experiencor / keras-yolo3

Training and Detecting Objects with YOLO3

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Training on single GPU doesn't work

skuehn1988 opened this issue · comments

I am now training on a VM but it only works on one CPU. When I enter GPU:0 in the config I thought it would train automatically on the GPU. But it works only on the CPU. Tensorflow GPU and CUDA are installed. On the Train File I tried some things, but without success.

My environment looks like this:
Tensorflow GPU: 1.15.0
Tensorflow Estimate: 1.15.1
Python: 3.5.5
Cuda: 11.0

The following CPUs and GPUs are available:
/device:GPU:0 but available devices are [ /job:localhost/replica:0/task:0/device:CPU:0, /job:localhost/replica:0/task:0/device:XLA_CPU:0, /job:localhost/replica:0/task:0/device:XLA_GPU:0 ]

How can I now point to the GPU device?

Try in config.json => "gpus": "0"