Convert your bulky Transformer models into lightweight high performance ONNX models!

How we converted our ALBERT model trained for text classification to ONNX runtime and how it suddenly increased to 358.3mb from 46.8mb of size( .bin weights file).