how use gpu to infer and generate mask readerer data

In ./recongnition/_tools/mask_render.py, II executed `CUDA_VISIBLE_DEVICES=0 python mask_renderer.py t`he utilization rate of GPU is 0, but the CPU is used to complete the data generation. How to use gpu to realize the generation of face mask data
In the readme document, the following method is to use gpu. I tried it, but to no avail。
`tool.prepare(ctx_id=0, det_size=(128,128)) #use gpu`

Start by inspecting ./recongnition/_tools/mask_render.py and the README example using tool.prepare(ctx_id=0, det_size=(128,128)). Reproduce the reported CUDA_VISIBLE_DEVICES=0 command, accounting for the mask_render.py versus mask_renderer.py naming shown in the issue, and verify that face-mask data generation uses the GPU rather than the CPU.

Written by the indexing model from the issue text.