Kaival Parikh

Ran some luceneutil benchmarks on Cohere vectors, 768d for various vector similarities x quantization bits: #### `dot_product` `main` ``` recall latency(ms) netCPU avgCpuCount nDoc topK fanout maxConn beamWidth quantized visited...

> It's hard for human eyes + brain to scan all those numbers and confirm there's no real difference Haha true :) I fed the raw data to an LLM...

@uschindler just wanted to ask, did you get a chance to look at these changes? Thanks!

Thanks @mikemccand, IIUC Uwe's main concern was to _not_ expose incubating / experimental APIs (i.e. the Vector API) publicly (which was true in this PR) -- I'm not sure what...

### File Layout Today, the `.vec` file is partitioned per-field, and looks like: ``` # field 1 begin: (vector for field 1, document d1) # position x0 (vector for field...

### Notes - Right now this is a crude implementation, rough and inefficient, only for demonstration purposes! - Basically a copy of `Lucene99FlatVectors*`, except that raw vectors are de-duped and...

+1 to this issue. Also, I wanted to ask about the motivation for this change, is it because the `pytorch` / `nvidia` channels have a restrictive license?