Summary: 47 instances, 37 unique

Text	Count
model.eval()  # TODO: decide where eval() should be called	1
# TODO: set int8 backend accordingly if needed	1
# TODO: After implementing `ms_deform_attn` CUDA kernels for FP16, we can remove the	1
# TODO make it support different-sized images	1
# TODO: merge with d2.engine.launch	1
# TODO: Support subsample n images	1
# TODO: decide if move vis_wrapper inside this interface	1
# TODO make this more general	1
# TODO this should probably be a separate loss, not hacked in this one here	2
# For backward compatibility, TODO: remove this function.	1
# TODO: patch other meta-archs defined in D2	1
# TODO: check why some parameters will have gradient	1
# TODO: add test_train	1
# TODO this is a hack	2
# TODO: potentially add this config	1
# FIXME should we update the area here if there are no boxes?	1
# TODO find a better way of exposing other arguments	1
# TODO prctl in spawned processes	2
# TODO use valid to mask invalid areas due to padding in loss	1
# FIXME h_boxes takes the last one computed, keep this in mind	1
# bottom_up and leaving other parts un-quantized. TODO (T109761730): However	1
# TODO cache ?	1
# TODO make top-100 as an option for non-focal-loss as well	1
# TODO: maybe we should apply this by default	1
# TODO: model.to(device) might not work for detection meta-arch, this function is the	1
# TODO: output of backbone is a dict and currently this will keep all output	1
# TODO: figure out how to validate outputs	1
# TODO: maybe suppress shape mismatch	1
# TODO: replace tbx_writter with Lightning's self.logger.experiment	1
# we use smaller numbers.  TODO: reduce proposals for test in .yaml directly.	8
# TODO: wrap the normalizer and make it quantizable	1
# TODO: the export may change model it self, need to fix this	1
# TODO: disallow passing mapper and use registry for all mapper registering	1
"data_processes_stats": {},  # TODO: add worker stats	1
# FIXME: DDPStrategy has been renamed to DDPStrategy, however internal version is	1
# TODO: entirely rely on PathManager for caching	1
class TemplateTestCase(unittest.TestCase):  # TODO: maybe subclass from TestMetaArch	1