XGBoost prediction using Webassembly

This is a Proof of Concept of how a model-based prediction can be run within the browser via Webassembly. Webassembly is a promising new technology allowing webapplications to run programs written in languages like Rust, C, C++, Java, or Golang and compiled into bytecode directly in the Browser.

This repository serves a simple web assembly (wasm) application to perform a prediction based on a pre-trained xgb model, using data from a table in the browser, which can be loaded as a delimited file by the user. We use the golang leaves leaves package to do the work. This is heavily inspired by this which uses golang and webassembly.

Flow

• The xgboost model has been trained with python (see python_modeling).
• Once the model has been trained, it is exported as binary file (model.bst).
• We then use the leaves golang package to compute predictions based on this model.
• We use the syscall/js package to interface the golang script with javascript
• We finally compile the golang script into Webassembly bytecode which can be run within the Browser.

As a matter of fact, intensive computing within the browser via Webassembly can be up to 20 times faster than the Javascript equivalent.

Local

you need golang installed on your machine.

1 - build the wasm.

$ cd xgb-wasm
$ make

2 - Add your own Go version specific wasm_exec.js file :

$ cp "$(go env GOROOT)/misc/wasm/wasm_exec.js" ./demo

3 - cd into the "demo" folder and execute sheret.exe (Windows users) to start a local server.

4 - open the browser at http://localhost:9999

Credits

Vanessa Sochat / Lawrence Livermore National Laboratory / vsoch.github.io