ȅi0':SSKJrJr SSKrSSKJr SSKJr SSKJ r J r J r SSK J r JrJr /SQr"SS\R"R$R&5r"S S \\R(5r"S S \\R*5r"S S\\R,5rg))ClassVarLiteralN) _FusedModule) _size_1_t _size_2_t _size_3_t)_pair_single_triple)Conv1dConv2dConv3dc \rSrSr%\\\RRR\ S'SS\ S\ S\ \ S4S\ \ S4S \ \ \ S4-S \ \ S4S \S \ \ S4S \ S\S\SSS4SjjrSr\SSj5rSrSrg)_ConvNd _FLOAT_MODULEN in_channels out_channels kernel_size.stridepaddingdilation transposedoutput_paddinggroupsbias padding_modezerosreflect replicatecircularreturncXS.n[RRRR"UUUUUUUUUU U U 4 0UD6 U (d [ S5eXlU RUS9Ulg)N)devicedtypez'qconfig must be provided for QAT module)factory_kwargs) nnmodulesconvr__init__AssertionErrorqconfigweightweight_fake_quant)selfrrrrrrrrrrrr-r%r&r's V/home/james-whalen/.local/lib/python3.13/site-packages/torch/ao/nn/qat/modules/conv.pyr+_ConvNd.__init__s"%+; ((               !JK K !(~!NclURXRUR5UR5$N _conv_forwardr/r.rr0inputs r1forward_ConvNd.forward7(!!%)?)? )LdiiXXr3c [U5URLaE[SURSURRS[U5R35e[ US5(d [S5eUR (d [S5e[ [U5[5(aUSnUR nU"URURURURURURURURSLUR US 9 nUR"UlURUlU$) zCreate a qat module from a float module Args: `mod`: a float module, either produced by torch.ao.quantization utilities or directly from user zqat.z.from_float only works for z, got r-z,Input float module must have qconfig definedz,Input float module must have a valid qconfigrN)rrrrrrr-)typerr,__name__hasattrr- issubclassrrrrrrrrrrr.)clsmoduse_precomputed_fake_quantr-qat_convs r1 from_float_ConvNd.from_float:s 9C-- - s||n$?$$--.fT#Y5G5G4HJ sI&& !OP P{{ !OP P d3i . .a&C++ OO    OO::KK\\::%))  ** r3c ,[U5nURURURURUR UR URURURSLUR5 n[RRURR55UlURb<[RRURR55Ul [!U["5(ayU/n[%US5(d['UR(S35eUR+5nUR-U5 UR."U6nUR1UR25 U$U$)zzThis works for both single qat conv, and the qat conv - relu modules to convert the qat module to a floating point module N_FLOAT_RELU_MODULEz' must have _FLOAT_RELU_MODULE attribute)r>_FLOAT_CONV_MODULErrrrrrrrrtorchr( Parameterr.detachrArr@r,r?rIappendrtraintraining)r0rBr*r)relufuseds r1to_float_ConvNd.to_float^s?4j%%          KK LL MM KK IIT !     hh((););)=> 99 **499+;+;+=>DI c< ( (fG3 455$||n$KL))+D NN4 %%w/E KK &LKr3)r-r/)NNNF)r? __module__ __qualname____firstlineno__rr>r(r)r*r__annotations__inttuplestrboolrr+r: staticmethodrFrS__static_attributes__r3r1rrsD!8!89::$O$O$O38_ $O c3h $O uS#X& $OS/$O$Oc3h$O$O$OIJ$O !$OLY!!F!r3rc^\rSrSr%Sr\R r\\ \R \ S'\R r \\ \R \ S'SS\ S\ S\ S \ S \\ -S \ S \ S \S\SSS4U4Sjjjr\SU4Sjj5rSrU=r$)r a: A Conv1d module attached with FakeQuantize modules for weight, used for quantization aware training. We adopt the same interface as :class:`~torch.nn.Conv1d` Similar to :class:`~torch.nn.Conv2d`, with FakeQuantize modules initialized to default. Attributes: weight_fake_quant: fake quant module for weight rrJNrrrrrrrrrrr#c >[U5n [U5n[U[5(aUO [U5n[U5n[TU]UUU UUUS[S5UUU U U U S9 gNFr) rrrrrrrrr-r%r&)r isinstancer\superr+r0rrrrrrrrrr-r%r& kernel_size_stride_padding_ dilation_ __class__s r1r+Conv1d.__init__z{+ &/(#667GG[TU]XUS9$N)rDrfrFrBrCrDrls r1rFConv1d.from_float w! 1K"  r3r` rrvrvTrNNNrU)r?rVrWrX__doc__r(r rrr>rYrJrZrr\r]rr+ classmethodrFr_ __classcell__rls@r1r r s 02yyM8DO,846IIbii1=#$MT" " "  "  " y " " " " IJ"  " " H  r3r c ^\rSrSr%Sr\R r\\ \R \ S'\R r \\ \R \ S'SS\ S\ S\ S \ S \\ -S \ S \ S \S\SSS4U4SjjjrSr\SU4Sjj5rSrU=r$)r a A Conv2d module attached with FakeQuantize modules for weight, used for quantization aware training. We adopt the same interface as `torch.nn.Conv2d`, please see https://pytorch.org/docs/stable/nn.html?highlight=conv2d#torch.nn.Conv2d for documentation. Similar to `torch.nn.Conv2d`, with FakeQuantize modules initialized to default. Attributes: weight_fake_quant: fake quant module for weight rrJNrrrrrrrrrrr#c >[U5n [U5n[U[5(aUO [U5n[U5n[TU]UUU UUUS[S5UUU U U U S9 grd)r rer\rfr+rgs r1r+Conv2d.__init__sx[) -(#667E'N(O      8%  r3clURXRUR5UR5$r5r6r8s r1r:Conv2d.forwardr<r3c >[TU]XUS9$rprqrrs r1rFConv2d.from_floatrtr3r`rurU)r?rVrWrXrwr(r rrr>rYrJrZrr\r]rr+r:rxrFr_ryrzs@r1r r  02yyM8DO,846IIbii1=#$MT" " "  "  " y " " " " IJ"  " " HY  r3r c ^\rSrSr%Sr\R r\\ \R \ S'\R r \\ \R \ S'SS\ S\ S\ S \ S \\ -S \ S \ S \S\SSS4U4SjjjrSr\SU4Sjj5rSrU=r$)ra A Conv3d module attached with FakeQuantize modules for weight, used for quantization aware training. We adopt the same interface as `torch.nn.Conv3d`, please see https://pytorch.org/docs/stable/nn.html?highlight=conv3d#torch.nn.Conv3d for documentation. Similar to `torch.nn.Conv3d`, with FakeQuantize modules initialized to default. Attributes: weight_fake_quant: fake quant module for weight rrJNrrrrrrrrrrr#c >[U5n [U5n[U[5(aUO [U5n[U5n[TU]UUU UUUS[S5UUU U U U S9 grd)r rer\rfr+rgs r1r+Conv3d.__init__rnr3clURXRUR5UR5$r5r6r8s r1r:Conv3d.forward6r<r3c >[TU]XUS9$rprqrrs r1rFConv3d.from_float9rtr3r`rurU)r?rVrWrXrwr(rrrr>rYrJrZrr\r]rr+r:rxrFr_ryrzs@r1rrrr3r)typingrrrKtorch.nnr(torch.ao.nn.intrinsicrtorch.nn.common_typesrrrtorch.nn.modules.utilsr r r __all__r)r*rr r rr`r3r1rsy$ .AA:: )qbjjoo%%qh9 Wbii9 x> Wbii> B> Wbii> r3