File size: 2,243 Bytes
80493f3 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 |
# GodeusAI - Conversations with the infinite, locally hosted.
GodeusAI is a spiritual assistant chatbot powered by a fine-tuned Mistral-7B language model with LoRA adapters. It provides answers to spiritual and philosophical questions, and can be run locally for private, offline use.
---
## Features
- Chatbot interface for spiritual Q&A
- Fine-tuned on custom spiritual/philosophical data
- LoRA adapter for efficient model adaptation
- Utilities for context-based question answering
- Example notebook for data preparation and training
---
## Quickstart
### 1. Install Dependencies
```bash
pip install -r requirements.txt
```
### 2. Run the Chatbot
```bash
python inference.py
```
You’ll see a prompt:
`💉 Welcome to GodeusAI — your spiritual assistant. Type 'exit' to quit.`
---
## Project Structure
| File/Folder | Purpose |
|--------------------------|--------------------------------------------------------------|
| `inference.py` | Main script to chat with the model |
| `rag_utils.py` | Utilities for loading model and context-based QA |
| `few_shots_qa.jsonl` | Example Q&A pairs for few-shot prompting |
| `adapter_model.safetensors` | LoRA adapter weights (required for inference) |
| `adapter_config.json` | Configuration for the LoRA adapter |
| `GodeusAI.ipynb` | Notebook for data prep, training, and experimentation |
| `requirements.txt` | Python dependencies |
---
## Data & Training
- **few_shots_qa.jsonl**: Example format for few-shot Q&A pairs.
- **GodeusAI.ipynb**:
- Prepares data from PDFs
- Chunks and formats into JSONL
- Trains LoRA adapters on Mistral-7B
---
## Model
- **Base model**: `mistralai/Mistral-7B-v0.1`
- **Adapter**: LoRA (config in `adapter_config.json`, weights in `adapter_model.safetensors`)
---
## License
This project is for research and educational purposes.
Refer to the base model and dataset licenses for usage restrictions. |