---
base_model: unsloth/gemma-4-E2B-it-unsloth-bnb-4bit
model:
- id: rajasingh012/vidya-gemma4-e2b-gguf
tags:
- unsloth
- fine-tuned
- gemma-4
- ncert
- socratic-tutoring
- education
- llm
library: llama.cpp
quantization: Q4_K_M
inference:
  parameters:
    temperature: 0.7
    top_k: 40
    top_p: 0.95
---

# Vidya - NCERT Socratic Learning Bot

Fine-tuned Gemma 4 E2B (2.3B params) on NCERT socratic dialogue data.

## Model Details

- **Base model**: unsloth/gemma-4-E2B-it-unsloth-bnb-4bit
- **Fine-tuning method**: QLoRA (rank=32, alpha=32)
- **Format**: GGUF Q4_K_M (~3.1 GB)
- **Context length**: 2048 tokens
- **Trained on**: 321 socratic dialogue samples (NCERT Class 6-10 Science)

## Usage

This model is compatible with llama.cpp and Ollama. Download the GGUF file from the Files tab, or use the HuggingFace Hub API:

```python
from huggingface_hub import hf_hub_download
path = hf_hub_download(repo_id='rajasingh012/vidya-gemma4-e2b-gguf', 
                       filename='gemma-4-e2b-it-pretrained.Q4_K_M.gguf')
```

## Mobile App

This model powers the Vidya Android app (NCERT Socratic tutoring chatbot).
See: https://github.com/rajasingh012/vidya-android

## Training

Trained via Unsloth FastModel API on Kaggle GPU (T4 x2).
Full training logs and data pipeline available in the ncert-wiki repository.