Data/output -- create pipelines directory with the related output processing

[nicolay-r]

В data/output нужно произвести рефакторинг имён и параметров

• Item handle, iter, map move into base pipeline folder
• Move and refactor the code below and place at common
  

  AREkit/arekit/contrib/networks/core/callback/utils_model_eval.py

  Lines 96 to 145 in fc007a0

  	def __convert_output_to_opinion_collections(exp_io, opin_ops, doc_ops, labels_scaler,
  	output_storage, data_type, epoch_index,
  	supported_collection_labels, label_calc_mode, labels_formatter):
  	assert(isinstance(opin_ops, OpinionOperations))
  	assert(isinstance(doc_ops, DocumentOperations))
  	assert(isinstance(labels_scaler, BaseLabelScaler))
  	assert(isinstance(exp_io, NetworkIOUtils))
  	assert(isinstance(data_type, DataType))
  	assert(isinstance(epoch_index, int))
  	assert(isinstance(label_calc_mode, LabelCalculationMode))
  	assert(isinstance(labels_formatter, StringLabelsFormatter))
  	cmp_doc_ids_set = set(doc_ops.iter_tagget_doc_ids(BaseDocumentTag.Compare))
  	output_view = MulticlassOutputView(labels_scaler=labels_scaler,
  	storage=output_storage)
  	# Opinion collections iterator pipeline.
  	collections_iter_pipeline = BasePipeline([
  	FilterPipelineItem(filter_func=lambda doc_id: doc_id in cmp_doc_ids_set),
  	MapPipelineItem(lambda doc_id:
  	(doc_id,
  	output_view.iter_opinion_linkages(
  	doc_id=doc_id,
  	opinions_view=exp_io.create_opinions_view(data_type)))
  	),
  	MapPipelineItem(lambda doc_id, linkages_iter:
  	(doc_id,
  	__create_opinion_collection(
  	linked_iter=linkages_iter,
  	supported_labels=supported_collection_labels,
  	create_opinion_collection=opin_ops.create_opinion_collection,
  	label_scaler=labels_scaler))),
  	])
  	# Executing pipeline.
  	pipeline_ctx = PipelineContext({"src": output_view.iter_doc_ids()})
  	collections_iter_pipeline.run(pipeline_ctx)
  	# Save collection.
  	for doc_id, collection in pipeline_ctx.provide("src"):
  	target = exp_io.create_result_opinion_collection_target(
  	data_type=data_type,
  	epoch_index=epoch_index,
  	doc_id=doc_id)
  	exp_io.write_opinion_collection(collection=collection,
  	labels_formatter=labels_formatter,
  	target=target)

[nicolay-r]

1. To Opinion Collections for example.