File size: 4,219 Bytes
9839b09 |
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 |
import numpy as np
from dataclasses import dataclass
from torch.utils.tensorboard.writer import SummaryWriter
from typing import Dict, List, Optional, Sequence, TypeVar
@dataclass
class Episode:
score: float = 0
length: int = 0
StatisticSelf = TypeVar("StatisticSelf", bound="Statistic")
@dataclass
class Statistic:
values: np.ndarray
round_digits: int = 2
@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 _diff(self: StatisticSelf, o: StatisticSelf) -> float:
return (self.mean - self.std) - (o.mean - o.std)
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)
std = round(self.std, self.round_digits)
if self.round_digits == 0:
mean = int(mean)
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:
episodes: Sequence[Episode]
simple: bool
score: Statistic
length: Statistic
def __init__(self, episodes: Sequence[Episode], simple: bool = False) -> None:
self.episodes = episodes
self.simple = simple
self.score = Statistic(np.array([e.score for e in episodes]))
self.length = Statistic(np.array([e.length for e in episodes]), round_digits=0)
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:
return (
f"Score: {self.score} ({round(self.score.mean - self.score.std, 2)}) | "
f"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.mean - self.score.std,
"n_episodes": len(self.episodes),
}
)
tb_writer.add_scalars(
main_tag,
stats,
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) -> None:
for idx, current in enumerate(self.current_episodes):
current.score += reward[idx]
current.length += 1
if done[idx]:
self._episodes.append(current)
self.on_done(idx, current)
self.current_episodes[idx] = Episode()
def __len__(self) -> int:
return len(self.episodes)
def on_done(self, ep_idx: int, episode: Episode) -> None:
pass
def stats(self) -> EpisodesStats:
return EpisodesStats(self.episodes)
|