Use logistic function from eigen (based on jachymb's PR) by SteveBronder · Pull Request #3160 · stan-dev/math · GitHub

Summary

Modifies inv_logit to use the logistic function from Eigen. Also makes a new reverse mode specialization of inv_logit for Eigen matrices.

Tests

No new tests.

Checklist

Copyright holder: Jáchym Barvínek and Steve Bronder

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

Can this use apply_vector_unary? That way the Eigen implementation will also be used for std::vector (and nested containers)

…gistic

@andrjohns do you mean something besides the definition in prim that uses apply_scalar_unary?

template <typename T, require_std_vector_t<T>* = nullptr>
inline auto inv_logit(T&& x) {
  return apply_scalar_unary<inv_logit_fun, std::decay_t<T>>::apply(
      std::forward<T>(x));
}

@andrjohns do you mean something besides the definition in prim that uses apply_scalar_unary?
template <typename T, require_std_vector_t<T>* = nullptr>
inline auto inv_logit(T&& x) {
  return apply_scalar_unary<inv_logit_fun, std::decay_t<T>>::apply(
      std::forward<T>(x));
}

Yeah the issue with the current definition is that it won't use the vectorised Eigen implementation for std::vector types (either std::vector<double> or std::vector<Eigen::...>). You can instead use the pattern from prim/fun/exp where apply_scalar_unary is used for non-arithmetic containers and apply_vector_unary is used for arithmetic containers:

template <typename Container, require_ad_container_t<Container>* = nullptr>
inline auto inv_logit(const Container& x) {
  return apply_scalar_unary<inv_logit_fun, Container>::apply(x);
}

template <typename Container,
          require_container_bt<std::is_arithmetic, Container>* = nullptr>
inline auto inv_logit(const Container& x) {
  return apply_vector_unary<Container>::apply(
      x, [](const auto& v) { return v.array().logistic(); });
}

This way the Eigen logistic function will be used for all arithmetic containers (even arbitrarily nested ones)

apply_vector_unary also handles returning an Eigen::Matrix vs Eigen::Array to match the input type, which would resolve the current unit test failures

Name	Old Result	New Result	Ratio	Performance change( 1 - new / old )
arma/arma.stan	0.39	0.42	0.92	-8.5% slower
low_dim_corr_gauss/low_dim_corr_gauss.stan	0.01	0.01	0.98	-1.8% slower
gp_regr/gen_gp_data.stan	0.03	0.03	0.98	-2.32% slower
gp_regr/gp_regr.stan	0.1	0.1	0.98	-1.86% slower
sir/sir.stan	76.6	74.73	1.03	2.44% faster
irt_2pl/irt_2pl.stan	4.97	4.93	1.01	0.82% faster
eight_schools/eight_schools.stan	0.06	0.06	1.05	4.46% faster
pkpd/sim_one_comp_mm_elim_abs.stan	0.28	0.28	0.99	-0.56% slower
pkpd/one_comp_mm_elim_abs.stan	21.06	20.4	1.03	3.12% faster
garch/garch.stan	0.51	0.48	1.06	5.55% faster
low_dim_gauss_mix/low_dim_gauss_mix.stan	2.93	3.17	0.92	-8.47% slower
arK/arK.stan	1.93	1.86	1.04	3.91% faster
gp_pois_regr/gp_pois_regr.stan	3.09	2.97	1.04	3.92% faster
low_dim_gauss_mix_collapse/low_dim_gauss_mix_collapse.stan	9.57	9.11	1.05	4.84% faster
performance.compilation	205.18	203.36	1.01	0.89% faster
Mean result: 1.006107982606331

Jenkins Console Log
Blue Ocean
Commit hash: fa89f3f2eeace783e24dc5d33e94f737d6e7f2f7

Machine information No LSB modules are available. Distributor ID: Ubuntu Description: Ubuntu 20.04.3 LTS Release: 20.04 Codename: focal

CPU:
Architecture: x86_64
CPU op-mode(s): 32-bit, 64-bit
Byte Order: Little Endian
Address sizes: 46 bits physical, 48 bits virtual
CPU(s): 80
On-line CPU(s) list: 0-79
Thread(s) per core: 2
Core(s) per socket: 20
Socket(s): 2
NUMA node(s): 2
Vendor ID: GenuineIntel
CPU family: 6
Model: 85
Model name: Intel(R) Xeon(R) Gold 6148 CPU @ 2.40GHz
Stepping: 4
CPU MHz: 3393.897
CPU max MHz: 3700.0000
CPU min MHz: 1000.0000
BogoMIPS: 4800.00
Virtualization: VT-x
L1d cache: 1.3 MiB
L1i cache: 1.3 MiB
L2 cache: 40 MiB
L3 cache: 55 MiB
NUMA node0 CPU(s): 0,2,4,6,8,10,12,14,16,18,20,22,24,26,28,30,32,34,36,38,40,42,44,46,48,50,52,54,56,58,60,62,64,66,68,70,72,74,76,78
NUMA node1 CPU(s): 1,3,5,7,9,11,13,15,17,19,21,23,25,27,29,31,33,35,37,39,41,43,45,47,49,51,53,55,57,59,61,63,65,67,69,71,73,75,77,79
Vulnerability Gather data sampling: Mitigation; Microcode
Vulnerability Itlb multihit: KVM: Mitigation: VMX disabled
Vulnerability L1tf: Mitigation; PTE Inversion; VMX conditional cache flushes, SMT vulnerable
Vulnerability Mds: Mitigation; Clear CPU buffers; SMT vulnerable
Vulnerability Meltdown: Mitigation; PTI
Vulnerability Mmio stale data: Mitigation; Clear CPU buffers; SMT vulnerable
Vulnerability Reg file data sampling: Not affected
Vulnerability Retbleed: Mitigation; IBRS
Vulnerability Spec rstack overflow: Not affected
Vulnerability Spec store bypass: Mitigation; Speculative Store Bypass disabled via prctl
Vulnerability Spectre v1: Mitigation; usercopy/swapgs barriers and __user pointer sanitization
Vulnerability Spectre v2: Mitigation; IBRS; IBPB conditional; STIBP conditional; RSB filling; PBRSB-eIBRS Not affected; BHI Not affected
Vulnerability Srbds: Not affected
Vulnerability Tsx async abort: Mitigation; Clear CPU buffers; SMT vulnerable
Flags: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx pdpe1gb rdtscp lm constant_tsc art arch_perfmon pebs bts rep_good nopl xtopology nonstop_tsc cpuid aperfmperf pni pclmulqdq dtes64 monitor ds_cpl vmx smx est tm2 ssse3 sdbg fma cx16 xtpr pdcm pcid dca sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand lahf_lm abm 3dnowprefetch cpuid_fault epb cat_l3 cdp_l3 invpcid_single pti intel_ppin ssbd mba ibrs ibpb stibp tpr_shadow vnmi flexpriority ept vpid ept_ad fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm cqm mpx rdt_a avx512f avx512dq rdseed adx smap clflushopt clwb intel_pt avx512cd avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local dtherm ida arat pln pts hwp hwp_act_window hwp_epp hwp_pkg_req pku ospke md_clear flush_l1d arch_capabilities

G++:
g++ (Ubuntu 9.4.0-1ubuntu1~20.04) 9.4.0
Copyright (C) 2019 Free Software Foundation, Inc.
This is free software; see the source for copying conditions. There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

Clang:
clang version 10.0.0-4ubuntu1
Target: x86_64-pc-linux-gnu
Thread model: posix
InstalledDir: /usr/bin

Just a couple of doc changes, but otherwise LGTM!

This overload would also be used for Eigen types with fvars I think

This overload would also be for std::vector with inner types of double or std::vector

Note there is also the Eigen::internal::scalar_logistic_op which can be used for the scalar (double) implementation instead of the custom implementation using exp calls. Not sure if there is any noticeable advantage besides shortening the code by a few lines.

@andrjohns ty those docs should be cleaned up now!

@jachymb personally I don't know what the benefit of using Eigen's logistic function for scalars would be. For the vectorized version of the code, using the Eigen logistic function makes sense so we can use their SIMD backend. But for scalars we would want to check the performance and precision before using it. I'm fine with hand waving the vectorized check because using an Eigen internal function is almost always faster for vectors.

LGTM!

Name	Old Result	New Result	Ratio	Performance change( 1 - new / old )
arma/arma.stan	0.3	0.3	0.99	-0.7% slower
low_dim_corr_gauss/low_dim_corr_gauss.stan	0.01	0.01	1.03	2.96% faster
gp_regr/gen_gp_data.stan	0.02	0.02	1.03	2.97% faster
gp_regr/gp_regr.stan	0.09	0.09	1.01	1.23% faster
sir/sir.stan	66.24	67.37	0.98	-1.69% slower
irt_2pl/irt_2pl.stan	3.98	4.0	0.99	-0.6% slower
eight_schools/eight_schools.stan	0.06	0.05	1.06	5.98% faster
pkpd/sim_one_comp_mm_elim_abs.stan	0.26	0.24	1.07	6.88% faster
pkpd/one_comp_mm_elim_abs.stan	18.56	18.73	0.99	-0.91% slower
garch/garch.stan	0.5	0.41	1.23	18.57% faster
low_dim_gauss_mix/low_dim_gauss_mix.stan	2.75	2.59	1.06	5.91% faster
arK/arK.stan	1.86	1.72	1.08	7.51% faster
gp_pois_regr/gp_pois_regr.stan	2.9	2.73	1.06	5.93% faster
low_dim_gauss_mix_collapse/low_dim_gauss_mix_collapse.stan	8.81	8.43	1.05	4.32% faster
performance.compilation	180.18	177.65	1.01	1.4% faster
Mean result: 1.044470247179379

Jenkins Console Log
Blue Ocean
Commit hash: 0d6bd650ebd331395197dd352b2d23d459fb7626

Machine information No LSB modules are available. Distributor ID: Ubuntu Description: Ubuntu 20.04.3 LTS Release: 20.04 Codename: focal

CPU:
Architecture: x86_64
CPU op-mode(s): 32-bit, 64-bit
Byte Order: Little Endian
Address sizes: 46 bits physical, 48 bits virtual
CPU(s): 80
On-line CPU(s) list: 0-79
Thread(s) per core: 2
Core(s) per socket: 20
Socket(s): 2
NUMA node(s): 2
Vendor ID: GenuineIntel
CPU family: 6
Model: 85
Model name: Intel(R) Xeon(R) Gold 6148 CPU @ 2.40GHz
Stepping: 4
CPU MHz: 2400.000
CPU max MHz: 3700.0000
CPU min MHz: 1000.0000
BogoMIPS: 4800.00
Virtualization: VT-x
L1d cache: 1.3 MiB
L1i cache: 1.3 MiB
L2 cache: 40 MiB
L3 cache: 55 MiB
NUMA node0 CPU(s): 0,2,4,6,8,10,12,14,16,18,20,22,24,26,28,30,32,34,36,38,40,42,44,46,48,50,52,54,56,58,60,62,64,66,68,70,72,74,76,78
NUMA node1 CPU(s): 1,3,5,7,9,11,13,15,17,19,21,23,25,27,29,31,33,35,37,39,41,43,45,47,49,51,53,55,57,59,61,63,65,67,69,71,73,75,77,79
Vulnerability Gather data sampling: Mitigation; Microcode
Vulnerability Itlb multihit: KVM: Mitigation: VMX disabled
Vulnerability L1tf: Mitigation; PTE Inversion; VMX conditional cache flushes, SMT vulnerable
Vulnerability Mds: Mitigation; Clear CPU buffers; SMT vulnerable
Vulnerability Meltdown: Mitigation; PTI
Vulnerability Mmio stale data: Mitigation; Clear CPU buffers; SMT vulnerable
Vulnerability Reg file data sampling: Not affected
Vulnerability Retbleed: Mitigation; IBRS
Vulnerability Spec rstack overflow: Not affected
Vulnerability Spec store bypass: Mitigation; Speculative Store Bypass disabled via prctl
Vulnerability Spectre v1: Mitigation; usercopy/swapgs barriers and __user pointer sanitization
Vulnerability Spectre v2: Mitigation; IBRS; IBPB conditional; STIBP conditional; RSB filling; PBRSB-eIBRS Not affected; BHI Not affected
Vulnerability Srbds: Not affected
Vulnerability Tsx async abort: Mitigation; Clear CPU buffers; SMT vulnerable
Flags: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx pdpe1gb rdtscp lm constant_tsc art arch_perfmon pebs bts rep_good nopl xtopology nonstop_tsc cpuid aperfmperf pni pclmulqdq dtes64 monitor ds_cpl vmx smx est tm2 ssse3 sdbg fma cx16 xtpr pdcm pcid dca sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand lahf_lm abm 3dnowprefetch cpuid_fault epb cat_l3 cdp_l3 invpcid_single pti intel_ppin ssbd mba ibrs ibpb stibp tpr_shadow vnmi flexpriority ept vpid ept_ad fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm cqm mpx rdt_a avx512f avx512dq rdseed adx smap clflushopt clwb intel_pt avx512cd avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local dtherm ida arat pln pts hwp hwp_act_window hwp_epp hwp_pkg_req pku ospke md_clear flush_l1d arch_capabilities

G++:
g++ (Ubuntu 9.4.0-1ubuntu1~20.04) 9.4.0
Copyright (C) 2019 Free Software Foundation, Inc.
This is free software; see the source for copying conditions. There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

Clang:
clang version 10.0.0-4ubuntu1
Target: x86_64-pc-linux-gnu
Thread model: posix
InstalledDir: /usr/bin

SteveBronder and others added 3 commits March 5, 2025 13:51

fixes signatures and rev for inv_logit

0258fca

docs

88eaa7c

[Jenkins] auto-formatting by clang-format version 10.0.0-4ubuntu1

5ec57f7

Merge remote-tracking branch 'origin/develop' into fix/jachymb-inv_lo…

3703953

…gistic

SteveBronder and others added 3 commits March 7, 2025 13:32

Allow Eigen to use the apply_scalar_unary framework

87bf545

Merge commit '0f1ad25835e12bfeccb357db4323ec35841fa837' into HEAD

2c71c06

[Jenkins] auto-formatting by clang-format version 10.0.0-4ubuntu1

79a929a

andrjohns requested changes Mar 8, 2025

View reviewed changes

SteveBronder and others added 3 commits March 13, 2025 16:52

update docs

d0921c8

Merge commit '8b8057ae220ba325e1ae38fb1adc48ceead52b0a' into HEAD

cb7f44d

[Jenkins] auto-formatting by clang-format version 10.0.0-4ubuntu1

b92cf29

andrjohns approved these changes Mar 14, 2025

View reviewed changes

SteveBronder merged commit 0717118 into develop Mar 14, 2025

WardBrian deleted the fix/jachymb-inv_logistic branch March 14, 2025 21:14

jachymb mentioned this pull request Mar 18, 2025

Use eigen logistic function for inv_logit #3154

Closed

+               * Vectorized version of inv_logit() for std::vector's containing ad types.
                *
-               * @tparam T type of container
-               * @param x container
+               * @tparam T type of std::vector
+               * @param x std::vector

+               * @tparam T A type of either `std::vector` whose inner type inherits from
+               * `Eigen::DenseBase` or a type that directly inherits from `Eigen::DenseBase`.
+               * The inner scalar type must not have a `var` scalar type.
+               * @param x Eigen expression

Use logistic function from eigen (based on jachymb's PR) by SteveBronder · Pull Request #3160 · stan-dev/math · GitHub

Uh oh!

Uh oh!

Use logistic function from eigen (based on jachymb's PR) - #3160

Use logistic function from eigen (based on jachymb's PR)#3160
SteveBronder merged 10 commits into
developfrom
fix/jachymb-inv_logistic

SteveBronder commented Mar 5, 2025

Uh oh!

andrjohns commented Mar 6, 2025

Uh oh!

SteveBronder commented Mar 6, 2025 •
edited
Loading

Uh oh!

Uh oh!

andrjohns commented Mar 7, 2025

Uh oh!

andrjohns commented Mar 7, 2025

Uh oh!

stan-buildbot commented Mar 8, 2025

Uh oh!

andrjohns left a comment

Uh oh!

andrjohns Mar 8, 2025

Uh oh!

andrjohns Mar 8, 2025

Uh oh!

jachymb commented Mar 10, 2025

Uh oh!

SteveBronder commented Mar 13, 2025

Uh oh!

andrjohns left a comment

Uh oh!

stan-buildbot commented Mar 14, 2025

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

Use logistic function from eigen (based on jachymb's PR) by SteveBronder · Pull Request #3160 · stan-dev/math · GitHub

Uh oh!

Uh oh!

Use logistic function from eigen (based on jachymb's PR) - #3160

Use logistic function from eigen (based on jachymb's PR)#3160SteveBronder merged 10 commits intodevelopstan-dev/math:developfrom fix/jachymb-inv_logisticstan-dev/math:fix/jachymb-inv_logisticCopy head branch name to clipboard

Conversation

SteveBronder commented Mar 5, 2025

Summary

Tests

Checklist

Uh oh!

andrjohns commented Mar 6, 2025

Uh oh!

SteveBronder commented Mar 6, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

andrjohns commented Mar 7, 2025

Uh oh!

andrjohns commented Mar 7, 2025

Uh oh!

stan-buildbot commented Mar 8, 2025

Uh oh!

andrjohns left a comment

Choose a reason for hiding this comment

Uh oh!

andrjohns Mar 8, 2025

Choose a reason for hiding this comment

Uh oh!

andrjohns Mar 8, 2025

Choose a reason for hiding this comment

Uh oh!

jachymb commented Mar 10, 2025

Uh oh!

SteveBronder commented Mar 13, 2025

Uh oh!

andrjohns left a comment

Choose a reason for hiding this comment

Uh oh!

stan-buildbot commented Mar 14, 2025

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

Use logistic function from eigen (based on jachymb's PR)#3160
SteveBronder merged 10 commits into
developfrom
fix/jachymb-inv_logistic

SteveBronder commented Mar 6, 2025 •
edited
Loading