Skip to content
GitLab
Menu
Projects
Groups
Snippets
Loading...
Help
Help
Support
Community forum
Keyboard shortcuts
?
Submit feedback
Contribute to GitLab
Sign in / Register
Toggle navigation
Menu
Open sidebar
gaoqiong
composable_kernel
Commits
5e858800b35ffb7b10f155b2b59bbda187b5274a
Switch branch/tag
composable_kernel
31 Aug, 2023
1 commit
gridwise dropout
· 5e858800
danyao12
authored
Aug 31, 2023
5e858800
29 Aug, 2023
11 commits
Merge pull request #870 from ROCmSoftwarePlatform/mha-train-bias-bwd-type2
· 4c8b47c0
Dan Yao
authored
Aug 29, 2023
Add bias to flash attention bwd
4c8b47c0
fix d0 load desc names
· 882b3328
letaoqin
authored
Aug 29, 2023
882b3328
fix d0load descriptor name
· c54a1014
letaoqin
authored
Aug 29, 2023
c54a1014
change d0_block_desc_n0_n1_m0_m1_m2_m3 to d0_block_desc_n0_n1_m0_m1_m2
· 703ef6d7
letaoqin
authored
Aug 29, 2023
703ef6d7
batched bwd add check for bias
· 1696ca42
letaoqin
authored
Aug 29, 2023
1696ca42
change biases to bias in batched mha
· 21cec2bb
letaoqin
authored
Aug 29, 2023
21cec2bb
grouped bwd change p_accx_bias to p_accx_bias_vec
· ff6d9e1f
letaoqin
authored
Aug 29, 2023
ff6d9e1f
remove _vec for bwd parameters
· eff268e6
letaoqin
authored
Aug 29, 2023
eff268e6
add comments for grouped host code
· 2464edd0
letaoqin
authored
Aug 29, 2023
2464edd0
add check code
· 28459058
letaoqin
authored
Aug 29, 2023
28459058
bwd biaes to bias
· d10f25a0
letaoqin
authored
Aug 29, 2023
d10f25a0
28 Aug, 2023
4 commits
fix p_acc1_biases size issue
· 127982f1
letaoqin
authored
Aug 28, 2023
127982f1
fix other call because interface change
· af3bec8f
letaoqin
authored
Aug 28, 2023
af3bec8f
v1 group finished
· 67e10a6a
letaoqin
authored
Aug 28, 2023
67e10a6a
v2 group finish
· 8efd67d8
letaoqin
authored
Aug 28, 2023
8efd67d8
25 Aug, 2023
4 commits
add group example
· 72539dbd
letaoqin
authored
Aug 25, 2023
72539dbd
start group
· 7cff2f4d
letaoqin
authored
Aug 25, 2023
7cff2f4d
remove debug code
· f4c4471f
letaoqin
authored
Aug 25, 2023
f4c4471f
v1 gridwise gemm
· e663a0d7
letaoqin
authored
Aug 25, 2023
e663a0d7
24 Aug, 2023
2 commits
v1 device complete
· 701879d0
letaoqin
authored
Aug 24, 2023
701879d0
fix for no bias
· 0539dbcd
letaoqin
authored
Aug 24, 2023
0539dbcd
23 Aug, 2023
7 commits
load D0 data only for 4 xdl
· 48a16339
letaoqin
authored
Aug 23, 2023
48a16339
change D0M name
· ab0d58b2
letaoqin
authored
Aug 23, 2023
ab0d58b2
gridwise gemm add template parameter D0BlockTransferSrcScalarPerVector
· 2220cf9a
letaoqin
authored
Aug 23, 2023
2220cf9a
multiple M block
· 289e1196
letaoqin
authored
Aug 23, 2023
289e1196
vector load
· a33c100d
letaoqin
authored
Aug 23, 2023
a33c100d
Merge pull request #859 from ROCmSoftwarePlatform/mha-train-develop-fix-itoa-issue
· 226355e7
ltqin
authored
Aug 23, 2023
fix to include <numeric>
226355e7
add #include <numeric>
· 224d81c1
ltqin
authored
Aug 23, 2023
224d81c1
22 Aug, 2023
1 commit
single block
· d4256471
letaoqin
authored
Aug 22, 2023
d4256471
19 Aug, 2023
1 commit
kernel add all code, need debug
· 763e26be
letaoqin
authored
Aug 19, 2023
763e26be
17 Aug, 2023
6 commits
load d0 to lds
· de53e421
letaoqin
authored
Aug 17, 2023
de53e421
Merge branch 'mha-train-develop' into mha-train-bias-bwd-type2
· ec2ad713
letaoqin
authored
Aug 17, 2023
ec2ad713
add d0_block_copy_global_to_lds
· e3eb4381
letaoqin
authored
Aug 17, 2023
e3eb4381
format
· 77df3ccb
letaoqin
authored
Aug 17, 2023
77df3ccb
add code to device
· 48f98948
letaoqin
authored
Aug 17, 2023
48f98948
fix z pointers empty issue
· e296ee56
ltqin
authored
Aug 17, 2023
e296ee56
16 Aug, 2023
3 commits
add code to device
· 79cf90f2
letaoqin
authored
Aug 16, 2023
79cf90f2
fix group d data type
· c9915508
letaoqin
authored
Aug 16, 2023
c9915508
bias data type convert
· 98df59c6
letaoqin
authored
Aug 16, 2023
98df59c6