Open lssam opened 5 years ago
I am trying to validate detection over ~87k classes with an FPN-faster-rcnn-101-2x on a validation set of 10000 Images. I keep getting an out of memory error, because of the inefficient way of validating in detectron with: https://github.com/facebookresearch/Detectron/blob/a6a835f5b8208c45d0dce217ce9bbda915f44df7/detectron/core/test_engine.py#L383
is there another option to validate with a memory friendly fashion ?
I am trying to validate detection over ~87k classes with an FPN-faster-rcnn-101-2x on a validation set of 10000 Images. I keep getting an out of memory error, because of the inefficient way of validating in detectron with: https://github.com/facebookresearch/Detectron/blob/a6a835f5b8208c45d0dce217ce9bbda915f44df7/detectron/core/test_engine.py#L383
is there another option to validate with a memory friendly fashion ?