stsb-xlm-r-multilingual-ro
| Property | Value |
|---|---|
| Parameter Count | 278M |
| Model Type | Sentence Transformer |
| Architecture | XLM-RoBERTa with Mean Pooling |
| Output Dimensions | 768 |
| Primary Language | Romanian |
What is stsb-xlm-r-multilingual-ro?
stsb-xlm-r-multilingual-ro is a specialized sentence transformer model designed for Romanian language processing. It's a fine-tuned version of XLM-RoBERTa that maps sentences and paragraphs to 768-dimensional dense vector space, optimized for semantic similarity tasks.
Implementation Details
The model utilizes a two-component architecture: an XLM-RoBERTa transformer followed by a mean pooling layer. It was trained using CosineSimilarityLoss with AdamW optimizer, featuring a learning rate of 2e-05 and 10 epochs of training on the Romanian STS dataset.
- Maximum sequence length: 128 tokens
- Training includes warmup steps: 100
- Implements weight decay: 0.01
- Uses batch size: 32
Core Capabilities
- Sentence and paragraph embedding generation
- Semantic similarity comparison
- Clustering applications
- Cross-lingual text understanding with Romanian optimization
Frequently Asked Questions
Q: What makes this model unique?
This model is specifically optimized for Romanian language processing while maintaining multilingual capabilities, making it particularly effective for Romanian text similarity tasks and semantic search applications.
Q: What are the recommended use cases?
The model is ideal for semantic similarity tasks, document clustering, and information retrieval in Romanian language contexts. It can be effectively used for both monolingual Romanian applications and cross-lingual scenarios.
