/external/tensorflow/tensorflow/core/kernels/ |
D | depthwise_conv_grad_op.cc | 115 int64 out_rows = 0, out_cols = 0, pad_rows = 0, pad_cols = 0; \ 121 padding_, &out_cols, &pad_cols)); \ 128 context, output_cols == out_cols, \ 131 "actual = ", output_cols, ", computed = ", out_cols)); \ 144 args.out_cols = out_cols; \ 151 << ", output: [" << batch << ", " << out_rows << ", " << out_cols \ 193 const int64 out_cols = args.out_cols; in CopyOutputBackpropRegion() local 201 const int64 out_c_end = std::min(out_cols - 1, (in_c + pad_cols) / stride); in CopyOutputBackpropRegion() 224 out_backprop + (out_r * args.out_cols + out_c) * args.out_depth; in CopyOutputBackpropRegion() 412 args.out_rows * args.out_cols * args.out_depth; in operator ()() [all …]
|
D | depthwise_conv_op.cc | 94 const int64 base_output_index = (out_r * args.out_cols + out_c) * out_depth; in Run() 193 args.out_rows * args.out_cols * args.out_depth; in operator ()() 214 for (int64 out_c = 0; out_c < args.out_cols; ++out_c) { in operator ()() 236 const int64 shard_cost = kCostMultiplier * args.out_cols * args.out_depth; in operator ()() 348 int64 out_rows = 0, out_cols = 0, pad_rows = 0, pad_cols = 0; in Compute() local 354 padding_, &out_cols, &pad_cols)); in Compute() 356 ShapeFromFormat(data_format_, batch, out_rows, out_cols, out_depth); in Compute() 381 << "]; Output: [" << batch << ", " << out_rows << ", " << out_cols in Compute() 424 args.out_cols = out_cols; in Compute()
|
D | dilation_ops.cc | 68 int64* out_rows, int64* out_cols) { in ParseSizes() argument 111 padding, out_cols, pad_left)); in ParseSizes() 129 int64 out_rows = 0, out_cols = 0; in Compute() local 132 &out_cols); in Compute() 138 const std::vector<int64> out_sizes = {batch, out_rows, out_cols, depth}; in Compute() 228 int64 out_rows = 0, out_cols = 0; in Compute() local 231 &out_cols); in Compute() 240 out_cols == out_backprop.dim_size(2) && in Compute() 348 int64 out_rows = 0, out_cols = 0; in Compute() local 351 &out_cols); in Compute() [all …]
|
D | deep_conv2d.h | 80 int out_cols; member 93 out_cols(0), in Conv2DArgs() 102 int out_rows, int out_cols);
|
D | conv_ops.cc | 174 int out_cols, int out_depth, int dilation_rows, in Run() argument 180 in_depth, out_depth, out_rows, out_cols)) { in Run() 194 args.out_cols = out_cols; in Run() 215 int out_cols, int out_depth, int stride_rows, int stride_cols, in Run() argument 229 int out_cols, int out_depth, int dilation_rows, in Run() argument 409 int64 out_rows = 0, out_cols = 0; in ComputeConv2DDimension() local 415 &out_cols, &pad_cols_before, &pad_cols_after)); in ComputeConv2DDimension() 430 dimensions->out_cols = out_cols; in ComputeConv2DDimension() 467 dimensions.out_cols, dimensions.out_depth); in Compute() 498 dimensions.out_cols, dimensions.out_depth, dimensions.dilation_rows, in Compute() [all …]
|
D | extract_image_patches_op.cc | 86 int64 out_rows = 0, out_cols = 0; in Compute() local 93 padding_, &out_cols, &pad_cols)); in Compute() 95 const std::vector<int64> out_sizes = {batch, out_rows, out_cols, in Compute()
|
D | extract_volume_patches_op.cc | 109 int64 out_planes = 0, out_rows = 0, out_cols = 0; in Compute() local 119 padding_, &out_cols, &pad_cols)); in Compute() 122 batch, out_planes, out_rows, out_cols, in Compute()
|
D | mkl_conv_ops.h | 390 int64 out_rows = 0, out_cols = 0, out_planes = 0; variable 412 padding_type, &out_cols, &pad_left, &pad_right)); 422 padding_, &out_cols, &pad_left, &pad_right)); 447 ? ShapeFromFormat(data_format_, out_batch, out_rows, out_cols, 450 {{out_planes, out_rows, out_cols}}, out_depth); 459 mkldnn_sizes[MklDnnDims::Dim_W] = static_cast<int>(out_cols); 467 mkldnn_sizes[MklDnnDims3D::Dim3d_W] = static_cast<int>(out_cols);
|
D | deep_conv2d.cc | 50 int out_depth, int out_rows, int out_cols) { in GetDeepConvCost() argument 66 const int64 col_tiles = (out_cols + out_tile_cols - 1) / out_tile_cols; in GetDeepConvCost() 75 int out_depth, int out_rows, int out_cols) { in GetDirectConvCost() argument 76 return filter_rows * filter_cols * in_depth * out_depth * out_rows * out_cols; in GetDirectConvCost() 99 int out_rows, int out_cols) { in CanUseDeepConv2D() argument 117 t.output_shape().cols, in_depth, out_depth, out_rows, out_cols); in CanUseDeepConv2D() 119 filter_rows, filter_cols, in_depth, out_depth, out_rows, out_cols); in CanUseDeepConv2D() 800 out_c_start < 0 || out_c_start >= args.out_cols) { in operator ()() 813 if (out_c >= args.out_cols) continue; in operator ()() 823 args.out_depth * (out_r * args.out_cols + out_c) + od; in operator ()() [all …]
|
D | pooling_ops_3d_sycl.h | 35 const int out_rows, const int out_cols, in SYCL3DPoolParams() 52 out_cols_(out_cols), in SYCL3DPoolParams() 126 const int out_rows, const int out_cols, in MaxPool3DSYCL() argument 133 out_cols, window, stride, padding), in MaxPool3DSYCL() 188 const int out_cols = GetTensorDim(*output, data_format, '2'); 208 out_planes, out_rows, out_cols, window, stride, 532 const int out_rows, const int out_cols, 539 out_cols, window, stride, padding), 594 const int out_cols = GetTensorDim(*output, data_format, '2'); 614 out_planes, out_rows, out_cols, window, stride,
|
D | fractional_avg_pool_op.cc | 246 const int64 out_cols = out_backprop.dim_size(2); in Compute() local 277 out_cols * out_rows * out_batch); in Compute() 288 for (int64 c = 0; c < out_cols; ++c) { in Compute() 296 const int64 out_index = (b * out_rows + r) * out_cols + c; in Compute()
|
D | quantized_conv_ops.cc | 540 int64 out_rows = 0, out_cols = 0, pad_rows = 0, pad_cols = 0; in Compute() local 546 padding_, &out_cols, &pad_cols)); in Compute() 549 CHECK_GT(out_cols, 0); in Compute() 551 TensorShape out_shape({batch, out_rows, out_cols, out_depth}); in Compute() 564 padding_, output->flat<T3>().data(), out_rows, out_cols, in Compute()
|
D | conv_ops_using_gemm.cc | 520 int64 out_rows = 0, out_cols = 0, pad_rows = 0, pad_cols = 0; in Compute() local 526 padding_, &out_cols, &pad_cols)); in Compute() 528 ShapeFromFormat(data_format_, batch, out_rows, out_cols, out_depth); in Compute() 552 output->flat<T>().data(), out_rows, out_cols); in Compute()
|
D | depthwise_conv_op_gpu.h | 48 args.in_cols == args.out_cols && args.pad_rows >= 0 && in CanLaunchDepthwiseConv2dGPUSmall() 61 args.in_cols == args.out_cols && args.pad_rows >= 0 && in CanLaunchDepthwiseConv2dBackpropFilterGPUSmall() 91 const int out_width = args.out_cols; in DepthwiseConv2dGPUKernelNHWC() 331 const int out_width = args.out_cols; in DepthwiseConv2dGPUKernelNCHW() 637 const int num_outputs = args.out_rows * args.out_cols * block_count; in LaunchDepthwiseConv2dGPUSmall() 760 args.batch * args.out_rows * args.out_cols * args.out_depth; 833 const int out_width = args.out_cols; 903 const int out_width = args.out_cols; 1053 const int out_width = args.out_cols; 1337 const int out_width = args.out_cols; [all …]
|
D | conv_ops_3d.cc | 227 int64 out_cols = GetTensorDim(*output, data_format, '2'); in launch() local 235 0, (out_cols - 1) * strides[2] + filter_cols - in_cols); in launch() 360 .set_spatial_dim(DimIndex::X, out_cols) in launch() 400 {{out_planes, out_rows, out_cols}}, out_depth), in launch()
|
D | depthwise_conv_op.h | 40 int out_cols; member 55 out_cols(0), in DepthwiseArgs()
|
D | conv_ops_fused_impl.h | 594 const int64 out_cols = GetTensorDim(*output, params.data_format, 'W'); 619 0, (out_cols - 1) * dimensions.stride_cols + 706 .set_width(out_cols) 729 out_cols, out_depths), 906 dimensions.out_cols, dimensions.out_depth);
|
D | conv_ops_fused_image_transform.cc | 825 int64 out_rows = 0, out_cols = 0, pad_rows = 0, pad_cols = 0; in Compute() local 831 padding_, &out_cols, &pad_cols)); in Compute() 833 ShapeFromFormat(FORMAT_NHWC, batch, out_rows, out_cols, out_depth); in Compute() 861 output->flat<T>().data(), out_rows, out_cols, st, top_padding, in Compute()
|
D | conv_ops.h | 99 int64 out_cols;
|
/external/tensorflow/tensorflow/python/framework/ |
D | common_shapes.py | 225 out_rows, out_cols = get2d_conv_output_size(in_rows, in_cols, filter_rows, 229 output_shape = [batch_size, out_rows, out_cols, depth_out] 286 out_rows, out_cols = get2d_conv_output_size(in_rows, in_cols, filter_rows, 290 return [tensor_shape.TensorShape([batch_size, out_rows, out_cols, depth_out])] 349 out_rows, out_cols = get2d_conv_output_size(in_rows, in_cols, filter_rows, 353 return [tensor_shape.TensorShape([batch_size, out_rows, out_cols, depth_out])] 412 out_rows, out_cols = get2d_conv_output_size(in_rows, in_cols, ksize_r, 416 output_shape = [batch_size, out_rows, out_cols, depth] 484 out_rows, out_cols = get2d_conv_output_size(in_rows, in_cols, ksize_r, 487 output_shape = [batch_size, out_rows, out_cols, depth]
|
/external/tensorflow/tensorflow/core/kernels/neon/ |
D | neon_depthwise_conv_op.cc | 90 int64 out_rows = 0, out_cols = 0, pad_rows = 0, pad_cols = 0; in Compute() local 96 padding_, &out_cols, &pad_cols)); in Compute() 97 TensorShape out_shape({batch, out_rows, out_cols, out_depth}); in Compute() 115 << out_rows << ", " << out_cols << ", " << out_depth << "]"; in Compute()
|
/external/tensorflow/tensorflow/core/api_def/base_api/ |
D | api_def_ExtractVolumePatches.pbtxt | 12 5-D Tensor with shape `[batch, out_planes, out_rows, out_cols, 15 in the "depth" dimension. Note `out_planes`, `out_rows` and `out_cols`
|
D | api_def_ExtractImagePatches.pbtxt | 12 4-D Tensor with shape `[batch, out_rows, out_cols, ksize_rows * 15 `out_rows` and `out_cols` are the dimensions of the output patches.
|
D | api_def_Conv3DBackpropInput.pbtxt | 19 Backprop signal of shape `[batch, out_depth, out_rows, out_cols,
|
D | api_def_Conv3DBackpropFilter.pbtxt | 19 Backprop signal of shape `[batch, out_depth, out_rows, out_cols,
|