TransformerXANES

[shelllbw]

Implement TransformerXANES in XANESNET
src code: ar-simplified-double-pdos.zip

• Implement transformer.py in the models/ dir.
  Note: Define the forward function as forward (self, batch) and unpack attributes within the function. This ensures that nn_learn and nn_predict can be applied directly to the model. Refer to the GNN model as an example.
• Add hybrid loss to switch.py
• Create datasets/ directory. Move the graph dataset data_graph into this dir, and implement XANESDataset class in the dir
• Add early stop option