Goto

Collaborating Authors

 return jax


A Appendix

Neural Information Processing Systems

Listing 1 demonstrates a maximum inner product search (MIPS) kernel implemented with Jax. When set to True emits the ExactRescoring kernel. Listing 2 is the Jax implementation of Euclidean space nearest neighbor search. We made a few adjustments to speed up the computation. The temporal locality refers to reusing previously accessed items.