[Fix] Convert loss_weights as same dtype of stacked losses

lululxvi / deepxde

A library for scientific machine learning and physics-informed learning

https://deepxde.readthedocs.io

GNU Lesser General Public License v2.1

2.47k stars 712 forks source link

[Fix] Convert loss_weights as same dtype of stacked losses #1780

Closed HydrogenSulfate closed 1 week ago

HydrogenSulfate commented 1 week ago

As descripted in https://github.com/PaddlePaddle/Paddle/pull/63842, dtype should be explicitly specified to float(16/32), or error will be raised. So setting dtype as same as stacked loss seems fine.

https://www.paddlepaddle.org.cn/documentation/docs/zh/develop/guides/advanced/auto_type_promotion_cn.html#sanfeijiangdeyinshileixingtishengshiyongfangfashuoming