iravikr/qwen3-0.6b-finance-india

This repository contains a fully fine-tuned and merged version of Qwen/Qwen3-0.6B, trained on the official Union Budget 2026–27 speech document.

The model has been instruction-tuned to answer questions related to:

• Fiscal policy
• Capital expenditure
• Sector-wise allocations
• Taxation changes
• Government initiatives and reforms

This is a standalone model.
No LoRA adapters or base model downloads are required separately.

---

🔍 Model Details

• Base model: Qwen/Qwen3-0.6B
• Fine-tuning method: QLoRA → merged into base model
• Model type: Causal Language Model
• Language: English
• Task: Instruction following, document-based Q&A
• Model size: ~0.6B parameters
• Deployment: CPU / GPU compatible

---

Training Dataset

The model was trained on an instruction-formatted dataset derived from the Union Budget 2026–27 PDF.

🔗 Dataset:
https://huggingface.co/datasets/your-username/qwen3-budget-2026-instructions

Dataset Structure

Each training sample includes:

• instruction: Task description
• input: Chunk of budget text
• output: Target response

---

Recommended

• Budget and policy question answering
• Government document assistants
• Educational and research use
• RAG + fine-tuning hybrid systems

Not Recommended

• Legal advice
• Financial or investment advice
• Real-time policy interpretation

---

How to Use

Load the Model

from transformers import AutoModelForCausalLM, AutoTokenizer

model_id = "your-username/qwen3-0.6b-budget-2026"

tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    device_map="auto"
)