rut5-base-absum
| Property | Value |
|---|---|
| Parameter Count | 244M |
| License | MIT |
| Framework | PyTorch |
| Training Datasets | IlyaGusev/gazeta, xlsum, mlsum, wiki_lingua |
What is rut5-base-absum?
rut5-base-absum is a specialized Russian language model designed for abstractive text summarization. Built upon the rut5-base-multitask architecture, this model has been fine-tuned on four diverse datasets to generate concise, meaningful summaries of Russian text. The model employs advanced transformer architecture with 244M parameters, utilizing the T5 framework optimized for text-to-text generation tasks.
Implementation Details
The model is implemented using PyTorch and the Transformers library, featuring a flexible API that allows users to control the summarization process through parameters such as word count and compression ratio. It supports both deterministic (beam search) and stochastic generation approaches.
- Supports custom summary length control through word count or compression ratio
- Uses beam search with configurable number of beams (default: 3)
- Implements repetition penalty for better summary quality
- Handles long-form text input with efficient processing
Core Capabilities
- Abstractive summarization of Russian text
- Controllable summary length through explicit parameters
- Support for both fixed word count and compression ratio-based summarization
- Efficient processing with F32 tensor type support
- Integration with popular ML frameworks and inference endpoints
Frequently Asked Questions
Q: What makes this model unique?
This model stands out for its specialized focus on Russian language summarization and its flexibility in controlling output length. The combination of four training datasets provides robust performance across different types of Russian text.
Q: What are the recommended use cases?
The model is ideal for automated news summarization, content condensation, and document abstracting in Russian. It's particularly useful for applications requiring controlled-length summaries of longer texts.
