openai / atari-reset Goto Github PK

I am running the code from https://github.com/uber-research/atari-reset, but since you have not enabled issues there I write here instead.

When I try to run your robustification code with the default parameters from https://github.com/uber-research/go-explore i get the following error:

*** The MPI_Comm_test_inter() function was called before MPI_INIT was invoked.
*** This is disallowed by the MPI standard.
*** Your MPI job will now abort.
[hampusa:2940] Local abort before MPI_INIT completed completed successfully, but am not able to aggregate error messages, and not able to guarantee that all other processes were killed!

I am trying to run it on a single machine, and I have tried if setting --nenvs=1 would help but there is no difference.

Fatal Python error: Cannot recover from stack overflow.
Current thread 0x00007fd2edffb700 (most recent call first):
File "/home/xwq/anaconda3/envs/goexp/lib/python3.7/site-packages/gym/core.py", line 238 in getattr
File "/home/xwq/anaconda3/envs/goexp/lib/python3.7/site-packages/gym/core.py", line 238 in getattr
File "/home/xwq/anaconda3/envs/goexp/lib/python3.7/site-packages/gym/core.py", line 238 in getattr
...

when it run to : env = SubprocVecEnv([make_env(i + nenvs * hvd.rank()) for i in range(nenvs)])
I am running in single machine

Hello, due to the large GPU resource needed to reproduce the experiment, it is difficult for us to check the performance in paper and make some relative research about your idea. Could you provide model trained for Pitfall and have positive score to validate results? I would be very appreciated for that !!

When I run
python3 train_atari.py --game=Pong

It crashes with this callstack:

Traceback (most recent call last):
  File "/usr/lib/python3.6/multiprocessing/process.py", line 258, in _bootstrap
    self.run()
  File "/usr/lib/python3.6/multiprocessing/process.py", line 93, in run
    self._target(*self._args, **self._kwargs)
  File "/home/fanbingbing/ML/atari-reset/atari_reset/wrappers.py", line 469, in worker
    env = env_fn_wrapper.x()
  File "train_atari.py", line 34, in env_fn
    env = ReplayResetEnv(env, demo_file_name='demos/'+game_name+'.demo', seed=rank, workers_per_sp=workers_per_sp)
  File "/home/fanbingbing/ML/atari-reset/atari_reset/wrappers.py", line 107, in __init__
    assert len(rewards) == len(self.actions)
AssertionError

When I run
python3 train_atari.py --game=MontezumaRevenge

it works fine

Info:
Ubuntu 18.04
Python 3.6.6
synced to latest atari-reset 0c1b112
synced to latest baselines 28aca63

Hi,

Congratulations on your results! I have a couple of questions about your code:

1. How much time did the training on 128 GPUs take?
2. Is there any chance of retraining your code with 1-4 cards? I'm also doing research on hard Atari games and I'm planning on building an 8 GPU system, since I'm bothered by the vast amount of time it takes to do these experiments.

Thank you very much.