thatguy11325

Results 2 comments of thatguy11325

You could try to vectorize the generation of samps -> leaves a la what was done in evaluate? Though I'm unsure if that'll work if the sz's can vary.

I think it'd look like ```python leaves = stacked_sample.reshape(len(flat_space), batch, *next(flat_space.values()).shape) ``` I assume since sz is the same across all flat spaces, shape will be the same too.