Update dygrae.py

[mzamini92]

move the model and input tensors to the GPU for faster computations. If you have multiple input samples, you can process them in batches using PyTorch's DataLoader to parallelize computations and take advantage of batch operations. This can significantly speed up the training process. Initialize the parameters of the GatedGraphConv and LSTM layers using appropriate initialization methods. Adding dropout regularization can help prevent overfitting and improve generalization. if the sequence length is fixed, we can use the LSTMCell module instead of the LSTM module to process each time step individually.