Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

#3127 Add log sum exp func #3131

Open
wants to merge 9 commits into
base: develop
Choose a base branch
from

Conversation

MichaScant
Copy link

Summary

As discussed in issue #3127, I've implemented a log sum exp function that computes the logarithm of the element-wise sum of exponentials and returns the result as a container.

I've implemented 3 files + a test file in both prim, rev, and fwd, handling different edge cases that may arise.

Tests

The tests are written in the following file path, I modelled them after the tests for log_sum_exp as a base example:
test/unit/math/mix/fun/log_add_exp_test.cpp

Side Effects

None

Release notes

log_add_exp will be available if merged

Checklist

  • Copyright holder: (fill in copyright holder information)

    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

@stan-buildbot
Copy link
Contributor


Name Old Result New Result Ratio Performance change( 1 - new / old )
arma/arma.stan 0.33 0.32 1.03 2.52% faster
low_dim_corr_gauss/low_dim_corr_gauss.stan 0.01 0.01 1.0 -0.06% slower
gp_regr/gen_gp_data.stan 0.03 0.03 1.03 2.81% faster
gp_regr/gp_regr.stan 0.1 0.09 1.03 2.73% faster
sir/sir.stan 70.52 70.01 1.01 0.72% faster
irt_2pl/irt_2pl.stan 4.29 4.14 1.04 3.42% faster
eight_schools/eight_schools.stan 0.06 0.06 0.99 -0.59% slower
pkpd/sim_one_comp_mm_elim_abs.stan 0.25 0.25 1.01 1.1% faster
pkpd/one_comp_mm_elim_abs.stan 19.41 19.63 0.99 -1.13% slower
garch/garch.stan 0.42 0.41 1.03 2.55% faster
low_dim_gauss_mix/low_dim_gauss_mix.stan 2.68 2.62 1.02 2.42% faster
arK/arK.stan 1.81 1.75 1.03 3.0% faster
gp_pois_regr/gp_pois_regr.stan 2.84 2.74 1.04 3.44% faster
low_dim_gauss_mix_collapse/low_dim_gauss_mix_collapse.stan 8.87 8.47 1.05 4.47% faster
performance.compilation 181.63 185.61 0.98 -2.19% slower
Mean result: 1.017452740630533

Jenkins Console Log
Blue Ocean
Commit hash: 994a6ce2385990c2ed28f9822e090202d7801a42


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

@wds15
Copy link
Contributor

wds15 commented Nov 26, 2024

What about the row wise or column wise summation of matrices? Would that get a different function name?

Just for my understanding: The intention of the function is to add two containers a and b using the log-sum-exp thing and returning again a container, right?

@WardBrian
Copy link
Member

The intention of the function is to add two containers a and b using the log-sum-exp thing and returning again a container, right?

Correct, that is what @spinkney requested at least

@MichaScant
Copy link
Author

Ok my apologies, I think i misinterpreted the original issue, I’ll work it on it tonight and incorporate the log_sum_exp into log_add_exp and ensure the tests reflect the change of only returning containers, should I assume the user is only allowed to pass in containers a and b, or should I account for different data types passed in. Thank you

@WardBrian
Copy link
Member

I think the only requirement should be that the containers passed must be the same size, but if one is a row vector and the other is a column vector I would expect that to be fine, I think?

@MichaScant
Copy link
Author

MichaScant commented Nov 27, 2024

Should I account for multi-dimensional matrix containers or assume that only vectors or 1 dimensional matrix's are accepted into the method regardless if they are column or row-wise?

@WardBrian
Copy link
Member

I think the most useful thing would probably be to work over arbitrary containers of the same size. I think the apply_scalar_binary helpers may automate that

@MichaScant
Copy link
Author

MichaScant commented Nov 27, 2024

I think the most useful thing would probably be to work over arbitrary containers of the same size. I think the apply_scalar_binary helpers may automate that

Been trying to implement this, however, I'm facing an issue where if I'm given 2 different matrix types in which the dimensions don't align, I am not able to pass them into apply_scalar_binary, I have tried a few solutions by transposing the matrix and casting it to ensure its type doesn't change, yet that didn't work as it caused vectors to start failing as a result. Perhaps we should restrict matrix containers if they are not both the same size and have the same containers? Or is there a solution I can attempt, that I haven't tried yet?

@WardBrian
Copy link
Member

I may be misunderstanding, but I think if both arguments are matricies I think it is correct to require them to be the same size. We don’t generally do any “broadcasting” besides maybe allowing a scalar and a container

@MichaScant
Copy link
Author

MichaScant commented Nov 27, 2024

Ok, thank you, in that case, I'll throw an exception if two matrix's passed and their dimensions are not aligned.

@WardBrian
Copy link
Member

I believe check_matching_dims does what we’d want it to do for that purpose

@MichaScant
Copy link
Author

MichaScant commented Nov 28, 2024

Finished implementing the above changes, to clarify though, I understand this method will be used in the row-wise or column-wise summation of matrices. Will this be done outside of the method based on how the user utilizes it (passing in pairs of rows and columns)? At the moment, the code can add two containers a and b using the log-sum-exp method and returns a container of the same type as the parameters passed in. Any matrices that are passed in have element-wise operations performed on them before returning a container with the result of the same type.

@stan-buildbot
Copy link
Contributor


Name Old Result New Result Ratio Performance change( 1 - new / old )
arma/arma.stan 0.33 0.3 1.09 8.23% faster
low_dim_corr_gauss/low_dim_corr_gauss.stan 0.01 0.01 1.06 5.22% faster
gp_regr/gen_gp_data.stan 0.03 0.03 1.04 3.43% faster
gp_regr/gp_regr.stan 0.09 0.09 1.07 6.67% faster
sir/sir.stan 70.32 70.31 1.0 0.01% faster
irt_2pl/irt_2pl.stan 4.21 4.31 0.98 -2.28% slower
eight_schools/eight_schools.stan 0.06 0.06 1.04 4.2% faster
pkpd/sim_one_comp_mm_elim_abs.stan 0.25 0.25 0.99 -0.92% slower
pkpd/one_comp_mm_elim_abs.stan 19.53 19.66 0.99 -0.66% slower
garch/garch.stan 0.43 0.47 0.92 -8.71% slower
low_dim_gauss_mix/low_dim_gauss_mix.stan 2.76 2.59 1.06 6.02% faster
arK/arK.stan 1.87 1.71 1.1 8.81% faster
gp_pois_regr/gp_pois_regr.stan 2.82 2.67 1.05 5.19% faster
low_dim_gauss_mix_collapse/low_dim_gauss_mix_collapse.stan 8.87 8.38 1.06 5.45% faster
performance.compilation 179.97 179.98 1.0 -0.01% slower
Mean result: 1.0300433622507947

Jenkins Console Log
Blue Ocean
Commit hash: 994a6ce2385990c2ed28f9822e090202d7801a42


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
Copy link
Collaborator

Sorry for my delay in looking at this. Bit of a busy week but I will try to look on Friday. Glancing at everything I didn't see anything that was a huge blocker

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

5 participants