B

Bart Base Cnn R2 19.4 D35 Hybrid

Developed by echarlaix
This is a pruned and optimized BART-base model specifically designed for summarization tasks, retaining 53% of the original model's weights.
Downloads 20
Release Time : 3/2/2022

Model Overview

This model is a summarization model fine-tuned from facebook/bart-base, optimized using the nn_pruning library, and excels on the CNN/DailyMail dataset.

Model Features

Efficient Pruning
Utilizes nn_pruning technology to retain only 35% of linear layer weights, preserving 53% of the original model's weights overall.
Attention Head Optimization
Removed 38 out of 216 attention heads (17.6%), enhancing model efficiency.
High-Performance Summarization
Achieves excellent Rouge scores on the CNN/DailyMail dataset.

Model Capabilities

Text Summarization
Long Text Compression
News Content Condensation

Use Cases

News Processing
News Summarization
Automatically generates concise summaries of news articles
Rouge-1:42.18, Rouge-2:19.44, Rouge-L:39.17
Content Analysis
Long Document Summarization
Extracts and summarizes key information from lengthy documents
Featured Recommended AI Models
AIbase
Empowering the Future, Your AI Solution Knowledge Base
© 2025AIbase