Phi-2 Fine-tuned Assistant Demo

This Space demonstrates a fine-tuned version of the Microsoft Phi-2 model, trained on the OpenAssistant dataset using QLoRA (Quantized Low-Rank Adaptation). The model is designed to provide helpful and informative responses to various types of queries and instructions.

Model Information

Base Model: Microsoft Phi-2
Fine-tuning Dataset: OpenAssistant
Training Method: QLoRA with 8-bit quantization
Model Card: jatingocodeo/phi2-finetuned-openassistant

How to Use

Enter your query in the text box
Adjust generation parameters (optional):
- Maximum Length (50-500): Controls response length
- Temperature (0.1-1.0): Controls randomness
- Top P (0.1-1.0): Controls token sampling

Example Prompts

Try these examples to see what the model can do:

"What is machine learning?"
"Write a short poem about artificial intelligence"
"Explain quantum computing to a 10-year-old"
"What are the best practices for writing clean code?"

Model Capabilities

The model is trained to:

Provide informative explanations
Answer questions clearly and concisely
Generate creative content
Give technical explanations
Follow instructions and complete tasks

Limitations

The model may occasionally generate incorrect information
Responses are limited by the training data
The model should not be used for critical applications without human oversight
Complex or ambiguous queries might receive simplified responses

Technical Details

The model uses:

8-bit quantization for efficient inference
Gradient checkpointing
Mixed precision training
LoRA fine-tuning techniques

License

This demo uses a model that inherits the license of the base Phi-2 model and the OpenAssistant dataset.