search

jungeun122333 / UVI-Net

Official implementation of "Data-Efficient Unsupervised Interpolation Without Any Intermediate Frame for 4D Medical Images" (CVPR 2024) in PyTorch.
MIT License
27 stars 2 forks source link

RuntimeWarning and NaN Values in Loss #1

Open AllenDiiaz opened 3 months ago

AllenDiiaz commented 3 months ago

Hi, I am a master's student from National Taiwan Univ, studying in the Department of Medical Engineering. I am impressed with your work and would like to seek some advice on a few issues I am encountering! Two issues while running the training script:

1. RuntimeWarning: I received the following warning during execution:

/home/allen/.conda/envs/UVI/lib/python3.8/site-packages/numpy/core/_methods.py:233: RuntimeWarning: invalid value encountered in subtract
  x = asanyarray(arr - arrmean)

2. NaN Values in loss_all/train: Additionally, the loss_all/train,loss_full/train metrics are consistently showing NaN values throughout the training process.

I have tried two different versions of the 4D-Lung-Preprocessed dataset (CVPR and update version), but the same problem persists.

Environment:

Hardware: Server: Supermicro GPU: NVIDIA A100 80GB PCIe

Software: Operating System: Windows 11 Remote Access Tool: MobaXterm 23.2 Python/PyTorch Version: Same as specified in the requirements.txt provided by the repository

Thank you for your patience and your prompt responses!

Best, Allen

螢幕擷取畫面 2024-08-03 180137

jungeun122333 commented 2 months ago

Dear Allen,

I apologize for the delayed response.

Have you had a chance to work with the cardiac dataset? I would like to determine whether the error is originating from the dataset itself or the code.