multi_gpu_model error

[guiyuliu]

I have 8 gpus on my machine and they all work well when use nvidia-smi to list
but I got errors below

the keras issue is here
keras-team/keras#11644

it seems like it cannot be fixed unless I downgrade tensorflow to 1.12
but it should use CUDA9.0

I was completely confused why this work used a very high version of CUDA , but use a very low version tf1.13 ?
Do you have other options ?

training_model = multi_gpu_model(model, gpus=multi_gpu)
File "/root/anaconda3/envs/crfnet2/lib/python3.5/site-packages/keras/utils/multi_gpu_utils.py", line 181, in multi_gpu_model
available_devices))
ValueError: To call multi_gpu_model with gpus=8, we expect the following devices to be available: ['/cpu:0', '/gpu:0', '/gpu:1', '/gpu:2', '/gpu:3', '/gpu:4', '/gpu:5', '/gpu:6', '/gpu:7']. However this m
achine only has: ['/cpu:0', '/xla_gpu:0', '/xla_cpu:0', '/gpu:0']. Try reducing gpus.