Skip to content

Latest commit

 

History

History
112 lines (80 loc) · 2.66 KB

README.md

File metadata and controls

112 lines (80 loc) · 2.66 KB

Benchmark

How to read the plots

  • the plot format follows the style of ann-benchmarks
  • X axis is recall@k, Y axis is queries/second
  • for the same recall@k, closer to topside edge is better
  • for the same queries/second, closer to the right side edge is better
  • each node on a curve represents a different n_probe, starting from 1 at the left most node, and multiplied by 2 at every next node. (n_probe = 1,2,4,8,16,32,64,128)

Dataset: Gist1M

  • number of training vectors: 500,000
  • number of base vectors: 1,000,000
  • number of query vectors: 1,000
  • vector dimentions: 960
  • runtime is average of 30 runs

Test device: 1 x Tesla T4 (Turing)

k = 1

Click here to see details:
IVF4096 PQ8
IVF4096 PQ16
IVF4096 PQ32
IVF4096 PQ64
IVF16384 PQ8
IVF16384 PQ16
IVF16384 PQ32
IVF16384 PQ64

k = 10

Click here to see details:
IVF4096 PQ8
IVF4096 PQ16
IVF4096 PQ32
IVF4096 PQ64
IVF16384 PQ8
IVF16384 PQ16
IVF16384 PQ32
IVF16384 PQ64

k = 100

Click here to see details:
IVF4096 PQ8
IVF4096 PQ16
IVF4096 PQ32
IVF4096 PQ64
IVF16384 PQ8
IVF16384 PQ16
IVF16384 PQ32
IVF16384 PQ64