[Layer] Add Self-Attention, RMSNorm benchmark #140
Conversation
|
|
||
| bool areArraysEqual(float array1[], float array2[], int size) { | ||
| for (int i = 0; i < size; ++i) { | ||
| if (array1[i] != array2[i]) { |
There was a problem hiding this comment.
The result of this benchmark is "Pass," indicating that the calculation result is correct. However, directly using the equality operator (==) to compare floating-point numbers may not be the best practice. This is because floating-point calculations can introduce tiny precision errors.
To more reliably compare floating-point numbers, it's advised to use a small error margin, for instance: MobileNetV3 example
|
|
||
| bool areArraysEqual(float array1[], float array2[], int size) { | ||
| for (int i = 0; i < size; ++i) { | ||
| if (array1[i] != array2[i]) { |
There was a problem hiding this comment.
This place is the same as above.
SamanthaWangdl
force-pushed
the
main
branch
from
46e69c1
to
c15514a
Compare
|
@ShiHaoGao The corresponding issue has been fixed. |
| @@ -0,0 +1,149 @@ | |||
| //===- GoogleBenchmarkMain.cpp---------------------------------------------===// | |||
There was a problem hiding this comment.
Recommand to centralize this title and keep the number of characters per line to 80.
| @@ -0,0 +1,147 @@ | |||
| //===- GoogleBenchmarkMain.cpp---------------------------------------------===// | |||
There was a problem hiding this comment.
Recommand to centralize this title and keep the number of characters per line to 80.
There was a problem hiding this comment.
In this version of the code standard, the title remains not centered, otherwise it will be too troublesome to check each file and centralize the tiles. Please help me change it back. Thanks~
benchmarks/DeepLearning/Layers/SelfAttention/buddy_selfattention_import.py
Show resolved
Hide resolved
|
All errors have been fixed. Ready to merge. |
|
Thanks! Just resolve the conflicts and make changes according to the code standards. |
Add