* host some of Cadene's weights on github instead of .fr for speed
* add my old port of ensemble adversarial inception resnet v2
* switch to my TF port of normal inception res v2 and change FC layer back to 'classif' for compat with ens_adv
* Support PyTorch native DDP as fallback if APEX not present
* Support SyncBN for both APEX and Torch native (if torch >= 1.1)
* EMA model does not appear to need DDP wrapper, no gradients, updated from wrapped model
* ModelEma class added to track an EMA set of weights for the model being trained
* EMA handling added to train, validation and clean_checkpoint scripts
* Add multi checkpoint or multi-model validation support to validate.py
* Add syncbn option (APEX) to train script for experimentation
* Cleanup interface of CheckpointSaver while adding ema functionality
* B0-B3 weights ported from TF with close to paper accuracy
* Renamed gen_mobilenet to gen_efficientnet since scaling params go well beyond 'mobile' specific
* Add Tensorflow preprocessing option for closer images to source repo
* Do mixup in custom collate fn if prefetcher enabled, reduces performance impact
* Move mixup code to own file
* Add arg to disable prefetcher
* Fix no cuda transfer when prefetcher off
* Random erasing when prefetcher off wasn't changed to match new args, fixed
* Default random erasing to off (prob = 0.) for train
* factor out data related constants to own file
* move data related config helpers to own file
* add a variant of RandomResizeCrop that randomizes interpolation method
* remove old Numpy version of RandomErasing
* cleanup torch version of RandomErasing and use it in either GPU loader batch mode or single image cpu Transform
* init square_avg with one instead of zero as per TF
* match TF order of ops for square_avg accumulation
* move LR scaling to momentum buffer accumulator as per TF
* add decoupled weight decay flag (not in TF)
Ross Wightman