tevatron-elastic-bert-retriever-mrl

A retriever trained with Tevatron-Elastic, which trains one checkpoint to serve many operating points along the depth / width / token compression axes. This checkpoint is an elastic width axis (Matryoshka / MRL): one checkpoint serves several embedding dimensions.

Base model: google-bert/bert-base-uncased
Task: retriever
Elastic axis: mrl
Training data: rlhn/rlhn-680K, max length 512, query:/passage: prefixes.

Full-point BEIR-15 nDCG@10: 0.445.

Load with the Tevatron-Elastic framework and select an operating point with prune_to / encode_at; see the repository for usage. Part of a release of 20 checkpoints (3 backbones, retrieval and reranking, all compression axes) accompanying the Tevatron-Elastic paper. Reported as a reproducibility resource, not a state-of-the-art claim.

Downloads last month: 6

Safetensors

Model size

0.1B params

Tensor type

BF16

Model tree for utahnlp/tevatron-elastic-bert-retriever-mrl

Base model

google-bert/bert-base-uncased

Finetuned

(6817)

this model

Collection including utahnlp/tevatron-elastic-bert-retriever-mrl

Tevatron-Elastic

Collection

checkpoints from Tevatron-Elastic framework • 20 items • Updated 5 days ago