• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 /* Copyright 2016 The TensorFlow Authors. All Rights Reserved.
2 
3 Licensed under the Apache License, Version 2.0 (the "License");
4 you may not use this file except in compliance with the License.
5 You may obtain a copy of the License at
6 
7     http://www.apache.org/licenses/LICENSE-2.0
8 
9 Unless required by applicable law or agreed to in writing, software
10 distributed under the License is distributed on an "AS IS" BASIS,
11 WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12 See the License for the specific language governing permissions and
13 limitations under the License.
14 ==============================================================================*/
15 
16 #include "tensorflow/core/framework/common_shape_fns.h"
17 #include "tensorflow/core/framework/op.h"
18 #include "tensorflow/core/framework/shape_inference.h"
19 
20 namespace tensorflow {
21 
22 using shape_inference::InferenceContext;
23 using shape_inference::ShapeHandle;
24 
25 // --------------------------------------------------------------------------
ApplySdcaOptimizerShapeFn(InferenceContext * c)26 static Status ApplySdcaOptimizerShapeFn(InferenceContext* c) {
27   std::vector<ShapeHandle> sparse_handles;
28   if (c->input("sparse_weights", &sparse_handles).ok()) {
29     TF_RETURN_IF_ERROR(
30         c->set_output("out_delta_sparse_weights", sparse_handles));
31   }
32   std::vector<ShapeHandle> dense_handles;
33   if (c->input("dense_weights", &dense_handles).ok()) {
34     TF_RETURN_IF_ERROR(c->set_output("out_delta_dense_weights", dense_handles));
35   }
36   return c->set_output(
37       "out_example_state_data",
38       {c->Matrix(InferenceContext::kUnknownDim, c->MakeDim(4))});
39 }
40 
41 REGISTER_OP("SdcaOptimizer")
42     .Attr(
43         "loss_type: {'logistic_loss', 'squared_loss', 'hinge_loss',"
44         "'smooth_hinge_loss', 'poisson_loss'}")
45     .Attr("adaptative : bool=false")
46     .Attr("num_sparse_features: int >= 0")
47     .Attr("num_sparse_features_with_values: int >= 0")
48     .Attr("num_dense_features: int >= 0")
49     .Attr("l1: float")
50     .Attr("l2: float")
51     .Attr("num_loss_partitions: int >= 1")
52     .Attr("num_inner_iterations: int >= 1")
53     .Input("sparse_example_indices: num_sparse_features * int64")
54     .Input("sparse_feature_indices: num_sparse_features * int64")
55     .Input("sparse_feature_values: num_sparse_features_with_values * float")
56     .Input("dense_features: num_dense_features * float")
57     .Input("example_weights: float")
58     .Input("example_labels: float")
59     .Input("sparse_indices: num_sparse_features * int64")
60     .Input("sparse_weights: num_sparse_features * float")
61     .Input("dense_weights: num_dense_features * float")
62     .Input("example_state_data: float")
63     .Output("out_example_state_data: float")
64     .Output("out_delta_sparse_weights: num_sparse_features * float")
65     .Output("out_delta_dense_weights: num_dense_features * float")
66     .SetShapeFn(ApplySdcaOptimizerShapeFn);
67 
68 // The SdcaOptimizerV2 op fixes the "adaptative" typo in v1.
69 REGISTER_OP("SdcaOptimizerV2")
70     .Attr(
71         "loss_type: {'logistic_loss', 'squared_loss', 'hinge_loss',"
72         "'smooth_hinge_loss', 'poisson_loss'}")
73     .Attr("adaptive : bool=false")
74     .Attr("num_sparse_features: int >= 0")
75     .Attr("num_sparse_features_with_values: int >= 0")
76     .Attr("num_dense_features: int >= 0")
77     .Attr("l1: float")
78     .Attr("l2: float")
79     .Attr("num_loss_partitions: int >= 1")
80     .Attr("num_inner_iterations: int >= 1")
81     .Input("sparse_example_indices: num_sparse_features * int64")
82     .Input("sparse_feature_indices: num_sparse_features * int64")
83     .Input("sparse_feature_values: num_sparse_features_with_values * float")
84     .Input("dense_features: num_dense_features * float")
85     .Input("example_weights: float")
86     .Input("example_labels: float")
87     .Input("sparse_indices: num_sparse_features * int64")
88     .Input("sparse_weights: num_sparse_features * float")
89     .Input("dense_weights: num_dense_features * float")
90     .Input("example_state_data: float")
91     .Output("out_example_state_data: float")
92     .Output("out_delta_sparse_weights: num_sparse_features * float")
93     .Output("out_delta_dense_weights: num_dense_features * float")
94     .SetShapeFn(ApplySdcaOptimizerShapeFn);
95 
96 REGISTER_OP("SdcaShrinkL1")
97     .Attr("num_features: int >= 0")
98     .Attr("l1: float")
99     .Attr("l2: float")
100     .Input("weights: Ref(num_features * float)")
101     .SetShapeFn(shape_inference::UnknownShape);
102 
103 REGISTER_OP("SdcaFprint")
104     .Input("input: string")
105     .Output("output: int64")
__anonac679a370102(InferenceContext* c) 106     .SetShapeFn([](InferenceContext* c) {
107       ShapeHandle handle;
108       TF_RETURN_IF_ERROR(c->WithRank(c->input(0), 1, &handle));
109       ShapeHandle output_shape;
110       TF_RETURN_IF_ERROR(c->Concatenate(handle, c->Vector(2), &output_shape));
111       c->set_output(0, output_shape);
112       return Status::OK();
113     });
114 
115 }  // namespace tensorflow
116