cookinai/Blitz-v0.2 Overview

cookinai/Blitz-v0.2 is a 7 billion parameter instruction-tuned language model, developed by cookinai. It is a fine-tuned version of the well-regarded Mistral-7B-Instruct-v0.2 base model.

Key Characteristics

• Base Model: Built upon Mistral-7B-Instruct-v0.2, inheriting its strong performance in various language tasks.
• Training Data: Fine-tuned using cookinai's proprietary Kugelblitz Dataset.
• Training Epochs: The model underwent 3 epochs of training, an increase from previous iterations, suggesting a more thorough adaptation to the fine-tuning data.
• Context Length: Supports a context window of 8192 tokens.

Intended Use Cases

This model is suitable for applications requiring a capable 7B instruction-following model. Its fine-tuning on the Kugelblitz Dataset aims to improve its general utility and responsiveness to instructions, making it a candidate for:

• General text generation
• Instruction-based tasks
• Conversational AI

Users should note that this is an alpha version, with further stability improvements anticipated in future releases like v0.3.