TensorFlow使用GPU（2）代码设置

1. 2. 问题

2.1 求助：TensorFlow指定使用GPU2,3却自动占用GPU0

问题：使用os.environ["CUDA_VISIBLE_DEVICES"] = "1"设置使用GPU1，却总是使用GPU0；，问题出在print("GPU状态:",tf.test.is_gpu_available())代码上，在设置使用哪块GPU之前，不能调用tensorflow的函数。

import os
# 用于保存训练后的检查点文件和日志文件路径
train_log_file = 'miniImageNet-better-cs.ckpt'
print("GPU状态:",tf.test.is_gpu_available())
os.environ["CUDA_DEVICE_ORDER"] = "PCI_BUS_ID"  #有效，但是运行结果不对
os.environ["CUDA_VISIBLE_DEVICES"] = gpu_flag
#os.environ["CUDA_VISIBLE_DEVICES"] = "1"

修改前的运行log:

2020-07-27 12:18:22.763620: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1471] Adding visible gpu devices: 0, 1, 2, 3, 4, 5, 6, 7
2020-07-27 12:18:22.772153: I tensorflow/core/common_runtime/gpu/gpu_device.cc:952] Device interconnect StreamExecutor with strength 1 edge matrix:
2020-07-27 12:18:22.772171: I tensorflow/core/common_runtime/gpu/gpu_device.cc:958]      0 1 2 3 4 5 6 7 
2020-07-27 12:18:22.772181: I tensorflow/core/common_runtime/gpu/gpu_device.cc:971] 0:   N N N N N N N N 
2020-07-27 12:18:22.772188: I tensorflow/core/common_runtime/gpu/gpu_device.cc:971] 1:   N N N N N N N N 
2020-07-27 12:18:22.772195: I tensorflow/core/common_runtime/gpu/gpu_device.cc:971] 2:   N N N N N N N N 
2020-07-27 12:18:22.772202: I tensorflow/core/common_runtime/gpu/gpu_device.cc:971] 3:   N N N N N N N N 
2020-07-27 12:18:22.772209: I tensorflow/core/common_runtime/gpu/gpu_device.cc:971] 4:   N N N N N N N N 
2020-07-27 12:18:22.772216: I tensorflow/core/common_runtime/gpu/gpu_device.cc:971] 5:   N N N N N N N N 
2020-07-27 12:18:22.772222: I tensorflow/core/common_runtime/gpu/gpu_device.cc:971] 6:   N N N N N N N N

修改如下，把tf.test.is_gpu_available()语句挪后即可：

import os
# 用于保存训练后的检查点文件和日志文件路径
train_log_file = 'miniImageNet-better-cs.ckpt'
print("GPU状态:",tf.test.is_gpu_available())
os.environ["CUDA_DEVICE_ORDER"] = "PCI_BUS_ID"  #有效，但是运行结果不对
os.environ["CUDA_VISIBLE_DEVICES"] = gpu_flag
#os.environ["CUDA_VISIBLE_DEVICES"] = "1"

修改后的运行log:

2020-07-27 12:25:40.070721: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1471] Adding visible gpu devices: 0
2020-07-27 12:25:40.070792: I tensorflow/core/common_runtime/gpu/gpu_device.cc:952] Device interconnect StreamExecutor with strength 1 edge matrix:
2020-07-27 12:25:40.070801: I tensorflow/core/common_runtime/gpu/gpu_device.cc:958]      0 
2020-07-27 12:25:40.070810: I tensorflow/core/common_runtime/gpu/gpu_device.cc:971] 0:   N 
2020-07-27 12:25:40.070915: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1084] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 10238 MB memory) -> physical GPU (device: 0, name: GeForce RTX 2080 Ti, pci bus id: 0000:88:00.0, compute capability: 7.5)
2020-07-27 12:25:42.117883: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1471] Adding visible gpu devices: 0
2020-07-27 12:25:42.117947: I tensorflow/core/common_runtime/gpu/gpu_device.cc:952] Device interconnect StreamExecutor with strength 1 edge matrix:
2020-07-27 12:25:42.117956: I tensorflow/core/common_runtime/gpu/gpu_device.cc:958]      0 
2020-07-27 12:25:42.117964: I tensorflow/core/common_runtime/gpu/gpu_device.cc:971] 0:   N 
2020-07-27 12:25:42.118029: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1084] Created TensorFlow device (/device:GPU:0 with 10238 MB memory) -> physical GPU (device: 0, name: GeForce RTX 2080 Ti, pci bus id: 0000:88:00.0, compute capability: 7.5)

参考资料

[1] tensorflow运行在gpu还是cpu
[2] tensorflow训练使用GPU和CPU的不同指定方法
[3] tensorflow指定使用哪块GPU运行程序
[4] 求助：TensorFlow指定使用GPU2,3却自动占用GPU0