICE_FORCE_INTEL'] = '1'
os.environ['MUJOCO_GL'] = 'egl'
import datetime
from absl import app, flags
from ml_collections import config_flags
from dataclasses import dataclass
from dataset_utils import make_env_and_dataset
from dataset_utils import SNSD4RLDataset,SNSD4RLMixedDataset, split_into_trajectories
import sys
import warnings
warnings.filterwarnings('ignore', category=DeprecationWarning)
import random
import numpy as np
import torch
from trainer import TrainerSNS as TrainerDILO
import dilo_utils
from models import TwinQ, ValueFunction, TwinV
from dilo import DILO
import json
from policy import GaussianPolicy, DeterministicPolicy
import time
from logging_utils.logx import EpochLogger
torch.backends.cudnn.benchmark = True
FLAGS = flags.FLAGS
flags.DEFINE_string('env_name', 'hopper-medium-v2', 'Environment name.')
flags.DEFINE_string('save_dir', './tmp/', 'Tensorboard logging dir.')
flags.DEFINE_string('exp_name', 'dump', 'Epoch logging dir.')
flags.DEFINE_integer('seed', 42, 'Random seed.')
flags.DEFINE_integer('expert_trajectories', 200, 'Number of expert trajectories')
flags.DEFINE_integer('eval_episodes', 10,
'Number of episodes used for evaluation.')
flags.DEFINE_integer('log_interval', 5000, 'Logging interval.')
flags.DEFINE_integer('eval_interval', 5000, 'Eval interval.')
flags.DEFINE_integer('batch_size', 1024, 'Mini batch size.')
flags.DEFINE_boolean('double', True, 'Use double q-learning')
flags.DEFINE_integer('max_steps', int(3e5), 'Number of training steps.')
flags.DEFINE_boolean('tqdm', True, 'Use tqdm progress bar.')
flags.DEFINE_float('lamda', 0.8, 'f-maximization strength')
flags.DEFINE_float('beta', 0.5, 'Mixture distribution ratio') # Only implemented for 0.5
flags.DEFINE_float('ita', 0.5, 'Orthogonalization strength')
flags.DEFINE_float('tau', 3.0/50, 'Policy Temperature')
flags.DEFINE_string('maximizer', 'smoothed_chi', 'Which f divergence to use')
flags.DEFINE_string('grad', 'full', 'Semi-gradient or full-gradient?')
config_flags.DEFINE_config_file(
'config',
'configs/mujoco_config.py',
'File path to the training hyperparameter configuration.',
lock_config=False)
def evaluate(agent, env,
num_episodes, device, verbose: bool = False,normalization_stats=None) :
stats = {'return': [], 'length': []}
for _ in range(num_episodes):
observation, done = env.reset(), False
while not done:
if 'obs_mean' in normalization_stats:
action = agent.act(torch.FloatTensor((observation-normalization_stats['obs_mean'])/normalization_stats['obs_std']).to(device), deterministic=True)
else:
action = agent.act(torch.FloatTensor(observation).to(device), deterministic=True)
observation, _, done, info = env.step(action.detach().cpu().numpy())
for k in stats.keys():
stats[k].append(info['episode'][k])
if verbose:
v = info['episode'][k]
print(f'{k}:{v}')
for k, v in stats.items():
stats[k] = np.mean(v)
return stats
def normalize(dataset):
trajs = split_into_trajectories(dataset.observations, dataset.actions,
dataset.rewards, dataset.masks,
dataset.dones_float,
dataset.next_observations)
def compute_returns(traj):
episode_return = 0
for _, _, rew, _, _, _ in traj:
episode_return += rew
return episode_return
trajs.sort(key=compute_returns)
dataset.rewards /= compute_returns(trajs[-1]) - compute_returns(trajs[0])
dataset.rewards *= 1000.0
def main(_):
# Set all seeds
torch.manual_seed(FLAGS.seed)
np.random.seed(FLAGS.seed)
random.seed(FLAGS.seed)
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
ts_str = datetime.datetime.fromtimestamp(time.time()).strftime("%Y-%m-%d_%H-%M-%S")
save_dir = os.path.join(FLAGS.save_dir, ts_str)
exp_id = f"results/dilo/{FLAGS.env_name}/{FLAGS.expert_trajectories}_expert/observations/" + FLAGS.exp_name
log_folder = exp_id + '/'+FLAGS.exp_name+'_s'+str(FLAGS.seed)
logger_kwargs={'output_dir':log_folder, 'exp_name':FLAGS.exp_name}
e_logger = EpochLogger(**logger_kwargs)
write_config = {}
# Iterate through all flags and add them to the dictionary
for flag_name in FLAGS.flags_by_module_dict()[sys.argv[0]]:
key = flag_name.serialize().split('=')[0]
if 'config' in key:
continue
value = flag_name.value
write_config[key] = value
write_config.update(dict(FLAGS.config))
with open(log_folder+"/config.json", "w") as outfile:
json.dump(write_config, outfile,indent=4)
hparam_str_dict = dict(seed=FLAGS.seed, env=FLAGS.env_name)
hparam_str = ','.join([
'%s=%s' % (k, str(hparam_str_dict[k]))
for k in sorted(hparam_str_dict.keys())
])
os.makedirs(save_dir, exist_ok=True)
env, dataset, expert_dataset, offline_min, offline_max, normalization_stats = make_env_and_dataset(FLAGS.env_name, FLAGS.seed,FLAGS.expert_trajectories,normalize_obs=False)
obs_dim = env.observation_space.shape[0]
act_dim = env.action_space.shape[0]
if FLAGS.grad=='semi':
agent = DILO(qf=TwinQ(state_dim=obs_dim,act_dim=obs_dim),vf = TwinV(state_dim=obs_dim),policy=GaussianPolicy(obs_dim,act_dim),
optimizer_factory=torch.optim.Adam,
lamda=FLAGS.lamda, maximizer=FLAGS.maximizer,beta=FLAGS.beta,ita=FLAGS.ita,tau=FLAGS.tau, gradient_type=FLAGS.grad, lr=3e-4, discount=0.99, alpha=0.005).to(device)
else:
agent = DILO(qf=TwinQ(state_dim=obs_dim,act_dim=obs_dim),vf = ValueFunction(state_dim=obs_dim),policy=DeterministicPolicy(obs_dim,act_dim),
optimizer_factory=torch.optim.Adam,
lamda=FLAGS.lamda, maximizer=FLAGS.maximizer,beta=FLAGS.beta,ita=FLAGS.ita,tau=FLAGS.tau, gradient_type=FLAGS.grad,use_twinV=True, lr=3e-4, discount=0.99, alpha=0.005).to(device)
trainer = TrainerDILO()
best_eval_returns = -np.inf
eval_returns = []
for i in range(1, FLAGS.max_steps + 1): # Remove TQDM
batch = dataset.sample(FLAGS.batch_size)
expert_batch = expert_dataset.sample(FLAGS.batch_size)
update_info, st = trainer.update(agent, batch, expert_batch)
if i % FLAGS.eval_interval == 0:
eval_stats = evaluate(agent.policy, env, FLAGS.eval_episodes, device,normalization_stats=normalization_stats)
if eval_stats['return'] >= best_eval_returns:
# Store best eval returns
best_eval_returns = eval_stats['return']
e_logger.log_tabular('Iterations', i)
e_logger.log_tabular('AverageNormalizedReturn', eval_stats['return'])
e_logger.log_tabular('SeenExpertV', update_info['expert_v_val'])
e_logger.log_tabular('SeenReplayV', update_info['replay_v_val'])
e_logger.log_tabular('UnseenExpertV', update_info['unseen_expert_v_val'])
e_logger.log_tabular('UnseenReplayV', update_info['unseen_replay_v_val'])
e_logger.log_tabular('UnseenExpertPolW', update_info['unseen_expert_pol_weight'])
e_logger.log_tabular('UnseenReplayPolW', update_info['unseen_replay_pol_weight'])
e_logger.log_tabular('Policy Loss', update_info['policy_loss'])
e_logger.dump_tabular()
eval_returns.append((i, eval_stats['return']))
print("Iterations: {} Average Return: {}".format(i,eval_stats['return']))
if __name__ == '__main__':
app.run(main)