diff options
author | Michele Di Giorgio <michele.digiorgio@arm.com> | 2021-03-09 14:09:08 +0000 |
---|---|---|
committer | Michele Di Giorgio <michele.digiorgio@arm.com> | 2021-03-31 17:08:51 +0000 |
commit | 33f41fabd30fb444aaa0cf3e65b61794d498d151 (patch) | |
tree | a381cff3096a3b05198b0cd311fee28e40fd5a4f /tests/validation/NEON/QLSTMLayerNormalization.cpp | |
parent | 5f91b5d7063462854b62d342f9d4e04ae647e9a6 (diff) | |
download | ComputeLibrary-33f41fabd30fb444aaa0cf3e65b61794d498d151.tar.gz |
Fix trademarks throughout the codebase
Resolves: COMPMID-4299
Change-Id: Ie6a52c1371b9a2a7b5bb4f019ecd5e70a2008567
Signed-off-by: Michele Di Giorgio <michele.digiorgio@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5338
Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Diffstat (limited to 'tests/validation/NEON/QLSTMLayerNormalization.cpp')
-rw-r--r-- | tests/validation/NEON/QLSTMLayerNormalization.cpp | 4 |
1 files changed, 2 insertions, 2 deletions
diff --git a/tests/validation/NEON/QLSTMLayerNormalization.cpp b/tests/validation/NEON/QLSTMLayerNormalization.cpp index 617f64ce1d..9738213114 100644 --- a/tests/validation/NEON/QLSTMLayerNormalization.cpp +++ b/tests/validation/NEON/QLSTMLayerNormalization.cpp @@ -167,7 +167,7 @@ TEST_SUITE(Quantized) TEST_SUITE(QSYMM16) /** Tests will be targetting - * - Comparison between Neon kernel and the exact same but scalar version of reference kernel + * - Comparison between optimized kernel and the exact same but scalar version of reference kernel * - Input shapes of 1D and 2D with the first dimension covers boundary values of 128-bit vector size (0~3 iterations) * - Weight and bias 1D shape that have same size as that of input shapes * - Quantization scale is greater and smaller than one. @@ -179,7 +179,7 @@ TEST_SUITE(QSYMM16) * - The algorithm has been sensitive to quantization scale but it is hard to fully test * the sensitivity due to aforementioned reason. * - Again, it is hard to fully test corner values due to the exact same algorithm of the - * reference kernel and the Neon kernel. + * reference kernel and the optimized kernel. */ constexpr uint32_t qsymm16_per_vector = vector_size_byte / sizeof(int16_t); |