Hypergrid.get_terminating_states_indices code

[younik]

I believe this function is not correct; it should be without argument and use self.sf (?)

torchgfn/src/gfn/gym/hypergrid.py

Lines 149 to 152 in c62b5b0

	def get_terminating_states_indices(
	self, states: DiscreteStates
	) -> TT["batch_shape", torch.long]:
	return self.get_states_indices(states)

[josephdviviano]

OK - actually, this is fine. It's supposed to return the indices of the final states as sampled. It's used in this file

torchgfn/src/gfn/utils/training.py

Line 20 in 8259a21

states_indices = env.get_terminating_states_indices(states).cpu().numpy().tolist()

to calculate the l1_dist with the true posterior.

however I'm going to patch up one or two things to make this more obvious and make the tests more robust.

[josephdviviano]

I've submitted a PR that hopefully addresses the confusion here.

[josephdviviano]

There's no need for this function when it just calls self.get_states_indices -- we should simply directly call this.

[josephdviviano]

Will follow up in new issue