search

csebuetnlp / xl-sum

This repository contains the code, data, and models of the paper titled "XL-Sum: Large-Scale Multilingual Abstractive Summarization for 44 Languages" published in Findings of the Association for Computational Linguistics: ACL-IJCNLP 2021.
https://aclanthology.org/2021.findings-acl.413/
256 stars 41 forks source link

mt5 small generating wrong predictions #8

Closed ashokurlana closed 2 years ago

ashokurlana commented 2 years ago

I am trying to finetune the mt5-small with Telugu corpus, all the generated summaries includes tokens, please suggest how to fixt it.

Example generated output:

హోమియోపతి కళాశాలను న్యూఢిల్లీ సెంట్రల్ కౌన్సిల్ ఆఫ్ ప్రత్యేక బృందం బుధవారం పరిశీలించింది. How to avoid this token in the summary? see this [issue](https://github.com/huggingface/transformers/issues/9792) for better understanding
abhik1505040 commented 2 years ago

Hi, are you using the scripts provided in our repository for fine-tuning and inference? If not, please provide a reference to the code.

ashokurlana commented 2 years ago

Yes, I am using the scripts provided in this repository.

Tahmid04 commented 2 years ago

Hi, this issue is pertinent to the small variant of the mT5 model, not because of any errors in our code. mT5-base or other multilingual models does not exhibit the generation of any special token.