[Feature] Use dataset in local path (#570)

* update commonsenseqa

* update drop

* update flores_first100

* update gsm8k

* update humaneval

* update lambda

* update obqa

* update piqa

* update race

* update siqa

* update story_cloze

* update strategyqa

* update tydiqa

* update winogrande

* update doc

* update hellaswag

* fix obqa

* update collections

* update .zip name

configs/datasets/race/race_ppl_a138cd.py

@@ -6,7 +6,10 @@ from opencompass.datasets import RaceDataset
 
 race_reader_cfg = dict(
     input_columns=['article', 'question', 'A', 'B', 'C', 'D'],
-    output_column='answer')
+    output_column='answer',
+    train_split="validation",
+    test_split="test"
+)
 
 race_infer_cfg = dict(
     prompt_template=dict(
@@ -29,17 +32,17 @@ race_eval_cfg = dict(evaluator=dict(type=AccEvaluator))
 
 race_datasets = [
     dict(
-        type=RaceDataset,
         abbr='race-middle',
-        path='race',
+        type=RaceDataset,
+        path='./data/race',
         name='middle',
         reader_cfg=race_reader_cfg,
         infer_cfg=race_infer_cfg,
         eval_cfg=race_eval_cfg),
     dict(
-        type=RaceDataset,
         abbr='race-high',
-        path='race',
+        type=RaceDataset,
+        path='./data/race',
         name='high',
         reader_cfg=race_reader_cfg,
         infer_cfg=race_infer_cfg,

configs/datasets/race/race_ppl_abed12.py

@@ -6,7 +6,10 @@ from opencompass.datasets import RaceDataset
 
 race_reader_cfg = dict(
     input_columns=['article', 'question', 'A', 'B', 'C', 'D'],
-    output_column='answer')
+    output_column='answer',
+    train_split="validation",
+    test_split="test"
+)
 
 race_infer_cfg = dict(
     prompt_template=dict(
@@ -28,17 +31,17 @@ race_eval_cfg = dict(evaluator=dict(type=AccEvaluator))
 
 race_datasets = [
     dict(
-        type=RaceDataset,
         abbr='race-middle',
-        path='race',
+        type=RaceDataset,
+        path='./data/race',
         name='middle',
         reader_cfg=race_reader_cfg,
         infer_cfg=race_infer_cfg,
         eval_cfg=race_eval_cfg),
     dict(
-        type=RaceDataset,
         abbr='race-high',
-        path='race',
+        type=RaceDataset,
+        path='./data/race',
         name='high',
         reader_cfg=race_reader_cfg,
         infer_cfg=race_infer_cfg,

configs/datasets/siqa/siqa_gen_e78df3.py

@@ -34,7 +34,7 @@ siqa_datasets = [
     dict(
         abbr="siqa",
         type=siqaDataset_V2,
-        path="social_i_qa",
+        path='./data/siqa',
         reader_cfg=siqa_reader_cfg,
         infer_cfg=siqa_infer_cfg,
         eval_cfg=siqa_eval_cfg)

configs/datasets/siqa/siqa_ppl_42bc6e.py

@@ -2,7 +2,7 @@ from opencompass.openicl.icl_prompt_template import PromptTemplate
 from opencompass.openicl.icl_retriever import ZeroRetriever
 from opencompass.openicl.icl_inferencer import PPLInferencer
 from opencompass.openicl.icl_evaluator import AccEvaluator
-from opencompass.datasets import HFDataset
+from opencompass.datasets import siqaDataset
 
 siqa_reader_cfg = dict(
     input_columns=['context', 'question', 'answerA', 'answerB', 'answerC'],
@@ -25,8 +25,8 @@ siqa_eval_cfg = dict(evaluator=dict(type=AccEvaluator))
 siqa_datasets = [
     dict(
         abbr="siqa",
-        type=HFDataset,
-        path='social_i_qa',
+        type=siqaDataset,
+        path='./data/siqa',
         reader_cfg=siqa_reader_cfg,
         infer_cfg=siqa_infer_cfg,
         eval_cfg=siqa_eval_cfg)

configs/datasets/siqa/siqa_ppl_7845b0.py

@@ -2,7 +2,7 @@ from opencompass.openicl.icl_prompt_template import PromptTemplate
 from opencompass.openicl.icl_retriever import ZeroRetriever
 from opencompass.openicl.icl_inferencer import PPLInferencer
 from opencompass.openicl.icl_evaluator import AccEvaluator
-from opencompass.datasets import HFDataset
+from opencompass.datasets import siqaDataset
 
 siqa_reader_cfg = dict(
     input_columns=['context', 'question', 'answerA', 'answerB', 'answerC'],
@@ -25,9 +25,8 @@ siqa_eval_cfg = dict(evaluator=dict(type=AccEvaluator))
 siqa_datasets = [
     dict(
         abbr="siqa",
-        type=HFDataset,
-        path='social_i_qa',
-        name='social_i_qa',
+        type=siqaDataset,
+        path='./data/siqa',
         reader_cfg=siqa_reader_cfg,
         infer_cfg=siqa_infer_cfg,
         eval_cfg=siqa_eval_cfg)

configs/datasets/siqa/siqa_ppl_ced5f6.py

@@ -2,7 +2,7 @@ from opencompass.openicl.icl_prompt_template import PromptTemplate
 from opencompass.openicl.icl_retriever import ZeroRetriever
 from opencompass.openicl.icl_inferencer import PPLInferencer
 from opencompass.openicl.icl_evaluator import AccEvaluator
-from opencompass.datasets import HFDataset
+from opencompass.datasets import siqaDataset
 
 siqa_reader_cfg = dict(
     input_columns=['context', 'question', 'answerA', 'answerB', 'answerC'],
@@ -37,8 +37,8 @@ siqa_eval_cfg = dict(evaluator=dict(type=AccEvaluator))
 siqa_datasets = [
     dict(
         abbr="siqa",
-        type=HFDataset,
-        path='social_i_qa',
+        type=siqaDataset,
+        path='./data/siqa',
         reader_cfg=siqa_reader_cfg,
         infer_cfg=siqa_infer_cfg,
         eval_cfg=siqa_eval_cfg)

configs/datasets/siqa/siqa_ppl_e8d8c5.py

@@ -2,7 +2,7 @@ from opencompass.openicl.icl_prompt_template import PromptTemplate
 from opencompass.openicl.icl_retriever import ZeroRetriever
 from opencompass.openicl.icl_inferencer import PPLInferencer
 from opencompass.openicl.icl_evaluator import AccEvaluator
-from opencompass.datasets import HFDataset
+from opencompass.datasets import siqaDataset
 
 siqa_reader_cfg = dict(
     input_columns=['context', 'question', 'answerA', 'answerB', 'answerC'],
@@ -37,8 +37,8 @@ siqa_eval_cfg = dict(evaluator=dict(type=AccEvaluator))
 siqa_datasets = [
     dict(
         abbr="siqa",
-        type=HFDataset,
-        path='social_i_qa',
+        type=siqaDataset,
+        path='./data/siqa',
         reader_cfg=siqa_reader_cfg,
         infer_cfg=siqa_infer_cfg,
         eval_cfg=siqa_eval_cfg)

configs/datasets/storycloze/storycloze_gen_7f656a.py

@@ -37,8 +37,8 @@ storycloze_datasets = [
     dict(
         abbr="story_cloze",
         type=storyclozeDataset_V2,
-        path="juletxara/xstory_cloze",
-        name="en",
+        path='./data/xstory_cloze',
+        lang='en',
         reader_cfg=storycloze_reader_cfg,
         infer_cfg=storycloze_infer_cfg,
         eval_cfg=storycloze_eval_cfg,

configs/datasets/storycloze/storycloze_ppl_496661.py

@@ -31,8 +31,8 @@ storycloze_datasets = [
     dict(
         abbr='story_cloze',
         type=storyclozeDataset,
-        path='juletxara/xstory_cloze',
-        name='en',
+        path='./data/xstory_cloze',
+        lang='en',
         reader_cfg=storycloze_reader_cfg,
         infer_cfg=storycloze_infer_cfg,
         eval_cfg=storycloze_eval_cfg)

configs/datasets/storycloze/storycloze_ppl_afd16f.py

@@ -28,8 +28,8 @@ storycloze_datasets = [
     dict(
         abbr='story_cloze',
         type=storyclozeDataset,
-        path='juletxara/xstory_cloze',
-        name='en',
+        path='./data/xstory_cloze',
+        lang='en',
         reader_cfg=storycloze_reader_cfg,
         infer_cfg=storycloze_infer_cfg,
         eval_cfg=storycloze_eval_cfg)

configs/datasets/strategyqa/strategyqa_gen_1180a7.py

@@ -2,7 +2,7 @@ from opencompass.openicl.icl_prompt_template import PromptTemplate
 from opencompass.openicl.icl_retriever import ZeroRetriever
 from opencompass.openicl.icl_inferencer import GenInferencer
 from opencompass.openicl.icl_evaluator import AccEvaluator
-from opencompass.datasets import HFDataset, strategyqa_pred_postprocess, strategyqa_dataset_postprocess
+from opencompass.datasets import StrategyQADataset, strategyqa_pred_postprocess, strategyqa_dataset_postprocess
 
 strategyqa_reader_cfg = dict(
     input_columns=['question'],
@@ -86,8 +86,8 @@ strategyqa_eval_cfg = dict(
 strategyqa_datasets = [
     dict(
         abbr='strategyqa',
-        type=HFDataset,
-        path='wics/strategy-qa',
+        type=StrategyQADataset,
+        path='./data/strategyqa/strategyQA_train.json',
         reader_cfg=strategyqa_reader_cfg,
         infer_cfg=strategyqa_infer_cfg,
         eval_cfg=strategyqa_eval_cfg)

configs/datasets/strategyqa/strategyqa_gen_934441.py

@@ -2,7 +2,7 @@ from opencompass.openicl.icl_prompt_template import PromptTemplate
 from opencompass.openicl.icl_retriever import ZeroRetriever
 from opencompass.openicl.icl_inferencer import GenInferencer
 from opencompass.openicl.icl_evaluator import AccEvaluator
-from opencompass.datasets import HFDataset, strategyqa_pred_postprocess, strategyqa_dataset_postprocess
+from opencompass.datasets import StrategyQADataset, strategyqa_pred_postprocess, strategyqa_dataset_postprocess
 
 strategyqa_reader_cfg = dict(
     input_columns=['question'],
@@ -50,8 +50,8 @@ strategyqa_eval_cfg = dict(
 strategyqa_datasets = [
     dict(
         abbr='strategyqa',
-        type=HFDataset,
-        path='wics/strategy-qa',
+        type=StrategyQADataset,
+        path='./data/strategyqa/strategyQA_train.json',
         reader_cfg=strategyqa_reader_cfg,
         infer_cfg=strategyqa_infer_cfg,
         eval_cfg=strategyqa_eval_cfg)

configs/datasets/tydiqa/tydiqa_gen_978d2a.py

@@ -6,9 +6,8 @@ from opencompass.datasets import TydiQADataset, TydiQAEvaluator
 # All configs are for TydiQA Goldp task
 tydiqa_reader_cfg = dict(
     input_columns=["passage_text", "question_text"],
-    output_column="answer",
-    test_split='validation',
-    train_split='validation',)
+    output_column="answer"
+)
 
 langs = ['arabic', 'bengali', 'english', 'finnish', 'indonesian', 'japanese', 'korean', 'russian', 'swahili', 'telugu', 'thai']
 
@@ -33,19 +32,25 @@ for _lang in langs:
         prompt_template=dict(
             type=PromptTemplate,
             template=f"{_hint[0]}\n\n</E>{_hint[1]}{{passage_text}}\n{_hint[2]} {{question_text}}\n{_hint[3]} {{answer}}" ,
-            ice_token='</E>'),
+            ice_token='</E>'
+        ),
         retriever=dict(type=ZeroRetriever),
-        inferencer=dict(type=GenInferencer), max_out_len=50)
+        inferencer=dict(type=GenInferencer), max_out_len=50
+    )
+
+    tydiqa_eval_cfg = dict(
+        evaluator=dict(type=TydiQAEvaluator),
+        ds_split='validation',
+        ds_column='answer',
+    )
 
-    tydiqa_eval_cfg = dict(evaluator=dict(type=TydiQAEvaluator),
-                        ds_split='validation',
-                        ds_column='answer',
-                        )
     tydiqa_datasets.append(
-    dict(abbr=f'tyidqa-goldp_{_lang}',
-        type=TydiQADataset,
-        path='khalidalt/tydiqa-goldp',
-        name=_lang,
-        reader_cfg=tydiqa_reader_cfg,
-        infer_cfg=tydiqa_infer_cfg,
-        eval_cfg=tydiqa_eval_cfg))
+        dict(abbr=f'tyidqa-goldp_{_lang}',
+            type=TydiQADataset,
+            path='./data/tydiqa',
+            lang=_lang,
+            reader_cfg=tydiqa_reader_cfg,
+            infer_cfg=tydiqa_infer_cfg,
+            eval_cfg=tydiqa_eval_cfg
\ No newline at end of file
+        )
+    )

configs/datasets/winogrande/winogrande_gen_a9ede5.py

@@ -7,8 +7,8 @@ from opencompass.utils.text_postprocessors import first_option_postprocess
 
 winogrande_reader_cfg = dict(
     input_columns=["opt1", "opt2"],
-    output_column="label",
-    test_split="validation")
+    output_column="answer",
+)
 
 winogrande_infer_cfg = dict(
     prompt_template=dict(
@@ -35,8 +35,7 @@ winogrande_datasets = [
     dict(
         abbr="winogrande",
         type=winograndeDataset_V2,
-        path="winogrande",
-        name="winogrande_xs",
+        path='./data/winogrande',
         reader_cfg=winogrande_reader_cfg,
         infer_cfg=winogrande_infer_cfg,
         eval_cfg=winogrande_eval_cfg,

configs/datasets/winogrande/winogrande_ppl_55a66e.py

@@ -7,8 +7,7 @@ from opencompass.datasets import winograndeDataset
 winogrande_reader_cfg = dict(
     input_columns=['opt1', 'opt2'],
     output_column='answer',
-    train_split='validation',
-    test_split='validation')
+)
 
 winogrande_infer_cfg = dict(
     prompt_template=dict(
@@ -28,8 +27,7 @@ winogrande_datasets = [
     dict(
         abbr='winogrande',
         type=winograndeDataset,
-        path='winogrande',
-        name='winogrande_xs',
+        path='./data/winogrande',
         reader_cfg=winogrande_reader_cfg,
         infer_cfg=winogrande_infer_cfg,
         eval_cfg=winogrande_eval_cfg)

configs/datasets/winogrande/winogrande_ppl_9307fd.py

@@ -7,8 +7,7 @@ from opencompass.datasets import winograndeDataset
 winogrande_reader_cfg = dict(
     input_columns=['opt1', 'opt2'],
     output_column='answer',
-    train_split='validation',
-    test_split='validation')
+)
 
 winogrande_infer_cfg = dict(
     prompt_template=dict(
@@ -26,8 +25,7 @@ winogrande_datasets = [
     dict(
         abbr='winogrande',
         type=winograndeDataset,
-        path='winogrande',
-        name='winogrande_xs',
+        path='./data/winogrande',
         reader_cfg=winogrande_reader_cfg,
         infer_cfg=winogrande_infer_cfg,
         eval_cfg=winogrande_eval_cfg)

configs/summarizers/medium.py

@@ -87,17 +87,6 @@ summarizer = dict(
         'eprstmt-dev',
         'lambada',
         'tnews-dev',
-        '--------- 安全 Safety ---------',  # category
-        # '偏见', # subcategory
-        'crows_pairs',
-        # '有毒性（判别）', # subcategory
-        'civil_comments',
-        # '有毒性（判别）多语言', # subcategory
-        'jigsaw_multilingual',
-        # '有毒性（生成）', # subcategory
-        'real-toxicity-prompts',
-        # '真实性/有用性', # subcategory
-        'truthful_qa',
     ],
     summary_groups=sum(
         [v for k, v in locals().items() if k.endswith("_summary_groups")], []),

configs/summarizers/small.py

@@ -56,8 +56,6 @@ summarizer = dict(
         'openbookqa_fact',
         'nq',
         'triviaqa',
-        '--- Security ---',
-        'crows_pairs',
     ],
     summary_groups=sum([v for k, v in locals().items() if k.endswith("_summary_groups")], []),
 )

docs/en/get_started/installation.md

@@ -66,10 +66,21 @@ Run the following commands to download and place the datasets in the `${OpenComp
 
 ```bash
 # Run in the OpenCompass directory
-wget https://github.com/open-compass/opencompass/releases/download/0.1.1/OpenCompassData.zip
-unzip OpenCompassData.zip
+wget https://github.com/open-compass/opencompass/releases/download/0.1.8.rc1/OpenCompassData-core-20231110.zip
+unzip OpenCompassData-core-20231110.zip
 ```
 
+If you need to use the more comprehensive dataset (~500M) provided by OpenCompass, You can download it using the following command:
+
+```bash
+wget https://github.com/open-compass/opencompass/releases/download/0.1.8.rc1/OpenCompassData-complete-20231110.zip
+unzip OpenCompassData-complete-20231110.zip
+cd ./data
+unzip *.zip
+```
+
+The list of datasets included in both `.zip` can be found [here](https://github.com/open-compass/opencompass/releases/tag/0.1.8.rc1)
+
 OpenCompass has supported most of the datasets commonly used for performance comparison, please refer to `configs/dataset` for the specific list of supported datasets.
 
 For next step, please read [Quick Start](./quick_start.md).

docs/zh_cn/get_started/installation.md

@@ -66,10 +66,21 @@ OpenCompass 支持的数据集主要包括两个部分：
 在 OpenCompass 项目根目录下运行下面命令，将数据集准备至 `${OpenCompass}/data` 目录下：
 
 ```bash
-wget https://github.com/open-compass/opencompass/releases/download/0.1.1/OpenCompassData.zip
-unzip OpenCompassData.zip
+wget https://github.com/open-compass/opencompass/releases/download/0.1.8.rc1/OpenCompassData-core-20231110.zip
+unzip OpenCompassData-core-20231110.zip
 ```
 
+如果需要使用 OpenCompass 提供的更加完整的数据集 (~500M)，可以使用下述命令进行下载：
+
+```bash
+wget https://github.com/open-compass/opencompass/releases/download/0.1.8.rc1/OpenCompassData-complete-20231110.zip
+unzip OpenCompassData-complete-20231110.zip
+cd ./data
+unzip *.zip
+```
+
+两个 `.zip` 中所含数据集列表如[此处](https://github.com/open-compass/opencompass/releases/tag/0.1.8.rc1)所示。
+
 OpenCompass 已经支持了大多数常用于性能比较的数据集，具体支持的数据集列表请直接在 `configs/datasets` 下进行查找。
 
 接下来，你可以阅读[快速上手](./quick_start.md)了解 OpenCompass 的基本用法。