Unity 机器学习代理工具包 (ML-Agents) 是一个开源项目,它使游戏和模拟能够作为训练智能代理的环境。
您最多选择25个主题 主题必须以中文或者字母或数字开头,可以包含连字符 (-),并且长度不得超过35个字符
 
 
 
 
 

46 行
1.5 KiB

from abc import ABC, abstractmethod
from typing import List
import uuid
from mlagents_envs.side_channel import IncomingMessage, OutgoingMessage
from mlagents_envs.logging_util import get_logger
logger = get_logger(__name__)
class SideChannel(ABC):
"""
The side channel just get access to a bytes buffer that will be shared
between C# and Python. For example, We will create a specific side channel
for properties that will be a list of string (fixed size) to float number,
that can be modified by both C# and Python. All side channels are passed
to the Env object at construction.
"""
def __init__(self, channel_id: uuid.UUID):
self._channel_id: uuid.UUID = channel_id
self.message_queue: List[bytearray] = []
def queue_message_to_send(self, msg: OutgoingMessage) -> None:
"""
Queues a message to be sent by the environment at the next call to
step.
"""
self.message_queue.append(msg.buffer)
@abstractmethod
def on_message_received(self, msg: IncomingMessage) -> None:
"""
Is called by the environment to the side channel. Can be called
multiple times per step if multiple messages are meant for that
SideChannel.
"""
pass
@property
def channel_id(self) -> uuid.UUID:
"""
:return:The type of side channel used. Will influence how the data is
processed in the environment.
"""
return self._channel_id