Instructions to use Bhaveshbhau44/Llama-3.2-3B-Gita-Vedic-Finetuned with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Bhaveshbhau44/Llama-3.2-3B-Gita-Vedic-Finetuned with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="Bhaveshbhau44/Llama-3.2-3B-Gita-Vedic-Finetuned")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForMultimodalLM

tokenizer = AutoTokenizer.from_pretrained("Bhaveshbhau44/Llama-3.2-3B-Gita-Vedic-Finetuned")
model = AutoModelForMultimodalLM.from_pretrained("Bhaveshbhau44/Llama-3.2-3B-Gita-Vedic-Finetuned")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use Bhaveshbhau44/Llama-3.2-3B-Gita-Vedic-Finetuned with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Bhaveshbhau44/Llama-3.2-3B-Gita-Vedic-Finetuned"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Bhaveshbhau44/Llama-3.2-3B-Gita-Vedic-Finetuned",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Bhaveshbhau44/Llama-3.2-3B-Gita-Vedic-Finetuned

SGLang

How to use Bhaveshbhau44/Llama-3.2-3B-Gita-Vedic-Finetuned with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "Bhaveshbhau44/Llama-3.2-3B-Gita-Vedic-Finetuned" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Bhaveshbhau44/Llama-3.2-3B-Gita-Vedic-Finetuned",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "Bhaveshbhau44/Llama-3.2-3B-Gita-Vedic-Finetuned" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Bhaveshbhau44/Llama-3.2-3B-Gita-Vedic-Finetuned",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Bhaveshbhau44/Llama-3.2-3B-Gita-Vedic-Finetuned with Docker Model Runner:
```
docker model run hf.co/Bhaveshbhau44/Llama-3.2-3B-Gita-Vedic-Finetuned
```

Llama 3.2 3B - Bhagavad Gita Fine-tuned (Fully Merged)

This is a fully merged and quantization-ready version of Llama 3.2 3B fine-tuned on the Bhagavad Gita corpus for understanding Vedic philosophy and [translate:Dharma].

🎯 Model Overview

Base Model: Meta's Llama 3.2 3B Instruct
Fine-tuned Dataset: Bhagavad Gita (701 verses)
Fine-tuning Method: LoRA (Low-Rank Adaptation) merged into full model
Total Parameters: 3.2 Billion
Model Size: ~6-7 GB (merged)
Training Hardware: Google Cloud TPU v6e-8 (8 cores)
Final Loss: 1.7188
Training Time: ~20 minutes

📊 Training Details

Dataset

Source: Bhagavad Gita Dataset (HuggingFace)
Samples: 701 verses
Language: English translations with Sanskrit references
Content: [translate:Bhagavad Gita verses covering Yoga, Dharma, Karma, and spiritual philosophy]

Fine-tuning Configuration

LoRA Configuration:

Rank (r): 16

Alpha: 32

Target Modules: q_proj, k_proj, v_proj, o_proj

Dropout: 0.05

Learning Rate: 1e-4

Batch Size: 4

Optimizer: AdamW

Max Length: 512 tokens

text

Hardware & Compute

Hardware: Google Cloud TPU v6e-8

8 TPU cores running in parallel

Distributed training across all cores

BF16 precision for efficiency

Compute Resources:

Training Duration: ~20 minutes

Samples per core: ~88-90

Total batches: ~22

text

🧠 Model Capabilities

This model excels at:

✅ Understanding [translate:Dharma] - Discussing duty, righteousness, and cosmic law
✅ Karma Yoga concepts - Explaining action without attachment
✅ Philosophical discussions - Deep understanding of ancient wisdom
✅ Sanskrit references - Recognizing and explaining Sanskrit terms
✅ Vedic knowledge - Providing context from Vedic literature
✅ Meditation & spirituality - Guidance on inner peace and enlightenment

🚀 Quick Start

Installation

pip install transformers torch

text

Basic Usage

import torch from transformers import AutoModelForCausalLM, AutoTokenizer Load model and tokenizer

model = AutoModelForCausalLM.from_pretrained( "Bhaveshbhau44/Llama-3.2-3B-Gita-Vedic-Finetuned", torch_dtype=torch.float16, device_map="auto" )

tokenizer = AutoTokenizer.from_pretrained( "Bhaveshbhau44/Llama-3.2-3B-Gita-Vedic-Finetuned" ) Generate text

prompt = "What is the meaning of Karma Yoga?" inputs = tokenizer(prompt, return_tensors="pt").to(model.device)

with torch.no_grad(): outputs = model.generate( **inputs, max_new_tokens=150, temperature=0.7, top_p=0.9, do_sample=True )

response = tokenizer.decode(outputs, skip_special_tokens=True) print(response)

text

📈 Example Outputs

Q: What is the purpose of life according to the Gita?

A: The purpose of life according to the Gita is to attain Moksha or Liberation. Moksha is the state of ultimate bliss and freedom from the cycle of birth and death. The Bhagavad Gita teaches that one can attain Moksha by performing one's duty without attachment and expectation of reward. This is achieved through the practice of Karma Yoga - the path of action...

Q: What does Krishna teach about duty?

A: Krishna teaches that duty ([translate:Dharma]) is the foundation of righteous living. One should perform their duties according to their nature and station in life without being attached to the fruits of their actions. The Gita emphasizes that it is better to perform one's own duty imperfectly than to perform another's duty perfectly...

Q: How can one achieve inner peace?

A: According to the Bhagavad Gita, inner peace is achieved through the practice of yoga and meditation. The mind must be trained to focus inward, away from the distractions and pleasures of the world. Through consistent practice and devotion, one can quiet the mind and experience the peace that comes from direct realization of the Divine...

🎓 Educational Applications

This model is ideal for:

📚 Educational platforms teaching Vedic philosophy
🧘 Meditation and wellness apps
📖 Spiritual and philosophical chatbots
🌍 Cultural understanding applications
💭 Thought-provoking Q&A systems

⚠️ Limitations

Fine-tuned specifically on English Bhagavad Gita translations
May have context limitations beyond 512 tokens
Best performance with philosophical/spiritual queries
Should not be used as sole source for religious/spiritual guidance

📝 Citation

If you use this model, please cite:

@misc{bhavesh_llama_gita_2025, title={Llama 3.2 3B - Bhagavad Gita Fine-tuned}, author={Bhavesh Varma}, year={2025}, publisher={HuggingFace Hub}, url={https://huggingface.co/Bhaveshbhau44/Llama-3.2-3B-Gita-Vedic-Finetuned} }

text

📞 Model Information

Developer: Bhavesh Varma
Model Type: Causal Language Model
Base Model License: Llama Community License
Training Framework: PyTorch, Transformers, PEFT, torch_xla
Hardware: Google Cloud TPU

🔗 Related Models

⭐ Acknowledgments

Meta for Llama 3.2 model
HuggingFace for Transformers and Hub
Google Cloud for TPU infrastructure
Bhagavad Gita Dataset contributors

Last Updated: November 4, 2025
Model Status: ✅ Active & Ready to Use

Downloads last month: 11

Safetensors

Model size

3B params

Tensor type

BF16