Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Paddle-Lite新需求收集(call for new feature request) #10452

Open
onecatcn opened this issue Feb 27, 2024 · 10 comments
Open

Paddle-Lite新需求收集(call for new feature request) #10452

onecatcn opened this issue Feb 27, 2024 · 10 comments
Assignees

Comments

@onecatcn
Copy link

问卷链接:https://iwenjuan.baidu.com/?code=suexg7

欢迎开发着提出自己的建议,一起优化Paddle-Lite

@Taichipeace
Copy link

希望能训推一体,避免出现训练和推理框架版本导致的推理问题。

@engineer1109
Copy link
Contributor

OpenCL int8 支持

@sprouteer
Copy link
Collaborator

OpenCL int8 支持

#8058 这个pr有支持fc的int8,在高通机型上测试性能提升不高只有20%~30%,你可以试试

@hong19860320
Copy link
Collaborator

OpenCL int8 支持

#8058 这个pr有支持fc的int8,在高通机型上测试性能提升不高只有20%~30%,你可以试试

wow~

@woodboard
Copy link

希望表格识别能和文字识别聚合起来做一个paddle-lite版本的

@z754688651
Copy link

希望后续可以支持动作分类模型进行端侧部署,支持 'temporal_shift' 算子和3d卷积'conv3d, pool3d'算子的转化

@fhowen
Copy link

fhowen commented Apr 7, 2024

希望后续可以支持YOLO模型进行端侧部署,目前YOLO5模型中'fill_constant' 算子和'flip'算子的转化有问题

@LitchiCheng
Copy link

#10485 模型转换报错 elementwise_add_bias_dims not equal to 1, fusion failed

@lg920810
Copy link

lg920810 commented Apr 9, 2024

#10469 希望提供opencl上全局限制Image2D的接口,通过config文件配置若想避免layout算子强制执行buffer2Image,需要把一连串的算子都指定到cpu,但推理就会很慢了

@AvaMins
Copy link

AvaMins commented May 20, 2024

希望后续可以支持动作分类模型进行端侧部署,支持 'temporal_shift' 算子和3d卷积'conv3d, pool3d'算子的转化,支持BMN、AttentionLSTM模型的压缩转化。

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests