Nilayan87/ocean_hazard_onnx

🌊 Ocean Hazard Detection (Quantized ONNX)

This repository contains a quantized ONNX version of the Ocean Hazard Detection model, optimized for faster inference and lower memory usage.

The model can classify social media posts into:

• Hazard Report
• Non-Hazard

📂 Files

• model_quantized.onnx : Quantized ONNX model
• config.json : Model configuration
• tokenizer.json, spiece.model : Tokenizer files
• special_tokens_map.json, tokenizer_config.json : Tokenizer configs
• ort_config.json : ONNX Runtime configuration

🚀 Usage Example

from optimum.onnxruntime import ORTModelForSequenceClassification
from transformers import AutoTokenizer
import torch

# Load model + tokenizer
model = ORTModelForSequenceClassification.from_pretrained("Nilayan87/ocean_hazard_onnx")
tokenizer = AutoTokenizer.from_pretrained("Nilayan87/ocean_hazard_onnx")

# Example input
text = "Cyclone alert near Vizag coast ⚠️ stay safe!"
inputs = tokenizer(text, return_tensors="pt")
outputs = model(**inputs)

# Prediction
pred = torch.argmax(outputs.logits, dim=1).item()
print("Prediction:", pred)