o
V h% @ s d dl mZ d dlmZmZmZmZ d dlZd dlmZm Z ddl
m Z ddl m
Z
ddlmZ d d
lmZmZmZ d d lmZ d d lmZmZmZ g d
ZG dd dejZ G dd dejZ ddedZ G dd deZ e ede j fdddddee d e!d ed e fd d Z"dS )! )partial)AnyCallableListOptionalN)nnTensor )Conv2dNormActivation)ImageClassification)_log_api_usage_once )register_modelWeights WeightsEnum)_IMAGENET_CATEGORIES)_make_divisible_ovewrite_named_paramhandle_legacy_interface) MobileNetV2MobileNet_V2_Weights mobilenet_v2c sZ e Zd Z ddededededeedejf ddf fd d
Zd e de fd d
Z
Z S )InvertedResidualNinpoupstride expand_ratio
norm_layer.returnc s t || _|dvrtd| |d u rtj}tt|| }| jdko*||k| _g }|dkr?| t
||d|tj d | t
|||||tj dtj
||ddddd||g tj| | _|| _|dk| _d S ) N)r
r z#stride should be 1 or 2 instead of r
kernel_sizer activation_layer)r groupsr r! r F)bias)super__init__r
ValueErrorr BatchNorm2dintrounduse_res_connectappendr
ReLU6extendConv2d
Sequentialconv out_channels_is_cn)selfr r r r r
hidden_dimlayers __class__ R/var/www/vscode/kcb/lib/python3.10/site-packages/torchvision/models/mobilenetv2.pyr% s:
zInvertedResidual.__init__xc C s | j r
|| | S | |S N)r* r0 r3 r: r8 r8 r9 forward< s
zInvertedResidual.forwardr; ) __name__
__module__ __qualname__r( r r r Moduler% r r=
__classcell__r8 r8 r6 r9 r s (r c s e Zd Z ddededeeee d ed
eed ej f d eed ej f d
eddf fdd
Z
de de fddZ de de fddZ
ZS )r ?N 皙? num_classes
width_multinverted_residual_setting
round_nearestblock.r dropoutr c s t t| |du rt}|du rtj}d}d} |du r5g dg dg dg dg dg d g d
g}t|d ksCt|d d krJtd
| t|| |}t| t d| || _
t d|d|tj dg}
|D ]*\} } }
}t| | |}t
|
D ]}|d kr|nd}|
||||| |d |}qzqi|
t || j
d|tj d tj|
| _ttj|dt| j
|| _| D ]M}t|tjrtjj|jdd |jdurtj|j qt|tjtj frtj |j tj|j qt|tjr tj |jd d tj|j qdS )aw
MobileNet V2 main class
Args:
num_classes (int): Number of classes
width_mult (float): Width multiplier - adjusts number of channels in each layer by this amount
inverted_residual_setting: Network structure
round_nearest (int): Round the number of channels in each layer to be a multiple of this number
Set to 1 to turn off rounding
block: Module specifying inverted residual building block for mobilenet
norm_layer: Module specifying the normalization layer to use
dropout (float): The droupout probability
N i )r
r
r
) r r )rO rM r )rO @ r )rO ` rQ r
)rO rQ r )rO i@ r
r
r rS zGinverted_residual_setting should be non-empty or a 4-element list, got rD rQ r )r r r! r
)r r r )pfan_out)modeg{Gz?) r$ r% r r r r' lenr& r max last_channelr
r, ranger+ r/ featuresDropoutLinear
classifiermodules
isinstancer. initkaiming_normal_weightr# zeros_ GroupNormones_normal_)r3 rG rH rI rJ rK r rL
input_channelr[ r] tcnsoutput_channelir mr6 r8 r9 r% D sp
zMobileNetV2.__init__r: c C s2 | |}tj|d}t|d}| |}|S )Nr
r
r
)r] r
functionaladaptive_avg_pool2dtorchflattenr` r< r8 r8 r9
_forward_impl s
zMobileNetV2._forward_implc C s
| |S r; )rw r< r8 r8 r9 r= s
zMobileNetV2.forward)rC rD NrE NNrF )r> r? r@ r( floatr r r r rA r% r rw r= rB r8 r8 r6 r9 r C s6 _
r iz5 rr )
num_paramsmin_size
categoriesc
@ sz e Zd Zedeeddi edddddid d
d d d
Zedeedddi edddddid ddd d
ZeZ dS )r z=https://download.pytorch.org/models/mobilenet_v2-b0353104.pth ) crop_sizezQhttps://github.com/pytorch/vision/tree/main/references/classification#mobilenetv2z ImageNet-1Kgx&1Q@gMV@)zacc@1zacc@5g$C?g\( +@zXThese weights reproduce closely the results of the paper using a simple training recipe.)recipe_metrics_ops
_file_size_docs)url
transformsmetaz=https://download.pytorch.org/models/mobilenet_v2-7ebf99e0.pth )r} resize_sizezHhttps://github.com/pytorch/vision/issues/3995#new-recipe-with-reg-tuningg`" R@gS㥛V@gV-2+@a$
These weights improve upon the results of the original paper by using a modified version of TorchVision's
`new training recipe
`_.
N)
r> r? r@ r r r _COMMON_META
IMAGENET1K_V1
IMAGENET1K_V2DEFAULTr8 r8 r8 r9 r sH
r
pretrained)weightsT)r progressr r kwargsr c K sV t | } | durt|dt| jd tdi |}| dur)|| j|dd |S )a MobileNetV2 architecture from the `MobileNetV2: Inverted Residuals and Linear
Bottlenecks `_ paper.
Args:
weights (:class:`~torchvision.models.MobileNet_V2_Weights`, optional): The
pretrained weights to use. See
:class:`~torchvision.models.MobileNet_V2_Weights` below for
more details, and possible values. By default, no pre-trained
weights are used.
progress (bool, optional): If True, displays a progress bar of the
download to stderr. Default is True.
**kwargs: parameters passed to the ``torchvision.models.mobilenetv2.MobileNetV2``
base class. Please refer to the `source code
`_
for more details about this class.
.. autoclass:: torchvision.models.MobileNet_V2_Weights
:members:
NrG r{ T)r
check_hashr8 )r verifyr rY r r load_state_dictget_state_dict)r r r modelr8 r8 r9 r s
r )# functoolsr typingr r r r ru r r ops.miscr
transforms._presetsr utilsr _apir r r _metar _utilsr r r __all__rA r r r r r boolr r8 r8 r8 r9 s< 0o*