PyTorch DataLoaders about bdl-benchmarks HOT 3 CLOSED

Hi @wjmaddox,

NumPy data loaders are already provided, which are framework-agnostic. Take a look at this colab, where a simple PyTorch model is trained using them!

As far as the other benchmarks go:

• Autonomous Vehicle's Scene Segmentation will be soon made public,
• Galaxy Zoo and Fishyscapes are owned by collaborators and due to this dependency, we cannot really guarantee anything.

from bdl-benchmarks.

Gotcha, thanks for the response! I'll check out that Colab notebook.

Additionally, is the suggested mechanism for using data augmentation in Pytorch to 1) use the numpy dataloaders 2) convert to pytorch and 3) perform data augmentation using torchvision on the pytorch tensors?

from bdl-benchmarks.

Hi @Filangel,

Thanks for the library! I was also wondering if you planned to implement / open source pytorch support? It would be very helpful to have it for things like data augmentation / data shuffling as noted by @wjmaddox. Or would you have any suggestions on what would be the best way to wrap it in pytorch dataloader?

Thank you!

from bdl-benchmarks.