FalconX80/NisoriV2llama7b

NisoriV2 LLaMA 7B

A fine-tuned version of Meta’s LLaMA 7B, created by Priyanshu Bhatt.
This model (NisoriV2) has been fine-tuned on Dev Bhoomi Uttarakhand University (DBUU) data for instruction-following and question answering in a conversational style.

---

🧠 Model Details

• Base Model: LLaMA 7B
• Fine-Tuned Model: NisoriV2 LLaMA 7B
• Fine-Tuned By: Priyanshu Bhatt (FalconX80)
• Dataset: Dev Bhoomi Uttarakhand University (DBUU) educational & institutional data
• Architecture: Decoder-only Transformer
• Libraries: Hugging Face Transformers

---

🎯 Intended Use

This model is designed for:

• Question Answering related to Dev Bhoomi Uttarakhand University
• Educational and Institutional Information Retrieval
• Conversational AI for DBUU context

---

📊 Example

Input Prompt

who is the chairman of dev bhoomi uttarakhand university?

Model Output

The Chairman of Dev Bhoomi Uttarakhand University is Shri Sanjay Bansal. 
He is a well-known businessman and philanthropist who has made significant contributions to the field of education. 
Under his leadership, the university has made tremendous progress and has established itself as a premier institution of higher learning in the region.

---

🚀 How to Use

import re
from transformers import AutoTokenizer, LlamaForCausalLM
import torch

model_name = "FalconX80/NisoriV2llama7b"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = LlamaForCausalLM.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")

def chat(prompt):
    inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
    outputs = model.generate(inputs.input_ids, max_length=200)
    text = tokenizer.decode(outputs[0], skip_special_tokens=True)

    # Cleaning (remove [INST] tokens and repeated prompt)
    if text.lower().startswith(prompt.lower()):
        text = text[len(prompt):].strip()
    text = re.sub(r"\[/?INST\]", "", text).strip()

    return text

print(chat("Tell me about Dev Bhoomi Uttarakhand University"))

---

🛠️ Training Details

• Fine-Tuning Task: Instruction Following & QA
• Framework: Hugging Face Transformers + PyTorch
• Hardware Used: GPU (Kaggle / Colab / custom environment)
• Dataset: Proprietary Dev Bhoomi Uttarakhand University (DBUU) dataset prepared for institutional knowledge and educational Q&A.

---

⚠️ Limitations & Biases

• The model knowledge is limited to DBUU data used during fine-tuning.
• May hallucinate facts outside the scope of DBUU.
• Not optimized for safety-critical applications.
• Responses may vary depending on prompt phrasing.

---

---

📬 Contact

Created by Priyanshu Bhatt

• 📧 Email: priyanshubhatt80@gmail.com