ANOMLEE: A Neat-O Machine Learning Execution Environment

💪 by Quilt's ewasm-rt

This repo demonstrates how a trained ML model could run inside an Ethereum 2 EE.

Prerequisits

Install LLVM

brew install llvm
echo 'export PATH="/usr/local/opt/llvm/bin:$PATH"' >> ~/.bash_profile

Install the WebAssembly Binary Toolkit

brew install wabt

Install Python Environment & Dependencies

python3 -m venv env
source ./env/bin/activate
pip3 install -r requirements.txt

Build

make all

Random Forest Classifier (iris model)

This model that is trained to determine which of the following types of iris flowers:

• setosa
• versicolor
• virginica

based on the following design variables:

• sepal length (cm)
• sepal width (cm)
• petal length (cm)
• petal width (cm)

Benchmarks:

• Python: 103307 microseconds
• eWasm: 551.849 microseconds
• C: 1 microsecond

Binary size:

• 826 bytes

Handwritten Digit Classifier (digit model)

This model uses a sequential neural network model to classify handwritten digits from the MINST database

TODO: finish

Run Benchmarks (all models)

$ make benchmark
########## Python Benchmark: ###########

Execution time: 103307 microseconds.


 0    setosa
Name: species, dtype: category
Categories (3, object): [setosa, versicolor, virginica] 

########## eWasm Benchmark: ###########
    Finished release [optimized] target(s) in 0.03s
     Running target/release/deps/anomlee-0de6000a8c14c88e

running 1 test

Execution Time: 551.849µs

test tests::test ... ok

test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out

########## C Benchmark: ##########

Execution time: 1 microseconds


Probabilities: 
1.000000 0.000000 0.000000 

Model Predicts: 
setosa