Toggle navigation
Toggle navigation
This project
Loading...
Sign in
2020-2-capstone-design1
/
HCS_project
Go to a project
Toggle navigation
Toggle navigation pinning
Projects
Groups
Snippets
Help
Project
Activity
Repository
Pipelines
Graphs
Issues
0
Merge Requests
0
Wiki
Snippets
Network
Create a new issue
Builds
Commits
Issue Boards
Authored by
김지훈
2020-11-25 19:28:57 +0900
Browse Files
Options
Browse Files
Download
Email Patches
Plain Diff
Commit
5b993ef820acf4c6aa335036fdbaacc0e1e8f226
5b993ef8
1 parent
e568103d
fix readme tabel 안보임
Hide whitespace changes
Inline
Side-by-side
Showing
1 changed file
with
29 additions
and
6 deletions
코드/연합학습/quantization/README.md
코드/연합학습/quantization/README.md
View file @
5b993ef
...
...
@@ -6,17 +6,40 @@
-
양자화 과정으로 학습된 모델은 pytorch model -> onnx -> tensorRT 변환이 불가능하여 gpu 에서 실행 불가능.
## TensorRT
-
Google Colab - install_tensorRT
-
양자화 학습을 사용하지 않고, 라이브러리를 활용하여 모델의 정밀도 감소 및 양자화 시도.
-
모델에 따라 속도 차이가 크고 아래 단계의 정밀도가 더 빠른 경우가 있었음
-
정확한 이해가 필요해 보임 (사용법 미숙, 입력 값은 float 등)
| Inference Time(msec) | Densenet - 32 packet | Ours - 1 packet |
| -------------------- | -------------------- | --------------- |
| Torch - float32 | 19.49 | 0.49 |
| TensorRT - float32 | 4.30 | 0.37 |
| TensorRT - float16 | 4.32 | 0.35 |
| TensorRT - int8 | 3.70 | 0.41 |
<table>
<tr
align=
"center"
>
<td
>
Inference Time(msec)
</td>
<td
>
Densenet - 32 packet
</td>
<td
>
Ours - 1 packet
</td>
</tr>
<tr
align=
"center"
>
<td
>
Torch - float32
</td>
<td
>
19.49
</td>
<td
>
0.49
</td>
</tr>
<tr
align=
"center"
>
<td
>
TensorRT - float32
</td>
<td
>
4.30
</td>
<td
>
0.37
</td>
</tr>
<tr
align=
"center"
>
<td
>
TensorRT - float16
</td>
<td
>
4.32
</td>
<td
>
0.35
</td>
</tr>
<tr
align=
"center"
>
<td
>
TensorRT - int8
</td>
<td
>
3.70
</td>
<td
>
0.41
</td>
</tr>
</table>
\ No newline at end of file
...
...
Please
register
or
login
to post a comment