提交 · 728defbec90a162ee5d7f8521106ded7797e72fa · Crayon鑫 / Paddle

12 11月, 2016 2 次提交
- H
  
  copy the data when createSparseMatrix · 728defbe
  由 Haonan 提交于 11月 11, 2016
  
  728defbe
- H
  
  multi_binary_cross_entropy when ids vector is provided · 069d0004
  由 Haonan 提交于 11月 04, 2016
  
  069d0004
17 10月, 2016 1 次提交

Fix sparse training for trainer_count=1 (#204) · 28bc05b1

由 emailweixu 提交于 10月 17, 2016

* Fix sparse training for trainer_count=1

For trainer_count=1, the gradient machine is NeuralNetwork, which does not create parameter buf for PARAMETER_GRADIENT for sparse update in Parameter::enableType. But gradient parameter buf is still used in SgdThreadUpdater.

* Minor update to comment

28bc05b1

27 9月, 2016 1 次提交

Correctly handling multiple inputs and integer inputs for recurrent_g… (#114) · b15a4783

由 emailweixu 提交于 9月 26, 2016

* Correctly handling multiple inputs and integer inputs for recurrent_group

* Fix ScatterAgentLayer for generation

* Revert sequence_(nest)_rnn.conf

b15a4783

17 9月, 2016 1 次提交
- X
  Further fix the memory for Hierarchical RNN · 05a97ab5
  由 xuwei06 提交于 9月 14, 2016
```
Sequences should be sorted according to the number of subsequences they have.
```
  05a97ab5
14 9月, 2016 3 次提交
- X
  Correctly handle memory in RecurrentGradientMachine for hirarchical RNN · 9a9de924
  由 xuwei06 提交于 9月 09, 2016
```
Change-Id: I8e0a8ea6fc2760652d9c76440a539c90860062d3
```
  9a9de924
- H
  use HPPL_STREAM_DEFAULT for layer computation · 0f91ea7e
  由 Haonan 提交于 9月 07, 2016
```
Change-Id: Id66da7b7f5bf9ec80cc19b347e4fb822a5a6f197
```
  0f91ea7e
- H
  fixed issues with synchronizing streams when copy from gpu to cpu · 688eeefa
  由 Haonan 提交于 9月 06, 2016
```
* by default, synchronize default_stream after resizeAndCopyFrom
* add sync in some places after resizeAndCopyFrom using other streams
```
  688eeefa
13 9月, 2016 1 次提交
- Y
  
  Fix linux compile · a238b11f
  由 Yu Yang 提交于 9月 13, 2016
  
  a238b11f
10 9月, 2016 1 次提交
- L
  
  Add main entry for unit test files and replace memalign by posix_memalign · 58f74e2c
  由 liaogang 提交于 9月 10, 2016
  
  58f74e2c
08 9月, 2016 1 次提交
- H
  Argument concat for subsequence start positions · 1e1a33b5
  由 Haonan 提交于 9月 06, 2016
```
Change-Id: Ia60c008a8c922f66e6b5e2ca3e488fc4625d6506
```
  1e1a33b5
30 8月, 2016 1 次提交

multi-gpu training with sparse input. · 2afe6609

由 hedaoyuan 提交于 8月 30, 2016

ISSUE=4586833

git-svn-id: https://svn.baidu.com/idl/trunk/paddle@1429 1ad973e4-5ce8-4261-8a94-b56d1f490c56

2afe6609

29 8月, 2016 1 次提交

fix dash and space bug, · b72beee4

由 zhangjinchao01 提交于 8月 29, 2016

ISSUE=4586495

git-svn-id: https://svn.baidu.com/idl/trunk/paddle@1408 1ad973e4-5ce8-4261-8a94-b56d1f490c56

b72beee4

Crayon鑫 / Paddle 与 Fork 源项目一致

Crayon鑫 / Paddle
与 Fork 源项目一致