Hugging Face Overview
LlaMaestra-3.2-1B-Translation Overview
LlaMaestra-3.2-1B-Translation is a compact 1 billion parameter language model, fine-tuned by Leonard Püttmann, designed for efficient text translation between English and Italian. Based on the meta-llama/Llama-3.2-1B-Instruct architecture, this model was trained on approximately 300,000 translation examples.
Key Capabilities
- Bidirectional Translation: Proficient in translating text from English to Italian and Italian to English.
- Direct Translation Output: Engineered to provide straightforward translations, minimizing additional explanatory text.
- Resource-Efficient: Due to its small parameter count, the model performs well on CPUs, making it accessible for various deployment scenarios.
Training Details
The model underwent about 10 hours of fine-tuning on an A10G Nvidia GPU. The training data consisted of sentence pairs sourced from tatoeba.com, ensuring a broad base for translation tasks.
Ideal Use Cases
- Applications requiring quick and direct English-Italian or Italian-English text translation.
- Environments with limited computational resources where larger models are impractical.
- Integration into systems where concise translation output is preferred over verbose explanations.