After using the new version, after CUDA10.0 and CUDA10.1 are installed at the same time, now, suddenly CUDA10.0 can't be used, no, it should be said that its corresponding tf-GPU can't be used. …
System: win10
tf version: 2.0.0
tf.config.experimental.list_physical_devices('GPU') returns as follows:
[Why is there only one graphics card??? My computer has a discrete graphics card, and the environment is all set up,...]
[PhysicalDevice(name='/physical_device:GPU:0', device_type='GPU')]
This tf version supports GPU:
print(tf.__version__)
print(tf.test.is_gpu_available())
print(tf.test.is_built_with_cuda())#是否支持NVIdia GPU
But it returns as follows:
2.0.0
True
True
tf.test.is_gpu_available() shows True but tf.config.experimental.list_physical_devices('GPU') does not have NVIDIA graphics card. . .
And the nvidia-smi command line is displayed as follows:
please comment from someone who comes over, thank you God.
[Already resolved]