====== Instructions for KSNN conversion tool ======

Introduction and guide on how to use the KSNN conversion tool.

===== Get the conversion tool =====

The conversion tool is integrated in the NPU SDK. You can download it from [[dl>products/vim3/tools/npu-sdk/| aml_npu_sdk]]

```shell
$ wget https://dl.khadas.com/products/vim3/tools/npu-sdk/aml_npu_sdk-ddk-6.4.8.7-250427.tgz
$ tar xvzf aml_npu_sdk-ddk-6.4.8.7-250427.tgz
$ ln -fs aml_npu_sdk-ddk-6.4.8.7-250427 aml_npu_sdk
```

The KSNN conversion tool is under ''acuity-toolkit/python'',

```shell
$ cd aml_npu_sdk/acuity-toolkit/python && ls
$ convert  data  outputs
```

===== Conversion example =====

Choose tensorflow MobileNet SSD as an example.

1. Get the frozen model,

```shell
$ cd aml_npu_sdk/acuity-toolkit/python
$ wget https://github.com/numbqq/models-zoo/raw/master/tensorflow/mobilenet_ssd/ssd_mobilenet_v1_coco_2017_11_17.pb
```

2. Convert,

```shell
# uint8
$ ./convert --model-name mobilenet_ssd \
           --platform tensorflow \
           --model ./ssd_mobilenet_v1_coco_2017_11_17.pb \
           --input-size-list '300,300,3' \
           --inputs FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_0/BatchNorm/batchnorm/mul_1 \
           --outputs "'concat concat_1'" \
           --mean-values '127.5 127.5 127.5 0.007843137' \
           --quantized-dtype asymmetric_affine \
           --source-files ./data/dataset/dataset0.txt \
           --kboard VIM3 --print-level 0

# int8
$ ./convert --model-name mobilenet_ssd \
           --platform tensorflow \
           --model ./ssd_mobilenet_v1_coco_2017_11_17.pb \
           --input-size-list '300,300,3' \
           --inputs FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_0/BatchNorm/batchnorm/mul_1 \
           --outputs "'concat concat_1'" \
           --mean-values '127.5 127.5 127.5 0.007843137' \
           --quantized-dtype dynamic_fixed_point \
           --qtype int8 \
           --source-files ./data/dataset/dataset0.txt \
           --kboard VIM3 --print-level 0

# int16
$ ./convert --model-name mobilenet_ssd \
           --platform tensorflow \
           --model ./ssd_mobilenet_v1_coco_2017_11_17.pb \
           --input-size-list '300,300,3' \
           --inputs FeatureExtractor/MobilenetV1/MobilenetV1/Conv2d_0/BatchNorm/batchnorm/mul_1 \
           --outputs "'concat concat_1'" \
           --mean-values '127.5 127.5 127.5 0.007843137' \
           --quantized-dtype dynamic_fixed_point \
           --qtype int16 \
           --source-files ./data/dataset/dataset0.txt \
           --kboard VIM3 --print-level 0
```

During conversion, if you need to view detailed information, you can modify ''--print-level 0'' to ''--print-level 1''.

3. Model file generated after conversion.

```shell
$ cd aml_npu_sdk/acuity-toolkit/python
$ ls outputs/mobilenet_ssd/
mobilenet_ssd.nb  libnn_mobilenet_ssd.so
```

===== More =====

  - [[ksnn-usage|KSNN Usage]]
  - [[ksnn-api|KSNN API Documentation]]