Jerry Ling

`pyhf` is free via PyCall, for example, this [hello world](https://pyhf.readthedocs.io/en/v0.6.3/) from the main doc page: ```julia using PyCall pyhf = pyimport("pyhf") model = pyhf.simplemodels.uncorrelated_background( signal=PyVector([12.0, 11.0]), bkg=PyVector([50.0, 52.0]), bkg_uncertainty=PyVector([3.0, 7.0])...

re-useing `pyhf`'s likelihood seems to be free: ```julia using PyCall, BAT pyhf = pyimport("pyhf") # spec is a JSON/Dict per pyhf standard pdf_pyhf = pyhf.Model(spec) data_pyhf = [v_data; pdf_pyhf.config.auxdata] function...

Still need to figure out how to define the correct rrule, right now it would crash if you try to sample the posterior

https://github.com/TuringLang/AdvancedHMC.jl/issues/6 also @willtebbutt mentioned in https://github.com/TuringLang/Turing.jl/issues/815#issuecomment-503138636

`notebook_cmd` is no longer working

bump, really need port and `--no-browser`

still hitting this with Julia 1.5, workaround of disabling GC still works. But any progress?

while this may be good for some cases, I personally don't like the idea of 'breaking' a CSV file; sure, you can skip in some libraries, but it's still a...

if you have a second row after the header, it will be parsed into the data in most default cases. Also, these types are Julia specific, which is doubly 'bad'

@tanmaykm probably can close this? and make a new breaking release maybe? for both HTCondor and `qsub` related overhaul in #153