gemma4-disaster-finetuned

Gemma4 Disaster Finetuned

This repository contains the fine-tuned LoRA adapters and the complete model weights for the Gemma-4-e4b model, customized for disaster management and preparedness conversational tasks.

Model Details

Model Description

This model has been fine-tuned on disaster response, management, and preparedness datasets to provide high-quality assistance during emergency scenarios. It helps users plan, respond, and find safety guidelines during natural or man-made disasters.

• Developed by: Sidhaarth Shree
• Model type: LoRA Adapter (PEFT) & Complete Model Weights
• Base model: unsloth/gemma-4-e4b-it-unsloth-bnb-4bit
• Language(s) (NLP): English
• License: MIT License

Fine-Tuning Notebook

You can find the interactive Jupyter/Kaggle notebook used for fine-tuning this model at:

• Kaggle Notebook: Disaster Management Preparedness Gemma 4 e4b

Repository Structure

The files in this repository are organized as follows:

• gemma4_e4b.litertlm: The full consolidated model weights (located at the root level).
• /lora: Subfolder containing all the fine-tuned LoRA adapters and configurations:
  • adapter_config.json
  • adapter_model.safetensors
  • chat_template.jinja
  • processor_config.json
  • tokenizer.json
  • tokenizer_config.json

Uses

Direct Use

This model is intended to be used directly in disaster preparedness dashboards, emergency communication simulators, or offline hazard assistants to offer immediate safety protocols and management tips.

How to Get Started with the Model

To load the fine-tuned LoRA adapters in python:

from peft import PeftModel, PeftConfig
from transformers import AutoModelForCausalLM, AutoTokenizer

config = PeftConfig.from_pretrained("Shree-007/gemma4-disaster-finetuned", subfolder="lora")
model = AutoModelForCausalLM.from_pretrained("unsloth/gemma-4-e4b-it-unsloth-bnb-4bit")
model = PeftModel.from_pretrained(model, "Shree-007/gemma4-disaster-finetuned", subfolder="lora")
tokenizer = AutoTokenizer.from_pretrained("Shree-007/gemma4-disaster-finetuned", subfolder="lora")

Framework Versions

• PEFT 0.18.1
• Transformers 4.40+
• PyTorch 2.0+
• Unsloth 2024.4+