Ck tile/layernorm: implement naive reduce, opt performance (#1784)

* add no welford

* enable output raw

* raw of int8

* fix build

* fix smoke test err

* [ck_tile]layernorm: fix welford ok, set int8 and bf16 small N as default and others open by generate

* [cktile]layernorm, fix err commit files and remove uselss

* fix quant 8192 err & change norm_reduce class and file name

---------

Co-authored-by: coderfeli <coderfeli@163.com>
Co-authored-by: carlushuang <carlus.huang@amd.com>
This commit is contained in:
feli
2025-01-03 14:28:59 +08:00
committed by GitHub
parent 17e8efb573
commit 4bc610416a
10 changed files with 250 additions and 167 deletions

View File

@@ -27,7 +27,8 @@ $EXE -prec_i=$pr_i -fadd=$fadd $fquant -m=7 -n=2734
$EXE -prec_i=$pr_i -fadd=$fadd $fquant -m=1 -n=3182
$EXE -prec_i=$pr_i -fadd=$fadd $fquant -m=9 -n=4096
$EXE -prec_i=$pr_i -fadd=$fadd $fquant -m=3 -n=8192
#$EXE -prec_i=$pr_i -fadd=$fadd $fquant -m=1 -n=10547
$EXE -prec_i=$pr_i -fadd=$fadd $fquant -m=3 -n=9120
$EXE -prec_i=$pr_i -fadd=$fadd $fquant -m=1 -n=10547
#$EXE -prec_i=$pr_i -fadd=$fadd $fquant -m=3 -n=17134
done
done