botbotrobotics/CabraMistral7b

TEXT GENERATIONConcurrency Cost:1Model Size:7BQuant:FP8Ctx Length:4kPublished:Mar 26, 2024License:ccArchitecture:Transformer0.0K Cold

CabraMistral7b is a 7 billion parameter instruction-tuned language model developed by botbotrobotics, based on the Mistral 7b Instruct 0.2 architecture with Grouped-Query Attention and Sliding-Window Attention. It is specifically fine-tuned for the Portuguese language using the internal Cabra 10k dataset, demonstrating improved performance on various Brazilian benchmarks compared to its base model. This model is primarily intended for research purposes, focusing on generative model investigation and understanding model limitations and biases.

Loading preview...

Cabra Mistral 7b v2: Portuguese-Optimized LLM

Cabra Mistral 7b v2 is a 7 billion parameter language model developed by botbotrobotics, fine-tuned from the Mistral 7b Instruct 0.2 base model. Its core differentiator is its optimization for the Portuguese language, achieved through training on the proprietary Cabra 10k dataset. This specialization results in enhanced performance across various Brazilian benchmarks, making it a strong candidate for Portuguese-centric NLP research.

Key Capabilities

  • Portuguese Language Proficiency: Significantly improved understanding and generation in Portuguese compared to the base Mistral 7b Instruct 0.2.
  • Mistral Architecture: Leverages advanced architectural features like Grouped-Query Attention and Sliding-Window Attention for efficient processing.
  • Quantized Versions Available: Offers various GGUF quantized versions in the "quantanization" branch for flexible deployment.
  • Research-Oriented: Designed for investigating generative models and understanding their limitations and biases.

Good for

  • Portuguese NLP Research: Ideal for academic and research projects focused on natural language processing in Portuguese.
  • Benchmarking: Useful for evaluating and comparing LLM performance on Brazilian-specific tasks and datasets.
  • Exploring Model Biases: Provides a tool for studying biases and limitations within generative models, particularly in a Portuguese context.

Note: This model is currently restricted to non-commercial research use only.

Finetunes
1 models