Closed BinhMinhs10 closed 2 years ago
Hi @BinhMinhs10, can you share the scripts to reproduce this?
https://github.com/huggingface/transformers/blob/main/examples/pytorch/summarization/run_summarization.py Mình dùng script này bỏ mỗi đoạn nltk chỗ hàm postprocess_text
Hi @BinhMinhs10 , sorry for the late reply. We have just updated our eval scripts here https://github.com/vietai/ViT5/blob/main/eval/Eval_vietnews_sum.ipynb
Please note that we fine-tuned the task with a vietnews:
prefix. We are working on an updated version without this prefix. For now you need to prepend a vietnews:
prefix in the input sequence.
Great, thanks to @justinphan3110 for pointing this out and suggesting a solution 🚀
mình thử evaluate trên dataset vietnew với code run_summarization của huggingface (đã set sourse_prefix "vietnews: " ) nhưng không hiểu sao rouge2, rougeL,.. rất thấp