Repository navigation
PerceptualLoss Bug #8412
Description
Describe the bug
When I use PerceptualLoss, if the loss is too low, it may occur "RuntimeError: Function ‘SqrtBackward0‘ returned nan values in its 0th output", I will provide the specific error location later.
Environment
MONAI version: 1.4.0
Numpy version: 1.26.4
Pytorch version: 2.4.0+cu118
MONAI flags: HAS_EXT = False, USE_COMPILED = False, USE_META_DICT = False
MONAI rev id: 46a5272
MONAI file: //miniconda3/envs/monai/lib/python3.10/site-packages/monai/init.py
Optional dependencies:
Pytorch Ignite version: NOT INSTALLED or UNKNOWN VERSION.
ITK version: NOT INSTALLED or UNKNOWN VERSION.
Nibabel version: NOT INSTALLED or UNKNOWN VERSION.
scikit-image version: NOT INSTALLED or UNKNOWN VERSION.
scipy version: 1.15.2
Pillow version: 11.1.0
Tensorboard version: 2.19.0
gdown version: NOT INSTALLED or UNKNOWN VERSION.
TorchVision version: 0.19.0+cu118
tqdm version: 4.67.1
lmdb version: NOT INSTALLED or UNKNOWN VERSION.
psutil version: 7.0.0
pandas version: NOT INSTALLED or UNKNOWN VERSION.
einops version: 0.8.1
transformers version: NOT INSTALLED or UNKNOWN VERSION.
mlflow version: NOT INSTALLED or UNKNOWN VERSION.
pynrrd version: NOT INSTALLED or UNKNOWN VERSION.
clearml version: NOT INSTALLED or UNKNOWN VERSION.
Activity
cvbourne commentedon Apr 7, 2025
Perhaps related to this issue? https://github.com/richzhang/PerceptualSimilarity/pull/139/files
@CugAgo Please provide an error location and if it's the same root cause, I'll submit a PRQ (if this is a 👍 issue for a first time contributor).
CugAgo commentedon Apr 7, 2025
yes, I'm sorry for forgetting to provide accurate error information yesterday. In the normalize_tensor function, I think this bug can be fixed through this method, good job!
cvbourne commentedon Apr 7, 2025
Great. #8414
CugAgo commentedon Apr 8, 2025
May I ask when this source code will be updated to the MONAI library? Can I update the corresponding version through pip
Fix NaN gradient in PerceptualLoss normalize_tensor (#8982)