解决UnknownError: Failed to get convolution algorithm. This is probably because cuDNN failed to initia

使用Tensorflow2.0-GPU版本时调用卷积层报错 UnknownError: Failed to get convolution algorithm. This is probably because cuDNN failed to initialize, so try looking to see if a warning log message was printed above. [Op:Conv2D]

原因是使用单个GPU设备运行Tensorflow自动占满显存，导致cuDNN无法运行。

解决方案：限制tf的显存占用

在文件开头加入以下代码

gpus = tf.config.experimental.list_physical_devices('GPU')
if gpus:
    try:
    	#创建虚拟显卡
        tf.config.experimental.set_virtual_device_configuration(
            gpus[0],
            [tf.config.experimental.VirtualDeviceConfiguration(memory_limit=4096)])#限制量/MB，4096为4G
        logical_gpus = tf.config.experimental.list_logical_devices('GPU')
        print(len(gpus), "Physical GPUs,", len(logical_gpus), "Logical GPUs")
    except RuntimeError as e:
        print(e)

参考资料

---

https://blog.csdn.net/jackhh1/article/details/102864524