Colabnotebook for UNET2D and 3D not utilizing GPU when perform training

[jinxsfe]

I am not sure it's my account problem, but when I try to different notebook for unet2D 3D, multilabel, for colab notbook, which can not using GPU colab during training, I double check the problem https://www.reddit.com/r/GoogleColab/comments/ql67h7/google_colab_not_using_gpu_properly/

also we can see the check button show I am utilize the GPU
said it's may be due to tenserflow problems, are you facing the same problems for training? how you solve it?

[jinxsfe]

@paxcalpt @IvanHCenalmor
@brunomsaraiva

[jinxsfe]

I double check the problem is tensorflow problems, at now the cuda verison update from 12.2 to 12.5, corresponding the tensorflow 2.18, seems tensorflow version must match the CUDA version, otherwise we have problems for training, before that you mentioned can downrgrade version for 2.18 into 2.15 to continue training progress. but 2.15 version~ at now only can rely CPU, when I changed back to 2.18,
googlecolab/colabtools#5086,

, this modified colab notebook version is https://colab.research.google.com/drive/1fTP-8_OPw9E8o8B1agQImTBzWATrvpc5?usp=drive_link

@esgomezm @paxcalpt

[jinxsfe]

I am sloved by down graded Cuda 12.5 to 12.2, if you update colab notebook version for tenserflow 12.8 or my notebook version please let me know