A error occoured when using slurm

I used srun to train example/pytorch_imagenet_resnet50.py on 8 GUPs and got this error:

$ srun --mpi=pmi2 -p Platform -n1 --gres=gpu:8 --ntasks-per-node=8 horovodrun -np 8 -H localhost:8 python pytorch_imagenet_resnet50.py                                                        
-------------------------------------------------------------------------- 
There are not enough slots available in the system to satisfy the 8 slots
that were requested by the application:
  python

Either request fewer slots for your application, or make more slots available
for use.
--------------------------------------------------------------------------

But it can work on 1 GPU with command

srun --mpi=pmi2 -p Platform -n1 --gres=gpu:1 --ntasks-per-node=1 horovodrun -np 1 -H localhost:1 python pytorch_imagenet_resnet50.py

Start with the reported srun/horovodrun commands and the example/pytorch_imagenet_resnet50.py training entry point, comparing the one-GPU and eight-GPU cases. Check the project's guidance for Slurm and distributed execution, and confirm whether the reported slot error is within Apex's scope. Done when the cause and supported multi-GPU invocation are established or the issue is redirected appropriately.