How can we train Multi Layer Perceptron(MLP) model with type float-16, float-8 and int8 through Pytorch?

Question

Currently, I am training an MLP using the float32 data type. I also attempted to train the MLP using the float-16 data type by utilizing Half(), but it resulted in a NAN error. Could someone please assist me in providing the PyTorch code for training the MLP model with float16, float8, and int8 data types?

Currently, I am training an MLP using the float32 data type. I also attempted to train the MLP using the float-16 data type by utilizing Half(), but it resulted in a NAN error.