How to speed up model Inference?

[whuhx]

Description

In my experiment, it takes 49 hours to infer an image with a size of 50000×50000. How to speed up?

sam. generate(
image, mask, batch=True, foreground=True, erosion_kernel=None, mask_multiplier=255
)

[mishagrol]

@whuhx hello! Do you utilize CPU or GPU ?

You may devide your initial image into blocks and run parallel

[giswqs]

It should not take that long with a GPU. You will need a large GPU to handle large images. Try out the SamGeo batch mode for large images.

[satas12825]

I have tried to run the model with GPU on docker nothing error but when I monitor the usage of GPU they seem like no one is using GPU, how can I know if my model using GPU to calculate?