lance-format / lance-format/lance
feat: improve GPU memory usage to train large IVF
Open
@eddyxu is already working on this.
Since Nov 13, 2023.
benchmark
enhancement
PyTorch
- Dominant language
- Rust
- Stars
- 7.1k
- Forks
- 852
- Avg merge
- 3d 18h
- Merged PRs (30d)
- 272
Description
Problem Statement
Improve IVF usage to train large k-means on GPU. For example, Lance usually requires sample 256 x #(IVF) of vectors to train kmeans. For a large IVF like IVF65536, it will result about 100GB just for training dataset.
We don't really need to keep all the data in memory.
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Assessment
This issue has not been assessed yet.