How to use the repo

python setup

cd mldb/container_files/drafts/rec
virtualenv rec-ve
source rec-ve/bin/activate
pip install -r requirements.txt

... or at mldb.ai (ask Jean for details)

pip install -U pip
pip install --no-index -r requirements.txt

training

For now I have only used the movielens ml-latest-small.zip dataset. Here is how to use it to train a model and test it.

# load the data into MLDB
MLDB_REC_CONFIG=<your_config.json> python movielens/create_datasets.py <your path to ml-latest-small.zip>
# train a model!
MLDB_REC_CONFIG=<your_config.json> python rec/train.py -d 2013-01-01 -n 2000
# test it
MLDB_REC_CONFIG=<your_config.json> python rec/test.py

There is a movielens_config_sample.json in the repo. This could in theory kind of work for BTR by changing the config but it's definitely not quite there yet.