From cfca87b91def4f455630f2094447dc0500b6256c Mon Sep 17 00:00:00 2001 From: Gunes Bayir Date: Tue, 9 Apr 2024 23:13:04 +0100 Subject: Add SME2 implementation of softmax for FP16 In addition to the softmax kernel, this patch fixes minor issues in the fp32 implementation. Resolves: COMPMID-6920 Change-Id: Ibbd9f0af5f2a93fba0e92d72ba437279c34149d3 Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/11402 Benchmark: Arm Jenkins Tested-by: Arm Jenkins Reviewed-by: Viet-Hoa Do Comments-Addressed: Arm Jenkins --- docs/user_guide/release_version_and_change_log.dox | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'docs') diff --git a/docs/user_guide/release_version_and_change_log.dox b/docs/user_guide/release_version_and_change_log.dox index b8910c9237..9da4956c43 100644 --- a/docs/user_guide/release_version_and_change_log.dox +++ b/docs/user_guide/release_version_and_change_log.dox @@ -45,7 +45,7 @@ v24.04 Public major release - Add Bfloat16 data type support for @ref NEMatMul. - Optimize start-up time of @ref NEConvolutionLayer for some input configurations where GeMM is selected as the convolution algorithm - Optimize @ref NEConvolutionLayer for input tensor size > 1e7 bytes and weight tensor height > 7 - - Add support for SoftMax in SME2 for FP32. + - Add support for SoftMax in SME2 for FP32 and FP16. - Performance optimizations: - Optimize @ref NESoftmaxLayer for axis != 0 by natively supporting higher axes up to axis 3. - Add support for in place accumulation to CPU GEMM kernels. -- cgit v1.2.1