pranav2711/ncu-smartllm-fp32

NCU SmartLLM (FP32) — Fine-Tuned Mistral-7B

A full-precision, fully fine-tuned version of Mistral-7B on domain-specific data from The NorthCap University (NCU) — fine-tuned to answer academic, administrative, and general university queries in an instructional format.

---

Model Summary

• Base Model: Mistral-7B
• Fine-tuning: Full fine-tuning (FP32)
• Architecture: Transformer Decoder (Causal LM)
• Dataset: NCU-specific instructions (approx. 1,100 records)
• Precision: Full precision (FP32)
• Trained on: Google Colab (T4 GPU, 4 Epochs)

---

Capabilities

The model has been trained to:

• Answer FAQs related to NCU (hostel, fees, scholarships, re-evaluation, etc.)
• Act as an academic assistant for students
• Handle general admin-related queries
• Demonstrate Mistral's capabilities in an Indian academic context

Fully plug-and-play with Hugging Face Transformers & inference APIs

---

Usage

Inference (Transformers)

pip install transformers accelerate

from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained("pranav2711/ncu-smartllm-fp32", device_map="auto")
tokenizer = AutoTokenizer.from_pretrained("pranav2711/ncu-smartllm-fp32")

prompt = "### Question:\nHow do I apply for hostel at NCU?\n\n### Answer:"
inputs = tokenizer(prompt, return_tensors="pt").to("cuda")
outputs = model.generate(**inputs, max_new_tokens=200)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))

---

Model Training Details

Detail	Value
Base model	mistralai/Mistral-7B-v0.1
Fine-tuned epochs	10
Batch size	2
Tokenizer max len	512
Output format	Instruction → Response
Loss (final epoch)	~1.405300

---

Dataset Format (Instruction-Tuned)

Each entry follows:

{
  "instruction": "How do I apply for a degree certificate?",
  "input": "I graduated in 2023.",
  "output": "You can apply for the degree certificate through the Registrar's Office. Submit your documents along with..."
}

Formatted as:

### Question:
[instruction + optional input]

### Answer:
[output]

---

Model Applications

• College ERP chatbots
• Student helpdesks
• Admission or exam queries
• EdTech integrations for personalized response systems

---

License

This model is released under the Apache 2.0 license. You are free to use, modify, and distribute it with attribution.

---

Author & Maintainer

• Pranav Singh
• The NorthCap University (NCU), Gurugram
• For collaboration: ping on Hugging Face or GitHub

---

How to Support

• Star this model on Hugging Face
• Try it on Spaces and share your feedback
• Contribute improvements or suggest datasets

---