hN< * d dl mZ d dlmZmZmZ d dlZd dlmZ d dlm Z ddl
m Z ddl m
Z
dd lmZmZmZ dd
lmZ dd lmZmZ g d Zd
e dede fdZ G d dej4 Z G d dej4 Z dee de dedede f
dZ dedd Z G d d e Z G d d e Z! G d! d"e Z" G d# d$e Z# e ed%e jH f& dd'd(dee de dede fd) Z% e ed%e!jH f& dd'd(dee! de dede fd* Z& e ed%e"jH f& dd'd(dee" de dede fd+ Z' e ed%e#jH f& dd'd(dee# de dede fd, Z(y)- )partial)AnyCallableOptionalN)Tensor )ImageClassification)_log_api_usage_once )register_modelWeights WeightsEnum)_IMAGENET_CATEGORIES)_ovewrite_named_paramhandle_legacy_interface) ShuffleNetV2ShuffleNet_V2_X0_5_WeightsShuffleNet_V2_X1_0_WeightsShuffleNet_V2_X1_5_WeightsShuffleNet_V2_X2_0_Weightsshufflenet_v2_x0_5shufflenet_v2_x1_0shufflenet_v2_x1_5shufflenet_v2_x2_0xgroupsreturnc | j \ }}}}||z }| j ||||| } t j | dd j } | j |||| } | S )Nr r )sizeviewtorch transpose
contiguous)r r batchsize num_channelsheightwidthchannels_per_groups ]/var/www/html/eduruby.in/venv/lib/python3.12/site-packages/torchvision/models/shufflenetv2.pychannel_shuffler* sp -.VVX*I|VU%/
y&"4feDA
1a ++-A
y,6A H c e Zd Zdedededdf fd Ze ddeded eded
ed edej fd Z d
e
de
fdZ xZ S )InvertedResidualinpoupstrider Nc t | d|cxk rdk st d t d || _ |dz }| j dk( r ||dz k7 rt d| d| d| d | j dkD rt j
| j
||d| j d t j | t j ||ddd
d t j | t j d
| _
nt j
| _
t j
t j | j dkD r|n||ddd
d t j | t j d
| j
||d| j d t j | t j ||ddd
d t j | t j d
| _ y )Nr zillegal stride valuer z Invalid combination of stride z, inp z and oup zB values. If stride == 1 then inp should be equal to oup // 2 << 1. kernel_sizer0 paddingr F)r4 r0 r5 biasTinplace) super__init__
ValueErrorr0 nn
Sequentialdepthwise_conv BatchNorm2dConv2dReLUbranch1branch2)selfr. r/ r0 branch_features __class__s r) r: zInvertedResidual.__init__, s
V q 34 4 ! 34 4 ( KK1 3/Q*>#> 0se9SE RT U
;;? ==##C!DKKYZ#[s# #AaQRY^_/% DL ==?DL}} II ao
NN? + GGD ! aX\XcXcmn o NN? + IIoAaYZaf g NN? + GGD !
r+ ior4 r5 r6 c : t j | |||||| S )N)r6 r )r< r@ )rG rH r4 r0 r5 r6 s r) r> z InvertedResidual.depthwise_convV s yyA{FG$qQQr+ r c ( | j dk( r?|j dd \ }}t j || j | fd }n7t j | j | | j | fd }t
|d }|S )Nr r )dim)r0 chunkr! catrC rB r* )rD r x1x2outs r) forwardzInvertedResidual.forward\ sx ;;! WWQAW&FB))R b!1 2:C))T\\!_dll1o >AFC c1%
r+ )r r F)
__name__
__module__ __qualname__intr: staticmethodboolr< r@ r> r rQ
__classcell__rF s @r) r- r- + s (
C (
c (
3 (
4 (
T Z_R RR%(R25RDGRSWR R R
F r+ r- c e Zd Zdefdee dee dededej f ddf
fd
Z d
e
de
fd Z d
e
de
fd Z xZ
S )
r i stages_repeatsstages_out_channels num_classesinverted_residual.r Nc
t | t | t | dk7 r t d t | dk7 r t d || _ d}| j
d }t
j t
j ||dddd t
j | t
j d
| _ |}t
j ddd | _
| | | d
D cg c] }d| }}t ||| j
dd D ]\ \ } }
} |||d g} t |
dz
D ] }| j! |||d t# | | t
j | |}^ | j
d }t
j t
j ||dddd t
j | t
j d
| _ t
j& || | _ y c c}w )Nr2 z2expected stages_repeats as list of 3 positive ints z7expected stages_out_channels as list of 5 positive intsr r r F)r6 Tr7 r3 )r r2 stage)r9 r: r
lenr; _stage_out_channelsr< r= r@ r? rA conv1 MaxPool2dmaxpoolziprangeappendsetattrconv5Linearfc)
rD r[ r\ r] r^ input_channelsoutput_channelsrG stage_namesnamerepeatsseqrF s
r) r: zShuffleNetV2.__init__i s D! ~ ! # QR R " #q ( VW W#6 2215]] IInoq!QU K NN? + GGD !
)||!QG
,56qqc{6 6.1+~tOgOghihjOk.l - *D'?$^_aHIC7Q;'
S
,_oqQR
S D$
s 3 4 ,N - 2226]] IInoq!QU K NN? + GGD !
))O[9 7s * G8r c | j | }| j | }| j | }| j | }| j | }| j | }|j
ddg }| j | }|S )Nr r2 )rf rh stage2stage3stage4rm meanro rD r s r)
_forward_implzShuffleNetV2._forward_impl ss JJqM LLO KKN KKN KKN JJqM
FFAq6N GGAJr+ c $ | j | S )N)r| r{ s r) rQ zShuffleNetV2.forward s !!!$$r+ )rR rS rT r- listrU r r< Moduler: r r| rQ rX rY s @r) r r h s|
6F .: S .: "#Y.: .:
$CN3 .:
.:`
v
&
% %F %r+ r weightsprogressargskwargsc | #t |dt | j d t |i |}| "|j | j |d |S )Nr]
categoriesT)r
check_hash)r rd metar load_state_dictget_state_dict)r r r r models r)
_shufflenetv2r s] fmS l9S5TU $ )& )E
g 4 4hSW 4 XY Lr+ )r r z2https://github.com/ericsun99/Shufflenet-v2-Pytorch)min_sizer recipec T e Zd Z ed eed i edddddid d
d d
ZeZy)r zDhttps://download.pytorch.org/models/shufflenetv2_x0.5-f707e7126e.pth crop_sizei ImageNet-1Kg-FN@g9voT@zacc@1zacc@5g{Gz?gT㥛 @VThese weights were trained from scratch to reproduce closely the results of the paper.
num_params_metrics_ops
_file_size_docsurl
transformsr N rR rS rT r
r r _COMMON_META
IMAGENET1K_V1DEFAULT r+ r) r r sT R.#>
! # # q
M$ Gr+ r c T e Zd Z ed eed i edddddid d
d d
ZeZy)r zBhttps://download.pytorch.org/models/shufflenetv2_x1-5666bf0f80.pthr r i" r gI +WQ@gNbX9V@r g(\?gE!@r r r Nr r r+ r) r r sT P.#>
! # # q
M$ Gr+ r c X e Zd Z ed eedd i eddddd d
id d d
d ZeZy)r zBhttps://download.pytorch.org/models/shufflenetv2_x1_5-3c479a10.pthr r resize_size+https://github.com/pytorch/vision/pull/5906iv5 r g9v?R@g/$V@r gl?gw/ +@
These weights were trained from scratch by using TorchVision's `new training recipe
`_.
r r r r r r r Nr r r+ r) r r [ P.#3O
C! # #
M* Gr+ r c X e Zd Z ed eedd i eddddd d
id d d
d ZeZy)r zBhttps://download.pytorch.org/models/shufflenetv2_x2_0-8be3c8ee.pthr r r r ip r g QS@gMb@W@r g-?g+n<@r r r Nr r r+ r) r r r r+ r
pretrained)r T)r r c R t j | } t | |g dg dfi |S )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:
ra ra ) 0 ` )r verifyr r r r s r) r r s, 4 )//8G (I7N YRX YYr+ c R t j | } t | |g dg dfi |S )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 r i r )r r r r s r) r r > , 4 )//8G (I7P [TZ [[r+ c R t j | } t | |g dg dfi |S )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) r r ] r r+ c R t j | } t | |g dg dfi |S )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) r r | r r+ )) functoolsr typingr r r r! torch.nnr< r transforms._presetsr utilsr
_apir r
r _metar _utilsr r __all__rU r* r r- r rW r r r r r r r r r r r r r+ r) r sn * * 5 ' 6 6 ' B
v
s
v
:ryy :z>%299 >%B
k
"
$ &B , , 2 2 ,0J0X0X!YZ7;dZ34ZGKZ^aZZ [ Z: ,0J0X0X!YZ7;d\34\GK\^a\\ [ \: ,0J0X0X!YZ7;d\34\GK\^a\\ [ \: ,0J0X0X!YZ7;d\34\GK\^a\\ [ \r+