Make `varmat` functions work with ldlt (Issue #2101) #2266

bbbales2 · 2020-12-18T21:51:44Z

Summary

Second attempt at making varmat work with ldlt

Side Effects

I removed LDLT_alloc and replaced LDLT_factor with an API incompatible replacement.

So this is a major version bump for Math right?

Release notes

Add varmat implementation of *_ldlt_* functions

Checklist

Math issue Make functions with custom autodiff var<mat> friendly #2101
Copyright holder: Columbia University

The copyright holder is typically you or your assignee, such as a university or company. By submitting this pull request, the copyright holder is agreeing to the license the submitted work under the following licenses:
- Code: BSD 3-clause (https://opensource.org/licenses/BSD-3-Clause)
- Documentation: CC-BY 4.0 (https://creativecommons.org/licenses/by/4.0/)
the basic tests are passing
- unit tests pass (to run, use: ./runTests.py test/unit)
- header checks pass, (make test-headers)
- dependencies checks pass, (make test-math-dependencies)
- docs build, (make doxygen)
- code passes the built in C++ standards checks (make cpplint)
the code is written in idiomatic C++ and changes are documented in the doxygen
the new changes are tested

…4.1 (tags/RELEASE_600/final)

…feature/varmat-ldlt2

…4.1 (tags/RELEASE_600/final)

bbbales2 · 2020-12-19T16:03:40Z

test/unit/math/mix/fun/trace_gen_inv_quad_form_ldlt_test.cpp

@@ -64,8 +55,8 @@ TEST(mathMixMatFun, traceGenInvQuadForm) {

  // exception tests
  // non-square second arg
-  Eigen::MatrixXd a34(3, 4);


Can't take an LDLT of a non-square matrix, so this fails before it gets to trace_gen_inv_quad_form_ldlt (which takes an LDLT as an argument)

bbbales2 · 2020-12-19T16:18:04Z

I ended up replacing the ldlt autodiff that was there, but it's still not great. There new LDLT_factor is incompatible with the old one. We should still probably remove this eventually. The new version is cleaner and simpler, but it's a kinda-sometimes autodiff object that doesn't really behave like the other autodiff things we have (that is why it takes a template parameter which says whether to allocate it in the arena or not).

If we're gonna have LDLT factors as autodiff variables, they need to be vari_value<LDLT> types or something and follow the rules (and similarly the functions that consume them need to not make any extra assumptions about the lifetime of objects, which they currently do). In the interim, we have this.

Still need add a three argument expect_ad_matvar so that trace_gen_inv_quad_form can be tested with varmat types.

Will also get some benchmarks to make sure the matvar stuff isn't slower. When I coded this stuff up before with reverse_pass_callback I remeber the existing stuff was pretty fast.

…4.1 (tags/RELEASE_600/final)

…feature/varmat-ldlt2

…4.1 (tags/RELEASE_600/final)

…feature/varmat-ldlt2

…lt` (Issue #2101)

…4.1 (tags/RELEASE_600/final)

bbbales2 · 2020-12-21T21:14:19Z

Here's benchmarks:

And here's the timing results of the one where performance gets bad. Not sure what happened there. I didn't spend time looking into it:

Edit: Benchmark code is basically the stuff here and here running with matvars. I had to modify them a bit for the new thing (to use stan::math::make_ldlt_factor instead of the LDLT_factor constructor), but that was it.

…feature/varmat-ldlt2

…4.1 (tags/RELEASE_600/final)

bbbales2 · 2021-01-01T20:37:09Z

This needs #2280 in before it is ready to go (there's a three argument function in this)

stan-buildbot · 2021-01-02T04:10:37Z

Name	Old Result	New Result	Ratio	Performance change( 1 - new / old )
gp_pois_regr/gp_pois_regr.stan	3.33	3.38	0.98	-1.62% slower
low_dim_corr_gauss/low_dim_corr_gauss.stan	0.02	0.02	0.91	-9.64% slower
eight_schools/eight_schools.stan	0.11	0.11	0.98	-1.98% slower
gp_regr/gp_regr.stan	0.15	0.15	1.01	1.24% faster
irt_2pl/irt_2pl.stan	5.22	5.21	1.0	0.3% faster
performance.compilation	91.96	89.8	1.02	2.35% faster
low_dim_gauss_mix_collapse/low_dim_gauss_mix_collapse.stan	8.65	8.63	1.0	0.17% faster
pkpd/one_comp_mm_elim_abs.stan	29.16	30.16	0.97	-3.42% slower
sir/sir.stan	133.61	137.78	0.97	-3.12% slower
gp_regr/gen_gp_data.stan	0.04	0.04	1.0	0.01% faster
low_dim_gauss_mix/low_dim_gauss_mix.stan	3.04	3.05	1.0	-0.21% slower
pkpd/sim_one_comp_mm_elim_abs.stan	0.4	0.37	1.07	6.33% faster
arK/arK.stan	2.54	2.53	1.01	0.51% faster
arma/arma.stan	0.61	0.61	1.01	0.52% faster
garch/garch.stan	0.67	0.67	1.0	-0.38% slower
Mean result: 0.995121488597

Jenkins Console Log
Blue Ocean
Commit hash: 294b7fa

Machine information

ProductName: Mac OS X ProductVersion: 10.11.6 BuildVersion: 15G22010

CPU:
Intel(R) Xeon(R) CPU E5-1680 v2 @ 3.00GHz

G++:
Configured with: --prefix=/Applications/Xcode.app/Contents/Developer/usr --with-gxx-include-dir=/usr/include/c++/4.2.1
Apple LLVM version 7.0.2 (clang-700.1.81)
Target: x86_64-apple-darwin15.6.0
Thread model: posix

Clang:
Apple LLVM version 7.0.2 (clang-700.1.81)
Target: x86_64-apple-darwin15.6.0
Thread model: posix

stan-buildbot · 2021-01-05T01:39:44Z

Name	Old Result	New Result	Ratio	Performance change( 1 - new / old )
gp_pois_regr/gp_pois_regr.stan	3.37	3.41	0.99	-1.32% slower
low_dim_corr_gauss/low_dim_corr_gauss.stan	0.02	0.02	0.97	-3.2% slower
eight_schools/eight_schools.stan	0.11	0.12	0.95	-5.35% slower
gp_regr/gp_regr.stan	0.16	0.15	1.03	2.87% faster
irt_2pl/irt_2pl.stan	5.16	5.19	1.0	-0.41% slower
performance.compilation	92.48	90.04	1.03	2.64% faster
low_dim_gauss_mix_collapse/low_dim_gauss_mix_collapse.stan	8.64	8.62	1.0	0.26% faster
pkpd/one_comp_mm_elim_abs.stan	30.07	28.28	1.06	5.96% faster
sir/sir.stan	135.68	135.11	1.0	0.42% faster
gp_regr/gen_gp_data.stan	0.05	0.04	1.09	8.44% faster
low_dim_gauss_mix/low_dim_gauss_mix.stan	3.08	3.07	1.0	0.26% faster
pkpd/sim_one_comp_mm_elim_abs.stan	0.37	0.4	0.92	-8.76% slower
arK/arK.stan	2.52	2.51	1.0	0.04% faster
arma/arma.stan	0.61	0.61	1.0	-0.48% slower
garch/garch.stan	0.67	0.68	0.98	-1.94% slower
Mean result: 1.00125016541

Jenkins Console Log
Blue Ocean
Commit hash: 294b7fa

Machine information

ProductName: Mac OS X ProductVersion: 10.11.6 BuildVersion: 15G22010

CPU:
Intel(R) Xeon(R) CPU E5-1680 v2 @ 3.00GHz

G++:
Configured with: --prefix=/Applications/Xcode.app/Contents/Developer/usr --with-gxx-include-dir=/usr/include/c++/4.2.1
Apple LLVM version 7.0.2 (clang-700.1.81)
Target: x86_64-apple-darwin15.6.0
Thread model: posix

Clang:
Apple LLVM version 7.0.2 (clang-700.1.81)
Target: x86_64-apple-darwin15.6.0
Thread model: posix

bbbales2 · 2021-01-05T18:14:52Z

@SteveBronder is is ready to review now that the three argument tests are in.

stan-buildbot · 2021-01-06T01:44:21Z

Name	Old Result	New Result	Ratio	Performance change( 1 - new / old )
gp_pois_regr/gp_pois_regr.stan	3.46	3.37	1.03	2.57% faster
low_dim_corr_gauss/low_dim_corr_gauss.stan	0.02	0.02	0.97	-3.16% slower
eight_schools/eight_schools.stan	0.11	0.11	1.01	0.81% faster
gp_regr/gp_regr.stan	0.16	0.15	1.06	5.34% faster
irt_2pl/irt_2pl.stan	5.22	5.2	1.01	0.52% faster
performance.compilation	91.14	89.74	1.02	1.54% faster
low_dim_gauss_mix_collapse/low_dim_gauss_mix_collapse.stan	8.64	8.66	1.0	-0.19% slower
pkpd/one_comp_mm_elim_abs.stan	27.49	27.75	0.99	-0.94% slower
sir/sir.stan	143.89	131.45	1.09	8.65% faster
gp_regr/gen_gp_data.stan	0.05	0.04	1.05	4.74% faster
low_dim_gauss_mix/low_dim_gauss_mix.stan	3.07	3.07	1.0	-0.25% slower
pkpd/sim_one_comp_mm_elim_abs.stan	0.43	0.38	1.14	12.35% faster
arK/arK.stan	2.53	2.54	0.99	-0.61% slower
arma/arma.stan	0.61	0.61	1.01	0.8% faster
garch/garch.stan	0.67	0.68	0.98	-1.82% slower
Mean result: 1.02247109706

Jenkins Console Log
Blue Ocean
Commit hash: dbff4ad

Machine information

ProductName: Mac OS X ProductVersion: 10.11.6 BuildVersion: 15G22010

CPU:
Intel(R) Xeon(R) CPU E5-1680 v2 @ 3.00GHz

G++:
Configured with: --prefix=/Applications/Xcode.app/Contents/Developer/usr --with-gxx-include-dir=/usr/include/c++/4.2.1
Apple LLVM version 7.0.2 (clang-700.1.81)
Target: x86_64-apple-darwin15.6.0
Thread model: posix

Clang:
Apple LLVM version 7.0.2 (clang-700.1.81)
Target: x86_64-apple-darwin15.6.0
Thread model: posix

stan-buildbot · 2021-01-07T20:23:54Z

Name	Old Result	New Result	Ratio	Performance change( 1 - new / old )
gp_pois_regr/gp_pois_regr.stan	3.34	3.33	1.0	0.46% faster
low_dim_corr_gauss/low_dim_corr_gauss.stan	0.02	0.02	1.01	0.67% faster
eight_schools/eight_schools.stan	0.11	0.12	0.97	-3.48% slower
gp_regr/gp_regr.stan	0.15	0.16	0.99	-1.21% slower
irt_2pl/irt_2pl.stan	5.24	5.14	1.02	1.88% faster
performance.compilation	91.48	89.67	1.02	1.97% faster
low_dim_gauss_mix_collapse/low_dim_gauss_mix_collapse.stan	8.66	8.7	0.99	-0.5% slower
pkpd/one_comp_mm_elim_abs.stan	28.54	29.48	0.97	-3.3% slower
sir/sir.stan	133.37	132.02	1.01	1.02% faster
gp_regr/gen_gp_data.stan	0.04	0.04	0.98	-1.54% slower
low_dim_gauss_mix/low_dim_gauss_mix.stan	3.05	3.04	1.0	0.09% faster
pkpd/sim_one_comp_mm_elim_abs.stan	0.37	0.38	0.97	-3.5% slower
arK/arK.stan	2.54	2.51	1.01	1.16% faster
arma/arma.stan	0.6	0.63	0.96	-4.12% slower
garch/garch.stan	0.67	0.69	0.97	-2.75% slower
Mean result: 0.991710391681

Jenkins Console Log
Blue Ocean
Commit hash: 4c59678

Machine information

ProductName: Mac OS X ProductVersion: 10.11.6 BuildVersion: 15G22010

CPU:
Intel(R) Xeon(R) CPU E5-1680 v2 @ 3.00GHz

G++:
Configured with: --prefix=/Applications/Xcode.app/Contents/Developer/usr --with-gxx-include-dir=/usr/include/c++/4.2.1
Apple LLVM version 7.0.2 (clang-700.1.81)
Target: x86_64-apple-darwin15.6.0
Thread model: posix

Clang:
Apple LLVM version 7.0.2 (clang-700.1.81)
Target: x86_64-apple-darwin15.6.0
Thread model: posix

SteveBronder

So reviewing this I didn't like that we had to make multiple copies as we passed around the ldlt factor. Trying to sort this out I think it was just easier to make a little review branch with some changes

feature/varmat-ldlt2...review/feature/varmat-ldlt2#diff-8eb184dd0a9220fb56a0aea61b945cfea318ecfbfe8f25fd7ac6050c5f19b724R14

There's only one thing I strongly don't like about the changes I made which is the forward deceleration of make_chainable_ptr() in prim. We could flip this around and just declare the method for chainable_ldlt_ptr() in prim

// in prim's LDLT_factor
inline auto* chainable_ldlt_ptr();

and then in rev actually define chainable_ldlt_ptr() for the prim version.

Though if you don't like what I have above then I think then you can ignore that and just use the other review comments

stan/math/prim/fun/LDLT_factor.hpp

SteveBronder · 2021-01-10T22:18:21Z

stan/math/prim/fun/mdivide_left_ldlt.hpp

+  return A.ldlt().solve(
      Eigen::Matrix<return_type_t<T, EigMat>, EigMat::RowsAtCompileTime,
                    EigMat::ColsAtCompileTime>(b));


[optional] if b is an rvalue forwarding could be better here

Good point. I'm gonna leave it as is though. This is some weird type promotion thing so the Eigen solves work. Like if T and EigMat don't match, it will make them match.

If we added another overload for when the scalar types are the same it would handle this. I grepped for where this function is used (it's an internal function not exposed to the Stan langage) and I don't think it's actually used anywhere (mdivide_right_ldlt uses this, but that is unused, and again that is an internal function), so I'm not gonna worry about this (though it would take about 5s to implement -- just seems like more code and complexity some future person has to decode).

stan/math/prim/prob/matrix_normal_prec_lpdf.hpp

stan/math/rev/fun/LDLT_factor.hpp

bbbales2 · 2021-01-11T00:46:14Z

Changes look good. I'll merge tomorrow. The arena_t thing on the var LDLT is a good catch. Thanks!

bbbales2 · 2021-01-11T14:19:50Z

I moved make_chainable_ptr back out of LDLT_factor. It was kinda awkward to make it work with the prim stuff. The only place we do this is mdivide_left_ldlt and I grepped around and I think everywhere we do this there will only be one copy anyway.

This is the grep, and trace_inv_quad_form_ldlt and trace_gen_inv_quad_form_ldlt are only used in one place each as well:

stan/math/prim/fun/trace_inv_quad_form_ldlt.hpp:  return B.cwiseProduct(mdivide_left_ldlt(A, B)).sum();
stan/math/prim/fun/trace_gen_inv_quad_form_ldlt.hpp:  return multiply(B, D.transpose()).cwiseProduct(mdivide_left_ldlt(A, B)).sum();
stan/math/prim/prob/wishart_lpdf.hpp:        mdivide_left_ldlt(ldlt_S, W_ref));
stan/math/prim/prob/inv_wishart_lpdf.hpp:        mdivide_left_ldlt(ldlt_W, S_ref));

bbbales2 · 2021-01-11T14:20:17Z

@SteveBronder should be ready for review again

…4.1 (tags/RELEASE_600/final)

SteveBronder

Good!

…4.1 (tags/RELEASE_600/final)

stan-buildbot · 2021-01-12T10:32:22Z

Name	Old Result	New Result	Ratio	Performance change( 1 - new / old )
gp_pois_regr/gp_pois_regr.stan	3.39	3.37	1.01	0.54% faster
low_dim_corr_gauss/low_dim_corr_gauss.stan	0.02	0.02	0.99	-0.58% slower
eight_schools/eight_schools.stan	0.11	0.11	1.02	2.06% faster
gp_regr/gp_regr.stan	0.15	0.15	1.0	-0.32% slower
irt_2pl/irt_2pl.stan	5.22	5.19	1.0	0.45% faster
performance.compilation	91.33	89.79	1.02	1.68% faster
low_dim_gauss_mix_collapse/low_dim_gauss_mix_collapse.stan	8.64	8.63	1.0	0.21% faster
pkpd/one_comp_mm_elim_abs.stan	29.21	29.17	1.0	0.13% faster
sir/sir.stan	134.85	137.58	0.98	-2.02% slower
gp_regr/gen_gp_data.stan	0.04	0.04	1.0	0.35% faster
low_dim_gauss_mix/low_dim_gauss_mix.stan	3.05	3.05	1.0	-0.01% slower
pkpd/sim_one_comp_mm_elim_abs.stan	0.37	0.38	0.98	-2.14% slower
arK/arK.stan	2.51	2.53	0.99	-0.98% slower
arma/arma.stan	0.6	0.6	1.0	0.15% faster
garch/garch.stan	0.57	0.57	1.0	-0.13% slower
Mean result: 0.999713098543

Jenkins Console Log
Blue Ocean
Commit hash: 8366815

Machine information

ProductName: Mac OS X ProductVersion: 10.11.6 BuildVersion: 15G22010

CPU:
Intel(R) Xeon(R) CPU E5-1680 v2 @ 3.00GHz

G++:
Configured with: --prefix=/Applications/Xcode.app/Contents/Developer/usr --with-gxx-include-dir=/usr/include/c++/4.2.1
Apple LLVM version 7.0.2 (clang-700.1.81)
Target: x86_64-apple-darwin15.6.0
Thread model: posix

Clang:
Apple LLVM version 7.0.2 (clang-700.1.81)
Target: x86_64-apple-darwin15.6.0
Thread model: posix

bbbales2 · 2021-01-12T12:19:32Z

@rok-cesnovar this was approved but then my test restart commit dismissed the review. Can you approve and merge?

bbbales2 and others added 9 commits December 9, 2020 08:04

Partial commit of another ldlt conversion (Issue #2101)

d794216

Partial commit of ldlt changes (Issue #2101)

8ce6c8c

Merge remote-tracking branch 'origin/develop' into feature/varmat-ldlt2

bab6da5

[Jenkins] auto-formatting by clang-format version 6.0.0-1ubuntu2~16.0…

906504f

…4.1 (tags/RELEASE_600/final)

Converted to new LDLT_factor (Issue #2101)

24b2e7c

Cleaned up new LDLT_factor and added docs (Issue #2101)

c76e914

Updated docs (Issue #2101)

95b92c6

Merge branch 'feature/varmat-ldlt2' of github.com:stan-dev/math into …

fe405fc

…feature/varmat-ldlt2

[Jenkins] auto-formatting by clang-format version 6.0.0-1ubuntu2~16.0…

36cdf30

…4.1 (tags/RELEASE_600/final)

bbbales2 commented Dec 19, 2020

View reviewed changes

bbbales2 and others added 13 commits December 20, 2020 17:01

Merge remote-tracking branch 'origin/develop' into feature/varmat-ldlt2

0dc4db9

Added missing explicits (Issue #2101)

181dc51

[Jenkins] auto-formatting by clang-format version 6.0.0-1ubuntu2~16.0…

518cda9

…4.1 (tags/RELEASE_600/final)

Updated LDLT_factor tests (Issue #2101)

00344d9

[Jenkins] auto-formatting by clang-format version 6.0.0-1ubuntu2~16.0…

10522d6

…4.1 (tags/RELEASE_600/final)

Updated matrix normal tests (Issue #2101)

2661baf

Merge branch 'feature/varmat-ldlt2' of github.com:stan-dev/math into …

16d06aa

…feature/varmat-ldlt2

Merge commit 'effd263008e18519c7490303b3db21fb13947738' into HEAD

d293bbd

[Jenkins] auto-formatting by clang-format version 6.0.0-1ubuntu2~16.0…

4ad4247

…4.1 (tags/RELEASE_600/final)

Merge remote-tracking branch 'origin/develop' into feature/varmat-ldlt2

9691925

Merge branch 'feature/varmat-ldlt2' of github.com:stan-dev/math into …

104750d

…feature/varmat-ldlt2

Updated tests and fixed problem with trace in `trace_inv_quad_form_ld…

9fdaf69

…lt` (Issue #2101)

[Jenkins] auto-formatting by clang-format version 6.0.0-1ubuntu2~16.0…

f2891e4

…4.1 (tags/RELEASE_600/final)

bbbales2 and others added 5 commits December 21, 2020 20:01

Updated inverse Wishart tests (Issue #2101)

ced30e0

Merge branch 'feature/varmat-ldlt2' of github.com:stan-dev/math into …

a1a7e0e

…feature/varmat-ldlt2

[Jenkins] auto-formatting by clang-format version 6.0.0-1ubuntu2~16.0…

099fd83

…4.1 (tags/RELEASE_600/final)

Updated Wishart lpdf to check for symmetric inputs (Issue #2101)

bdf98fd

[Jenkins] auto-formatting by clang-format version 6.0.0-1ubuntu2~16.0…

29dd029

…4.1 (tags/RELEASE_600/final)

This was referenced Jan 1, 2021

Make multivariate distributions work with varmat #2284

Merged

2.26 release varmat meta-issue stan-dev/cmdstan#958

Closed

Merge remote-tracking branch 'origin/develop' into feature/varmat-ldlt2

dbff4ad

Merge remote-tracking branch 'origin/develop' into feature/varmat-ldlt2

4c59678

make a chainable_ldlt_ptr() method for LDLT_factor()

a2fbfe9

SteveBronder reviewed Jan 11, 2021

View reviewed changes

Removed make_chainable_ptr stuff from LDLT_factor classes (Issue #2101)

0a89494

yashikno and others added 2 commits January 11, 2021 14:29

Merge commit '222e1ad9426ba13c364a8a65d2f3958104047b54' into HEAD

9de6fd1

[Jenkins] auto-formatting by clang-format version 6.0.0-1ubuntu2~16.0…

a5532a1

…4.1 (tags/RELEASE_600/final)

SteveBronder previously approved these changes Jan 11, 2021

View reviewed changes

Restart tests (Issue #2101)

76733bd

bbbales2 dismissed SteveBronder’s stale review via 76733bd January 11, 2021 18:02

[Jenkins] auto-formatting by clang-format version 6.0.0-1ubuntu2~16.0…

8366815

…4.1 (tags/RELEASE_600/final)

rok-cesnovar approved these changes Jan 12, 2021

View reviewed changes

rok-cesnovar merged commit 7252021 into develop Jan 12, 2021

rok-cesnovar deleted the feature/varmat-ldlt2 branch January 12, 2021 12:22

rok-cesnovar mentioned this pull request Jan 19, 2021

Feature freeze for Stan 2.26 #2327

Closed

bbbales2 mentioned this pull request Feb 3, 2021

2.28 Varmat issue stan-dev/cmdstan#980

Open

13 tasks

hsbadr mentioned this pull request Mar 1, 2021

Issues with the experimental version of 2.26 from PR #887 stan-dev/rstan#899

Open

5 tasks

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Make `varmat` functions work with ldlt (Issue #2101) #2266

Make `varmat` functions work with ldlt (Issue #2101) #2266

bbbales2 commented Dec 18, 2020 •

edited

Loading

bbbales2 Dec 19, 2020

bbbales2 commented Dec 19, 2020

bbbales2 commented Dec 21, 2020 •

edited

Loading

bbbales2 commented Jan 1, 2021 •

edited

Loading

stan-buildbot commented Jan 2, 2021

stan-buildbot commented Jan 5, 2021

bbbales2 commented Jan 5, 2021

stan-buildbot commented Jan 6, 2021

stan-buildbot commented Jan 7, 2021

SteveBronder left a comment

SteveBronder Jan 10, 2021

bbbales2 Jan 11, 2021

bbbales2 commented Jan 11, 2021

bbbales2 commented Jan 11, 2021

bbbales2 commented Jan 11, 2021

SteveBronder left a comment

stan-buildbot commented Jan 12, 2021

bbbales2 commented Jan 12, 2021

Make varmat functions work with ldlt (Issue #2101) #2266

Make varmat functions work with ldlt (Issue #2101) #2266

Conversation

bbbales2 commented Dec 18, 2020 • edited Loading

Summary

Side Effects

Release notes

Checklist

bbbales2 Dec 19, 2020

Choose a reason for hiding this comment

bbbales2 commented Dec 19, 2020

bbbales2 commented Dec 21, 2020 • edited Loading

bbbales2 commented Jan 1, 2021 • edited Loading

stan-buildbot commented Jan 2, 2021

stan-buildbot commented Jan 5, 2021

bbbales2 commented Jan 5, 2021

stan-buildbot commented Jan 6, 2021

stan-buildbot commented Jan 7, 2021

SteveBronder left a comment

Choose a reason for hiding this comment

SteveBronder Jan 10, 2021

Choose a reason for hiding this comment

bbbales2 Jan 11, 2021

Choose a reason for hiding this comment

bbbales2 commented Jan 11, 2021

bbbales2 commented Jan 11, 2021

bbbales2 commented Jan 11, 2021

SteveBronder left a comment

Choose a reason for hiding this comment

stan-buildbot commented Jan 12, 2021

bbbales2 commented Jan 12, 2021

Make `varmat` functions work with ldlt (Issue #2101) #2266

Make `varmat` functions work with ldlt (Issue #2101) #2266

bbbales2 commented Dec 18, 2020 •

edited

Loading

bbbales2 commented Dec 21, 2020 •

edited

Loading

bbbales2 commented Jan 1, 2021 •

edited

Loading