Update datasets_info & hf_path

dataset-index.yml

@@ -128,6 +128,24 @@
     paper: https://arxiv.org/abs/2501.18362
     configpath: opencompass/configs/datasets/MedXpertQA/MedXpertQA_gen.py
     configpath_llmjudge: opencompass/configs/datasets/MedXpertQA/MedXpertQA_llmjudge_gen.py
+- ClinicBench:
+    name: ClinicBench
+    category: Knowledge / Medicine
+    paper: https://arxiv.org/abs/2405.00716
+    configpath: ''
+    configpath_llmjudge: opencompass/configs/datasets/ClinicBench/ClinicBench_llmjudge_gen.py
+- ScienceQA:
+    name: ScienceQA
+    category: Knowledge / Medicine
+    paper: https://arxiv.org/abs/2209.09513
+    configpath: ''
+    configpath_llmjudge: opencompass/configs/datasets/ScienceQA/ScienceQA_llmjudge_gen.py
+- PubMedQA:
+    name: PubMedQA
+    category: Knowledge / Medicine
+    paper: https://arxiv.org/abs/1909.06146
+    configpath: ''
+    configpath_llmjudge: opencompass/configs/datasets/PubMedQA/PubMedQA_llmjudge_gen.py
 - musr:
     name: MuSR
     category: Reasoning

opencompass/configs/datasets/ClinicBench/ClinicBench_llmjudge_gen_d09668.py

@@ -45,7 +45,6 @@ ClinicBench_datasets = []
 ClinicBench_reader_cfg = dict(
     input_columns=['question', 'choices'],
     output_column='label',
-    test_split='validation',
 )
 
 ClinicBench_infer_cfg = dict(

opencompass/configs/datasets/PubMedQA/PubMedQA_llmjudge_gen_f00302.py

@@ -39,7 +39,6 @@ PubMedQA_datasets = []
 PubMedQA_reader_cfg = dict(
     input_columns=['question', 'choices'],
     output_column='label',
-    test_split='validation',
 )
 
 PubMedQA_infer_cfg = dict(

opencompass/configs/datasets/ScienceQA/ScienceQA_llmjudge_gen_f00302.py

@@ -39,7 +39,6 @@ ScienceQA_datasets = []
 ScienceQA_reader_cfg = dict(
     input_columns=['question', 'choices'],
     output_column='label',
-    test_split='validation',
 )
 
 ScienceQA_infer_cfg = dict(

opencompass/datasets/ClinicBench.py

@@ -1,6 +1,7 @@
-from datasets import Dataset, DatasetDict, load_dataset
+from datasets import load_dataset
 
 from opencompass.registry import LOAD_DATASET
+from opencompass.utils import get_data_path
 
 from .base import BaseDataset
 
@@ -9,16 +10,12 @@ from .base import BaseDataset
 class ClinicBenchDataset(BaseDataset):
 
     @staticmethod
-    def load_single():
-        dataset = load_dataset('xuxuxuxuxu/Pharmacology-QA')['train']
+    def load_single(path):
+        dataset = load_dataset(path)['train']
         return dataset
 
     @staticmethod
     def load(path):
-        train_dataset = Dataset.from_list([])
-        val_dataset = ClinicBenchDataset.load_single()
-        dataset = DatasetDict({
-            'train': train_dataset,
-            'validation': val_dataset
-        })
+        path = get_data_path(path)
+        dataset = ClinicBenchDataset.load_single(path)
         return dataset

opencompass/datasets/PubMedQA.py

@@ -1,6 +1,7 @@
-from datasets import Dataset, DatasetDict, load_dataset
+from datasets import Dataset, load_dataset
 
 from opencompass.registry import LOAD_DATASET
+from opencompass.utils import get_data_path
 
 from .base import BaseDataset
 
@@ -9,9 +10,9 @@ from .base import BaseDataset
 class PubMedQADataset(BaseDataset):
 
     @staticmethod
-    def load_single():
+    def load_single(path):
         dataset = []
-        ds = load_dataset('qiaojin/PubMedQA', 'pqa_labeled')
+        ds = load_dataset(path, 'pqa_labeled')
         for data in ds['train']:
             data['question'] = (f"CONTEXTS: {data['context']}\n"
                                 f"QUESTION: {data['question']}")
@@ -30,10 +31,6 @@ class PubMedQADataset(BaseDataset):
 
     @staticmethod
     def load(path):
-        train_dataset = Dataset.from_list([])
-        val_dataset = PubMedQADataset.load_single()
-        dataset = DatasetDict({
-            'train': train_dataset,
-            'validation': val_dataset
-        })
+        path = get_data_path(path)
+        dataset = PubMedQADataset.load_single(path)
         return dataset

opencompass/datasets/ScienceQA.py

@@ -1,6 +1,7 @@
-from datasets import Dataset, DatasetDict, load_dataset
+from datasets import Dataset, load_dataset
 
 from opencompass.registry import LOAD_DATASET
+from opencompass.utils import get_data_path
 
 from .base import BaseDataset
 
@@ -9,9 +10,9 @@ from .base import BaseDataset
 class ScienceQADataset(BaseDataset):
 
     @staticmethod
-    def load_single():
+    def load_single(path):
         dataset = []
-        ds = load_dataset('derek-thomas/ScienceQA')
+        ds = load_dataset(path)
         for data in ds['test']:
             if data['image'] is None:
                 data['label'] = chr(65 + data['answer']
@@ -28,10 +29,6 @@ class ScienceQADataset(BaseDataset):
 
     @staticmethod
     def load(path):
-        train_dataset = Dataset.from_list([])
-        val_dataset = ScienceQADataset.load_single()
-        dataset = DatasetDict({
-            'train': train_dataset,
-            'validation': val_dataset
-        })
+        path = get_data_path(path)
+        dataset = ScienceQADataset.load_single(path)
         return dataset

opencompass/utils/datasets_info.py

@@ -446,6 +446,21 @@ DATASETS_MAPPING = {
         "hf_id": "",
         "local": "./data/ChemBench4K",
     },
+     "opencompass/ClinicBench": {
+        "ms_id": "",
+        "hf_id": "xuxuxuxuxu/Pharmacology-QA",
+        "local": "",
+    },
+     "opencompass/ScienceQA": {
+        "ms_id": "",
+        "hf_id": "derek-thomas/ScienceQA",
+        "local": "",
+    },
+     "opencompass/PubMedQA": {
+        "ms_id": "",
+        "hf_id": "qiaojin/PubMedQA",
+        "local": "",
+    },
 
 }