|
|
--- |
|
|
license: mit |
|
|
tags: |
|
|
- text-generation |
|
|
- study-assistant |
|
|
- emotionally-intelligent |
|
|
- casual-ai |
|
|
inference: true |
|
|
language: |
|
|
- en |
|
|
library_name: transformers |
|
|
pipeline_tag: text-generation |
|
|
widget: |
|
|
- text: "Can you help me study math?" |
|
|
--- |
|
|
|
|
|
|
|
|
--- |
|
|
base_model: microsoft/phi-2 |
|
|
library_name: peft |
|
|
tags: |
|
|
- study-assistant |
|
|
- education |
|
|
- phi-2 |
|
|
- fine-tuned |
|
|
- emotional-intelligence |
|
|
license: mit |
|
|
inference: true |
|
|
--- |
|
|
|
|
|
# Study Buddy AI (Lite) |
|
|
|
|
|
A fine-tuned version of Microsoft's Phi-2, trained to act as a casual, emotionally intelligent study assistant that helps users learn deeply through conversation. Designed for real-time interaction inside a lightweight web app. |
|
|
|
|
|
 <!-- Replace this with your logo when ready --> |
|
|
|
|
|
## Model Details |
|
|
|
|
|
### Model Description |
|
|
|
|
|
Study Buddy AI Lite is built on `microsoft/phi-2` and fine-tuned using PEFT (Parameter-Efficient Fine-Tuning). It was trained to behave like a study partner — casual, kind, encouraging, and deeply helpful for academic subjects and personal growth. |
|
|
|
|
|
### Features |
|
|
|
|
|
- Friendly tone tailored for Gen Z and students |
|
|
- Explains complex topics using simple language |
|
|
- Emotionally intelligent responses — supportive and motivational |
|
|
- Great for studying, reviewing, or breaking down hard concepts |
|
|
|
|
|
### Use Cases |
|
|
|
|
|
- Personal tutoring via chat |
|
|
- Homework help and study review |
|
|
- Emotional motivation during learning sessions |
|
|
- Supplementing lessons with a relatable assistant |
|
|
|
|
|
### Limitations |
|
|
|
|
|
- Doesn’t replace real tutors or teachers |
|
|
- May hallucinate or simplify info if pushed too far |
|
|
- Best used in short back-and-forths, not long essays |
|
|
|
|
|
## Training & Fine-Tuning |
|
|
|
|
|
- **Base model**: `microsoft/phi-2` |
|
|
- **Fine-tuning method**: LoRA via the `peft` library |
|
|
- **Custom dataset**: 3,000+ curated prompts focused on simplified teaching, emotional tone, and helpful behavior |
|
|
- **Hosted via**: Hugging Face Inference Endpoints (GPU-backed) |
|
|
|
|
|
## How It Works |
|
|
|
|
|
Frontend ➝ Flask API ➝ Hugging Face Model ➝ Response ➝ KnowBetter Logic (to clean output) |
|
|
|
|
|
Built to run inside a React + Flask stack with CORS enabled and token-secured backend access to Hugging Face. |
|
|
|
|
|
## Model Creator |
|
|
|
|
|
**Creator**: Micah Miles |
|
|
**GitHub/Contact**: redmint-him |
|
|
**Team**: Redmint |
|
|
**Project**: Study Buddy Lite |
|
|
|
|
|
## License |
|
|
|
|
|
This model is licensed under the [MIT License](https://opensource.org/licenses/MIT). You can use, modify, or remix it with attribution. |
|
|
|
|
|
--- |
|
|
|
|
|
✨ “Never back down. Never give up.” — This is for every student who wanted to learn differently. |
|
|
|
|
|
I'm just a kid with a dream. |
