ó àÆ÷Xc@`sJdZddlmZmZmZddlZddlmZmZddl Z ddl Z ddl m Z m Z ddl mZmZddlmZddlmZmZdd lmZddlZddlZddlZy6dd lmZmZmZdd lmZe Z!Wne"k r1e#Z!nXd Z$ej%d ƒZ&dd„Z(dd„Z)dd„Z*d„Z+dd„Z,d„Z-dd„Z.d„Z/dd„Z0dddd7e d8d„Z1dddd9e d:d„Z2ddd;e d<d„Z3ddd=e d>d„Z4ddd?e d@d„Z5dddAe dBd„Z6e d „Z7e d!„Z8dde d"„Z9d#efd$„ƒYZ:d%e:fd&„ƒYZ;d'e;fd(„ƒYZ<d)e;fd*„ƒYZ=d+e:fd,„ƒYZ>d-e>fd.„ƒYZ?d/e>fd0„ƒYZ@d1e:fd2„ƒYZAd3eAfd4„ƒYZBd5eAfd6„ƒYZCdS(Cs Abstract conv interface i(tabsolute_importtprint_functiontdivisionN(treraiset integer_types(tas_tensor_variabletpatternbroadcast(tget_scalar_constant_valuetNotScalarConstantError(tAssert(tApplytOp(txrange(t _valfrommodet_bvalfromboundarytconvolve(t _convolve2dsrestructuredtext ens theano.tensor.nnet.abstract_convc`sÜ|d|d}‰|d|d}‰ˆdkrTtjtˆƒddƒ‰ntˆtƒršt‡‡‡‡‡fd†ttˆƒƒDƒƒ}n4t‡‡‡‡‡fd†ttˆƒƒDƒƒ}||f|S(s8 This function compute the output shape of convolution operation. Parameters ---------- image_shape: tuple of int (symbolic or numeric) corresponding to the input image shape. Its four (or five) element must correspond respectively to: batch size, number of input channels, height and width (and possibly depth) of the image. None where undefined. kernel_shape: tuple of int (symbolic or numeric) corresponding to the kernel shape. Its four (or five) elements must correspond respectively to: number of output channels, number of input channels, height and width (and possibly depth) of the kernel. None where undefined. border_mode: string, int (symbolic or numeric) or tuple of int (symbolic or numeric). If it is a string, it must be 'valid', 'half' or 'full'. If it is a tuple, its two (or three) elements respectively correspond to the padding on height and width (and possibly depth) axis. subsample: tuple of int (symbolic or numeric). Its two or three elements espectively correspond to the subsampling on height and width (and possibly depth) axis. filter_dilation: tuple of int (symbolic or numeric). Its two or three elements correspond respectively to the dilation on height and width axis. Returns ------- output_shape: tuple of int corresponding to the output image shape. Its four element must correspond respectively to: batch size, number of output channels, height and width of the image. None where undefined. iitdtypetintc3`s;|]1}tˆ|ˆ|ˆ|ˆ|ˆ|ƒVqdS(N(tget_conv_shape_1axis(t.0ti(t border_modetfilter_dilationtimshptkshpt subsample(s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pys Ksc3`s7|]-}tˆ|ˆ|ˆˆ|ˆ|ƒVqdS(N(R(RR(RRRRR(s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pys OsN(tNonetnumpytonestlent isinstancettupletrange(t image_shapet kernel_shapeRRRtbsizetnkerntout_shp((RRRRRs@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pytget_conv_output_shape#s! icC`sëd|||||gkrdS|d|d}|dkrJ|d}nO|dkrc|d}n6|dkrxd}n!|}|dkr™tdƒ‚n|dkr²||}n|d||}|dkrÝ||}n|d}|S( s1 This function compute the output shape of convolution operation. Parameters ---------- image_shape: int or None. Corresponds to the input image shape on a given axis. None if undefined. kernel_shape: int or None. Corresponds to the kernel shape on a given axis. None if undefined. border_mode: string or int. If it is a string, it must be 'valid', 'half' or 'full'. If it is an integer, it must correspond to the padding on the considered axis. subsample: int. It must correspond to the subsampling on the considered axis. dilation: int. It must correspond to the dilation on the considered axis. Returns ------- out_shp: int corresponding to the output image shape on the considered axis. None if undefined. ithalfitfulltvalidisborder_mode must be >= 0N(Rt ValueError(R"R#RRtdilationtdil_kernel_shapetpadR&((s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pyRWs(             c`sÜ|d|d}‰|d|d}‰ˆdkrTtjtˆƒddƒ‰ntˆtƒršt‡‡‡‡‡fd†ttˆƒƒDƒƒ}n4t‡‡‡‡‡fd†ttˆƒƒDƒƒ}||f|S(s This function tries to compute the kernel shape of convolution gradWeights. The weights shape can only be computed exactly when subsample is 1 and border_mode is not 'half'. If subsample is not 1 or border_mode is 'half', this function will return None. Parameters ---------- image_shape: tuple of int corresponding to the input image shape. Its four (or five) elements must correspond respectively to: batch size, number of output channels, height and width of the image. None where undefined. top_shape: tuple of int (symbolic or numeric) corresponding to the top image shape. Its four (or five) element must correspond respectively to: batch size, number of output channels, height and width (and possibly depth) of the image. None where undefined. border_mode: string, int (symbolic or numeric) or tuple of int (symbolic or numeric). If it is a string, it must be 'valid', 'half' or 'full'. If it is a tuple, its two (or three) elements respectively correspond to the padding on height and width (and possibly depth) axis. subsample: tuple of int (symbolic or numeric). Its two or three elements respectively correspond to the subsampling on height and width (and possibly depth) axis. filter_dilation: tuple of int (symbolic or numeric). Its two or three elements correspond respectively to the dilation on height and width axis. Returns ------- kernel_shape: tuple of int (symbolic or numeric) corresponding to the kernel shape. Its four (or five) elements correspond respectively to: number of output channels, number of input channels, height and width (and possibly depth) of the kernel. None where undefined. iiRRc3`s;|]1}tˆ|ˆ|ˆ|ˆ|ˆ|ƒVqdS(N(t get_conv_gradweights_shape_1axis(RR(RRRRttopshp(s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pys »sc3`s7|]-}tˆ|ˆ|ˆˆ|ˆ|ƒVqdS(N(R/(RR(RRRRR0(s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pys ¿sN(RRRRRR R!(R"t top_shapeRRRR%tnchanR&((RRRRR0s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pytget_conv_gradweights_shapes' cC`s»d|||||gkrdS|dks7|dkr;dS|dkrT||}nF|dkrm||}n-|dkrˆtdƒ‚n|d||}|dkr³||}n|dS( só This function tries to compute the image shape of convolution gradWeights. The weights shape can only be computed exactly when subsample is 1 and border_mode is not 'half'. If subsample is not 1 or border_mode is 'half', this function will return None. Parameters ---------- image_shape: int or None. Corresponds to the input image shape on a given axis. None if undefined. top_shape: int or None. Corresponds to the top shape on a given axis. None if undefined. border_mode: string or int. If it is a string, it must be 'valid', 'half' or 'full'. If it is an integer, it must correspond to the padding on the considered axis. subsample: int. It must correspond to the subsampling on the considered axis. dilation: int. It must correspond to the dilation on the considered axis. Returns ------- kernel_shape: int or None. Corresponds to the kernel shape on a given axis. None if undefined. iR(R)R*isborder_mode must be >= 0iN(RR+(R"R1RRR,R#((s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pyR/Ås        c`sÜ|d|d}‰|d|d}‰ˆdkrTtjtˆƒddƒ‰ntˆtƒršt‡‡‡‡‡fd†ttˆƒƒDƒƒ}n4t‡‡‡‡‡fd†ttˆƒƒDƒƒ}||f|S( sî This function tries to compute the image shape of convolution gradInputs. The image shape can only be computed exactly when subsample is 1. If subsample for a dimension is not 1, this function will return None for that dimension. Parameters ---------- kernel_shape: tuple of int (symbolic or numeric) corresponding to the kernel shape. Its four (or five) elements must correspond respectively to: number of output channels, number of input channels, height and width (and possibly depth) of the kernel. None where undefined. top_shape: tuple of int (symbolic or numeric) corresponding to the top image shape. Its four (or five) element must correspond respectively to: batch size, number of output channels, height and width (and possibly depth) of the image. None where undefined. border_mode: string, int (symbolic or numeric) or tuple of int (symbolic or numeric). If it is a string, it must be 'valid', 'half' or 'full'. If it is a tuple, its two (or three) elements respectively correspond to the padding on height and width (and possibly depth) axis. subsample: tuple of int (symbolic or numeric). Its two or three elements respectively correspond to the subsampling on height and width (and possibly depth) axis. filter_dilation: tuple of int (symbolic or numeric). Its two or three elements correspond respectively to the dilation on height and width axis. Returns ------- image_shape: tuple of int corresponding to the input image shape. Its four element must correspond respectively to: batch size, number of output channels, height and width of the image. None where undefined. iiiRRc3`s;|]1}tˆ|ˆ|ˆ|ˆ|ˆ|ƒVqdS(N(tget_conv_gradinputs_shape_1axis(RR(RRRRR0(s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pys $sc3`s7|]-}tˆ|ˆ|ˆˆ|ˆ|ƒVqdS(N(R4(RR(RRRRR0(s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pys (sN(RRRRRR R!(R#R1RRRR$R%R&((RRRRR0s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pytget_conv_gradinputs_shape÷s& cC`sàd|||||gkrdS|dkr/dS|d|d}|dkrZ|d}nO|dkrs|d}n6|dkrˆd}n!|}|dkr©tdƒ‚n|dkrÆ||d}n|d||d}|S( sµ This function tries to compute the image shape of convolution gradInputs. The image shape can only be computed exactly when subsample is 1. If subsample is not 1, this function will return None. Parameters ---------- kernel_shape: int or None. Corresponds to the kernel shape on a given axis. None if undefined. top_shape: int or None. Corresponds to the top shape on a given axis. None if undefined. border_mode: string or int. If it is a string, it must be 'valid', 'half' or 'full'. If it is an integer, it must correspond to the padding on the considered axis. subsample: int. It must correspond to the subsampling on the considered axis. dilation: int. It must correspond to the dilation on the considered axis. Returns ------- image_shape: int or None. Corresponds to the input image shape on a given axis. None if undefined. iR(iR)R*isborder_mode must be >= 0N(RR+(R#R1RRR,R-R.R"((s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pyR4.s&          c`sèt|ƒ}t|ƒ}t|ƒ}t|ƒt|ƒksTt|ƒt|ƒkrXtSt|ƒdt|ƒkrxtS|dk r¤t|ƒdt|ƒkr¤tSt|||||ƒ}d„‰t‡fd†t||ƒDƒƒS(sÑ This function checks if the given image shapes are consistent. Parameters ---------- image_shape: tuple of int (symbolic or numeric) corresponding to the input image shape. Its four (or five) element must correspond respectively to: batch size, number of input channels, height and width (and possibly depth) of the image. None where undefined. kernel_shape: tuple of int (symbolic or numeric) corresponding to the kernel shape. Its four (or five) elements must correspond respectively to: number of output channels, number of input channels, height and width (and possibly depth) of the kernel. None where undefined. output_shape: tuple of int (symbolic or numeric) corresponding to the output shape. Its four (or five) elements must correspond respectively to: batch size, number of output channels, height and width (and possibly depth) of the output. None where undefined. border_mode: string, int (symbolic or numeric) or tuple of int (symbolic or numeric). If it is a string, it must be 'valid', 'half' or 'full'. If it is a tuple, its two (or three) elements respectively correspond to the padding on height and width (and possibly depth) axis. subsample: tuple of int (symbolic or numeric). Its two or three elements respectively correspond to the subsampling on height and width (and possibly depth) axis. filter_dilation: tuple of int (symbolic or numeric). Its two or three elements correspond respectively to the dilation on height and width axis. Returns ------- Returns False if a convolution with the given input shape, kernel shape and parameters would not have produced the given output shape. Returns True in all other cases: if the given output shape matches the computed output shape, but also if the shape could not be checked because because the shape contains symbolic values. icS`sg|dks|dkrtSy2t|ƒ}t|ƒ}t|ƒt|ƒkSWntk rbtSXdS(N(RtTrueRRR(tgiventcomputed((s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pyt check_dim¡s   c3`s$|]\}}ˆ||ƒVqdS(N((RR7R8(R9(s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pys ¬sN(R RtFalseRR'talltzip(R"R#t output_shapeRRRtcomputed_output_shape((R9s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pytcheck_conv_gradinputs_shapehs)   0( cC`s0g}xt|ƒD]\}}yvt|ƒ}|dkrb|dkr‡td||fƒ‚q‡n%|dkr‡td||fƒ‚n|j|ƒWqtk r!|dkrétd|ƒ}|j||tjj|dƒƒƒq"td|ƒ}|j||tjj |dƒƒƒqXqWt |ƒS(s¼This function adds Assert nodes that check if shape is a valid convolution shape. The first two dimensions should be larger than or equal to zero. The convolution dimensions should be larger than zero. Parameters ---------- shape: tuple of int (symbolic or numeric) corresponding to the input, output or kernel shape of a convolution. For input and output, the first elements should should be the batch size and number of channels. For kernels, the first and second elements should contain the number of input and output channels. The remaining dimensions are the convolution dimensions. Returns ------- Returns a tuple similar to the given `shape`. For constant elements in `shape`, the function checks the value and raises a `ValueError` if the dimension is invalid. The elements that are not constant are wrapped in an `Assert` op that checks the dimension at run time. iisAThe convolution would produce an invalid shape (dim[%d]: %d < 0).sBThe convolution would produce an invalid shape (dim[%d]: %d <= 0).s=The convolution would produce an invalid shape (dim[%d] < 0).s>The convolution would produce an invalid shape (dim[%d] <= 0).( t enumerateRR+tappendRR ttheanottensortgetgtR (tshapet out_shapeRtntconst_nt assert_shp((s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pytassert_conv_shape°s"      (-sUnexpected shape.cC`s¡|dkstjjj r |S|j}g}xNt|jƒD]=}||dk r?|jtj j ||||ƒƒq?q?W|r™t |ƒ||ŒS|SdS(sWraps `x` in an `Assert` to check its shape. Parameters ---------- x : Tensor x will be wrapped in an `Assert`. expected_shape : tuple or list The expected shape of `x`. The size of a dimension can be None, which means it will not be checked. msg : str The error message of the `Assert`. Returns ------- Tensor `x` wrapped in an `Assert`. At execution time, this will throw an AssertionError if the shape of `x` does not match `expected_shape`. If `expected_shape` is None or contains only Nones, the function will return `x` directly. N( RRBtconfigtconvt assert_shapeRFR!tndimRARCteqR (txtexpected_shapetmsgRFttestsR((s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pyRNÚs +R*c C`sRt|ƒ}t|ƒ}td|d|d|d|d|d|ƒ}|||ƒS(s1This function will build the symbolic graph for convolving a mini-batch of a stack of 2D inputs with a set of 2D filters. The implementation is modelled after Convolutional Neural Networks (CNN). Refer to :func:`nnet.conv2d ` for a more detailed documentation. RRRRt filter_flipR(RtAbstractConv2d( tinputtfilterst input_shapet filter_shapeRRRURtconv_op((s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pytconv2dýs    c C`sRt|ƒ}t|ƒ}td|d|d|d|d|d|ƒ}|||ƒS(s This function will build the symbolic graph for convolving a mini-batch of a stack of 3D inputs with a set of 3D filters. The implementation is modelled after Convolutional Neural Networks (CNN). Parameters ---------- input: symbolic 5D tensor Mini-batch of feature map stacks, of shape (batch size, input channels, input depth, input rows, input columns). See the optional parameter ``input_shape``. filters: symbolic 5D tensor Set of filters used in CNN layer of shape (output channels, input channels, filter depth, filter rows, filter columns). See the optional parameter ``filter_shape``. input_shape: None, tuple/list of len 5 of int or Constant variable The shape of the input parameter. Optional, possibly used to choose an optimal implementation. You can give ``None`` for any element of the list to specify that this element is not known at compile time. filter_shape: None, tuple/list of len 5 of int or Constant variable The shape of the filters parameter. Optional, possibly used to choose an optimal implementation. You can give ``None`` for any element of the list to specify that this element is not known at compile time. border_mode: str, int or tuple of three int Either of the following: ``'valid'``: apply filter wherever it completely overlaps with the input. Generates output of shape: input shape - filter shape + 1 ``'full'``: apply filter wherever it partly overlaps with the input. Generates output of shape: input shape + filter shape - 1 ``'half'``: pad input with a symmetric border of ``filter // 2``, then perform a valid convolution. For filters with an odd number of slices, rows and columns, this leads to the output shape being equal to the input shape. ``int``: pad input with a symmetric border of zeros of the given width, then perform a valid convolution. ``(int1, int2, int3)`` pad input with a symmetric border of ``int1``, ``int2`` and ``int3`` columns, then perform a valid convolution. subsample: tuple of len 3 Factor by which to subsample the output. Also called strides elsewhere. filter_flip: bool If ``True``, will flip the filter x, y and z dimensions before sliding them over the input. This operation is normally referred to as a convolution, and this is the default. If ``False``, the filters are not flipped and the operation is referred to as a cross-correlation. filter_dilation: tuple of len 3 Factor by which to subsample (stride) the input. Also called dilation elsewhere. Returns ------- Symbolic 5D tensor Set of feature maps generated by convolutional layer. Tensor is is of shape (batch size, output channels, output depth, output rows, output columns) Notes ----- If cuDNN is available, it will be used on the GPU. Otherwise, it is the *Corr3dMM* convolution that will be used "caffe style convolution". This is only supported in Theano 0.8 or the development version until it is released. RRRRRUR(RtAbstractConv3d( RWRXRYRZRRRURR[((s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pytconv3dsX    c C`s†t|ƒ}t|ƒ}xBddgD]4}t||tjjttd ƒfƒs%t‚q%WxBddgD]4}t||tjj tjjtfƒsjt‚qjW|d k rüxKddddgD]4}t||tjjttd ƒfƒsÁt‚qÁWnt |ƒ} x:ddgD],}t||tjj ƒrd | |t‚|d|dd}||d|d9}||d|d|d9}|Std|jƒ‚dS(s6 Useful with the hack in profiling to print the MFlopsiiiis$flops not implemented for convdim={}N(R‘RatNotImplementedError(R˜tinptoutptinputsRXtoutputstflops((s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pyR£¼s  ic`s tstdƒ‚n|dkr9tdj|ƒƒ‚ntˆtƒr[ˆfˆj‰ntˆƒˆjkrŽtdjˆˆjƒƒ‚nt|j ˆj |dgˆjˆƒ}t j |d|j ƒ}ˆj ˆj t ‡‡‡fd†tˆjƒDƒƒ}t j |dˆj ƒ}ˆ|tdƒtdƒft ‡fd †tˆjƒDƒƒ<ˆjd krQt|ƒ} td ƒ} tjƒºtjd t jƒxŸt|j d ƒD]Š} xtˆj d ƒD]l} xct|j dƒD]N} || | dfct|| | df|| | dfd| | d ƒ7ésc3`s%|]}tddˆ|ƒVqdS(N(tsliceR(RR(R,(s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pys ísitfilltignorei.Nis*only 2D and 3D convolution are implemented(svalidsfull(timported_scipy_signalRžR+R“RRR‘RR'RFRtzerosRR R!R¤RR Rtwarningstcatch_warningst simplefiltertComplexWarningR RR(R˜timgRxtmodeR,RGtoutt dil_kern_shpt dilated_kerntvaltbvaltbRHtim0((R,RxR˜s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pyRMÐsR   ?   &'& (sconvdims border_modes subsamples filter_flipsimshpskshpsfilter_dilationN( t__name__t __module__t__doc__R:tcheck_broadcastt __props__RR6R›RR£RM(((s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pyR<s5 ?  t AbstractConvcB`sMeZdZddddedd„Zd„Zd„Zd„Zd„Z RS(s¯ Abstract Op for the forward convolution. Refer to :func:`BaseAbstractConv ` for a more detailed documentation. R*cC`sAtt|ƒjd|d|d|d|d|d|d|ƒdS(NR‘RRRRRUR(tsuperR»R›(R˜R‘RRRRRUR((s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pyR›s  cC`s[t|tjƒs!t|ƒ}nt|tjƒsBt|ƒ}n|jjd|jd|jƒ}|j|ƒ}|jj d|j kr¥t dd|j ƒ‚n|jj d|j krØt dd|j ƒ‚nt ||j dƒ}t ||jdƒ}|jd|jdgtg|j }|jjd|ƒƒ}t|||g|gƒS( NRt broadcastableisimg must be %dD tensorskern must be %dD tensorsGAbstractConv shape mismatch: shape of image does not match given imshp.sHAbstractConv shape mismatch: shape of filters does not match given kshp.i(RRBtVariableRR`tcloneRR½tfilter_variableROR‘R€RNRRR:R (R˜R­RxtktypeR½toutput((s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pyt make_nodes&      c`s«|\‰‰tjˆƒ‰tjˆƒ‰t‡‡fd†tˆjƒDƒƒ‰|\}ˆj‰tˆtƒr…tˆƒdkpŽˆdks©tdj ˆƒƒ‚nˆdkrÝt‡fd†tˆjƒDƒƒ‰n4ˆdkrt‡fd†tˆjƒDƒƒ‰ntˆtƒrèt‡fd †tˆjƒDƒƒ‰d‰tj ˆj dˆj d ft‡‡fd †tˆjƒDƒƒd ˆj ƒ}ˆ|t dƒt dƒft‡‡fd †tˆjƒDƒƒ<|‰nˆjs*ˆt dƒt dƒft dddƒfˆj‰nˆjˆˆdddˆjƒ}|t dƒt dƒft‡fd†tˆjƒDƒƒ}|jdjj|ƒ|d=siR*R)R(sginvalid border_mode {}, which must be either "valid", "full", "half", an integer or a tuple of integersc3`s|]}ˆ|dVqdS(iN((RR(t dil_kernshp(s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pys Jsc3`s|]}ˆ|dVqdS(iN((RR(RÄ(s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pys Lsc3`s|]}tˆ|ƒVqdS(N(R(RR(R®(s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pys Nsic3`s,|]"}ˆj|ddˆ|VqdS(iN(RF(RR(R­R.(s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pys QsRc3`s5|]+}tˆ|ˆj|dˆ|ƒVqdS(iN(R¤RF(RR(R­R.(s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pys UsiÿÿÿÿR®R,c3`s(|]}tddˆj|ƒVqdS(N(R¤RR(RR(R˜(s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pys \s(svalidsfullshalf(RtasarrayR R!R‘RRR•R+R“R¨RFRR¤RRURMRR¢R`tfilter(R˜RœRŸtout_totnew_imgtconv_out((RÄR­RxR®R.R˜s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pytperform9s@   !  ( (% B  9!cC`s©d}|ddk r:|j|d|dƒjd}n|ddk r¢|dkrz|j|d|dƒjd}q¢||j|d|dƒjd7}n|gS(Nii(RRÃR¢(R˜R¡t eval_pointstrval((s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pytR_opas$ $(cC`sô|d}|d}|jdk rpgtd|jƒD]0}|j|dkrZ||n |j|^q7}n|jdk rÌgtd|jƒD]0}|j|dkr¶||n |j|^q“}nt|||j|j|jƒ}|gS(Niii( RRR!R‘RR'RRR(R˜Rœt input_shapesRRRtres((s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pyt infer_shapels  JJN( R¶R·R¸RR6R›RÃRËRÎRÑ(((s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pyR» s   ( RVcB`s2eZdZddddedd„Zd„ZRS(s¯ Abstract Op for the forward convolution. Refer to :func:`BaseAbstractConv ` for a more detailed documentation. R*icC`sAtt|ƒjddd|d|d|d|d|d|ƒdS( NR‘iRRRRRUR(R¼RVR›(R˜RRRRRUR((s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pyR›‚s  cC`só|\}}|\}t|j|j|j|j|j|jƒ|||jddtƒ}t |j|j|j|j|j|jƒ|||jddtƒ}t ||j ƒ}|j j |ƒ}t ||j ƒ}|j j |ƒ}||fS(Niþÿÿÿtadd_assert_shape(RdRRRRRURRFR:RlRR½R`RÀ(R˜RŸtgradstbottomtweightsttoptd_bottomt d_weights((s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pytgrads&    N(ii(ii(R¶R·R¸RR6R›RÙ(((s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pyRV|s R]cB`s2eZdZddddedd„Zd„ZRS(s¯ Abstract Op for the forward convolution. Refer to :func:`BaseAbstractConv ` for a more detailed documentation. R*icC`sAtt|ƒjddd|d|d|d|d|d|ƒdS( NR‘iRRRRRUR(R¼R]R›(R˜RRRRRUR((s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pyR›´s  cC`sç|\}}|\}t|j|j|j|j|j|jƒ|||jdƒ}t|j|j|j|j|j|jƒ|||jdƒ}t ||j ƒ}|j j |ƒ}t ||j ƒ}|j j |ƒ}||fS(Niýÿÿÿ( RjRRRRRURRFRpRR½R`RÀ(R˜RŸRÓRÔRÕRÖR×RØ((s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pyRÙÂs&    N(iii(iii(R¶R·R¸RR6R›RÙ(((s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pyR]®s tAbstractConv_gradWeightscB`sPeZdZddddedd„Zed„Zd„Zd„Zd„Z RS(sTGradient wrt. filters for `AbstractConv`. Refer to :func:`BaseAbstractConv ` for a more detailed documentation. :note: You will not want to use this directly, but rely on Theano's automatic differentiation or graph optimization to use it as needed. R*cC`sAtt|ƒjd|d|d|d|d|d|d|ƒdS(NR‘RRRRRUR(R¼RÚR›(R˜R‘RRRRRUR((s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pyR›és  cC`s^t|tjƒs!t|ƒ}nt|tjƒsBt|ƒ}n|jjd|jd|jƒ}|j|ƒ}|jj d|j kr¥t dd|j ƒ‚n|jj d|j krØt dd|j ƒ‚n|röt ||j dƒ}nt|ƒ}|jd|jdgtg|j }|jjd|ƒƒ}t||||g|gƒS(NRR½isimg must be %dD tensorstopgrad must be %dD tensorsSAbstractConv_gradWeights shape mismatch: shape of image does not match given imshp.i(RRBR¾RR`R¿RR½RÀROR‘R€RNRR:R (R˜R­ttopgradRFRÒtgtypeR½RÂ((s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pyRÃùs&     c `sò|\‰}‰tjˆƒ‰tj|ƒ}|\}ˆj‰tˆtƒr`tˆƒdkpiˆdks„tdjˆƒƒ‚nt‡‡fd†tˆj ƒDƒƒ‰ˆdkràt‡fd†tˆj ƒDƒƒ‰n4ˆdkrt‡fd†tˆj ƒDƒƒ‰ntˆtƒrët‡fd †tˆj ƒDƒƒ‰d‰tj ˆj dˆj d ft‡‡fd †tˆj ƒDƒƒd ˆj ƒ}ˆ|t dƒt dƒft‡‡fd †tˆj ƒDƒƒ<|‰nt‡fd†tˆj ƒDƒƒr°|j d|j d ft‡‡fd†tˆj ƒDƒƒ}tj |d |j ƒ}||t dƒt dƒft‡fd†tˆj ƒDƒƒ<|}ndttdˆj dƒƒ} t dƒt dƒft dddƒfˆj } |j| ƒ| }ˆj| ƒ‰ˆjˆ|ddƒ} t‡fd†tˆj ƒDƒƒr£| t dƒt dƒft‡fd†tˆj ƒDƒƒ} nˆjrÂ| j| ƒ| } n| j| ƒ} |jdjj| ƒ|d!sc3`s|]}ˆ|dVqdS(iN((RR(t dil_shape(s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pys %sc3`s|]}ˆ|dVqdS(iN((RR(RÝ(s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pys 'sc3`s|]}tˆ|ƒVqdS(N(R(RR(R®(s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pys )sic3`s,|]"}ˆj|ddˆ|VqdS(iN(RF(RR(R­R.(s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pys -sRc3`s5|]+}tˆ|ˆj|dˆ|ƒVqdS(iN(R¤RF(RR(R­R.(s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pys 1sc3`s"|]}ˆj|dkVqdS(iN(R(RR(R˜(s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pys 5sc3`s,|]"}ˆj|dˆ|dVqdS(iiN(RF(RR(RÝR­(s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pys 7sc3`s(|]}tddˆj|ƒVqdS(N(R¤RR(RR(R˜(s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pys ;siiÿÿÿÿR®c3`s"|]}ˆj|dkVqdS(iN(R(RR(R˜(s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pys Esc3`s(|]}tddˆj|ƒVqdS(N(R¤RR(RR(R˜(s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pys Gs(svalidsfullshalf(ii(RRÅRRR R•R+R“R!R‘R¨RFRR¤Rtanyt transposeRMRUR¢R`RÆ( R˜RœRŸRÇRÛRÈRÉt new_shapet new_topgradt axes_ordert flip_filtersRx((RÝR­R®R.R˜RFs@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pyRËs\  !  ( (% B %?  % cC`sdgdgdggS(Nii((R˜Rœ((s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pytconnection_patternOscC`sË|d}|d}|jdk r-|jndgd|j}|d|dggt|jƒD]}|jd|^qb}gtd|jƒD]*}||dkr´||n||^q”}|gS(Niii(RRR‘R!R¡(R˜RœRÏRR0RRt fallback_kshp((s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pyRÑRs  -.AN( R¶R·R¸RR6R›RÃRËRäRÑ(((s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pyRÚßs   = RlcB`s2eZdZddddedd„Zd„ZRS(sVGradient wrt. filters for `AbstractConv2d`. Refer to :func:`BaseAbstractConv ` for a more detailed documentation. :note: You will not want to use this directly, but rely on Theano's automatic differentiation or graph optimization to use it as needed. R*icC`sAtt|ƒjddd|d|d|d|d|d|ƒdS( NR‘iRRRRRUR(R¼RlR›(R˜RRRRRUR((s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pyR›ks  c C`sú|d \}}|\}t|j|j|j|j|j|jƒ|||jdƒ}t|j|j|j|j|j|jƒ||ƒ}t ||j ƒ}|j j |ƒ}t ||j ƒ}|j j |ƒ}t jjƒƒf}||f|S(Niiþÿÿÿ(RdRRRRRURRFRVRR½R`RÀRBtgradienttDisconnectedType( R˜RŸRÓRÔRÖRÕR×td_toptd_height_width((s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pyRÙys*   N(ii(ii(R¶R·R¸RR6R›RÙ(((s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pyRlas  RpcB`s2eZdZddddedd„Zd„ZRS(sVGradient wrt. filters for `AbstractConv3d`. Refer to :func:`BaseAbstractConv ` for a more detailed documentation. :note: You will not want to use this directly, but rely on Theano's automatic differentiation or graph optimization to use it as needed. R*icC`sAtt|ƒjddd|d|d|d|d|d|ƒdS( NR‘iRRRRRUR(R¼RpR›(R˜RRRRRUR((s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pyR›¡s  c C`sú|d \}}|\}t|j|j|j|j|j|jƒ|||jdƒ}t|j|j|j|j|j|jƒ||ƒ}t ||j ƒ}|j j |ƒ}t ||j ƒ}|j j |ƒ}t jjƒƒf}||f|S(Niiýÿÿÿ(RjRRRRRURRFR]RR½R`RÀRBRæRç( R˜RŸRÓRÔRÖRÕR×Rètd_depth_height_width((s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pyRÙ¯s*   N(iii(iii(R¶R·R¸RR6R›RÙ(((s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pyRp—s  tAbstractConv_gradInputscB`sPeZdZddddedd„Zed„Zd„Zd„Zd„Z RS(sSGradient wrt. inputs for `AbstractConv`. Refer to :func:`BaseAbstractConv ` for a more detailed documentation. :note: You will not want to use this directly, but rely on Theano's automatic differentiation or graph optimization to use it as needed. R*cC`sAtt|ƒjd|d|d|d|d|d|d|ƒdS(NR‘RRRRRUR(R¼RëR›(R˜R‘RRRRRUR((s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pyR›Øs  cC`sdt|tjƒs!t|ƒ}nt|tjƒsBt|ƒ}n|jjd|jd|jƒ}|j|ƒ}|jj d|j kr¥t dd|j ƒ‚n|jj d|j krØt dd|j ƒ‚n|röt ||j dƒ}nt|ƒ}|jjd|jjdgtg|j }|jjd|ƒƒ}t||||g|gƒS( NRR½iskern must be %dD tensorstopgrad must be %dD tensorsSAbstractConv_gradInputs shape mismatch: shape of filters does not match given kshp.ii(RRBR¾RR`R¿RR½RÀROR‘R€RNRR:R (R˜RxRÛRFRÒRÜR½RÂ((s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pyRÃès&     !c`sb|\‰}‰tjˆƒ‰tj|ƒ}|\}ˆj‰tˆtƒr`tˆƒdkpiˆdks„tdjˆƒƒ‚nˆjdk rˆjndgdˆj }|j dˆj dggt ˆj ƒD]}ˆ|^qØ}gt dˆj ƒD]*}||dkr#||n||^q}t |ˆj ˆjˆjˆjƒ} t| ƒt|j ƒksœtdjt| ƒt|j ƒƒƒ‚nt‡‡fd †t ˆj ƒDƒƒ‰dˆj ‰ˆdkrt‡fd †t ˆj ƒDƒƒ‰nkˆdkr9t‡fd †t ˆj ƒDƒƒ‰n7tˆtƒrpt‡fd †t ˆj ƒDƒƒ‰nt‡fd †t ˆj ƒDƒƒr8|j d|j dft‡‡‡fd†t ˆj ƒDƒƒ} tj| d|jƒ} || tdƒtdƒft‡fd†t ˆj ƒDƒƒ<| }ndtt dˆj dƒƒ} tdƒtdƒftdddƒfˆj } ˆj| ƒ‰ˆjr¯|| }nˆj|ˆdddˆjƒ‰ˆjræˆ| ‰ntd„ˆDƒƒrAˆtdƒtdƒft‡‡fd†t ˆj ƒDƒƒ‰n|jdjjˆƒ|dsc3`s|]}ˆ|dVqdS(iN((RR(RÄ(s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pys #sc3`s|]}ˆ|dVqdS(iN((RR(RÄ(s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pys %sc3`s|]}ˆ|VqdS(N((RR(R®(s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pys 'sc3`s"|]}ˆj|dkVqdS(iN(R(RR(R˜(s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pys (sc3`s1|]'}ˆ|dˆ|ˆ|dVqdS(iiN((RR(RÄR.RF(s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pys *sRc3`s(|]}tddˆj|ƒVqdS(N(R¤RR(RR(R˜(s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pys .siÿÿÿÿR®R,cs`s|]}|dkVqdS(iN((Rtp((s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pys ;sc3`s5|]+}tˆ|ˆj|dˆ|ƒVqdS(iN(R¤RF(RR(R­R.(s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pys =s(svalidsfullshalf(i(ii(RRÅRRR R•R+R“RRR‘RFR!R'RRRÞR¨RR¤RßRURMR¢R`RÆ(R˜RœRŸRÇRÛRÈRRtfallback_imshptexpected_topgrad_shapeRàRáRâRã((RÄR­RxR®R.R˜RFs@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pyRËsj  ! -'A    ( ((%?    !  cC`sdgdgdggS(Nii((R˜Rœ((s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pyRäAscC`sË|d}|d}|jdk r-|jndgd|j}|d|dggt|jƒD]}|jd|^qb}gtd|jƒD]*}||dkr´||n||^q”}|gS(Niii(RRR‘R!R¡(R˜RœRÏRR0RRRí((s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pyRÑDs  -.AN( R¶R·R¸RR6R›RÃRËRäRÑ(((s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pyRëÍs   ? RdcB`s2eZdZddddedd„Zd„ZRS(sUGradient wrt. inputs for `AbstractConv2d`. Refer to :func:`BaseAbstractConv ` for a more detailed documentation. :note: You will not want to use this directly, but rely on Theano's automatic differentiation or graph optimization to use it as needed. R*icC`sAtt|ƒjddd|d|d|d|d|d|ƒdS( NR‘iRRRRRUR(R¼RdR›(R˜RRRRRUR((s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pyR›^s  c C`sú|d \}}|\}t|j|j|j|j|j|jƒ|||jdƒ}t|j|j|j|j|j|jƒ||ƒ}t ||j ƒ}|j j |ƒ}t ||j ƒ}|j j |ƒ}t jjƒƒf}||f|S(Niiþÿÿÿ(RlRRRRRURRFRVRR½R`RÀRBRæRç( R˜RŸRÓRÕRÖRÔRØRèRé((s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pyRÙls(  N(ii(ii(R¶R·R¸RR6R›RÙ(((s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pyRdSs  RjcB`s2eZdZddddedd„Zd„ZRS(sUGradient wrt. inputs for `AbstractConv3d`. Refer to :func:`BaseAbstractConv ` for a more detailed documentation. :note: You will not want to use this directly, but rely on Theano's automatic differentiation or graph optimization to use it as needed. R*icC`sAtt|ƒjddd|d|d|d|d|d|ƒdS( NR‘iRRRRRUR(R¼RjR›(R˜RRRRRUR((s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pyR›”s  c C`sú|d \}}|\}t|j|j|j|j|j|jƒ|||jdƒ}t|j|j|j|j|j|jƒ||ƒ}t ||j ƒ}|j j |ƒ}t ||j ƒ}|j j |ƒ}t jjƒƒf}||f|S(Niiýÿÿÿ(RpRRRRRURRFR]RR½R`RÀRBRæRç( R˜RŸRÓRÕRÖRÔRØRèRê((s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pyRÙ¢s& N(iii(iii(R¶R·R¸RR6R›RÙ(((s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pyRj‰s  (ii(ii(iii(iii(ii(ii(iii(iii(ii(ii(iii(iii(DR¸t __future__RRRtloggingtsixRRR–RBt theano.tensorRRRRttheano.tensor.optR t theano.gofR R t six.movesR R©RR‚tscipy.signal.signaltoolsR RRtscipy.signal.sigtoolsRR6R§t ImportErrorR:t __docformat__t getLoggert_loggerRR'RR3R/R5R4R?RKRNR\R^RiRkRoRqRyRtRRR»RVR]RÚRlRpRëRdRj(((s@/tmp/pip-build-X4mzal/theano/theano/tensor/nnet/abstract_conv.pytsš          3 7 6 4 5 < F * %  _ ƒ ƒ { w  # tÍs21‚66†6