mT5-SEAHORSE models - Githubissues

google-research-datasets / seahorse

Seahorse is a dataset for multilingual, multi-faceted summarization evaluation. It consists of 96K summaries with human ratings along 6 quality dimensions: comprehensibility, repetition, grammar, attribution, main idea(s), and conciseness, covering 6 languages, 9 systems and 4 datasets.

85 stars 11 forks source link

mT5-SEAHORSE models #1

Open dlvp opened 1 year ago

dlvp commented 1 year ago

Congratulations for the great work! Is there a plan to release the 6 mT5_SEAHORSE models?

RibinMTC commented 1 year ago

Any updates on releasing the mt5_SEAHORSE models?

BiliBraker commented 1 year ago

ankparikh commented 1 year ago

Hi, sorry for the delay. We have released the models here: https://huggingface.co/google

The model names are (for Large and XXL sized models respectively): google/seahorse-large-q1 google/seahorse-large-q2 google/seahorse-large-q3 google/seahorse-large-q4 google/seahorse-large-q5 google/seahorse-large-q6

google/seahorse-xxl-q1 google/seahorse-xxl-q2 google/seahorse-xxl-q3 google/seahorse-xxl-q4 google/seahorse-xxl-q5 google/seahorse-xxl-q6