submit failed

[orange-111222]

When I test directly with Try-out Algorithm it succeeds, but when I submit on submit page it reports the following error:
numpy.core._exceptions._ArrayMemoryError: Unable to allocate 16.6 GiB for an array with shape (49, 271, 410, 410) and data type float64

[orange-111222]

[LucaLumetti]

Hi @orange-111222,

Thanks for reaching out. I'm not familiar with the specific details behind GrandChallenge or the environment used by the "Try-out Algorithm" to run the provided Docker. However, here are a few suggestions that might help:

• You can try increasing the required CPU memory during the Docker upload on your algorithm page. I am not aware if there are any constraints on the maximum memory that you can set.

• Your output currently consists of 49 channels, while we only use 42 labels. If the 7 additional channels are not essential for your algorithm (e.g., for post-processing), consider removing them.

• If you are processing the input volume patch-wise (and as it seems that your algorithm is based on nnUNet, probably it is), using float32 instead of float64 can significantly reduce memory usage without compromising performance in most cases.

Best regards,
Luca Lumetti

[orange-111222]

Dear Luca Lumetti,

• I've used up to 32 gigs of CPU memory so far.

• Since I'm currently using a dataset I downloaded around May, I'll try to download the lasted dataset for training.

• yes，my algorithm was built based on nn-Unet, and I noticed that nn-Unet itself suffers from a large amount of RAM used for inference, so it's possible that nn-Unet is not suitable for ToothFairy.

• Thank you and all the best.

Best regards,
ChiZhang