-
Notifications
You must be signed in to change notification settings - Fork 5.6k
2017 09 28
Tao Luo edited this page Dec 9, 2019
·
1 revision
- Refactoring InferShape
- fix nv_library
- add namespace and reduce warning
- fix SoftmaxWithCrossEntropyOp
- Reduce elementwise warning
- Adding support for the sigmoid_cross_entropy_with_logits operator
- survey on computation graph
- Add expand operator
- Add Skeleton of Python API design
- https://github.com/PaddlePaddle/Paddle/pull/4052
- Remove OperatorBase::InferShape
- https://github.com/PaddlePaddle/Paddle/pull/4458
- Feature/pybind for protobuf desc
- rename LoDTensor::Slice to Shrink
- trying out Sextant e2e in virtual machine cluster with help of Wuyi
- Documented comparison of Sextant and Tectonic https://github.com/PaddlePaddle/Paddle/blob/develop/doc/survey/cluster_bootstrapping_tools.md
- Discussion with Helin regarding paddle cloud
- DS2
- Evaluate VoxForge dataset
https://github.com/PaddlePaddle/models/issues/314 - Evaluate Aishell dataset
https://github.com/PaddlePaddle/models/issues/316 - Fix bugs for demo_server.py
https://github.com/PaddlePaddle/models/pull/341 - Tune none weight share model
- Evaluate VoxForge dataset
- Op porting
- Refine gradient checker
https://github.com/PaddlePaddle/Paddle/pull/4333
- Refine gradient checker
-
pr
- https://github.com/PaddlePaddle/Mobile/pull/10 add rounding doc(merged)
- https://github.com/PaddlePaddle/Mobile/pull/8 add the merge batch normalization tools （merged）
-
test about protobuf-lite https://github.com/PaddlePaddle/Paddle/pull/4357
-
write documents and examples for https://github.com/PaddlePaddle/Mobile (in process)
- Deep speech 2:
- Merge & refine tuning script
- Add the script to parse tuning log
- Fix bugs in setup
- PR Review:
- Op porting:
- Merge Multiplex Op
- Update MarginRankLoss Op
- Mobile Review
- Bug fix and review
- fc layer sequence input may cause param dim error
- fix doc error in cos_sim layer
- Add pool op
- Fix max pooling and unit test
- fix Op Attr Checker（To support the detection of the vector Attr）
- [WIP] Add max pool op (with index)
- [WIP]Add vol2col functor
- Reviews
- Cudnn conv op
- LSTM unit op merged
- Re-read some Tensorflow codes, now implementing gather/scatter updates on GPU kernels;
- Yu Yang: Prelu stability https://github.com/PaddlePaddle/Paddle/pull/4201
- Jiayi: modified huber loss https://github.com/PaddlePaddle/Paddle/pull/4407
- Chunwei: participate in beam-search and dynamic-ran implementation discussion;
- Refactorization:
- Discuss Multi-node env of SendRecv/ Block/ Session. https://github.com/PaddlePaddle/Paddle/issues/4031#issuecomment-330915690
- Multi-GPU/ Multi-node support: send recv OP Implement
- Reviews
- add block execution doc
- Mxnet executor survey
- computation graph
- OpPorting: bilinear operator
- OpPorting: sigmoid_cross_entropy
- OpPorting: crop_op
- fix design doc grammar
- fix book demo
- Update FAQ:
- add dense_sequence in capi
- doing
- refine ctc models for STR
- nest sequence input for text classification
- PR
- review
- others
- fix models/site https://github.com/PaddlePaddle/models/issues/322
Survey:
- survey on computation graph
- Mxnet executor survey
Block design:
- Update: add block execution doc
- Use stridecpy instead of eigen in PadGradKernel.
- Fix doc of clip op.
- Add sequence_expand Operator based LoDTensor[WIP]
- Compelete
ProgramDesc,BlockDesc,VarDescandOpDesc. And expose them to Python. with @yuyang - tiny fix:
- review:
- Gradient checker
- Identify precision related problems: https://github.com/PaddlePaddle/Paddle/issues/4283
- Code review: https://github.com/PaddlePaddle/Paddle/pull/4333
- PR
- Fix a few typos in docs: https://github.com/PaddlePaddle/Paddle/pull/4220 (merged)
- Fix an error in computation graph doc: https://github.com/PaddlePaddle/Paddle/pull/4287