Edouard Leurent

Note that with configure, it's only after applying reset() that the observation space is changed, as expected: 

So without changing the training code, I just added this missing reset in the features_extration.create_environment as follows:  and things seem fine now: 

> I have done what you first suggest, do `env.reset()` before training (it doesn't even let you train without doing it). The problem is afterwards when I try to run...

I don't think there should be a difference, but I did my tests with the version at head, which you can try by uninstalling your current version and installing the...

Can you run this colab (copy of your notebook)? https://colab.research.google.com/drive/1y_xoKFiO_kRiF-_nkRAlN2wOmiuDEgt-?usp=sharing Everything seems to run fine...

I imported your featyres_extraction.ipynb into colab and it also works, can you try it? https://colab.research.google.com/drive/1KZR7wiX_jf56p0hoE863j_aFD5hKl-Vs?usp=sharing You have to upload the required files:  You should get something like this: 

Hi, sorry about that. This issue was fixed [~1 year ago](https://github.com/Farama-Foundation/HighwayEnv/commit/116c23feb686eea00e74a4756389d70352443ad3) on dev, but there hasn't been any stable release since. I'll make a new release with the patch, in...

You can replace the call to `configure` by `env = gym.make('highway-v0', config=config)`. I'll update the docs accordingly.

Cc @pseudo-rnd-thoughts any recommendation?