long time before deepmd running

[fire-bot]

Sent by Gitee (no-reply@mailer.oschina.net). Created by fire.

---

通知

skyinner 新建了一个任务 #I48G0T long time before deepmd run...

@Gitee.com

如果你不想再收到类似邮件，可以到 这里 取消

点击这里取消订阅
click to unsubscribe

---

---

I am training a potential with deepmd 2.01 and TF 2.4. After training about half an hour. I found that the potential still doesn't begin training. And the train.log shows:

WARNING:tensorflow:From /apps/soft/tensorflow-2.4.0/lib/python3.7/site-packages/tensorflow/python/compat/v2_compat.py:96: disable_resource_variables (from tensorflow.python.ops.variable_scope) is deprecated and will be removed in a future version.
Instructions for updating:
non-resource variables are not supported in the long term
Switch to serial execution due to lack of horovod module.

When I trained the potential with depend 1.2.2 and TF 1.2. The beginning process had not been so long. Is their any suggestions to help me accelerate this process?

Thanks.

[njzjz]

@denghuilu It seems that we are calling NeighborStat twice during training?

[denghuilu]

@denghuilu It seems that we are calling NeighborStat twice during training?

NeighborStat will be used twice only when the user does not set the sel parameter

[njzjz]

NeighborStat will be used twice only when the user does not set the sel parameter

No. In #914, I've enabled the sel checking for every users.

[njzjz]

Reopen as I think we still need to conduct some optimization.