clatfd / GNN-ART-LABEL

Code and dataset for the MICCAI 2020 paper: Automated Intracranial Artery Labeling using a Graph Neural Network and Hierarchical Refinement
Apache License 2.0
27 stars 4 forks source link

I have some errors... Could you help me? #6

Open JinyongChung opened 2 years ago

JinyongChung commented 2 years ago

I am really interested in your work. I tried to use this tool, However, I had troubles and could not solve these.

ArtLabel_errors.pdf

Here are my results. I could replicate the results until [14]. But then some warnings occurred and null results for the trainings. Finally, I got errors with "Blas GEMM launch failed".

How can I solve it?

Thank you.

clatfd commented 2 years ago

Hi Jinyong, Thanks for your interest. It looks like a package compatibility issue. Are you following the package versions suggested in the notebook? I would recommend you create a new virtual environment and run again. Please try again using the following codes, which I just tested and had no problem. I am runing on Windows Anaconda platform. conda create --name gnn activate gnn conda install tensorflow-gpu==1.15.0 nb_conda jupyter pip install graph_nets matplotlib scipy "tensorflow>=1.15,<2" "dm-sonnet<2" "tensorflow_probability<0.9" run the notebook

JinyongChung commented 2 years ago

Thank you for your help. I could solve the error, but still have problems. As addressed in the previous issue, I also have nan results for training and test loss. image

"zqaz999" said that it was a GPU acceleration problem, how can I check this?

clatfd commented 2 years ago

Sorry, I still cannot reproduce the nan error. Please contact zqaz999 to see how that was solved.