Use large, local streaming dataset to finetune object detection model

Hi! Can you please show me how to adapt the code in this script: transformers/examples/pytorch/object-detection/run_object_detection.py at main · huggingface/transformers · GitHub to train using a custom streaming dataset?

I have my data loaded into a Dataset, and can instantiate it with streaming=True. Is there a simple way to adjust the Trainer so that it expects the streaming dataset?

1 Like