nflows
nflows copied to clipboard
bayesiains
Update sample method docstring to describe shapes with more verbosity
I assume the existing ... parts in the comments are because we don't always assume the samples are necessarily a vector of length features, but could take other shapes. In which case replacing ... with features would not be appropriate. @arturbekasov ?
Yes, ... implies an arbitrary shape, and we make great use of this when working with, say, images, where inputs are of shape [B,C,H,W]. So indeed we shouldn't replace ... with features. IIRC the same is true for context (left a comment on the line), but would have to check.
I suspected as much with ... implying arbitrary shape, though I do think it's still helpful to add meaning to that shape within the output. Do you think it's enough to say something like "[context_size, ...], where ... represents an arbitrary shape feature vector"?
Hey @phinate, sorry for delay in response. Are you still interested in getting this merged in?
Do you think it's enough to say something like "
[context_size, ...], where...represents an arbitrary shape feature vector"?
Yes, this sounds good.
Hey @phinate, sorry for delay in response. Are you still interested in getting this merged in?
Do you think it's enough to say something like "
[context_size, ...], where...represents an arbitrary shape feature vector"?Yes, this sounds good.
Hi @arturbekasov, I've made those changes now. Let me know if this is good -- I vaguely remember there still could be other places in the codebase that re-use these docstrings.