1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
|
# Copyright (c) Microsoft Corporation.
# Licensed under the MIT License.
from __future__ import annotations
from typing import Tuple, Union
import numpy as np
from onnx import TensorProto
from onnx.helper import make_tensor
from onnxscript import script
from onnxscript.onnx_opset import opset15 as op
from onnxscript.onnx_types import INT32, INT64
from tests.common.onnx_script_test_case import FunctionTestParams
x = np.array([[0, 1, 2], [3, 4, 5], [6, 7, 8], [9, 10, 11]], dtype=np.int32)
zero = np.array(0, dtype=np.int64)
# Inputs/Outputs of test-cases are specified as numpy arrays, or tuples of numpy arrays,
# or as a list of values, e.g. [0, 1, 2], converted to an int32 numpy array.
IOType = Union[np.ndarray, Tuple[np.ndarray, ...], list]
def wrap_input_output(x: IOType) -> list(np.ndarray):
if isinstance(x, np.ndarray):
return [x]
elif isinstance(x, tuple):
return list(x)
else:
return [np.array(x, dtype=np.int32)]
def test(f, input: IOType, output: IOType) -> FunctionTestParams:
return FunctionTestParams(f, wrap_input_output(input), wrap_input_output(output))
@script(default_opset=op)
def get_first_row(A: INT32[...]) -> INT32[...]:
return A[0]
get_first_row_test = test(get_first_row, input=x, output=[0, 1, 2])
@script(default_opset=op)
def get_last_row(A: INT32[...]) -> INT32[...]:
return A[-1]
get_last_row_test = test(get_last_row, input=x, output=[9, 10, 11])
@script(default_opset=op)
def get_column(A: INT32[...]) -> INT32[...]:
return A[:, 1]
get_column_test = test(get_column, input=x, output=[1, 4, 7, 10])
@script(default_opset=op)
def get_unknown_row(A: INT32[...], i: INT64[...]) -> INT32[...]:
return A[i]
get_unknown_row_test = test(get_unknown_row, input=(x, zero), output=[0, 1, 2])
@script(default_opset=op)
def get_computed_row(A: INT32[...]) -> INT32[...]:
scalar_zero = op.Constant(value=make_tensor("scalar_zero", TensorProto.INT64, [], [0]))
return A[scalar_zero + 1]
get_computed_row_test = test(get_computed_row, input=x, output=[3, 4, 5])
@script(default_opset=op)
def slice_from_1_to_2(A: INT32[...]) -> INT32[...]:
return A[1:2]
slice_from_to = test(slice_from_1_to_2, input=x, output=[[3, 4, 5]])
@script(default_opset=op)
def slice_from_1(A: INT32[...]) -> INT32[...]:
return A[1:]
slice_from = test(slice_from_1, input=x, output=[[3, 4, 5], [6, 7, 8], [9, 10, 11]])
@script(default_opset=op)
def slice_to_2(A: INT32[...]) -> INT32[...]:
return A[:2]
slice_to = test(slice_to_2, input=x, output=[[0, 1, 2], [3, 4, 5]])
@script(default_opset=op)
def slice_step_minus1(A: INT32[...]) -> INT32[...]:
return A[::-1]
slice_neg_step = test(
slice_step_minus1, input=x, output=[[9, 10, 11], [6, 7, 8], [3, 4, 5], [0, 1, 2]]
)
@script(default_opset=op)
def slice_from_1_to_minus1(A: INT32[...]) -> INT32[...]:
return A[1:-1]
slice_from_neg_to = test(slice_from_1_to_minus1, input=x, output=[[3, 4, 5], [6, 7, 8]])
@script(default_opset=op)
def slice_to_0_step_minus1(A: INT32[...]) -> INT32[...]:
return A[:0:-1]
slice_to_neg_step = test(slice_to_0_step_minus1, input=x, output=x[:0:-1])
@script(default_opset=op)
def get_row_slice_column(A: INT32[...]) -> INT32[...]:
return A[0, :0:-1]
index_and_slice = test(get_row_slice_column, input=x, output=x[0, :0:-1])
@script(default_opset=op)
def slice_row_get_column(A: INT32[...]) -> INT32[...]:
return A[:2, 0]
slice_and_index = test(slice_row_get_column, input=x, output=x[:2, 0])
@script(default_opset=op)
def slice_row_and_column(A: INT32[...]) -> INT32[...]:
return A[:2, :1]
slice_and_slice = test(slice_row_and_column, input=x, output=x[:2, :1])
@script(default_opset=op)
def slice_from_2_to_0_step_minus1(A: INT32[...]) -> INT32[...]:
return A[2:0:-1]
slice_from_to_neg_step = test(slice_from_2_to_0_step_minus1, input=x, output=x[2:0:-1])
@script()
def slice_computed_range(A: INT32[...]) -> INT32[...]:
scalar_zero = op.Constant(value=make_tensor("scalar_zero", TensorProto.INT64, [], [0]))
return A[scalar_zero + 1 : scalar_zero + 2]
slice_computed_range_test = test(slice_computed_range, input=x, output=x[1:2])
@script(default_opset=op)
def nested_expr(A: INT32[...]) -> INT32[...]:
r = (A + 1)[0]
return r
nested_expr_test = test(nested_expr, input=x, output=[1, 2, 3])
# This notation is not possible with ONNX but is allowed by numpy.
# @script()
# def getitem_i_slice_right_step(A):
# r = A[1::-1]
# return r
|
