+
jN< ^ RI Ht ^ RIHtHtHt ^ RIt^ RIHt ^ RIH t ^RI
H t ^RI H
t
^RIHtHtHt ^RIHt ^R IHtHt . R+OtR R lt ! R R]P2 4 t ! R R
]P2 4 tR R lt RR,R]RR/t ! R R ]4 t ! R R ]4 t ! R R
]4 t ! R R]4 t!]! 4 ]! R ] PD 3R 7 R RR!R"/R# R$ ll4 4 t#]! 4 ]! R ] PD 3R 7 R RR!R"/R% R& ll4 4 t$]! 4 ]! R ] PD 3R 7 R RR!R"/R' R( ll4 4 t%]! 4 ]! R ]!PD 3R 7 R RR!R"/R) R* ll4 4 t&R# )- )partial)AnyCallableOptionalNTensor)ImageClassification)_log_api_usage_once)register_modelWeights WeightsEnum)_IMAGENET_CATEGORIES)_ovewrite_named_paramhandle_legacy_interface ShuffleNetV2ShuffleNet_V2_X0_5_WeightsShuffleNet_V2_X1_0_WeightsShuffleNet_V2_X1_5_WeightsShuffleNet_V2_X2_0_Weightsc < V ^8 d QhR\ R\ R\ /# ) xgroupsreturn)r int)formats "`/home/luca/screen-checking/.venv/lib/python3.14/site-packages/torchvision/models/shufflenetv2.py __annotate__r s!
v
s
v
c V P 4 w r#rEW1, pV P W!WdV4 p \ P ! V ^^4 P 4 p V P W#WE4 p V # ) )sizeviewtorch transpose
contiguous)r r batchsize num_channelsheightwidthchannels_per_groups && r channel_shuffler, sb -.VVX*IV%/
y"4eDA
1a ++-A
y6A Hr c p a a ] tR t^+t oV3R lV 3R llt]RV3R lR ll4 tV3R lR ltRtVt V ;t
# ) InvertedResidualc 0 < V ^8 d QhRS[ RS[ RS[ RR/# )r inpoupstrider N)r )r
__classdict__s "r r InvertedResidual.__annotate__, s) (
(
C (
c (
3 (
4 (
r c 2 < \ SV ` 4 ^Tu;8: d ^8: g M \ R4 hW0n V^, pV P ^8X d# W^, 8w d \ RV RV RV R24 hV P ^8 d \ P
! V P
W^V P ^R7 \ P ! V4 \ P ! W^^^ RR7 \ P ! V4 \ P ! R R
7 4 V n
M\ P
! 4 V n
\ P
! \ P ! V P ^8 d TMTV^^^ RR7 \ P ! V4 \ P ! R R
7 V P
WD^V P ^R7 \ P ! V4 \ P ! WD^^^ RR7 \ P ! V4 \ P ! R R
7 4 V n R # ) r! zillegal stride valuez Invalid combination of stride z, inp z and oup zB values. If stride == 1 then inp should be equal to oup // 2 << 1. kernel_sizer2 paddingF)r7 r2 r8 biasTinplaceN) super__init__
ValueErrorr2 nn
Sequentialdepthwise_conv BatchNorm2dConv2dReLUbranch1branch2)selfr0 r1 r2 branch_features __class__s &&&& r r= InvertedResidual.__init__, s
V q 34 4 ( KK1 3Q*>#> 0se9SE RT U
;;? ==##C!DKKYZ#[s# #AaQRY^_/% DL ==?DL}} II ao
NN? + GGD ! aX\XcXcmn o NN? + IIoAaYZaf g NN? + GGD !
r c X < V ^8 d QhRS[ RS[ RS[ RS[ RS[ RS[RS[P /# )r ior7 r2 r8 r9 r )r boolr? rC )r r3 s "r r r4 W sR R R RR%(R25RDGRSWR Rr c
6 \ P ! WW#WEV R 7 # ))r9 r )r? rC )rL rM r7 r2 r8 r9 s &&&&&&r rA InvertedResidual.depthwise_convV s yy{GqQQr c & < V ^8 d QhRS[ RS[ /# r r r r )r r3 s "r r r4 \ s F r c , V P ^8X d? VP ^^R7 w r#\ P ! W P V4 3^R7 pM8\ P ! V P V4 V P V4 3^R7 p\
V^4 pV# )r! )dim)r2 chunkr$ catrF rE r, )rG r x1x2outs && r forwardInvertedResidual.forward\ st ;;! WWQAW&FB))R b!1 2:C))T\\!_dll1o >AFC c1%
r )rE rF r2 )r! r F) __name__
__module__ __qualname____firstlineno__r= staticmethodrA rZ __static_attributes____classdictcell__
__classcell__rI r3 s @@r r. r. + s; (
(
T R R R
r r. c j a a ] tR t^ht oR]3V3R lV 3R llltV3R lR ltV3R lR ltRtVt V ;t
# ) r i c
~ < V ^8 d QhRS[ S[, RS[ S[, RS[RS[RS[P 3, RR/# )r stages_repeatsstages_out_channels num_classesinverted_residual.r N)listr r r? Module)r r3 s "r r ShuffleNetV2.__annotate__i sR .: .: S .: "#Y.: .:
$CN3 .:
.:r c 6 < \ S V ` 4 \ V 4 \ V4 ^8w d \ R4 h\ V4 ^8w d \ R4 hW n ^pV P
^ , p\ P ! \ P ! WV^^^RR7 \ P ! V4 \ P ! RR7 4 V n Tp\ P ! ^^^R7 V n
V V V R Uu. uF pRV 2NK
pp\ WV P
R , 4 Fa w rpV! WV^4 .p \ V
^,
4 F pV P! V! Wf^4 4 K \# W \ P ! V ! 4 TpKc V P
R , p\ P ! \ P ! WV^^^ RR7 \ P ! V4 \ P ! RR7 4 V n \ P& ! Wc4 V n R
# u upi )
z2expected stages_repeats as list of 3 positive intsz7expected stages_out_channels as list of 5 positive intsF)r9 Tr: r6 stage:r! NNN)r ro )r< r= r
lenr> _stage_out_channelsr? r@ rC rB rD conv1 MaxPool2dmaxpoolziprangeappendsetattrconv5Linearfc)
rG rg rh ri rj input_channelsoutput_channelsrL stage_namesnamerepeatsseqrI s
&&&&& r r= ShuffleNetV2.__init__i s D! ~ ! # QR R " #q ( VW W#6 2215]] IInq!QU K NN? + GGD !
)||!QG
,56Iqqc{I 6.1+tOgOghjOk.l *D?$^aHIC7Q;'
,_qQR ( D
s 3 4 ,N /m 2226]] IInq!QU K NN? + GGD !
))O9 7s 5Hc & < V ^8 d QhRS[ RS[ /# rR r )r r3 s "r r rm s
v
&
r c V P V4 pV P V4 pV P V4 pV P V4 pV P V4 pV P V4 pVP
^^.4 pV P V4 pV# )r )ru rw stage2stage3stage4r| meanr~ rG r s &&r
_forward_implShuffleNetV2._forward_impl ss JJqM LLO KKN KKN KKN JJqM
FFAq6N GGAJr c & < V ^8 d QhRS[ RS[ /# rR r )r r3 s "r r rm s % % %F %r c $ V P V4 # )N)r r s &&r rZ ShuffleNetV2.forward s !!!$$r )rt ru r| r~ rw ) r\ r] r^ r_ r. r= r rZ ra rb rc rd s @@r r r h s1
6F .: .:`
% % %r c
j V ^8 d QhR\ \ , R\ R\ R\ R\ /# )r weightsprogressargskwargsr )r r
rN r r )r s "r r r s?
k
"
r c V e( \ VR\ V P R, 4 4 \ V/ VB pV e# VP V P VRR7 4 V# )Nri
categoriesT)r
check_hash)r rs metar load_state_dictget_state_dict)r r r r models &&*, r
_shufflenetv2r s] fmS l9S5TU $ )& )E
g 4 4hSW 4 XY Lr min_sizer recipez2https://github.com/ericsun99/Shufflenet-v2-Pytorchc h ] tR t^t]! R]! ]^R7 / ]CRRRRRRR R
//R R R
RRR/CR7 t]t Rt
R# )r zDhttps://download.pytorch.org/models/shufflenetv2_x0.5-f707e7126e.pth crop_size
num_paramsi _metrics ImageNet-1Kacc@1g-FN@acc@5g9voT@_opsg{Gz?
_file_sizegT㥛 @_docsVThese weights were trained from scratch to reproduce closely the results of the paper.url
transformsr N r\ r] r^ r_ r r r _COMMON_META
IMAGENET1K_V1DEFAULTra r r r r r sd R.#>
' VV
D % q
M$ Gr c h ] tR t^t]! R]! ]^R7 / ]CRRRRRRR R
//R R R
RRR/CR7 t]t Rt
R# )r zBhttps://download.pytorch.org/models/shufflenetv2_x1-5666bf0f80.pthr r i" r r r gI +WQ@r gNbX9V@r g(\?r gE!@r r r r Nr r r r r r sd P.#>
' VV
E % q
M$ Gr c n ] tR t^t]! R]! ]^^R7 / ]CRRRRRRR R
R R //R
RRRRR/CR7 t]t Rt
R# )r zBhttps://download.pytorch.org/models/shufflenetv2_x1_5-3c479a10.pthr resize_sizer +https://github.com/pytorch/vision/pull/5906r iv5 r r r g9v?R@r g/$V@r gl?r gw/ +@r
These weights were trained from scratch by using TorchVision's `new training recipe
`_.
r r Nr r r r r r n P.#3O
C ' VV
E &
M* Gr c n ] tR tRt]! R]! ]^^R7 / ]CRRRRRR R
R R R
//RRRRRR/CR7 t]t Rt
R# )r i zBhttps://download.pytorch.org/models/shufflenetv2_x2_0-8be3c8ee.pthr r r r ip r r r g QS@r gMb@W@r g-?r g+n<@r r r r Nr r r r r r r r
pretrained)r r r Tc ^ V ^8 d QhR\ \ , R\ R\ R\ /# r r r r r )r r rN r r )r s "r r r ! s8 Z Z34ZGKZ^aZZr c P \ P V 4 p \ W. RO. RO3/ VB # )a
Constructs a ShuffleNetV2 architecture with 0.5x output channels, as described in
`ShuffleNet V2: Practical Guidelines for Efficient CNN Architecture Design
`__.
Args:
weights (:class:`~torchvision.models.ShuffleNet_V2_X0_5_Weights`, optional): The
pretrained weights to use. See
:class:`~torchvision.models.ShuffleNet_V2_X0_5_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.shufflenetv2.ShuffleNetV2``
base class. Please refer to the `source code
`_
for more details about this class.
.. autoclass:: torchvision.models.ShuffleNet_V2_X0_5_Weights
:members:
rq rq ) 0 ` )r verifyr r r r s $$,r shufflenet_v2_x0_5r s* 4 )//8G I7N YRX YYr c ^ V ^8 d QhR\ \ , R\ R\ R\ /# r )r r rN r r )r s "r r r @ 8 \ \34\GK\^a\\r c P \ P V 4 p \ W. RO. RO3/ VB # )a
Constructs a ShuffleNetV2 architecture with 1.0x output channels, as described in
`ShuffleNet V2: Practical Guidelines for Efficient CNN Architecture Design
`__.
Args:
weights (:class:`~torchvision.models.ShuffleNet_V2_X1_0_Weights`, optional): The
pretrained weights to use. See
:class:`~torchvision.models.ShuffleNet_V2_X1_0_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.shufflenetv2.ShuffleNetV2``
base class. Please refer to the `source code
`_
for more details about this class.
.. autoclass:: torchvision.models.ShuffleNet_V2_X1_0_Weights
:members:
r )r t i r )r r r r s $$,r shufflenet_v2_x1_0r > * 4 )//8G I7P [TZ [[r c ^ V ^8 d QhR\ \ , R\ R\ R\ /# r )r r rN r r )r s "r r r _ r r c P \ P V 4 p \ W. RO. RO3/ VB # )a
Constructs a ShuffleNetV2 architecture with 1.5x output channels, as described in
`ShuffleNet V2: Practical Guidelines for Efficient CNN Architecture Design
`__.
Args:
weights (:class:`~torchvision.models.ShuffleNet_V2_X1_5_Weights`, optional): The
pretrained weights to use. See
:class:`~torchvision.models.ShuffleNet_V2_X1_5_Weights` below for
more details, and possible values. By default, no pre-trained
weights are used.
progress (bool, optional): If True, displays a progress bar of the
download to stderr. Default is True.
**kwargs: parameters passed to the ``torchvision.models.shufflenetv2.ShuffleNetV2``
base class. Please refer to the `source code
`_
for more details about this class.
.. autoclass:: torchvision.models.ShuffleNet_V2_X1_5_Weights
:members:
r )r i` i r )r r r r s $$,r shufflenet_v2_x1_5r ] r r c ^ V ^8 d QhR\ \ , R\ R\ R\ /# r )r r rN r r )r s "r r r ~ r r c P \ P V 4 p \ W. RO. RO3/ VB # )a
Constructs a ShuffleNetV2 architecture with 2.0x output channels, as described in
`ShuffleNet V2: Practical Guidelines for Efficient CNN Architecture Design
`__.
Args:
weights (:class:`~torchvision.models.ShuffleNet_V2_X2_0_Weights`, optional): The
pretrained weights to use. See
:class:`~torchvision.models.ShuffleNet_V2_X2_0_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.shufflenetv2.ShuffleNetV2``
base class. Please refer to the `source code
`_
for more details about this class.
.. autoclass:: torchvision.models.ShuffleNet_V2_X2_0_Weights
:members:
r )r i i i )r r r r s $$,r shufflenet_v2_x2_0r | r r ) r r r r r r r r r )r! r! )' functoolsr typingr r r r$ torch.nnr? r transforms._presetsr utilsr
_apir r r
_metar _utilsr r __all__r, rl r. r r r r r r r r r r r r r r r r s * * 5 ' 6 6 ' B
:ryy :z>%299 >%B$ & B , , 2 2 ,0J0X0X!YZZ7;ZNRZ [ Z: ,0J0X0X!YZ\7;\NR\ [ \: ,0J0X0X!YZ\7;\NR\ [ \: ,0J0X0X!YZ\7;\NR\ [ \r