Ezra-Yu

> ping @Ezra-Yu , can close? We can finish the PR today.

> docs is still off, I can't figure out how does the formatting works for now can you open this [link](https://mmclassification--935.org.readthedocs.build/zh_CN/935/api/generated/mmcls.models.RegNet.html#mmcls.models.RegNet) 

Thanks for your contribution and we appreciate it a lot. We will review it soon.

```GropuNorm(num_group=1)``` also acts as `LayerNorm2d`, which is used in [`PoolFormer`](https://github.com/open-mmlab/mmclassification/blob/71ef7bae854a9b85d83864f0e4af72c1e9fc7357/mmcls/models/backbones/poolformer.py#L286).

Your dataset is small, try to use a smaller batchsize. And use a smaller `log.interval` to observe the losses: ``` log_config = dict(interval=1, hooks=[dict(type='TextLoggerHook')]) ```

Thanks for your suggestion, MobileOne is in our plan.

@dsp6414 We have supported the MobileOne backbone in #966. Please try it. That PR will be merged into the master branch in nearly future.

> Hope to add TrivialAugmentation，https://github.com/automl/trivialaugment Yes， this is in our latest plan.

> Is there any plan for ArcFace loss? Yes, We will add it in the near future.

It looks like there is an error with the transforms. Some your image shapes are smaller than (32, 32) Try to Resize your images bigger than (32*32) before `RandomCrop`, eg....