GPU Utilization code

Use this code to manage GPU usage

• os.environ["CUDA_VISIBLE_DEVICES"]="0" or “1” to set GPU
• config.gpu_options.per_process_gpu_memory_fraction = 1. (0.1,0.2,…1.0) to set GPU usage. This will be help full in usage of GPU memory
• This code will make your program to wait until GPU gets free. Once GPU is free than automatically your program starts execution.

#================================================= import os import tensorflow as tf from GPUtil import getGPUs os.environ["CUDA_DEVICE_ORDER"]="PCI_BUS_ID"

The GPU id to use, usually either "0" or "1"

os.environ["CUDA_VISIBLE_DEVICES"]="0" gpunum = int(os.environ["CUDA_VISIBLE_DEVICES"])

Wait until GPU is free

while True: gpudata = getGPUs() if (gpudata[gpunum].memoryUtil <= 0.2): break #================================================= #SET Utilization GPU #================================================= from keras import backend as K config = tf.ConfigProto() config.gpu_options.per_process_gpu_memory_fraction = 1. # set % of usage of GPU session = tf.Session(config=config) K.set_session(session) #=================================================