Skip to content
Projects
Groups
Snippets
Help
This project
Loading...
Sign in / Register
Toggle navigation
M
Model-Transfer-Adaptability
Overview
Overview
Details
Activity
Cycle Analytics
Repository
Repository
Files
Commits
Branches
Tags
Contributors
Graph
Compare
Charts
Issues
0
Issues
0
List
Board
Labels
Milestones
Merge Requests
0
Merge Requests
0
CI / CD
CI / CD
Pipelines
Jobs
Schedules
Charts
Wiki
Wiki
Snippets
Snippets
Members
Members
Collapse sidebar
Close sidebar
Activity
Graph
Charts
Create a new issue
Jobs
Commits
Issue Boards
Open sidebar
haoyifan
Model-Transfer-Adaptability
Commits
1f5b02d5
Commit
1f5b02d5
authored
Apr 16, 2023
by
Klin
Browse files
Options
Browse Files
Download
Email Patches
Plain Diff
doc: modify markdown
parent
cea929c5
Hide whitespace changes
Inline
Side-by-side
Showing
1 changed file
with
2 additions
and
1 deletions
+2
-1
ykl/README.md
+2
-1
No files found.
ykl/README.md
View file @
1f5b02d5
...
...
@@ -6,5 +6,5 @@
+
ptq.py中计算js_param笔误,应由flop_ratio改为par_ratio。否则flops和param拟合没有区别
+
module.py中bias_qmax方法,应当为float类型传参num_bits为16,e_bits为7.
+
这里主要关注e_bits,拟合离群点主要为FLOAT_7_E5 / FLOAT_8_E5 / FLOAT_8_E6,其表现为bias两极分布,与之前int量化bias溢出的问题现象相似。
+
原先指定e_bits为5,由于bias的scale为input和weight的scale乘积,bias量化范围应当大致为x和weight量化范围的平方倍。目前代码支持的最高x和weight量化范围大致为
$2^{2^6}$,因此bias范围应当近似取到$2^{2^7}$
,即将e_bits指定为7
+
原先指定e_bits为5,由于bias的scale为input和weight的scale乘积,bias量化范围应当大致为x和weight量化范围的平方倍。目前代码支持的最高x和weight量化范围大致为
2的2的6次方 ,因此bias范围应当近似取到2的2的7次方
,即将e_bits指定为7
+
改动之后,离群点消失,拟合效果显著提高
\ No newline at end of file
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