Colab Notebook (baseline submission)

Hello, is it possible to share a dummy submission notebook that can be run (train, eval, submit) by google colab?
Thanks.

Hi – yes, we’re working on adding this. There are a few technical quirks to sort out.

3 Likes

Thank you, saves me a lot of trouble!
And while you are at it, I am looking for a way to see the ttyrec2 files to visually inspect the agent. I save them now to my Google drive and try to watch them in termrec on my pc, but the files are empty. I was hoping a virtual environment (pyvirtualdisplay, piglet, ffmpeg) would do the trick but no luck so far. Maybe you have a solution already?

One more :slight_smile:
if I run this little piece of code in Colab I get the same output of the position of the agent, while if I render the screen I do see that the agent moved north (e.g. visually checked by rendering the screen). Is it a bug, or am I making a mistake?

Show the map (‘glyphs’, ‘chars’, ‘colors’)

env = gym.make(“NetHackChallenge-v0”, savedir=None)
obs=env.reset() # each reset generates a new dungeon
print (‘The position of the agent is row:’, obs[‘blstats’][1], ’ and column: ', obs[‘blstats’][0] )
obs, reward, done, info = env.step(0) # move north
print (‘The position of the agent is row:’, obs[‘blstats’][1], ’ and column: ', obs[‘blstats’][0] )

Gets me:
The position of the agent is row: 6 and column: 14
The position of the agent is row: 6 and column: 14
(while I didn’t hit a wall or something obvious)

Edit: I see this one is a known issue and already solved in the master but not yet in the “nle.git@eric” branch. Would be nice if it is solved there as well because I like the way the notebook renders.
Below is how my Colab notebook code looks like:
‘’’
%%capture
!pip install -U cmake
!apt update -qq && apt install -qq -y flex bison libbz2-dev libglib2.0 libsm6 libxext6
!pip install -U pip
#!pip install git+https://github.com/facebookresearch/nle.git@eric/notebook-render # this can render notebooks
!pip install git+https://github.com/facebookresearch/nle.git # latest version
‘’’

Hello @charel_van_hoof

You probably want to post this in https://discourse.aicrowd.com/c/neurips-2021-nethack-challenge/