Transformer-Optimization TODO: Mixed precision in lightning (7secs to 3secs) model to onnx inference code adding proper readme