Converting FP16 to FP32 while exporting pytorch model to ONNX

1

I have trained the pytorch model on half_precision, now can I use FP32 when I am trying to export it in ONNX format?