ValueError: Need either a `state_dict` or a `save_folder` containing offloaded weights.

Mac with M1 Pro

`from pandalm import EvaluationPipeline
pipeline = EvaluationPipeline(
    candidate_paths=["hfl/llama-3-chinese-8b-instruct-v2", "hfl/llama-3-chinese-8b-instruct-v3"],
    input_data_path="data/pipeline-sanity-check.json",
)
print(pipeline.evaluate())`

ValueError: Need either a `state_dict` or a `save_folder` containing offloaded weights.