2 # Copyright (C) 2018 The Android Open Source Project
4 # Licensed under the Apache License, Version 2.0 (the "License");
5 # you may not use this file except in compliance with the License.
6 # You may obtain a copy of the License at
8 # http://www.apache.org/licenses/LICENSE-2.0
10 # Unless required by applicable law or agreed to in writing, software
11 # distributed under the License is distributed on an "AS IS" BASIS,
12 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13 # See the License for the specific language governing permissions and
14 # limitations under the License.
18 TestCase = collections.namedtuple("TestCase", [
19 "inp", "inp_data", "begin", "begin_data", "size", "size_data", "output",
25 inp=Input("input", "TENSOR_FLOAT32", "{4}"),
26 inp_data=[1, 2, 3, 4],
27 begin=Input("begin", "TENSOR_INT32", "{1}"),
29 size=Input("size", "TENSOR_INT32", "{1}"),
31 output=Output("output", "TENSOR_FLOAT32", "{2}"),
34 inp=Input("input", "TENSOR_FLOAT32", "{2,3}"),
35 inp_data=[1, 2, 3, 4, 5, 6],
36 begin=Input("begin", "TENSOR_INT32", "{2}"),
38 size=Input("size", "TENSOR_INT32", "{2}"),
40 output=Output("output", "TENSOR_FLOAT32", "{1, 2}"),
43 inp=Input("input", "TENSOR_FLOAT32", "{2,3,2}"),
44 inp_data=[1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12],
45 begin=Input("begin", "TENSOR_INT32", "{3}"),
47 size=Input("size", "TENSOR_INT32", "{3}"),
49 output=Output("output", "TENSOR_FLOAT32", "{2, 3, 2}"),
50 output_data=[1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12]),
52 inp=Input("input", "TENSOR_FLOAT32", "{4, 1, 1, 1}"),
53 inp_data=[1, 2, 3, 4],
54 begin=Input("begin", "TENSOR_INT32", "{4}"),
55 begin_data=[1, 0, 0, 0],
56 size=Input("size", "TENSOR_INT32", "{4}"),
57 size_data=[3, 1, 1, 1],
58 output=Output("output", "TENSOR_FLOAT32", "{3, 1, 1, 1}"),
59 output_data=[2, 3, 4]),
61 inp=Input("input", "TENSOR_INT32", "{3, 2, 3, 1}"),
62 inp_data=[1, 1, 1, 2, 2, 2, 3, 3, 3, 4, 4, 4, 5, 5, 5, 6, 6, 6],
63 begin=Input("begin", "TENSOR_INT32", "{4}"),
64 begin_data=[1, 0, 0, 0],
65 size=Input("size", "TENSOR_INT32", "{4}"),
66 size_data=[1, 1, 3, 1],
67 output=Output("output", "TENSOR_INT32", "{1, 1, 3, 1}"),
68 output_data=[3, 3, 3]),
70 inp=Input("input", "TENSOR_INT32", "{3, 2, 3, 1}"),
71 inp_data=[1, 1, 1, 2, 2, 2, 3, 3, 3, 4, 4, 4, 5, 5, 5, 6, 6, 6],
72 begin=Input("begin", "TENSOR_INT32", "{4}"),
73 begin_data=[1, 0, 0, 0],
74 size=Input("size", "TENSOR_INT32", "{4}"),
75 size_data=[2, 1, 3, 1],
76 output=Output("output", "TENSOR_INT32", "{2, 1, 3, 1}"),
77 output_data=[3, 3, 3, 5, 5, 5]),
79 inp=Input("input", "TENSOR_QUANT8_ASYMM", "{3, 2, 3, 1}, 2.0, 128"),
80 inp_data=[1, 1, 1, 2, 2, 2, 3, 3, 3, 4, 4, 4, 5, 5, 5, 6, 6, 6],
81 begin=Input("begin", "TENSOR_INT32", "{4}"),
82 begin_data=[1, 0, 0, 0],
83 size=Input("size", "TENSOR_INT32", "{4}"),
84 size_data=[2, 1, 3, 1],
85 output=Output("output", "TENSOR_QUANT8_ASYMM", "{2, 1, 3, 1}, 2.0, 128"),
86 output_data=[3, 3, 3, 5, 5, 5]),
88 inp=Input("input", "TENSOR_INT32", "{3, 2, 3, 1}"),
89 inp_data=[1, 1, 1, 2, 2, 2, 3, 3, 3, 4, 4, 4, 5, 5, 5, 6, 6, 6],
90 begin=Input("begin", "TENSOR_INT32", "{4}"),
91 begin_data=[1, 0, 0, 0],
92 size=Input("size", "TENSOR_INT32", "{4}"),
93 size_data=[2, 1, -1, 1],
94 output=Output("output", "TENSOR_INT32", "{2, 1, 3, 1}"),
95 output_data=[3, 3, 3, 5, 5, 5]),
98 for test_case in test_cases:
99 model = Model().Operation("SLICE", test_case.inp, test_case.begin,
100 test_case.size).To(test_case.output)
102 test_case.inp: test_case.inp_data,
103 test_case.begin: test_case.begin_data,
104 test_case.size: test_case.size_data,
105 test_case.output: test_case.output_data,
107 model=model).AddVariations("relaxed", "float16")
112 # Use BOX_WITH_NMS_LIMIT op to generate a zero-sized internal tensor for box cooridnates.
113 p1 = Parameter("scores", "TENSOR_FLOAT32", "{1, 2}", [0.90, 0.10]) # scores
114 p2 = Parameter("roi", "TENSOR_FLOAT32", "{1, 8}", [1, 1, 10, 10, 0, 0, 10, 10]) # roi
115 o1 = Output("scoresOut", "TENSOR_FLOAT32", "{0}") # scores out
116 o2 = Output("classesOut", "TENSOR_INT32", "{0}") # classes out
117 tmp1 = Internal("roiOut", "TENSOR_FLOAT32", "{0, 4}") # roi out
118 tmp2 = Internal("batchSplitOut", "TENSOR_INT32", "{0}") # batch split out
119 model = Model("zero_sized").Operation("BOX_WITH_NMS_LIMIT", p1, p2, [0], 0.3, -1, 0, 0.4, 1.0, 0.3).To(o1, tmp1, o2, tmp2)
121 # Use ROI_ALIGN op to convert into zero-sized feature map.
122 layout = BoolScalar("layout", False) # NHWC
123 i1 = Input("in", "TENSOR_FLOAT32", "{1, 1, 1, 1}")
124 zero_sized = Internal("featureMap", "TENSOR_FLOAT32", "{0, 2, 2, 1}")
125 model = model.Operation("ROI_ALIGN", i1, tmp1, tmp2, 2, 2, 2.0, 2.0, 4, 4, layout).To(zero_sized)
127 # SLICE op with numBatches = 0.
128 o3 = Output("out", "TENSOR_FLOAT32", "{0, 1, 1, 1}") # out
129 model = model.Operation("SLICE", zero_sized, [0, 1, 1, 0], [-1, 1, -1, 1]).To(o3)
131 quant8 = DataTypeConverter().Identify({
132 p1: ("TENSOR_QUANT8_ASYMM", 0.1, 128),
133 p2: ("TENSOR_QUANT16_ASYMM", 0.125, 0),
134 o1: ("TENSOR_QUANT8_ASYMM", 0.1, 128),
135 tmp1: ("TENSOR_QUANT16_ASYMM", 0.125, 0),
136 i1: ("TENSOR_QUANT8_ASYMM", 0.1, 128),
137 zero_sized: ("TENSOR_QUANT8_ASYMM", 0.1, 128),
138 o3: ("TENSOR_QUANT8_ASYMM", 0.1, 128)
141 # Create test case with dummy values.
147 }).AddVariations("relaxed", quant8, "float16")