The problem of time

[KeshuaiNiu]

When using Docker for runtime, the model is taking a very long time at the Bias correction (N4) stage. I am using an A6000 GPU. Is this situation normal?

[YueSun814]

Yes, it is very normal for the N4 bias correction step to be slow, especially when the image size is large (e.g., > 400 × 400 × 400), even with an A6000 GPU. N4 runs on the CPU rather than the GPU, so the GPU does not accelerate this stage.