R

Rubert Tiny2

Developed by cointegrated
A compact BERT-based Russian encoder capable of generating high-quality sentence embeddings
Downloads 585.48k
Release Time : 3/2/2022

Model Overview

This is an upgraded version of rubert-tiny, specialized for Russian language processing, suitable for generating sentence embeddings or fine-tuning for downstream tasks.

Model Features

Expanded vocabulary
Vocabulary increased from 29,564 to 83,828 tokens, enhancing model expressiveness
Long sequence support
Maximum supported sequence length extended from 512 to 2048
High-quality sentence embeddings
Sentence embeddings closer to LaBSE performance
Optimized segment embeddings
Tuned for NLI tasks with meaningful segment embeddings
Specialized for Russian
The model is specifically optimized for Russian language processing

Model Capabilities

Generate sentence embeddings
Short text classification
Sentence similarity calculation
Masked language modeling

Use Cases

Text processing
Short text classification
Classify short texts using methods like KNN
Semantic search
Perform semantic similarity searches based on sentence embeddings
Featured Recommended AI Models
AIbase
Empowering the Future, Your AI Solution Knowledge Base
© 2025AIbase