MCPcopy Create free account
hub / github.com/OpenBitSys/BitDistiller / _process_doc

Method _process_doc

test/general/lm_eval/tasks/qasper.py:146–165  ·  view source on GitHub ↗

Given a `doc`, flatten it out so that each JSON blob contains exactly one question and one answer. Logic taken from the reference implementation available at https://github.com/allenai/qasper-led-baseline/blob/main/scripts/evaluator.py

(self, doc)

Source from the content-addressed store, hash-verified

144 yield from self._process_doc(doc)
145
146 def _process_doc(self, doc):
147 """Given a `doc`, flatten it out so that each JSON blob
148 contains exactly one question and one answer. Logic taken from
149 the reference implementation available at
150 https://github.com/allenai/qasper-led-baseline/blob/main/scripts/evaluator.py
151 """
152 obs_list = []
153 for question, answer_list in zip(doc["qas"]["question"], doc["qas"]["answers"]):
154 for answer_blob in answer_list["answer"]:
155 answer, answer_type = categorise_answer(answer_blob)
156 obs_list.append(
157 {
158 "title": doc["title"],
159 "abstract": doc["abstract"],
160 "question": question,
161 "answer": answer,
162 "answer_type": answer_type,
163 }
164 )
165 return obs_list
166
167 def process_results(self, doc, results):
168 # TODO: Calculate a score for extractive spans once a request type for generating

Callers 2

training_docsMethod · 0.95
validation_docsMethod · 0.95

Calls 1

categorise_answerFunction · 0.85

Tested by

no test coverage detected