Open mjeensung opened 3 months ago
cc @ArthurZucker
I am a beginner, I will try my best to solve this issue, If possible, assign me to it. I will try my very best to read the research paper mentioned, and add model description for XTR
Hi @Souradip121, great that you want to contribute! We don't assign issues to people, as we find that this can lead to issues becoming stale, even if they're highly requested. Instead, we will prioritise based on PRs being open. Adding a model can be a big task - great if you want to tackle straight away, but if you'd like to start on something smaller, issues tagged with "Good First Issue" are a great place to start
@amyeroberts Okay I will focus on Good First Issues first. Thanks for the suggestion!!!
Model description
XTR (ConteXtualized Token Retriever) is a multi-vector retrieval model that improves efficiency by focusing on retrieving and ranking the most important document tokens. Details are described at https://arxiv.org/abs/2304.01982.
Open source status
Provide useful links for the implementation
Implementation for PyTorch: https://github.com/mjeensung/xtr-pytorch Weight: https://huggingface.co/google/xtr-base-en Author: @jhyuklee