Skip to content
体验新版
项目
组织
正在加载...
登录
切换导航
打开侧边栏
PaddlePaddle
Paddle-Lite
提交
40764c5e
P
Paddle-Lite
项目概览
PaddlePaddle
/
Paddle-Lite
通知
331
Star
4
Fork
1
代码
文件
提交
分支
Tags
贡献者
分支图
Diff
Issue
271
列表
看板
标记
里程碑
合并请求
78
Wiki
0
Wiki
分析
仓库
DevOps
项目成员
Pages
P
Paddle-Lite
项目概览
项目概览
详情
发布
仓库
仓库
文件
提交
分支
标签
贡献者
分支图
比较
Issue
271
Issue
271
列表
看板
标记
里程碑
合并请求
78
合并请求
78
Pages
分析
分析
仓库分析
DevOps
Wiki
0
Wiki
成员
成员
收起侧边栏
关闭侧边栏
动态
分支图
创建新Issue
提交
Issue看板
体验新版 GitCode,发现更多精彩内容 >>
未验证
提交
40764c5e
编写于
2月 28, 2020
作者:
C
cc
提交者:
GitHub
2月 28, 2020
浏览文件
操作
浏览文件
下载
电子邮件补丁
差异文件
Move all docs from release2.3 to develop, test=develop, test=document_fix (#3037)
上级
0775140a
变更
3
显示空白变更内容
内联
并排
Showing
3 changed file
with
39 addition
and
73 deletion
+39
-73
docs/advanced_user_guides/support_operation_list.md
docs/advanced_user_guides/support_operation_list.md
+1
-12
docs/benchmark/benchmark.md
docs/benchmark/benchmark.md
+31
-59
docs/introduction/roadmap.md
docs/introduction/roadmap.md
+7
-2
未找到文件。
docs/advanced_user_guides/support_operation_list.md
浏览文件 @
40764c5e
...
...
@@ -110,6 +110,7 @@
-
equal
-
gather
-
generate_proposals
-
graph_op
-
greater_equal
-
greater_than
-
gru
...
...
@@ -153,7 +154,6 @@
-
sequence_expand
-
sequence_expand_as
-
sequence_pool
-
sequence_pool_concat
-
sequence_reshape
-
sequence_reverse
-
sequence_softmax
...
...
@@ -226,7 +226,6 @@
-
generate_proposals
-
greater_equal
-
greater_than
-
grid_sampler
-
gru
-
gru_unit
-
hard_sigmoid
...
...
@@ -312,9 +311,6 @@
-
gelu
-
gru
-
layer_norm
-
leaky_relu
-
lookup_table
-
lookup_table_v2
-
match_matrix_tensor
-
matmul
-
mul
...
...
@@ -392,11 +388,9 @@
-
yolo_box
### OpenCL kernels
-
concat
-
conv2d
-
depthwise_conv2d
-
elementwise_add
-
elementwise_mul
-
fc
-
fusion_elementwise_add_activation
-
layout
...
...
@@ -404,10 +398,5 @@
-
io_copy
-
io_copy_once
-
mul
-
nearest_interp
-
pool2d
-
relu
-
reshape
-
reshape2
-
scale
-
sigmoid
docs/benchmark/benchmark.md
浏览文件 @
40764c5e
...
...
@@ -15,14 +15,12 @@
*
int8模型
*
mobilenet_v1
*
mobilenet_v2
*
resnet50
*
测试机器(android ndk ndk-r17c)
*
骁龙855
*
xiaomi mi9, snapdragon 855
*
4xA76(1@2.84GHz + 3@2.4GHz) + 4xA55@1.78GHz
*
骁龙845
*
xiaomi mi8, 845
*
2.8GHz(大四核),1.7GHz(小四核)
...
...
@@ -31,19 +29,11 @@
*
xiaomi mix2, snapdragon 835
*
2.45GHz(大四核),1.9GHz(小四核)
*
骁龙625
*
oppo R9s, snapdragon625
*
A53 x 8, big core@2.0GHz
*
骁龙653
*
360 N5, snapdragon 653
*
4 x A73@2.0GHz + 4 x A53@1.4GHz
*
麒麟970
*
HUAWEI Mate10
*
测试说明
*
branch: release/
2.0
.0
*
branch: release/
v2.3
.0
*
warmup=10, repeats=30,统计平均时间,单位是ms
*
当线程数为1时,
```DeviceInfo::Global().SetRunMode```
设置LITE_POWER_HIGH,否者设置LITE_POWER_NO_BIND
*
模型的输入图像的维度是{1, 3, 224, 224},输入图像的每一位数值是1
...
...
@@ -55,78 +45,59 @@
#### paddlepaddle model
骁龙855|armv7 | armv7 | armv7 |armv8 | armv8 |armv8
----| ---- | ---- | ---- | ---- |---- |----
threads num|1 |2 |4 |1 |2 |4
mobilenet_v1 |32.19 |18.81 |10.90 |30.92 |18.31 |10.15
mobilenet_v2 |22.91 |13.75 |8.64 |21.15 |12.79 |7.84
shufflenet_v2 |4.67 |3.37 |2.65 |4.43 |3.15 |2.66
squeezenet_v1.1 |25.10 |15.93 |9.68 |23.28 |14.61 |8.71
mnasnet |21.84 |13.14 |7.96 |19.61 |11.88 |7.55
mobilenet_v1 |33.27 |19.52 |11.14 |31.72 |18.76 |10.24 |
mobilenet_v2 |29.08 |15.79 |9.25 |25.89 |14.17 |8.38 |
shufflenet_v2 |4.40 |3.09 |2.30 |4.28 |3.02 |2.35 |
squeezenet_v1.1 |19.96 |12.61 |8.76 |18.25 |11.46 |7.97 |
mnasnet |21.00 |12.54 |7.28 |19.65 |11.65 |6.96 |
骁龙8
3
5|armv7 | armv7 | armv7 |armv8 | armv8 |armv8
骁龙8
4
5|armv7 | armv7 | armv7 |armv8 | armv8 |armv8
----| ---- | ---- | ---- | ---- |---- |----
threads num|1 |2 |4 |1 |2 |4
mobilenet_v1 |
94.13 |52.17 |30.68 |88.28 |47.58 |26.64
mobilenet_v2 |
61.24 |34.64 |22.36 |56.66 |32.19 |19.63
shufflenet_v2 |
10.87 |6.92 |5.12 |10.41 |6.76 |4.97
squeezenet_v1.1 |
73.61 |42.25 |24.44 |64.87 |38.43 |23.06
mnasnet |
58.22 |33.43 |20.44 |53.43 |30.20 |18.09
mobilenet_v1 |
66.36 |35.97 |19.45 |62.66 |33.87 |17.85 |
mobilenet_v2 |
45.86 |25.53 |14.6 |41.58 |23.24 |13.39 |
shufflenet_v2 |
7.58 |4.89 |3.41 |7.44 |4.91 |3.58 |
squeezenet_v1.1 |
37.15 |22.74 |13.51 |34.69 |21.27 |12.74 |
mnasnet |
40.09 |21.73 |11.91 |38.19 |21.02 |12.11 |
麒麟980|armv7 | armv7 | armv7 |armv8 | armv8 |armv8
----| ---- | ---- | ---- | ---- |---- |----
threads num|1 |2 |4 |1 |2 |4
mobilenet_v1 |55.11 |28.24 |13.27 |34.24 |17.74 |12.41
mobilenet_v2 |37.03 |19.80 |51.94 |23.64 |12.98 |9.38
shufflenet_v2 |7.26 |4.94 |15.06 |5.32 |3.33 |2.82
squeezenet_v1.1 |42.73 |23.66 |57.39 |26.03 |14.53 |13.66
mnasnet |36.87 |20.15 |46.04 |21.85 |12.06 |8.68
麒麟970|armv7 | armv7 | armv7 |armv8 | armv8 |armv8
骁龙835|armv7 | armv7 | armv7 |armv8 | armv8 |armv8
----| ---- | ---- | ---- | ---- |---- |----
threads num|1 |2 |4 |1 |2 |4
mobilenet_v1 |9
7.80 |52.64 |34.46 |94.51 |49.36 |28.43
mobilenet_v2 |6
6.55 |38.52 |23.19 |62.89 |34.93 |21.53
shufflenet_v2 |1
3.78 |8.11 |5.93 |11.95 |7.90 |5.91
squeezenet_v1.1 |
77.64 |43.67 |25.72 |69.91 |40.66 |24.62
mnasnet |
61.86 |34.62 |22.68 |59.61 |32.79 |19.56
mobilenet_v1 |9
6.98 |53.92 |32.24 |89.31 |48.02 |27.58 |
mobilenet_v2 |6
7.72 |37.66 |23.82 |60.10 |34.36 |21.05 |
shufflenet_v2 |1
0.72 |6.62 |4.63 |10.10 |6.44 |4.63 |
squeezenet_v1.1 |
53.89 |33.28 |20.73 |50.83 |32.31 |19.51 |
mnasnet |
59.55 |33.53 |20.32 |56.21 |31.58 |19.06 |
#### caffe model
骁龙855|armv7 | armv7 | armv7 |armv8 | armv8 |armv8
----| ---- | ---- | ---- | ---- |---- |----
threads num|1 |2 |4 |1 |2 |4 |
mobilenet_v1 |3
2.42 |18.68 |10.86 |30.92 |18.35 |10.07
|
mobilenet_v2 |
29.53 |17.76 |10.89 |27.19 |16.53 |9.75
|
shufflenet_v2 |4.
61 |3.29 |2.61 |4.36 |3.11 |2.51
|
mobilenet_v1 |3
3.36 |19.45 |11.26 |31.63 |18.74 |10.31
|
mobilenet_v2 |
31.63 |19.21 |11.61 |28.34 |17.14 |10.16
|
shufflenet_v2 |4.
46 |3.08 |2.32 |4.26 |2.98 |2.35
|
骁龙8
3
5|armv7 | armv7 | armv7 |armv8 | armv8 |armv8
骁龙8
4
5|armv7 | armv7 | armv7 |armv8 | armv8 |armv8
----| ---- | ---- | ---- | ---- |---- |----
threads num|1 |2 |4 |1 |2 |4 |
mobilenet_v1 |92.52 |52.34 |30.37 |88.31 |49.75 |27.29 |
mobilenet_v2 |79.50 |45.67 |28.79 |76.13 |44.01 |26.13 |
shufflenet_v2 |10.94 |7.08 |5.16 |10.64 |6.83 |5.01 |
mobilenet_v1 |66.32 |35.83 |19.56 |62.52 |33.79 |17.91 |
mobilenet_v2 |58.46 |32.69 |18.56 |53.72 |29.86 |16.80 |
shufflenet_v2 |7.65 |4.82 |3.46 |7.55 |4.97 |3.62 |
麒麟980|armv7 | armv7 | armv7 |armv8 | armv8 |armv8
----| ---- | ---- | ---- | ---- |---- |----
threads num|1 |2 |4 |1 |2 |4 |
mobilenet_v1 |55.36 |28.18 |13.31 |34.42 |17.93 |12.52 |
mobilenet_v2 |49.17 |26.10 |65.49 |30.50 |16.66 |11.72 |
shufflenet_v2 |8.45 |5.00 |15.65 |4.58 |3.14 |2.83 |
麒麟970|armv7 | armv7 | armv7 |armv8 | armv8 |armv8
骁龙835|armv7 | armv7 | armv7 |armv8 | armv8 |armv8
----| ---- | ---- | ---- | ---- |---- |----
threads num|1 |2 |4 |1 |2 |4 |
mobilenet_v1 |9
7.85 |53.38 |33.85 |94.29 |49.42 |28.29
|
mobilenet_v2 |8
7.40 |50.25 |31.85 |85.55 |48.11 |28.24
|
shufflenet_v2 |1
2.16 |8.39 |6.21 |12.21 |8.33 |6.32
|
mobilenet_v1 |9
5.38 |54.09 |32.03 |95.05 |48.33 |27.54
|
mobilenet_v2 |8
8.46 |48.98 |30.23 |79.28 |44.64 |27.10
|
shufflenet_v2 |1
0.07 |6.51 |4.61 |10.31 |6.50 |4.66
|
#### int8量化模型测试数据
...
...
@@ -136,6 +107,7 @@ threads num|1 |2 |4 |1 |2 |4 |
mobilenet_v1 |36.80 |21.58 |11.12 | 14.01 |8.13 |4.32 |
mobilenet_v2 |28.72 |19.08 |12.49 | 17.24 |11.55 |7.82 |
骁龙835|armv7 | armv7 | armv7 |armv8 | armv8 |armv8
----| ---- | ---- | ---- | ---- |---- |----
threads num|1 |2 |4 |1 |2 |4 |
...
...
docs/introduction/roadmap.md
浏览文件 @
40764c5e
...
...
@@ -23,5 +23,10 @@
-
增加 INT8 量化,从 PaddleSlim 训练到 PaddleLite 部署完整案例
-
支持内存中加载模型,以支持 APP 的简易加密
## 2.0.0-rc ?
预计发布
*2019-9-16~7days*
## 2.3
[
v2.3 project
](
https://github.com/PaddlePaddle/Paddle-Lite/milestone/3?closed=1
)
## 2.6
[
v2.6 project
](
https://github.com/PaddlePaddle/Paddle-Lite/milestones/v2.6
)
编辑
预览
Markdown
is supported
0%
请重试
或
添加新附件
.
添加附件
取消
You are about to add
0
people
to the discussion. Proceed with caution.
先完成此消息的编辑!
取消
想要评论请
注册
或
登录