MCMCChain → MCMChains Changeover (#695)

* Changed MCMCChain to MCMCChains.

* Rewrite of Turing native Chains functions.

* Add MCMCChains to REQUIRE file.

* Import save! and resume functions for the samplers.

* Fix save! and resume functionality.

* Removed names2inds import.

* Cleaning up.

* Changed :logevidence to :lp, as the previous functionality to handle this was removed.

* Modified SMC and IS to store logevidence, previously this was exp.(logevidence).

* The save! function now works with the NamedTuple info field.

* Adapted tests to use chn.logevidence, used to be chn[:logevidence].

* Updated REQUIRE to MCMCChains v0.3.0

* Removed save!, replaced with save as Chains are immutable.

* Made save function immutable.

* Removed unneeded test set, all are tested at MCMCChains.

* Added serialization tests to save/resume suite.

* Reincluded the io.jl test set for resumeable sampling.

* Removed vcat from utilities/io.jl

* Remove chain read/write tests.

* Changed logevidence in PMMH to -Inf

* Removed exp.(...) from PG

* Removed the correct exp.(...)

REQUIRE

@@ -4,7 +4,7 @@ Requires 0.5.0
 Reexport 0.2.0
 Distributions 0.16.0
 ForwardDiff 0.8.0
-MCMCChain 0.1.1
+MCMCChains 0.3.0
 Libtask 0.2.5
 Flux 0.6.7
 MacroTools

docs/src/_docs/get-started.md

@@ -46,7 +46,7 @@ end
 #  Run sampler, collect results
 chn = sample(gdemo(1.5, 2), HMC(1000, 0.1, 5))
 
-# Summarise results (currently requires the master branch from MCMCChain)
+# Summarise results (currently requires the master branch from MCMCChains)
 describe(chn)
 
 # Plot and save results

docs/src/_docs/guide.md

@@ -50,7 +50,7 @@ c5 = sample(gdemo(1.5, 2), HMCDA(1000, 0.15, 0.65))
 c6 = sample(gdemo(1.5, 2), NUTS(1000,  0.65))
 ```
 
-The `MCMCChain` module (which is re-exported by Turing) provides plotting tools for the `Chain` objects returned by a `sample` function. See the [MCMCChain](https://github.com/TuringLang/MCMCChain.jl) repository for more information on the suite of tools available for diagnosing MCMC chains.
+The `MCMCChains` module (which is re-exported by Turing) provides plotting tools for the `Chain` objects returned by a `sample` function. See the [MCMCChains](https://github.com/TuringLang/MCMCChains.jl) repository for more information on the suite of tools available for diagnosing MCMC chains.
 
 ```julia
 # Summarise results
@@ -315,18 +315,18 @@ The `Gibbs` sampler can be used to specify unique automatic differentation backe
 
 For more details of compositional sampling in Turing.jl, please check the corresponding [paper](http://xuk.ai/assets/aistats2018-turing.pdf).
 
-### Working with MCMCChain.jl
+### Working with MCMCChains.jl
 
-Turing.jl wraps its samples using `MCMCChain.Chain` so that all the functions working for `MCMCChain.Chain` can be re-used in Turing.jl. Two typical functions are `MCMCChain.describe` and `MCMCChain.plot`, which can be used as follows for an obtained chain `chn`. For more information on `MCMCChain`, please see the [GitHub repository](https://github.com/TuringLang/MCMCChain.jl).
+Turing.jl wraps its samples using `MCMCChains.Chain` so that all the functions working for `MCMCChains.Chain` can be re-used in Turing.jl. Two typical functions are `MCMCChains.describe` and `MCMCChains.plot`, which can be used as follows for an obtained chain `chn`. For more information on `MCMCChains`, please see the [GitHub repository](https://github.com/TuringLang/MCMCChains.jl).
 
 ```julia
-using MCMCChain: describe, plot
+using MCMCChains: describe, plot
 
 describe(chn) # Lists statistics of the samples.
 plot(chn) # Plots statistics of the samples.
 ```
 
-There are numerous functions in addition to `describe` and `plot` in the `MCMCChain` package, such as those used in convergence diagnostics. For more information on the package, please see the [GitHub repository](https://github.com/TuringLang/MCMCChain.jl).
+There are numerous functions in addition to `describe` and `plot` in the `MCMCChains` package, such as those used in convergence diagnostics. For more information on the package, please see the [GitHub repository](https://github.com/TuringLang/MCMCChains.jl).
 
 ### Working with Libtask.jl
 

docs/src/_docs/quick-start.md

@@ -5,7 +5,7 @@ permalink: /docs/quick-start/
 
 If you are already well-versed in probabalistic programming and just want to take a quick look at how Turing's syntax works or otherwise just want a model to start with, we have provided a Bayesian coin-flipping model to play with.
 
-This example can be run on however you have Julia installed (see [Getting Started](get-started.md)), but you will need to install the packages `Turing`, `Distributions`, `MCMCChain`, and `StatsPlots` if you have not done so already.
+This example can be run on however you have Julia installed (see [Getting Started](get-started.md)), but you will need to install the packages `Turing`, `Distributions`, `MCMCChains`, and `StatsPlots` if you have not done so already.
 
 This is an excerpt from a more formal example introducing probabalistic programming which can be found in Jupyter notebook form [here](https://nbviewer.jupyter.org/github/TuringLang/TuringTutorials/blob/master/0_Introduction.ipynb) or as part of the documentation website [here](../../tutorials/0-introduction/).
 

src/Turing.jl

@@ -12,12 +12,12 @@ using Requires, Reexport, ForwardDiff
 using Bijectors, StatsFuns, SpecialFunctions
 using Statistics, LinearAlgebra, ProgressMeter
 using Markdown, Libtask, MacroTools
-@reexport using Distributions, MCMCChain, Libtask
+@reexport using Distributions, MCMCChains, Libtask
 using Flux.Tracker: Tracker
 
 import Base: ~, convert, promote_rule, rand, getindex, setindex!
 import Distributions: sample
-import MCMCChain: AbstractChains, Chains
+import MCMCChains: AbstractChains, Chains
 
 const PROGRESS = Ref(true)
 function turnprogress(switch::Bool)

src/inference/Inference.jl

@@ -10,7 +10,7 @@ using StatsFuns: logsumexp
 
 import Distributions: sample
 import ..Core: getchunksize, getADtype
-import ..Utilities: Sample
+import ..Utilities: Sample, save, resume
 
 export  InferenceAlgorithm,
         Hamiltonian,
@@ -21,20 +21,20 @@ export  InferenceAlgorithm,
         HamiltonianRobustInit,
         SampleFromPrior,
         AnySampler,
-        MH, 
+        MH,
         Gibbs,      # classic sampling
-        HMC, 
-        SGLD, 
-        SGHMC, 
-        HMCDA, 
+        HMC,
+        SGLD,
+        SGHMC,
+        HMCDA,
         NUTS,       # Hamiltonian-like sampling
         DynamicNUTS,
-        IS, 
-        SMC, 
-        CSMC, 
-        PG, 
-        PIMH, 
-        PMMH, 
+        IS,
+        SMC,
+        CSMC,
+        PG,
+        PIMH,
+        PMMH,
         IPMCMC,  # particle-based sampling
         getspace,
         assume,

src/inference/gibbs.jl

@@ -178,12 +178,12 @@ function sample(
     @info("  Running time    = $time_total;")
 
     if resume_from != nothing   # concat samples
-        pushfirst!(samples, resume_from.value2...)
+        pushfirst!(samples, resume_from.info[:samples]...)
     end
     c = Chain(0.0, samples)       # wrap the result by Chain
 
     if save_state               # save state
-        save!(c, spl, model, varInfo)
+        c = save(c, spl, model, varInfo, samples)
     end
 
     return c

src/inference/hmc.jl

@@ -104,7 +104,7 @@ function sample(model::Model, alg::Hamiltonian;
                                 reuse_spl_n=0,                      # flag for spl re-using
                                 adapt_conf=STAN_DEFAULT_ADAPT_CONF, # adapt configuration
                 )
-    
+
     spl = reuse_spl_n > 0 ?
           resume_from.info[:spl] :
           Sampler(alg, adapt_conf)
@@ -184,13 +184,13 @@ function sample(model::Model, alg::Hamiltonian;
     end
 
     if resume_from != nothing   # concat samples
-        pushfirst!(samples, resume_from.value2...)
+        pushfirst!(samples, resume_from.info[:samples]...)
     end
     c = Chain(0.0, samples)       # wrap the result by Chain
     if save_state               # save state
         # Convert vi back to X if vi is required to be saved
         if spl.alg.gid == 0 invlink!(vi, spl) end
-        save!(c, spl, model, vi)
+        c = save(c, spl, model, vi, samples)
     end
     return c
 end

src/inference/is.jl

@@ -53,7 +53,7 @@ function sample(model::Model, alg::IS)
 
     le = logsumexp(map(x->x[:lp], samples)) - log(n)
 
-    Chain(exp.(le), samples)
+    Chain(le, samples)
 end
 
 function assume(spl::Sampler{<:IS}, dist::Distribution, vn::VarName, vi::VarInfo)

src/inference/mh.jl

@@ -165,11 +165,11 @@ function sample(model::Model, alg::MH;
   println("  Accept rate         = $accept_rate;")
 
   if resume_from != nothing   # concat samples
-    pushfirst!(samples, resume_from.value2...)
+    pushfirst!(samples, resume_from.info[:samples]...)
   end
   c = Chain(0.0, samples)       # wrap the result by Chain
   if save_state               # save state
-    save!(c, spl, model, vi)
+    c = save(c, spl, model, vi, samples)
   end
 
   c

src/inference/pgibbs.jl

@@ -82,13 +82,13 @@ function step(model, spl::Sampler{<:PG}, vi::VarInfo)
     return particles[indx].vi, true
 end
 
-function sample(  model::Model, 
+function sample(  model::Model,
                   alg::PG;
                   save_state=false,         # flag for state saving
                   resume_from=nothing,      # chain to continue
                   reuse_spl_n=0             # flag for spl re-using
                 )
-    
+
     spl = reuse_spl_n > 0 ?
           resume_from.info[:spl] :
           Sampler(alg)
@@ -128,24 +128,24 @@ function sample(  model::Model,
 
     loge = exp.(mean(spl.info[:logevidence]))
     if resume_from != nothing   # concat samples
-        pushfirst!(samples, resume_from.value2...)
-        pre_loge = resume_from.weight
+        pushfirst!(samples, resume_from.info[:samples]...)
+        pre_loge = exp.(resume_from.logevidence)
         # Calculate new log-evidence
-        pre_n = length(resume_from.value2)
-        loge = exp.((log(pre_loge) * pre_n + log(loge) * n) / (pre_n + n))
+        pre_n = length(resume_from.info[:samples])
+        loge = (log(pre_loge) * pre_n + log(loge) * n) / (pre_n + n)
     end
     c = Chain(loge, samples)       # wrap the result by Chain
 
     if save_state               # save state
-        save!(c, spl, model, vi)
+        c = save(c, spl, model, vi, samples)
     end
 
     return c
 end
 
-function assume(  spl::Sampler{T}, 
-                  dist::Distribution, 
-                  vn::VarName, 
+function assume(  spl::Sampler{T},
+                  dist::Distribution,
+                  vn::VarName,
                   _::VarInfo
                 ) where T<:Union{PG,SMC}
 
@@ -177,10 +177,10 @@ function assume(  spl::Sampler{T},
     return r, zero(Real)
 end
 
-function assume(  spl::Sampler{A}, 
-                  dists::Vector{D}, 
-                  vn::VarName, 
-                  var::Any, 
+function assume(  spl::Sampler{A},
+                  dists::Vector{D},
+                  vn::VarName,
+                  var::Any,
                   vi::VarInfo
                 ) where {A<:Union{PG,SMC},D<:Distribution}
     error("[Turing] PG and SMC doesn't support vectorizing assume statement")
@@ -191,9 +191,9 @@ function observe(spl::Sampler{T}, dist::Distribution, value, vi) where T<:Union{
     return zero(Real)
 end
 
-function observe( spl::Sampler{A}, 
-                  ds::Vector{D}, 
-                  value::Any, 
+function observe( spl::Sampler{A},
+                  ds::Vector{D},
+                  value::Any,
                   vi::VarInfo
                 ) where {A<:Union{PG,SMC},D<:Distribution}
     error("[Turing] PG and SMC doesn't support vectorizing observe statement")

src/inference/pmmh.jl

@@ -113,7 +113,7 @@ function step(model, spl::Sampler{<:PMMH}, vi::VarInfo, is_first::Bool)
     return vi, is_accept
 end
 
-function sample(  model::Model, 
+function sample(  model::Model,
                   alg::PMMH;
                   save_state=false,         # flag for state saving
                   resume_from=nothing,      # chain to continue
@@ -170,12 +170,12 @@ function sample(  model::Model,
     println("  Accept rate         = $accept_rate;")
 
     if resume_from != nothing   # concat samples
-      pushfirst!(samples, resume_from.value2...)
+      pushfirst!(samples, resume_from.info[:samples]...)
     end
-    c = Chain(0.0, samples)       # wrap the result by Chain
+    c = Chain(-Inf, samples)       # wrap the result by Chain
 
     if save_state               # save state
-      save!(c, spl, model, vi)
+      c = save(c, spl, model, vi, samples)
     end
 
     c

src/inference/smc.jl

@@ -85,6 +85,6 @@ function sample(model::Model, alg::SMC)
       end
     end
     w, samples = getsample(particles)
-    res = Chain(w, samples)
+    res = Chain(log(w), samples)
     return res
 end

src/utilities/Utilities.jl

@@ -3,7 +3,7 @@ module Utilities
 using ..Turing: Sampler
 using Distributions, Bijectors
 using StatsFuns, SpecialFunctions
-using MCMCChain: AbstractChains, Chains, names2inds
+using MCMCChains: AbstractChains, Chains, setinfo
 import Distributions: sample
 
 export  resample,