Workflow

Provides a full pipeline from fetching the training data, training the model to submitting the predictions.

class workflow.Workflow(*args, **kwargs)[source]

A luigi task pipeline that fetches the most recent data, trains a model on said data and does a submission to the numer.ai website.

Param:output_path (str): path where the data shall be stored, defaults to ./data/.
Param:public_id (str): public_id from the API credentials
Param:secret (str): secret from the API credentials
requires()[source]

Formulates the incoming dependencies, in this case the retrieval of the data for the current tournament. Returns the results of the tasks.numerai_train_and_predict.TrainAndPredict task as a way to forcefully call the task without depending on it.

run()[source]

After the incoming dependencies (tasks.numerai_fetch_training_data.FetchAndExtractData, and tasks.numerai_train_and_predict.TrainAndPredict) are done, the only thing that’s left is the submission of the predictions. This method is taking care of this.