DeepHyperX
DeepHyperX copied to clipboard
Parallelize data loading
Currently, the torch DataLoader uses blocking data loading. Although loading is very fast (we store the NumPy arrays in-memory), transfer to GPU and data augmentation (which is done on CPU) can slow things done.
Using workers > 0 would make data loading asynchronous and workers > 1 could increase speed somewhat.
TODO:
- [ ] Benchmark speed gain using asynchronous data loading
- [x] Implement asynchronous data loading for all DataLoader objects
- [x] Add a user-input option to define the number of jobs