aa1cee113d
!1279 test(custom_op): clean up custom op test case.
...
Merge pull request !1279 from gongchen/fix_custom_op_case
5 years ago
44458865b6
test(custom_op): Clean up custom op test case.
5 years ago
6ffb3c97cd
!1058 add host cpu kernel: GatherV2/Concat/Slice/SliceGrad
...
Merge pull request !1058 from sunsuodong/lstm_ops
5 years ago
2703ac5ba4
fix bert percision problem
5 years ago
366364ba38
Add custom op testcases.
5 years ago
5a914994ba
clean pylint
5 years ago
28241d0293
lstm ops
5 years ago
bcfaff97f9
clean pylint
5 years ago
981b013f81
Fix CSE bug for some operations like `DropoutGenMask` which should not
...
be optimized as it will generate different values each time.
5 years ago
2907cf4488
remove some context param
5 years ago
1991a89f40
LayerNormGrad fix & codex
5 years ago
bb374ebc28
!1015 Support weights to be compiled according to shape
...
Merge pull request !1015 from candanzg/me_with_shape
5 years ago
32c1d558f4
SummaryRecord as context manager
5 years ago
2cc85bdc93
Support weight compile according to shape
...
Signed-off-by: candanzg <zhangshucheng@huawei.com>
5 years ago
049d3796a6
!1091 gpu support Cast/RealDiv/Mul/Sub/Softmax kernels enforcement
...
Merge pull request !1091 from chenweifeng/cast
5 years ago
6f386c5782
!1082 Support 'break', 'continue' and 'pass' statements
...
Merge pull request !1082 from hewei/support_cont_break
5 years ago
33fa90efc9
Support 'break', 'continue' and 'pass'
...
To handle 'break' and 'continue' statement, a loop context is pushed
to a stack before we parse the loop body, and pop it after body parsed.
When a 'break', 'continue' statement is encountered, we retrieve current
loop contex from the stack, and let the current block jump to the end
block or header block;
For 'break' statement, we added an extra 'end_block' follow the 'after_block',
because 'after_block' is called from a ContionalJump in 'header_block', it can
not be set as jump target from other place. to support 'break', we let loop
body jump to the 'end_block' at the 'break' point. and 'after_block'
maybe a good place to handle loop 'else' clause in the future.
Handle 'pass' is simple, just bypass it when doing parse.
5 years ago
391a060f21
remove two context param
5 years ago
59c4cf256c
gpu support broadcast kernels
5 years ago
d40e89b1bc
clean pylint warning in st/ops/ascend
5 years ago
680ce090a3
!1057 matmul support fp16
...
Merge pull request !1057 from chenweifeng/matmul
5 years ago
00383fc760
!1040 base attention aggregator for Graph Attention Networks
...
Merge pull request !1040 from zhangdengcheng/master
5 years ago
0edc6d254a
!370 Gpu Support UnsortedSegmentSum kernel
...
Merge pull request !370 from chenweifeng/unsorted_segment_sum
5 years ago
907b609b05
!994 gpu broadcast kernel support different dims
...
Merge pull request !994 from chenweifeng/broadcast_unequal_dims
5 years ago
b5096e1f6c
!1021 gpu support MinimumGrad & MaximumGrad kernel
...
Merge pull request !1021 from chenweifeng/broadcast_grad
5 years ago
da7054645a
!948 gpu support LogSoftmax & LogSoftmaxGrad kernel
...
Merge pull request !948 from chenweifeng/logsoftmax
5 years ago
b56572bb89
matmul support fp16
5 years ago
825d9740fa
Fixed the bug that mean aggregator argument can not pass to base class and add attention head for GAT
5 years ago
af7c54b12a
!920 [MS][Bert]Modify example script of bert model.
...
Merge pull request !920 from wsc/lossscale_script
5 years ago
46c175a574
Modify example scripts of bert model
5 years ago
05e8d95e7f
optimize the gpu context switch
5 years ago
00e78bf6c4
gpu support MinimumGrad & MaximumGrad kernel
5 years ago
31f3611f9a
gpu support UnsortedSegmentSum kernel
5 years ago
6146424596
refactor vm module for multigraph sink
5 years ago
0a1195ddf5
broadcast kernel support unqual dims & half
5 years ago
d3936b9f2a
GPU kernels adapt with special dimensions.
5 years ago
1eb60df5d4
gpu support logsoftmax & logsoftmaxgrad kernel
5 years ago
bda4ebd591
!322 Gpu Support RMSProp kernel
...
Merge pull request !322 from chenweifeng/rmsprop
5 years ago
f602970990
!323 Gpu Concat support 4 inputs
...
Merge pull request !323 from chenweifeng/concat
5 years ago
4e25fec769
!324 Gpu Slice kernel performance improve
...
Merge pull request !324 from chenweifeng/slice
5 years ago
378a7122a5
!372 Gpu support BatchMatMul kernel
...
Merge pull request !372 from chenweifeng/batchmatmul
5 years ago
97d21ba014
!502 Gpu Support Gelu & GeluGrad
...
Merge pull request !502 from chenweifeng/gelu
5 years ago
a97f30ba7d
!516 Gpu support Tanh & TanhGrad kernel
...
Merge pull request !516 from chenweifeng/tanh
5 years ago
38c56fd1a5
!945 gpu queue support Sqrt & Rsqrt kernel
...
Merge pull request !945 from chenweifeng/unary
5 years ago
8a484dbd0b
!812 [CR] add lenet train and eval st case
...
Merge pull request !812 from jinyaohui/train_eval
5 years ago
8b49a4cf0e
add lenet train eval gpu st case
5 years ago
a304304c30
gpu support Gelu & GeluGrad kernels
5 years ago
311bf41e6d
gpu support tanh & tanhgrad kernel
5 years ago
67a0cc3bf1
gpu queue support unary
5 years ago
16f0688230
gpu support broadcast kernels
5 years ago
mindspore-ci-bot
gongchen
wanghua
jinyaohui
sunsuodong
seatea
wilfChen
mindspore-ci-bot
李鸿章
candanzg
He Wei
zhangdengcheng
wsc
limingqi107
rick_sanchez
ZPaC