Trained by multiple GPUs,Improve the training speed

[githubueser]

Hello author, Thank you very much for uploading the source code of your work project. Can the program be trained by calling multiple GPUs at the same time to improve the training speed?If I use the parameters in the project, 500 epochs, each round is 30 minutes, then 500 rounds is too long, can I modify the parameters somewhere, can I improve the training speed?We look forward to hearing from you, thank you