diff options
| author | Sangwon Ha <sangwon.ha@arm.com> | 2023-10-31 12:12:41 +0000 |
|---|---|---|
| committer | Adnan AlSinan <adnan.alsinan@arm.com> | 2023-11-01 16:44:52 +0000 |
| commit | c259aa5e04714bb3a8d23a6903161c240c279743 (patch) | |
| tree | 84e6c3aa88bf3f3e860d82411071c919ecd54bec /tests | |
| parent | a7ddd60527e6027e6d619821b6cf91721d94f101 (diff) | |
| download | ComputeLibrary-c259aa5e04714bb3a8d23a6903161c240c279743.tar.gz | |
Increase tolerance for MatMul in FP16
Resolves: COMPMID-6585
Change-Id: Ibfc5412ab23f10026c872eab99a056eb682d77a1
Signed-off-by: Sangwon Ha <sangwon.ha@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/10577
Benchmark: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: SiCong Li <sicong.li@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Diffstat (limited to 'tests')
| -rw-r--r-- | tests/validation/CL/MatMulNativeMMULKernel.cpp | 2 |
1 files changed, 1 insertions, 1 deletions
diff --git a/tests/validation/CL/MatMulNativeMMULKernel.cpp b/tests/validation/CL/MatMulNativeMMULKernel.cpp index bfc36c07ab..655dd354dc 100644 --- a/tests/validation/CL/MatMulNativeMMULKernel.cpp +++ b/tests/validation/CL/MatMulNativeMMULKernel.cpp @@ -46,7 +46,7 @@ RelativeTolerance<float> tolerance_f32(0.001f); /**< Tolerance value for compari constexpr float abs_tolerance_f32( 0.0001f); /**< Absolute tolerance value for comparing reference's output against implementation's output for floating point data types in case using relative tolerance fails because of small values */ constexpr float abs_tolerance_f16( - 0.001f); /**< Absolute tolerance value for comparing reference's output against implementation's output for fp16 data types in case using relative tolerance fails because of small values */ + 0.02f); /**< Absolute tolerance value for comparing reference's output against implementation's output for fp16 data types in case using relative tolerance fails because of small values */ RelativeTolerance<half_float::half> tolerance_f16(half(0.02)); /**< Tolerance value for comparing reference's output against implementation's output for floating point data types */ } // namespace |