Fixing the issue that the recall from the first search step is being cached in streaming evaluation.

harsha-simhadri / big-ann-benchmarks

Framework for evaluating ANNS algorithms on billion scale datasets.

https://big-ann-benchmarks.com

MIT License

313 stars 103 forks source link

Fixing the issue that the recall from the first search step is being cached in streaming evaluation. #280

Closed arron2003 closed 4 months ago

arron2003 commented 4 months ago

Illustrates and fixes the issue that the recall from the first search step is being cached in streaming evaluation.

fixes harsha-simhadri/big-ann-benchmarks#279

harsha-simhadri commented 4 months ago

@arron2003 How does the recall of algorithms on streaming track change with this fix?

arron2003 commented 4 months ago

@arron2003 How does the recall of algorithms on streaming track change with this fix?

I thought it might be easier if you have the original hdf5 files? If you don't have bandwidth then I guess I can create a Azure account to do the re-computation in a few days...

maumueller commented 4 months ago

I am merging this for now. Let's continue the discussion on the influence of this change in #279