Updated HMC implementation for new AHMC version

[torfjelde]

This PR makes Turing.jl compatible with the AdvancedHMC.jl's master branch, which will soon become a new release.

[yebai]

@torfjelde It seems that the CI time increased due to some recent PRs, which leads to more often timeouts. Maybe splitting the Turing-CI workflow into two workflows to resolve the time limits of a single workflow?

[torfjelde]

@torfjelde It seems that the CI time increased due to some recent PRs, which leads to more often timeouts. Maybe splitting the Turing-CI workflow into two workflows to resolve the time limits of a single workflow?

Am fully aware; been trying to fix it for a couple of days now. I'm also not certain it's due to any of the recent PRs; it might be some other dependencies that slow things down.

[torfjelde]

E.g. integration tests in DPPL passed within an hour in our last PR-merge just 3 days ago: https://github.com/TuringLang/DynamicPPL.jl/actions/runs/1031412449. Then it suddenly started timing out without any changes made to master: https://github.com/TuringLang/DynamicPPL.jl/actions/runs/1034731795.

And Turing's full test-suite started timing out just 2 days ago: https://github.com/TuringLang/DynamicPPL.jl/actions/runs/1034731795

[torfjelde]

Uhhh I have no idea why but it seems like JuliaStats/Distributions.jl#1362 is causing the freeze. I noticed it would halt when testing

Turing.jl/test/test_utils/models.jl

Lines 73 to 77 in d029198

	@model function gdemo3(x = 10 * ones(2))
	# Multivariate `assume` and `observe`
	m ~ MvNormal(length(x), 1.0)
	x ~ MvNormal(m, 0.5 * ones(length(x)))
	end

for ESS sampler. Upper-bounding Distributions.jl to 0.12.10 and everything runs as per usual, it seems.

[codecov]

Codecov Report

Merging #1660 (041e620) into master (0168d84) will increase coverage by 11.06%.
The diff coverage is 85.71%.

@@             Coverage Diff             @@
##           master    #1660       +/-   ##
===========================================
+ Coverage   68.49%   79.56%   +11.06%     
===========================================
  Files          26       23        -3     
  Lines        1479     1414       -65     
===========================================
+ Hits         1013     1125      +112     
+ Misses        466      289      -177     

Impacted Files	Coverage Δ
src/core/Core.jl	100.00% <ø> (ø)
src/stdlib/RandomMeasures.jl	34.28% <0.00%> (-3.46%)	⬇️
src/variational/VariationalInference.jl	53.84% <37.50%> (-35.52%)	⬇️
src/stdlib/distributions.jl	56.98% <58.13%> (+4.60%)	⬆️
src/core/container.jl	76.19% <76.19%> (-15.40%)	⬇️
src/variational/advi.jl	61.11% <76.92%> (-2.13%)	⬇️
src/inference/Inference.jl	84.55% <80.35%> (-6.79%)	⬇️
src/core/compat/reversediff.jl	95.34% <83.33%> (+8.86%)	⬆️
src/modes/ModeEstimation.jl	84.29% <84.74%> (+15.46%)	⬆️
src/inference/hmc.jl	76.33% <85.39%> (-8.54%)	⬇️
... and 26 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update d029198...041e620. Read the comment docs.

[yebai]

Uhhh I have no idea why but it seems like JuliaStats/Distributions.jl#1362 is causing the freeze. I noticed it would halt when testing

cc @devmotion

[torfjelde]

David is one vacation so not sure he'll respond 😕

Also, instead look at #1663 where I've added some more info.

Should we maybe upper-bound Turing.jl to Distributions@0.12.10 too, rather than just the tests? Worried peoples code might be running into similar scenarios (though I still don't understand why it fails).

[yebai]

Should we maybe upper-bound Turing.jl to Distributions@0.12.10 too, rather than just the tests? Worried peoples code might be running into similar scenarios (though I still don't understand why it fails).

Happy to upper-bound Distributions before we figure out why CI times out.