| import json |
|
|
| import crafter |
| import elements |
| import embodied |
| import numpy as np |
|
|
|
|
| class Crafter(embodied.Env): |
|
|
| def __init__(self, task, size=(64, 64), logs=False, logdir=None, seed=None): |
| assert task in ('reward', 'noreward') |
| self._env = crafter.Env(size=size, reward=(task == 'reward'), seed=seed) |
| self._logs = logs |
| self._logdir = logdir and elements.Path(logdir) |
| self._logdir and self._logdir.mkdir() |
| self._episode = 0 |
| self._length = None |
| self._reward = None |
| self._achievements = crafter.constants.achievements.copy() |
| self._done = True |
|
|
| @property |
| def obs_space(self): |
| spaces = { |
| 'image': elements.Space(np.uint8, self._env.observation_space.shape), |
| 'reward': elements.Space(np.float32), |
| 'is_first': elements.Space(bool), |
| 'is_last': elements.Space(bool), |
| 'is_terminal': elements.Space(bool), |
| 'log/reward': elements.Space(np.float32), |
| } |
| if self._logs: |
| spaces.update({ |
| f'log/achievement_{k}': elements.Space(np.int32) |
| for k in self._achievements}) |
| return spaces |
|
|
| @property |
| def act_space(self): |
| return { |
| 'action': elements.Space(np.int32, (), 0, self._env.action_space.n), |
| 'reset': elements.Space(bool), |
| } |
|
|
| def step(self, action): |
| if action['reset'] or self._done: |
| self._episode += 1 |
| self._length = 0 |
| self._reward = 0 |
| self._done = False |
| image = self._env.reset() |
| return self._obs(image, 0.0, {}, is_first=True) |
| image, reward, self._done, info = self._env.step(action['action']) |
| self._reward += reward |
| self._length += 1 |
| if self._done and self._logdir: |
| self._write_stats(self._length, self._reward, info) |
| return self._obs( |
| image, reward, info, |
| is_last=self._done, |
| is_terminal=info['discount'] == 0) |
|
|
| def _obs( |
| self, image, reward, info, |
| is_first=False, is_last=False, is_terminal=False): |
| obs = dict( |
| image=image, |
| reward=np.float32(reward), |
| is_first=is_first, |
| is_last=is_last, |
| is_terminal=is_terminal, |
| **{'log/reward': np.float32(info['reward'] if info else 0.0)}, |
| ) |
| if self._logs: |
| log_achievements = { |
| f'log/achievement_{k}': info['achievements'][k] if info else 0 |
| for k in self._achievements} |
| obs.update({k: np.int32(v) for k, v in log_achievements.items()}) |
| return obs |
|
|
| def _write_stats(self, length, reward, info): |
| stats = { |
| 'episode': self._episode, |
| 'length': length, |
| 'reward': round(reward, 1), |
| **{f'achievement_{k}': v for k, v in info['achievements'].items()}, |
| } |
| filename = self._logdir / 'stats.jsonl' |
| lines = filename.read() if filename.exists() else '' |
| lines += json.dumps(stats) + '\n' |
| filename.write(lines, mode='w') |
| print(f'Wrote stats: {filename}') |
|
|
