Spaces:
Sleeping
Sleeping
File size: 11,491 Bytes
eaf2e33 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 |
from collections import deque, OrderedDict
from src.rlkit.core.eval_util import create_stats_ordered_dict
from src.rlkit.samplers.rollout_functions import rollout, multitask_rollout, ensemble_rollout, ensemble_eval_rollout
from src.rlkit.samplers.rollout_functions import ensemble_ucb_rollout
from src.rlkit.samplers.data_collector.base import PathCollector
class MdpPathCollector(PathCollector):
def __init__(
self,
env,
policy,
noise_flag=0,
max_num_epoch_paths_saved=None,
render=False,
render_kwargs=None,
):
if render_kwargs is None:
render_kwargs = {}
self._env = env
self._policy = policy
self._max_num_epoch_paths_saved = max_num_epoch_paths_saved
self._epoch_paths = deque(maxlen=self._max_num_epoch_paths_saved)
self._render = render
self._render_kwargs = render_kwargs
self._noise_flag = noise_flag
self._num_steps_total = 0
self._num_paths_total = 0
def collect_new_paths(
self,
max_path_length,
num_steps,
discard_incomplete_paths,
):
paths = []
num_steps_collected = 0
while num_steps_collected < num_steps:
max_path_length_this_loop = min( # Do not go over num_steps
max_path_length,
num_steps - num_steps_collected,
)
path = rollout(
self._env,
self._policy,
noise_flag=self._noise_flag,
max_path_length=max_path_length_this_loop,
)
path_len = len(path['actions'])
if (
path_len != max_path_length
and not path['terminals'][-1]
and discard_incomplete_paths
):
break
num_steps_collected += path_len
paths.append(path)
self._num_paths_total += len(paths)
self._num_steps_total += num_steps_collected
self._epoch_paths.extend(paths)
return paths
def collect_normalized_new_paths(
self,
max_path_length,
num_steps,
discard_incomplete_paths,
input_mean,
input_std,
):
paths = []
num_steps_collected = 0
while num_steps_collected < num_steps:
max_path_length_this_loop = min( # Do not go over num_steps
max_path_length,
num_steps - num_steps_collected,
)
path = normalized_rollout(
self._env,
self._policy,
noise_flag=self._noise_flag,
max_path_length=max_path_length_this_loop,
input_mean=input_mean,
input_std=input_std,
)
path_len = len(path['actions'])
if (
path_len != max_path_length
and not path['terminals'][-1]
and discard_incomplete_paths
):
break
num_steps_collected += path_len
paths.append(path)
self._num_paths_total += len(paths)
self._num_steps_total += num_steps_collected
self._epoch_paths.extend(paths)
return paths
def get_epoch_paths(self):
return self._epoch_paths
def end_epoch(self, epoch):
self._epoch_paths = deque(maxlen=self._max_num_epoch_paths_saved)
def get_diagnostics(self):
path_lens = [len(path['actions']) for path in self._epoch_paths]
stats = OrderedDict([
('num steps total', self._num_steps_total),
('num paths total', self._num_paths_total),
])
stats.update(create_stats_ordered_dict(
"path length",
path_lens,
always_show_all_stats=True,
))
return stats
def get_snapshot(self):
return dict(
env=self._env,
policy=self._policy,
)
class EnsembleMdpPathCollector(PathCollector):
def __init__(
self,
env,
policy,
num_ensemble,
noise_flag=0,
ber_mean=0.5,
eval_flag=False,
max_num_epoch_paths_saved=None,
render=False,
render_kwargs=None,
critic1=None,
critic2=None,
inference_type=0.0,
feedback_type=1,
):
if render_kwargs is None:
render_kwargs = {}
self._env = env
self._policy = policy
self._max_num_epoch_paths_saved = max_num_epoch_paths_saved
self._epoch_paths = deque(maxlen=self._max_num_epoch_paths_saved)
self._render = render
self._render_kwargs = render_kwargs
self.num_ensemble = num_ensemble
self.eval_flag = eval_flag
self.ber_mean = ber_mean
self.critic1 = critic1
self.critic2 = critic2
self.inference_type = inference_type
self.feedback_type = feedback_type
self._noise_flag = noise_flag
self._num_steps_total = 0
self._num_paths_total = 0
def collect_new_paths(
self,
max_path_length,
num_steps,
discard_incomplete_paths,
):
paths = []
num_steps_collected = 0
while num_steps_collected < num_steps:
max_path_length_this_loop = min( # Do not go over num_steps
max_path_length,
num_steps - num_steps_collected,
)
if self.eval_flag:
path = ensemble_eval_rollout(
self._env,
self._policy,
self.num_ensemble,
max_path_length=max_path_length_this_loop,
)
else:
if self.inference_type > 0: # UCB
path = ensemble_ucb_rollout(
self._env,
self._policy,
critic1=self.critic1,
critic2=self.critic2,
inference_type=self.inference_type,
feedback_type=self.feedback_type,
num_ensemble=self.num_ensemble,
noise_flag=self._noise_flag,
max_path_length=max_path_length_this_loop,
ber_mean=self.ber_mean,
)
else:
path = ensemble_rollout(
self._env,
self._policy,
self.num_ensemble,
noise_flag=self._noise_flag,
max_path_length=max_path_length_this_loop,
ber_mean=self.ber_mean,
)
path_len = len(path['actions'])
if (
path_len != max_path_length
and not path['terminals'][-1]
and discard_incomplete_paths
):
break
num_steps_collected += path_len
paths.append(path)
self._num_paths_total += len(paths)
self._num_steps_total += num_steps_collected
self._epoch_paths.extend(paths)
return paths
def get_epoch_paths(self):
return self._epoch_paths
def end_epoch(self, epoch):
self._epoch_paths = deque(maxlen=self._max_num_epoch_paths_saved)
def get_diagnostics(self):
path_lens = [len(path['actions']) for path in self._epoch_paths]
stats = OrderedDict([
('num steps total', self._num_steps_total),
('num paths total', self._num_paths_total),
])
stats.update(create_stats_ordered_dict(
"path length",
path_lens,
always_show_all_stats=True,
))
return stats
def get_snapshot(self):
return dict(
env=self._env,
policy=self._policy,
)
class AsyncEnsembleMdpPathCollector(PathCollector):
def __init__(
self,
env,
policy,
num_ensemble,
noise_flag=0,
ber_mean=0.5,
eval_flag=False,
max_num_epoch_paths_saved=None,
render=False,
render_kwargs=None,
critic1=None,
critic2=None,
inference_type=0.0,
feedback_type=1,
):
if render_kwargs is None:
render_kwargs = {}
self._env = env
self._policy = policy
self._max_num_epoch_paths_saved = max_num_epoch_paths_saved
self._epoch_paths = deque(maxlen=self._max_num_epoch_paths_saved)
self._render = render
self._render_kwargs = render_kwargs
self.num_ensemble = num_ensemble
self.eval_flag = eval_flag
self.ber_mean = ber_mean
self.critic1 = critic1
self.critic2 = critic2
self.inference_type = inference_type
self.feedback_type = feedback_type
self._noise_flag = noise_flag
self._num_steps_total = 0
self._num_paths_total = 0
def collect_new_paths(
self,
max_path_length,
num_steps,
discard_incomplete_paths,
):
paths = []
num_steps_collected = 0
while num_steps_collected < num_steps:
max_path_length_this_loop = min( # Do not go over num_steps
max_path_length,
num_steps - num_steps_collected,
)
if self.eval_flag:
path = ensemble_eval_rollout(
self._env,
self._policy,
self.num_ensemble,
max_path_length=max_path_length_this_loop,
)
else:
if self.inference_type > 0: # UCB
path = ensemble_ucb_rollout(
self._env,
self._policy,
critic1=self.critic1,
critic2=self.critic2,
inference_type=self.inference_type,
feedback_type=self.feedback_type,
num_ensemble=self.num_ensemble,
noise_flag=self._noise_flag,
max_path_length=max_path_length_this_loop,
ber_mean=self.ber_mean,
)
else:
path = ensemble_rollout(
self._env,
self._policy,
self.num_ensemble,
noise_flag=self._noise_flag,
max_path_length=max_path_length_this_loop,
ber_mean=self.ber_mean,
)
path_len = len(path['actions'])
if (
path_len != max_path_length
and not path['terminals'][-1]
and discard_incomplete_paths
):
break
num_steps_collected += path_len
paths.append(path)
self._num_paths_total += len(paths)
self._num_steps_total += num_steps_collected
self._epoch_paths.extend(paths)
return paths
|