To support int8 model deployment on AIoT devices, we provide some universal post training quantization tools which can convert the Float32 tmfile model to Int8/UInt8 tmfile model.
sudo apt install libopencv-dev
git clone https://github.com/OAID/Tengine.git tengine-lite
cd tengine-lite
mkdir build
cd build
cmake -DTENGINE_BUILD_QUANT_TOOL=ON ..
make && make install
Those quantization tools should be in ./install/bin/ directory
$ tree install/bin/
install/bin/
├── quant_tool_int8
├── quant_tool_uint8
├── ......
| Type | Note |
|---|---|
| Adaptive | TENGINE_MODE_INT8 |
| Activation data | Int8 |
| Weight date | Int8 |
| Bias date | Int32 |
| Example | tm_classification_int8.c |
| Execution environment | Ubuntu 18.04 |
$ ./quant_tool_int8 -h
---- Tengine Post Training Quantization Tool ----
Version : v1.2, 15:20:21 Jul 25 2021
Status : int8, per-channel, symmetric
[Quant Tools Info]: The input file of Float32 tmfile file not specified!
[Quant Tools Info]: optional arguments:
-h help show this help message and exit
-m input model path to input float32 tmfile
-i image dir path to calibration images folder
-f scale file path to calibration scale file
-o output model path to output int8 tmfile
-a algorithm the type of quant algorithm(0:min-max, 1:kl, 2:aciq, default is 0)
-g size the size of input image(using the resize the original image,default is 3,224,224)
-w mean value of mean (mean value, default is 104.0,117.0,123.0)
-s scale value of normalize (scale value, default is 1.0,1.0,1.0)
-b swapRB flag which indicates that swap first and last channels in 3-channel image is necessary(0:OFF, 1:ON, default is 1)
-c center crop flag which indicates that center crop process image is necessary(0:OFF, 1:ON, default is 0)
-y letter box the size of letter box process image is necessary([rows, cols], default is [0, 0])
-k focus flag which indicates that focus process image is necessary(maybe using for YOLOv5, 0:OFF, 1:ON, default is 0)
-t num thread count of processing threads(default is 1)
[Quant Tools Info]: example arguments:
./quant_tool_int8 -m ./mobilenet_fp32.tmfile -i ./dataset -o ./mobilenet_int8.tmfile -g 3,224,224 -w 104.007,116.669,122.679 -s 0.017,0.017,0.017
Before use the quant tool, you need Float32 tmfile and Calibration Dataset, the image num of calibration dataset we suggest to use 500-1000.
$ .quant_tool_int8 -m ./mobilenet_fp32.tmfile -i ./dataset -o ./mobilenet_int8.tmfile -g 3,224,224 -w 104.007,116.669,122.679 -s 0.017,0.017,0.017 -z 1
---- Tengine Post Training Quantization Tool ----
Version : v1.1, 15:46:24 Mar 14 2021
Status : int8, per-channel, symmetric
Input model : ./mobilenet_fp32.tmfile
Output model: ./mobilenet_int8.tmfile
Calib images: ./dataset
Algorithm : KL
Dims : 3 224 224
Mean : 104.007 116.669 122.679
Scale : 0.017 0.017 0.017
BGR2RGB : ON
Center crop : OFF
Letter box : OFF
Thread num : 1
[Quant Tools Info]: Step 0, load FP32 tmfile.
[Quant Tools Info]: Step 0, load FP32 tmfile done.
[Quant Tools Info]: Step 0, load calibration image files.
[Quant Tools Info]: Step 0, load calibration image files done, image num is 55.
[Quant Tools Info]: Step 1, find original calibration table.
[Quant Tools Info]: Step 1, find original calibration table done, output ./table_minmax.scale
[Quant Tools Info]: Step 2, find calibration table.
[Quant Tools Info]: Step 2, find calibration table done, output ./table_kl.scale
[Quant Tools Info]: Thread 1, image nums 55, total time 1964.24 ms, avg time 35.71 ms
[Quant Tools Info]: Calibration file is using table_kl.scale
[Quant Tools Info]: Step 3, load FP32 tmfile once again
[Quant Tools Info]: Step 3, load FP32 tmfile once again done.
[Quant Tools Info]: Step 3, load calibration table file table_kl.scale.
[Quant Tools Info]: Step 4, optimize the calibration table.
[Quant Tools Info]: Step 4, quantize activation tensor done.
[Quant Tools Info]: Step 5, quantize weight tensor done.
[Quant Tools Info]: Step 6, save Int8 tmfile done, ./mobilenet_int8.tmfile
[Quant Tools Info]: Step Evaluate, evaluate quantitative losses
cosin 0 32 avg 0.995317 ### 0.000000 0.953895 0.998249 0.969256 ...
cosin 1 32 avg 0.982403 ### 0.000000 0.902383 0.964436 0.873998 ...
cosin 2 64 avg 0.976753 ### 0.952854 0.932301 0.982766 0.958503 ...
cosin 3 64 avg 0.981889 ### 0.976637 0.981754 0.987276 0.970671 ...
cosin 4 128 avg 0.979728 ### 0.993999 0.991858 0.990438 0.992766 ...
cosin 5 128 avg 0.970351 ### 0.772556 0.989541 0.986996 0.989563 ...
cosin 6 128 avg 0.954545 ### 0.950125 0.922964 0.946804 0.972852 ...
cosin 7 128 avg 0.977192 ### 0.994728 0.972071 0.995353 0.992700 ...
cosin 8 256 avg 0.977426 ### 0.968429 0.991248 0.991274 0.994450 ...
cosin 9 256 avg 0.962224 ### 0.985255 0.969171 0.958762 0.967461 ...
cosin 10 256 avg 0.954253 ### 0.984353 0.935643 0.656188 0.929778 ...
cosin 11 256 avg 0.971987 ### 0.997596 0.967681 0.476525 0.999115 ...
cosin 12 512 avg 0.972861 ### 0.968920 0.905907 0.993918 0.622953 ...
cosin 13 512 avg 0.959161 ### 0.935686 0.000000 0.642560 0.994388 ...
cosin 14 512 avg 0.963903 ### 0.979613 0.957169 0.976440 0.902512 ...
cosin 15 512 avg 0.963226 ### 0.977065 0.965819 0.998149 0.905297 ...
cosin 16 512 avg 0.960935 ### 0.861674 0.972926 0.950579 0.987609 ...
cosin 17 512 avg 0.961057 ### 0.738472 0.987884 0.999124 0.995397 ...
cosin 18 512 avg 0.960127 ### 0.935455 0.968909 0.970831 0.981240 ...
cosin 19 512 avg 0.963755 ### 0.972628 0.992305 0.999518 0.799737 ...
cosin 20 512 avg 0.949364 ### 0.922776 0.896038 0.945079 0.971338 ...
cosin 21 512 avg 0.961256 ### 0.902256 0.896438 0.923361 0.973974 ...
cosin 22 512 avg 0.946552 ### 0.963806 0.982075 0.878965 0.929992 ...
cosin 23 512 avg 0.953677 ### 0.953880 0.996364 0.936540 0.930796 ...
cosin 24 1024 avg 0.941197 ### 0.000000 0.992507 1.000000 0.994460 ...
cosin 25 1024 avg 0.973546 ### 1.000000 0.889181 0.000000 0.998084 ...
cosin 26 1024 avg 0.869351 ### 0.522966 0.000000 0.987009 0.000000 ...
cosin 27 1 avg 0.974982 ### 0.974982
cosin 28 1 avg 0.974982 ### 0.974982
cosin 29 1 avg 0.974982 ### 0.974982
cosin 30 1 avg 0.978486 ### 0.978486
---- Tengine Int8 tmfile create success, best wish for your INT8 inference has a low accuracy loss...\(^0^)/ ----
| Type | Note |
|---|---|
| Adaptive | TENGINE_MODE_UINT8 |
| Activation data | UInt8 |
| Weight date | UInt8 |
| Bias date | Int32 |
| Example | tm_classification_uint8.c |
| Execution environment | Ubuntu 18.04 |
$ ./quant_tool_uint8 -h
---- Tengine Post Training Quantization Tool ----
Version : v1.2, 15:20:08 Jul 25 2021
Status : uint8, per-layer, asymmetric
[Quant Tools Info]: The input file of Float32 tmfile file not specified!
[Quant Tools Info]: optional arguments:
-h help show this help message and exit
-m input model path to input float32 tmfile
-i image dir path to calibration images folder
-f scale file path to calibration scale file
-o output model path to output uint8 tmfile
-a algorithm the type of quant algorithm(0:min-max, 1:kl, 2:aciq, default is 0)
-g size the size of input image(using the resize the original image,default is 3,224,224)
-w mean value of mean (mean value, default is 104.0,117.0,123.0)
-s scale value of normalize (scale value, default is 1.0,1.0,1.0)
-b swapRB flag which indicates that swap first and last channels in 3-channel image is necessary(0:OFF, 1:ON, default is 1)
-c center crop flag which indicates that center crop process image is necessary(0:OFF, 1:ON, default is 0)
-y letter box the size of letter box process image is necessary([rows, cols], default is [0, 0])
-k focus flag which indicates that focus process image is necessary(maybe using for YOLOv5, 0:OFF, 1:ON, default is 0)
-t num thread count of processing threads(default is 1)
[Quant Tools Info]: example arguments:
./quant_tool_uint8 -m ./mobilenet_fp32.tmfile -i ./dataset -o ./mobilenet_uint8.tmfile -g 3,224,224 -w 104.007,116.669,122.679 -s 0.017,0.017,0.017
Before use the quant tool, you need Float32 tmfile and Calibration Dataset, the image num of calibration dataset we suggest to use 500-1000.
$ .quant_tool_uint8 -m ./mobilenet_fp32.tmfile -i ./dataset -o ./mobilenet_uint8.tmfile -g 3,224,224 -w 104.007,116.669,122.679 -s 0.017,0.017,0.017
---- Tengine Post Training Quantization Tool ----
Version : v1.2, 18:32:53 May 30 2021
Status : uint8, per-layer, asymmetric
Input model : ./mobilenet_fp32.tmfile
Output model: ./mobilenet_uint8.tmfile
Calib images: ./dataset
Scale file : NULL
Algorithm : MIN MAX
Dims : 3 224 224
Mean : 104.000 117.000 123.000
Scale : 0.017 0.017 0.017
BGR2RGB : ON
Center crop : OFF
Letter box : 0 0
YOLOv5 focus: OFF
Thread num : 4
[Quant Tools Info]: Step 0, load FP32 tmfile.
[Quant Tools Info]: Step 0, load FP32 tmfile done.
[Quant Tools Info]: Step 0, load calibration image files.
[Quant Tools Info]: Step 0, load calibration image files done, image num is 5.
[Quant Tools Info]: Step 1, find original calibration table.
[Quant Tools Info]: Step 1, images 00005 / 00005
[Quant Tools Info]: Step 1, find original calibration table done, output ./table_minmax.scale
[Quant Tools Info]: Thread 4, image nums 5, total time 37.23 ms, avg time 87.45 ms
[Quant Tools Info]: Calibration file is using table_minmax.scale
[Quant Tools Info]: Step 3, load FP32 tmfile once again
[Quant Tools Info]: Step 3, load FP32 tmfile once again done.
[Quant Tools Info]: Step 3, load calibration table file table_minmax.scale.
[Quant Tools Info]: Step 4, optimize the calibration table.
[Quant Tools Info]: Step 4, quantize activation tensor done.
[Quant Tools Info]: Step 5, quantize weight tensor done.
[Quant Tools Info]: Step 6, save Int8 tmfile done, mobilenet_uint8.tmfile
---- Tengine Int8 tmfile create success, best wish for your INT8 inference has a low accuracy loss...\(^0^)/ ----