How improve the speed of training after using Transforms.Resize()?

When I used Transforms.Resize(), I found the training of the model gets slow. I created an issue in Github about it, someone gave me an answer that Resize() takes CPU resource. So I want to know how to increase the speed of training or make Resize() to use GPU resources.

You could use e.g. DALI for image processing on the GPU.
Depending on your system, this could yield a speedup.
Alternatively, you could use PIL-SIMD.

PS: I would recommend to stick to one thread, as this question is highly related to this one.