| /external/tensorflow/tensorflow/contrib/learn/python/learn/ops/ |
| losses_ops.py | 32 def mean_squared_error_regressor(tensor_in, labels, weights, biases, name=None):
35 [tensor_in, labels]):
36 predictions = nn.xw_plus_b(tensor_in, weights, biases)
44 def softmax_classifier(tensor_in,
58 tensor_in: Input tensor, [batch_size, feature_size], features.
71 with ops.name_scope(name, 'softmax_classifier', [tensor_in, labels]):
72 logits = nn.xw_plus_b(tensor_in, weights, biases)
|
| embeddings_ops.py | 69 def categorical_variable(tensor_in, n_classes, embedding_size, name):
73 tensor_in: Input tensor with class identifier (can be batch or
88 return embedding_lookup(embeddings, tensor_in)
|
| /external/tensorflow/tensorflow/core/kernels/ |
| cudnn_pooling_gpu.h | 45 TensorFormat data_format, const Tensor& tensor_in,
62 const Tensor* tensor_in, const Tensor* tensor_out,
|
| pooling_ops_common_gpu.h | 46 TensorFormat data_format, const Tensor& tensor_in,
61 TensorFormat data_format, const Tensor* tensor_in,
|
| cudnn_pooling_gpu.cc | 39 const Tensor& tensor_in, Tensor* output) {
40 const auto in_shape = tensor_in.shape();
43 const int64 in_batch = GetTensorDim(tensor_in, data_format, 'N');
44 const int64 in_features = GetTensorDim(tensor_in, data_format, 'C');
50 ShapeFromFormat(FORMAT_NCHW, tensor_in.shape(),
54 tensor_in.tensor<T, 5>(),
57 transformed_input = tensor_in;
86 GetTensorDim(tensor_in, data_format, '2' - i));
124 const Tensor* tensor_in, const Tensor* tensor_out, Tensor* input_backprop) {
126 (tensor_in && tensor_out)
[all...] |
| quantized_pooling_ops.cc | 56 const Tensor& tensor_in = context->input(0); variable
58 padding_, FORMAT_NHWC, tensor_in.shape()};
70 OP_REQUIRES(context, tensor_in.dims() == 4,
71 errors::InvalidArgument("tensor_in must be 4-dimensional"));
83 Tensor int32_input(DT_INT32, tensor_in.shape());
84 int32_input.flat<int32>() = tensor_in.flat<T>().template cast<int32>();
|
| pooling_ops_3d.cc | 58 // For maxpooling, tensor_in should have 4 dimensions.
60 errors::InvalidArgument("tensor_in must be 4-dimensional"));
102 static void launch(OpKernelContext* context, const Tensor& tensor_in,
109 Eigen::CuboidAvgPooling(tensor_in.tensor<T, 5>(), window[0], window[1],
117 static void launch(OpKernelContext* context, const Tensor& tensor_in,
124 Eigen::CuboidMaxPooling(tensor_in.tensor<T, 5>(), window[0], window[1],
167 const Tensor& tensor_in = context->input(0); variable
169 OP_REQUIRES(context, tensor_in.dims() == 5,
170 errors::InvalidArgument("tensor_in must be 5-dimensional"));
171 const int64 depth = GetTensorDim(tensor_in, data_format_, 'C')
340 const Tensor& tensor_in = context->input(0); variable
684 const Tensor& tensor_in = context->input(0); variable
[all...] |
| pooling_ops_common.cc | 36 // For maxpooling, tensor_in should have 2 spatial dimensions.
145 Padding padding, TensorFormat data_format, const Tensor& tensor_in,
150 if (tensor_in.shape().num_elements() == 0) {
155 padding, data_format, tensor_in.shape()};
166 ShapeFromFormat(FORMAT_NCHW, tensor_in.shape(),
170 tensor_in.tensor<T, 4>(),
173 transformed_input = tensor_in;
242 Padding padding, TensorFormat data_format, const Tensor* tensor_in,
246 (tensor_in && tensor_out))
247 << "For MaxPoolGrad, both tensor_in and tensor_out needs to be
[all...] |
| maxpooling_op.cc | 58 Tensor* input_backprop, const Tensor& tensor_in, const Tensor& out_backprop,
68 tensor_in.flat<T>().data(), params.depth,
233 const Tensor& tensor_in = context->input(0); variable
237 // For maxpooling, tensor_in should have 4 dimensions.
238 OP_REQUIRES(context, tensor_in.dims() == 4,
239 errors::InvalidArgument("tensor_in must be 4-dimensional"));
246 const TensorShape& output_shape = tensor_in.shape();
285 padding_, FORMAT_NHWC, tensor_in.shape()};
295 context, &tensor_out_dup, &tensor_out_arg_max, output, tensor_in,
311 const std::vector<int32>& stride, Padding padding, const Tensor* tensor_in,
365 const Tensor& tensor_in = context->input(0); variable
471 const Tensor& tensor_in = context->input(0); variable
668 const Tensor& tensor_in = context->input(0); variable
767 const Tensor& tensor_in = context->input(0); variable
822 const Tensor& tensor_in = context->input(0); variable
895 const Tensor& tensor_in = context->input(0); variable
951 const Tensor& tensor_in = context->input(0); variable
1001 const Tensor& tensor_in = context->input(0); variable
1060 const Tensor& tensor_in = context->input(0); variable
1146 const Tensor& tensor_in = context->input(0); variable
[all...] |
| pooling_ops_common.h | 110 const Tensor& tensor_in = context->input(0); variable
112 padding_, FORMAT_NHWC, tensor_in.shape()};
133 DepthwiseMaxPool(context, output, tensor_in, params);
135 SpatialMaxPool(context, output, tensor_in, params, padding_);
147 const Tensor& tensor_in, const PoolParameters& params) {
149 in_by_pool(tensor_in.flat<T>().data(), params.depth_window,
150 tensor_in.NumElements() / params.depth_window);
157 const Tensor& tensor_in, const PoolParameters& params,
168 tensor_in.tensor<T, 4>(), params.window_rows, params.window_cols,
176 ConstEigenMatrixMap in_mat(tensor_in.flat<T>().data(), params.depth
323 const Tensor& tensor_in = context->input(0); variable
[all...] |
| fractional_max_pool_op.cc | 78 const Tensor& tensor_in = context->input(0); variable
79 OP_REQUIRES(context, tensor_in.dims() == tensor_in_and_out_dims,
80 errors::InvalidArgument("tensor_in must be 4-dimensional"));
85 input_size[i] = tensor_in.dim_size(i);
125 ConstEigenMatrixMap in_mat(tensor_in.flat<T>().data(), input_size[3],
230 const Tensor& tensor_in = context->input(0); variable
241 input_size[i] = tensor_in.dim_size(i);
260 tensor_in.flat<T>().data(), input_size[3],
339 {0}, 0, tensor_in.shape(), &output));
|
| pooling_ops_3d_sycl.h | 179 static void launch(OpKernelContext* context, const Tensor& tensor_in,
189 const int batch = GetTensorDim(tensor_in, data_format, 'N');
190 const int in_planes = GetTensorDim(tensor_in, data_format, '0');
191 const int in_rows = GetTensorDim(tensor_in, data_format, '1');
192 const int in_cols = GetTensorDim(tensor_in, data_format, '2');
193 const int depth = GetTensorDim(tensor_in, data_format, 'C');
198 device.get_sycl_buffer(tensor_in.template flat<T>().data());
347 static void launch(OpKernelContext* context, const Tensor& tensor_in,
355 const int batch = GetTensorDim(tensor_in, data_format, 'N');
356 const int in_planes = GetTensorDim(tensor_in, data_format, '0')
[all...] |
| avgpooling_op.cc | 77 const Tensor& tensor_in = context->input(0); variable
79 padding_, data_format_, tensor_in.shape()};
87 // For avgpooling, tensor_in should have 4 dimensions.
88 OP_REQUIRES(context, tensor_in.dims() == 4,
89 errors::InvalidArgument("tensor_in must be 4-dimensional"));
95 SpatialAvgPool<Device, T>(context, output, tensor_in, params, padding_);
139 const Tensor& tensor_in = context->input(0); variable
141 padding_, data_format_, tensor_in.shape()};
149 // For avgpooling, tensor_in should have 4 dimensions.
150 OP_REQUIRES(context, tensor_in.dims() == 4
[all...] |
| mkl_maxpooling_op.cc | 78 const Tensor& tensor_in = MklGetInput(context, 0); variable
86 tensor_in.shape());
134 const_cast<void*>(static_cast<const void*>(tensor_in.flat<T>().data()));
235 const Tensor& tensor_in = MklGetInput(context, 0); variable
245 mkl_context.params.in_dim = tensor_in.dims();
252 tensor_in.shape());
389 const Tensor& tensor_in = MklGetInput(context, 0); local
403 tensor_in.flat<T>().data())),
412 static_cast<const void*>(tensor_in.flat<T>().data())),
418 static_cast<const void*>(tensor_in.flat<T>().data()))
[all...] |
| fractional_avg_pool_op.cc | 75 const Tensor& tensor_in = context->input(0); variable
76 OP_REQUIRES(context, tensor_in.dims() == tensor_in_and_out_dims,
77 errors::InvalidArgument("tensor_in must be 4-dimensional"));
82 input_size[i] = tensor_in.dim_size(i);
119 ConstEigenMatrixMap in_mat(tensor_in.flat<T>().data(), input_size[3],
|
| mkl_avgpooling_op.cc | 70 const Tensor& tensor_in = MklGetInput(context, 0); variable
75 mkl_context.params.in_dim = tensor_in.dims();
82 tensor_in.shape());
106 static_cast<void*>(const_cast<T*>(tensor_in.flat<T>().data())),
113 static_cast<void*>(const_cast<T*>(tensor_in.flat<T>().data())),
119 static_cast<void*>(const_cast<T*>(tensor_in.flat<T>().data()));
|
