GFNOrg/torchgfn

Fix `intro_gfn_continuous_line.ipynb`

Opened this issue · 1 comments

Running the second code cell returns TypeError: Can't instantiate abstract class Line with abstract methods backward_step, step

This is likely due to the recent changes in the environment API

Ah, so I also thought it was due to recent changes in the environment API, but it isn't -- it was never a functional notebook and is left as a todo for me.

I've added a note at the top of the notebook to this effect.

The goal of this notebook is to demonstrate how to handle continuous policies with more complex action spaces, e.g., sampling the likelihood of exiting at each step, sampling from a mixture beta distributions (instead of the other notebook where all trajectories are the same length, and the action space is just a single gaussian). The _simple.ipynb notebook is the Gaussian version (the one we presented at the workshop).

This notebook will be completed in a future PR, so I'll leave this open.