SWAHR seems to need much time for training 300 epochs if using 1 GPU

greatlog / SWAHR-HumanPose

Bottom-up Human Pose Estimation

127 stars 20 forks source link

SWAHR seems to need much time for training 300 epochs if using 1 GPU #16

Open AlanLuSun opened 1 year ago

AlanLuSun commented 1 year ago

Hi authors,

When I train the source code with one GPU of TitanX which does not have distributed training, I calculate that the training for 300 epochs at least takes 30.5days. Do you have suggestions that make it faster? I would be very appreciate if you could response me.

Thanks~~