How do I force training on GPU?
Currently it's only using CPU even when I run dlib.DLIB_USE_CUDA and it says true.
It also says 1 when I run print(dlib.cuda.get_num_devices())
Here's the attached image that shows that there's nothing running on GPU when in fact I am running the code:
NOTE: GPU was set as RUn
Comment:
Apparently as what I've tested this wasn't a training error but rather it is loading error. It takes so much time and ram to load ibug-300W files. Is there any way to load this faster?
If someone ever stumbled upon this issue or problem on google colab (Slow training time).
The way to load this faster is to transfer the dataset directly on the vm/content of colab. Because the transfer speed between Drive and Colab is slow.
PS: You need atleast 14-15GB of ram to load ibug-300W files.
