mirror of
https://github.com/ROCm/composable_kernel.git
synced 2026-04-20 06:49:15 +00:00
update layernorm (#1570)
* port layernorm
* change warp_welford.hpp
* Update warpshuffle
* 1. Add save mean and save std back
2. Move construction of tensor_view and tile_window to operator()
* refine welford max count calculation
* unify layernorm api
* Rename file
* Remove save mean and inv std
* Revert "refine welford max count calculation"
This reverts commit 022365802b.
* Fix order of parameter
* refine welford max count calculation again
* Remove fp32 instances
* Fix bug of padding
* refactor api
* Support bf16
* Extract common function
* Refine arg of operator()
* Add kMThreadPerBlock to template parameter
* clang format
* Refine variable name
* Refine file name
* remove redundant line
* refactor layernorm2d pipeline and add block-per-block utility
* fix name
* rename more
* add more block-per-tile instance
* remove duplicated define
* update instance for 2048, 1024 case
* support up to 2048 now
* opt loading
* add n1536
* Add two pass pipeline
* format
* Fix incorrect type
* parallel compilation
* Use smaller N
* fix 2p pass
* Support Repeat_M in distribution
* Refine nameing
* Add reduce example
---------
Co-authored-by: letaoqin <letaoqin@amd.com>
Co-authored-by: aska-0096 <haocwang@amd.com>
Co-authored-by: rocking <ChunYu.Lai@amd.com>
Co-authored-by: carlushuang <carlus.huang@amd.com>
This commit is contained in:
38
example/ck_tile/02_layernorm2d/script/perf_test.sh
Executable file
38
example/ck_tile/02_layernorm2d/script/perf_test.sh
Executable file
@@ -0,0 +1,38 @@
|
||||
|
||||
# run from top of ck folder
|
||||
EXE=build/bin/tile_example_layernorm2d_fwd
|
||||
|
||||
$EXE -m=1 -n=1 -e=1e-12 -v=1 -prec=bf16 -repeat=1000
|
||||
$EXE -m=700 -n=80 -e=1e-12 -v=1 -prec=bf16 -repeat=1000
|
||||
$EXE -m=700 -n=128 -e=1e-12 -v=1 -prec=bf16 -repeat=1000
|
||||
$EXE -m=700 -n=144 -e=1e-12 -v=1 -prec=bf16 -repeat=1000
|
||||
$EXE -m=700 -n=168 -e=1e-12 -v=1 -prec=bf16 -repeat=1000
|
||||
$EXE -m=700 -n=184 -e=1e-12 -v=1 -prec=bf16 -repeat=1000
|
||||
$EXE -m=700 -n=256 -e=1e-12 -v=1 -prec=bf16 -repeat=1000
|
||||
$EXE -m=700 -n=288 -e=1e-12 -v=1 -prec=bf16 -repeat=1000
|
||||
$EXE -m=700 -n=344 -e=1e-12 -v=1 -prec=bf16 -repeat=1000
|
||||
$EXE -m=700 -n=376 -e=1e-12 -v=1 -prec=bf16 -repeat=1000
|
||||
$EXE -m=700 -n=448 -e=1e-12 -v=1 -prec=bf16 -repeat=1000
|
||||
$EXE -m=700 -n=512 -e=1e-12 -v=1 -prec=bf16 -repeat=1000
|
||||
$EXE -m=700 -n=924 -e=1e-12 -v=1 -prec=bf16 -repeat=1000
|
||||
$EXE -m=700 -n=1024 -e=1e-12 -v=1 -prec=bf16 -repeat=1000
|
||||
$EXE -m=700 -n=1078 -e=1e-12 -v=1 -prec=bf16 -repeat=1000
|
||||
$EXE -m=700 -n=1996 -e=1e-12 -v=1 -prec=bf16 -repeat=1000
|
||||
$EXE -m=700 -n=4080 -e=1e-12 -v=1 -prec=bf16 -repeat=1000
|
||||
|
||||
$EXE -m=700 -n=80 -e=1e-12 -v=1 -prec=fp16 -repeat=1000
|
||||
$EXE -m=700 -n=128 -e=1e-12 -v=1 -prec=fp16 -repeat=1000
|
||||
$EXE -m=700 -n=144 -e=1e-12 -v=1 -prec=fp16 -repeat=1000
|
||||
$EXE -m=700 -n=168 -e=1e-12 -v=1 -prec=fp16 -repeat=1000
|
||||
$EXE -m=700 -n=184 -e=1e-12 -v=1 -prec=fp16 -repeat=1000
|
||||
$EXE -m=700 -n=256 -e=1e-12 -v=1 -prec=fp16 -repeat=1000
|
||||
$EXE -m=700 -n=288 -e=1e-12 -v=1 -prec=fp16 -repeat=1000
|
||||
$EXE -m=700 -n=344 -e=1e-12 -v=1 -prec=fp16 -repeat=1000
|
||||
$EXE -m=700 -n=376 -e=1e-12 -v=1 -prec=fp16 -repeat=1000
|
||||
$EXE -m=700 -n=448 -e=1e-12 -v=1 -prec=fp16 -repeat=1000
|
||||
$EXE -m=700 -n=512 -e=1e-12 -v=1 -prec=fp16 -repeat=1000
|
||||
$EXE -m=700 -n=924 -e=1e-12 -v=1 -prec=fp16 -repeat=1000
|
||||
$EXE -m=700 -n=1024 -e=1e-12 -v=1 -prec=fp16 -repeat=1000
|
||||
$EXE -m=700 -n=1078 -e=1e-12 -v=1 -prec=fp16 -repeat=1000
|
||||
$EXE -m=700 -n=1996 -e=1e-12 -v=1 -prec=fp16 -repeat=1000
|
||||
$EXE -m=700 -n=4080 -e=1e-12 -v=1 -prec=fp16 -repeat=1000
|
||||
31
example/ck_tile/02_layernorm2d/script/smoke_test.sh
Executable file
31
example/ck_tile/02_layernorm2d/script/smoke_test.sh
Executable file
@@ -0,0 +1,31 @@
|
||||
#!/bin/sh
|
||||
# call from top of CK folder
|
||||
EXE=./build/bin/tile_example_layernorm2d_fwd
|
||||
|
||||
for pr_i in "fp16" "bf16" ; do
|
||||
$EXE -prec=$pr_i -m=99 -n=13
|
||||
$EXE -prec=$pr_i -m=17 -n=16
|
||||
$EXE -prec=$pr_i -m=1 -n=100
|
||||
$EXE -prec=$pr_i -m=4 -n=128
|
||||
$EXE -prec=$pr_i -m=80 -n=127
|
||||
$EXE -prec=$pr_i -m=22 -n=255 -stride=256
|
||||
$EXE -prec=$pr_i -m=7 -n=599
|
||||
$EXE -prec=$pr_i -m=19 -n=512
|
||||
$EXE -prec=$pr_i -m=33 -n=313 -stride=1000
|
||||
$EXE -prec=$pr_i -m=11 -n=510
|
||||
$EXE -prec=$pr_i -m=171 -n=676 -stride=818
|
||||
$EXE -prec=$pr_i -m=91 -n=636
|
||||
$EXE -prec=$pr_i -m=12 -n=768 -stride=800
|
||||
$EXE -prec=$pr_i -m=100 -n=766 -stride=812
|
||||
$EXE -prec=$pr_i -m=31 -n=1024
|
||||
$EXE -prec=$pr_i -m=64 -n=1000 -stride=1004
|
||||
$EXE -prec=$pr_i -m=8 -n=1501
|
||||
$EXE -prec=$pr_i -m=3 -n=1826
|
||||
$EXE -prec=$pr_i -m=5 -n=2040
|
||||
$EXE -prec=$pr_i -m=7 -n=2734
|
||||
$EXE -prec=$pr_i -m=1 -n=3182
|
||||
$EXE -prec=$pr_i -m=9 -n=4096
|
||||
$EXE -prec=$pr_i -m=3 -n=8192
|
||||
$EXE -prec=$pr_i -m=1 -n=10547
|
||||
$EXE -prec=$pr_i -m=3 -n=17134
|
||||
done
|
||||
Reference in New Issue
Block a user