[Feature] re-implement ceval load dataset (#446)

configs/summarizers/groups/ceval.py

@@ -22,3 +22,26 @@ ceval_summary_groups.append({'name': 'ceval-hard', 'subsets': _ceval_hard})
 
 _ceval_all = _ceval_stem + _ceval_social_science + _ceval_humanities + _ceval_other
 ceval_summary_groups.append({'name': 'ceval', 'subsets': _ceval_all})
+
+_ceval_stem = ['computer_network', 'operating_system', 'computer_architecture', 'college_programming', 'college_physics', 'college_chemistry', 'advanced_mathematics', 'probability_and_statistics', 'discrete_mathematics', 'electrical_engineer', 'metrology_engineer', 'high_school_mathematics', 'high_school_physics', 'high_school_chemistry', 'high_school_biology', 'middle_school_mathematics', 'middle_school_biology', 'middle_school_physics', 'middle_school_chemistry', 'veterinary_medicine']
+_ceval_stem = ['ceval-test-' + s for s in _ceval_stem]
+ceval_summary_groups.append({'name': 'ceval-test-stem', 'subsets': _ceval_stem})
+
+_ceval_social_science = ['college_economics', 'business_administration', 'marxism', 'mao_zedong_thought', 'education_science', 'teacher_qualification', 'high_school_politics', 'high_school_geography', 'middle_school_politics', 'middle_school_geography']
+_ceval_social_science = ['ceval-test-' + s for s in _ceval_social_science]
+ceval_summary_groups.append({'name': 'ceval-test-social-science', 'subsets': _ceval_social_science})
+
+_ceval_humanities = ['modern_chinese_history', 'ideological_and_moral_cultivation', 'logic', 'law', 'chinese_language_and_literature', 'art_studies', 'professional_tour_guide', 'legal_professional', 'high_school_chinese', 'high_school_history', 'middle_school_history']
+_ceval_humanities = ['ceval-test-' + s for s in _ceval_humanities]
+ceval_summary_groups.append({'name': 'ceval-test-humanities', 'subsets': _ceval_humanities})
+
+_ceval_other = ['civil_servant', 'sports_science', 'plant_protection', 'basic_medicine', 'clinical_medicine', 'urban_and_rural_planner', 'accountant', 'fire_engineer', 'environmental_impact_assessment_engineer', 'tax_accountant', 'physician']
+_ceval_other = ['ceval-test-' + s for s in _ceval_other]
+ceval_summary_groups.append({'name': 'ceval-test-other', 'subsets': _ceval_other})
+
+_ceval_hard = ['advanced_mathematics', 'discrete_mathematics', 'probability_and_statistics', 'college_chemistry', 'college_physics', 'high_school_mathematics', 'high_school_chemistry', 'high_school_physics']
+_ceval_hard = ['ceval-test-' + s for s in _ceval_hard]
+ceval_summary_groups.append({'name': 'ceval-test-hard', 'subsets': _ceval_hard})
+
+_ceval_all = _ceval_stem + _ceval_social_science + _ceval_humanities + _ceval_other
+ceval_summary_groups.append({'name': 'ceval-test', 'subsets': _ceval_all})

opencompass/datasets/ceval.py

+import csv
 import os.path as osp
 
-from datasets import DatasetDict, load_dataset
+from datasets import Dataset, DatasetDict
 
 from opencompass.registry import LOAD_DATASET
 
@@ -12,26 +13,15 @@ class CEvalDataset(BaseDataset):
 
     @staticmethod
     def load(path: str, name: str):
-        dev_dataset = load_dataset('csv',
-                                   data_files=osp.join(path, 'dev',
-                                                       f'{name}_dev.csv'),
-                                   split='train')
-        val_dataset = load_dataset('csv',
-                                   data_files=osp.join(path, 'val',
-                                                       f'{name}_val.csv'),
-                                   split='train')
-        val_dataset = val_dataset.add_column('explanation',
-                                             [''] * len(val_dataset))
-        test_dataset = load_dataset('csv',
-                                    data_files=osp.join(
-                                        path, 'test', f'{name}_test.csv'),
-                                    split='train')
-        test_dataset = test_dataset.add_column(
-            'answer',
-            [''] * len(test_dataset)).add_column('explanation',
-                                                 [''] * len(test_dataset))
-        return DatasetDict({
-            'val': val_dataset,
-            'dev': dev_dataset,
-            'test': test_dataset
-        })
+        dataset = {}
+        for split in ['dev', 'val', 'test']:
+            with open(osp.join(path, split, f'{name}_{split}.csv')) as f:
+                reader = csv.reader(f)
+                header = next(reader)
+                for row in reader:
+                    item = dict(zip(header, row))
+                    item.setdefault('explanation', '')
+                    item.setdefault('answer', '')
+                    dataset.setdefault(split, []).append(item)
+        dataset = {i: Dataset.from_list(dataset[i]) for i in dataset}
+        return DatasetDict(dataset)