Can we use multiple GPU on inference?

[tamyxgopenx]

🚀 Feature

Prediction using multiple GPU without

Motivation

I want to speed up the prediction time with more GPUs

Pitch

It should be easier if you can use config for this, like config.CPU_DEVICE_COUNT = 2
Is there anyway to do it? i have seen python tools/train_net.py --num-gpus 4 --config-file configs/modanet.yaml but i guess it for training, i only want to predict frame.

[ppwwyyxx]

See #716, #586, #33, etc.

[tamyxgopenx]

@ppwwyyxx thanks for your answer but is there a function that make gpu parallel computing one frame only?

[ppwwyyxx]

there isn't