Commit Graph

61021 Commits

Author SHA1 Message Date
mindspore-ci-bot 062b744b19 !12 Fix dtype bug for loss_scale and weight_decay
Merge pull request !12 from seatea/dynamic-loss-scale
2020-03-30 22:06:35 +08:00
mindspore-ci-bot 3ab402e110 !7 adapt relu6grad
Merge pull request !7 from zhaozhenlong/adapte-relu6grad
2020-03-30 21:58:39 +08:00
lichenever b4d34973bc fix_cast_bug 2020-03-30 21:13:45 +08:00
helloway 66714da29f Add issue and PR templates for gitee and github 2020-03-30 20:17:13 +08:00
chenhaozhe cab5503280 use find instead of equal to distinguish training graph 2020-03-30 20:10:56 +08:00
mindspore-ci-bot 1a3c06a948 !6 add bert example
Merge pull request !6 from yoonlee666/master
2020-03-30 19:52:51 +08:00
mindspore-ci-bot 44cd0c1f90 !13 Check input shape for `NMSWithMask` op
Merge pull request !13 from seatea/NMSWithMask-check-shape
2020-03-30 19:52:41 +08:00
buxue 0da0bdcf40 Fix bug structure output when there is depend whose first input is constant in outputs 2020-03-30 19:49:46 +08:00
Ziyan 4cbcd8e907 enable use float type learning rate in lars optimizer 2020-03-30 19:05:33 +08:00
zhaozhenlong 9862dea3cf adapt relu6grad and graphengine modified 2020-03-30 19:01:03 +08:00
shibeiji 468e257a14 add log for kernel runtime in order to trace performance 2020-03-30 17:55:28 +08:00
jonyguo 34e42bd6f9 1. add more log info for dataset & mindrecord, 2. add two new testcase for MindDataset 2020-03-30 17:24:24 +08:00
seatea 6c03542eec Fix dtype bug for loss_scale and weight_decay.
1.Change dtype of scale to dtype of grad in loss_scale.py;
2.Change dtype of weight_decay to dtype of weight in optimizer.py.
2020-03-30 17:14:52 +08:00
mindspore-ci-bot c24252b2cc !11 Add dockerfile to build MindSpore docker images
Merge pull request !11 from leonwanghui/dockerfile
2020-03-30 16:52:11 +08:00
yoonlee666 c6d261b277 add bert script to master 2020-03-30 16:21:57 +08:00
leonwanghui 545ff417d0 Add dockerfile to build MindSpore docker images
Signed-off-by: leonwanghui <leon.wanghui@huawei.com>
2020-03-30 15:58:37 +08:00
mindspore-ci-bot 04eab166e5 !9 change parallel complie num: 32->16
Merge pull request !9 from jjfeing/master
2020-03-30 15:11:34 +08:00
seatea 7b7a6a45a0 Check if the shape of the input of NMSWithMask is (N, 5). 2020-03-30 12:10:35 +08:00
jjfeing 86f5c69995 change parallel complie num: 32->16 2020-03-30 10:54:53 +08:00
GinFung 468dbc3557 Add matmul biasadd fusion pass 2020-03-30 10:12:50 +08:00
zhunaipan 930a1fb0a8 initial version
Signed-off-by: leonwanghui <leon.wanghui@huawei.com>
2020-03-27 22:54:54 +08:00