94bfb32cc1
1. add policy.eval() in all test scripts' "watch performance" 2. remove dict return support for collector preprocess_fn 3. add `__contains__` and `pop` in batch: `key in batch`, `batch.pop(key, deft)` 4. exact n_episode for a list of n_episode limitation and save fake data in cache_buffer when self.buffer is None (#184) 5. fix tensorboard logging: h-axis stands for env step instead of gradient step; add test results into tensorboard 6. add test_returns (both GAE and nstep) 7. change the type-checking order in batch.py and converter.py in order to meet the most often case first 8. fix shape inconsistency for torch.Tensor in replay buffer 9. remove `**kwargs` in ReplayBuffer 10. remove default value in batch.split() and add merge_last argument (#185) 11. improve nstep efficiency 12. add max_batchsize in onpolicy algorithms 13. potential bugfix for subproc.wait 14. fix RecurrentActorProb 15. improve the code-coverage (from 90% to 95%) and remove the dead code 16. fix some incorrect type annotation The above improvement also increases the training FPS: on my computer, the previous version is only ~1800 FPS and after that, it can reach ~2050 (faster than v0.2.4.post1).
55 lines
1.7 KiB
Python
55 lines
1.7 KiB
Python
import gym
|
|
import numpy as np
|
|
from typing import List, Callable, Tuple, Optional, Any
|
|
|
|
from tianshou.env.worker import EnvWorker
|
|
|
|
try:
|
|
import ray
|
|
except ImportError:
|
|
pass
|
|
|
|
|
|
class RayEnvWorker(EnvWorker):
|
|
"""Ray worker used in RayVectorEnv."""
|
|
|
|
def __init__(self, env_fn: Callable[[], gym.Env]) -> None:
|
|
super().__init__(env_fn)
|
|
self.env = ray.remote(gym.Wrapper).options(num_cpus=0).remote(env_fn())
|
|
|
|
def __getattr__(self, key: str) -> Any:
|
|
return ray.get(self.env.__getattr__.remote(key))
|
|
|
|
def reset(self) -> Any:
|
|
return ray.get(self.env.reset.remote())
|
|
|
|
@staticmethod
|
|
def wait(workers: List['RayEnvWorker'],
|
|
wait_num: int,
|
|
timeout: Optional[float] = None) -> List['RayEnvWorker']:
|
|
results = [x.result for x in workers]
|
|
ready_results, _ = ray.wait(results,
|
|
num_returns=wait_num, timeout=timeout)
|
|
return [workers[results.index(result)] for result in ready_results]
|
|
|
|
def send_action(self, action: np.ndarray) -> None:
|
|
# self.action is actually a handle
|
|
self.result = self.env.step.remote(action)
|
|
|
|
def get_result(self) -> Tuple[
|
|
np.ndarray, np.ndarray, np.ndarray, np.ndarray]:
|
|
return ray.get(self.result)
|
|
|
|
def seed(self, seed: Optional[int] = None) -> List[int]:
|
|
if hasattr(self.env, 'seed'):
|
|
return ray.get(self.env.seed.remote(seed))
|
|
return None
|
|
|
|
def render(self, **kwargs) -> Any:
|
|
if hasattr(self.env, 'render'):
|
|
return ray.get(self.env.render.remote(**kwargs))
|
|
return None
|
|
|
|
def close_env(self) -> None:
|
|
ray.get(self.env.close.remote())
|