Stuck on an issue?

Lightrun Answers was designed to reduce the constant googling that comes with debugging 3rd party libraries. It collects links to all the places you might be looking at while hunting down a tough bug.

And, if you’re still stuck at the end, we’re happy to hop on a call to see how we can help out.

Processes not terminated

See original GitHub issue

BUG REPORT

High Level Description I use SMARTS for training imitation learning agents with NGSIM dataset. When training is finished, SMARTS will leave some unterminated processes.

SMARTS version 0.4.15

Steps to reproduce the bug I use the multiprocessing package to open multiple SMARTS instances for parallel sampling.

And I created each simulator using:

self.smarts = SMARTS(
    agent_interfaces={},
    traffic_sim=None,
    visdom=None,
    envision=None,
    zoo_addrs=None,
)

Resulting and expected behaviour There are multiple processes like smarts/zoo/worker.py --port xxxxx not terminated after the training stops.

Error logs and screenshots 截屏2021-06-02 11 16 23

System information

Operating system & Environment [Please note if using Docker]: Ubuntu 16.04
Python version [Please note if using Jupyter] 3.7.3
SUMO version 1.8.0

Issue Analytics

State:
Created 2 years ago
Comments:6 (6 by maintainers)

Top GitHub Comments

1reaction

Adaickalavancommented, Sep 16, 2021

Hi @Zbzhu99,

We have now created ParallelEnv class to simulate multiple SMARTS environments in parallel, with synchronous or asynchronous episodes, to collect observations at a faster rate.

See pull request #1087 for details.

You may consider using ParallelEnv class to run mulitple SMARTS for parallel rollouts, instead of writing your own code. We hope this new feature would be helpful and simplify your usage of SMARTS.

0reactions

Adaickalavancommented, Jul 2, 2021

Closed by pull request #924 .