CorticalStack/mistral-7b-alpaca-sft Overview

This model is a 7 billion parameter instruction-tuned variant of the Mistral architecture, specifically fine-tuned from unsloth/mistral-7b-bnb-4bit. The fine-tuning process utilized the yahma/alpaca-cleaned dataset, which is known for its diverse instruction-following examples.

Key Characteristics

• Base Model: Mistral-7B, known for its strong performance in its size class.
• Fine-tuning: Supervised Fine-Tuning (SFT) on the Alpaca-cleaned dataset, enhancing its ability to follow instructions and engage in conversational tasks.
• Efficiency: Trained with Unsloth and Huggingface's TRL library, indicating an emphasis on efficient training and potentially efficient inference.
• Training Configuration: Employed LoRA with specific parameters (r: 256, alpha: 128, dropout: 0.0) and 4-bit BNB quantization during training, optimizing for resource usage.

Ideal Use Cases

• Instruction Following: Excels at responding to prompts and executing specific instructions due to its Alpaca-based fine-tuning.
• General Conversational AI: Suitable for chatbots and interactive applications where clear and coherent responses are needed.
• Resource-Constrained Environments: The 7B parameter count and 4-bit quantization make it a good candidate for deployment where computational resources are a consideration.