T0b1/wasmtime
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
Files
main
wasmtime/crates/wasi-nn/README.md
Andrew Brown a61f068c64 Add an initial wasi-nn implementation for Wasmtime (#2208) 
* Add an initial wasi-nn implementation for Wasmtime

This change adds a crate, `wasmtime-wasi-nn`, that uses `wiggle` to expose the current state of the wasi-nn API and `openvino` to implement the exposed functions. It includes an end-to-end test demonstrating how to do classification using wasi-nn:
 - `crates/wasi-nn/tests/classification-example` contains Rust code that is compiled to the `wasm32-wasi` target and run with a Wasmtime embedding that exposes the wasi-nn calls
 - the example uses Rust bindings for wasi-nn contained in `crates/wasi-nn/tests/wasi-nn-rust-bindings`; this crate contains code generated by `witx-bindgen` and eventually should be its own standalone crate

* Test wasi-nn as a CI step

This change adds:
 - a GitHub action for installing OpenVINO
 - a script, `ci/run-wasi-nn-example.sh`, to run the classification example
2020-11-16 12:54:00 -06:00

1.5 KiB
Raw Permalink Blame History

wasmtime-wasi-nn

This crate enables support for the wasi-nn API in Wasmtime. Currently it contains an implementation of wasi-nn using OpenVINO™ but in the future it could support multiple machine learning backends. Since the wasi-nn API is expected to be an optional feature of WASI, this crate is currently separate from the wasi-common crate. This crate is experimental and its API, functionality, and location could quickly change.

Use

Use the Wasmtime APIs to instantiate a Wasm module and link in the WasiNn implementation as follows:

let wasi_nn = WasiNn::new(&store, WasiNnCtx::new()?);
wasi_nn.add_to_linker(&mut linker)?;

Build

This crate should build as usual (i.e. cargo build) but note that using an existing installation of OpenVINO™, rather than building from source, will drastically improve the build times. See the openvino crate for more information

Example

An end-to-end example demonstrating ML classification is included in examples:

  • tests/wasi-nn-rust-bindings contains ergonomic bindings for writing Rust code against the wasi-nn APIs
  • tests/classification-example contains a standalone Rust project that uses the wasi-nn APIs and is compiled to the wasm32-wasi target using the wasi-nn-rust-bindings

Run the example from the Wasmtime project directory:

ci/run-wasi-nn-example.sh