feat: Enable OpenVINO backend for WASI-NN

[hydai]

Motivation

In the previous ticket #343, we already have some WASI-NN implementation. However, the backend is using ONNX. In this ticket, we would like to have both ONNX and OpenVINO backend.

Machine Learning is a big topic nowadays. WasmEdge already provides a set of TensorFlow host functions to enable the ML inference in WebAssembly. However, these TensorFlow host functions are defined by us and they are just a Wasm function binding from the TensorFlow C API.

Here comes a standard, the WASI-NN proposal provides a new way to perform neural network inferencing by using a runtime-provided implementation that can leverage host native optimizations, CPU multi-threading, or powerful hardware devices such as GPUs or TPUs.

Specification

1. Get familiar with WASI-NN proposal.
2. Modify the current WASI-NN implementation for enabling multiple backend support.
3. Take OpenVINO as another WASI-NN backend and create a demo kit that is similar to Wasmtime
4. Create unit tests and integrations tests to verify the WASI-NN component.

Minimum qualifications

To apply this LFX mentorship, there are some minimum qualifications:

1. Experience with C++ programming language.
2. Finish the challenge LFX Mentorship 2022 Spring Challenge - Proposal #1221

Appendix

1. WasmEdge TensorFlow rust SDK: https://github.com/second-state/wasmedge_tensorflow_interface
2. WasmEdge TensorFlow host functions: https://github.com/second-state/WasmEdge-tensorflow
3. TensorFlow C API: https://github.com/tensorflow/tensorflow/blob/master/tensorflow/c/c_api.h
4. WASI-NN proposal: https://github.com/WebAssembly/wasi-nn
5. Implementing-wasi-nn-in-wasmtime: https://bytecodealliance.org/articles/implementing-wasi-nn-in-wasmtime

[wangyuan249]

Hi ~ @hydai, I am ziyuan Wang, currently study at ZheJiang University and will graduate at June 30th this summer. I would like to apply for LFX mentorship in this year and work on it.

First, I have open source experience through last year's internship in AliYun while working with CNCF ToC Zhanglei's Team and I made contribution to oam/KubeVela.
And I also familiar with Deep learning and Model Inference in Edge Computing. In this area, I have trained a face recognition model through Caffe, building self-defined layer through C++. For another, I have used tensorflow C API to inference and detect object, converting tf model to ONNX and OM model(HuaWei MindSpore model), and deployed on the Ascend 310 chip.

Thus, I think I'm fit for this task and I would be grateful if you could guide me to relevant resources. Thanks~

[hydai]

Hi @wangyuan249,

I just updated the Minimum qualifications. If you are interested in this issue, please finish the challenge. It will help you to get familiar with the WasmEdge data structure and how to interact with the WasmEdge.

[gusye1234]

Hi @hydai,
I'm curious, for those who want to apply for this LFX mentorship, is there anything else we can do besides waiting after applying on the mentorship.lfx.linuxfoundation.org website and finishing the challenge?
I would really appreciate it if your could answer😊.

[q82419]

Implementation: #1340 #1440 #1461
Unit test: #1446
Docs: #1414
Examples: second-state/WasmEdge-WASINN-examples#1