o )iA@sddlmZddlmZmZmZmZddlZddlm Z ddlm Z ddl m Z ddl mZdd lmZmZmZdd lmZdd lmZmZdd l mZmZmZmZd dlmZmZmZgdZ Gddde j!Z"Gddde j#Z$dee%dee%deede&de&dede$fddZ'deddd d!Z(Gd"d#d#eZ)Gd$d%d%eZ*Gd&d'd'eZ+Gd(d)d)eZ,ed*d+ed,d-d.fd/dd0d1d2deee)efde&de&dede$f d3d4Z-ed5d+ed,d6d.fd/dd0d1d2deee*efde&de&dede$f d7d8Z.ed9d+ed,d:d.fd/dd0d1d2deee+efde&de&dede$f d;d<Z/ed=d+ed,d>d.fd/dd0d1d2deee,efde&de&dede$f d?d@Z0dS)A)partial)AnyListOptionalUnionN)Tensor) shufflenetv2)ImageClassification)register_modelWeights WeightsEnum)_IMAGENET_CATEGORIES)_ovewrite_named_paramhandle_legacy_interface)ShuffleNet_V2_X0_5_WeightsShuffleNet_V2_X1_0_WeightsShuffleNet_V2_X1_5_WeightsShuffleNet_V2_X2_0_Weights) _fuse_modules _replace_reluquantize_model) QuantizableShuffleNetV2#ShuffleNet_V2_X0_5_QuantizedWeights#ShuffleNet_V2_X1_0_QuantizedWeights#ShuffleNet_V2_X1_5_QuantizedWeights#ShuffleNet_V2_X2_0_QuantizedWeightsshufflenet_v2_x0_5shufflenet_v2_x1_0shufflenet_v2_x1_5shufflenet_v2_x2_0cs<eZdZdededdffdd Zdedefdd ZZS) QuantizableInvertedResidualargskwargsreturnNcs"tj|i|tj|_dSN)super__init__nnZ quantizedZFloatFunctionalcatselfr$r% __class__{/var/www/html/eduruby.in/lip-sync/lip-sync-env/lib/python3.10/site-packages/torchvision/models/quantization/shufflenetv2.pyr)$sz$QuantizableInvertedResidual.__init__xcCsh|jdkr|jddd\}}|jj|||gdd}n|jj||||gdd}t|d}|S)Nrr )dim)Zstridechunkr+branch2branch1rZchannel_shuffle)r-r2x1Zx2outr0r0r1forward(s   z#QuantizableInvertedResidual.forward)__name__ __module__ __qualname__rr)rr9 __classcell__r0r0r.r1r##sr#csTeZdZdededdffdd Zdedefdd Zd d eeddfd d Z Z S)rr$r%r&Ncs6tj|dti|tjj|_tjj|_ dS)NZinverted_residual) r(r)r#torchZaoZ quantizationZ QuantStubquantZ DeQuantStubdequantr,r.r0r1r)6sz QuantizableShuffleNetV2.__init__r2cCs"||}||}||}|Sr')r?Z _forward_implr@)r-r2r0r0r1r9;s   zQuantizableShuffleNetV2.forwardis_qatcCs|jD]\}}|dvr|durt|gdg|ddq|D]3}t|turTt|jjdkrBt|jddggd g|ddt|jgdd d ggd g|ddq!dS) aBFuse conv/bn/relu modules in shufflenetv2 model Fuse conv+bn+relu/ conv+relu/conv+bn modules to prepare for quantization. Model is modified in place. .. note:: Note that this operation does not change numerics and the model after modification is in floating point )Zconv1Zconv5N)012T)ZinplacerrBrC)rD34rErF)567) Z_modulesitemsrmodulestyper#lenr6r5)r-rAnamemr0r0r1 fuse_modelAs    z"QuantizableShuffleNetV2.fuse_modelr') r:r;r<rr)rr9rboolrPr=r0r0r.r1r4s rstages_repeatsstages_out_channelsweightsprogressquantizer%r&cKs|durt|dt|jdd|jvrt|d|jd|dd}t||fi|}t||r7t|||durE||j|dd|S)NZ num_classes categoriesbackendfbgemmT)rUZ check_hash) rrMmetapoprrrZload_state_dictZget_state_dict)rRrSrTrUrVr%rXmodelr0r0r1 _shufflenetv2Zs    r])rrrYzdhttps://github.com/pytorch/vision/tree/main/references/classification#post-training-quantized-modelsz These weights were produced by doing Post Training Quantization (eager mode) on top of the unquantized weights listed below. )Zmin_sizerWrXrecipeZ_docsc @FeZdZedeeddiedejddddid d d d Z e Z d S)rzShttps://download.pytorch.org/models/quantized/shufflenetv2_x0.5_fbgemm-00845098.pth crop_sizei ImageNet-1Kg#~jL@gRS@zacc@1zacc@5g{Gz?gjt? num_params unquantized_metrics_ops _file_sizeurlZ transformsrZN) r:r;r<r rr _COMMON_METAr IMAGENET1K_V1IMAGENET1K_FBGEMM_V1DEFAULTr0r0r0r1r& rc @r_)rzQhttps://download.pytorch.org/models/quantized/shufflenetv2_x1_fbgemm-1e62bb32.pthr`rai"rcgףp= Q@gh|?U@rdg(\?gy&1@rerkN) r:r;r<r rr rmrrnrorpr0r0r0r1rrqrc @JeZdZedeedddieddejddd d id d d dZ e Z dS)rzShttps://download.pytorch.org/models/quantized/shufflenetv2_x1_5_fbgemm-d7401f05.pthr`rbZ resize_size+https://github.com/pytorch/vision/pull/5906iv5rcgSR@g̬V@rdgl?gK7A` @r^rfrgrhrirjrkN) r:r;r<r rr rmrrnrorpr0r0r0r1r( rc @rr)rzShttps://download.pytorch.org/models/quantized/shufflenetv2_x2_0_fbgemm-5cac526c.pthr`rsrtruiprcg-R@gZd;W@rdg-?g|?5@rvrkN) r:r;r<r rr rmrrnrorpr0r0r0r1rrwrZquantized_shufflenet_v2_x0_5)rNZ pretrainedcC|ddr tjStjSNrVF)getrrorrnr%r0r0r1 r|)rTTFrTrUrVcK4|rtnt|}tgdgdf|||d|S)aQ Constructs a ShuffleNetV2 with 0.5x output channels, as described in `ShuffleNet V2: Practical Guidelines for Efficient CNN Architecture Design `__. .. note:: Note that ``quantize = True`` returns a quantized model with 8 bit weights. Quantized models only support inference and run on CPUs. GPU inference is not yet supported. Args: weights (:class:`~torchvision.models.quantization.ShuffleNet_V2_X0_5_QuantizedWeights` or :class:`~torchvision.models.ShuffleNet_V2_X0_5_Weights`, optional): The pretrained weights for the model. See :class:`~torchvision.models.quantization.ShuffleNet_V2_X0_5_QuantizedWeights` 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. quantize (bool, optional): If True, return a quantized version of the model. Default is False. **kwargs: parameters passed to the ``torchvision.models.quantization.ShuffleNet_V2_X0_5_QuantizedWeights`` base class. Please refer to the `source code `_ for more details about this class. .. autoclass:: torchvision.models.quantization.ShuffleNet_V2_X0_5_QuantizedWeights :members: .. autoclass:: torchvision.models.ShuffleNet_V2_X0_5_Weights :members: :noindex: r)0`r~)rrverifyr]rTrUrVr%r0r0r1r0 rZquantized_shufflenet_v2_x1_0cCrxry)rzrrorrnr{r0r0r1r|r}cKr)aQ Constructs a ShuffleNetV2 with 1.0x output channels, as described in `ShuffleNet V2: Practical Guidelines for Efficient CNN Architecture Design `__. .. note:: Note that ``quantize = True`` returns a quantized model with 8 bit weights. Quantized models only support inference and run on CPUs. GPU inference is not yet supported. Args: weights (:class:`~torchvision.models.quantization.ShuffleNet_V2_X1_0_QuantizedWeights` or :class:`~torchvision.models.ShuffleNet_V2_X1_0_Weights`, optional): The pretrained weights for the model. See :class:`~torchvision.models.quantization.ShuffleNet_V2_X1_0_QuantizedWeights` 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. quantize (bool, optional): If True, return a quantized version of the model. Default is False. **kwargs: parameters passed to the ``torchvision.models.quantization.ShuffleNet_V2_X1_0_QuantizedWeights`` base class. Please refer to the `source code `_ for more details about this class. .. autoclass:: torchvision.models.quantization.ShuffleNet_V2_X1_0_QuantizedWeights :members: .. autoclass:: torchvision.models.ShuffleNet_V2_X1_0_Weights :members: :noindex: r)rtrsirr~)rrrr]rr0r0r1r rr Zquantized_shufflenet_v2_x1_5cCrxry)rzrrorrnr{r0r0r1r|Fr}cKr)aQ Constructs a ShuffleNetV2 with 1.5x output channels, as described in `ShuffleNet V2: Practical Guidelines for Efficient CNN Architecture Design `__. .. note:: Note that ``quantize = True`` returns a quantized model with 8 bit weights. Quantized models only support inference and run on CPUs. GPU inference is not yet supported. Args: weights (:class:`~torchvision.models.quantization.ShuffleNet_V2_X1_5_QuantizedWeights` or :class:`~torchvision.models.ShuffleNet_V2_X1_5_Weights`, optional): The pretrained weights for the model. See :class:`~torchvision.models.quantization.ShuffleNet_V2_X1_5_QuantizedWeights` 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. quantize (bool, optional): If True, return a quantized version of the model. Default is False. **kwargs: parameters passed to the ``torchvision.models.quantization.ShuffleNet_V2_X1_5_QuantizedWeights`` base class. Please refer to the `source code `_ for more details about this class. .. autoclass:: torchvision.models.quantization.ShuffleNet_V2_X1_5_QuantizedWeights :members: .. autoclass:: torchvision.models.ShuffleNet_V2_X1_5_Weights :members: :noindex: r)ri`irr~)rrrr]rr0r0r1r!Brr!Zquantized_shufflenet_v2_x2_0cCrxry)rzrrorrnr{r0r0r1r||r}cKr)aQ Constructs a ShuffleNetV2 with 2.0x output channels, as described in `ShuffleNet V2: Practical Guidelines for Efficient CNN Architecture Design `__. .. note:: Note that ``quantize = True`` returns a quantized model with 8 bit weights. Quantized models only support inference and run on CPUs. GPU inference is not yet supported. Args: weights (:class:`~torchvision.models.quantization.ShuffleNet_V2_X2_0_QuantizedWeights` or :class:`~torchvision.models.ShuffleNet_V2_X2_0_Weights`, optional): The pretrained weights for the model. See :class:`~torchvision.models.quantization.ShuffleNet_V2_X2_0_QuantizedWeights` 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. quantize (bool, optional): If True, return a quantized version of the model. Default is False. **kwargs: parameters passed to the ``torchvision.models.quantization.ShuffleNet_V2_X2_0_QuantizedWeights`` base class. Please refer to the `source code `_ for more details about this class. .. autoclass:: torchvision.models.quantization.ShuffleNet_V2_X2_0_QuantizedWeights :members: .. autoclass:: torchvision.models.ShuffleNet_V2_X2_0_Weights :members: :noindex: r)riiir~)rrrr]rr0r0r1r"xrr")1 functoolsrtypingrrrrr>Ztorch.nnr*rZtorchvision.modelsrZtransforms._presetsr Z_apir r r_metar_utilsrrrrrrutilsrrr__all__ZInvertedResidualr#Z ShuffleNetV2rintrQr]rmrrrrrr r!r"r0r0r0r1s       &   - - -