[NeurIPS 2023 Filter Track] faissplus

[PUITAR]

Our submission of algorithm 1: faissplus.

Using some parameters to optimize faiss.

[maumueller]

@PUITAR This seems to have the same problems as mentioned in #223 . Please update your pull request to avoid changing the existing code base.

[maumueller]

I'll try again, but please also fix the other issues where you remove files from our codebase.

[maumueller]

It still doesn't work: https://github.com/harsha-simhadri/big-ann-benchmarks/actions/runs/6706568601/job/18223474904?pr=222#step:4:1858

[PUITAR]

It still doesn't work: https://github.com/harsha-simhadri/big-ann-benchmarks/actions/runs/6706568601/job/18223474904?pr=222#step:4:1858

I‘m sorry we have so many problems with our code. I have just add a file to our faissplus, so run the action of ’install requirement‘ again may works.

[maumueller]

@PUITAR I don't see a new commit after the failing CI run that I am pointing to.

[PUITAR]

@maumueller I just realized some mistake. Since you might used the same docker to run the CI test, it wrong again. you should see this, in the INTALL REQUIREMENT action of the CI test, there is an error

------
Dockerfile:15
--------------------
  13 |     RUN python3 -m pip install -r requirements_conda.txt 
  14 |     
  15 | >>> COPY neurips23/filter/faiss/bow_id_selector.swig ./
  16 |     
  17 |     RUN swig -c++ -python -I$CONDA_PREFIX/include -Ifaiss bow_id_selector.swig 
--------------------
ERROR: failed to solve: failed to compute cache key: failed to calculate checksum of ref 5de61ec8-c85c-4b78-898d-5c657cc86999::96cbz8xkkinfbz3umqxm8tzf3: "/neurips23/filter/faiss/bow_id_selector.swig": not found

This means the file(bow_id_selector.swig) lost, and i had add the file to my PR. So please run the intall requirement again. Thank you very much. orz

[maumueller]

@PUITAR This looks much better.

I got the following results:

faissplus,"Faiss(('IVF11264,SQ8', {'nprobe': 30, 'mt_threshold': 0.00033}))",yfcc-10M,10,3840.5696040105754,0.0,573.4747545719147,4905644.0,0,0,filter,0.895063
faissplus,"Faiss(('IVF11264,SQ8', {'nprobe': 32, 'mt_threshold': 0.0003}))",yfcc-10M,10,3764.112198957016,0.0,573.4747545719147,4905644.0,0,0,filter,0.8968999999999999
faissplus,"Faiss(('IVF11264,SQ8', {'nprobe': 32, 'mt_threshold': 0.00031}))",yfcc-10M,10,3756.703088341436,0.0,573.4747545719147,4905644.0,0,0,filter,0.897869
faissplus,"Faiss(('IVF11264,SQ8', {'nprobe': 32, 'mt_threshold': 0.00033}))",yfcc-10M,10,3737.604998681685,0.0,573.4747545719147,4905644.0,0,0,filter,0.8996660000000001
faissplus,"Faiss(('IVF11264,SQ8', {'nprobe': 34, 'mt_threshold': 0.0003}))",yfcc-10M,10,3673.341554167496,0.0,573.4747545719147,4905644.0,0,0,filter,0.901073
faissplus,"Faiss(('IVF11264,SQ8', {'nprobe': 32, 'mt_threshold': 0.00035}))",yfcc-10M,10,3735.4637292410794,0.0,573.4747545719147,4905644.0,0,0,filter,0.901529
faissplus,"Faiss(('IVF11264,SQ8', {'nprobe': 34, 'mt_threshold': 0.00031}))",yfcc-10M,10,3652.6801971503146,0.0,573.4747545719147,4905644.0,0,0,filter,0.902008
faissplus,"Faiss(('IVF11264,SQ8', {'nprobe': 34, 'mt_threshold': 0.00033}))",yfcc-10M,10,3655.287962959017,0.0,573.4747545719147,4905644.0,0,0,filter,0.903754
faissplus,"Faiss(('IVF11264,SQ8', {'nprobe': 34, 'mt_threshold': 0.00035}))",yfcc-10M,10,3635.117445737324,0.0,573.4747545719147,4905644.0,0,0,filter,0.9055630000000001
faissplus,"Faiss(('IVF11264,SQ8', {'nprobe': 40, 'mt_threshold': 0.0003}))",yfcc-10M,10,3453.915576495494,0.0,573.4747545719147,4905644.0,0,0,filter,0.9119970000000001

Is this ready to be merged?