[Fix] update hubconf.py

[Bing-su]

2.0.0 출시를 축하드립니다.

Problem (Why?)

1. load schedulers

2. torch.hub.list에서 func, optimizer 감추기

>>> import torch

>>> torch.hub.list("kozistr/pytorch_optimizer")
...
 'adan',
 'adapnm',
 'diffgrad',
 'diffrgrad',
 'func', # <<<<<<
 'lamb',
 'lars',
 'madgrad',
 'nero',
 'optimizer', # <<<<<<
 'pnm',
 'radam',
 'ralamb',
 'ranger',
 'ranger21',
 'sgdp',
 'shampoo']

3. torch.load.help 타겟을 class 자체에서 __init__으로 변경

• before

>>> print(torch.hub.help("kozistr/pytorch_optimizer", "ranger"))

    Reference : https://github.com/lessw2020/Ranger-Deep-Learning-Optimizer
    Example :
        from pytorch_optimizer import Ranger
        ...
        model = YourModel()
        optimizer = Ranger(model.parameters())
        ...
        for input, output in data:
          optimizer.zero_grad()
          loss = loss_function(output, model(input))
          loss.backward()
          optimizer.step()

• after

>>> print(torch.hub.help("Bing-su/pytorch_optimizer:hubconf", "ranger"))

Ranger optimizer
        :param params: PARAMETERS. iterable of parameters to optimize or dicts defining parameter groups
        :param lr: float. learning rate
        :param betas: BETAS. coefficients used for computing running averages of gradient and the squared hessian trace
        :param weight_decay: float. weight decay (L2 penalty)
        :param n_sma_threshold: int. (recommended is 5)
        :param use_gc: bool. use Gradient Centralization (both convolution & fc layers)
        :param gc_conv_only: bool. use Gradient Centralization (only convolution layer)
        :param adamd_debias_term: bool. Only correct the denominator to avoid inflating step sizes early in training
        :param eps: float. term added to the denominator to improve numerical stability

Solution (What/How?)

거창한 설명에 비해 코드 수정을 별로 많지 않습니다. 아래 변경점을 확인해주세요.

Notes

3번 help 타겟을 .__init__으로 변경한건 사람마다 생각이 다를 것 같네요...

[codecov-commenter]

Codecov Report

Merging #81 (becf7cd) into main (2258885) will not change coverage.
The diff coverage is n/a.

@@           Coverage Diff           @@
##             main      #81   +/-   ##
=======================================
  Coverage   96.22%   96.22%           
=======================================
  Files          30       30           
  Lines        2171     2171           
=======================================
  Hits         2089     2089           
  Misses         82       82           

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

[kozistr]

LGTM! Thanks for your contribution :)

[kozistr]

저도 빙수님 생각처럼 class 자체 usage를 보여주는 것보단 parameter에 대한 설명을 보여주는 게 (init) 더 좋지 않을까 생각해요! (usage 자체는 직관적이고 README나 docs에서 확인할 수 있지(어야 하지) 않을까 생각해서요)

+) 제가 아직 class docstring를 어느 부분에 어떻게 쓸 지 고민을 많이 못해서 그런데, 더 고민해서 언젠가 바꿔볼게요!