38 lines
1.5 KiB
Markdown
38 lines
1.5 KiB
Markdown
|
# How to use model_pruning function
|
||
|
|
|
||
|
|
## Model Source
|
||
|
|
The model used in this example come from the following open source projects:
|
||
|
|
https://github.com/shicai/MobileNet-Caffe
|
||
|
|
|
||
|
|
## Script Usage
|
||
|
|
*Usage:*
|
||
|
|
```
|
||
|
|
python test.py
|
||
|
|
```
|
||
|
|
*Description:*
|
||
|
|
- The default target platform in script is 'rk3566', please modify the 'target_platform' parameter of 'rknn.config' according to the actual platform.
|
||
|
|
- If connecting board is required, please add the 'target' parameter in 'rknn.init_runtime'.
|
||
|
|
- The 'model_pruning' parameter of 'rknn.config' is set to True.
|
||
|
|
- When verbose is set to True, the following similar prompts will appear during the build process,
|
||
|
|
indicating that model pruning has been effective for this model. (This means that approximately 6.9%
|
||
|
|
of the weights have been removed, resulting in a saving of about 13.4% of the computational workload.)
|
||
|
|
Please note that not all models can be pruned, only models with sparse weights are likely to benefit from pruning.
|
||
|
|
```
|
||
|
|
I model_pruning ...
|
||
|
|
I model_pruning results:
|
||
|
|
I -1.12144 MB (-6.9%)
|
||
|
|
I -0.00016 T (-13.4%)
|
||
|
|
I model_pruning done.
|
||
|
|
```
|
||
|
|
|
||
|
|
## Expected Results
|
||
|
|
This example will print the TOP5 labels and corresponding scores of the test image classification results, as follows:
|
||
|
|
```
|
||
|
|
-----TOP 5-----
|
||
|
|
[155]: 0.724609375
|
||
|
|
[154]: 0.1920166015625
|
||
|
|
[204]: 0.0509033203125
|
||
|
|
[284]: 0.004177093505859375
|
||
|
|
[252 283]: 0.0038623809814453125
|
||
|
|
```
|
||
|
|
- Note: Different platforms, different versions of tools and drivers may have slightly different results.
|