Using TensorFlow version 2.4.1 We will restrict TensorFlow to max 8GB GPU RAM then RAPIDS can use 8GB GPU RAM

import os
os.environ['CUDA_VISIBLE_DEVICES']='0'

import tensorflow as tf
import tensorflow.keras.backend as K
print('Using TensorFlow version',tf.__version__)

# RESTRICT TENSORFLOW TO 8GB OF GPU RAM
# SO THAT WE HAVE 8GB RAM FOR RAPIDS
LIMIT = 8
gpus = tf.config.experimental.list_physical_devices('GPU')
if gpus:
  try:
    tf.config.experimental.set_virtual_device_configuration(
        gpus[0],
        [tf.config.experimental.VirtualDeviceConfiguration(memory_limit=1024*LIMIT)])
    logical_gpus = tf.config.experimental.list_logical_devices('GPU')
  except RuntimeError as e:
    print(e)
print('We will restrict TensorFlow to max %iGB GPU RAM'%LIMIT)
print('then RAPIDS can use %iGB GPU RAM'%(16-LIMIT))
Restrict TensorFlow GPU Memory Usage for RAPIDS

原文地址: http://www.cveoy.top/t/topic/oFS8 著作权归作者所有。请勿转载和采集!

免费AI点我,无需注册和登录