Powerful unsupervised domain adaptation method for dense retrieval. Requires only unlabeled corpus and yields massive improvement: "GPL: Generative Pseudo Labeling for Unsupervised Domain Adaptation of Dense Retrieval" https://arxiv.org/abs/2112.07577
Apache License 2.0
321
stars
37
forks
source link
Evaluation data and generation code is not released ? #41
First of all thank you for sharing this incredible work ! It truly is amazing that you've shared your code, some models weights and some generated data.
For my end-of-word studies I'd like to adapt this work with the LoRa technique. I came upon realization that the evaluation data is not shared (test.tsv) and the creation of the test.tsv file in unavailable in the repo, is there a reason for this ?
Hi,
First of all thank you for sharing this incredible work ! It truly is amazing that you've shared your code, some models weights and some generated data.
For my end-of-word studies I'd like to adapt this work with the LoRa technique. I came upon realization that the evaluation data is not shared (test.tsv) and the creation of the test.tsv file in unavailable in the repo, is there a reason for this ?