Add Autodiff compatable emulator and Barker gradient-based sampler (#339

)

* add Forward diff and selection of gradient-based Samplers

* added ESJD and test for it

* format

* new protocols to modularize the autodiff packages

* typo

* allow kwargs and duals in Emulator arguments

* add abstract GPs as an emulator option, using provieded lengthscales etc as keywords

* add kwargs into RF build_models too

* add ew deps for emulators

* add convenience to build AGP from GPJL and implement 1D->1D unit test

* format and typo

* format

* add tests for 2d->2d GP and improve interface for AbstractGP

* rtol->atol

* add test for RWM and PCN with AGP and no derivatives

* updated proposals from Kotas latest CES repo, also added target_acc to find stepsize

* framework to test the ForwardDiff schemes

* typo

* Seemingly sign errors in derivative calculations

* one test set per method

* add L into HMC as parameter, sign changes for autodiff.

* typo

* remove all non-barker gradient methods for now

* format

* add docs Manifest to inc. AbstractGP

* docs API fix

* rm long comment

* format

* add reversediff compatability

* Project toml and format

* improve code-cov except for hessians

* refined error messages

* better stepsize guess to reduce (Mac) runtest time

Project.toml

@@ -4,13 +4,16 @@ authors = ["CLIMA contributors <clima-software@caltech.edu>"]
 version = "0.6.1"
 
 [deps]
+AbstractGPs = "99985d1d-32ba-4be9-9821-2ec096f28918"
 AbstractMCMC = "80f14c24-f653-4e6a-9b94-39d6b0f70001"
 AdvancedMH = "5b7e9947-ddc0-4b3f-9b55-0d8042f74170"
 Conda = "8f4d0f93-b110-5947-807f-2305c1781a2d"
 Distributions = "31c24e10-a181-5473-b8eb-7969acd0382f"
 DocStringExtensions = "ffbed154-4ef7-542d-bbb7-c09d3a79fcae"
 EnsembleKalmanProcesses = "aa8a2aa5-91d8-4396-bcef-d4f2ec43552d"
+ForwardDiff = "f6369f11-7733-5829-9624-2563aa707210"
 GaussianProcesses = "891a1506-143c-57d2-908e-e1f8e92e6de9"
+KernelFunctions = "ec8451be-7e33-11e9-00cf-bbf324bd1392"
 LinearAlgebra = "37e2e46d-f89d-539d-b4ee-838fcccc9c8e"
 MCMCChains = "c7f686f2-ff18-58e9-bc7b-31028e88f75d"
 Pkg = "44cfe95a-1eb2-52ea-b672-e2afdf69b78f"
@@ -19,25 +22,30 @@ ProgressBars = "49802e3a-d2f1-5c88-81d8-b72133a6f568"
 PyCall = "438e738f-606a-5dbb-bf0a-cddfbfd45ab0"
 Random = "9a3f8284-a2c9-5f02-9a11-845980a1fd5c"
 RandomFeatures = "36c3bae2-c0c3-419d-b3b4-eebadd35c5e5"
+ReverseDiff = "37e2e3b7-166d-5795-8a7a-e32c996b4267"
 ScikitLearn = "3646fa90-6ef7-5e7e-9f22-8aca16db6324"
 StableRNGs = "860ef19b-820b-49d6-a774-d7a799459cd3"
 Statistics = "10745b16-79ce-11e8-11f9-7d13ad32a3b2"
 StatsBase = "2913bbd2-ae8a-5f71-8c99-4fb6c76f3a91"
 
 [compat]
+AbstractGPs = "0.5.21"
 AbstractMCMC = "3.3, 4, 5"
 AdvancedMH = "0.6, 0.7, 0.8"
 Conda = "1.7"
 Distributions = "0.24, 0.25"
 DocStringExtensions = "0.8, 0.9"
 EnsembleKalmanProcesses = "2"
+ForwardDiff = "0.10.38"
 GaussianProcesses = "0.12"
+KernelFunctions = "0.10.64"
 MCMCChains = "4.14, 5, 6"
 Printf = "1"
 ProgressBars = "1"
 PyCall = "1.93"
 Random = "1"
 RandomFeatures = "0.3"
+ReverseDiff = "1.15.3"
 ScikitLearn = "0.6, 0.7"
 StableRNGs = "1"
 Statistics = "1"