-
Notifications
You must be signed in to change notification settings - Fork 5.6k
2017 01 18
Tao Luo edited this page Dec 9, 2019
·
1 revision
Paddle on Kubernetes
- https://github.com/PaddlePaddle/Paddle/pull/1161#pullrequestreview-16886955
- https://github.com/PaddlePaddle/Paddle/pull/1161#pullrequestreview-16888342
- https://github.com/PaddlePaddle/Paddle/pull/1161#pullrequestreview-16906039
- https://github.com/PaddlePaddle/Paddle/pull/1176#pullrequestreview-17091984
Paddle API
- https://github.com/PaddlePaddle/Paddle/pull/1126#pullrequestreview-16906287
- https://github.com/reyoung/Paddle/pull/3#pullrequestreview-16908280
- https://github.com/PaddlePaddle/Paddle/pull/1149#pullrequestreview-17126500
- https://github.com/PaddlePaddle/Paddle/pull/1149#pullrequestreview-17266694
Fix Issue:
- CentOS 7 下编译出错 #1151
- compile error #1140
- undefined reference #1124
- another undefined reference #1120
- Segment fault before main Bug #1113
- 在使用develop分支中提供的dockerfile进行安装Paddle时,安装失败 #1154
- protobuf与glog一些引用找不到 #1175
PR:
- Detect CentOS 7 in CMake #1155
- Add external openblas #1150
- Enable cuda dynamic link libs #1144
- Update Dockerfile for develop branch #1169
Doing:
- C-API for predict #1062
- C-API is used well in Baidu some product now.
- Try to learn dlmopen, to link multiple version of protobuf.
- ErrorHandling in Paddle #1149
- Interface basically done.
- Add LRS in Paddle settings #1170
- Feature/python.paddle.v2 #1108
- Should we extract this PR into multiples?
- Data API
- Layer API
- Model API
- Should we extract this PR into multiples?
Done:
- Make sign compare as a compile warning not error.#1148
- Fix bug in DenseScanner of DataProviderConverter. #1121
- Fix bugs in config_helpers unittest. #1163
- merge machine translation #23
- carefully review tutorials
- Code review: fix default value of cos_sim scale #1156
- OCR Model
- The problem about accuray has been solved.
- Use AdaDelt and average the gradient in the backward of MultiClassCrossEntropy.
- Tune some hyper parameters.
- The accuray of both CTC model and attention model is slightly better than Torch. (PaddlePaddle-86.13% vs Torch-85.16%(from Email) )
- PR View
- Update PR
- Bug fix
- read code of config_parser and add a layer abstruction impliment #1108
- prepare a docker build environment, learn to use docker and try build and use paddlepaddle with docker.
- Study carefully tutorials
- Investigate visualization for livebook
- Discuss with ui student about livebook visualization
- Will push the report of livebook visualization to github
Paddle on Kubernetes tutorial improvements.
- https://github.com/PaddlePaddle/Paddle/pull/1176
- https://github.com/PaddlePaddle/Paddle/pull/1161
- https://github.com/PaddlePaddle/Paddle/pull/1142
PR
Add a FunctionTest for unit test of Function
ISSUE
Complement the description of Function
- https://github.com/PaddlePaddle/Paddle/issues/892#issuecomment-272082011
- https://github.com/PaddlePaddle/Paddle/issues/1165
REVIEW
Open PR
Matrix::MUL operators (finished 60%)
Merged PR
Context Projection -- follow comments
PR reviewed
Daoyuan’s Function Test
- InterestingExamples Simshang https://github.com/PaddlePaddle/Paddle/issues/958
- InterestingExamples https://github.com/iooiiooii/malasong
- write 1st class document
- review and merge pr of paddlebook
- fix problems with recommendation code
- recoded the download part of data api
- coding the preprocess part of data api