An alignment problem for xnnpack EP on Intel/AMD CPUs on PC platforms.Though the API is available in all Windows 10 versions, some sandbox environments block using the API. It is because in 1.15.0 we introduced a Windows API call to SetThreadDescription. Native onnxruntime library not loading in Azure App Service.A build issue that build with onnxruntime_ENABLE_MEMORY_PROFILE was broken in 1.15.0 release.(In other words: with TensorFloat-32 support) ( #15981) The error only occurs when CUDA version = 11.8 and the GPU's CUDA Compute capability >=80. A misaligned error in cublasGemmBatchedHelper function.
Without this change, some DNNL/OpenVino tests would fail on some AMD CPUs. A coding problem in test/shared_lib/test_ that it should use ASSERT_NEAR to test float values instead of ASSERT_EQ.
0 kommentar(er)
