Describe the current behavior / 问题描述 (Mandatory / 必填)

Allegro网络单卡训练loss=nan,性能不达标

Environment / 环境信息 (Mandatory / 必填)

• Hardware Environment(Ascend/GPU/CPU) / 硬件环境:

Please delete the backend not involved / 请删除不涉及的后端:
/device Ascend910B

• Software Environment / 软件环境 (Mandatory / 必填):
  -- MindSpore version (e.g., 1.7.0.Bxxx) :2.3.0rc1
  -- Python version (e.g., Python 3.7.5) :3.8.0
  -- OS platform and distribution (e.g., Linux Ubuntu 16.04):Linux 10-90-66-192 4.15.0-45-generic #48-Ubuntu
  -- GCC/Compiler version (if compiled from source): 7.5.0

• Excute Mode / 执行模式 (Mandatory / 必填)(PyNative/Graph):

Please delete the mode not involved / 请删除不涉及的模式:
/mode graph

Related testcase / 关联用例 (Mandatory / 必填)

暂无

Steps to reproduce the issue / 重现步骤 (Mandatory / 必填)

1.git clone https://gitee.com/mindspore/mindscience.git
2.cd MindChemistry/applications/allegro
3.python {allegro_path}train.py

Describe the expected behavior / 预期结果 (Mandatory / 必填)

Allegro网络训练loss正常,Epoch2到Epoch5500每个epoch小于24s

Related log / screenshot / 日志 / 截图 (Mandatory / 必填)

train loss: nan, time gap: 26.6766

Special notes for this issue/备注 (Optional / 选填)