pydarknet.is_compiled_with_gpu returns false

[Harsharma2308]

I have followed the steps to run with gpu and opencv. But on importing pydarknet,
Both pydarknet.is_compiled_with_gpu and pydarknet.is_compiled_with_opencv return False

[frfrrfrrr]

Same thing for me.
And pydarknet.set_cuda_device(0) returns:
module 'pydarknet' has no attribute 'set_cuda_device'

[dquig]

I was running into the same issue trying to install from source with a custom darknet installation. It looks like the GPU environment variables is ignored if DARKNET_HOME is also set. T

YOLO3-4-Py/setup.py

Line 28 in 062de22

if "GPU" in os.environ:

if "GPU" in os.environ:
    if "DARKNET_HOME" in os.environ:
        logging.warning("GPU environment variable is skipped since DARKNET_HOME is specified")
    else:
        if int(os.environ["GPU"]) == 1:
            logging.info("Darknet will be compiled with GPU support")
            USE_GPU = True
        else:
            logging.info("Darknet will be compiled without GPU support")
            USE_GPU = False

This causes pydarknet to be compiled without the ability to set cuda device.

YOLO3-4-Py/pydarknet.pxd

Line 63 in 062de22

IF USE_GPU == 1:

IF USE_GPU == 1:
--
64 | void cuda_set_device(int n)

It looks to me like the USE_GPU variable should still be set in the case that DARKNE_HOME is also set.

[madhawav]

@dquig Thank you for showing the cause of issue. I will go through the setup script again and correct it.

[kartikpaigwar]

I am getting the same thing. Have you fixed that.
pydarknet.set_cuda_device(0) returns:
module 'pydarknet' has no attribute 'set_cuda_device'

[fspider]

Hello. I am going to run several processes.. One on Gpu 1 and another one on GPU 2.
So i am going to use set_gpu_device() function.
But it is showing this error.

pydarknet.set_cuda_device(1)
AttributeError: module 'pydarknet' has no attribute 'set_cuda_device'

How can i do this?

[madhawav]

Fixed issue of ignoring GPU environment variable when custom DARKNET_HOME is set.
Refer #118