Hugging Face Overview
Tess-2.0-Llama-3-70B Overview
Tess-2.0-Llama-3-70B, named after "Tesoro" (Treasure), is a 70 billion parameter general-purpose large language model developed by migtissera. It is built upon the robust meta-llama/Meta-Llama-3-70B base model, leveraging its strong foundational capabilities.
Key Characteristics
- Base Model: Fine-tuned from
meta-llama/Meta-Llama-3-70B. - Training Methodology: Utilizes LIMA (Less-Is-More) principles, focusing on a curated dataset for efficient learning.
- Dataset: Trained on the Tess-2.0 dataset, comprising approximately 100,000 high-quality code and general training samples.
- Instruction Following: Designed to be highly uncensored and consistently follow instructions, making it versatile for various applications.
- Training Depth: Fine-tuned for only two epochs with a low learning rate to preserve the base model's entropy.
- Prompt Format: Employs the Llama-3 prompt format for optimal interaction.
Intended Use Cases
This model is well-suited for applications requiring a powerful, general-purpose language model that adheres closely to user instructions. Its uncensored nature means it will attempt to fulfill requests without refusal, making it a strong candidate for:
- General conversational AI.
- Content generation across diverse topics.
- Code generation and assistance.
- Applications where consistent instruction following is paramount.
Limitations
As an uncensored model, Tess-2.0-Llama-3-70B may generate inappropriate, biased, or offensive content. Users should exercise caution and verify information, as the model can occasionally produce inaccurate or misleading results.