Commit Graph

127 Commits

Author SHA1 Message Date
guoqi 8fccec4c20 insert overflow check operaters according to the 'gradients' scope 2021-08-31 16:44:02 +08:00
gaojing fa02606348 step train modified 2021-08-29 23:04:55 -04:00
baihuawei a9694a9230 ascend add nontask sink mode 2021-08-25 14:45:58 +08:00
ms_yan 36a8886ca2 Revert "[feat] [assistant] [I3T96T] add new Dataset operator CMUARCTICDataset"
This reverts commit b077aa1cab.

Revert "[feat] [assistant] [I3T96X] add new Dataset operator LibriSpeechDataset"

This reverts commit 4e6f7dc97d.

delete pass_registry_test.cc

comment  hiai_nlu_model_multi.pb related  line
2021-08-23 01:46:38 +08:00
djc b077aa1cab [feat] [assistant] [I3T96T] add new Dataset operator CMUARCTICDataset 2021-08-22 16:26:45 +08:00
djc 4e6f7dc97d [feat] [assistant] [I3T96X] add new Dataset operator LibriSpeechDataset 2021-08-22 13:39:37 +08:00
yanghaitao1 8fc11cb676 adapt delete libms_profiler_fwk.a 2021-08-19 19:33:53 +08:00
caifubi dfe0e94466 Fix PyNative get_rank_id/get_rank_size 2021-08-18 16:39:18 +08:00
i-robot 6afcd815d2 !21362 add pynative profiling codes based on ascend and gpu
Merge pull request !21362 from lvchangquan/profiling_refactor
2021-08-13 06:18:31 +00:00
zhoufeng 03a56f2bb0 alltoall exception handle
Signed-off-by: zhoufeng <zhoufeng54@huawei.com>
2021-08-11 17:30:49 +08:00
lby a5029f061c ascend kernel build refactory 2021-08-10 23:40:49 +08:00
lvchangquan e8d9803258 add profiling codes based on ascend and gpu in pynative mode 2021-08-10 20:46:51 +08:00
lby e6cdf098db op tiling compute interface replace 2021-08-06 11:26:24 +08:00
baihuawei 41de02a58c ascend support nontask sink 2021-08-03 20:55:43 +08:00
i-robot 69c5021bb5 !20995 pyfunc cpu kernel
Merge pull request !20995 from chenweifeng/cpu-dynamic-input
2021-08-03 04:51:20 +00:00
wilfChen d6fffdad6e support dynamic inputs & outputs 2021-08-02 20:05:46 +08:00
yanghaoran 0364650eae Upgrade Ascend packages 28 Jul 21, with testcases removed 2021-07-31 18:55:55 +08:00
dayschan 3ab53dd26d Send compilation attrs to akg
1. Add a new message type "AKG/ATTR" in AkgKernelBuilder.
   the attrs was sent before the kernel infos.
2. Send "online_tuning" attribute when the flag is not zero,
   but error occurs in the latest akg submodule.
3. Send "repository_path" attribute when the flag is not empty.
4. Add a new value "compute_capability" into kernel info when the processor is GPU.
2021-07-16 16:52:18 +08:00
dingpeifei 24ff7ab8b4 upgrade_ascend_0705 2021-07-06 11:31:45 +08:00
zhoufeng b9378e36c3 dynamic default memory size for vnpu
Signed-off-by: zhoufeng <zhoufeng54@huawei.com>
2021-06-29 19:23:04 +08:00
dingpeifei 63784e49f5 upgrade_ascend_0626_mindspore 2021-06-28 22:03:19 +08:00
i-robot 708e56f659 !17899 Fix compile cache bug for resent50
Merge pull request !17899 from LiangZhibo/cache
2021-06-24 13:29:58 +00:00
l00591931 8ae5d7cc84 Fix compile cache for resnet50 2021-06-24 14:36:24 +08:00
lishanni513 8c98146c76 Add register limit constraint 2021-06-23 17:39:22 +08:00
caifubi 61efa2c23b remove some code 2021-06-08 10:55:41 +08:00
yanghaitao1 127e4d4068 fix profiler pclint&codex 2021-06-07 02:17:23 -04:00
yujianfeng 7cea682c7d Add func_graph caching after validate action 2021-05-30 13:00:35 +08:00
zhoufeng fee35368d1 modify defination of hccl origin function in hccl plugin
Signed-off-by: zhoufeng <zhoufeng54@huawei.com>
2021-05-17 14:19:35 +08:00
zhoufeng a76d58b52d hccl decouple
Signed-off-by: zhoufeng <zhoufeng54@huawei.com>
2021-05-11 22:25:00 +08:00
zhoufeng 3d38bcf791 use internal ge runtime
Signed-off-by: zhoufeng <zhoufeng54@huawei.com>
2021-04-26 18:56:33 +08:00
caifubi 58f3ea002d clean codex 2021-03-31 15:41:37 +08:00
dayschan 3c6c30024c dump graph_kernel_split info 2021-03-29 17:15:02 +08:00
liubuyu a03452ff2c dynamic shape bug fix 2021-03-19 17:57:31 +08:00
zhoufeng b7e5f956e5 workspace of comm op can be reused
Signed-off-by: zhoufeng <zhoufeng54@huawei.com>
2021-03-16 14:42:52 +08:00
ms_yan b4b4577b85 adapter profiling with new ascend lib 2021-03-14 11:38:20 +08:00
caifubi 171b468bb3 PyNative AllReduce Bucket 2021-03-03 15:47:18 +08:00
jjfeing 502be04491 upgrade 0204 2021-02-06 10:51:43 +08:00
yangzhenzhang 7303c3d3b8 add group ckpt 2021-01-21 14:41:05 +08:00
yanghaoran b8345d03b6 Synchronize latest Ascend software 18 Dec 2020, with profiler fixes 2020-12-24 15:11:01 +08:00
zhoufeng ee7e412def add reset device, fix runtime delete ring buffer failed
Signed-off-by: zhoufeng <zhoufeng54@huawei.com>
2020-12-14 20:58:13 +08:00
yanghaoran 306bbe0e65 Synchronize latest Ascend software suite 09 Dec 2020, with refactored GraphEngine 2020-12-09 18:29:29 +08:00
caifubi 04d8cd5d8b Support PyNative Data Parallel 2020-12-01 20:55:39 +08:00
zhoufeng d532af3a9a new hccl interface
Signed-off-by: zhoufeng <zhoufeng54@huawei.com>
2020-11-24 17:34:35 +08:00
caifubi 12643a33dd Add profiling reporter 2020-11-23 16:53:11 +08:00
wilfChen e4e9362bd0 gpu suppor dynamic shape 2020-11-06 14:16:27 +08:00
Yi Huaijie d7faa77b5e support int64 shape 2020-11-05 21:15:49 +08:00
jjfeing 56f1bb1b66 add dump data when task error 2020-11-02 09:42:53 +08:00
caifubi d3b978147f Ascend Dynamic Shape 2020-10-10 19:11:47 +08:00
caifubi b17b4abf7d Fix async data dump failed 2020-09-30 11:39:19 +08:00
jjfeing 7dda95d247 set soc version 2020-09-28 21:04:07 +08:00
caifubi 39039974b2 fix profiling start up fail 2020-09-18 08:56:13 +08:00
wuweikang e24f534002 sync-from-trunk-to-blue-zone-0912-c75b220 2020-09-16 11:28:22 +08:00
wuxuejian bd527a331d update aicpu proto and update module: graphengine
Support Dynamic Shape Aicpu Run Package

adapt tensorengin modify, fix ub fusion
2020-09-08 11:38:44 +08:00
zhousiyi c25e37e7bf make backend/optimizer pybind free 2020-08-28 08:56:58 +00:00
yao_yf 60a9fb0001 add_tensor_layout_in_stra_ckpt 2020-08-06 15:57:38 +08:00
liubuyu d81862a916 decoupling core and context 2020-08-05 08:15:14 +08:00
zhoufeng 000a6c73a5 op debug
Signed-off-by: zhoufeng <zhoufeng54@huawei.com>
2020-07-23 11:12:02 +08:00
yanghaoran 859acc6d2a synchronize latest Ascend software suite 18 Jul 2020, and merging branches 2020-07-18 21:09:55 +08:00
changzherui f4cb445ea8 syn code for 0715 2020-07-16 17:18:21 +08:00
caifubi cb8b5dbd76 Remove build option -S 2020-07-16 14:51:48 +08:00
liubuyu 43c79eb853 mindspore path adjust 2020-07-14 18:07:28 +08:00
caifubi c577952c9a Async Data Dump 2020-07-13 09:42:37 +08:00
yanghaoran 9ce5c97204 synchronize Ascend software suite 07 Jul 2020 2020-07-08 21:49:09 +08:00
gukecai c703c4d7c8 hcom one stream 2020-06-24 08:59:36 +08:00
zhoufeng bbbfaa2441 enable new control sink
Signed-off-by: zhoufeng <zhoufeng54@huawei.com>
2020-06-20 14:48:21 +08:00
caifubi a49c6f0b13 Fix profiling stream id bug 2020-06-19 14:34:20 +08:00
caifubi 5db0d8129c Support hccl profiling 2020-06-17 15:14:30 +08:00
leopz c45f4d225d dump binary ir 2020-06-12 13:19:41 +08:00
yanghaoran 8da4c1a763 synchronize latest ascend software 04 Jun 2020 2020-06-09 23:34:13 +08:00
lizhenyu c5ac2cc38c add memory copy module 2020-05-23 13:04:50 +08:00
zhoufeng af5019b94f link child graphs
Signed-off-by: zhoufeng <zhoufeng54@huawei.com>
2020-05-22 15:37:19 +08:00
wenchunjiang 729ea8cc55 fix bug of loadding ctrl input tensors failed in control sink mode 2020-05-20 10:45:28 +08:00
zhoufeng c99f904276 assign label resource for new control sink
Signed-off-by: zhoufeng <zhoufeng54@huawei.com>
2020-05-18 21:44:52 +08:00
yao_yf 6cde5f6d91 auto parallel strategy checkpoint 2020-04-28 15:48:41 +08:00
gukecai f8208c7c52 Support GetNext Parallel 2020-04-21 11:39:07 +08:00
jojobugfree effdb483d6 profiling feature enhancement 2020-04-08 15:12:06 +08:00
zhunaipan 930a1fb0a8 initial version
Signed-off-by: leonwanghui <leon.wanghui@huawei.com>
2020-03-27 22:54:54 +08:00