lucianosb/boto-9B
lucianosb/boto-9B is a 9 billion parameter language model fine-tuned from Gemma2-9B specifically for the Portuguese language, developed by lucianosb. This model is optimized for generating verbose responses in Portuguese, leveraging a 16384-token context length. It is particularly suited for applications requiring extensive text generation and understanding in Portuguese, as evidenced by its performance on the Open Portuguese LLM Leaderboard.
Loading preview...
Overview
lucianosb/boto-9B is a 9 billion parameter language model, fine-tuned from Google's Gemma2-9B architecture by lucianosb. Its primary focus is on the Portuguese language, utilizing the cetacean-ptbr dataset for training. The model is noted for its tendency to produce verbose and often lengthy responses, making it suitable for applications where detailed textual output is desired.
Key Capabilities & Features
- Portuguese Language Specialization: Fine-tuned specifically for generating and understanding text in Portuguese.
- Verbose Output: Designed to provide comprehensive and detailed responses.
- Gemma2-9B Base: Built upon the robust Gemma2-9B foundation, enhanced with a 16384-token context length.
- Efficient Training: Trained using Unsloth and Huggingface's TRL library, enabling faster fine-tuning.
Performance Highlights
Evaluated on the Open Portuguese LLM Leaderboard, boto-9B achieved an average score of 68.45. Notable scores include:
- ENEM Challenge (No Images): 75.02
- Assin2 RTE: 89.38
- Assin2 STS: 76.59
Considerations
Users should be aware that the model does not include content moderation mechanisms and may inadvertently reproduce social stereotypes or generate content inconsistent with reality due to biases in its training data. It is advised not to rely solely on the model for critical decisions.