“实战深度学习”-缺陷：ran out of memory错误解决办法

---

问题描述

tensorflow报以下错误：Allocator (GPU_0_bfc) ran out of memory

详细错误代码：

Allocator (GPU_0_bfc) ran out of memory trying to allocate 200.00MiB (rounded to 209715200).  Current allocation summary follows.

Resource exhausted: OOM when allocating tensor with shape[51200,1024] and type float on /job:localhost/replica:0/task:0/device:GPU:0 by allocator GPU_0_bfc.

tensorflow.python.framework.errors_impl.ResourceExhaustedError: 2 root error(s) found.
  (0) Resource exhausted: OOM when allocating tensor with shape[51200,1024] and type

---

原因分析：

查看此时GPU的使用情况，如下图：

由上图可知，主要原因是显存不足

---

解决方案：

1. 限制gpu的最大增长容量

import tensorflow as tf
#限制消耗固定大小的显存（程序不会超出限定的显存大小，若超出的报错）。
physical_gpus = tf.config.list_physical_devices("GPU")
tf.config.experimental.set_virtual_device_configuration(
    physical_gpus[0], #对0号gpu进行限制
    [tf.config.experimental.VirtualDeviceConfiguration(memory_limit=5000)] 
)
logical_gpus = tf.config.list_logical_devices("GPU")

2. 设置较小的batch size

3. 买更大的显卡

4. 调整代码在Data这部分使用cpu