README.md

metadata

datasets:
  - FIdo-AI/ua-news
language:
  - uk
metrics:
  - rouge
library_name: transformers
pipeline_tag: summarization
tags:
  - news

Model Card for Model ID

Model Summary

The mT5-multilingual-XLSum model was fine-tuned on the UA-News dataset to generate concise and accurate news headlines in Ukrainian language.

Model Details

• Model Type: Summarization
• Language: Multilingual/Ukrainian
• Library: Transformers

Dataset

• Name: UA-News
• Source: FIdo-AI
• Description: The UA-News dataset contains a diverse collection of Ukrainian news articles, covering various topics including politics, economics, culture, and sports. The dataset is curated to provide high-quality training data for summarization tasks.

Training

• Fine-Tuning: The model was fine-tuned on the UA-News dataset using the Hugging Face Transformers library.
• Epochs: 4
• Batch Size: 4
• Learning Rate: 4e-5

Evaluation

• Metrics: The model's performance was evaluated using the ROUGE metric.
  • ROUGE-1: 0.2452
  • ROUGE-2: 0.1075
  • ROUGE-L: 0.2348
  • BERTScore: 0.7573

Usage

• Pipeline Tag: Summarization
• How to Use: The model can be used with the Hugging Face pipeline for summarization. Here's an example:

  from transformers import pipeline
  
  summarizer = pipeline("summarization", model="yelyah/mT5-XLSUM-ua-news ")
  article = "Your news article text here."
  summary = summarizer(article)
  print(summary)