NVIDIA / caffe

Caffe: a fast open framework for deep learning.
http://caffe.berkeleyvision.org/
Other
672 stars 263 forks source link

How can I use NVCaffe container with IB? #590

Open liuchintao opened 3 years ago

liuchintao commented 3 years ago

I followed this doc and run a NVCaffe container with command docker run -it --cap-add=IPC_LOCK --device=/dev/infiniband/uverbs1 --network host --name=my-verbs-nonprvlg myofed451image bash.

Container started and work well, but when I try to use command ibv_devinfo -l, I got a error as below:

ibv_devinfo -l
libibverbs: Warning: no userspace device-specific driver found for /sys/class/infiniband_verbs/uverbs4
libibverbs: Warning: no userspace device-specific driver found for /sys/class/infiniband_verbs/uverbs3
libibverbs: Warning: no userspace device-specific driver found for /sys/class/infiniband_verbs/uverbs2
libibverbs: Warning: no userspace device-specific driver found for /sys/class/infiniband_verbs/uverbs1
libibverbs: Warning: no userspace device-specific driver found for /sys/class/infiniband_verbs/uverbs0
0 HCAs found:

I tried my best to resolve it, but it did not work.

I found that NVCaffe Container includes MLNX_OFED from release note.

So, what can I do to use IB devices in NVCaffe container?