Created by: yihuaxu

It will can not use the AVX intrinsic function without 8-bytes aligned issue when used the GRU operator. Just re-implement the relevant functions to fix the issue.

Platform: Intel(R) Xeon(R) Gold 6140 CPU @ 2.30GHz Model Path: models/fluid/chinese_ner Batch Size: 6 Command: python infer.py --device CPU --profile Data Source: Use original date provided in the model Chinese_NER in the github.

The following is the comparison with the different scenarios.