TuringLang / Turing.jl / 9953741319 / 3
76%
main: 84%

LAST BUILD BRANCH: py/dppl-0.38
DEFAULT BRANCH: main
Ran
Files 22
Run time 3s
Badge
Embed ▾
README BADGES
x

If you need to use a raster PNG badge, change the '.svg' to '.png' in the link

Markdown

Textile

RDoc

HTML

Rst
coverage: 0.0%. Remained the same
9953741319.3

push

github

web-flow 
Fixes and improvements to experimental `Gibbs` (#2231)

* moved new Gibbs tests all into a single block

* initial work on making Gibbs work with `externalsampler`

* removed references to Setfield.jl

* fixed crucial bug in experimental Gibbs sampler

* added ground-truth comparison for Gibbs sampler on demo models

* added convenience method for performing two sample KS test

* use thinning to avoid OOM issues

* removed incredibly slow testset that didn't really add much

* removed now-redundant testset

* use Anderson-Darling test instead of Kolomogorov-Smirnov to better
capture tail differences + remove subsampling of chains since it
doesn't really matter that when we're using aggressive thinning and
test statistics based on comparing order stats

* more work on testing

* fixed tests

* make failures of `two_sample_ad_tests` a bit more informative

* make failrues of `two_sample_ad_test` produce more informative logs

* additional information upon `two_sample_ad_test` failure

* rename `two_sample_ad_test` to `two_sample_test` and use KS test instead

* added minor test for externalsampler usage

* also test AdvancedHMC samplers with Gibbs

* forgot to add updates to src/mcmc/abstractmcmc.jl in previous commits

* removed usage of `timeit_testset` macro

* added temporary fix for externalsampler that needs to be removed once
DPPL has been updated

* minor reorg of two testsets

* set random seeds more aggressively in an attempt to make tests more reproducible

* removed hack, awaiting PR to DynamicPPL

* renamed `_getmodel` to `getmodel`, `_setmodel` to `setmodel`, and
`_varinfo` to `varinfo_from_logdensityfn`

* missed some instances during rnenaming

* fixed missing merge in initial step for experimental `Gibbs`

* Always reconstruct `ADGradientWrapper` using the `adype` available in `ExternalSampler`

* Test Gibbs with different adtype in externalsampler to ensure that works

* Up... (continued)

0 of 1560 relevant lines covered (0.0%)

0.0 hits per line

Source Files on job run-mcmc/abstractmcmc.jl-1.7-windows-latest-x64-2 - 9953741319.3