Tooka S B E R T

Introduction

Tooka-SBERT is a Sentence Transformer model optimized for Persian language tasks such as semantic textual similarity, semantic search, paraphrase mining, text classification, and clustering. It is based on the TookaBERT-Large model and trained to map sentences and paragraphs into a dense vector space.

Architecture

• Model Type: Sentence Transformer
• Base Model: TookaBERT-Large
• Maximum Sequence Length: 512 tokens
• Output Dimensionality: 1024 tokens
• Similarity Function: Cosine Similarity
• Language: Persian

Training

The Tooka-SBERT model utilizes a Siamese network structure to compute sentence embeddings and is trained using the CachedMultipleNegativesRankingLoss function. It is designed to efficiently capture semantic relationships in text, especially in the Persian language.

Guide: Running Locally

To run the Tooka-SBERT model locally, follow these steps:

1. Install Sentence Transformers:

   pip install -U sentence-transformers
   

2. Load the Model and Run Inference:

   from sentence_transformers import SentenceTransformer
   
   # Load the model from Hugging Face Hub
   model = SentenceTransformer("PartAI/Tooka-SBERT")
   
   # Define sentences for encoding
   sentences = [
       'درنا از پرندگان مهاجر با پاهای بلند و گردن دراز است.',
       'درناها با قامتی بلند و بال‌های پهن، از زیباترین پرندگان مهاجر به شمار می‌روند.',
       'درناها پرندگانی کوچک با پاهای کوتاه هستند که مهاجرت نمی‌کنند.'
   ]
   
   # Encode sentences
   embeddings = model.encode(sentences)
   print(embeddings.shape)
   
   # Compute similarity scores
   similarities = model.similarity(embeddings, embeddings)
   print(similarities.shape)
   

3. Suggestions for Cloud GPUs:

   • Use cloud platforms like AWS, Google Cloud, or Azure that provide GPU resources to enhance the performance of model inference.

License

The Tooka-SBERT model is released under the Apache 2.0 License. This permits use, distribution, and modification under the terms specified in the license.