search

PaddlePaddle / Paddle

PArallel Distributed Deep LEarning: Machine Learning Framework from Industrial Practice （『飞桨』核心框架，深度学习&机器学习高性能单机、分布式训练和跨平台部署）
http://www.paddlepaddle.org/
Apache License 2.0
22.13k stars 5.55k forks source link

[Custom Device] Custom Device在FP16混合精度下不能使用融合算子Pass #60542

Closed engineer1109 closed 9 months ago

engineer1109 commented 9 months ago

bug描述 Describe the Bug

paddle/fluid/inference/api/analysis_predictor.cc

这段代码是我很久以前提交的，过去未考虑现在的情况。 现在Custom Device已经能支持融合算子，只需要完成对应的Kernel。 ClearPass会导致在FP16下，Custom Device的融合算子Pass失效

修复PR https://github.com/PaddlePaddle/Paddle/pull/60541

其他补充信息 Additional Supplementary Information

No response

YanhuiDua commented 9 months ago

你好，感谢贡献，PR已经approve

engineer1109 commented 9 months ago

merged close