-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathplay_task.py
154 lines (138 loc) · 5.01 KB
/
play_task.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
import gym
from gym.utils.play import play
import random
import sys
sys.path.insert(0, 'language/')
import language.nl_wrapper as nlw
# import atari wrappers
from stable_baselines3.common.atari_wrappers import AtariWrapper
from stable_baselines3.common.vec_env import VecFrameStack
from language.task_wrapper import TaskWrapper
from language.tasks import *
#from language.play import play
import numpy as np
from ale_py import ALEInterface
import pickle
import time
from stable_baselines3 import PPO
from stable_baselines3.common.env_util import make_atari_env
from stable_baselines3.common.vec_env import VecFrameStack
from stable_baselines3.common.evaluation import evaluate_policy
# parse arguments
import argparse
task_dict = {0: DownLadderJumpRight, 1: ClimbDownRightLadder, 2: JumpSkullReachLadder,
3: JumpSkullGetKey, 4: ClimbLadderGetKey, 5: ClimbDownGoRightClimbUp, 6: JumpMiddleClimbReachLeftDoor}
parser = argparse.ArgumentParser()
parser.add_argument('--task', type=int, default=1, help='task number 0-6'+str(task_dict))
#parser.add_argument('--lang_rewards', action=argparse.BooleanOptionalAction)
parser.add_argument('--lang_rewards', type=str, default='true', help='use language rewards')
parser.add_argument('--timesteps', type=int, default=500000, help='number of timesteps to play')
parser.add_argument('--render', type=str, default='false', help='use language rewards')
parser.add_argument('--instr', type=str, default='none', help='instruction type')
parser.add_argument('--lang_coef', type=float, default=0.2, help='language reward coefficient')
# save path arg
parser.add_argument('--save_folder', type=str, default='data/train_log', help='save path')
parser.add_argument('--device', type=str, default='cuda', help='save path')
args = parser.parse_args()
print("lang rewards: ", args.lang_rewards)
print("task: ", args.task)
print("timesteps: ", args.timesteps)
print("render: ", args.render)
print("instr: ", args.instr)
print("lang_coef: ", args.lang_coef)
#log_save_path = 'data/train_log/task-{}-lang-{}.npy'.format(args.task, args.lang_rewards)
log_save_path = '{}/task-{}-lang-{}.npy'.format(args.save_folder, args.task, args.lang_rewards)
#env = gym.make("ALE/MontezumaRevenge-v5", render_mode='human')
#height, width, channels = env.observation_space.shape
#actions = env.action_space.n
if args.render == 'true':
env = gym.make("ALE/MontezumaRevenge-v5", render_mode='human')
else:
env = gym.make("ALE/MontezumaRevenge-v5")
#env = AtariWrapper(env)
if args.lang_rewards == 'true':
if args.instr == 'none':
env = nlw.Translearner(env, args=args)
else:
env = nlw.Translearner(env, args=args)
env = TaskWrapper(env, save_data=True, save_path=log_save_path)
task = task_dict[args.task](env)
env.assign_task(task)
env.reset()
#task = ClimbLadderGetKey(env)
# play(env, zoom=5)
# 2048
# env = VecFrameStack(env, n_stack=4)
# model = PPO("CnnPolicy", env, verbose=1, tensorboard_log="data/tensorboard/", device=args.device)
# model.learn(total_timesteps=args.timesteps)
# model.save("models/PPO-task-{}-lang-{}".format(args.task, args.lang_rewards))
print('key: ', env.has_key())
play(env, zoom=5)
print('key: ', env.has_key())
# print(env.n_steps)
# print(env.successes)
# print(env.successes_array)
#env.save_data_file()
# env.close()
# episodes = 100
# # time_steps = 0
# max_time = 1000
# log_interval = 1000
# num_finished = 0
# while True:
# if env.n_steps > max_time:
# break
# task.reset()
# start_lives = task.env.lives
# env.step(0)
# done = False
# score = 0
# dead = False
# while not done:
# if env.lives < start_lives:
# done = True
# env.render()
# action = env.env.action_space.sample()
# n_state, reward, done, info = env.step(action)
# score+=reward
# time.sleep(0.01)
# num_finished = (num_finished + 1) if task.finished() else num_finished
# done = env.finished() or done
# if env.finished():
# print("finished task")
# time.sleep(2)
# if env.n_steps >= max_time:
# break
# #print(reward)
# # print('Episode:{} Score:{}'.format(episode, score))
# # print(env.agent_pos())
# # print(env.has_key())
# # print(env.room())
# # print('Finished: {}/{}'.format(num_finished, episode))
# print(env.n_steps)
# print(num_finished)
# env.save_data_file()
# env.close()
# for episode in range(1, episodes+1):
# state = env.restore_state(newState)
# start_lives = 6
# env.step(0)
# done = False
# score = 0
# dead = False
# while not (done or dead):
# if env.lives < 6:
# dead = True
# print(env.lives)
# env.render()
# action = env.env.action_space.sample()
# n_state, reward, done, info = env.step(action)
# score+=reward
# time_steps += 1
# #time.sleep(0.01)
# print('Episode:{} Score:{}'.format(episode, score))
# print(env.agent_pos())
# print(env.has_key())
# print(env.room())
# print(time_steps)
# env.close()