rubert-tiny-turbo
| Property | Value |
|---|---|
| Model Size | 111MB |
| Embedding Dimension | 312 |
| Context Length | 2048 tokens |
| CPU Inference Time | 5.51ms |
| GPU Inference Time | 3.25ms |
| Base Model | rubert-tiny2 |
What is rubert-tiny-turbo?
rubert-tiny-turbo is a highly optimized Russian language model designed for generating sentence embeddings. Based on the architecture of rubert-tiny2, this model delivers impressive performance while maintaining a compact size of just 111MB. It produces 312-dimensional embeddings and can handle sequences up to 2048 tokens in length.
Implementation Details
The model excels in both speed and accuracy metrics. On CPU, it processes sentences in just 5.51ms, while GPU processing time is even faster at 3.25ms. According to the encodechka benchmark, it achieves a mean score of 0.749 for semantic tasks, making it competitive with much larger models.
- Optimized for Russian language understanding
- Excellent balance between size and performance
- Easy integration with SentenceTransformers library
- Significant improvement over base rubert-tiny2 model
Core Capabilities
- Semantic Textual Similarity (STS) score: 0.828
- Strong performance in classification tasks
- Efficient document retrieval capabilities
- Competitive results on ruMTEB benchmark
Frequently Asked Questions
Q: What makes this model unique?
The model stands out for its exceptional speed-to-performance ratio, being 20-100x faster than larger models while maintaining competitive accuracy. It's particularly useful for applications requiring real-time processing of Russian text.
Q: What are the recommended use cases?
The model is ideal for semantic search, document classification, clustering, and similarity comparison tasks in Russian language applications where processing speed is crucial. It's particularly effective for systems with limited computational resources.
