The habanoz/tinyllama-oasst1-top1-instruct-full-lr1-5-v0.1 is a 1.1 billion parameter instruction-tuned language model, fine-tuned from TinyLlama-1.1B-intermediate-step-715k-1.5T. Developed by habanoz, this model leverages the OpenAssistant/oasst_top1_2023-08-25 dataset for instruction following. It is designed for general conversational AI tasks, offering a compact solution for applications requiring a smaller footprint.
Model Overview
The habanoz/tinyllama-oasst1-top1-instruct-full-lr1-5-v0.1 is a 1.1 billion parameter instruction-tuned language model. It is built upon the TinyLlama-1.1B-intermediate-step-715k-1.5T base model and has been fine-tuned using the OpenAssistant/oasst_top1_2023-08-25 dataset, which focuses on high-quality, human-annotated conversational data.
Key Capabilities
- Instruction Following: The model is specifically fine-tuned to understand and respond to user instructions, making it suitable for chatbot-like interactions.
- Compact Size: With 1.1 billion parameters, it offers a relatively small footprint, which can be beneficial for deployment in resource-constrained environments or for faster inference.
- General Conversational AI: Its training on the OpenAssistant dataset suggests a capability for engaging in general-purpose conversations and answering a variety of prompts.
Performance Metrics
Evaluation on the Open LLM Leaderboard shows the following average scores:
- Avg. Score: 35.58
- AI2 Reasoning Challenge (25-Shot): 32.85
- HellaSwag (10-Shot): 58.16
- MMLU (5-Shot): 25.96
- TruthfulQA (0-shot): 38.35
- Winogrande (5-shot): 57.70
- GSM8k (5-shot): 0.45
These scores indicate its performance across various reasoning, common sense, and knowledge-based tasks, positioning it as a capable model within its size class.
Good For
- Lightweight Chatbots: Ideal for applications where a smaller model size is preferred without sacrificing basic instruction-following capabilities.
- Prototyping: A good choice for quickly developing and testing conversational AI features.
- Educational Purposes: Its manageable size makes it accessible for learning about and experimenting with instruction-tuned LLMs.