TwinModels

TwinModels is an open source library of two-tower matching models, built with stunning features in configurability, tunability, and reproducibility.

Model List

Get Started

Run the demo

The code workflow is structured as follows:

# Set the data config and model config
feature_cols = [{...}] # define feature columns
label_col = {...} # define label column
params = {...} # set data params and model params

# Set the feature encoding specs
feature_encoder = FeatureEncoder(feature_cols, label_col, ...) # define the feature encoder
datasets.build_dataset(feature_encoder, ...) # fit feature_encoder and build dataset 

# Load data generators
train_gen, valid_gen, test_gen = h5_generator(feature_encoder, ...)

# Define a model
model = SimpleX(...)

# Train the model
model.fit(train_gen, valid_gen, ...)

# Evaluation
model.evaluate(test_gen)

Run the benchmark

For reproducing the experiment result, you can run the benchmarking script with the corresponding config file as follows.

• --config: The config file that defines the hyper-parameter space.
• --gpu: The gpu index used for experiment, and -1 for CPU.

cd benchmarks
python run_param_tuner.py --config Yelp18/SimpleX_yelp18_x0/SimpleX_yelp18_x0_tuner_config.yaml --gpu 0