Commits · 7cf73b1e248ec64a476f3c2347d3a4284cce247b · NiuTrans / NiuTrans.Tensor · GitLab

28 Sep, 2018 3 commits
- bug fix · 7cf73b1e
  xiaotong committed Sep 28, 2018
  
  7cf73b1e Browse Files
- remplementation of pos embedding · 288916ca
  xiaotong committed Sep 28, 2018
  
  288916ca Browse Files
- remove </s> for transformer lm · 543320bc
  xiaotong committed Sep 28, 2018
  
  543320bc Browse Files
27 Sep, 2018 4 commits
- remove temp dumping · 98ac897b
  xiaotong committed Sep 27, 2018
  
  98ac897b Browse Files
- improve the implementation of softmax · 63e2cfa7
  xiaotong committed Sep 27, 2018
  
  63e2cfa7 Browse Files
- rename the model for dumping · 7f483801
  xiaotong committed Sep 27, 2018
  
  7f483801 Browse Files
- remove a space :) · cb51dc13
  xiaotong committed Sep 27, 2018
  
  cb51dc13 Browse Files
26 Sep, 2018 1 commit
- bug fixes of attention and layer normalization · dd0ad421
  xiaotong committed Sep 26, 2018
  
  dd0ad421 Browse Files
25 Sep, 2018 2 commits
- improve the implementation of the overload of = · fe37006f
  xiaotong committed Sep 25, 2018
  
  fe37006f Browse Files
- test code · c090457f
  xiaotong committed Sep 25, 2018
  
  c090457f Browse Files
24 Sep, 2018 2 commits
- revise the initialization method · 907f7d2a
  xiaotong committed Sep 24, 2018
  
  907f7d2a Browse Files
- improve the way of using buf tensors · 0ec51854
  xiaotong committed Sep 24, 2018
  
  0ec51854 Browse Files
23 Sep, 2018 1 commit
- improve coding · 0e1074ff
  xiaotong committed Sep 23, 2018
  
  0e1074ff Browse Files
22 Sep, 2018 1 commit
- bug fixes · efe32603
  xiaotong committed Sep 22, 2018
  
  efe32603 Browse Files
20 Sep, 2018 3 commits
- add label smoothing · a0a38702
  xiaotong committed Sep 21, 2018
  
  a0a38702 Browse Files
- improve the code · c8cb9219
  xiaotong committed Sep 20, 2018
  
  c8cb9219 Browse Files
- bug fixes · 20e9678d
  xiaotong committed Sep 20, 2018
  
  20e9678d Browse Files
19 Sep, 2018 2 commits
- add dropout to fnn and attention sub-layers in t2t · be552310
  xiaotong committed Sep 20, 2018
  
  be552310 Browse Files
- improve the t2t implementation · 8b0e06ab
  xiaotong committed Sep 19, 2018
  
  8b0e06ab Browse Files
18 Sep, 2018 4 commits
- fix a bug · 6fb9ad1b
  xiaotong committed Sep 18, 2018
  
  6fb9ad1b Browse Files
- improve the space management · baad6629
  xiaotong committed Sep 18, 2018
  
  baad6629 Browse Files
- add length-based sorting · 6ea64b51
  xiaotong committed Sep 18, 2018
  
  6ea64b51 Browse Files
- rename SetTMP · ad3fc86f
  xiaotong committed Sep 18, 2018
  
  ad3fc86f Browse Files
17 Sep, 2018 5 commits
- better allocation of the memory · 22df3e17
  xiaotong committed Sep 17, 2018
  
  22df3e17 Browse Files
- fix the bug of computing the size of a memory piece · 194a2c28
  xiaotong committed Sep 17, 2018
  
  194a2c28 Browse Files
- Merge branch 'xuchen' into xiaotong-working · 1b50554a
  xuchen committed Sep 17, 2018
  
  1b50554a Browse Files
- better code for dropout function (by broadcasting) · 102db468
  xuchen committed Sep 17, 2018
  
  102db468 Browse Files
- new code · cf43c58c
  xiaotong committed Sep 17, 2018
  
  cf43c58c Browse Files
16 Sep, 2018 4 commits
- faster dropout implementation · 56326405
  xiaotong committed Sep 16, 2018
  
  56326405 Browse Files
- improve XMem for less memory footprint · 7f9c0b47
  xiaotong committed Sep 16, 2018
  
  7f9c0b47 Browse Files
- add embedding dropout · c90e83fe
  xiaotong committed Sep 16, 2018
  
  c90e83fe Browse Files
- add dropout to transformer · cecbceb9
  xiaotong committed Sep 16, 2018
  
  cecbceb9 Browse Files
15 Sep, 2018 5 commits
- better code of dropout · 2e20824a
  xiaotong committed Sep 15, 2018
  
  2e20824a Browse Files
- better code of multi-gpu runs · df76b612
  xiaotong committed Sep 15, 2018
  
  df76b612 Browse Files
- Merge branch 'xuchen' into xiaotong-working · 182633ac
  xuchen committed Sep 15, 2018
  
  182633ac Browse Files
- 1. add dropout function 2. add some function in tensor/core/math/unary file 3.… · 4336f2f9
```
1. add dropout function 2. add some function in tensor/core/math/unary file 3. merge with xiaotong-working branch
```
  xuchen committed Sep 15, 2018
  4336f2f9 Browse Files
- better use of enum in cuda · afa887d4
  xiaotong committed Sep 15, 2018
  
  afa887d4 Browse Files
14 Sep, 2018 2 commits
- remove some code · bb278bfa
  xiaotong committed Sep 14, 2018
  
  bb278bfa Browse Files
- kill the unneccessary allocatoin on devices that are never be used · 197fac6d
  xiaotong committed Sep 14, 2018
  
  197fac6d Browse Files
13 Sep, 2018 1 commit
- bug of empty head of the first block · 860980dd
  xiaotong committed Sep 13, 2018
  
  860980dd Browse Files