This is a fine-tuned version of Microsoft's Phi-2 model using QLoRA (Quantized Low-Rank Adaptation). The model has been trained to provide helpful responses for various tasks including coding, writing, and general assistance.
Model Details
Base Model: Microsoft Phi-2 (2.7B parameters)
Fine-tuning Method: QLoRA (4-bit quantization)
Training Data: Custom dataset focused on programming and professional communication
Hardware Used: NVIDIA RTX 4090 (24GB VRAM)
Usage
You can interact with the model through the Gradio interface by visiting the "Spaces" tab of this repository.
Local Installation
To run the model locally:
Clone this repository
Install dependencies:
pip install -r requirements.txt
Run the Gradio app:
python gradio_app.py
Parameters
Max Length: Controls the maximum length of the generated response (64-1024 tokens)
Temperature: Controls randomness in generation (0.1-1.0)
Top P: Controls diversity of generated responses (0.1-1.0)
Example Prompts
"Write a Python function to calculate the factorial of a number"
"Explain the concept of machine learning in simple terms"
"Write a professional email requesting a meeting with a client"
Limitations
The model works best with English language input
Response quality may vary depending on the complexity of the prompt
Maximum context length is limited to 2048 tokens
License
This model is subject to the Microsoft Phi-2 license terms and conditions.
Acknowledgments
Microsoft for the Phi-2 base model
Hugging Face for the transformers library and model hosting
The QLoRA paper authors for the quantization technique
