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
OpenDAS
apex
Commits
77f9d73c
Unverified
Commit
77f9d73c
authored
Apr 29, 2022
by
yjk21
Committed by
GitHub
Apr 29, 2022
Browse files
[FastLayerNorm] Support hidden dim of 14336 (#1368)
parent
f9305e75
Changes
3
Hide whitespace changes
Inline
Side-by-side
Showing
3 changed files
with
13 additions
and
0 deletions
+13
-0
apex/contrib/csrc/layer_norm/ln_bwd_semi_cuda_kernel.cu
apex/contrib/csrc/layer_norm/ln_bwd_semi_cuda_kernel.cu
+6
-0
apex/contrib/csrc/layer_norm/ln_fwd_cuda_kernel.cu
apex/contrib/csrc/layer_norm/ln_fwd_cuda_kernel.cu
+6
-0
apex/contrib/test/layer_norm/test_fast_layer_norm.py
apex/contrib/test/layer_norm/test_fast_layer_norm.py
+1
-0
No files found.
apex/contrib/csrc/layer_norm/ln_bwd_semi_cuda_kernel.cu
View file @
77f9d73c
...
...
@@ -166,6 +166,12 @@ REGISTER_BWD_LAUNCHER(12800, fp16, fp32, fp16, fp32, 5, 1, 4, 16, 4);
REGISTER_BWD_LAUNCHER
(
12800
,
bf16
,
bf16
,
bf16
,
fp32
,
5
,
1
,
4
,
8
,
4
);
REGISTER_BWD_LAUNCHER
(
12800
,
bf16
,
fp32
,
bf16
,
fp32
,
5
,
1
,
4
,
16
,
4
);
REGISTER_BWD_LAUNCHER
(
14336
,
fp32
,
fp32
,
fp32
,
fp32
,
4
,
1
,
4
,
8
,
4
);
REGISTER_BWD_LAUNCHER
(
14336
,
fp16
,
fp16
,
fp16
,
fp32
,
4
,
1
,
4
,
8
,
4
);
REGISTER_BWD_LAUNCHER
(
14336
,
fp16
,
fp32
,
fp16
,
fp32
,
4
,
1
,
4
,
8
,
4
);
REGISTER_BWD_LAUNCHER
(
14336
,
bf16
,
bf16
,
bf16
,
fp32
,
4
,
1
,
4
,
8
,
4
);
REGISTER_BWD_LAUNCHER
(
14336
,
bf16
,
fp32
,
bf16
,
fp32
,
4
,
1
,
4
,
8
,
4
);
REGISTER_BWD_LAUNCHER
(
15360
,
fp32
,
fp32
,
fp32
,
fp32
,
4
,
1
,
4
,
8
,
4
);
REGISTER_BWD_LAUNCHER
(
15360
,
fp16
,
fp16
,
fp16
,
fp32
,
4
,
1
,
4
,
4
,
4
);
REGISTER_BWD_LAUNCHER
(
15360
,
fp16
,
fp32
,
fp16
,
fp32
,
4
,
1
,
4
,
8
,
4
);
...
...
apex/contrib/csrc/layer_norm/ln_fwd_cuda_kernel.cu
View file @
77f9d73c
...
...
@@ -154,6 +154,12 @@ REGISTER_FWD_LAUNCHER(12800, fp16, fp32, fp16, fp32, 2, 1, 4, 4);
REGISTER_FWD_LAUNCHER
(
12800
,
bf16
,
bf16
,
bf16
,
fp32
,
2
,
1
,
4
,
4
);
REGISTER_FWD_LAUNCHER
(
12800
,
bf16
,
fp32
,
bf16
,
fp32
,
2
,
1
,
4
,
4
);
REGISTER_FWD_LAUNCHER
(
14336
,
fp32
,
fp32
,
fp32
,
fp32
,
2
,
1
,
4
,
16
);
REGISTER_FWD_LAUNCHER
(
14336
,
fp16
,
fp16
,
fp16
,
fp32
,
2
,
1
,
4
,
16
);
REGISTER_FWD_LAUNCHER
(
14336
,
fp16
,
fp32
,
fp16
,
fp32
,
2
,
1
,
4
,
16
);
REGISTER_FWD_LAUNCHER
(
14336
,
bf16
,
bf16
,
bf16
,
fp32
,
2
,
1
,
4
,
16
);
REGISTER_FWD_LAUNCHER
(
14336
,
bf16
,
fp32
,
bf16
,
fp32
,
2
,
1
,
4
,
8
);
REGISTER_FWD_LAUNCHER
(
15360
,
fp32
,
fp32
,
fp32
,
fp32
,
2
,
1
,
4
,
8
);
REGISTER_FWD_LAUNCHER
(
15360
,
fp16
,
fp16
,
fp16
,
fp32
,
2
,
1
,
4
,
8
);
REGISTER_FWD_LAUNCHER
(
15360
,
fp16
,
fp32
,
fp16
,
fp32
,
2
,
1
,
4
,
8
);
...
...
apex/contrib/test/layer_norm/test_fast_layer_norm.py
View file @
77f9d73c
...
...
@@ -216,6 +216,7 @@ class TestFastLayerNorm(unittest.TestCase):
10240
,
12288
,
12800
,
14336
,
15360
,
16384
,
18432
,
...
...
Write
Preview
Markdown
is supported
0%
Try again
or
attach a new file
.
Attach a file
Cancel
You are about to add
0
people
to the discussion. Proceed with caution.
Finish editing this message first!
Cancel
Please
register
or
sign in
to comment