Home

Libevo

Welcome to Evo Page

{% assign date = '2020-04-13T10:20:00Z' %}

• Original date - {{ date }}

• With timeago filter - {{ date | timeago }}

• libevo is a Inference engine used in TinyML write by pure C.

• Keyword: Lite > High Performance > Easy to Use > Compatibility
• Try online

1 Docs & PPT

1.1 Docs

about	doc	desc
model	evo-mdl.md	model load
operator	evo-op.md	operator lib
bindle	evo-bind.md	Bindle API
tools	evo-tool.md	Useful Tools
profile	evo-profile.md	profiler
quant	evo-quant.md	quant
reference	evo-ref.md	reference

1.1 PPT

about	ppt	desc
Infer Engine	ppt-01	Engine
Domin Engine	ppt-02	Engine
Dynamic Eng	ppt-04	Engine
Model Quant	ppt-05	Model
Infer Intro	ppt-06	Engine
Runtime Opt	ppt-07	Runtime
Memory Org	ppt-08	Memory

2 Build

• To build the item, you should:

make

3 Demo

• Here is a c demo for libevo :

#include <evo.h>

int main() {
    // ...
    serializer_t * sez = serializer_get("onnx");
    model_t * mdl = sez->load_file(sez, "model/mnist_8/model.onnx");
    tensor_t * t1 = model_get_tensor(mdl, "Input3");
    tensor_dump(t1);
    // ...
    return 0;
}

4 Performance

• Task: Embedding Deep Learning Model
• Compare to: ORT, TFLM, TVM and so on
• BenchMark: MLPerf

5 Other Features

• Code scale:

make line   # calculate lines of code