search

Paper99 / SRFBN_CVPR19

Pytorch code for our paper "Feedback Network for Image Super-Resolution" (CVPR2019)
MIT License
551 stars 126 forks source link

the meaning of "split_batch"? #28

Open Grassyue opened 5 years ago

Grassyue commented 5 years ago

Hi, thank you for your sharing and sorry to bother you, i am a new guy in super-resolution, can you explain the meaning of "split_batch" in the code? thank you sooo much!

Paper99 commented 5 years ago

Each batch will first split into N (depending on the value of split_batch) sub-batches before feeding to the network. The larger split_batch will lead to less GPU memory usage during training.

Grassyue commented 5 years ago

Thanks for your reply，but why not directly set the batch size to a small number? what is the difference?

Paper99 commented 5 years ago

When you set a large batchsize, the option split_batch make it possible to update the parameters only once.

DamperHa commented 5 years ago

When I use the option split_batch, the accuracy was much worse on the test set, what is the reason for this?