Closed genmeblog closed 5 years ago
Looks like part 1 of https://github.com/probcomp/metaprob/issues/143#issue-433660738 was fixed in #146.
@genmeblog, what do you mean by part 2 of https://github.com/probcomp/metaprob/issues/143#issue-433660738? I don't see a place where n-particles
is missing, can you link to a line of the code?
Ah I see. So you're suggesting to change that line to :n-particles n-particles))))
, otherwise importance sampling will always use the default value of 1?
Well... it's a syntax error. importance-resampling
expects map (for associative destructuring). When you call this function in current shape you'll get an execution error (IllegalArgumentException
).
So yes, this line has to be :n-particles n-particles))))
to make it work.
Thanks, @genmeblog.
@bzinberg FYI, this has been fixed since March 30 on the autodiff branch: https://github.com/probcomp/metaprob/commit/ca437c4a5a34d262eeda82713e7abe4e6ca57e81#diff-6b7158b9913b99f823c5cb9670ba6072R166
Ah, got it. Thanks @genmeblog, I'm new to Clojure :slightly_smiling_face:
Alex, good to know. Looks like there are a few miscellaneous fixes in that branch which should probably get separate PRs.
Hi, I was trying to run examples and there are some minor issues which can be easily fixed:
infer-and-score
(aide or curve-fitting)eq-importance-assay
: missing:n-particles
key