Highlights
- Arctic Code Vault Contributor
- Developer Program Member
Create your own GitHub profile
Sign up for your own profile on GitHub, the best place to host code, manage projects, and build software alongside 50 million developers.
Sign up
Pinned
1,526 contributions in the last year
Contribution activity
November 1, 2020
October 2020
Created a pull request in PaddlePaddle/Paddle-Lite that received 2 comments
[OPENCL][KERNEL] add leaky relu in conv for opencl. test=develop
状态:等待review 主要内容 增加这个op的原因是因为yolov3中有leaky relu。 增加conv和leaky relu的融合。在cl_common中增加对这种激活函数的支持,修改conv_image_compute加入对该方法的支持; 单测支持leaky relu。本地loop…
- 【arm-cv】修复bra rotate 90 and resize error problem
- [OpenCL][Backend] Fix opencl version macro
- [quant] Add post_quant_dynamic to opt
- [OpenCL][Profiling] Fix invalid opencl profiling setting
- [OpenCL][Profiling] Fix invalid opencl profiling setting
- [Doc][OpenCL]Update opencl.md
- [OpenCL][Kernel] Tune concat image impl when num of inputs > 4
- [arm] fix bgr/bgr to gray convert compute error
- [cherry-pick] fix conv+conv compute error
- [OpenCL] Add opencl func wrapper: clGetCommandQueueInfo
- [cherry-pick] fix a53_valid long time run
- [arm] fix a53_valid long time run problem
- add sin, cos ops and completion pow, elementwise_pow ops test=develop
Created an issue in ysh329/OpenCL-101 that received 7 comments
TensorFlow Lite GPU OpenCL WorkGroup TuningType策略
最近发现TensorFlow Lite在GPU方面的性能有不小提升,先前了解到起初是支持的OpenGL来完成计算,猜想可能是考虑到GL的更广阔的的兼容性(不同的GPU版本,兼容的新老的库版本),但后续这次对GPU以OpenCL进行支持,应该考虑的更多是计算性能,也是与TFLite的相关竞品,如…