nnUNet icon indicating copy to clipboard operation
nnUNet copied to clipboard

inference error!

Open wade0604 opened this issue 1 year ago • 7 comments

70e81006454463834ac32de9d792a1f I am reasoning with the command !CUDA_VISIBLE_DEVICES=0 nnUNetv2_predict -i $INPUT_FOLDER -o $OUTPUT_FOLDER -d 1 -c 3d_lowres --save_probabilities -f 0 I get the following error: RuntimeError: Background workers died. Look for the error message further up! Use fewer workers or get more RAM in that case! But I checked the cpu and gpu utilization is 0

wade0604 avatar Jul 05 '23 06:07 wade0604

image

ancestor-mithril avatar Jul 05 '23 06:07 ancestor-mithril

@ancestor-mithril How to set MKL_SERVICEFORCE_INTEL? Please tell me in detail thanks!

wade0604 avatar Jul 05 '23 06:07 wade0604

https://stackoverflow.com/q/66932956

ancestor-mithril avatar Jul 05 '23 07:07 ancestor-mithril

@ancestor-mithril thank you very much! I update the version of numpy, and it's work now! But there is another error as follows: 1688542088856 How can I pre-process the images so that they have the same size? My image format is nii.gz.

wade0604 avatar Jul 05 '23 07:07 wade0604

You can use skimage.transform.resize, but I do not understand why it says "There are 1 cases in the source folder".

ancestor-mithril avatar Jul 05 '23 07:07 ancestor-mithril

@ancestor-mithril because i only use one-fold

wade0604 avatar Jul 05 '23 08:07 wade0604

Hi wade0604, How many cases do you have in that validation fold? It seems for me that nnUNet identifies 10 cases as 1 multi-channel data and that's why you get this error. Is it possible that your file naming is not correct?

Kobalt93 avatar Jan 24 '24 14:01 Kobalt93