WikiHow-Mistral-Instruct-7B Overview

ajibawa-2023's WikiHow-Mistral-Instruct-7B is a 7 billion parameter language model, fine-tuned from the Mistral-7B-Instruct-v0.2 base model. Its primary distinction lies in its specialized training on a comprehensive WikiHow dataset, enabling it to generate detailed and instructional tutorials.

Key Capabilities

• Tutorial Generation: Excels at creating long, in-depth tutorials in the style of WikiHow, providing step-by-step explanations and practical tips.
• Informative Content: Designed to produce highly informative and instructional text across diverse subjects.
• Contextual Understanding: Leverages its training on a vast repository of practical knowledge to offer insightful guidance.

Training Details

The model was fully fine-tuned for 3 epochs on 4 x A100 80GB GPUs, utilizing the Axolotl codebase. The training process took over 15 hours. Quantized versions (GGUF and Exllama v2) are also available, thanks to Bartowski.

Performance Benchmarks

Evaluated on the Open LLM Leaderboard, the model achieved an average score of 61.25. Notable scores include:

• AI2 Reasoning Challenge (25-Shot): 60.92
• HellaSwag (10-Shot): 80.99
• MMLU (5-Shot): 58.57
• TruthfulQA (0-shot): 62.16

Recommended Use Cases

This model is particularly well-suited for applications requiring the generation of instructional content, how-to guides, and detailed explanations for various tasks and skills. Users should employ the specified ChatML prompt format for optimal results.