add continuous action space

README.md

@@ -22,7 +22,7 @@ MAPPO原版代码对于环境的封装过于复杂，本项目直接将环境封
 
 ## 用法
 
-- 环境部分是一个空的的实现，文件`light_mappo/envs/env_wrappers.py`里面环境部分的实现：[Code](https://github.com/tinyzqh/light_mappo/blob/main/envs/env_wrappers.py)
+- 环境部分是一个空的的实现，文件`light_mappo/envs/env_core.py`里面环境部分的实现：[Code](https://github.com/tinyzqh/light_mappo/blob/main/envs/env_wrappers.py)
 
 ```python
 class Env(object):
@@ -63,7 +63,7 @@ class Env(object):
 ```
 
 
-只需要编写这一部分的代码，就可以无缝衔接MAPPO。初始版本，后期这一部分会单独提出来。
+只需要编写这一部分的代码，就可以无缝衔接MAPPO。在env_core.py之后，单独提出来了两个文件env_discrete.py和env_continuous.py这两个文件用于封装处理动作空间和离散动作空间。在algorithms/utils/act.py中elif self.continuous_action:这个判断逻辑也是用来处理连续动作空间的。和runner/shared/env_runner.py部分的# TODO 这里改造成自己环境需要的形式即可都是用来处理连续动作空间的。
 
 
 ## Related Efforts

algorithms/utils/act.py

@@ -14,11 +14,13 @@ class ACTLayer(nn.Module):
         super(ACTLayer, self).__init__()
         self.mixed_action = False
         self.multi_discrete = False
+        self.continuous_action = False
 
         if action_space.__class__.__name__ == "Discrete":
             action_dim = action_space.n
             self.action_out = Categorical(inputs_dim, action_dim, use_orthogonal, gain)
         elif action_space.__class__.__name__ == "Box":
+            self.continuous_action = True
             action_dim = action_space.shape[0]
             self.action_out = DiagGaussian(inputs_dim, action_dim, use_orthogonal, gain)
         elif action_space.__class__.__name__ == "MultiBinary":
@@ -49,7 +51,7 @@ class ACTLayer(nn.Module):
         :return actions: (torch.Tensor) actions to take.
         :return action_log_probs: (torch.Tensor) log probabilities of taken actions.
         """
-        if self.mixed_action :
+        if self.mixed_action:
             actions = []
             action_log_probs = []
             for action_out in self.action_outs:
@@ -74,7 +76,16 @@ class ACTLayer(nn.Module):
 
             actions = torch.cat(actions, -1)
             action_log_probs = torch.cat(action_log_probs, -1)
-        
+        elif self.continuous_action:
+            # actions = []
+            # action_log_probs = []
+            action_logit = self.action_out(x)
+            actions = action_logit.mode() if deterministic else action_logit.sample()
+            action_log_probs = action_logit.log_probs(actions)
+            # actions.append(action.float())
+            # action_log_probs.append(action_log_prob)
+            # actions = torch.cat(actions, -1)
+            # action_log_probs = torch.sum(torch.cat(action_log_probs, -1), -1, keepdim=True)
         else:
             action_logits = self.action_out(x, available_actions)
             actions = action_logits.mode() if deterministic else action_logits.sample() 
@@ -150,7 +161,28 @@ class ACTLayer(nn.Module):
 
             action_log_probs = torch.cat(action_log_probs, -1) # ! could be wrong
             dist_entropy = torch.tensor(dist_entropy).mean()
-        
+
+        elif self.continuous_action:
+            # a, b = action.split((2, 1), -1)
+            # b = b.long()
+            # action = [a, b]
+            action_log_probs = []
+            dist_entropy = []
+            # for action_out, act in zip(self.action_outs, action):
+            action_logit = self.action_out(x)
+            action_log_probs.append(action_logit.log_probs(action))
+            if active_masks is not None:
+                if len(action_logit.entropy().shape) == len(active_masks.shape):
+                    dist_entropy.append((action_logit.entropy() * active_masks).sum() / active_masks.sum())
+                else:
+                    dist_entropy.append(
+                        (action_logit.entropy() * active_masks.squeeze(-1)).sum() / active_masks.sum())
+            else:
+                dist_entropy.append(action_logit.entropy().mean())
+
+            action_log_probs = torch.sum(torch.cat(action_log_probs, -1), -1, keepdim=True)
+            dist_entropy = dist_entropy[0] # / 2.0 + dist_entropy[1] / 0.98  # ! dosen't make sense
+
         else:
             action_logits = self.action_out(x, available_actions)
             action_log_probs = action_logits.log_probs(action)

config.py

@@ -176,7 +176,6 @@ def get_config():
     parser.add_argument("--num_env_steps", type=int, default=10e6,
                         help='Number of environment steps to train (default: 10e6)')
     parser.add_argument("--user_name", type=str, default='marl',help="[for wandb usage], to specify user's name for simply collecting training data.")
-    parser.add_argument("--use_wandb", action='store_false', default=False, help="[for wandb usage], by default True, will log date to wandb server. or else will use tensorboard to log data.")
 
     # env parameters
     parser.add_argument("--env_name", type=str, default='MyEnv', help="specify the name of environment")

envs/env.py

-"""
-# @Time    : 2021/7/2 5:22 下午
-# @Author  : hezhiqiang01
-# @Email   : hezhiqiang01@baidu.com
-# @File    : env.py
-"""
-
-import numpy as np
-
-
-class Env(object):
-    """
-    # 环境中的智能体
-    """
-    def __init__(self, i):
-        self.agent_num = 2  # 设置智能体(小飞机)的个数，这里设置为两个
-        self.obs_dim = 14  # 设置智能体的观测纬度
-        self.action_dim = 5  # 设置智能体的动作纬度，这里假定为一个五个纬度的
-
-    def reset(self):
-        """
-        # self.agent_num设定为2个智能体时，返回值为一个list，每个list里面为一个shape = (self.obs_dim, )的观测数据
-        """
-        sub_agent_obs = []
-        for i in range(self.agent_num):
-            sub_obs = np.random.random(size=(14, ))
-            sub_agent_obs.append(sub_obs)
-        return sub_agent_obs
-
-    def step(self, actions):
-        """
-        # self.agent_num设定为2个智能体时，actions的输入为一个2纬的list，每个list里面为一个shape = (self.action_dim, )的动作数据
-        # 默认参数情况下，输入为一个list，里面含有两个元素，因为动作纬度为5，所里每个元素shape = (5, )
-        """
-        sub_agent_obs = []
-        sub_agent_reward = []
-        sub_agent_done = []
-        sub_agent_info = []
-        for i in range(self.agent_num):
-            sub_agent_obs.append(np.random.random(size=(14,)))
-            sub_agent_reward.append([np.random.rand()])
-            sub_agent_done.append(False)
-            sub_agent_info.append({})
-
-        return [sub_agent_obs, sub_agent_reward, sub_agent_done, sub_agent_info]
\ No newline at end of file

envs/env_discrete.py

+"""
+# @Time    : 2021/7/2 5:22 下午
+# @Author  : hezhiqiang
+# @Email   : tinyzqh@163.com
+# @File    : env_discrete.py
+"""
+
+import gym
+from gym import spaces
+import numpy as np
+from envs.env_core import EnvCore
+
+
+class DiscreteActionEnv(object):
+    """对于离散动作环境的封装"""
+    def __init__(self):
+        self.env = EnvCore()
+        self.num_agent = self.env.agent_num
+
+        self.signal_obs_dim = self.env.obs_dim
+        self.signal_action_dim = self.env.action_dim
+
+        # if true, action is a number 0...N, otherwise action is a one-hot N-dimensional vector
+        self.discrete_action_input = False
+
+        self.movable = True
+
+        # configure spaces
+        self.action_space = []
+        self.observation_space = []
+        self.share_observation_space = []
+
+        share_obs_dim = 0
+        for agent in range(self.num_agent):
+            total_action_space = []
+
+            # physical action space
+            u_action_space = spaces.Discrete(self.signal_action_dim)  # 5个离散的动作
+
+            if self.movable:
+                total_action_space.append(u_action_space)
+
+            # total action space
+            if len(total_action_space) > 1:
+                # all action spaces are discrete, so simplify to MultiDiscrete action space
+                if all([isinstance(act_space, spaces.Discrete) for act_space in total_action_space]):
+                    act_space = MultiDiscrete([[0, act_space.n - 1] for act_space in total_action_space])
+                else:
+                    act_space = spaces.Tuple(total_action_space)
+                self.action_space.append(act_space)
+            else:
+                self.action_space.append(total_action_space[0])
+
+            # observation space
+            share_obs_dim += self.signal_obs_dim
+            self.observation_space.append(spaces.Box(low=-np.inf, high=+np.inf, shape=(self.signal_obs_dim,),
+                                                     dtype=np.float32))  # [-inf,inf]
+
+        self.share_observation_space = [spaces.Box(low=-np.inf, high=+np.inf, shape=(share_obs_dim,),
+                                                   dtype=np.float32) for _ in range(self.num_agent)]
+
+    def step(self, actions):
+        """
+        输入actions纬度假设：
+        # actions shape = (5, 2, 5)
+        # 5个线程的环境，里面有2个智能体，每个智能体的动作是一个one_hot的5维编码
+        """
+
+        results = self.env.step(actions)
+        obs, rews, dones, infos = results
+        return np.stack(obs), np.stack(rews), np.stack(dones), infos
+
+    def reset(self):
+        obs = self.env.reset()
+        return np.stack(obs)
+
+    def close(self):
+        pass
+
+    def render(self, mode="rgb_array"):
+        pass
+
+    def seed(self, seed):
+        pass
+
+class MultiDiscrete(gym.Space):
+    """
+    - The multi-discrete action space consists of a series of discrete action spaces with different parameters
+    - It can be adapted to both a Discrete action space or a continuous (Box) action space
+    - It is useful to represent game controllers or keyboards where each key can be represented as a discrete action space
+    - It is parametrized by passing an array of arrays containing [min, max] for each discrete action space
+       where the discrete action space can take any integers from `min` to `max` (both inclusive)
+    Note: A value of 0 always need to represent the NOOP action.
+    e.g. Nintendo Game Controller
+    - Can be conceptualized as 3 discrete action spaces:
+        1) Arrow Keys: Discrete 5  - NOOP[0], UP[1], RIGHT[2], DOWN[3], LEFT[4]  - params: min: 0, max: 4
+        2) Button A:   Discrete 2  - NOOP[0], Pressed[1] - params: min: 0, max: 1
+        3) Button B:   Discrete 2  - NOOP[0], Pressed[1] - params: min: 0, max: 1
+    - Can be initialized as
+        MultiDiscrete([ [0,4], [0,1], [0,1] ])
+    """
+
+    def __init__(self, array_of_param_array):
+        super().__init__()
+        self.low = np.array([x[0] for x in array_of_param_array])
+        self.high = np.array([x[1] for x in array_of_param_array])
+        self.num_discrete_space = self.low.shape[0]
+        self.n = np.sum(self.high) + 2
+
+    def sample(self):
+        """ Returns a array with one sample from each discrete action space """
+        # For each row: round(random .* (max - min) + min, 0)
+        random_array = np.random.rand(self.num_discrete_space)
+        return [int(x) for x in np.floor(np.multiply((self.high - self.low + 1.), random_array) + self.low)]
+
+    def contains(self, x):
+        return len(x) == self.num_discrete_space and (np.array(x) >= self.low).all() and (
+                    np.array(x) <= self.high).all()
+
+    @property
+    def shape(self):
+        return self.num_discrete_space
+
+    def __repr__(self):
+        return "MultiDiscrete" + str(self.num_discrete_space)
+
+    def __eq__(self, other):
+        return np.array_equal(self.low, other.low) and np.array_equal(self.high, other.high)
+
+
+if __name__ == "__main__":
+    DiscreteActionEnv().step(actions=None)
\ No newline at end of file

envs/env_wrappers.py

@@ -7,214 +7,294 @@ Modified from OpenAI Baselines code to work with multi-agent envs
 """
 
 import numpy as np
-import gym
-from gym import spaces
-from envs.env import Env
+import torch
+from multiprocessing import Process, Pipe
+from abc import ABC, abstractmethod
 
-
-class MultiDiscrete(gym.Space):
+def tile_images(img_nhwc):
     """
-    - The multi-discrete action space consists of a series of discrete action spaces with different parameters
-    - It can be adapted to both a Discrete action space or a continuous (Box) action space
-    - It is useful to represent game controllers or keyboards where each key can be represented as a discrete action space
-    - It is parametrized by passing an array of arrays containing [min, max] for each discrete action space
-       where the discrete action space can take any integers from `min` to `max` (both inclusive)
-    Note: A value of 0 always need to represent the NOOP action.
-    e.g. Nintendo Game Controller
-    - Can be conceptualized as 3 discrete action spaces:
-        1) Arrow Keys: Discrete 5  - NOOP[0], UP[1], RIGHT[2], DOWN[3], LEFT[4]  - params: min: 0, max: 4
-        2) Button A:   Discrete 2  - NOOP[0], Pressed[1] - params: min: 0, max: 1
-        3) Button B:   Discrete 2  - NOOP[0], Pressed[1] - params: min: 0, max: 1
-    - Can be initialized as
-        MultiDiscrete([ [0,4], [0,1], [0,1] ])
+    Tile N images into one big PxQ image
+    (P,Q) are chosen to be as close as possible, and if N
+    is square, then P=Q.
+    input: img_nhwc, list or array of images, ndim=4 once turned into array
+        n = batch index, h = height, w = width, c = channel
+    returns:
+        bigim_HWc, ndarray with ndim=3
+    """
+    img_nhwc = np.asarray(img_nhwc)
+    N, h, w, c = img_nhwc.shape
+    H = int(np.ceil(np.sqrt(N)))
+    W = int(np.ceil(float(N)/H))
+    img_nhwc = np.array(list(img_nhwc) + [img_nhwc[0]*0 for _ in range(N, H*W)])
+    img_HWhwc = img_nhwc.reshape(H, W, h, w, c)
+    img_HhWwc = img_HWhwc.transpose(0, 2, 1, 3, 4)
+    img_Hh_Ww_c = img_HhWwc.reshape(H*h, W*w, c)
+    return img_Hh_Ww_c
+
+class CloudpickleWrapper(object):
+    """
+    Uses cloudpickle to serialize contents (otherwise multiprocessing tries to use pickle)
     """
 
-    def __init__(self, array_of_param_array):
-        super().__init__()
-        self.low = np.array([x[0] for x in array_of_param_array])
-        self.high = np.array([x[1] for x in array_of_param_array])
-        self.num_discrete_space = self.low.shape[0]
-        self.n = np.sum(self.high) + 2
-
-    def sample(self):
-        """ Returns a array with one sample from each discrete action space """
-        # For each row: round(random .* (max - min) + min, 0)
-        random_array = np.random.rand(self.num_discrete_space)
-        return [int(x) for x in np.floor(np.multiply((self.high - self.low + 1.), random_array) + self.low)]
-
-    def contains(self, x):
-        return len(x) == self.num_discrete_space and (np.array(x) >= self.low).all() and (
-                    np.array(x) <= self.high).all()
-
-    @property
-    def shape(self):
-        return self.num_discrete_space
-
-    def __repr__(self):
-        return "MultiDiscrete" + str(self.num_discrete_space)
+    def __init__(self, x):
+        self.x = x
+
+    def __getstate__(self):
+        import cloudpickle
+        return cloudpickle.dumps(self.x)
+
+    def __setstate__(self, ob):
+        import pickle
+        self.x = pickle.loads(ob)
+
+def worker(remote, parent_remote, env_fn_wrapper):
+    parent_remote.close()
+    env = env_fn_wrapper.x()
+    while True:
+        cmd, data = remote.recv()
+        if cmd == 'step':
+            ob, reward, done, info = env.step(data)
+            if 'bool' in done.__class__.__name__:
+                if done:
+                    ob = env.reset()
+            else:
+                if np.all(done):
+                    ob = env.reset()
+
+            remote.send((ob, reward, done, info))
+        elif cmd == 'reset':
+            ob = env.reset()
+            remote.send((ob))
+        elif cmd == 'render':
+            if data == "rgb_array":
+                fr = env.render(mode=data)
+                remote.send(fr)
+            elif data == "human":
+                env.render(mode=data)
+        elif cmd == 'reset_task':
+            ob = env.reset_task()
+            remote.send(ob)
+        elif cmd == 'close':
+            env.close()
+            remote.close()
+            break
+        elif cmd == 'get_spaces':
+            remote.send((env.observation_space, env.share_observation_space, env.action_space))
+        else:
+            raise NotImplementedError
+
+class ShareVecEnv(ABC):
+    """
+    An abstract asynchronous, vectorized environment.
+    Used to batch data from multiple copies of an environment, so that
+    each observation becomes an batch of observations, and expected action is a batch of actions to
+    be applied per-environment.
+    """
+    closed = False
+    viewer = None
 
-    def __eq__(self, other):
-        return np.array_equal(self.low, other.low) and np.array_equal(self.high, other.high)
+    metadata = {
+        'render.modes': ['human', 'rgb_array']
+    }
 
+    def __init__(self, num_envs, observation_space, share_observation_space, action_space):
+        self.num_envs = num_envs
+        self.observation_space = observation_space
+        self.share_observation_space = share_observation_space
+        self.action_space = action_space
 
-class SubprocVecEnv(object):
-    def __init__(self, all_args):
+    @abstractmethod
+    def reset(self):
         """
-        envs: list of gym environments to run in subprocesses
+        Reset all the environments and return an array of
+        observations, or a dict of observation arrays.
+        If step_async is still doing work, that work will
+        be cancelled and step_wait() should not be called
+        until step_async() is invoked again.
         """
+        pass
 
-        self.env_list = [Env(i) for i in range(all_args.n_rollout_threads)]
-        self.num_envs = all_args.n_rollout_threads
-
-        self.num_agent = self.env_list[0].agent_num
-        self.signal_obs_dim = self.env_list[0].obs_dim
-        self.signal_action_dim = self.env_list[0].action_dim
-
-        self.u_range = 1.0  # control range for continuous control
-        self.movable = True
-
-        # environment parameters
-        # self.discrete_action_space = True
-        self.discrete_action_space = True
-
-        # if true, action is a number 0...N, otherwise action is a one-hot N-dimensional vector
-        self.discrete_action_input = False
-        # if true, even the action is continuous, action will be performed discretely
-        self.force_discrete_action = False
-
-        # configure spaces
-        self.action_space = []
-        self.observation_space = []
-        self.share_observation_space = []
-        share_obs_dim = 0
-        for agent in range(self.num_agent):
-            total_action_space = []
+    @abstractmethod
+    def step_async(self, actions):
+        """
+        Tell all the environments to start taking a step
+        with the given actions.
+        Call step_wait() to get the results of the step.
+        You should not call this if a step_async run is
+        already pending.
+        """
+        pass
 
-            # physical action space
-            if self.discrete_action_space:
-                u_action_space = spaces.Discrete(self.signal_action_dim)  # 5个离散的动作
-            else:
-                u_action_space = spaces.Box(low=-self.u_range, high=+self.u_range, shape=(2,), dtype=np.float32)  # [-1,1]
-            if self.movable:
-                total_action_space.append(u_action_space)
-
-            # total action space
-            if len(total_action_space) > 1:
-                # all action spaces are discrete, so simplify to MultiDiscrete action space
-                if all([isinstance(act_space, spaces.Discrete) for act_space in total_action_space]):
-                    act_space = MultiDiscrete([[0, act_space.n - 1] for act_space in total_action_space])
-                else:
-                    act_space = spaces.Tuple(total_action_space)
-                self.action_space.append(act_space)
-            else:
-                self.action_space.append(total_action_space[0])
+    @abstractmethod
+    def step_wait(self):
+        """
+        Wait for the step taken with step_async().
+        Returns (obs, rews, dones, infos):
+         - obs: an array of observations, or a dict of
+                arrays of observations.
+         - rews: an array of rewards
+         - dones: an array of "episode done" booleans
+         - infos: a sequence of info objects
+        """
+        pass
 
-            # observation space
-            share_obs_dim += self.signal_obs_dim
-            self.observation_space.append(spaces.Box(low=-np.inf, high=+np.inf, shape=(self.signal_obs_dim,),
-                                                     dtype=np.float32))  # [-inf,inf]
+    def close_extras(self):
+        """
+        Clean up the  extra resources, beyond what's in this base class.
+        Only runs when not self.closed.
+        """
+        pass
 
-        self.share_observation_space = [spaces.Box(low=-np.inf, high=+np.inf, shape=(share_obs_dim,),
-                                                   dtype=np.float32) for _ in range(self.num_agent)]
+    def close(self):
+        if self.closed:
+            return
+        if self.viewer is not None:
+            self.viewer.close()
+        self.close_extras()
+        self.closed = True
 
     def step(self, actions):
         """
-        输入actions纬度假设：
-        # actions shape = (5, 2, 5)
-        # 5个线程的环境，里面有2个智能体，每个智能体的动作是一个one_hot的5维编码
+        Step the environments synchronously.
+        This is available for backwards compatibility.
+        """
+        self.step_async(actions)
+        return self.step_wait()
+
+    def render(self, mode='human'):
+        imgs = self.get_images()
+        bigimg = tile_images(imgs)
+        if mode == 'human':
+            self.get_viewer().imshow(bigimg)
+            return self.get_viewer().isopen
+        elif mode == 'rgb_array':
+            return bigimg
+        else:
+            raise NotImplementedError
+
+    def get_images(self):
+        """
+        Return RGB images from each environment
         """
+        raise NotImplementedError
 
-        results = [env.step(action) for env, action in zip(self.env_list, actions)]
+    @property
+    def unwrapped(self):
+        if isinstance(self, VecEnvWrapper):
+            return self.venv.unwrapped
+        else:
+            return self
+
+    def get_viewer(self):
+        if self.viewer is None:
+            from gym.envs.classic_control import rendering
+            self.viewer = rendering.SimpleImageViewer()
+        return self.viewer
+
+class SubprocVecEnv(ShareVecEnv):
+    def __init__(self, env_fns, spaces=None):
+        """
+        envs: list of gym environments to run in subprocesses
+        """
+        self.waiting = False
+        self.closed = False
+        nenvs = len(env_fns)
+        self.remotes, self.work_remotes = zip(*[Pipe() for _ in range(nenvs)])
+        self.ps = [Process(target=worker, args=(work_remote, remote, CloudpickleWrapper(env_fn)))
+                   for (work_remote, remote, env_fn) in zip(self.work_remotes, self.remotes, env_fns)]
+        for p in self.ps:
+            p.daemon = True  # if the main process crashes, we should not cause things to hang
+            p.start()
+        for remote in self.work_remotes:
+            remote.close()
+
+        self.remotes[0].send(('get_spaces', None))
+        observation_space, share_observation_space, action_space = self.remotes[0].recv()
+        ShareVecEnv.__init__(self, len(env_fns), observation_space,
+                             share_observation_space, action_space)
+
+    def step_async(self, actions):
+        for remote, action in zip(self.remotes, actions):
+            remote.send(('step', action))
+        self.waiting = True
+
+    def step_wait(self):
+        results = [remote.recv() for remote in self.remotes]
+        self.waiting = False
         obs, rews, dones, infos = zip(*results)
         return np.stack(obs), np.stack(rews), np.stack(dones), infos
 
     def reset(self):
-        obs = [env.reset() for env in self.env_list]
+        for remote in self.remotes:
+            remote.send(('reset', None))
+        obs = [remote.recv() for remote in self.remotes]
         return np.stack(obs)
 
+
+    def reset_task(self):
+        for remote in self.remotes:
+            remote.send(('reset_task', None))
+        return np.stack([remote.recv() for remote in self.remotes])
+
     def close(self):
-        pass
+        if self.closed:
+            return
+        if self.waiting:
+            for remote in self.remotes:
+                remote.recv()
+        for remote in self.remotes:
+            remote.send(('close', None))
+        for p in self.ps:
+            p.join()
+        self.closed = True
 
     def render(self, mode="rgb_array"):
-        pass
-
+        for remote in self.remotes:
+            remote.send(('render', mode))
+        if mode == "rgb_array":
+            frame = [remote.recv() for remote in self.remotes]
+            return np.stack(frame)
 
 # single env
-class DummyVecEnv(object):
-    def __init__(self, all_args):
-        """
-        envs: list of gym environments to run in subprocesses
-        """
-
-        self.env_list = [Env(i) for i in range(all_args.n_eval_rollout_threadss)]
-        self.num_envs = all_args.n_rollout_threads
-
-        self.num_agent = self.env_list[0].agent_num
-
-        self.u_range = 1.0  # control range for continuous control
-        self.movable = True
-
-        # environment parameters
-        self.discrete_action_space = True
-
-        # if true, action is a number 0...N, otherwise action is a one-hot N-dimensional vector
-        self.discrete_action_input = False
-        # if true, even the action is continuous, action will be performed discretely
-        self.force_discrete_action = False
-        # in this env, force_discrete_action == False��because world do not have discrete_action
-
-        # configure spaces
-        self.action_space = []
-        self.observation_space = []
-        self.share_observation_space = []
-        share_obs_dim = 0
-        for agent_num in range(self.num_agent):
-            total_action_space = []
-
-            # physical action space
-            if self.discrete_action_space:
-                u_action_space = spaces.Discrete(5)  # 5个离散的动作
+class DummyVecEnv(ShareVecEnv):
+    def __init__(self, env_fns):
+        self.envs = [fn() for fn in env_fns]
+        env = self.envs[0]
+        ShareVecEnv.__init__(self, len(
+            env_fns), env.observation_space, env.share_observation_space, env.action_space)
+        self.actions = None
+
+    def step_async(self, actions):
+        self.actions = actions
+
+    def step_wait(self):
+        results = [env.step(a) for (a, env) in zip(self.actions, self.envs)]
+        obs, rews, dones, infos = map(np.array, zip(*results))
+
+        for (i, done) in enumerate(dones):
+            if 'bool' in done.__class__.__name__:
+                if done:
+                    obs[i] = self.envs[i].reset()
             else:
-                u_action_space = spaces.Box(low=-self.u_range, high=+self.u_range, shape=(2,), dtype=np.float32)  # [-1,1]
-            if self.movable:
-                total_action_space.append(u_action_space)
-
-            # total action space
-            if len(total_action_space) > 1:
-                # all action spaces are discrete, so simplify to MultiDiscrete action space
-                if all([isinstance(act_space, spaces.Discrete) for act_space in total_action_space]):
-                    act_space = MultiDiscrete([[0, act_space.n - 1] for act_space in total_action_space])
-                else:
-                    act_space = spaces.Tuple(total_action_space)
-                self.action_space.append(act_space)
-            else:
-                self.action_space.append(total_action_space[0])
-
-            # observation space
-            obs_dim = 14  # 单个智能体的观测维度
-            share_obs_dim += obs_dim
-            self.observation_space.append(spaces.Box(low=-np.inf, high=+np.inf, shape=(obs_dim,), dtype=np.float32))  # [-inf,inf]
-
-        self.share_observation_space = [spaces.Box(low=-np.inf, high=+np.inf, shape=(share_obs_dim,),
-                                                   dtype=np.float32) for _ in range(self.num_agent)]
-
-    def step(self, actions):
-        """
-        输入actions纬度假设：
-        # actions shape = (5, 2, 5)
-        # 5个线程的环境，里面有2个智能体，每个智能体的动作是一个one_hot的5维编码
-        """
+                if np.all(done):
+                    obs[i] = self.envs[i].reset()
 
-        results = [env.step(action) for env, action in zip(self.env_list, actions)]
-        obs, rews, dones, infos = zip(*results)
-        return np.stack(obs), np.stack(rews), np.stack(dones), infos
+        self.actions = None
+        return obs, rews, dones, infos
 
     def reset(self):
-        obs = [env.reset() for env in self.env_list]
-        return np.stack(obs)
+        obs = [env.reset() for env in self.envs]
+        return np.array(obs)
 
     def close(self):
-        pass
-
-    def render(self, mode="rgb_array"):
-        pass
+        for env in self.envs:
+            env.close()
+
+    def render(self, mode="human"):
+        if mode == "rgb_array":
+            return np.array([env.render(mode=mode) for env in self.envs])
+        elif mode == "human":
+            for env in self.envs:
+                env.render(mode=mode)
+        else:
+            raise NotImplementedError
\ No newline at end of file

runner/shared/base_runner.py

-import wandb
 import os
 import numpy as np
 import torch
@@ -37,7 +36,6 @@ class Runner(object):
         self.n_render_rollout_threads = self.all_args.n_render_rollout_threads
         self.use_linear_lr_decay = self.all_args.use_linear_lr_decay
         self.hidden_size = self.all_args.hidden_size
-        self.use_wandb = self.all_args.use_wandb
         self.use_render = self.all_args.use_render
         self.recurrent_N = self.all_args.recurrent_N
 
@@ -50,18 +48,14 @@ class Runner(object):
         # dir
         self.model_dir = self.all_args.model_dir
 
-        if self.use_wandb:
-            self.save_dir = str(wandb.run.dir)
-            self.run_dir = str(wandb.run.dir)
-        else:
-            self.run_dir = config["run_dir"]
-            self.log_dir = str(self.run_dir / 'logs')
-            if not os.path.exists(self.log_dir):
-                os.makedirs(self.log_dir)
-            self.writter = SummaryWriter(self.log_dir)
-            self.save_dir = str(self.run_dir / 'models')
-            if not os.path.exists(self.save_dir):
-                os.makedirs(self.save_dir)
+        self.run_dir = config["run_dir"]
+        self.log_dir = str(self.run_dir / 'logs')
+        if not os.path.exists(self.log_dir):
+            os.makedirs(self.log_dir)
+        self.writter = SummaryWriter(self.log_dir)
+        self.save_dir = str(self.run_dir / 'models')
+        if not os.path.exists(self.save_dir):
+            os.makedirs(self.save_dir)
 
         from algorithms.algorithm.r_mappo import RMAPPO as TrainAlgo
         from algorithms.algorithm.rMAPPOPolicy import RMAPPOPolicy as Policy
@@ -146,10 +140,7 @@ class Runner(object):
         :param total_num_steps: (int) total number of training env steps.
         """
         for k, v in train_infos.items():
-            if self.use_wandb:
-                wandb.log({k: v}, step=total_num_steps)
-            else:
-                self.writter.add_scalars(k, {k: v}, total_num_steps)
+            self.writter.add_scalars(k, {k: v}, total_num_steps)
 
     def log_env(self, env_infos, total_num_steps):
         """
@@ -159,7 +150,4 @@ class Runner(object):
         """
         for k, v in env_infos.items():
             if len(v)>0:
-                if self.use_wandb:
-                    wandb.log({k: np.mean(v)}, step=total_num_steps)
-                else:
-                    self.writter.add_scalars(k, {k: np.mean(v)}, total_num_steps)
+                self.writter.add_scalars(k, {k: np.mean(v)}, total_num_steps)

runner/shared/env_runner.py

@@ -16,7 +16,6 @@ import time
 import numpy as np
 import torch
 from runner.shared.base_runner import Runner
-import wandb
 import imageio
 
 
@@ -133,9 +132,12 @@ class EnvRunner(Runner):
                 else:
                     actions_env = np.concatenate((actions_env, uc_actions_env), axis=2)
         elif self.envs.action_space[0].__class__.__name__ == 'Discrete':
+            # actions  --> actions_env : shape:[10, 1] --> [5, 2, 5]
             actions_env = np.squeeze(np.eye(self.envs.action_space[0].n)[actions], 2)
         else:
-            raise NotImplementedError
+            # TODO 这里改造成自己环境需要的形式即可
+            actions_env = actions
+            # raise NotImplementedError
 
         return values, actions, action_log_probs, rnn_states, rnn_states_critic, actions_env
 

train/train.py

 """
 # @Time    : 2021/6/30 10:07 下午
-# @Author  : hezhiqiang01
-# @Email   : hezhiqiang01@baidu.com
+# @Author  : hezhiqiang
+# @Email   : tinyzqh@163.com
 # @File    : train.py
 """
 
 # !/usr/bin/env python
 import sys
 import os
-import wandb
 import socket
 import setproctitle
 import numpy as np
@@ -21,11 +20,27 @@ from envs.env_wrappers import SubprocVecEnv, DummyVecEnv
 
 
 def make_train_env(all_args):
-    return SubprocVecEnv(all_args)
+    def get_env_fn(rank):
+        def init_env():
+            # from envs.env_continuous import ContinuousActionEnv
+            # env = ContinuousActionEnv()
+            from envs.env_discrete import DiscreteActionEnv
+            env = DiscreteActionEnv()
+            env.seed(all_args.seed + rank * 1000)
+            return env
+        return init_env
+    return DummyVecEnv([get_env_fn(i) for i in range(all_args.n_rollout_threads)])
 
 
 def make_eval_env(all_args):
-    return DummyVecEnv(all_args)
+    def get_env_fn(rank):
+        def init_env():
+            from envs.env_discrete import DiscreteActionEnv
+            env = DiscreteActionEnv()
+            env.seed(all_args.seed + rank * 1000)
+            return env
+        return init_env
+    return DummyVecEnv([get_env_fn(i) for i in range(all_args.n_rollout_threads)])
 
 
 def parse_args(args, parser):
@@ -72,32 +87,18 @@ def main(args):
     if not run_dir.exists():
         os.makedirs(str(run_dir))
 
-    # wandb
-    if all_args.use_wandb:
-        run = wandb.init(config=all_args,
-                         project=all_args.env_name,
-                         entity=all_args.user_name,
-                         notes=socket.gethostname(),
-                         name=str(all_args.algorithm_name) + "_" +
-                              str(all_args.experiment_name) +
-                              "_seed" + str(all_args.seed),
-                         group=all_args.scenario_name,
-                         dir=str(run_dir),
-                         job_type="training",
-                         reinit=True)
+    if not run_dir.exists():
+        curr_run = 'run1'
     else:
-        if not run_dir.exists():
+        exst_run_nums = [int(str(folder.name).split('run')[1]) for folder in run_dir.iterdir() if
+                         str(folder.name).startswith('run')]
+        if len(exst_run_nums) == 0:
             curr_run = 'run1'
         else:
-            exst_run_nums = [int(str(folder.name).split('run')[1]) for folder in run_dir.iterdir() if
-                             str(folder.name).startswith('run')]
-            if len(exst_run_nums) == 0:
-                curr_run = 'run1'
-            else:
-                curr_run = 'run%i' % (max(exst_run_nums) + 1)
-        run_dir = run_dir / curr_run
-        if not run_dir.exists():
-            os.makedirs(str(run_dir))
+            curr_run = 'run%i' % (max(exst_run_nums) + 1)
+    run_dir = run_dir / curr_run
+    if not run_dir.exists():
+        os.makedirs(str(run_dir))
 
     setproctitle.setproctitle(str(all_args.algorithm_name) + "-" + \
                               str(all_args.env_name) + "-" + str(all_args.experiment_name) + "@" + str(
@@ -136,11 +137,8 @@ def main(args):
     if all_args.use_eval and eval_envs is not envs:
         eval_envs.close()
 
-    if all_args.use_wandb:
-        run.finish()
-    else:
-        runner.writter.export_scalars_to_json(str(runner.log_dir + '/summary.json'))
-        runner.writter.close()
+    runner.writter.export_scalars_to_json(str(runner.log_dir + '/summary.json'))
+    runner.writter.close()
 
 
 if __name__ == "__main__":