Front page Forums 🤖 🚗 AI Driving Olympics 🏆 Embodied tasks 600 Timeout when submit

This topic contains 5 replies, has 3 voices, and was last updated by  heyt0ny 1 month ago.

Viewing 6 posts - 1 through 6 (of 6 total)
  • Author
    Posts
  • #29618

    heyt0ny
    Participant

    Hi!

    I’m getting 600 timeout error – https://challenges.duckietown.org/v3/humans/submissions/832
    What may this mean?

    Thank you so much for answering the questions)

    #29619

    Bhairav Mehta
    Participant

    I believe this is the same issue as your other post I answered this morning.
    Can you post a link to the solution.py you are using?

    #29628

    heyt0ny
    Participant

    So it’s really strange error and it’s not about solution file

    So I have 2 models (A and B). They both return the same type and dimensions of action. The both run locally successfully.

    But the model A is failed on step1-simulation. I dind’t change the architecture of model A for at least 1 week (I only changed weights). So I decided to use old weights from successful submits (~ 3 days ago) — and it still fails.

    Then I decided to turn off weights load — and the model A is submitted successfully. But I’m 100% sure it’s not the problem with weights load, because model B uses the same code for weights load.

    I’m really confused about it.

    Is there any way I can download my submits to make sure I didn’t change anything.. ?

    #29629

    lavoiems
    Participant

    Hey Tony, so my extremly wild guess would be that the docker image that your are using is using python2 and your weight were saved, using pickle, on python3.

    If this is the case, you can fix it by changing the first line of your Dockerfile to FROM pytorch/pytorch:0.4.1-cuda9-cudnn7-runtime

    Cheers

    #29872

    heyt0ny
    Participant

    Thank you lavoiems, good point!

    Tried it, but the same failure (

    #29877

    heyt0ny
    Participant

    seems like I was able to fix this, by adding frame skip:

    
        nested_break = False
        try:
            cis.info('Reset environment')
            observation = env.reset()
            while True:
                action = model.predict(observation)
    
                if debug:
                    env.render()
                for i in range(config['frame_skip']):
                    observation, reward, done, info = env.step(action)
                    if 'simulation_done' in info:
                        cis.info('simulation_done received.')
                        nested_break = True
                        break
                    if done:
                        cis.info('Episode done; calling reset()')
                        env.reset()
                        break
                if nested_break:
                    break
    
        finally:
            # release CPU/GPU resources, let's be friendly with other users that may need them
            cis.info('Releasing resources')
            try:
                model.close()
            except Exception as e:
                cis.error(e)
                msg = 'Could not call model.close():\n%s' % traceback.format_exc()
                cis.error(msg)
        cis.info('Graceful exit of solve()')
    

    magic !

Viewing 6 posts - 1 through 6 (of 6 total)

You must be logged in to reply to this topic.

Close Menu