Add generatedtests for TransposeConv (#3710)

This patch adds generatedtests for TransposeConv.

Signed-off-by: b.devansh <b.devansh@samsung.com>

diff --git a/runtimes/tests/neural_networks_test/generated/all_generated_tests.cpp b/runtimes/tests/neural_networks_test/generated/all_generated_tests.cpp
index 8a94e80..440610d 100644 (file)
--- a/runtimes/tests/neural_networks_test/generated/all_generated_tests.cpp
+++ b/runtimes/tests/neural_networks_test/generated/all_generated_tests.cpp
@@ -4005,6 +4005,48 @@ TEST_F(GeneratedTests, topk_v2_ex_2D_quant8) {
             topk_v2_ex_2D_quant8::examples);
 }
 
+namespace transpose_conv_ex_float_1 {
+std::vector<MixedTypedExample> examples = {
+// Generated transpose_conv_ex_float_1 test
+#include "generated/examples/transpose_conv_ex_float_1.example.cpp"
+};
+// Generated model constructor
+#include "generated/models/transpose_conv_ex_float_1.model.cpp"
+} // namespace transpose_conv_ex_float_1
+TEST_F(GeneratedTests, transpose_conv_ex_float_1) {
+    execute(transpose_conv_ex_float_1::CreateModel,
+            transpose_conv_ex_float_1::is_ignored,
+            transpose_conv_ex_float_1::examples);
+}
+
+namespace transpose_conv_ex_float_2 {
+std::vector<MixedTypedExample> examples = {
+// Generated transpose_conv_ex_float_2 test
+#include "generated/examples/transpose_conv_ex_float_2.example.cpp"
+};
+// Generated model constructor
+#include "generated/models/transpose_conv_ex_float_2.model.cpp"
+} // namespace transpose_conv_ex_float_2
+TEST_F(GeneratedTests, transpose_conv_ex_float_2) {
+    execute(transpose_conv_ex_float_2::CreateModel,
+            transpose_conv_ex_float_2::is_ignored,
+            transpose_conv_ex_float_2::examples);
+}
+
+namespace transpose_conv_ex_float_3 {
+std::vector<MixedTypedExample> examples = {
+// Generated transpose_conv_ex_float_3 test
+#include "generated/examples/transpose_conv_ex_float_3.example.cpp"
+};
+// Generated model constructor
+#include "generated/models/transpose_conv_ex_float_3.model.cpp"
+} // namespace transpose_conv_ex_float_3
+TEST_F(GeneratedTests, transpose_conv_ex_float_3) {
+    execute(transpose_conv_ex_float_3::CreateModel,
+            transpose_conv_ex_float_3::is_ignored,
+            transpose_conv_ex_float_3::examples);
+}
+
 namespace unpack_ex_3D_float2 {
 std::vector<MixedTypedExample> examples = {
 // Generated unpack_ex_3D_float2 test

diff --git a/runtimes/tests/neural_networks_test/generated/examples/transpose_conv_ex_float_1.example.cpp b/runtimes/tests/neural_networks_test/generated/examples/transpose_conv_ex_float_1.example.cpp
new file mode 100644 (file)
index 0000000..0661187
--- /dev/null
+++ b/runtimes/tests/neural_networks_test/generated/examples/transpose_conv_ex_float_1.example.cpp
@@ -0,0 +1,22 @@
+// Generated file (from: transpose_conv_ex_float_1.mod.py). Do not edit
+// Begin of an example
+{
+//Input(s)
+{ // See tools/test_generator/include/TestHarness.h:MixedTyped
+  // int -> FLOAT32 map
+  {{0, {1, 4, 4, 1}}, {1, {1, 2, 3, 4, 5, 6, 7, 8, 9}}, {2, {1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16}}},
+  // int -> INT32 map
+  {},
+  // int -> QUANT8_ASYMM map
+  {}
+},
+//Output(s)
+{ // See tools/test_generator/include/TestHarness.h:MixedTyped
+  // int -> FLOAT32 map
+  {{0, {29, 62, 83, 75, 99, 192, 237, 198, 207, 372, 417, 330, 263, 446, 485, 365}}},
+  // int -> INT32 map
+  {},
+  // int -> QUANT8_ASYMM map
+  {}
+}
+}, // End of an example

diff --git a/runtimes/tests/neural_networks_test/generated/examples/transpose_conv_ex_float_2.example.cpp b/runtimes/tests/neural_networks_test/generated/examples/transpose_conv_ex_float_2.example.cpp
new file mode 100644 (file)
index 0000000..d0103ae
--- /dev/null
+++ b/runtimes/tests/neural_networks_test/generated/examples/transpose_conv_ex_float_2.example.cpp
@@ -0,0 +1,22 @@
+// Generated file (from: transpose_conv_ex_float_2.mod.py). Do not edit
+// Begin of an example
+{
+//Input(s)
+{ // See tools/test_generator/include/TestHarness.h:MixedTyped
+  // int -> FLOAT32 map
+  {{0, {1, 5, 5, 1}}, {1, {1, 2, 3, 4, 5, 6, 7, 8, 9}}, {2, {1, 2, 3, 4}}},
+  // int -> INT32 map
+  {},
+  // int -> QUANT8_ASYMM map
+  {}
+},
+//Output(s)
+{ // See tools/test_generator/include/TestHarness.h:MixedTyped
+  // int -> FLOAT32 map
+  {{0, {1, 2, 5, 4, 6, 4, 5, 14, 10, 12, 10, 14, 36, 24, 30, 12, 15, 34, 20, 24, 21, 24, 55, 32, 36}}},
+  // int -> INT32 map
+  {},
+  // int -> QUANT8_ASYMM map
+  {}
+}
+}, // End of an example

diff --git a/runtimes/tests/neural_networks_test/generated/examples/transpose_conv_ex_float_3.example.cpp b/runtimes/tests/neural_networks_test/generated/examples/transpose_conv_ex_float_3.example.cpp
new file mode 100644 (file)
index 0000000..a291edb
--- /dev/null
+++ b/runtimes/tests/neural_networks_test/generated/examples/transpose_conv_ex_float_3.example.cpp
@@ -0,0 +1,22 @@
+// Generated file (from: transpose_conv_ex_float_3.mod.py). Do not edit
+// Begin of an example
+{
+//Input(s)
+{ // See tools/test_generator/include/TestHarness.h:MixedTyped
+  // int -> FLOAT32 map
+  {{0, {1, 5, 5, 2}}, {1, {1, 3, 5, 7, 9, 11, 13, 15, 17, 2, 4, 6, 8, 10, 12, 14, 16, 18}}, {2, {1, 2, 3, 4}}},
+  // int -> INT32 map
+  {},
+  // int -> QUANT8_ASYMM map
+  {}
+},
+//Output(s)
+{ // See tools/test_generator/include/TestHarness.h:MixedTyped
+  // int -> FLOAT32 map
+  {{0, {1, 2, 3, 4, 7, 10, 6, 8, 10, 12, 7, 8, 9, 10, 25, 28, 18, 20, 22, 24, 16, 20, 24, 28, 62, 72, 42, 48, 54, 60, 21, 24, 27, 30, 61, 68, 36, 40, 44, 48, 39, 42, 45, 48, 103, 110, 60, 64, 68, 72}}},
+  // int -> INT32 map
+  {},
+  // int -> QUANT8_ASYMM map
+  {}
+}
+}, // End of an example

diff --git a/runtimes/tests/neural_networks_test/generated/models/transpose_conv_ex_float_1.model.cpp b/runtimes/tests/neural_networks_test/generated/models/transpose_conv_ex_float_1.model.cpp
new file mode 100644 (file)
index 0000000..f31d02c
--- /dev/null
+++ b/runtimes/tests/neural_networks_test/generated/models/transpose_conv_ex_float_1.model.cpp
@@ -0,0 +1,33 @@
+// Generated file (from: transpose_conv_ex_float_1.mod.py). Do not edit
+void CreateModel(Model *model) {
+  OperandType type3(Type::INT32, {});
+  OperandType type1(Type::TENSOR_FLOAT32, {1, 3, 3, 1});
+  OperandType type2(Type::TENSOR_FLOAT32, {1, 4, 4, 1});
+  OperandType type0(Type::TENSOR_FLOAT32, {4});
+  // Phase 1, operands
+  auto op_shape = model->addOperand(&type0);
+  auto ker = model->addOperand(&type1);
+  auto in = model->addOperand(&type2);
+  auto pad_same = model->addOperand(&type3);
+  auto stride_x = model->addOperand(&type3);
+  auto stride_y = model->addOperand(&type3);
+  auto op = model->addOperand(&type2);
+  // Phase 2, operations
+  static int32_t pad_same_init[] = {1};
+  model->setOperandValue(pad_same, pad_same_init, sizeof(int32_t) * 1);
+  static int32_t stride_x_init[] = {1};
+  model->setOperandValue(stride_x, stride_x_init, sizeof(int32_t) * 1);
+  static int32_t stride_y_init[] = {1};
+  model->setOperandValue(stride_y, stride_y_init, sizeof(int32_t) * 1);
+  model->addOperationEx(ANEURALNETWORKS_TRANSPOSE_CONV_EX, {op_shape, ker, in, pad_same, stride_x, stride_y}, {op});
+  // Phase 3, inputs and outputs
+  model->identifyInputsAndOutputs(
+    {op_shape, ker, in},
+    {op});
+  assert(model->isValid());
+}
+
+bool is_ignored(int i) {
+  static std::set<int> ignore = {};
+  return ignore.find(i) != ignore.end();
+}

diff --git a/runtimes/tests/neural_networks_test/generated/models/transpose_conv_ex_float_2.model.cpp b/runtimes/tests/neural_networks_test/generated/models/transpose_conv_ex_float_2.model.cpp
new file mode 100644 (file)
index 0000000..18b9443
--- /dev/null
+++ b/runtimes/tests/neural_networks_test/generated/models/transpose_conv_ex_float_2.model.cpp
@@ -0,0 +1,34 @@
+// Generated file (from: transpose_conv_ex_float_2.mod.py). Do not edit
+void CreateModel(Model *model) {
+  OperandType type3(Type::INT32, {});
+  OperandType type2(Type::TENSOR_FLOAT32, {1, 2, 2, 1});
+  OperandType type1(Type::TENSOR_FLOAT32, {1, 3, 3, 1});
+  OperandType type4(Type::TENSOR_FLOAT32, {1, 5, 5, 1});
+  OperandType type0(Type::TENSOR_FLOAT32, {4});
+  // Phase 1, operands
+  auto op_shape = model->addOperand(&type0);
+  auto ker = model->addOperand(&type1);
+  auto in = model->addOperand(&type2);
+  auto pad_valid = model->addOperand(&type3);
+  auto stride_x = model->addOperand(&type3);
+  auto stride_y = model->addOperand(&type3);
+  auto op = model->addOperand(&type4);
+  // Phase 2, operations
+  static int32_t pad_valid_init[] = {2};
+  model->setOperandValue(pad_valid, pad_valid_init, sizeof(int32_t) * 1);
+  static int32_t stride_x_init[] = {2};
+  model->setOperandValue(stride_x, stride_x_init, sizeof(int32_t) * 1);
+  static int32_t stride_y_init[] = {2};
+  model->setOperandValue(stride_y, stride_y_init, sizeof(int32_t) * 1);
+  model->addOperationEx(ANEURALNETWORKS_TRANSPOSE_CONV_EX, {op_shape, ker, in, pad_valid, stride_x, stride_y}, {op});
+  // Phase 3, inputs and outputs
+  model->identifyInputsAndOutputs(
+    {op_shape, ker, in},
+    {op});
+  assert(model->isValid());
+}
+
+bool is_ignored(int i) {
+  static std::set<int> ignore = {};
+  return ignore.find(i) != ignore.end();
+}

diff --git a/runtimes/tests/neural_networks_test/generated/models/transpose_conv_ex_float_3.model.cpp b/runtimes/tests/neural_networks_test/generated/models/transpose_conv_ex_float_3.model.cpp
new file mode 100644 (file)
index 0000000..138a594
--- /dev/null
+++ b/runtimes/tests/neural_networks_test/generated/models/transpose_conv_ex_float_3.model.cpp
@@ -0,0 +1,34 @@
+// Generated file (from: transpose_conv_ex_float_3.mod.py). Do not edit
+void CreateModel(Model *model) {
+  OperandType type3(Type::INT32, {});
+  OperandType type2(Type::TENSOR_FLOAT32, {1, 2, 2, 1});
+  OperandType type4(Type::TENSOR_FLOAT32, {1, 5, 5, 2});
+  OperandType type1(Type::TENSOR_FLOAT32, {2, 3, 3, 1});
+  OperandType type0(Type::TENSOR_FLOAT32, {4});
+  // Phase 1, operands
+  auto op_shape = model->addOperand(&type0);
+  auto ker = model->addOperand(&type1);
+  auto in = model->addOperand(&type2);
+  auto pad_valid = model->addOperand(&type3);
+  auto stride_x = model->addOperand(&type3);
+  auto stride_y = model->addOperand(&type3);
+  auto op = model->addOperand(&type4);
+  // Phase 2, operations
+  static int32_t pad_valid_init[] = {2};
+  model->setOperandValue(pad_valid, pad_valid_init, sizeof(int32_t) * 1);
+  static int32_t stride_x_init[] = {2};
+  model->setOperandValue(stride_x, stride_x_init, sizeof(int32_t) * 1);
+  static int32_t stride_y_init[] = {2};
+  model->setOperandValue(stride_y, stride_y_init, sizeof(int32_t) * 1);
+  model->addOperationEx(ANEURALNETWORKS_TRANSPOSE_CONV_EX, {op_shape, ker, in, pad_valid, stride_x, stride_y}, {op});
+  // Phase 3, inputs and outputs
+  model->identifyInputsAndOutputs(
+    {op_shape, ker, in},
+    {op});
+  assert(model->isValid());
+}
+
+bool is_ignored(int i) {
+  static std::set<int> ignore = {};
+  return ignore.find(i) != ignore.end();
+}

diff --git a/runtimes/tests/neural_networks_test/runtime_run_android_nn_test.skip.armv7l-linux.neurun b/runtimes/tests/neural_networks_test/runtime_run_android_nn_test.skip.armv7l-linux.neurun
index 534de1f..9b5c359 100644 (file)
--- a/runtimes/tests/neural_networks_test/runtime_run_android_nn_test.skip.armv7l-linux.neurun
+++ b/runtimes/tests/neural_networks_test/runtime_run_android_nn_test.skip.armv7l-linux.neurun
@@ -116,5 +116,6 @@ GeneratedTests.conv_quant8_2
 GeneratedTests.max_pool_quant8_4
 GeneratedTests.softmax*
 GeneratedTests.split*
+GeneratedTests.transpose_conv*
 GeneratedTests.pack*
 GeneratedTests.unpack*

diff --git a/runtimes/tests/neural_networks_test/runtime_run_android_nn_test.skip.armv7l-tizen b/runtimes/tests/neural_networks_test/runtime_run_android_nn_test.skip.armv7l-tizen
index 4596409..e068f44 100644 (file)
--- a/runtimes/tests/neural_networks_test/runtime_run_android_nn_test.skip.armv7l-tizen
+++ b/runtimes/tests/neural_networks_test/runtime_run_android_nn_test.skip.armv7l-tizen
@@ -111,5 +111,6 @@ GeneratedTests.logistic_quant8_2
 GeneratedTests.mul_broadcast_quant8
 GeneratedTests.mul_quant8
 GeneratedTests.split*
+GeneratedTests.transpose_conv*
 GeneratedTests.sub_broadcast_float
 GeneratedTests.pack_ex_3D_float2

diff --git a/runtimes/tests/neural_networks_test/runtime_run_android_nn_test.skip.x86_64-linux b/runtimes/tests/neural_networks_test/runtime_run_android_nn_test.skip.x86_64-linux
index dd18702..ab29975 100644 (file)
--- a/runtimes/tests/neural_networks_test/runtime_run_android_nn_test.skip.x86_64-linux
+++ b/runtimes/tests/neural_networks_test/runtime_run_android_nn_test.skip.x86_64-linux
@@ -81,5 +81,6 @@ ValidationTestExecution.*
 GeneratedTests.gather_ex_1D_quant8
 GeneratedTests.gather_ex_2D_quant8
 GeneratedTests.split*
+GeneratedTests.transpose_conv*
 GeneratedTests.pack*
 GeneratedTests.unpack*

diff --git a/runtimes/tests/neural_networks_test/specs/Ex/transpose_conv_ex_float_1.mod.py b/runtimes/tests/neural_networks_test/specs/Ex/transpose_conv_ex_float_1.mod.py
new file mode 100644 (file)
index 0000000..40be027
--- /dev/null
+++ b/runtimes/tests/neural_networks_test/specs/Ex/transpose_conv_ex_float_1.mod.py
@@ -0,0 +1,25 @@
+# model
+model = Model()
+i0 = Input("op_shape", "TENSOR_FLOAT32", "{4}")
+weights = Input("ker", "TENSOR_FLOAT32", "{1, 3, 3, 1}")
+i1 = Input("in", "TENSOR_FLOAT32", "{1, 4, 4, 1}" )
+pad = Int32Scalar("pad_same", 1)
+s_x = Int32Scalar("stride_x", 1)
+s_y = Int32Scalar("stride_y", 1)
+i2 = Output("op", "TENSOR_FLOAT32", "{1, 4, 4, 1}")
+model = model.Operation("TRANSPOSE_CONV_EX", i0, weights, i1, pad, s_x, s_y).To(i2)
+
+# Example 1. Input in operand 0,
+input0 = {i0:  # output shape
+          [1, 4, 4, 1],
+          i1:  # input 0
+          [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16],
+          weights:  # input 1
+          [1, 2, 3, 4, 5, 6, 7, 8, 9]}
+
+output0 = {i2:  # output 0
+           [29, 62, 83, 75, 99, 192, 237, 198, 207, 372,
+                                417, 330, 263, 446, 485, 365]}
+
+# Instantiate an example
+Example((input0, output0))

diff --git a/runtimes/tests/neural_networks_test/specs/Ex/transpose_conv_ex_float_2.mod.py b/runtimes/tests/neural_networks_test/specs/Ex/transpose_conv_ex_float_2.mod.py
new file mode 100644 (file)
index 0000000..d52d82c
--- /dev/null
+++ b/runtimes/tests/neural_networks_test/specs/Ex/transpose_conv_ex_float_2.mod.py
@@ -0,0 +1,25 @@
+# model
+model = Model()
+i0 = Input("op_shape", "TENSOR_FLOAT32", "{4}")
+weights = Input("ker", "TENSOR_FLOAT32", "{1, 3, 3, 1}")
+i1 = Input("in", "TENSOR_FLOAT32", "{1, 2, 2, 1}" )
+pad = Int32Scalar("pad_valid", 2)
+s_x = Int32Scalar("stride_x", 2)
+s_y = Int32Scalar("stride_y", 2)
+i2 = Output("op", "TENSOR_FLOAT32", "{1, 5, 5, 1}")
+model = model.Operation("TRANSPOSE_CONV_EX", i0, weights, i1, pad, s_x, s_y).To(i2)
+
+# Example 1. Input in operand 0,
+input0 = {i0:  # output shape
+          [1, 5, 5, 1],
+          i1:  # input 0
+          [1, 2, 3, 4],
+          weights:  # input 1
+          [1, 2, 3, 4, 5, 6, 7, 8, 9]}
+
+output0 = {i2:  # output 0
+           [1,  2,  5,  4,  6,  4,  5,  14, 10, 12, 10, 14, 36,
+                        24, 30, 12, 15, 34, 20, 24, 21, 24, 55, 32, 36]}
+
+# Instantiate an example
+Example((input0, output0))

diff --git a/runtimes/tests/neural_networks_test/specs/Ex/transpose_conv_ex_float_3.mod.py b/runtimes/tests/neural_networks_test/specs/Ex/transpose_conv_ex_float_3.mod.py
new file mode 100644 (file)
index 0000000..03cfaed
--- /dev/null
+++ b/runtimes/tests/neural_networks_test/specs/Ex/transpose_conv_ex_float_3.mod.py
@@ -0,0 +1,27 @@
+# model
+model = Model()
+i0 = Input("op_shape", "TENSOR_FLOAT32", "{4}")
+weights = Input("ker", "TENSOR_FLOAT32", "{2, 3, 3, 1}")
+i1 = Input("in", "TENSOR_FLOAT32", "{1, 2, 2, 1}" )
+pad = Int32Scalar("pad_valid", 2)
+s_x = Int32Scalar("stride_x", 2)
+s_y = Int32Scalar("stride_y", 2)
+i2 = Output("op", "TENSOR_FLOAT32", "{1, 5, 5, 2}")
+model = model.Operation("TRANSPOSE_CONV_EX", i0, weights, i1, pad, s_x, s_y).To(i2)
+
+# Example 1. Input in operand 0,
+input0 = {i0:  # output shape
+          [1, 5, 5, 2],
+          i1:  # input 0
+          [1, 2, 3, 4],
+          weights:  # input 1
+          [1, 3, 5, 7, 9, 11, 13, 15, 17, 2, 4, 6, 8, 10, 12, 14, 16, 18]}
+
+output0 = {i2:  # output 0
+           [1,  2,  3,  4,  7,  10,  6,   8,  10, 12, 7,  8,  9,
+                        10, 25, 28, 18, 20, 22,  24,  16, 20, 24, 28, 62, 72,
+                        42, 48, 54, 60, 21, 24,  27,  30, 61, 68, 36, 40, 44,
+                        48, 39, 42, 45, 48, 103, 110, 60, 64, 68, 72]}
+
+# Instantiate an example
+Example((input0, output0))