Commit Graph

796 Commits

Author SHA1 Message Date
huangyong 84140de08f upgrade ascend 20221129 master
cherry pick 1.10 update

upgrade ascend 20221122

filter out matmul if it is fp16->fp32 with a fp32 bias

modify jjfeng comment
2022-12-20 10:03:35 +08:00
Zhu Guodong 5847b11963 [MSLITE] modify opencl cmake to support cross-platform like windows. 2022-12-15 17:22:39 +08:00
i-robot 7b87b4971f
!46549 [MSLITE] Change ascend ge to plugin and supports independent session.
Merge pull request !46549 from wangshaocong/bugfix
2022-12-13 03:46:55 +00:00
i-robot b3dd74cb76
!44563 [MS][MHA] Adding support in FP16 and cross for Multi Head Attention
Merge pull request !44563 from Nizzan/export_nizzan
2022-12-09 09:12:29 +00:00
i-robot f8dafdc71d
!46514 Add converter.h API docs and add it to release pkg
Merge pull request !46514 from 刘崇鸣/add_converter_header_to_pkg
2022-12-09 06:04:00 +00:00
zhengyuanhua 9688c8cb5e [bugfix] fix cloud compile 2022-12-08 16:02:00 +08:00
wang_shaocong 78642ec8d4 [MSLITE] ascend ge supports independent session. 2022-12-08 11:51:15 +08:00
liuchongming d2380f0755 Add converter.h header to release pkg and supply api docs. 2022-12-07 10:17:39 +08:00
nizzan dff877dbd3 Adding support for FP16, cross & T5 MHA 2022-12-06 09:56:12 +02:00
i-robot 09fb9ad966
!46032 include notice in whl
Merge pull request !46032 from Henry Shi/notice_whl
2022-11-29 02:26:39 +00:00
Henry 156f9a4c56 include notice 2022-11-25 15:14:28 +08:00
liuchongming 0fa1e0d69d Adjust release package structure and remove unused header. Besides, fix the early release problem of pass plugin. 2022-11-19 17:06:32 +08:00
zhoufeng b25cc0256f support all cuda version in a whl package
Signed-off-by: zhoufeng <zhoufeng54@huawei.com>
2022-11-17 20:30:31 +08:00
i-robot eeb3b4642d
!45421 third-party packages generate more consistent checksums to optimize cache reuse
Merge pull request !45421 from yanghaoran/thirdparty_checksum
2022-11-16 09:34:14 +00:00
yanghaoran f03f70c433 third-party packages generate more consistent checksums to optimize cache reuse 2022-11-15 20:40:29 +08:00
ZPaC b7df5799b2 Delete gpu_collective so for cloud side. 2022-11-13 12:28:36 +08:00
zhoufeng f2adc0109b gpu/cpu use glibcxx_abi=0
Signed-off-by: zhoufeng <zhoufeng54@huawei.com>
2022-11-09 11:41:35 +08:00
i-robot dfec441054
!45029 cmake md5 to sha256
Merge pull request !45029 from 范吉斌/md5tosha256
2022-11-08 02:07:04 +00:00
i-robot b0b30f6447
!45006 [MS][LITE]add dvpp to lite minddata
Merge pull request !45006 from 张学同/new_api
2022-11-04 03:36:00 +00:00
zhangxuetong de1af9a2af add dvpp to lite minddata 2022-11-03 20:00:58 +08:00
fanjibin 650b93eaad cmake md5 to sha256 2022-11-02 19:49:32 +08:00
i-robot af16ffe913
!44723 Make kernel as a plugin
Merge pull request !44723 from JuiceZ/dev_kernel
2022-11-01 01:18:23 +00:00
i-robot 6ba24976b2
!44849 NCCL库增加安全编译选项-DFORTIFY_SOURCE=2 -O2
Merge pull request !44849 from zuochuanyong/nccl_cmake
2022-10-31 06:53:30 +00:00
i-robot 0abd5a56b8
!44502 kernel executor merge from enterprise
Merge pull request !44502 from 王平安/kernel_executor_api
2022-10-31 06:25:43 +00:00
sjtujayyyy 28d6781416 kernel plugin 2022-10-31 11:30:25 +08:00
zuochuanyong 9fb02cb36e add -DFORTIFY_SOURCE=2 2022-10-31 09:05:07 +08:00
i-robot b919f94ac5
!44617 fix lite VS compile error.
Merge pull request !44617 from 王平安/windows
2022-10-27 07:35:53 +00:00
i-robot 891da18e68
!44548 Make converter as a plugin
Merge pull request !44548 from JuiceZ/dev_plugin
2022-10-27 02:48:38 +00:00
wangpingan2 853da3ab71 fix lite vs compile error. 2022-10-26 17:15:14 +08:00
i-robot 04fab5011f
!43958 Add cucollections cmake
Merge pull request !43958 from zyli2020/dynamic_embedding_compile
2022-10-25 14:46:42 +00:00
i-robot 98793e09db
!44524 fix libpython3.so not found on gpu
Merge pull request !44524 from zhoufeng/xiu-ba-ge
2022-10-25 13:22:33 +00:00
sjtujayyyy da4519e726 converter plugin 2022-10-25 19:17:47 +08:00
lizhenyu 0f643987eb Add cucollections cmake 2022-10-25 14:52:19 +08:00
zhoufeng d0b9ba4f26 fix libpython3.so not found on gpu
Signed-off-by: zhoufeng <zhoufeng54@huawei.com>
2022-10-25 12:06:00 +08:00
wangpingan2 20d6596026 sync kernel executor code. 2022-10-25 10:31:13 +08:00
qiuzhongya 4e33db0365 support for msvc debug, add pdb file in package 2022-10-24 21:17:42 +08:00
changzherui 6cbad3f58c fix CVE-2022-1941 2022-10-15 12:28:31 +08:00
i-robot aae892cc5e
!43753 [MS][lite]modify extendrt package
Merge pull request !43753 from 张学同/new_api
2022-10-13 07:57:19 +00:00
i-robot c937113df8
!43765 support compile in msvc, enable download robin from gitee
Merge pull request !43765 from qiuzhongya/qiuzhongya_msvc9
2022-10-13 01:25:21 +00:00
qiuzhongya 1239493ed7 enable robin download from gitee 2022-10-12 20:29:00 +08:00
zhangxuetong c7235c7e7e modify extendrt package 2022-10-12 15:00:34 +08:00
qiuzhongya bd2a545d1b support for windows msvc compile, download dirent from gitee mirrors 2022-10-12 09:58:56 +08:00
i-robot 6bdf223d92
!43431 support fof windows msvc debug mode, postfix "lib" is not necessary for lib name
Merge pull request !43431 from qiuzhongya/msvc_debug2
2022-10-10 12:08:54 +00:00
i-robot 4fcbbfa2e2
!42843 ready for adapt to rocm
Merge pull request !42843 from lsder/master
2022-10-10 06:54:04 +00:00
i-robot 7c1607a678
!43405 lite support helper
Merge pull request !43405 from zhengyuanhua/br3
2022-10-10 03:14:24 +00:00
zhengyuanhua a306f5f8d2 lite support helper 2022-10-09 21:49:05 +08:00
qiuzhongya 57da978728 support for windows msvc debug mode 2022-10-09 19:56:02 +08:00
i-robot 5d2d11e950
!43369 cut down compiler threads num to meet msvc compiling under Windows
Merge pull request !43369 from iambowen1984/master
2022-10-09 06:21:18 +00:00
i-robot b2c39cc90c
!43377 support compile in msvc debug mode
Merge pull request !43377 from qiuzhongya/msvc_debug1
2022-10-09 03:28:39 +00:00
yanghaoran e2df5559cf fix liboptiling.so path as future cann packages are changing it 2022-10-08 20:43:11 +08:00