multi-cards training

[gqsmmz]

Do you have code for single player multi card training? I couldn't find it in the train. py and dataset code, only the torch run instruction