search

openai / large-scale-curiosity

Code for the paper "Large-Scale Study of Curiosity-Driven Learning"
https://arxiv.org/abs/1808.04355
805 stars 180 forks source link

noisy TV unity environment #5

Open qdrn opened 6 years ago

qdrn commented 6 years ago

Hi, Would you consider releasing the noisy TV unity environment that you used in your experiments at the end of the paper ? Thanks

yburda commented 6 years ago

cc @pathak22 - what are your thoughts?

qdrn commented 5 years ago

Any news ?

qqadssp commented 5 years ago

also need noisy TV environment, where can I find it? @christopherhesse @pathak22

hcwang95 commented 5 years ago

It will be great to have the vizdoom with the noisy TV. Thanks a lot!

utanashati commented 4 years ago

@pathak22 @yburda Really, why would you want to keep it from the community? It'd be a nice baseline for people to build upon.

FoxForge commented 2 years ago

Please could we see this environment, it would be fantastic to write c# wrappers and APIs for intrinsic learning for game devs. Doesn't matter if it's outdated by now, just seeing how it interfaces is the key point

luchris429 commented 2 years ago

I created the original environment for this paper and released it here a few years ago here: https://github.com/luchris429/tv_maze

Hopefully you all find the environment useful!

For context: I was 19 when I made it. I was told I would be made an author if I made the environment, which then didn't happen since I was then told it was not a significant contribution to the paper...

qdrn commented 2 years ago

I created the original environment for this paper and released it here a few years ago here: https://github.com/luchris429/tv_maze

Thanks for providing the environment!

For context: I was 19 when I made it. I was told I would be made an author if I made the environment, which then didn't happen since I was then told it was not a significant contribution to the paper...

Ouch. That hurts. Unfortunately this happens more often than not. Unfortunate that these sort of things happen at OpenAI.

luchris429 commented 2 years ago

While I am still unhappy about what happened years ago, my intention wasn't to call out the authors -- it was just to pre-empt the questions of: "Why does this random guy have the original environment?" and "Is it the legitimate environment?"

FoxForge commented 2 years ago

Hopefully you all find the environment useful!

This is just what i was looking for! This will be mega helpful, you deserve massive credit here for bridging the gap to make this applicable in modern day games for AI.

luchris429 commented 2 years ago

The unity-python interface part (mostly) isn't my code! It's from an old version of this:

https://github.com/Unity-Technologies/ml-agents

(I interned at Unity and helped test the first versions of that project a long time ago -- their current repo is likely much more useful and up-to-date if you want an interface that works with current Unity versions)

I did spend quite a bit of time building the actual tv maze environment in Unity though.