aboutsummaryrefslogtreecommitdiff
path: root/tests/validation/CL/DepthwiseConvolutionLayerNative.cpp
diff options
context:
space:
mode:
Diffstat (limited to 'tests/validation/CL/DepthwiseConvolutionLayerNative.cpp')
-rw-r--r--tests/validation/CL/DepthwiseConvolutionLayerNative.cpp11
1 files changed, 6 insertions, 5 deletions
diff --git a/tests/validation/CL/DepthwiseConvolutionLayerNative.cpp b/tests/validation/CL/DepthwiseConvolutionLayerNative.cpp
index bbcded9267..6c9b8e355f 100644
--- a/tests/validation/CL/DepthwiseConvolutionLayerNative.cpp
+++ b/tests/validation/CL/DepthwiseConvolutionLayerNative.cpp
@@ -59,7 +59,8 @@ namespace
RelativeTolerance<float> rel_tolerance_f32(0.001f);
constexpr float abs_tolerance_f32(0.0001f);
-RelativeTolerance<half_float::half> tolerance_f16(half_float::half(0.01));
+RelativeTolerance<half_float::half> rel_tolerance_f16(half_float::half(0.01f));
+const float abs_tolerance_f16(half_float::half(0.02f));
/** Width values to test - Precommit */
const auto width_values_precommit = framework::dataset::make("width", { 37U } );
@@ -187,7 +188,7 @@ FIXTURE_DATA_TEST_CASE(RunSmall, CLDepthwiseConvolutionLayerNativeFixture<half>,
n0_values_precommit))
{
// Validate output
- validate(CLAccessor(_target), _reference, tolerance_f16);
+ validate(CLAccessor(_target), _reference, rel_tolerance_f16);
}
FIXTURE_DATA_TEST_CASE(RunLarge, CLDepthwiseConvolutionLayerNativeFixture<half>, framework::DatasetMode::NIGHTLY,
@@ -207,7 +208,7 @@ FIXTURE_DATA_TEST_CASE(RunLarge, CLDepthwiseConvolutionLayerNativeFixture<half>,
n0_values_nightly))
{
// Validate output
- validate(CLAccessor(_target), _reference, tolerance_f16);
+ validate(CLAccessor(_target), _reference, rel_tolerance_f16, 0.f, abs_tolerance_f16);
}
TEST_SUITE_END() // FP16
TEST_SUITE_END() // Float
@@ -273,7 +274,7 @@ FIXTURE_DATA_TEST_CASE(RunSmall, CLDepthwiseConvolutionLayerNativeFixture<half>,
framework::dataset::make("N0", 1)))
{
// Validate output
- validate(CLAccessor(_target), _reference, tolerance_f16);
+ validate(CLAccessor(_target), _reference, rel_tolerance_f16);
}
FIXTURE_DATA_TEST_CASE(RunLarge, CLDepthwiseConvolutionLayerNativeFixture<half>, framework::DatasetMode::NIGHTLY,
@@ -293,7 +294,7 @@ FIXTURE_DATA_TEST_CASE(RunLarge, CLDepthwiseConvolutionLayerNativeFixture<half>,
framework::dataset::make("N0", 1)))
{
// Validate output
- validate(CLAccessor(_target), _reference, tolerance_f16);
+ validate(CLAccessor(_target), _reference, rel_tolerance_f16, 0.f, abs_tolerance_f16);
}
TEST_SUITE_END() // FP16
TEST_SUITE_END() // Float