Generalize operator-like functions #1628

t4c1 · 2020-01-20T09:48:26Z

Summary

Generalizes function signatures of functions that act kind of like operators (add, subtract, multiply, divide, elt_multiply, elt_divide). That is they can now accept and return general Eigen expressions.
EDIT: I also had to generalize functions that were called directely by return value of one of generalized functions. Those were: trace, trace_gen_quad_form, transpose and matrix_exp.

In many cases I replaced function specializations with overloads. This results in more requires, but it also allowed me to replace multiple specializations with one general overload, reducing total amount of code.

There are also some changes to autodiff testing framework that allow it to work with Eigen expressions. I did some of those in earlier PR #1471, but I missed some places.

Tests

This is mostly a refactor so no new tests.

Side Effects

None.

Checklist

Math issue Generalize matrix function signatures #1470
Copyright holder: Tadej Ciglarič

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)
- 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

…stable/2017-11-14)

t4c1 · 2020-01-20T09:57:35Z

stan/math/fwd/fun/multiply.hpp

+    = require_t<bool_constant<T1::ColsAtCompileTime == T2::RowsAtCompileTime>>,
+    typename = require_not_t<
+        conjunction<is_eigen_row_vector<T1>, is_eigen_col_vector<T2>>>,
+    int = 0>


This int template parameter is required to make compiler treat these as different overloads (if they have same arguments and template parameters it treats them as same overload and gives a lot of errors about redefining default template parameter values.

! O that's how you do that. I was fighting with that compile error like a week ago. Thank you. :)

To me this solution seems a bit hacky, but I don't know any better one.

I'm not a fan of this approach. It's feels like it fights against the language. I think it would be good to post a stackoverflow Q about this

I found a nice solution here: https://stackoverflow.com/questions/38305222/default-template-argument-when-using-stdenable-if-as-templ-param-why-ok-wit

The trick is to use requires as type of template parameters instead of defaults to typename parameters - require_t<...>* = nullptr instead of typename = require_t<...>

stan/math/fwd/fun/multiply.hpp

t4c1 · 2020-01-20T10:02:19Z

stan/math/prim/fun/dot_product.hpp

-  check_vector("dot_product", "v1", v1);
-  check_vector("dot_product", "v2", v2);
+template <typename T1, typename T2,
+          typename = require_all_eigen_vector_t<T1, T2>>
+inline auto dot_product(const T1 &v1, const T2 &v2) {
  check_matching_sizes("dot_product", "v1", v1, "v2", v2);
  return v1.dot(v2);


These check_vector checks were redundant, as Eigen's .dot() is only defined for compile time vectors, so there is no way this could compile and check fail.

stan/math/prim/fun/multiply.hpp

t4c1 · 2020-01-20T12:57:50Z

stan/math/prim/fun/trace_gen_quad_form.hpp

@@ -68,10 +60,11 @@ inline return_type_t<TD, TA, TB> trace_gen_quad_form(
 * @throw std::domain_error if A cannot be multiplied by B or B cannot
 * be multiplied by D.
 */
-template <int RD, int CD, int RA, int CA, typename TB, int RB, int CB>


typename TB is never used and therefore can not be deduced. Due to the bug this overload was never chosen.

t4c1 · 2020-01-20T12:59:37Z

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

-  tols.hessian_hessian_ = 1e-1;
-  tols.hessian_fvar_hessian_ = 1e-1;
+  tols.hessian_hessian_ = 0.15;
+  tols.hessian_fvar_hessian_ = 0.15;


I fixed a bug that prevented all-double overload of trace_gen_quad_form() to be ever chosen. While faster, that overload seems to be a bit less accurate, so I had to increase these tolerances.

…xp1~20180509124008.99 (branches/release_50)

SteveBronder

Just looked over the metaprogramming. I like it! I'll sit down and look at more once the tests pass

SteveBronder · 2020-01-21T16:14:03Z

stan/math/fwd/fun/multiply.hpp

-    const Eigen::Matrix<fvar<T>, R2, C2>& m2) {
+template <typename T1, typename T2,
+          typename = require_all_eigen_vt<is_fvar, T1, T2>,
+          typename = require_same_t<typename value_type_t<T1>::Scalar,


There is require_same_st and require_same_vt

https://github.com/stan-dev/math/blob/develop/stan/math/prim/meta/require_generics.hpp#L283

The problem here is that scalar_t<T1> is fvar<X> and I want to compare those Xes.

Whoops, I was wrong. Of course require_same_vt is fine.

SteveBronder · 2020-01-21T16:17:43Z

stan/math/fwd/fun/multiply.hpp

+          typename = require_same_t<typename value_type_t<T1>::Scalar,
+                                    typename value_type_t<T2>::Scalar>,
+          typename = require_not_t<
+              conjunction<is_eigen_row_vector<T1>, is_eigen_col_vector<T2>>>,


[optional]
We could have something like `require_eigen_dot_product (I'm not sure if there is some other math word for general ops with a row and column vector.

Also I think this could use require_any_not_t

Also names like RowVec and ColVec may be good here

Here we have require_NOT, so these are not row and col vec.

SteveBronder · 2020-01-21T16:21:43Z

stan/math/fwd/fun/multiply.hpp

-    const Eigen::Matrix<double, R2, C2>& m2) {
+template <typename T1, typename T2, typename = require_all_eigen_t<T1, T2>,
+          typename = require_fvar_t<value_type_t<T1>>,
+          typename = require_same_vt<double, T2>,


Could shorten this up with

template <typename Mat1, typename Mat2, typename = require_eigen_vt<is_fvar, Mat1>, typename = require_eigen_vt<std::is_floating_point, Mat2>

SteveBronder · 2020-01-21T16:23:06Z

stan/math/fwd/fun/multiply.hpp

+          typename = require_not_t<
+              conjunction<is_eigen_row_vector<T1>, is_eigen_col_vector<T2>>>,


Maybe just a meta function for is_dot_product for these instead of a requires specialization

stan/math/fwd/fun/multiply.hpp

SteveBronder · 2020-01-21T16:44:25Z

stan/math/prim/fun/trace_gen_quad_form.hpp

+          typename = require_all_not_var_t<value_type_t<TD>, value_type_t<TA>,
+                                           value_type_t<TB>>,
+          typename = require_any_not_same_t<double, value_type_t<TD>,
+                                            value_type_t<TA>, value_type_t<TB>>,


It might be a weird way to express it but one way to look at this is

require_not_same_t<double, return_type_t<TD, TA, TB>>

While this is shorter, I find it less clear.

SteveBronder · 2020-01-21T16:47:13Z

stan/math/prim/fun/trace_gen_quad_form.hpp

+template <typename TD, typename TA, typename TB,
+          typename = require_all_same_t<double, value_type_t<TD>,
+                                        value_type_t<TA>, value_type_t<TB>>,
+          typename = require_all_eigen_t<TD, TA, TB>>


require_all_eigen_vt<std::is_floating_point, TD, TA, TB>

If we ever worry about mixing types of floats we could also have

require_same_same_vt<TD, TA, TB>

Though the bottom may be better as a static assert to the user

Static assert is not what we want here - if that require is not satisfied, other overload must be chosen.

Anyway I would not complicate here, since this overload can be removed once transpose is generalized too.

stan/math/prim/fun/transpose.hpp

stan/math/rev/fun/divide.hpp

SteveBronder · 2020-01-21T16:50:11Z

stan/math/rev/fun/divide.hpp

-      = new internal::matrix_scalar_divide_vd_vari<R, C>(m, c);
-  Eigen::Matrix<var, R, C> result(m.rows(), m.cols());
+template <typename T, typename = require_eigen_vt<is_var, T>>
+inline Eigen::Matrix<var, T::RowsAtCompileTime, T::ColsAtCompileTime> divide(


Does auto not work here?

It does. I missed that one.

…stable/2017-11-14)

into generalize_operators # Conflicts: # stan/math/fwd/fun/multiply.hpp

into generalize_operators

…stable/2017-11-14)

SteveBronder · 2020-01-22T20:02:27Z

For the errors up in the stan repo I started working on generalizing the rvalue and assign functions. Should those all take in Eigen::MatrixBase<> objects (or just a EigMat template with a require_*?) and call eval if it's being assigned to or we need to do coefficient level access?

https://github.com/stan-dev/stan/blob/refactor/rvalue-pass/src/stan/model/indexing/rvalue.hpp#L114

If that branch is helpful and you want to make any changes to that branch feel free to push straight to it

SteveBronder · 2020-01-22T20:03:42Z

I think the lockstep here is updating the stan repo first then merged this

t4c1 · 2020-01-23T09:40:46Z

I agree. Eigen::MatrixBase<> is not as general. I think we should either have a general template and require or Eigen::EigenBase<>. Let's just go with template + require.

Looking over the branch you linked I see many changes, none of which seem like they could fix test failures in this PR. Did you link the right branch?

SteveBronder · 2020-01-23T17:46:24Z

Looking over the branch you linked I see many changes, none of which seem like they could fix test failures in this PR. Did you link the right branch?

Crap actually sorry this branch doesn't cover that. I thought I had a branch with these but looking on my local I must not have saved it

If we use the generic template with require should we have inside each function that does coeff access like

if (is_eigen_expression<EigMat>::value) {
  the_mat.eval();
}

Or is there some better pattern?

t4c1 · 2020-01-23T20:33:07Z

What you suggested has multiple issues. The correct pattern would be:

const auto& mat = expr_argument.eval();

where expr_argument is function argument and mat is used for reminder of the function (if is not necessary). However we can do even better and use Eigen::Ref to avoid evaling expression if it is just a sub block operation, which allows element access. So the pattern becomes:

const Eigen::Ref<const typename T::PlainObject>& mat = expr_argument;

where T is the type of expr_argument.

…stable/2017-11-14)

stan-buildbot · 2020-02-03T17:25:12Z

Name	Old Result	New Result	Ratio	Performance change( 1 - new / old )
gp_pois_regr/gp_pois_regr.stan	4.89	4.98	0.98	-1.8% slower
low_dim_corr_gauss/low_dim_corr_gauss.stan	0.02	0.02	0.99	-1.42% slower
eight_schools/eight_schools.stan	0.09	0.09	1.01	0.9% faster
gp_regr/gp_regr.stan	0.22	0.23	0.96	-3.82% slower
irt_2pl/irt_2pl.stan	6.08	6.07	1.0	0.3% faster
performance.compilation	89.16	87.17	1.02	2.23% faster
low_dim_gauss_mix_collapse/low_dim_gauss_mix_collapse.stan	7.48	7.32	1.02	2.13% faster
pkpd/one_comp_mm_elim_abs.stan	21.66	21.07	1.03	2.72% faster
sir/sir.stan	92.92	91.67	1.01	1.35% faster
gp_regr/gen_gp_data.stan	0.05	0.05	1.01	1.13% faster
low_dim_gauss_mix/low_dim_gauss_mix.stan	2.99	2.99	1.0	-0.25% slower
pkpd/sim_one_comp_mm_elim_abs.stan	0.33	0.3	1.08	7.32% faster
arK/arK.stan	1.76	1.76	1.0	0.19% faster
arma/arma.stan	0.8	0.74	1.08	7.76% faster
garch/garch.stan	0.62	0.63	0.99	-1.12% slower
Mean result: 1.01283994857

Jenkins Console Log
Blue Ocean
Commit hash: b160720

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

# Conflicts: # test/unit/math/mix/fun/trace_gen_quad_form_test.cpp

stan-buildbot · 2020-02-10T13:20:00Z

Name	Old Result	New Result	Ratio	Performance change( 1 - new / old )
gp_pois_regr/gp_pois_regr.stan	4.86	4.87	1.0	-0.19% slower
low_dim_corr_gauss/low_dim_corr_gauss.stan	0.02	0.02	0.96	-4.6% slower
eight_schools/eight_schools.stan	0.09	0.09	0.99	-0.88% slower
gp_regr/gp_regr.stan	0.22	0.22	1.0	-0.33% slower
irt_2pl/irt_2pl.stan	6.08	6.05	1.0	0.33% faster
performance.compilation	88.12	86.97	1.01	1.31% faster
low_dim_gauss_mix_collapse/low_dim_gauss_mix_collapse.stan	7.49	7.48	1.0	0.09% faster
pkpd/one_comp_mm_elim_abs.stan	22.46	21.4	1.05	4.73% faster
sir/sir.stan	93.01	98.56	0.94	-5.96% slower
gp_regr/gen_gp_data.stan	0.05	0.05	1.0	0.03% faster
low_dim_gauss_mix/low_dim_gauss_mix.stan	3.0	2.99	1.0	0.24% faster
pkpd/sim_one_comp_mm_elim_abs.stan	0.32	0.32	1.0	-0.03% slower
arK/arK.stan	1.74	1.74	1.0	-0.0% slower
arma/arma.stan	0.79	0.79	1.01	0.5% faster
garch/garch.stan	0.59	0.58	1.0	0.16% faster
Mean result: 0.997466204586

Jenkins Console Log
Blue Ocean
Commit hash: f88d89a

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

t4c1 · 2020-02-10T13:48:00Z

@SteveBronder This is still waiting for a review.

SteveBronder · 2020-02-11T06:17:09Z

Sorry for the delay it's been a v busy week. I should have time tomorrow night to look at this

# Conflicts: # stan/math/fwd/fun/multiply.hpp # stan/math/prim/fun/multiply.hpp # stan/math/rev/fun/multiply.hpp

SteveBronder · 2020-02-12T04:08:41Z

stan/math/fwd/fun/multiply.hpp

+template <typename Mat1, typename Mat2,
+          typename = require_eigen_vt<is_fvar, Mat1>,
+          typename = require_eigen_vt<std::is_floating_point, Mat2>,
+          typename = require_not_eigen_row_and_col_t<Mat1, Mat2>, int = 0>
+inline auto multiply(const Mat1& m1, const Mat2& m2) {


I think we need to do these multiplies differently. This and the ones in prim are pretty confusing

SteveBronder · 2020-02-12T04:11:42Z

stan/math/prim/fun/divide.hpp

+template <typename Mat, typename Scal, typename = require_eigen_t<Mat>,
+          typename = require_stan_scalar_t<Scal>,
+          typename = require_all_not_var_t<scalar_type_t<Mat>, Scal>>
+inline auto divide(const Mat& m, Scal c) {


Is this so that divide works for fvar?

This one works for fvar. I just generalized it the same way as other functions.

SteveBronder · 2020-02-12T04:14:01Z

stan/math/prim/fun/multiply.hpp

+          typename = require_any_not_same_t<double, value_type_t<Mat1>,
+                                            value_type_t<Mat2>>,
+          typename = require_not_eigen_row_and_col_t<Mat1, Mat2>>


[sidenote]

We should probably have a require_*_vt for checking the value types of any arbitrary container

SteveBronder · 2020-02-12T04:15:28Z

stan/math/prim/fun/trace_gen_quad_form.hpp

-                                  const Eigen::Matrix<double, RB, CB> &B) {
+template <typename TD, typename TA, typename TB,
+          typename = require_all_eigen_t<TD, TA, TB>,
+          typename = require_all_same_vt<double, TD, TA, TB>>


Should we not do this for all arithmetics?

You are right. Done.

stan/math/prim/fun/transpose.hpp

SteveBronder · 2020-02-12T04:23:55Z

I looked this over and like a lot of it! But the multiply needs cleaned up. The templates there are v complicated, the requires stuff is pretty complicated to read for those. Do you have any ideas on how else to do those? I can look at this again on Thursday and see if I can hack up something else

SteveBronder · 2020-02-12T04:32:59Z

If @syclik or @bob-carpenter sign off on the multiply templates then I'm fine with it but I'm not sure about them

stan-buildbot · 2020-02-13T04:29:51Z

Name	Old Result	New Result	Ratio	Performance change( 1 - new / old )
gp_pois_regr/gp_pois_regr.stan	4.94	4.93	1.0	0.22% faster
low_dim_corr_gauss/low_dim_corr_gauss.stan	0.02	0.02	0.97	-3.37% slower
eight_schools/eight_schools.stan	0.09	0.09	1.0	0.41% faster
gp_regr/gp_regr.stan	0.22	0.22	1.01	0.76% faster
irt_2pl/irt_2pl.stan	6.11	6.05	1.01	0.99% faster
performance.compilation	88.37	87.51	1.01	0.97% faster
low_dim_gauss_mix_collapse/low_dim_gauss_mix_collapse.stan	7.5	7.49	1.0	0.2% faster
pkpd/one_comp_mm_elim_abs.stan	20.16	21.02	0.96	-4.26% slower
sir/sir.stan	93.71	91.84	1.02	2.0% faster
gp_regr/gen_gp_data.stan	0.05	0.05	1.01	0.9% faster
low_dim_gauss_mix/low_dim_gauss_mix.stan	3.01	3.0	1.01	0.64% faster
pkpd/sim_one_comp_mm_elim_abs.stan	0.32	0.35	0.9	-11.44% slower
arK/arK.stan	1.75	1.73	1.01	1.18% faster
arma/arma.stan	0.79	0.8	0.99	-0.66% slower
garch/garch.stan	0.59	0.58	1.01	0.78% faster
Mean result: 0.993918240423

Jenkins Console Log
Blue Ocean
Commit hash: 9529ccc

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

t4c1 · 2020-02-13T11:12:30Z

@SteveBronder Now everything you mentioned is fixed, including multiply templates, and tests passed so this is ready for next review.

SteveBronder

lgtm!

t4c1 and others added 4 commits January 20, 2020 10:06

generalized operator-like functions

84b5ab6

optimize fwd multiply

7d7779e

Merge commit '10cc6ba675743f09832c6749fbb1a92d74888bd2' into HEAD

4b50381

[Jenkins] auto-formatting by clang-format version 6.0.0 (tags/google/…

3c23987

…stable/2017-11-14)

t4c1 commented Jan 20, 2020

View reviewed changes

t4c1 added 2 commits January 20, 2020 12:52

fixed call sites.

7cd74ef

fixed a test.

8263fdb

t4c1 commented Jan 20, 2020

View reviewed changes

t4c1 and others added 5 commits January 21, 2020 08:45

generalized transpose.hpp() input

fb2de8a

fixed transpose

1fedd01

[Jenkins] auto-formatting by clang-format version 5.0.2-svn328729-1~e…

9029b14

…xp1~20180509124008.99 (branches/release_50)

removed compile time multiplicable requirements

d73d0c8

Merge commit '12cbb63adfcfb64440ade29e97a25db30c5ac8e2' into HEAD

45319be

SteveBronder mentioned this pull request Jan 21, 2020

[WIP] Add specialization for getting cols and cleanup of rvalue/lvalue stan-dev/stan#2880

Closed

3 tasks

SteveBronder requested changes Jan 21, 2020

View reviewed changes

t4c1 and others added 10 commits January 22, 2020 09:01

fix unnecessary vector to matrix conversions

f989b11

Merge commit '81248fdcd9bbd107f63cf8211333c2acbdbe5ed8' into HEAD

01a87f4

[Jenkins] auto-formatting by clang-format version 6.0.0 (tags/google/…

d34d6fe

…stable/2017-11-14)

addressed review comments

c68504a

fix cpplint

aaa8c17

[Jenkins] auto-formatting by clang-format version 6.0.0 (tags/google/…

b98a5a0

…stable/2017-11-14)

Merge branch 'generalize_operators' of https://github.com/bstatcomp/math

9445e07

into generalize_operators # Conflicts: # stan/math/fwd/fun/multiply.hpp

generalized matrix_exp

9180fc6

Merge branch 'generalize_operators' of https://github.com/bstatcomp/math

ed9b73a

into generalize_operators

[Jenkins] auto-formatting by clang-format version 6.0.0 (tags/google/…

f2b4e13

…stable/2017-11-14)

stan-buildbot and others added 3 commits February 3, 2020 07:27

[Jenkins] auto-formatting by clang-format version 6.0.0 (tags/google/…

a9490b3

…stable/2017-11-14)

added missing include

c7162fd

one more header

b160720

Merge branch 'develop' into generalize_operators

f88d89a

# Conflicts: # test/unit/math/mix/fun/trace_gen_quad_form_test.cpp

Merge branch 'develop' into generalize_operators

d9284e8

# Conflicts: # stan/math/fwd/fun/multiply.hpp # stan/math/prim/fun/multiply.hpp # stan/math/rev/fun/multiply.hpp

SteveBronder reviewed Feb 12, 2020

View reviewed changes

stan/math/prim/fun/transpose.hpp Show resolved Hide resolved

t4c1 added 4 commits February 12, 2020 10:45

fixed size zero multiply

cf2b72d

Merge branch 'develop' into generalize_operators

fb619c0

fixed multiply templates

5290f6b

addressed remaining review comments

9529ccc

rok-cesnovar mentioned this pull request Feb 13, 2020

double version of trace_gen_quad_form is never used #1709

Closed

mcol linked an issue Feb 13, 2020 that may be closed by this pull request

double version of trace_gen_quad_form is never used #1709

Closed

SteveBronder approved these changes Feb 14, 2020

View reviewed changes

t4c1 merged commit dde9b47 into stan-dev:develop Feb 14, 2020

rok-cesnovar mentioned this pull request Feb 14, 2020

fix develop #1712

Merged

4 tasks

SteveBronder mentioned this pull request Apr 16, 2020

Stan Math 3.2 release #1826

Closed

bbbales2 mentioned this pull request Apr 20, 2020

Stanc3 release for Cmdstan 2.23 stan-dev/stanc3#498

Closed

t4c1 deleted the generalize_operators branch November 30, 2020 09:25

		typename = require_not_t<
		conjunction<is_eigen_row_vector<T1>, is_eigen_col_vector<T2>>>,

Generalize operator-like functions #1628

Generalize operator-like functions #1628

Conversation

t4c1 commented Jan 20, 2020 • edited Loading

Summary

Tests

Side Effects

Checklist

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

SteveBronder left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

SteveBronder commented Jan 22, 2020

SteveBronder commented Jan 22, 2020

t4c1 commented Jan 23, 2020

SteveBronder commented Jan 23, 2020

t4c1 commented Jan 23, 2020

stan-buildbot commented Feb 3, 2020

stan-buildbot commented Feb 10, 2020

t4c1 commented Feb 10, 2020

SteveBronder commented Feb 11, 2020

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

t4c1 Feb 12, 2020 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

SteveBronder commented Feb 12, 2020

SteveBronder commented Feb 12, 2020

stan-buildbot commented Feb 13, 2020

t4c1 commented Feb 13, 2020

SteveBronder left a comment

Choose a reason for hiding this comment

t4c1 commented Jan 20, 2020 •

edited

Loading

t4c1 Feb 12, 2020 •

edited

Loading