-
Notifications
You must be signed in to change notification settings - Fork 2
/
run.py
166 lines (141 loc) · 8.29 KB
/
run.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
import os
import gym
import time
import numpy as np
import torch
from torch.utils.tensorboard import SummaryWriter
from mjrl.algos.npg_cg import NPG
from mjrl.algos.behavior_cloning import BC
from mjrl.policies.gaussian_mlp import MLP
from mjrl.baselines.mlp_baseline import MLPBaseline
from mjrl.utils.gym_env import GymEnv
from mjrl.utils.train_agent import train_agent
from mjrl.samplers.core import sample_paths, sample_data_batch
import milo.gym_env
from milo.utils import *
from milo.gym_env import model_based_env
from milo.dataset import OfflineDataset
from milo.cost import RBFLinearCost
from milo.dynamics_model import DynamicsEnsemble
def main():
args = get_args()
dirs, ids, ensemble_checkpoint, logger, writer, device = setup(args, ask_prompt=True)
# ======== Dataset Setup ==========
offline_db_path = os.path.join(args.data_path, 'offline_data', args.offline_db)
expert_db_path = os.path.join(args.data_path, 'expert_data', args.expert_db)
offline_state, offline_action, offline_next_state = get_db_mjrl(offline_db_path, 'all')
if args.subsample_expert:
expert_state, expert_action, expert_next_state = get_db_mjrl_randomize(expert_db_path, args.num_samples) # Expert DB
else:
expert_state, expert_action, expert_next_state = get_db_mjrl(expert_db_path, args.num_trajs) # Expert DB
offline_dataset = OfflineDataset(args.env, offline_state, offline_action, offline_next_state, device=device)
# Compute Expert DB stats
max_expert_norm = torch.max(torch.norm(expert_state, p=2, dim=1))
# ========= Create Model Ensemble =========
model_ensemble = DynamicsEnsemble(args.env, args.n_models, offline_dataset, hidden_sizes=[1024, 1024], base_seed=args.seed)
if ensemble_checkpoint is not None:
logger.info(f">>>>> Loading Dynamics ensemble with id: {ids['dynamics_id']}")
model_ensemble.load(ensemble_checkpoint)
else:
logger.info(f">>>>> Training Dynamics ensemble with id: {ids['dynamics_id']}")
model_ensemble.train(n_epochs=args.n_epochs, logger=logger, log_epoch=False, grad_clip=args.grad_clip)
logger.info(f">>>>> Saving ensemble weights in {dirs['dynamics_path']}")
model_ensemble.save(dirs['dynamics_path'])
model_ensemble.compute_threshold()
logger.info(f">>>>> Computed Maximum Discrepancy for Ensemble: {model_ensemble.threshold}")
# ======== ENV SETUP ========
logger.info(">>>>> Creating Environments")
inf_env = GymEnv(gym.make(args.env))
mb_env = GymEnv(model_based_env(gym.make(args.env), model_ensemble, init_state_buffer=expert_state.numpy(),\
norm_thresh = args.norm_thresh_coeff*max_expert_norm, device=device))
# ====== Cost Setup =======
cost_function = RBFLinearCost(torch.cat([expert_state, expert_action], dim=1), feature_dim=args.feature_dim, \
bw_quantile=args.bw_quantile, lambda_b=args.lambda_b, seed=args.seed)
# ============= INIT AGENT =============
policy = MLP(inf_env.spec, hidden_sizes=tuple(args.actor_model_hidden), seed=args.seed,
init_log_std=args.policy_init_log, min_log_std=args.policy_min_log)
baseline = MLPBaseline(inf_env.spec, reg_coef=args.vf_reg_coef, batch_size=args.vf_batch_size, \
hidden_sizes=tuple(args.critic_model_hidden), epochs=args.vf_iters, learn_rate=args.vf_lr)
# =============== BC Warmstart =================
if args.bc_epochs > 0:
logger.info(f">>>>> BC Warmstart for {args.bc_epochs} epochs")
offline_paths = get_paths_mjrl(offline_db_path, 'all')
bc_agent = BC(offline_paths, policy=policy, epochs=args.bc_epochs, batch_size=64, lr=1e-3)
bc_agent.train()
# Reinit Policy Std
policy_params = policy.get_param_values()
action_dim = inf_env.env.action_space.shape[0]
policy_params[-1*action_dim:] = args.policy_init_log
policy.set_param_values(policy_params, set_new=True, set_old=True)
# ============== Policy Gradient Init =============
if args.subsample_expert:
expert_paths = get_paths_mjrl_randomize(expert_db_path, args.num_samples)
else:
expert_paths = get_paths_mjrl(expert_db_path, args.num_trajs)
bc_reg_args = {'flag': args.do_bc_reg, 'reg_coeff': args.bc_reg_coeff, 'expert_paths': expert_paths[0]}
if args.planner == 'trpo':
cg_args = {'iters': args.cg_iter, 'damping': args.cg_damping}
planner_agent = NPG(mb_env, policy, baseline, normalized_step_size=args.kl_dist, \
hvp_sample_frac=args.hvp_sample_frac, seed=args.seed, FIM_invert_args=cg_args, \
bc_args=bc_reg_args, save_logs=True)
else:
raise NotImplementedError('Chosen Planner not yet supported')
# ==============================================
# ============== MAIN LOOP START ===============
# ==============================================
n_iter = 0
best_policy_score = -float('inf')
greedy_scores, sample_scores, greedy_mmds, sample_mmds = [], [], [], []
while n_iter<args.n_iter:
logger.info(f"{'='*10} Main Episode {n_iter+1} {'='*10}")
# ============= Evaluate, Save, Plot ===============
scores, mmds = evaluate(n_iter, logger, writer, args, inf_env, \
planner_agent.policy, cost_function, num_traj=50)
save_and_plot(n_iter, args, dirs, scores, mmds)
if scores['greedy'] > best_policy_score:
best_policy_score = scores['greedy']
save_checkpoint(dirs, planner_agent, cost_function, 'best', agent_type=args.planner)
if (n_iter+1) % args.save_iter == 0:
save_checkpoint(dirs, planner_agent, cost_function, n_iter+1, agent_type=args.planner)
# =============== DO PG STEPS =================
logger.info('=== PG Planning Start ===')
best_baseline_optim, best_baseline = None, None
best_policy = None
curr_max_reward, curr_min_vloss = -float('inf'), float('inf')
for i in range(args.pg_iter):
reward_kwargs = dict(reward_func=cost_function, ensemble=model_ensemble, device=device)
planner_args = dict(N=args.samples_per_step, env=mb_env, sample_mode='model_based', \
gamma=args.gamma, gae_lambda=args.gae_lambda, num_cpu=4, \
reward_kwargs=reward_kwargs)
r_mean, r_std, r_min, r_max, _, infos = planner_agent.train_step(**planner_args)
# Baseline Heuristic
if infos['vf_loss_end'] < curr_min_vloss:
curr_min_vloss = infos['vf_loss_end']
best_baseline = planner_agent.baseline.model.state_dict()
best_baseline_optim = planner_agent.baseline.optimizer.state_dict()
# Stderr Logging
reward_mean = np.array(infos['reward']).mean()
int_mean = np.array(infos['int']).mean()
ext_mean = np.array(infos['ext']).mean()
len_mean = np.array(infos['ep_len']).mean()
ground_truth_mean = np.array(infos['ground_truth_reward']).mean()
logger.info(f'Model MMD: {infos["mb_mmd"]}')
logger.info(f'Bonus MMD: {infos["bonus_mmd"]}')
logger.info(f'Model Ground Truth Reward: {ground_truth_mean}')
logger.info('PG Iteration {} reward | int | ext | ep_len ---- {:.2f} | {:.2f} | {:.2f} | {:.2f}' \
.format(i+1, reward_mean, int_mean, ext_mean, len_mean))
# Tensorboard Logging
step_count = n_iter*args.pg_iter + i
writer.add_scalar('data/reward_mean', reward_mean, step_count)
writer.add_scalar('data/ext_reward_mean', ext_mean, step_count)
writer.add_scalar('data/int_reward_mean', int_mean, step_count)
writer.add_scalar('data/ep_len_mean', len_mean, step_count)
writer.add_scalar('data/true_reward_mean', ground_truth_mean, step_count)
writer.add_scalar('data/value_loss', infos['vf_loss_end'], step_count)
writer.add_scalar('data/mb_mmd', infos['mb_mmd'], step_count)
writer.add_scalar('data/bonus_mmd', infos['bonus_mmd'], step_count)
planner_agent.baseline.model.load_state_dict(best_baseline)
planner_agent.baseline.optimizer.load_state_dict(best_baseline_optim)
n_iter += 1
if __name__ == '__main__':
main()