Training time

Open TachyonicSpace opened this issue 6 years ago • 2 comments

i am having a sample size that is signifigantly smaller than the ideal sample soze for my final project, and am having 5 days to finish the training with this smaller data set, is there any way to speed up training time with multithreading or something else of the like?

Oct 16 '19 04:10 TachyonicSpace

Are you using TensorFlow-GPU for the ImageAI backend, or are you using regular TensorFlow?

Nov 19 '19 02:11 cody-jarrett

depending if you run code on CPU or GPU : i used 1800 images, on CPU it takes 10 days to finish 30 epochs batchsize 8 , with GPU, it takes 8 hours to finish 100 epochs batchsize 16

May 21 '23 13:05 Amine1309