docker swarm with gpu only one replica can get the gpu core

I have tried docker swarm and create service with gpu:
after setting the deamon.josn and config.toml, I successfully create the service with gpu, but only one replica can get the gpu core,the rest of them are held pending. Is there any way can make all replicas share the one gpu core??thanks!!!!

[email protected]:~/camera$ docker service create --generic-resource "gpu=1" 
--replicas 2 --name tt2 nvidia/cuda:9.0-cudnn7-devel-ubuntu16.04 nvidia-smi -l
overall progress: 1 out of 2 tasks 
1/2: no suitable node (insufficient resources on 1 node) 
2/2: running   [==================================================>] 
Operation continuing in background.
Use `docker service ps pcrsjk569yj8pt0n1cpfsjgsj` to check progress.
[email protected]:~/camera$ docker service ls
ID                  NAME                MODE                REPLICAS        
IMAGE                                      PORTS
pcrsjk569yj8        tt2                 replicated          1/2                 
[email protected]:~/camera$ docker service ps tt2
ID                  NAME                IMAGE                                      
NODE                DESIRED STATE       CURRENT STATE            ERROR                              
0cmg2mewjy66        tt2.1               nvidia/cuda:9.0-cudnn7-devel- 
ubuntu16.04         Running             Pending 22 seconds ago   "no suitable node (insufficien…"   
pnrr7p4si9j7        tt2.2               nvidia/cuda:9.0-cudnn7-devel- 
ubuntu16.04         Running             Running 19 seconds ago                 

Source: StackOverflow