linux系统，cuda10.0，anaconda虚拟环境中安装pytorch1.0.0，import torch出现libmkldnn.so.0: undefined symbol: cblas_sgemm_alloc问题。
解决方法：在虚拟环境中卸载mkl，mkl-service两个包，重新安装pytorch1.0.0，再使用conda install -c anaconda mkl命令安装mkl。
参考：
https://segmentfault.com/a/1190000019516715?utm_source=tag-newesthttps://segmentfault.com/a/1190000019516715?utm_source=tag-newest
https://blog.csdn.net/Lstar_/article/details/118658610?utm_medium=distribute.pc_aggpage_search_result.none-task-blog-2~aggregatepage~first_rank_ecpm_v1~rank_aggregation-13-118658610.pc_agg_rank_aggregation&utm_term=cblas_sgemm&spm=1000.2123.3001.4430https://blog.csdn.net/Lstar_/article/details/118658610?utm_medium=distribute.pc_aggpage_search_result.none-task-blog-2~aggregatepage~first_rank_ecpm_v1~rank_aggregation-13-118658610.pc_agg_rank_aggregation&utm_term=cblas_sgemm&spm=1000.2123.3001.4430
