未验证 提交 1ec626b1 编写于 作者: J joanna.wozna.intel 提交者: GitHub

Revert md-in-tensor refactoring (#43564)

上级 13ad8bde
......@@ -47,8 +47,12 @@ class SoftmaxMKLDNNHandler
platform::errors::InvalidArgument(
"The shape of input and output tensor must be identical."));
this->AcquireForwardPrimitiveDescriptor(prop_kind::forward_scoring,
input->mem_desc(), axis);
auto softmax_tz = phi::vectorize(input->dims());
auto md = memory::desc(softmax_tz, platform::MKLDNNGetDataType<T>(),
input->format());
this->AcquireForwardPrimitiveDescriptor(prop_kind::forward_scoring, md,
axis);
}
SoftmaxMKLDNNHandler(const framework::ExecutionContext& ctx,
......
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册