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.
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 https://github.com/keras-team/keras/issues/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
withgpus=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 reducinggpus
.