diff options
| author | Mohammed Suhail Munshi <MohammedSuhail.Munshi@arm.com> | 2023-10-26 00:14:36 +0100 |
|---|---|---|
| committer | Mohmun02 <MohammedSuhail.Munshi@arm.com> | 2023-11-03 14:01:37 +0000 |
| commit | 02c452fe1ec17c3941272a07b5cae1f32d614c56 (patch) | |
| tree | 3fc26399885bfefe6b9b086ea2b7fefc7eff54e2 /tests/validation/Helpers.h | |
| parent | c259aa5e04714bb3a8d23a6903161c240c279743 (diff) | |
| download | ComputeLibrary-02c452fe1ec17c3941272a07b5cae1f32d614c56.tar.gz | |
Add Dynamic Quantization tests to Fully Connected Layer
This patch calculates the output quantization info based on the inputs'
quantization information.
The previous approach was using the same quantization information for
input, weights and output.
This implementation does not cover the cases where we have fused
activation function.
Resolves: [COMPMID-6484]
Signed-off-by: Mohammed Suhail Munshi <MohammedSuhail.Munshi@arm.com>
Change-Id: Ib58143165191e82ae8547e661ac7c8d077bda200
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/10539
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: SiCong Li <sicong.li@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>
Diffstat (limited to 'tests/validation/Helpers.h')
| -rw-r--r-- | tests/validation/Helpers.h | 5 |
1 files changed, 3 insertions, 2 deletions
diff --git a/tests/validation/Helpers.h b/tests/validation/Helpers.h index 5a1e69afbd..647adcdb69 100644 --- a/tests/validation/Helpers.h +++ b/tests/validation/Helpers.h @@ -286,12 +286,13 @@ QuantizationHint suggest_matmul_dst_q_info_and_bias(const QuantizationInfo &lhs_ * @param[in] k number of accumulations taking place in the sum, i.e. c_k = sum_k(a_k * b_k) * @param[in] data_type data type, only QASYMM8, QASYMM8_SIGNED are supported * @param[in] bias_fraction the fraction of bias amplitude compared to integer accummulation. + * @param[in] num_sd (Optional) number of standard deviations we allow from the mean. Default value is 2. * * @return QuantizationHint object containing the suggested output quantization info and min/max bias range */ QuantizationHint suggest_mac_dst_q_info_and_bias(const QuantizationInfo &lhs_q_info, - const QuantizationInfo &rhs_q_info, int32_t k, DataType data_type, - float bias_fraction); + const QuantizationInfo &rhs_q_info, int32_t k, DataType data_type, float bias_fraction, + int num_sd = 2); } // namespace validation } // namespace test } // namespace arm_compute |