提交 9a25f289 编写于 作者: T tensor-tang

update the default cpu memory with MKLDNN

上级 f1f8327c
......@@ -30,7 +30,9 @@ DEFINE_double(fraction_of_cpu_memory_to_use, 1,
DEFINE_uint64(initial_cpu_memory_in_mb,
#ifdef PADDLE_WITH_MKLDNN
1000,
/* Aligned with mozga-intel, MKLDNN need at least 5000 MB
* to obtain the best performance*/
5000,
#else
500,
#endif
......
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