L2-1 Compass评测书生大模型实践

C-Eval数据集的评测

• 运行时遇到如下报错：

returnnext(self.gen)File"/root/.conda/envs/opencompass10/lib/python3.10/site-packages/modelscope/msdatasets/utils/hf_datasets_util.py", line1492,inload_dataset_with_ctx dataset_res=DatasetsWrapperHF.load_dataset(*args, **kwargs)File"/root/.conda/envs/opencompass10/lib/python3.10/site-packages/modelscope/msdatasets/utils/hf_datasets_util.py", line1008,inload_dataset builder_instance=DatasetsWrapperHF.load_dataset_builder(File"/root/.conda/envs/opencompass10/lib/python3.10/site-packages/modelscope/msdatasets/utils/hf_datasets_util.py", line1142,inload_dataset_builder dataset_module=DatasetsWrapperHF.dataset_module_factory(File"/root/.conda/envs/opencompass10/lib/python3.10/site-packages/modelscope/msdatasets/utils/hf_datasets_util.py", line1447,indataset_module_factory raise e1 from None File"/root/.conda/envs/opencompass10/lib/python3.10/site-packages/modelscope/msdatasets/utils/hf_datasets_util.py", line1401,indataset_module_factory).get_module()File"/root/.conda/envs/opencompass10/lib/python3.10/site-packages/modelscope/msdatasets/utils/hf_datasets_util.py", line914,inget_module_with_script raise ValueError(ValueError: Loading opencompass/ceval-exam requires you to execute the dataset scriptinthat repo on yourlocalmachine. Make sure you havereadthe code there to avoid malicious use,thensetthe optiontrust_remote_code=True to remove this error.

判断是参数的问题，通过修改若干如下位置的代码：

把trust_remote_code = trust_remote_code 改成如下=True，完成运行

运行结果

自建数据集测评

运行结果

多模态数据集测评