OneHot - 9 vs 11#

Next section compares an older to a newer version of the same operator after both definition are converted into markdown text. Green means an addition to the newer version, red means a deletion. Anything else is unchanged.

Files changed (1) hide show
  1. OneHot9 → OneHot11 +10 -17
OneHot9 → OneHot11 RENAMED
@@ -1 +1 @@
1
1
  Produces a one-hot tensor based on inputs.
2
2
  The locations represented by the index values in the 'indices' input tensor will have 'on_value'
3
3
  and the other locations will have 'off_value' in the output tensor, where 'on_value' and 'off_value'
4
4
  are specified as part of required input argument 'values', which is a two-element tensor of format
5
5
  [off_value, on_value]. The rank of the output tensor will be one greater than the rank of the
6
6
  input tensor. The additional dimension is for one-hot representation. The additional dimension will
7
7
  be inserted at the position specified by 'axis'. If 'axis' is not specified then then additional
8
8
  dimension will be inserted as the innermost dimension, i.e. axis=-1. The size of the additional
9
9
  dimension is specified by required scalar input 'depth'. The type of the output tensor is the same
10
10
  as the type of the 'values' input. Any entries in the 'indices' input tensor with values outside
11
- the range [-depth, depth-1] will result in one-hot representation with all 'off_value' values in the
11
+ the range [0, depth) will result in one-hot representation with all 'off_value' values in the
12
12
  output tensor.
13
-
14
- when axis = 0:
15
- output[input[i, j, k], i, j, k] = 1 for all i, j, k and 0 otherwise.
16
-
17
- when axis = -1:
18
- output[i, j, k, input[i, j, k]] = 1 for all i, j, k and 0 otherwise.
19
13
  **Attributes**
20
14
  * **axis**:
21
15
  (Optional) Axis along which one-hot representation in added.
22
16
  Default: axis=-1. axis=-1 means that the additional dimension will
23
17
  be inserted as the innermost/last dimension in the output tensor.
24
- Negative value means counting dimensions from the back. Accepted
25
- range is [-r-1, r] where r = rank(indices).
26
18
  **Inputs**
27
19
  * **indices** (heterogeneous) - **T1**:
28
- Input tensor containing indices. Any entries in the 'indices' input
20
+ Input tensor containing indices. The values must be non-negative
29
- tensor with values outside the range [-depth, depth-1] will result
30
- in one-hot representation with all 'off_value' values in the output
31
- tensor.In case 'indices' is of non-integer type, the values will be
21
+ integers. Any entries in the 'indices' input tensor with values
22
+ outside the range [0, depth) will result in one-hot representation
23
+ with all 'off_value' values in the output tensor.In case 'indices'
32
- casted to int64 before use.
24
+ is of non-integer type, the values will be casted to int64 before
25
+ use.
33
26
  * **depth** (heterogeneous) - **T2**:
34
27
  Scalar specifying the number of classes in one-hot tensor. This is
35
28
  also the size of the one-hot dimension (specified by 'axis'
36
29
  attribute) added on in the output tensor. The values in the
37
- 'indices' input tensor are expected to be in the range [-depth,
30
+ 'indices' input tensor are expected to be in the range [0, depth).
38
- depth-1]. In case 'depth' is of non-integer type, it will be casted
31
+ In case 'depth' is of non-integer type, it will be casted to int64
39
- to int64 before use.
32
+ before use.
40
33
  * **values** (heterogeneous) - **T3**:
41
34
  Rank 1 tensor containing exactly two elements, in the format
42
35
  [off_value, on_value], where 'on_value' is the value used for
43
36
  filling locations specified in 'indices' input tensor, and
44
37
  'off_value' is the value used for filling locations other than those
45
38
  specified in 'indices' input tensor.
46
39
  **Outputs**
47
40
  * **output** (heterogeneous) - **T3**:
48
41
  Tensor of rank one greater than input tensor 'indices', i.e.
49
42
  rank(output) = rank(indices) + 1. The data type for the elements of
50
43
  the output tensor is the same as the type of input 'values' is used.
51
44
  **Type Constraints**
52
45
  * **T1** in (
53
46
  tensor(double),
54
47
  tensor(float),
55
48
  tensor(float16),
56
49
  tensor(int16),
57
50
  tensor(int32),
58
51
  tensor(int64),
59
52
  tensor(int8),
60
53
  tensor(uint16),
61
54
  tensor(uint32),
62
55
  tensor(uint64),
63
56
  tensor(uint8)
64
57
  ):
65
58
  Constrain input to only numeric types.
66
59
  * **T2** in (
67
60
  tensor(double),
68
61
  tensor(float),
69
62
  tensor(float16),
70
63
  tensor(int16),
71
64
  tensor(int32),
72
65
  tensor(int64),
73
66
  tensor(int8),
74
67
  tensor(uint16),
75
68
  tensor(uint32),
76
69
  tensor(uint64),
77
70
  tensor(uint8)
78
71
  ):
79
72
  Constrain input to only numeric types.
80
73
  * **T3** in (
81
74
  tensor(bool),
82
75
  tensor(complex128),
83
76
  tensor(complex64),
84
77
  tensor(double),
85
78
  tensor(float),
86
79
  tensor(float16),
87
80
  tensor(int16),
88
81
  tensor(int32),
89
82
  tensor(int64),
90
83
  tensor(int8),
91
84
  tensor(string),
92
85
  tensor(uint16),
93
86
  tensor(uint32),
94
87
  tensor(uint64),
95
88
  tensor(uint8)
96
89
  ):
97
90
  Constrain to any tensor type.