metadata
license: apache-2.0
language:
- en
tags:
- text-generation
- text2text-generation
- summarization
pipeline_tag: text2text-generation
widget:
- text: >-
Summarize: You may want to stick it to your boss and leave your job, but
don't do it if these are your reasons.
example_title: Example1
- text: >-
Summarize: Jorge Alfaro drove in two runs, Aaron Nola pitched seven
innings of two-hit ball and the Philadelphia Phillies beat the Los Angeles
Dodgers 2-1 Thursday, spoiling Clayton Kershaw's first start in almost a
month. Hitting out of the No. 8 spot in the ...
example_title: Example2
model-index:
- name: RUCAIBox/mtl-summarization
results:
- task:
type: summarization
name: Summarization
dataset:
name: samsum
type: samsum
config: samsum
split: test
metrics:
- name: ROUGE-1
type: rouge
value: 27.3604
verified: true
- name: ROUGE-2
type: rouge
value: 7.2277
verified: true
- name: ROUGE-L
type: rouge
value: 22.4688
verified: true
- name: ROUGE-LSUM
type: rouge
value: 24.6476
verified: true
- name: loss
type: loss
value: 2.2192304134368896
verified: true
- name: gen_len
type: gen_len
value: 17.2418
verified: true
MTL-summarization
The MTL-summarization model was proposed in MVP: Multi-task Supervised Pre-training for Natural Language Generation by Tianyi Tang, Junyi Li, Wayne Xin Zhao and Ji-Rong Wen.
The detailed information and instructions can be found https://github.com/RUCAIBox/MVP.
Model Description
MTL-summarization is supervised pre-trained using a mixture of labeled summarization datasets. It is a variant (Single) of our main MVP model. It follows a standard Transformer encoder-decoder architecture.
MTL-summarization is specially designed for summarization tasks, such as new summarization (CNN/DailyMail, XSum) and dialog summarization (SAMSum).
Example
>>> from transformers import MvpTokenizer, MvpForConditionalGeneration
>>> tokenizer = MvpTokenizer.from_pretrained("RUCAIBox/mvp")
>>> model = MvpForConditionalGeneration.from_pretrained("RUCAIBox/mtl-summarization")
>>> inputs = tokenizer(
... "Summarize: You may want to stick it to your boss and leave your job, but don't do it if these are your reasons.",
... return_tensors="pt",
... )
>>> generated_ids = model.generate(**inputs)
>>> tokenizer.batch_decode(generated_ids, skip_special_tokens=True)
["Don't do it if these are your reasons"]
Related Models
MVP: https://huggingface.co/RUCAIBox/mvp.
Prompt-based models:
- MVP-multi-task: https://huggingface.co/RUCAIBox/mvp-multi-task.
- MVP-summarization: https://huggingface.co/RUCAIBox/mvp-summarization.
- MVP-open-dialog: https://huggingface.co/RUCAIBox/mvp-open-dialog.
- MVP-data-to-text: https://huggingface.co/RUCAIBox/mvp-data-to-text.
- MVP-story: https://huggingface.co/RUCAIBox/mvp-story.
- MVP-question-answering: https://huggingface.co/RUCAIBox/mvp-question-answering.
- MVP-question-generation: https://huggingface.co/RUCAIBox/mvp-question-generation.
- MVP-task-dialog: https://huggingface.co/RUCAIBox/mvp-task-dialog.
Multi-task models:
- MTL-summarization: https://huggingface.co/RUCAIBox/mtl-summarization.
- MTL-open-dialog: https://huggingface.co/RUCAIBox/mtl-open-dialog.
- MTL-data-to-text: https://huggingface.co/RUCAIBox/mtl-data-to-text.
- MTL-story: https://huggingface.co/RUCAIBox/mtl-story.
- MTL-question-answering: https://huggingface.co/RUCAIBox/mtl-question-answering.
- MTL-question-generation: https://huggingface.co/RUCAIBox/mtl-question-generation.
- MTL-task-dialog: https://huggingface.co/RUCAIBox/mtl-task-dialog.
Citation
@article{tang2022mvp,
title={MVP: Multi-task Supervised Pre-training for Natural Language Generation},
author={Tang, Tianyi and Li, Junyi and Zhao, Wayne Xin and Wen, Ji-Rong},
journal={arXiv preprint arXiv:2206.12131},
year={2022},
url={https://arxiv.org/abs/2206.12131},
}