It seems you have implemented multiple gpus for tensorflow backend:

from imageai.Detection.Custom import DetectionModelTrainer
trainer = DetectionModelTrainer()
trainer.setModelTypeAsYOLOv3()
trainer.setDataDirectory(data_directory="/content/gdrive/MyDrive/ai-training/test")
trainer.setGpuUsage(1)
trainer.setTrainConfig(object_names_array=["Item 1", "Sample 2"], batch_size=4, num_experiments=200)
trainer.trainModel()

However, this has not yet been implemented for the pytorch backend. Is this something you intend to add in future?

Thank you!