search

experiencor / keras-yolo2

Easy training on custom dataset. Various backends (MobileNet and SqueezeNet) supported. A YOLO demo to detect raccoon run entirely in brower is accessible at https://git.io/vF7vI (not on Windows).
MIT License
1.73k stars 784 forks source link

How to run on GPU #416

Closed harshitazilen closed 5 years ago

harshitazilen commented 5 years ago

I tried configuring below parameters but it still does not use GPU. Please guide:

os.environ["CUDA_DEVICE_ORDER"] = "PCI:1:0:0"
os.environ["CUDA_VISIBLE_DEVICES"] = "0"

Thanks

rodrigo2019 commented 5 years ago

if you installed tensorflow correctly, it will use GPU by default

look4pritam commented 5 years ago

Install cuda 10 Install cudnn for cuda-10 Install tensorflow- GPU. Then it will work.

harshitazilen commented 5 years ago

Thanks. I tried by creating new virtual environment and re-configured everything and it is working now.