AttributeError: 'AveragedHausdorffLoss' object has no attribute '_backward_hooks'

[lg2578]

the AveragedHausdorffLoss is copied from https://github.com/HaipengXiong/weighted-hausdorff-loss/blob/master/object-locator/losses.py

class AveragedHausdorffLoss(nn.Module):
    def __init__(self):
        super(nn.Module, self).__init__()

    def forward(self, set1, set2):
        """
        Compute the Averaged Hausdorff Distance function
         between two unordered sets of points (the function is symmetric).
         Batches are not supported, so squeeze your inputs first!
        :param set1: Tensor where each row is an N-dimensional point.
        :param set2: Tensor where each row is an N-dimensional point.
        :return: The Averaged Hausdorff Distance between set1 and set2.
        """

        assert set1.ndimension() == 2, 'got %s' % set1.ndimension()
        assert set2.ndimension() == 2, 'got %s' % set2.ndimension()

        assert set1.size()[1] == set2.size()[1], \
            'The points in both sets must have the same number of dimensions, got %s and %s.'\
            % (set2.size()[1], set2.size()[1])

        d2_matrix = cdist(set1, set2)

        # Modified Chamfer Loss
        term_1 = torch.mean(torch.min(d2_matrix, 1)[0])
        term_2 = torch.mean(torch.min(d2_matrix, 0)[0])

        res = term_1 + term_2

        return res

and new trainer's code:

from nnunet.training.network_training.nnUNetTrainerV2 import nnUNetTrainerV2
from nnunet.training.loss_functions.hausdroff_loss import AveragedHausdorffLoss
from nnunet.utilities.nd_softmax import softmax_helper



class nnUNetTrainerV2_Loss_HD(nnUNetTrainerV2):
    def __init__(self, plans_file, fold, output_folder=None, dataset_directory=None, batch_dice=True, stage=None,
                 unpack_data=True, deterministic=True, fp16=False):
        super().__init__(plans_file, fold, output_folder, dataset_directory, batch_dice, stage, unpack_data,
                         deterministic, fp16)
        self.initial_lr = 1e-3
        self.loss = AveragedHausdorffLoss() 

error message:
epoch: 0
Traceback (most recent call last):
File "/root/miniconda3/envs/myconda/bin/nnUNet_train", line 33, in
sys.exit(load_entry_point('nnunet', 'console_scripts', 'nnUNet_train')())
File "/mnt/nnUNet/nnunet/run/run_training.py", line 179, in main
trainer.run_training()
File "/mnt/nnUNet/nnunet/training/network_training/nnUNetTrainerV2.py", line 440, in run_training
ret = super().run_training()
File "/mnt/nnUNet/nnunet/training/network_training/nnUNetTrainer.py", line 317, in run_training
super(nnUNetTrainer, self).run_training()
File "/mnt/nnUNet/nnunet/training/network_training/network_trainer.py", line 456, in run_training
l = self.run_iteration(self.tr_gen, True)
File "/mnt/nnUNet/nnunet/training/network_training/nnUNetTrainerV2.py", line 249, in run_iteration
l = self.loss(output, target)
File "/root/miniconda3/envs/myconda/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1110, in _call_impl
return forward_call(*input, **kwargs)
File "/mnt/nnUNet/nnunet/training/loss_functions/deep_supervision.py", line 39, in forward
l = weights[0] * self.loss(x[0], y[0])
File "/root/miniconda3/envs/myconda/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1108, in _call_impl
if not (self._backward_hooks or self._forward_hooks or self._forward_pre_hooks or _global_backward_hooks
File "/root/miniconda3/envs/myconda/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1185, in getattr
raise AttributeError("'{}' object has no attribute '{}'".format(
AttributeError: 'AveragedHausdorffLoss' object has no attribute '_backward_hooks'
Exception in thread Thread-4:
Traceback (most recent call last):
File "/root/miniconda3/envs/myconda/lib/python3.8/threading.py", line 932, in _bootstrap_inner
self.run()
File "/root/miniconda3/envs/myconda/lib/python3.8/threading.py", line 870, in run
self._target(*self._args, **self._kwargs)
File "/root/miniconda3/envs/myconda/lib/python3.8/site-packages/batchgenerators/dataloading/multi_threaded_augmenter.py", line 92, in results_loop
raise RuntimeError("Abort event was set. So someone died and we should end this madness. \nIMPORTANT: "
RuntimeError: Abort event was set. So someone died and we should end this madness.
IMPORTANT: This is not the actual error message! Look further up to see what caused the error. Please also check whether your RAM was full
Exception in thread Thread-5:
Traceback (most recent call last):
File "/root/miniconda3/envs/myconda/lib/python3.8/threading.py", line 932, in _bootstrap_inner
self.run()
File "/root/miniconda3/envs/myconda/lib/python3.8/threading.py", line 870, in run
self._target(*self._args, **self._kwargs)
File "/root/miniconda3/envs/myconda/lib/python3.8/site-packages/batchgenerators/dataloading/multi_threaded_augmenter.py", line 92, in results_loop
raise RuntimeError("Abort event was set. So someone died and we should end this madness. \nIMPORTANT: "
RuntimeError: Abort event was set. So someone died and we should end this madness.
IMPORTANT: This is not the actual error message! Look further up to see what caused the error. Please also check whether your RAM was full

[FabianIsensee]

Hey, this is an issue that you should post on the repository you took the loss from. You may also want to take a look at this repo: https://github.com/JunMa11/SegLoss

[lg2578]

thanks very much.I will try the SegLoss repo.