---
language:
- en
tags:
- gpt2
- keras-nlp
- text-generation
- bhagavad-gita
license: mit
datasets:
- custom
pipeline_tag: text-generation
---

# 📖 Fine-tuned GPT-2 on the Bhagavad Gita

This repository contains a **fine-tuned GPT-2 model** trained on the *Bhagavad Gita* (English meanings/verses).  
It generates text inspired by the teachings and style of the Bhagavad Gita.

---

## 🧾 Model Card

| Attribute        | Details |
|------------------|---------|
| **Base Model**   | [GPT-2 (124M)](https://huggingface.co/openai-community/gpt2) |
| **Architecture** | Decoder-only Transformer |
| **Framework**    | TensorFlow / KerasNLP |
| **Dataset**      | Bhagavad Gita (English meanings) |
| **Languages**    | English |
| **Dataset Size** | ~700 verses |
| **Tokenizer**    | GPT-2 tokenizer (inherited vocabulary) |

---

## 📊 Training Details

- **Epochs**: 3  
- **Batch Size**: 8  
- **Learning Rate**: 3e-5  
- **Optimizer**: AdamW  
- **Loss Function**: Causal Language Modeling (CrossEntropy)  
- **Preprocessing**: Cleaned, tokenized, and formatted into text sequences  

---

## 🚀 Usage

You can load and run the model using Hugging Face `transformers`:

```python
from transformers import AutoTokenizer, AutoModelForCausalLM

# Load tokenizer & model
tokenizer = AutoTokenizer.from_pretrained("AP6621/Bhagawatgitagpt")
model = AutoModelForCausalLM.from_pretrained("AP6621/Bhagawatgitagpt")

# Generate text
inputs = tokenizer("Arjuna asked:", return_tensors="pt")
outputs = model.generate(**inputs, max_length=100, do_sample=True, top_p=0.9, temperature=0.8)

print(tokenizer.decode(outputs[0], skip_special_tokens=True))
```

---

## ✨ Example

**Prompt**:
```
Arjuna asked:
```

**Generated Response**:
*"O Arjuna, when the mind is detached from desires, the soul attains peace and wisdom. Such a person sees all beings with equal vision."*

---

## ⚖️ Limitations & Bias

- Small dataset → may **repeat phrases** or **hallucinate content**  
- Not an official or scholarly translation  
- Should not be used as a **religious authority**  

---

## ✅ Intended Use

✔️ Educational experiments  
✔️ Creative/spiritual-inspired text generation  
✔️ AI-assisted storytelling  

❌ Not for religious/spiritual authority  
❌ Not for official translations  
❌ Not for sensitive decision-making  

---

## 📌 Citation

If you use this model, please cite:

```bibtex
@misc{gpt2-bhagavadgita,
  title        = {Fine-tuned GPT-2 on Bhagavad Gita Dataset},
  author       = {Your Name},
  year         = {2025},
  publisher    = {Hugging Face},
  howpublished = {\url{https://huggingface.co/your-username/gpt2-bhagavad-gita}}
}
```

---

## 🙏 Acknowledgements

- [OpenAI](https://huggingface.co/openai-community) for GPT-2  
- Public Bhagavad Gita dataset  
- Hugging Face community for tools & inspiration  

---