MobileOne

class mmpretrain.models.backbones.MobileOne(arch, in_channels=3, out_indices=(3,), frozen_stages=-1, conv_cfg=None, norm_cfg={'type': 'BN'}, act_cfg={'type': 'ReLU'}, se_cfg={'ratio': 16}, deploy=False, norm_eval=False, init_cfg=[{'type': 'Kaiming', 'layer': ['Conv2d']}, {'type': 'Constant', 'val': 1, 'layer': ['_BatchNorm']}])

MobileOne backbone.

A PyTorch impl of : An Improved One millisecond Mobile Backbone

参数:

• arch (str | dict) –

  MobileOne architecture. If use string, choose from ‘s0’, ‘s1’, ‘s2’, ‘s3’ and ‘s4’. If use dict, it should have below keys:

  • num_blocks (Sequence[int]): Number of blocks in each stage.

  • width_factor (Sequence[float]): Width factor in each stage.

  • num_conv_branches (Sequence[int]): Number of conv branches in each stage.

  • num_se_blocks (Sequence[int]): Number of SE layers in each stage, all the SE layers are placed in the subsequent order in each stage.

  Defaults to ‘s0’.

• in_channels (int) – Number of input image channels. Default: 3.

• out_indices (Sequence[int] | int) – Output from which stages. Defaults to (3, ).

• frozen_stages (int) – Stages to be frozen (all param fixed). -1 means not freezing any parameters. Defaults to -1.

• conv_cfg (dict | None) – The config dict for conv layers. Defaults to None.

• norm_cfg (dict) – The config dict for norm layers. Defaults to dict(type='BN').

• act_cfg (dict) – Config dict for activation layer. Defaults to dict(type='ReLU').

• deploy (bool) – Whether to switch the model structure to deployment mode. Defaults to False.

• norm_eval (bool) – Whether to set norm layers to eval mode, namely, freeze running stats (mean and var). Note: Effect on Batch Norm and its variants only. Defaults to False.

• init_cfg (dict or list[dict], optional) – Initialization config dict.

示例

>>> from mmpretrain.models import MobileOne
>>> import torch
>>> x = torch.rand(1, 3, 224, 224)
>>> model = MobileOne("s0", out_indices=(0, 1, 2, 3))
>>> model.eval()
>>> outputs = model(x)
>>> for out in outputs:
...     print(tuple(out.shape))
(1, 48, 56, 56)
(1, 128, 28, 28)
(1, 256, 14, 14)
(1, 1024, 7, 7)

switch_to_deploy()

switch the model to deploy mode, which has smaller amount of parameters and calculations.

train(mode=True)

switch the mobile to train mode or not.