-
-
Notifications
You must be signed in to change notification settings - Fork 4
/
Copy pathload.py
188 lines (180 loc) · 6.18 KB
/
load.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 logging
from typing import Optional
import torch
from core.algorithms.evaluator import Evaluator
from core.resnet import TurboZeroResnet
from core.test.tester import TesterConfig, Tester, TwoPlayerTesterConfig, TwoPlayerTester
from core.train.collector import Collector
from core.train.trainer import Trainer, TrainerConfig
from core.utils.history import TrainingMetrics
from envs._2048.collector import _2048Collector
from envs._2048.tester import _2048Tester
from envs._2048.trainer import _2048Trainer
from envs.connect_x.collector import ConnectXCollector
from envs.connect_x.env import ConnectXConfig, ConnectXEnv
from envs.connect_x.tester import ConnectXTester
from envs.connect_x.trainer import ConnectXTrainer
from envs.othello.collector import OthelloCollector
from envs.othello.tester import OthelloTester
from envs.othello.trainer import OthelloTrainer
from .othello.env import OthelloEnv, OthelloEnvConfig
from ._2048.env import _2048Env, _2048EnvConfig
def init_env(device: torch.device, parallel_envs: int, env_config: dict, debug: bool):
env_type = env_config['env_type']
if env_type == 'othello':
config = OthelloEnvConfig(**env_config)
return OthelloEnv(parallel_envs, config, device, debug)
elif env_type == '2048':
config = _2048EnvConfig(**env_config)
return _2048Env(parallel_envs, config, device, debug)
elif env_type == 'connect_x':
config = ConnectXConfig(**env_config)
return ConnectXEnv(parallel_envs, config, device, debug)
else:
raise NotImplementedError(f'Environment {env_type} not implemented')
def init_collector(episode_memory_device: torch.device, env_type: str, evaluator: Evaluator):
if env_type == 'othello':
return OthelloCollector(
evaluator=evaluator,
episode_memory_device=episode_memory_device
)
elif env_type == '2048':
return _2048Collector(
evaluator=evaluator,
episode_memory_device=episode_memory_device
)
elif env_type == 'connect_x':
return ConnectXCollector(
evaluator=evaluator,
episode_memory_device=episode_memory_device
)
else:
raise NotImplementedError(f'Collector for environment {env_type} not supported')
def init_tester(
test_config: dict,
env_type: str,
collector: Collector,
model: torch.nn.Module,
history: TrainingMetrics,
optimizer: Optional[torch.optim.Optimizer],
log_results: bool,
debug: bool
):
if env_type == 'othello':
return OthelloTester(
config=TwoPlayerTesterConfig(**test_config),
collector=collector,
model=model,
optimizer=optimizer,
history=history,
log_results=log_results,
debug=debug
)
elif env_type == '2048':
return _2048Tester(
config=TesterConfig(**test_config),
collector=collector,
model=model,
optimizer=optimizer,
history=history,
log_results=log_results,
debug=debug
)
elif env_type == 'connect_x':
return ConnectXTester(
config=TwoPlayerTesterConfig(**test_config),
collector=collector,
model=model,
optimizer=optimizer,
history=history,
log_results=log_results,
debug=debug
)
else:
raise NotImplementedError(f'Tester for {env_type} not supported')
def init_trainer(
device: torch.device,
env_type: str,
collector: Collector,
tester: Tester,
model: TurboZeroResnet,
optimizer: torch.optim.Optimizer,
train_config: dict,
raw_env_config: dict,
history: TrainingMetrics,
log_results: bool,
interactive: bool,
run_tag: str = '',
debug: bool = False
):
trainer_config = TrainerConfig(**train_config)
if env_type == 'othello':
assert isinstance(collector, OthelloCollector)
assert isinstance(tester, TwoPlayerTester)
return OthelloTrainer(
config = trainer_config,
collector = collector,
tester = tester,
model = model,
optimizer = optimizer,
device = device,
raw_train_config = train_config,
raw_env_config = raw_env_config,
history = history,
log_results=log_results,
interactive=interactive,
run_tag = run_tag,
debug = debug
)
elif env_type == '2048':
assert isinstance(collector, _2048Collector)
return _2048Trainer(
config = trainer_config,
collector = collector,
tester = tester,
model = model,
optimizer = optimizer,
device = device,
raw_train_config = train_config,
raw_env_config = raw_env_config,
history = history,
log_results=log_results,
interactive=interactive,
run_tag = run_tag,
debug = debug
)
elif env_type == 'connect_x':
assert isinstance(collector, ConnectXCollector)
assert isinstance(tester, TwoPlayerTester)
return ConnectXTrainer(
config = trainer_config,
collector = collector,
tester = tester,
model = model,
optimizer = optimizer,
device = device,
raw_train_config = train_config,
raw_env_config = raw_env_config,
history = history,
log_results=log_results,
interactive=interactive,
run_tag = run_tag,
debug = debug
)
else:
logging.warn(f'No trainer found for environment {env_type}')
return Trainer(
config = trainer_config,
collector = collector,
tester = tester,
model = model,
optimizer = optimizer,
device = device,
raw_train_config = train_config,
raw_env_config = raw_env_config,
history = history,
log_results=log_results,
interactive=interactive,
run_tag = run_tag,
debug = debug
)