multiple GPU mode

[DashingPanda]

I really appreciate the project. Unfortunately, I'm restricted by CUDA memory and limited skills. Could you provide files about the training and inferencing in multiple GPU modes?

[apavani2]

while training the model, you can set the num_gpus to the number of gpus you have. Pytorch will automatically do the memory allocation for you

For example,
python train.py --config ./configs/zju.json --data_root ./data/zju_mocap --num_gpus 4