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_ROCM
Commits
7796fc738b315d276318be2bef409ee270874610
Switch branch/tag
composable_kernel_rocm
15 Feb, 2025
2 commits
fix gemm2 scale, gemm2 ok now
· 7796fc73
coderfeli
authored
Feb 15, 2025
7796fc73
fix moe gemm2
· 61e3c238
coderfeli
authored
Feb 15, 2025
61e3c238
14 Feb, 2025
7 commits
hotfix:gemm1 use real tokens and gemm2 ok
· db53dba4
coderfeli
authored
Feb 14, 2025
db53dba4
fix topk id
· 58db931e
coderfeli
authored
Feb 14, 2025
58db931e
merge max_token_id and fix err
· 84b27d75
coderfeli
authored
Feb 14, 2025
84b27d75
fix ref gemm no padding
· b3ae04f8
coderfeli
authored
Feb 14, 2025
b3ae04f8
add max_token_id
· 83be79ba
coderfeli
authored
Feb 14, 2025
83be79ba
add logics and debug
· 1078d229
coderfeli
authored
Feb 14, 2025
1078d229
add codes for a scatter
· d4b8f1e3
coderfeli
authored
Feb 14, 2025
d4b8f1e3
13 Feb, 2025
1 commit
change cshuffle cluster, mi300x reach roofline
· 82e1f1b9
coderfeli
authored
Feb 13, 2025
82e1f1b9
12 Feb, 2025
3 commits
fix mtile 64,128 for gemm1
· 568ad1e1
coderfeli
authored
Feb 12, 2025
568ad1e1
remove d2 for gemm1
· 59f3e009
coderfeli
authored
Feb 12, 2025
59f3e009
moe gemm1 scaleready
· 418baed3
coderfeli
authored
Feb 12, 2025
418baed3
11 Feb, 2025
4 commits
gemm1 scale debug
· b02c0b82
coderfeli
authored
Feb 11, 2025
b02c0b82
moe gemm2 scales ok
· e4ca61f9
coderfeli
authored
Feb 11, 2025
e4ca61f9
impl topk weight scatter
· 66d08ea3
coderfeli
authored
Feb 11, 2025
66d08ea3
fix warnings and impl scale for gemm2, build ok
· a8a82e0c
coderfeli
authored
Feb 11, 2025
a8a82e0c
10 Feb, 2025
6 commits
impl 3ds epilog ok
· 69f54ee8
coderfeli
authored
Feb 10, 2025
69f54ee8
merge gemm1 gemm2 together and run ok
· 72752420
coderfeli
authored
Feb 10, 2025
72752420
merge gemm1 and gemm2
· 66cff910
coderfeli
authored
Feb 10, 2025
66cff910
add moegemm in device and grid
· aa15c49a
coderfeli
authored
Feb 10, 2025
aa15c49a
skip empty expert
· 2e53f972
coderfeli
authored
Feb 10, 2025
2e53f972
add skip expert
· fcf6106b
coderfeli
authored
Feb 10, 2025
fcf6106b
09 Feb, 2025
2 commits
moegemm2 ok
· e21f36fc
coderfeli
authored
Feb 09, 2025
e21f36fc
gemm2 result ok
· 12301455
coderfeli
authored
Feb 09, 2025
12301455
08 Feb, 2025
3 commits
one tile ok
· 7ba5bff4
coderfeli
authored
Feb 08, 2025
7ba5bff4
add files , build and run ok
· 8a5bb9f3
coderfeli
authored
Feb 08, 2025
8a5bb9f3
add empty expert jump
· bd64a30b
coderfeli
authored
Feb 08, 2025
bd64a30b
07 Feb, 2025
8 commits
fp8 ok
· 6b71f3d8
coderfeli
authored
Feb 07, 2025
6b71f3d8
tileM 32,64,128 ok
· 7822382b
coderfeli
authored
Feb 07, 2025
7822382b
tile m = 64 ok
· e15351ca
coderfeli
authored
Feb 07, 2025
e15351ca
a 16x16 ok
· 48d87d9c
coderfeli
authored
Feb 07, 2025
48d87d9c
add ret logit for empty expert
· 24734db8
coderfeli
authored
Feb 07, 2025
24734db8
debug 16x16 load
· 965c9f0c
coderfeli
authored
Feb 07, 2025
965c9f0c
fix hack in oob
· 83970cbe
coderfeli
authored
Feb 07, 2025
83970cbe
use offsets in transfer ok
· f9abcf80
coderfeli
authored
Feb 07, 2025
f9abcf80
05 Feb, 2025
1 commit
save outputs
· e947d11e
coderfeli
authored
Feb 05, 2025
e947d11e
04 Feb, 2025
3 commits
perf ok
· 9afc4a0b
coderfeli
authored
Feb 04, 2025
9afc4a0b
add others
· f8d15f2a
coderfeli
authored
Feb 04, 2025
f8d15f2a
results ok
· 00627fed
coderfeli
authored
Feb 04, 2025
00627fed