Sheikhaei/llama-3.2-1b-english-persian-translator

Hugging Face
TEXT GENERATIONConcurrency Cost:1Model Size:1BQuant:BF16Ctx Length:32kPublished:Jul 28, 2025License:apache-2.0Architecture:Transformer0.0K Open Weights Warm

Sheikhaei/llama-3.2-1b-english-persian-translator is a 1 billion parameter Llama 3.2 model fine-tuned by Sheikhaei for bidirectional English-Persian translation. This model excels at translating text between English and Persian, supporting both English to Persian and Persian to English directions. It was trained on a custom dataset of approximately 640,000 sentence pairs, achieving BLEU scores of 0.47 (En-Fa) and 0.58 (Fa-En).

Loading preview...

Overview

Sheikhaei/llama-3.2-1b-english-persian-translator is a specialized 1 billion parameter model derived from meta-llama/Llama-3.2-1B. Its primary function is bidirectional translation between English and Persian.

Key Capabilities

  • English to Persian Translation: Translates English text into Persian.
  • Persian to English Translation: Translates Persian text into English.
  • Specific Prompt Format: Requires input in a structured format, e.g., ### English:\nThe children were playing in the park.\n### Persian: or ### Persian:\nکودکان در پارک بازی می‌کردند.\n### English:.

Training and Performance

The model was fine-tuned on a custom dataset comprising around 640,000 English-Persian parallel sentence pairs. This dataset was compiled from Tatoeba and expanded using the Gemma-3-12B model. Evaluation metrics demonstrate strong performance for a model of its size:

  • English → Persian: Achieved a BLEU score of 0.47 and a COMET score of 0.89.
  • Persian → English: Achieved a BLEU score of 0.58 and a COMET score of 0.91.

Good for

  • Developers and applications requiring efficient and accurate English-Persian translation.
  • Integrating translation capabilities into systems where a smaller, specialized model is preferred over larger, general-purpose LLMs.
  • Use cases focused specifically on text translation between these two languages.