o 2hy8@sddlZddlZddlmZddlmZddlmZddlm Z ddZ dd Z dd d Z  d ddZ ddZddZddZddZd!ddZedddZdS)"N)tree) keras_exportcanonicalize_axis)to_tuple_or_listc Cs,t|}t|}|}|}t|t|kr!dgt|t||}t|t|kr6dgt|t||}t|}tt|D]S}||dkrO||||<q@||durd||dkr]dn||||<q@||dksx||dusx||||kr||||<q@td||d||d|d|d |S)aSBroadcast input shapes to a unified shape. Convert to list for mutability. Args: shape1: A tuple or list of integers. shape2: A tuple or list of integers. Returns: output_shape (list of integers or `None`): The broadcasted shape. Example: >>> broadcast_shapes((5, 3), (1, 3)) [5, 3] Nz2Cannot broadcast shape, the failure dim has value z!, which cannot be broadcasted to z. Input shapes are:  and .)listlenrange ValueError)shape1shape2 origin_shape1 origin_shape2 output_shapeirX/var/www/html/chatgem/venv/lib/python3.10/site-packages/keras/src/ops/operation_utils.pybroadcast_shapes s8  (rcsjt|}dur t|ttt|fddDt|fddtD}t|S)aCompute the output shape for the `expand_dims` operation. Args: input_shape: Input shape. axis: int or sequence of ints for the axis to expand. Returns: Tuple of ints: The output shape after the `expand_dims` operation. Ncsg|]}t|qSrr).0a)out_ndimrr Fsz4compute_expand_dims_output_shape..cs g|] }|vr dntqSr)nextrax)axis shape_iterrrrHs)r r riterr tuple) input_shaper new_shaper)rrr r compute_expand_dims_output_shape7s  r%valid channels_lastc Csp|dur|n|}t|}t|}|dkr|dd}n|dd}g}tt|D]}||dur;d||<||q*t|}|dkrot|||d} tt| D]}||vrm| |dkrmtd|d |d qVn|d krt|d|d} ntd |d d| D} |D]}d| |<qt| } |dkr|df| |df} | S|d|df| } | S)a3Computes the output shape of pooling operations. Args: input_shape: Input shape. Must be a tuple of integers. pool_size: Size of the pooling operation. Must be a tuple of integers. strides: Stride of the pooling operation. Must be a tuple of integers. Defaults to `pool_size`. padding: Padding method. Available methods are `"valid"` or `"same"`. Defaults to `"valid"`. data_format: String, either `"channels_last"` or `"channels_first"`. The ordering of the dimensions in the inputs. `"channels_last"` corresponds to inputs with shape `(batch, height, width, channels)` while `"channels_first"` corresponds to inputs with shape `(batch, channels, height, weight)`. Defaults to `"channels_last"`. Returns: Tuple of ints: The output shape of the pooling operation. Examples: # Basic usage with square pooling on a single image >>> compute_pooling_output_shape((1, 4, 4, 1), (2, 2)) (1, 2, 2, 1) # Strided pooling on a single image with strides different from pool_size >>> compute_pooling_output_shape((1, 4, 4, 1), (2, 2), strides=(1, 1)) (1, 3, 3, 1) # Pooling on a batch of images >>> compute_pooling_output_shape((32, 4, 4, 3), (2, 2)) (32, 2, 2, 3) Nr'rr&rz@Computed output size would be negative. Received: `inputs.shape=z` and `pool_size=`.samezGArgument `padding` must be either 'valid' or 'same'. Received: padding=cSg|]}t|qSrintrrrrrrz0compute_pooling_output_shape..) r nparrayr r appendfloorr r") r# pool_sizestridespadding data_formatinput_shape_origin spatial_shape none_dimsroutput_spatial_shaperrrrcompute_pooling_output_shapeNsh'       r=rc Cs$|dkr|dd}||d|f}n|dd}||d|f}t|t|kr4td|d|dt|tr@|ft|}t|trL|ft|}t|t|kr_td |d |dg} t|}tt|D]} || dur}d|| <| | qlt|dd } t|}|d krt||| dd|d} tt| D]} | | vr| | d krtd|d|d|dqn|dks|dkrt|d|d} ntd|ddd| D} | D]} d| | <qt | } |dkr|d f| |df} | S|d |df| } | S)z%Compute the output shape of conv ops.r'rr(r)NzNKernel shape must have the same length as input, but received kernel of shape z and input of shape r zlDilation must be None, scalar or tuple/list of length of inputs' spatial shape, but received `dilation_rate=z` and input of shape r&rz?Computed output size would be negative. Received `inputs shape=z`, `kernel shape=z`, `dilation_rate=r*r+causalz9`padding` must be either `'valid'` or `'same'`. Received cSr,rr-r/rrrrr0z-compute_conv_output_shape..) r r isinstancer.r1r2r r3r4r")r#filters kernel_sizer6r7r8 dilation_rater: kernel_shaper;rkernel_spatial_shaper<rrrrcompute_conv_output_shapes          rFcCst|dkr d|df}t|dkr|ddf}|ddur7|ddur7|d|dkr7td|d|dt|dd|dd}|d|dg}||}t|dkrY|d=t|dkrb|d=t|S) zCompute the output shape of a `matmul` operation. Args: shape1: Shape of the left operand. shape2: Shape of the right operand. Returns: Tuple of ints: The output shape for the `matmul` operation. rrr(Nr>z[Inner dimensions (`x1.shape[-1]` and `x2.shape[-2]`) must be equal, but received `x1.shape=z` and `x2.shape=r*)r r rr")rr leading_shapelast_2_dims_shaperrrrcompute_matmul_output_shapes,       rIc Cs|d}|dkrtd|d|d|dd|vr$tdd |DSt|}|d krC|t|krAtd |d |d||Sd}d}t|D]\}}|dkrV|}qK||9}qK|d kse||d krutd |d|d |d|t|} ||| |<t| S)zConverts `-1` in `newshape` to either an actual dimension or `None`. This utility does not special case the 0th dimension (batch size). r(rz4There must be at most one unknown dimension (-1) in z . Received: =r Ncss |] }|dkr |ndVqdS)r(Nr)rdimrrr (sz/compute_reshape_output_shape..rzFThe total size of the tensor must be unchanged. Received: input_shape=z, zyThe total size of the tensor must be unchanged, however, the input size cannot by divided by the specified dimensions in z. Received: input_shape=)countr r"mathprod enumerater ) r#newshapenewshape_arg_nameunknown_dim_count input_sizeknown_output_sizeunknown_dim_indexindexrKrrrrcompute_reshape_output_shapes^    rXcsft|durtdddSt|tkr(tdtdt|dtfdd|DS)a Compute the output shape for the `transpose` operation. Args: input_shape: Input shape. axes: Permutation of the dimensions for the `transpose` operation. Returns: Tuple of ints: The output shape after the `transpose` operation. Nr(zDaxis must be a list of the same length as the input shape, expected z, but received r c3s|]}|VqdSNrrr#rrrL]z1compute_transpose_output_shape..)r r"r r )r#axesrrZrcompute_transpose_output_shapeJs r]cCsvt|}t|}|durd|vrdgntt|g}t|t|kr.td|d|d||||<t||}|S)NzM`x` and `indices` must have the same number of dimensions, but receive shape rr )r r.r1rOr r r)r# indices_shaperrrrr$compute_take_along_axis_output_shape`s   r_FcCs|t|}|dur|rtdd|DStgSt|tr|f}|r.|D]}d||<q#t|St|ddD]}||=q4t|S)NcSsg|]}dqSrr)r_rrrrwsz reduce_shape..rT)reverse)r r"r@r.sorted)shaperkeepdimsrrrr reduce_shapess  rezkeras.utils.get_source_inputscst|ds|S|j\}}}|r|js|gS|j|}|jr#t|jSg}|jD]}t|}|D]t fdd|DrB| q0q(|S)zReturns the list of input tensors necessary to compute `tensor`. Output will always be a list of tensors (potentially with 1 element). Args: tensor: The tensor to start from. Returns: List of input tensors. _keras_historyc3s|]}|uVqdSrYr)rtxrrrLr[z$get_source_inputs..) hasattrrf_inbound_nodesis_inputrflattenoutput_tensors input_tensorsget_source_inputsallr3)tensor operation node_indexr`nodesource_tensorsprevious_sourcesrrhrrps"       rp)r&r')rr&r'r)NF)rNnumpyr1 keras.srcrkeras.src.api_exportr&keras.src.backend.common.backend_utilsrrrr%r=rFrIrXr]r_rerprrrrs.    , ] P#0