get_model_descr

get_model_descr(model, source='kipoi')

Get model description

Arguments

  • model: model's relative path/name in the source. 2nd column in the kipoi.list_models()pd.DataFrame`.
  • source: Model source. 1st column in the kipoi.list_models() pd.DataFrame.

get_model

get_model(model, source='kipoi', with_dataloader=True)

Load the model from source, as well as the default dataloder to model.default_dataloder.

Arguments

  • model (str): model name
  • source (str): source name
  • with_dataloader (bool): if True, the default dataloader is loaded to model.default_dataloader and the pipeline at model.pipeline enabled.

Returns

  • Instance of class inheriting from kipoi.models.BaseModel (like kipoi.models.KerasModel) decorated with additional attributes.

Methods

  • predict_on_batch(x): Make model predictions given a batch of data x

Appended attributes

  • type (str): model type (class name)
  • args (dict): model args used to instantiate the model class
  • info (kipoi.specs.Info): information about the author (etc)
  • schema (kipoi.specs.ModelSchema): information about the input/outputdata modalities
  • dependencies (kipoi.specs.Dependencies): class specifying the dependencies. (implements install method for running the installation)
  • default_dataloader (class inheriting from kipoi.data.BaseDataLoader): default dataloader. None if with_dataloader=False was used.
  • name (str): model name
  • source (str): model source
  • source_dir (str): local path to model source storage
  • postprocessing (dict): dictionary of loaded plugin specifications
  • pipeline (kipoi.pipeline.Pipeline): handle to a Pipeline object