remove AD code, update docs, bump to 2.0 (#99)

* remove AD code, update docs, bump to 2.0

Author: tpapp

Project.toml

@@ -1,36 +1,22 @@
 name = "LogDensityProblems"
 uuid = "6fdf6af0-433a-55f7-b3ed-c6c6e0b8df7c"
 authors = ["Tamas K. Papp <tkpapp@gmail.com>"]
-version = "1.0.3"
+version = "2.0.0"
 
 [deps]
 ArgCheck = "dce04be8-c92d-5529-be00-80e4d2c0e197"
 DocStringExtensions = "ffbed154-4ef7-542d-bbb7-c09d3a79fcae"
 Random = "9a3f8284-a2c9-5f02-9a11-845980a1fd5c"
-Requires = "ae029012-a4dd-5104-9daa-d747884805df"
 UnPack = "3a884ed6-31ef-47d7-9d2a-63182c4928ed"
 
 [compat]
 ArgCheck = "1, 2"
-BenchmarkTools = "1"
 DocStringExtensions = "0.8, 0.9"
-Requires = "0.5, 1"
 UnPack = "0.1, 1"
 julia = "1.6"
 
 [extras]
-BenchmarkTools = "6e4b80f9-dd63-53aa-95a3-0cdb28fa8baf"
-DiffResults = "163ba53b-c6d8-5494-b064-1a9d43ac40c5"
-Distributions = "31c24e10-a181-5473-b8eb-7969acd0382f"
-Documenter = "e30172f5-a6a5-5a46-863b-614d45cd2de4"
-ForwardDiff = "f6369f11-7733-5829-9624-2563aa707210"
-Pkg = "44cfe95a-1eb2-52ea-b672-e2afdf69b78f"
-ReverseDiff = "37e2e3b7-166d-5795-8a7a-e32c996b4267"
-StatsBase = "2913bbd2-ae8a-5f71-8c99-4fb6c76f3a91"
-StatsFuns = "4c63d2b9-4356-54db-8cca-17b64c39e42c"
 Test = "8dfed614-e22c-5e08-85e1-65c5234f0b40"
-Tracker = "9f7883ad-71c0-57eb-9f7f-b5c9e6d3789c"
-Zygote = "e88e6eb3-aa80-5325-afca-941959d7151f"
 
 [targets]
-test = ["BenchmarkTools", "Distributions", "Documenter", "ForwardDiff", "Pkg", "ReverseDiff", "StatsBase", "StatsFuns", "Test", "Tracker", "Zygote"]
+test = ["Test"]

README.md

@@ -11,14 +11,18 @@ A common framework for implementing and using log densities for inference, provi
 
 1. The [`logdensity`](https://tamaspapp.eu/LogDensityProblems.jl/dev/#LogDensityProblems.logdensity) method with corresponding interface, which can be used by other packages that operate on (log) densities and need to evaluate the log densities or the gradients (eg [MCMC](https://en.wikipedia.org/wiki/Markov_chain_Monte_Carlo), [MAP](https://en.wikipedia.org/wiki/Maximum_a_posteriori_estimation), [ML](https://en.wikipedia.org/wiki/Maximum_likelihood_estimation) or similar methods).
 
-2. The [`ADgradient`](https://tamaspapp.eu/LogDensityProblems.jl/dev/#LogDensityProblems.ADgradient) which makes objects that support `logdensity` to calculate log density *values* calculate log density *gradients* using various automatic differentiation packages.
-
-3. Various utility functions for debugging and testing log densities.
+2. Various utility functions for debugging and testing log densities.
 
 **NOTE** As of version 1.0, transformed log densities have been moved to [TransformedLogDensities.jl](https://github.com/tpapp/TransformedLogDensities.jl). Existing code that uses `TransformedLogDensity` should add
 ```
 using TransformedLogDensities
 ```
 or equivalent.
 
+**NOTE**: As of version 2.0, automatic differentiation backends have been moved to [https://github.com/tpapp/LogDensityProblemsAD.jl](https://github.com/tpapp/LogDensityProblemsAD.jl "LogDensityProblemsAD.jl"). If your code uses `ADgradient`, simply add
+```julia
+using LogDensityProblemsAD
+```
+or equivalent.
+
 See the [documentation](https://tpapp.github.io/LogDensityProblems.jl/dev) for details.

docs/Project.toml

@@ -2,6 +2,7 @@
 BenchmarkTools = "6e4b80f9-dd63-53aa-95a3-0cdb28fa8baf"
 Documenter = "e30172f5-a6a5-5a46-863b-614d45cd2de4"
 ForwardDiff = "f6369f11-7733-5829-9624-2563aa707210"
+LogDensityProblemsAD = "996a588d-648d-4e1f-a8f0-a84b347e47b1"
 Tracker = "9f7883ad-71c0-57eb-9f7f-b5c9e6d3789c"
 TransformVariables = "84d833dd-6860-57f9-a1a7-6da5db126cff"
 TransformedLogDensities = "f9bc47f6-f3f8-4f3b-ab21-f8bc73906f26"

docs/make.jl

@@ -1,4 +1,5 @@
-using Documenter, LogDensityProblems, ForwardDiff, Tracker, Zygote, BenchmarkTools, TransformedLogDensities
+using Documenter, LogDensityProblems, ForwardDiff, Tracker, Zygote, BenchmarkTools,
+    TransformedLogDensities, LogDensityProblemsAD
 
 makedocs(
     sitename = "LogDensityProblems.jl",

docs/src/index.md

@@ -131,19 +131,20 @@ Here we use the exponential function to transform from ``\mathbb{R}`` to the pos
 
 ## Automatic differentiation
 
-Using either definition, you can now transform to another object which is capable of evaluating the *gradient*, using automatic differentiation. The wrapper for this is
-```@docs
-ADgradient
-```
+Using either definition, you can transform to another object which is capable of evaluating the *gradient*, using automatic differentiation. For this, you need the [LogDensityProblemsAD.jl](https://github.com/tpapp/LogDensityProblemsAD.jl) package.
 
 Now observe that we can obtain gradients, too:
 ```@repl 1
 import ForwardDiff
+using LogDensityProblemsAD
 ∇ℓ = ADgradient(:ForwardDiff, ℓ)
 LogDensityProblems.capabilities(∇ℓ)
 LogDensityProblems.logdensity_and_gradient(∇ℓ, zeros(2))
 ```
 
+!!! note
+    Before version 2.0, `ADgradient` was part of this package. To update older code, just add `using LogDensityProblemsAD`.
+
 ## Manually calculated derivatives
 
 If you prefer not to use automatic differentiation, you can wrap your own derivatives following the template
@@ -171,7 +172,6 @@ You may find these utilities useful for debugging and optimization.
 
 ```@docs
 LogDensityProblems.stresstest
-LogDensityProblems.benchmark_ForwardDiff_chunks
 ```
 
 # [Log densities API](@id log-density-api)