Ei% z S SK Jr S SKJrJrJr S SKrS SKJrJr SSK J
r
SSK J r SSK
Jr S S
KJrJrJr S S KJr S S KJrJrJr / S
Qr " S S\R4 5 r " S S\R4 5 r SS\S.r " S S\5 r \" 5 \" S\ R> 4S9SSS.S\\ S \ S \S \ 4S jj5 5 r!g) )partial)AnyCallableOptionalN)nnTensor )Conv2dNormActivation)ImageClassification)_log_api_usage_once )register_modelWeights WeightsEnum)_IMAGENET_CATEGORIES)_make_divisible_ovewrite_named_paramhandle_legacy_interface) MobileNetV2MobileNet_V2_Weights mobilenet_v2c ^ \ rS rSr SS\S\S\S\S\\S\R 4 S S4 U 4S
jjjr S \
S \
4S jr S
r U =r
$ )InvertedResidual Ninpoupstride expand_ratio
norm_layer.returnc , > [ TU ] 5 X0l US; a [ SU 35 eUc [ R
n[
[ X- 5 5 nU R S:H =( a X:H U l / nUS:w a) UR [ XSU[ R S95 UR [ UUUUU[ R S9[ R " XbSSSSS9U" U5 /5 [ R " U6 U l X l US: U l g ) 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__s [/home/mande/repo/quber/.venv/lib/python3.13/site-packages/torchvision/models/mobilenetv2.pyr( InvertedResidual.__init__ s B6( KL L Js 123
#{{a / >CJ "$ 1 MM$S!PZmomumuv
% !% )%'XX
*1a?3
MM6* qj xc l U R ( a XR U5 - $ U R U5 $ N)r- r3 r6 r= s r: forwardInvertedResidual.forward< s* yy|# #99Q< r< )r5 r3 r4 r r- r? )__name__
__module__ __qualname____firstlineno__r+ r r r Moduler( r rA __static_attributes__
__classcell__r9 s @r: r r sq sw&!&! &!*-&!=@&!NVW_`cegenen`nWoNp&!
&! &!P F r< r c ^ \ rS rSr SS\S\S\\\\ S\S\\S\ R 4 S \\S\ R 4 S
\S S4U 4S jjjr S
\ S \ 4S jr
S
\ S \ 4S jrSrU =r$ )r C N num_classes
width_multinverted_residual_setting
round_nearestblock.r dropoutr c > [ TU ] 5 [ U 5 Uc [ nUc [ R
nSnSn Uc / SQ/ SQ/ SQ/ SQ/ SQ/ S Q/ S
Q/n[
U5 S :X d [
US 5 S :w a [ S
U 35 e[ X- U5 n[ U [ SU5 - U5 U l
[ SUSU[ R S9/n
U HI u pp[ X- U5 n[ U
5 H& nUS :X a UOSnU
R U" XUXS95 UnM( MK U
R [ XR SU[ R S95 [ R " U
6 U l [ R " [ R" " US9[ R$ " U R U5 5 U l U R) 5 GH n[+ U[ R, 5 ( ab [ R. R1 UR2 SS9 UR4 b+ [ R. R7 UR4 5 M M [+ U[ R
[ R8 45 ( aU [ R. R; UR2 5 [ R. R7 UR4 5 GM [+ U[ R$ 5 ( d GM+ [ R. R= UR2 S S5 [ R. R7 UR4 5 GM g)a
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 )rV rT r )rV @ r )rV ` rX r
)rV rX r )rV i@ r
r
r rZ zGinverted_residual_setting should be non-empty or a 4-element list, got ?rX 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. r2 featuresDropoutLinear
classifiermodules
isinstancer1 initkaiming_normal_weightr& zeros_ GroupNormones_normal_)r6 rM rN rO rP rQ r rR
input_channelrc re tcnsoutput_channelir mr9 s r: r( MobileNetV2.__init__D s 0 D! =$E J
$ , ) % ( )Q .#6OPQ6R2SWX2X YZsYtu
(
(BMR
+L3sJ;O,OQ^ _ M!
egemem n%
4JA! ,Q^] KN1X 1f!mVZ[ st .
4 0 0aJikiqiq
x0
-- JJ ! IId'' 5
A!RYY''''y'A66%GGNN166* &A =>>
ahh'qvv&Aryy))!T2qvv& r< r= c U R U5 n[ R R US5 n[ R
" US5 nU R
U5 nU$ )Nr
r
r
)re r
functionaladaptive_avg_pool2dtorchflattenrh r@ s r:
_forward_implMobileNetV2._forward_impl sK
MM! MM - -a 8 MM!Q OOA r< c $ U R U5 $ r? )r r@ s r: rA MobileNetV2.forward s !!!$$r< )rh re rc )i r] N NNg?)rC rD rE rF r+ floatr listr r rG r( r r rA rH rI rJ s @r: r r C s ?C 489= ]']' ]' $,DcO#< ]'
]' bii 01
]' Xc299n 56]' ]'
]' ]'~v & % %F % %r< r iz5 r| )
num_paramsmin_size
categoriesc \ rS rSr\" S\" \SS90 \ESSSSS .0S
S S S
.ES9r\" S\" \SSS90 \ESSSSS .0S
SSS
.ES9r \ r
Sr g)r z=https://download.pytorch.org/models/mobilenet_v2-b0353104.pth ) crop_sizezQhttps://github.com/pytorch/vision/tree/main/references/classification#mobilenetv2z ImageNet-1Kgx&1Q@gMV@)zacc@1zacc@5g$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) rC rD rE rF r r r _COMMON_META
IMAGENET1K_V1
IMAGENET1K_V2DEFAULTrH r r< r: r r s K.#>
i # # s
M" K.#3O
` # #
M* Gr< r
pretrained)weightsT)r progressr r kwargsr c [ R U 5 n U b# [ US[ U R S 5 5 [ S0 UD6nU b UR
U R USS95 U$ )aq 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:
rM r T)r
check_hashr )r verifyr ra r r load_state_dictget_state_dict)r r r models r: r r sk 0 #))'2G fmS l9S5TU !& !E
g 4 4hSW 4 XY Lr< )" functoolsr typingr r r r r r ops.miscr
transforms._presetsr utilsr _apir r r _metar _utilsr r r __all__rG r r r r r boolr r r< r: r s * * + 5 ' 6 6 ' S S B- ryy - `k%")) k%^ & ' ; ' T ,0D0R0R!ST15 -. AE X[ U r<