-
Notifications
You must be signed in to change notification settings - Fork 136
Open
Labels
bugSomething isn't workingSomething isn't working
Description
System Info
No available worker for model 'QwenQwen2.5-Math-PRM-7B'. Controller response: {'address': ''}
in this code value_list = self.rm_call(input_list, lm_step_tag=self.lm_call.lm_step_tag)
Who can help?
No response
Information
- The official example scripts
- My own modified scripts
Tasks
- An officially supported task in the codebase (such as scrips/, ...)
- My own task or dataset (give details below)
Reproduction
def analyze_output(self, problem_inst: Dict[str, str], gen_answers: List[str]):
extracted_groundtruth = self._task.extract_groundtruth(problem_inst["answer"])
if len(gen_answers) > 1:
input_list = [(problem_inst["question"], txt) for txt in gen_answers]
# XXX(ziyu): for tree search methods with value_fn, should not call rm
# to compute it again
value_list = self.rm_call(input_list, lm_step_tag=self.lm_call.lm_step_tag)
else:
value_list = [[0]]
output_list = [
{"path_idx": i, "text": txt, "value": v}
for i, (txt, v) in enumerate(zip(gen_answers, value_list))
]
res = {
agg_method: judge_ans(
problem_inst["question"],
extracted_groundtruth,
gen_answers,
value_list,
agg_method,
self._task.extract_answer,
self._task.judge_correct,
)
for agg_method in (
CHOSEN_AGGR_METHODS if len(gen_answers) > 1 else [MAJORITY_VOTE]
)
}
return res, output_list
Expected behavior
please help me
Metadata
Metadata
Assignees
Labels
bugSomething isn't workingSomething isn't working
