def run_pipeline(model,
fictitious_entities,
sentences,
config,
number_of_entity_trials,
logger,
temporal=False):
if not temporal:
dataset = proc.prepare_masked_instances(
sentences=sentences,
config=config,
fictitious_entities=fictitious_entities,
num_entity_trials=number_of_entity_trials)
else:
dataset = proc.prepare_masked_instances_temporal(
sentences=sentences,
config=config,
fictitious_entities=fictitious_entities,
num_entity_trials=number_of_entity_trials)
logger.info("finished creating dataset")
perf = utils.fair_seq_masked_word_prediction(
masked_examples=dataset,
model=model,
gpu_available=torch.cuda.is_available(),
top_n=100,
logger=logger)
logger.info("finished evaluating dataset")
output_df = utils.convert_bi_statistic_results_into_df(perf)
return output_df
def run_pipeline(model, tokenizer, possible_chars, sentences, number_of_trials, logger):
dataset = proc.prepare_truism_data_for_sentence_scoring(sentences,
possible_chars,
tokenizer,
number_of_trials)
logger.info("finished creating dataset")
perf = utils.generative_truism_reasoning_test(dataset, model, torch.cuda.is_available(), logger)
logger.info("finished evaluating dataset")
output_df = utils.convert_bi_statistic_results_into_df(perf)
return output_df
def run_pipeline(model, tokenizer, fictitious_entities, sentences, config,
number_of_entity_trials, logger):
dataset = proc.prepare_masked_instances(
sentences=sentences,
config=config,
fictitious_entities=fictitious_entities,
num_entity_trials=number_of_entity_trials)
logger.info("finished creating dataset")
perf = utils.happy_transformer_masked_word_prediction(
masked_examples=dataset, model=model, top_n=100, logger=logger)
logger.info("finished evaluating dataset")
output_df = utils.convert_bi_statistic_results_into_df(perf)
return output_df
