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
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
|
Description: ARM32 has a limited support (see docs/Roadmap.md)
Several tests fail on this architecture, so skipping them for now.
Author: Dylan Aïssi <daissi@debian.org>
Forwarded: not-needed
--- a/onnxruntime/test/optimizer/qdq_transformer_test.cc
+++ b/onnxruntime/test/optimizer/qdq_transformer_test.cc
@@ -138,11 +138,13 @@
QDQTransformerConvTests<int8_t, int8_t, int32_t, uint8_t>();
}
+#if !defined(__arm__)
TEST(QDQTransformerTests, Conv_S8X8S8) {
// input not uint8_t and output not uint8_t
QDQTransformerConvTests<int8_t, uint8_t, int32_t, int8_t>();
QDQTransformerConvTests<int8_t, int8_t, int32_t, int8_t>();
}
+#endif // #if !defined(__arm__)
TEST(QDQTransformerTests, ConvMaxPoolReshape_UInt8) {
auto test_case = [&](const std::vector<int64_t>& input_shape, const std::vector<int64_t>& weights_shape,
@@ -215,6 +217,7 @@
test_case({1, 22, 11, 13, 15}, {30, 22, 5, 3, 3}, 19, true); // Use com.microsoft QDQ ops
}
+#if !defined(__arm__)
TEST(QDQTransformerTests, ConvMaxPoolReshape_Int8) {
auto test_case = [&](const std::vector<int64_t>& input_shape, const std::vector<int64_t>& weights_shape,
bool use_contrib_qdq = false) {
@@ -689,10 +692,12 @@
QDQTransformerMatMulTests<uint8_t, int8_t, uint8_t>(true);
}
+#if !defined(__arm__)
TEST(QDQTransformerTests, MatMul_S8S8S8) {
QDQTransformerMatMulTests<int8_t, int8_t, int8_t>(false);
QDQTransformerMatMulTests<int8_t, int8_t, int8_t>(true);
}
+#endif // #if !defined(__arm__)
TEST(QDQTransformerTests, MatMul_S8U8U8) {
QDQTransformerMatMulTests<int8_t, uint8_t, uint8_t>(false);
@@ -704,10 +709,12 @@
QDQTransformerMatMulTests<int8_t, uint8_t, int8_t>(true);
}
+#if !defined(__arm__)
TEST(QDQTransformerTests, MatMul_S8S8U8) {
QDQTransformerMatMulTests<int8_t, int8_t, uint8_t>(false);
QDQTransformerMatMulTests<int8_t, int8_t, uint8_t>(true);
}
+#endif // #if !defined(__arm__)
template <typename Input1Type, typename Input2Type, typename OutputType, typename BiasType = int32_t>
void QDQTransformerGemmTests(bool has_output_q, bool has_bias, bool beta_not_one = false) {
@@ -874,10 +881,12 @@
QDQTransformerGemmTests<uint8_t, int8_t, uint8_t, uint8_t>();
}
+#if !defined(__arm__)
TEST(QDQTransformerTests, Gemm_S8S8S8) {
QDQTransformerGemmTests<int8_t, int8_t, int8_t>();
QDQTransformerGemmTests<int8_t, int8_t, int8_t, uint8_t>();
}
+#endif // #if !defined(__arm__)
TEST(QDQTransformerTests, Gemm_S8U8U8) {
QDQTransformerGemmTests<int8_t, uint8_t, uint8_t>();
@@ -889,10 +898,12 @@
QDQTransformerGemmTests<int8_t, uint8_t, int8_t, uint8_t>();
}
+#if !defined(__arm__)
TEST(QDQTransformerTests, Gemm_S8S8U8) {
QDQTransformerGemmTests<int8_t, int8_t, uint8_t>();
QDQTransformerGemmTests<int8_t, int8_t, uint8_t, uint8_t>();
}
+#endif // #if !defined(__arm__)
// Runs a test case that checks if Q/DQ nodes are dropped from DQ -> Gather -> Q.
template <typename QuantType>
@@ -2435,6 +2446,7 @@
test_case({1, 12, 37}, {32, 12, 5}, true /*use_contrib_qdq*/);
}
+#if !defined(__arm__)
TEST(QDQTransformerTests, ConvAveragePoolReshape_Int8) {
auto test_case = [&](const std::vector<int64_t>& input_shape, const std::vector<int64_t>& weights_shape,
bool use_contrib_qdq) {
@@ -2577,6 +2589,7 @@
test_case({1, 22, 11, 13, 15}, {30, 22, 5, 3, 3}, false /*use_contrib_qdq*/);
test_case({1, 12, 37}, {32, 12, 5}, true /*use_contrib_qdq*/);
}
+#endif // #if !defined(__arm__)
template <typename InputType, typename OutputType>
void QDQTransformerLeakyReluTests() {
@@ -2753,6 +2766,7 @@
QDQTransformerSigmoidTests<uint8_t, int8_t>();
}
+#if !defined(__arm__)
TEST(QDQTransformerTests, ConvTranspose_QBackward) {
auto test_case = [&](const std::vector<int64_t>& input_shape, const std::vector<int64_t>& weights_shape,
const std::vector<int64_t>& perms, bool use_contrib_qdq) {
@@ -3611,6 +3625,7 @@
test_case({1, 13, 13, 23}, 4, {0, 3, 1, 2}, true, true, true /*use_contrib_qdq*/);
#endif
}
+#endif // #if !defined(__arm__)
TEST(QDQTransformerTests, QDQPropagation_StopAtOtherQDQ) {
auto test_case = [&](const std::vector<int64_t>& input_shape, bool same_scale, bool same_zp,
@@ -3971,6 +3986,7 @@
run_test_case(/*slice_has_graph_output*/ false);
run_test_case(/*slice_has_graph_output*/ true);
}
+#endif // #if !defined(__arm__)
#if defined(USE_NNAPI) || defined(USE_QNN) || defined(USE_XNNPACK)
static void VerifyIODef(const NodeUnitIODef& io_def, const Node& node) {
--- a/onnxruntime/test/contrib_ops/math/matmul_sparse_test.cc
+++ b/onnxruntime/test/contrib_ops/math/matmul_sparse_test.cc
@@ -140,6 +140,7 @@
}
*/
#if !defined(DISABLE_SPARSE_TENSORS)
+#if !defined(__arm__)
TEST(SparseToDenseMatMul, TestCsr) {
constexpr int64_t rows = 9;
constexpr int64_t cols = 9;
@@ -260,6 +261,7 @@
tester.Run(OpTester::ExpectResult::kExpectSuccess);
}
}
+#endif // #if !defined(__arm__)
TEST(SparseToDenseMatMul, TestCoo) {
constexpr int64_t rows = 9;
@@ -381,4 +383,4 @@
#endif // !defined(DISABLE_SPARSE_TENSORS)
} // namespace test
-} // namespace onnxruntime
\ No newline at end of file
+} // namespace onnxruntime
--- a/onnxruntime/test/contrib_ops/quant_gemm_test.cc
+++ b/onnxruntime/test/contrib_ops/quant_gemm_test.cc
@@ -180,6 +180,7 @@
false /*per_column*/);
}
+#if !defined(__arm__)
TEST(QuantGemmTest, Scalar) {
RunQuantGemmTestBatch(1, 1, 32);
RunQuantGemmTestBatch(1, 1, 260);
@@ -194,6 +195,7 @@
RunQuantGemmTestBatch(1, 8, 400);
RunQuantGemmTestBatch(1, 512, 1024);
}
+#endif // #if !defined(__arm__)
TEST(QuantGemmTest, GEMM) {
RunQuantGemmTestBatch(2, 2, 40);
--- a/onnxruntime/test/providers/cpu/nn/qlinearconv_op_test.cc
+++ b/onnxruntime/test/providers/cpu/nn/qlinearconv_op_test.cc
@@ -1080,6 +1080,7 @@
}
}
+#if !defined(__arm__)
TEST(QLinearConvTest, Conv1D_S8S8) {
QLinearConvOpTester<int8_t, int8_t> test;
test.GenerateRandomInput({3, 24, 15}, .05f, 4);
@@ -1089,6 +1090,7 @@
test.SetOutputScaleAndZeroPoint(.55f, 54);
test.Run();
}
+#endif // #if !defined(__arm__)
TEST(QLinearConvTest, Conv2D_S8S8_Sym_M64_C64) {
QLinearConvOpTester<int8_t, int8_t> test;
@@ -1100,6 +1102,7 @@
test.Run();
}
+#if !defined(__arm__)
TEST(QLinearConvTest, Conv2D_S8S8_Sym_M16_C4_Bias) {
QLinearConvOpTester<int8_t, int8_t> test;
test.GenerateRandomInput({1, 4, 3, 3}, .05f, 4);
@@ -1119,6 +1122,7 @@
test.SetOutputScaleAndZeroPoint(.55f, 54);
test.Run();
}
+#endif // #if !defined(__arm__)
TEST(QLinearConvTest, Conv2D_S8S8_Sym_M48_C48_Bias_Pads) {
QLinearConvOpTester<int8_t, int8_t> test;
@@ -1140,6 +1144,7 @@
test.Run();
}
+#if !defined(__arm__)
TEST(QLinearConvTest, Conv2D_S8S8) {
QLinearConvOpTester<int8_t, int8_t> test;
test.GenerateRandomInput({3, 24, 15, 11}, .05f, 4);
@@ -1149,6 +1154,7 @@
test.SetOutputScaleAndZeroPoint(.55f, 54);
test.Run();
}
+#endif // #if !defined(__arm__)
TEST(QLinearConvTest, Conv3D_S8S8) {
// TODO: Unskip when fixed #41968513
@@ -1165,6 +1171,7 @@
test.Run();
}
+#if !defined(__arm__)
TEST(QLinearConvTest, Conv1D_S8S8_Pointwise) {
QLinearConvOpTester<int8_t, int8_t> test;
test.GenerateRandomInput({3, 24, 15}, .05f, 4);
@@ -1191,6 +1198,7 @@
test.SetOutputScaleAndZeroPoint(.75f, -14);
test.Run();
}
+#endif // #if !defined(__arm__)
TEST(QLinearConvTest, Conv3D_S8S8_Pointwise) {
// TODO: Unskip when fixed #41968513
@@ -1206,6 +1214,7 @@
test.Run();
}
+#if !defined(__arm__)
TEST(QLinearConvTest, Conv1D_S8S8_Dilations) {
QLinearConvOpTester<int8_t, int8_t> test;
test.GenerateRandomInput({1, 4, 19}, .02f, 20);
@@ -1214,6 +1223,7 @@
test.SetOutputScaleAndZeroPoint(.24f, -15);
test.Run();
}
+#endif // #if !defined(__arm__)
TEST(QLinearConvTest, Conv2D_S8S8_Dilations) {
QLinearConvOpTester<int8_t, int8_t> test;
@@ -1238,6 +1248,7 @@
test.Run();
}
+#if !defined(__arm__)
TEST(QLinearConvTest, Conv1D_S8S8_Strides) {
QLinearConvOpTester<int8_t, int8_t> test;
test.GenerateRandomInput({1, 7, 18}, .04f, 16);
@@ -1349,6 +1360,7 @@
test.SetOutputScaleAndZeroPoint(.26f, 8);
test.Run();
}
+#endif // #if !defined(__arm__)
TEST(QLinearConvTest, Conv1D_S8S8_Depthwise) {
for (int8_t weight_zero_point : std::initializer_list<int8_t>{-2, 0, 2}) {
@@ -1462,6 +1474,7 @@
}
}
+#if !defined(__arm__)
TEST(QLinearConvTest, Conv2D_S8S8_Requantize_NoBias) {
for (int64_t channels = 1; channels <= 32; channels++) {
QLinearConvOpTester<int8_t, int8_t> test;
@@ -1506,6 +1519,7 @@
test.Run();
}
}
+#endif // #if !defined(__arm__)
TEST(QLinearConvTest, Conv2D_S8S8_Depthwise_Kernelsize) {
TestQLinearConv2dDepthwiseKernelsize<int8_t, int8_t>();
|
