Getting Broken Pipe error when trying to implement my actions into the environment

[jpjayminpatel]

Hello,

I am trying to implement human input into the actions taken where I input the following code in the Tianshou>Data>collector.py file.

            try:
                print("Waiting for 3 seconds to get user input")
                sleep(3)
                if random:
                    try:
                        act_sample = [
                            self._action_space[i].sample() for i in ready_env_ids
                        ]
                    except TypeError:  # envpool's action space is not for per-env
                        act_sample = [self._action_space.sample() for _ in ready_env_ids]
                    act_sample = self.policy.map_action_inverse(act_sample)  # type: ignore
                    self.data.update(act=act_sample)
                else:
                    if no_grad:
                        with torch.no_grad():  # faster than retain_grad version
                            # self.data.obs will be used by agent to get result
                            result = self.policy(self.data, last_state)
                    else:
                        result = self.policy(self.data, last_state)
                    # update state / act / policy into self.data
                    policy = result.get("policy", Batch())
                    assert isinstance(policy, Batch)
                    state = result.get("state", None)
                    if state is not None:
                        policy.hidden_state = state  # save state into buffer
                    act = to_numpy(result.act)
                    if self.exploration_noise:
                        act = self.policy.exploration_noise(act, self.data)
                    self.data.update(policy=policy, act=act)

            except KeyboardInterrupt:
                action_input = int(input("Write action you want to input : "))
                action_input = [action_input]
                action_input = self.policy.map_action_inverse(action_input)
                self.data.update(act=action_input)

When i do this I am getting Broken Pipe error as follows:

Traceback (most recent call last):
  File "/home/jpatel/offworld/code/offworld-gym/examples/sim/qrdqn_monolith_discrete_sim.py", line 247, in <module>
    train_qrdqn(get_args())
  File "/home/jpatel/offworld/code/offworld-gym/examples/sim/qrdqn_monolith_discrete_sim.py", line 222, in train_qrdqn
    train_collector.collect(n_step=args.batch_size * args.training_num)
  File "/home/jpatel/qrdqn/lib/python3.10/site-packages/tianshou/data/collector.py", line 261, in collect
    result = self.env.step(action_remap, ready_env_ids)  # type: ignore
  File "/home/jpatel/qrdqn/lib/python3.10/site-packages/tianshou/env/venvs.py", line 248, in step
    self.workers[j].send(action[i])
  File "/home/jpatel/qrdqn/lib/python3.10/site-packages/tianshou/env/worker/subproc.py", line 190, in send
    self.parent_remote.send(["step", action])
  File "/home/jpatel/.pyenv/versions/3.10.4/lib/python3.10/multiprocessing/connection.py", line 211, in send
    self._send_bytes(_ForkingPickler.dumps(obj))
  File "/home/jpatel/.pyenv/versions/3.10.4/lib/python3.10/multiprocessing/connection.py", line 416, in _send_bytes
    self._send(header + buf)
  File "/home/jpatel/.pyenv/versions/3.10.4/lib/python3.10/multiprocessing/connection.py", line 373, in _send
    n = write(self._handle, buf)
BrokenPipeError: [Errno 32] Broken pipe

I would highly appreciate some help regarding the same.

[Trinkle23897]

Have you tried DummyVectorEnv instead of SubprocVectorEnv?