Not able to bring lambada templates
dayeonki opened this issue · comments
Hello, I am running bigscience-workshop/bigscience-lm-evaluation-harness.
I wanted to add new task: LAMBADA.
I respected and followed the "new_prompt_source_task" to create one.
But I am keep getting this kind of error message:
raise ValueError(f"No original task templates found for {task_name}")
ValueError: No original task templates found for lambada
task > lambada.py is defined as:
class LAMBADA(PromptSourceTask):
VERSION = 0
DATASET_PATH = inspect.getfile(lm_eval.datasets.lambada.lambada)
DATASET_NAME = None
def has_training_docs(self):
return False
def has_validation_docs(self):
return True
def has_test_docs(self):
return False
def training_docs(self):
pass
def validation_docs(self):
return self.dataset["validation"]
def test_docs(self):
pass
def doc_to_text(self, doc):
return doc["text"].rsplit(" ", 1)[0]
def should_decontaminate(self):
return True
def doc_to_decontamination_query(self, doc):
return doc["text"]
def doc_to_target(self, doc):
return " " + doc["text"].rsplit(" ", 1)[1]
def construct_requests(self, doc, ctx):
ll, is_greedy = rf.loglikelihood(ctx, self.doc_to_target(doc))
return ll, is_greedy
def process_results(self, doc, results):
ll, is_greedy = results
return {"ppl": ll, "acc": int(is_greedy)}
def aggregation(self):
return {"ppl": perplexity, "acc": mean}
def higher_is_better(self):
return {"ppl": False, "acc": True}
Hi @dayeonki, this might be better posted in bigscience-lm-evaluation-harness, but I can try to help with debugging. Can you post the entire backtrace? You could also try to load LAMBADA in promptsource following the example in the promptsource README.