From c6feb3cf6ef3a525ce23945ef30eb3c0d99dee37 Mon Sep 17 00:00:00 2001 From: Jose Lamego Date: Tue, 11 Dec 2018 17:23:40 -0600 Subject: [PATCH] releasenote.md: Add note for no-avx-support message A message stating that the TensorFlow used by the stacks-dlrs-mkl image does not support AVX instructions can be missleading for users. This change adds a note clarifying how the performance optimization is handled in this image. Signed-off-by: Jose Lamego --- stacks/dlrs/releasenote.md | 3 +++ 1 file changed, 3 insertions(+) diff --git a/stacks/dlrs/releasenote.md b/stacks/dlrs/releasenote.md index b227f7e..d64ebc1 100644 --- a/stacks/dlrs/releasenote.md +++ b/stacks/dlrs/releasenote.md @@ -33,6 +33,9 @@ The release includes: * Runtimes (python) * TensorFlow 1.12 optimized using Intel® Math Kernel Library for Deep Neural Networks (Intel® MKL-DNN) primitives. +> **Note:** + When using the Deep Learning Reference Stack with Intel(r ) MKL-DNN version, you may see this warning message: "tensorflow/core/platform/cpu_feature_guard.cc:141] Your CPU supports instructions that this TensorFlow binary was not compiled to use: SSE4.1 SSE4.2 AVX AVX2 AVX512F FMA". This is because this version of the Deep Learning Reference Stack is using Intel(r ) MKL-DNN for performance optimization rather than Intel(r ) Advanced Vector Extensions 512, and is expected for this version of the Deep Learning Reference Stack. + ## How to get the Deep Learning Reference Stack