Why test_caltech.py runs so slow? Low gpu occupation but high gpu occupation？

[TOM-tym]

@liuwei16 Thanks for your excellent work! I am trying to run test_caltech.py but it seems that it's running too slow. I am using single GPU GTX 1080 Ti and modified test_caltech.py in line 11 to use it. Modified line is showed below.

os.environ["CUDA_VISIBLE_DEVICES"] = '0'

However, while running test_caltech.py , it just takes up about 150 Mb GPU memory but takes up almost all CPU resources. And it runs every slow. How can I fully use my GPU? Thanks for your attention.

[Boisa]

I would like to answer the question .Because you're not calling the GPU to run test_caltech.py you should to add the follow : C.gpu_ids='0' or other GPU number depend on your computer os.environ["CUDA_VISIBLE_DEVICES"] = C.gpu_ids so you must query for the correct GPU number Can you answer my question? It is ‘FPPI’

[TOM-tym]

@Boisa thanks for your comment. But I am a little bit confused. C.gpu_ids='0' os.environ["CUDA_VISIBLE_DEVICES"] = C.gpu_ids and os.environ["CUDA_VISIBLE_DEVICES"] = '0' Are there any difference between these two modifications?

[Boisa]

There is no any difference when you only have one GPU.You should check if your GPU number is ‘0’

[TOM-tym]

Thanks. I am pretty sure that my GPU numer is 0.