Cleanup benchmarks (#1937)

* Re-organise benchmarks

* Delete MicroBenchmarks.yml

* Update README.md

Author: yebai

.github/workflows/MicroBenchmarks.yml

@@ -1,48 +1,18 @@
-# MicroBenchmarks for Turing.jl
+# Benchmarks for Turing.jl
 
 ## Write a benchmark
 
 Generally, you can run any code in a benchmark. We use package
 `BenchmarkTools.jl` to organize our benchmarks, and
-`BenchmarkHelper.BenchmarkSuite` is the root benchmark group, so you
+`BenchmarkSuite` is the root benchmark group, so you
 can write a benchmark like this:
 
 ```julia
-using Turing, BenchmarkHelper, BenchmarkTools
+using Turing, BenchmarkTools
 
-BenchmarkSuite["dummy"] = BenchmarkGroup(["tag1", "tag2"])
+BenchmarkSuite["example_benchmark"] = BenchmarkGroup(["tag1", "tag2"])
 
-BenchmarkSuite["dummy"]["benchmark_1"] = @benchmarkable a_time_consuming_computing()
+BenchmarkSuite["example_benchmark"]["benchmark_1"] = @benchmarkable a_time_consuming_computing()
 ```
-More details can be found in the documents of `BenchmarkTools.jl`.
-
-After its run, the results will be saved under the directory
-`benchmarks/output/`.
-
-## Run on Github
 
-All benchmarks under this directory can run on Github, triggered by
-commenting on a Pull Request (a.k.a Code Review Comment).
-
-You must **@BayesBot** and include a command in pattern
-`runbenchmarks(benchamrk_tags, vs=branch)` in the comment text to
-specify the branches on which the benchmarks will run and trigger the
-benchmarks. This command pattern is borrowed from `Nanosoldier.jl`,
-please find more detail from [its
-documents](https://github.com/TuringLang/Nanosoldier.jl/blob/turing/README.md#trigger-syntax). Here
-is an examples:
-
-```
-@BayesBot `runbenchmarks(ALL, vs=":master")`
-```
-
-After these triggered benchmarks finish, the reports will be committed
-to
-[TuringLang/BenchmarkReports](https://github.com/TuringLang/BenchmarkReports),
-then a reply comment contains the reports link will be made on the
-original pull request.
-
-### Notes
-
-- Put a Regex pattern to the `EXCLUDES` array in
-  `benchmarks/config.toml` to exclude paricular benchmark files
+More details can be found in the documents of `BenchmarkTools.jl`.

benchmarks/BenchmarkHelper.jl

@@ -1,9 +1,18 @@
-using Turing, BenchmarkTools, BenchmarkHelper
+using Turing, BenchmarkTools
 using LinearAlgebra
 
-## Dummny benchmarks
+const BenchmarkSuite = BenchmarkTools.BenchmarkGroup()
 
-BenchmarkSuite["dummy"] = BenchmarkGroup(["dummy"])
+#
+# Add models to benchmarks 
+#
+
+include("models/hlr.jl")
+include("models/lr.jl")
+include("models/sv_nuts.jl")
+
+# constrained 
+BenchmarkSuite["constrained"] = BenchmarkGroup(["constrained"])
 
 data = [0, 1, 0, 1, 1, 1, 1, 1, 1, 1]
 
@@ -17,7 +26,7 @@ data = [0, 1, 0, 1, 1, 1, 1, 1, 1, 1]
 end
 
 
-BenchmarkSuite["dummy"]["dummy"] = @benchmarkable sample($(constrained_test(data)), $(HMC(0.01, 2)), 2000)
+BenchmarkSuite["constrained"]["constrained"] = @benchmarkable sample($(constrained_test(data)), $(HMC(0.01, 2)), 2000)
 
 
 ## gdemo
@@ -35,9 +44,8 @@ end
 BenchmarkSuite["gdemo"]["hmc"] = @benchmarkable sample($(gdemo(1.5, 2.0)), $(HMC(0.01, 2)), 2000)
 
 
-##
+
 ## MvNormal
-##
 
 BenchmarkSuite["mnormal"] = BenchmarkGroup(["mnormal"])
 

benchmarks/README.md

@@ -1,8 +1,6 @@
 using Turing
 using LinearAlgebra
 
-using BenchmarkHelper
-
 include("lr_helper.jl")
 
 if !haskey(BenchmarkSuite, "nuts")

benchmarks/benchmarks.jl renamed to benchmarks/benchmarks_suite.jl

@@ -1,8 +1,6 @@
 using Turing
 using LinearAlgebra
 
-using BenchmarkHelper
-
 include("lr_helper.jl")
 
 if !haskey(BenchmarkSuite, "nuts")