File size: 5,999 Bytes
1e1c086 b18ddcc 1e1c086 b18ddcc 1e1c086 b18ddcc 3cc5c1d b18ddcc 3cc5c1d b18ddcc 3cc5c1d b18ddcc 3cc5c1d b18ddcc 3cc5c1d b18ddcc 3cc5c1d b18ddcc 1e1c086 3cc5c1d b18ddcc 3cc5c1d b18ddcc 3cc5c1d b18ddcc 1e1c086 b18ddcc 1e1c086 b18ddcc 1e1c086 b18ddcc 1e1c086 b18ddcc |
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 |
import dataclasses
from collections import defaultdict
from dataclasses import dataclass
from typing import Any, Dict, List, Optional, Sequence, TypeVar, Union
import numpy as np
from torch.utils.tensorboard.writer import SummaryWriter
@dataclass
class Episode:
score: float = 0
length: int = 0
info: Dict[str, Dict[str, Any]] = dataclasses.field(default_factory=dict)
StatisticSelf = TypeVar("StatisticSelf", bound="Statistic")
@dataclass
class Statistic:
values: np.ndarray
round_digits: int = 2
score_function: str = "mean-std"
@property
def mean(self) -> float:
return np.mean(self.values).item()
@property
def std(self) -> float:
return np.std(self.values).item()
@property
def min(self) -> float:
return np.min(self.values).item()
@property
def max(self) -> float:
return np.max(self.values).item()
def sum(self) -> float:
return np.sum(self.values).item()
def __len__(self) -> int:
return len(self.values)
def score(self) -> float:
if self.score_function == "mean-std":
return self.mean - self.std
elif self.score_function == "mean":
return self.mean
else:
raise NotImplemented(
f"Only mean-std and mean score_functions supported ({self.score_function})"
)
def _diff(self: StatisticSelf, o: StatisticSelf) -> float:
return self.score() - o.score()
def __gt__(self: StatisticSelf, o: StatisticSelf) -> bool:
return self._diff(o) > 0
def __ge__(self: StatisticSelf, o: StatisticSelf) -> bool:
return self._diff(o) >= 0
def __repr__(self) -> str:
mean = round(self.mean, self.round_digits)
if self.round_digits == 0:
mean = int(mean)
if self.score_function == "mean":
return f"{mean}"
std = round(self.std, self.round_digits)
if self.round_digits == 0:
std = int(std)
return f"{mean} +/- {std}"
def to_dict(self) -> Dict[str, float]:
return {
"mean": self.mean,
"std": self.std,
"min": self.min,
"max": self.max,
}
EpisodesStatsSelf = TypeVar("EpisodesStatsSelf", bound="EpisodesStats")
class EpisodesStats:
def __init__(
self,
episodes: Sequence[Episode],
simple: bool = False,
score_function: str = "mean-std",
) -> None:
self.episodes = episodes
self.simple = simple
self.score = Statistic(
np.array([e.score for e in episodes]), score_function=score_function
)
self.length = Statistic(np.array([e.length for e in episodes]), round_digits=0)
additional_values = defaultdict(list)
for e in self.episodes:
if e.info:
for k, v in e.info.items():
if isinstance(v, dict):
for k2, v2 in v.items():
additional_values[f"{k}_{k2}"].append(v2)
else:
additional_values[k].append(v)
self.additional_stats = {
k: Statistic(np.array(values)) for k, values in additional_values.items()
}
self.score_function = score_function
def __gt__(self: EpisodesStatsSelf, o: EpisodesStatsSelf) -> bool:
return self.score > o.score
def __ge__(self: EpisodesStatsSelf, o: EpisodesStatsSelf) -> bool:
return self.score >= o.score
def __repr__(self) -> str:
mean = self.score.mean
score = self.score.score()
if mean != score:
return f"Score: {self.score} ({round(score)}) | Length: {self.length}"
else:
return f"Score: {self.score} | Length: {self.length}"
def __len__(self) -> int:
return len(self.episodes)
def _asdict(self) -> dict:
return {
"n_episodes": len(self.episodes),
"score": self.score.to_dict(),
"length": self.length.to_dict(),
}
def write_to_tensorboard(
self, tb_writer: SummaryWriter, main_tag: str, global_step: Optional[int] = None
) -> None:
stats = {"mean": self.score.mean}
if not self.simple:
stats.update(
{
"min": self.score.min,
"max": self.score.max,
"result": self.score.score(),
"n_episodes": len(self.episodes),
"length": self.length.mean,
}
)
for k, addl_stats in self.additional_stats.items():
stats[k] = addl_stats.mean
for name, value in stats.items():
tb_writer.add_scalar(f"{main_tag}/{name}", value, global_step=global_step)
class EpisodeAccumulator:
def __init__(self, num_envs: int):
self._episodes = []
self.current_episodes = [Episode() for _ in range(num_envs)]
@property
def episodes(self) -> List[Episode]:
return self._episodes
def step(self, reward: np.ndarray, done: np.ndarray, info: List[Dict]) -> None:
for idx, current in enumerate(self.current_episodes):
current.score += reward[idx]
current.length += 1
if done[idx]:
self._episodes.append(current)
self.current_episodes[idx] = Episode()
self.on_done(idx, current, info[idx])
def __len__(self) -> int:
return len(self.episodes)
def on_done(self, ep_idx: int, episode: Episode, info: Dict) -> None:
pass
def stats(self) -> EpisodesStats:
return EpisodesStats(self.episodes)
def log_scalars(
tb_writer: SummaryWriter,
main_tag: str,
tag_scalar_dict: Dict[str, Union[int, float]],
global_step: int,
) -> None:
for tag, value in tag_scalar_dict.items():
tb_writer.add_scalar(f"{main_tag}/{tag}", value, global_step)
|