[FEATURE] how to reproduce ResNet baseline #1097
Description
great project.
hi, the code I run according to Training Examples:
- https://rwightman.github.io/pytorch-image-models/training_hparam_examples/
- ResNet50 with JSD loss and RandAugment (clean + 2x RA augs) - 79.04 top-1, 94.39 top-5
in the paper, ResNet strikes back: An improved training procedure in timm.
how can I reproduce the following training procedure?
