Joshua Z. Zhang

Yes, I agree that it’s not a good practice to add an option to disable warning. However in this particular case, if user acknowledged that they understand the risk of...

I need more opinions, my personal feeling is that is much convenient to use the switch to turn off the warnings

I would say it might be extra burden for users who don't know exactly what layers are inited or uninited and it may not be feasible to control all of...

This change will break all legacy build systems and documents with USE_CUDA=1 enabled. I suggest keep USE_CUDA and add USE_GPU as a dispatcher to turn on either USE_CUDA or USE_HIP...

Sounds like a remotely served tensorboard with fancy UI. I am upvoting for it.

- The default `from_folders` expects the train/val/test root folders to be pre-split into subdirectories similar to this example here: https://auto.gluon.ai/stable/tutorials/image_prediction/dataset.html#load-an-image-directory - If the structure of the folders are not designed...

How did you get GPU to work using this repo? I think the makefile don't support GPU yet.

Can you profile the runtime?

model is inconsistant with the new version, you have to use the latest model I think

see #6