Sweep code for studying model population stats

This is a major update and introduces powerful new functionality to pycls.

The pycls codebase now provides powerful support for studying design spaces and more generally population statistics of models as introduced in On Network Design Spaces for Visual Recognition and Designing Network Design Spaces. This idea is that instead of planning a single pycls job (e.g., testing a specific model configuration), one can study the behavior of an entire population of models. This allows for quite powerful and succinct experimental design, and elevates the study of individual model behavior to the study of the behavior of model populations. Please see SWEEP_INFO for details.

This code was co-authored by Piotr Dollar (@pdollar) and Raj Prateek Kosaraju (@rajprateek).

Initial release of Model Zoo

We have added a large set of baseline results and pretrained models available for download in the pycls Model Zoo; including the simple, fast, and effective RegNet models that we hope can serve as solid baselines across a wide range of flop regimes.

New features included in this release:

• Cache model weight URLs provided in configs locally | 4e470e2
• Allow optional weight decay fine-tuning for BN params, changes the default BN weight decay | 4e470e2
• Support Squeeze & Excitation in AnyNet | ec18863
• RegNet model abstraction | 4f5b5da, 708d429, 24a2805
• Run precise time in test_net and count acts for a model | ad01b81

Other changes:

• Updated README | 4acac2b
• New Model Zoo | 4acac2b, 4f5b5da
• New configs for models from Designing Network Design Spaces paper | 44bcfa7, e9c1ef4