U
ʹj% @ 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 d
ddgZG dd dejZ G dd
d
ejZ ddedZ G dd deZ e ede j fddddee e!ee dd dZ"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 sL e Zd Zd eeeeeedejf dd fdd
Ze e dddZ
Z S )
InvertedResidualN.)inpoupstride expand_ratio
norm_layerreturnc s t || _|dkr&td| |d kr4tj}tt|| }| jdkoT||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
Z kernel_sizer activation_layer)r groupsr r r F)bias)super__init__r
ValueErrorr BatchNorm2dintrounduse_res_connectappendr
ReLU6extendConv2d
SequentialconvZ out_channelsZ_is_cn)selfr r r r r Z
hidden_dimZlayers __class__ ]/var/www/html/bwave/python/venv/lib/python3.8/site-packages/torchvision/models/mobilenetv2.pyr$ s:
zInvertedResidual.__init__xr c C s" | j r|| | S | |S d S N)r) r/ r0 r6 r3 r3 r4 forward< s zInvertedResidual.forward)N) __name__
__module__ __qualname__r' r r r Moduler$ r r9
__classcell__r3 r3 r1 r4 r s (r c
sz e Zd Zdeeeeee eeedej f eedej f edd fdd
Z
e e d
d d Z e e d
d
dZ
ZS )r ?N 皙?.) num_classes
width_multinverted_residual_setting
round_nearestblockr dropoutr c s@ t t| |dkr t}|dkr,tj}d}d} |dkrddddgddddgddd dgdd
d dgdd d dgdd
d dgddddgg}t|dkst|d d krtd| t|| |}t| t d| || _
t d |d|tj dg}
|D ]X\} } }
}t| | |}t
|
D ]4}|dkr"|nd}|
||||| |d |}qq|
t || j
d|tj d tj|
| _ttj|dt| j
|| _| D ]}t|tjrtjj|jdd |jdk r8tj|j n`t|tjtj fr
tj |j tj|j n.t|tjrtj |jdd 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 @ ` i@ r zGinverted_residual_setting should be non-empty or a 4-element list, got r@ )r r r )r r r )pZfan_out)modeg{Gz?) r# r$ r r r r& lenr% r max last_channelr
r+ ranger* r. featuresZDropoutZLinear
classifiermodules
isinstancer- initZkaiming_normal_weightr" Zzeros_Z GroupNormZones_Znormal_)r0 rC rD rE rF rG r rH Z
input_channelrV rX tcnsZoutput_channelir mr1 r3 r4 r$ D sp
zMobileNetV2.__init__r5 c C s2 | |}tj|d}t|d}| |}|S )Nr
r
r
)rX r Z
functionalZadaptive_avg_pool2dtorchflattenrY r8 r3 r3 r4
_forward_impl s
zMobileNetV2._forward_implc C s
| |S r7 )rg r8 r3 r3 r4 r9 s zMobileNetV2.forward)r? r@ NrA NNrB )r: r; r< r' floatr r r r r= r$ r rg r9 r> r3 r3 r1 r4 r C s& _
iz5 rd )Z
num_paramsZmin_size
categoriesc
@ sr e Zd Zedeeddedddddid d
d d d
Zedeedddedddddid 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.)ZrecipeZ_metricsZ_ops
_file_sizeZ_docs)urlZ
transformsmetaz=https://download.pytorch.org/models/mobilenet_v2-7ebf99e0.pth )rk Z 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_V1Z
IMAGENET1K_V2DEFAULTr3 r3 r3 r4 r s>
Z
pretrained)weightsT)rs progress)rs rt kwargsr c K sR t | } | dk r(t|dt| jd tf |}| dk rN|| 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:
NrC ri T)rt Z
check_hash)r verifyr rT rn r Zload_state_dictZget_state_dict)rs rt ru modelr3 r3 r4 r s
)# functoolsr typingr r r r re r r Zops.miscr
Ztransforms._presetsr utilsr Z_apir r r _metar Z_utilsr r r __all__r= r r rp r rq boolr r3 r3 r3 r4 s6
0o*