---
title: Microsoft Phi-2 Neuroscience (CPU)
emoji: 🧠
colorFrom: blue
colorTo: indigo
sdk: gradio
sdk_version: 5.20.0
app_file: app.py
pinned: false
hardware:
  cpu: true
---

# Neuroscience Fine-tuned Phi-2 Model

This is a Gradio app to interact with a neuroscience-specific fine-tuned version of Microsoft's Phi-2 model.

## Model Description

This model is fine-tuned from Microsoft's Phi-2 using LoRA (Low-Rank Adaptation) to specialize in neuroscience domain content. It's designed to assist with:

- Neuroscience research questions
- Explanations of neurological concepts
- Summaries of recent findings in the field
- Assistance with neuroscience-related topics

## Training Details

- **Base Model**: Microsoft Phi-2 (2.7B parameters)
- **Training Type**: LoRA fine-tuning
- **Training Dataset**: BrainGPT/train_valid_split_pmc_neuroscience_2002-2022_filtered_subset
- **Parameter-Efficient Fine-Tuning**: Used LoRA with r=16, alpha=32

## Space Features

This Gradio interface allows you to:

- Input prompts related to neuroscience
- Adjust generation parameters (temperature, top-p, etc.)
- Try example prompts to see the model's capabilities

## Model Repository

The model weights are available at [alaamostafa/Microsoft-Phi-2](https://huggingface.co/alaamostafa/Microsoft-Phi-2)

Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference