Pipeline

Pipeline(self, model, dataloader_cls)

Runs model predictions from raw files:

raw files --(dataloader)--> data batches --(model)--> prediction

Arguments

• model: model returned by kipoi.get_model
• dataloader_cls: dataloader class returned by kipoi.get_dataloader_factory of kipoi.get_model().default_dataloader

predict_example

Pipeline.predict_example(self, batch_size=32, output_file=None, **kwargs)

Run model prediction for the example file

Arguments

• batch_size: batch_size
• output_file: if not None, inputs and predictions are stored to output_file path
• **kwargs: Further arguments passed to batch_iter

predict

Pipeline.predict(self, dataloader_kwargs, batch_size=32, **kwargs)

Arguments

• dataloader_kwargs: Keyword arguments passed to the pre-processor
• **kwargs: Further arguments passed to batch_iter

Returns

np.array, dict, list: Predict the whole array

predict_generator

Pipeline.predict_generator(self, dataloader_kwargs, batch_size=32, layer=None, **kwargs)

Prediction generator

Arguments

• dataloader_kwargs: Keyword arguments passed to the dataloader
• batch_size: Size of batches produced by the dataloader
• layer: If not None activation of specified layer will be returned. Only possible for models that are a subclass of LayerActivationMixin.
• **kwargs: Further arguments passed to batch_iter

Yields

• dict: model batch prediction

predict_to_file

Pipeline.predict_to_file(self, output_file, dataloader_kwargs, batch_size=32, keep_inputs=False, keep_metadata=False, **kwargs)

Make predictions and write them iteratively to a file

Arguments

• output_file: output file path. File format is inferred from the file path ending. Available file formats are: 'bed', 'h5', 'hdf5', 'tsv'
• dataloader_kwargs: Keyword arguments passed to the dataloader
• batch_size: Batch size used for the dataloader
• keep_inputs: if True, inputs and targets will also be written to the output file.
• keep_metadata: if True, metadata will also be written to the output file.
• **kwargs: Further arguments passed to batch_iter

input_grad

Pipeline.input_grad(self, dataloader_kwargs, batch_size=32, filter_idx=None, avg_func=None, layer=None, final_layer=True, selected_fwd_node=None, pre_nonlinearity=False, **kwargs)

Get input gradients

Arguments

• dataloader_kwargs: Keyword arguments passed to the dataloader
• batch_size: Batch size used for the dataloader
• filter_idx: filter index of layer for which the gradient should be returned
• avg_func: String name of averaging function to be applied across filters in layer layer
• layer: layer from which backwards the gradient should be calculated
• final_layer: Use the final (classification) layer as layer
• selected_fwd_node: None - not supported by KerasModel at the moment
• pre_nonlinearity: Try to use the layer output prior to activation (will not always be possible in an automatic way)
• **kwargs: Further arguments passed to input_grad

Returns

dict: A dictionary of all model inputs and the gradients. Gradients are stored in key 'grads'

input_grad_generator

Pipeline.input_grad_generator(self, dataloader_kwargs, batch_size=32, filter_idx=None, avg_func=None, layer=None, final_layer=True, selected_fwd_node=None, pre_nonlinearity=False, **kwargs)

Get input gradients

Arguments

• dataloader_kwargs: Keyword arguments passed to the dataloader
• batch_size: Batch size used for the dataloader
• filter_idx: filter index of layer for which the gradient should be returned
• avg_func: String name of averaging function to be applied across filters in layer layer
• layer: layer from which backwards the gradient should be calculated
• final_layer: Use the final (classification) layer as layer
• selected_fwd_node: None - not supported by KerasModel at the moment
• pre_nonlinearity: Try to use the layer output prior to activation (will not always be possible in an automatic way)
• **kwargs: Further arguments passed to input_grad

Yields

• dict: A dictionary of all model inputs and the gradients. Gradients are stored in key 'grads'