C. George Williams
C. George Williams
Are you referring to our existing formula for measuring throughput? (latency to search batch of queries) / ( number of queries in a batch ) On Mon, Aug 23, 2021...
@anaruse. OK, we synced over the new config and re-ran the evaluation. Please verify the following: https://github.com/harsha-simhadri/big-ann-benchmarks/blob/gw/T3/t3/cuanns_ivfpq/algos.yaml https://github.com/harsha-simhadri/big-ann-benchmarks/blob/gw/T3/t3/eval_2021/cuanns_ivfpq/EvalPublic.ipynb
@anaruse OK, we synced your new algos yaml and reran. It looks like a slight improvement to text2image-1B. Please verify the config and results notebook below: https://github.com/harsha-simhadri/big-ann-benchmarks/blob/gw/T3/t3/cuanns_ivfpq/algos.yaml https://github.com/harsha-simhadri/big-ann-benchmarks/blob/gw/T3/t3/eval_2021/cuanns_ivfpq/EvalPublic.ipynb
@anaruse Great.
@anaruse OK, we re-evaluated your submission. The "anomaly" count was reduced successfully from 7 to 5. Thanks. Can you verify the following? new config we synced: https://github.com/harsha-simhadri/big-ann-benchmarks/blob/gw/T3/t3/cuanns_ivfpq/algos.yaml eval notebook -...
@yong-wang Great. We sync'ed the latest config file and ran the eval. Please verify the config we used and the analysis: https://github.com/harsha-simhadri/big-ann-benchmarks/blob/gw/T3/t3/cuanns_multigpu/algos.yaml https://github.com/harsha-simhadri/big-ann-benchmarks/blob/gw/T3/t3/eval_2021/cuanns_multigpu/EvalPublic.ipynb
@mdouze @ingberam I went ahead and prototyped the new/updated candidate virtual methods in BaseANN. See the Files Changed tab for the proposals.
> @sourcesync Thanks so much for going ahead with this! As I understood, there is no hybrid query, but only a dense _query_, _add_, _remove_ (possible with metadata) and a...
@Martin Aumüller ***@***.***> I agree unless Mark sees a way to "resolve the merge conflicts" as he offered. But even then, I'm not sure how useful that would be. I...
@Martin Aumüller ***@***.***> Thanks for the diff. I don't see anything in there worth keeping for version 2. I might recommend we keep or archive the branch and just close...