README.md

metadata

language: en
tags:
  - llama
  - llama-3.2
  - function-calling
  - instruction-tuning
  - conversational
license: llama2

NeuralTau Functions v1

This is a full version of the model fine-tuned on the full dataset. The model is trained to understand and follow complex instructions, providing detailed explanations and performing function-like operations in a conversational manner.

Model Variants Available

• 16-bit full model
• GGUF Q4_K_M quantized version (recommended for most use cases)
• GGUF Q8_0 quantized version (higher quality, larger size)

Training Details

• Base Model: unsloth/Llama-3.2-3B-Instruct
• Training Dataset: 0xroyce/NeuralTau-With-Functions-chat (https://huggingface.co/datasets/0xroyce/NeuralTau-With-Functions-chat)
• Training Method: LoRA fine-tuning with r=16
• Library Used: Unsloth
• Parameters: 3 billion

Usage

The model follows the Llama chat format. You can interact with it using:

messages = [
    {"role": "user", "content": "Your instruction or question here"},
]

Function calling example:

>>> how do i do a function for weather? use <tool_call> </tool_call>
<tool_call>
{"arguments": {"location": "Los Angeles", "time_period": "current"}, "name": "get_weather_data"}
</tool_call>

Model Capabilities

• Understanding and following complex instructions
• Providing detailed explanations and analysis
• Breaking down complex topics into understandable components
• Function-like operations and systematic problem-solving
• Maintaining context in multi-turn conversations
• Generating clear and structured responses

License

This model is subject to the Llama 2 license.