def post_processing_function(examples, features, predictions, stage="eval"):
# Post-processing: we match the start logits and end logits to answers in the original context.
predictions, scores_diff_json = postprocess_qa_predictions_with_beam_search(
examples=examples,
features=features,
predictions=predictions,
version_2_with_negative=data_args.version_2_with_negative,
n_best_size=data_args.n_best_size,
max_answer_length=data_args.max_answer_length,
start_n_top=model.config.start_n_top,
end_n_top=model.config.end_n_top,
output_dir=training_args.output_dir,
is_world_process_zero=trainer.is_world_process_zero(),
prefix=stage,
)
# Format the result to the format the metric expects.
if data_args.version_2_with_negative:
formatted_predictions = [
{"id": k, "prediction_text": v, "no_answer_probability": scores_diff_json[k]}
for k, v in predictions.items()
]
else:
formatted_predictions = [{"id": k, "prediction_text": v} for k, v in predictions.items()]
references = [{"id": ex["id"], "answers": ex[answer_column_name]} for ex in examples]
return EvalPrediction(predictions=formatted_predictions, label_ids=references)
def post_processing_function(examples,
features,
predictions,
stage="eval"):
# Post-processing: we match the start logits and end logits to
# answers in the original context.
if data_args.beam_search:
predictions, scores_diff_json = \
postprocess_qa_predictions_with_beam_search(
examples=examples,
features=features,
predictions=predictions,
version_2_with_negative=data_args.version_2_with_negative,
n_best_size=data_args.n_best_size,
max_answer_length=data_args.max_answer_length,
start_n_top=model.config.start_n_top,
end_n_top=model.config.end_n_top,
output_dir=training_args.output_dir,
# log_level=log_level,
prefix=stage,
)
else:
predictions = postprocess_qa_predictions(
examples=examples,
features=features,
predictions=predictions,
version_2_with_negative=data_args.version_2_with_negative,
n_best_size=data_args.n_best_size,
max_answer_length=data_args.max_answer_length,
output_dir=training_args.output_dir,
prefix=stage,
)
if data_args.version_2_with_negative:
if data_args.beam_search:
formatted_predictions = [
{
"id": k,
"prediction_text": v,
"no_answer_probability": scores_diff_json[k]
} # noqa E501
for k, v in predictions.items()
]
else:
formatted_predictions = [
{
"id": k,
"prediction_text": v,
"no_answer_probability": 0.0
} for k, v in predictions.items() # noqa E501
]
else:
formatted_predictions = [{
"id": k,
"prediction_text": v
} for k, v in predictions.items()] # noqa E501
references = [{
"id": ex["id"],
"answers": ex[answer_column_name]
} for ex in examples] # noqa E501
return EvalPrediction(predictions=formatted_predictions,
label_ids=references)
