evaluation/ceval/ceval.py

@@ -0,0 +1,163 @@
+# Copyright 2025 the LlamaFactory team.
+# Copyright 2020 The HuggingFace Datasets Authors and the current dataset script contributor.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import os
+
+import datasets
+import pandas as pd
+
+
+_CITATION = """\
+@article{huang2023ceval,
+  title={C-Eval: A Multi-Level Multi-Discipline Chinese Evaluation Suite for Foundation Models},
+  author={Huang, Yuzhen and Bai, Yuzhuo and Zhu, Zhihao and others},
+  journal={arXiv preprint arXiv:2305.08322},
+  year={2023}
+}
+"""
+
+_DESCRIPTION = """\
+C-Eval is a comprehensive Chinese evaluation suite for foundation models.
+It consists of 13948 multi-choice questions spanning 52 diverse disciplines and four difficulty levels.
+"""
+
+_HOMEPAGE = "https://cevalbenchmark.com"
+
+_LICENSE = "Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License"
+
+_URL = "ceval.zip"
+
+task_list = [
+    "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",
+    "college_economics",
+    "business_administration",
+    "marxism",
+    "mao_zedong_thought",
+    "education_science",
+    "teacher_qualification",
+    "high_school_politics",
+    "high_school_geography",
+    "middle_school_politics",
+    "middle_school_geography",
+    "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",
+    "civil_servant",
+    "sports_science",
+    "plant_protection",
+    "basic_medicine",
+    "clinical_medicine",
+    "urban_and_rural_planner",
+    "accountant",
+    "fire_engineer",
+    "environmental_impact_assessment_engineer",
+    "tax_accountant",
+    "physician",
+]
+
+
+class CevalConfig(datasets.BuilderConfig):
+    def __init__(self, **kwargs):
+        super().__init__(version=datasets.Version("1.0.0"), **kwargs)
+
+
+class Ceval(datasets.GeneratorBasedBuilder):
+    BUILDER_CONFIGS = [
+        CevalConfig(
+            name=task_name,
+        )
+        for task_name in task_list
+    ]
+
+    def _info(self):
+        features = datasets.Features(
+            {
+                "id": datasets.Value("int32"),
+                "question": datasets.Value("string"),
+                "A": datasets.Value("string"),
+                "B": datasets.Value("string"),
+                "C": datasets.Value("string"),
+                "D": datasets.Value("string"),
+                "answer": datasets.Value("string"),
+                "explanation": datasets.Value("string"),
+            }
+        )
+        return datasets.DatasetInfo(
+            description=_DESCRIPTION,
+            features=features,
+            homepage=_HOMEPAGE,
+            license=_LICENSE,
+            citation=_CITATION,
+        )
+
+    def _split_generators(self, dl_manager):
+        data_dir = dl_manager.download_and_extract(_URL)
+        task_name = self.config.name
+        return [
+            datasets.SplitGenerator(
+                name=datasets.Split.TEST,
+                gen_kwargs={
+                    "filepath": os.path.join(data_dir, "test", f"{task_name}_test.csv"),
+                },
+            ),
+            datasets.SplitGenerator(
+                name=datasets.Split.VALIDATION,
+                gen_kwargs={
+                    "filepath": os.path.join(data_dir, "val", f"{task_name}_val.csv"),
+                },
+            ),
+            datasets.SplitGenerator(
+                name=datasets.Split.TRAIN,
+                gen_kwargs={
+                    "filepath": os.path.join(data_dir, "dev", f"{task_name}_dev.csv"),
+                },
+            ),
+        ]
+
+    def _generate_examples(self, filepath):
+        df = pd.read_csv(filepath, encoding="utf-8")
+        for i, instance in enumerate(df.to_dict(orient="records")):
+            if "answer" not in instance.keys():
+                instance["answer"] = ""
+            if "explanation" not in instance.keys():
+                instance["explanation"] = ""
+            yield i, instance

evaluation/ceval/mapping.json

@@ -0,0 +1,210 @@
+{
+  "accountant": {
+    "name": "注册会计师",
+    "category": "Other"
+  },
+  "advanced_mathematics": {
+    "name": "高等数学",
+    "category": "STEM"
+  },
+  "art_studies": {
+    "name": "艺术学",
+    "category": "Humanities"
+  },
+  "basic_medicine": {
+    "name": "基础医学",
+    "category": "Other"
+  },
+  "business_administration": {
+    "name": "工商管理",
+    "category": "Social Sciences"
+  },
+  "chinese_language_and_literature": {
+    "name": "中国语言文学",
+    "category": "Humanities"
+  },
+  "civil_servant": {
+    "name": "公务员",
+    "category": "Other"
+  },
+  "clinical_medicine": {
+    "name": "临床医学",
+    "category": "Other"
+  },
+  "college_chemistry": {
+    "name": "大学化学",
+    "category": "STEM"
+  },
+  "college_economics": {
+    "name": "大学经济学",
+    "category": "Social Sciences"
+  },
+  "college_physics": {
+    "name": "大学物理",
+    "category": "STEM"
+  },
+  "college_programming": {
+    "name": "大学编程",
+    "category": "STEM"
+  },
+  "computer_architecture": {
+    "name": "计算机组成",
+    "category": "STEM"
+  },
+  "computer_network": {
+    "name": "计算机网络",
+    "category": "STEM"
+  },
+  "discrete_mathematics": {
+    "name": "离散数学",
+    "category": "STEM"
+  },
+  "education_science": {
+    "name": "教育学",
+    "category": "Social Sciences"
+  },
+  "electrical_engineer": {
+    "name": "注册电气工程师",
+    "category": "STEM"
+  },
+  "environmental_impact_assessment_engineer": {
+    "name": "环境影响评价工程师",
+    "category": "Other"
+  },
+  "fire_engineer": {
+    "name": "注册消防工程师",
+    "category": "Other"
+  },
+  "high_school_biology": {
+    "name": "高中生物",
+    "category": "STEM"
+  },
+  "high_school_chemistry": {
+    "name": "高中化学",
+    "category": "STEM"
+  },
+  "high_school_chinese": {
+    "name": "高中语文",
+    "category": "Humanities"
+  },
+  "high_school_geography": {
+    "name": "高中地理",
+    "category": "Social Sciences"
+  },
+  "high_school_history": {
+    "name": "高中历史",
+    "category": "Humanities"
+  },
+  "high_school_mathematics": {
+    "name": "高中数学",
+    "category": "STEM"
+  },
+  "high_school_physics": {
+    "name": "高中物理",
+    "category": "STEM"
+  },
+  "high_school_politics": {
+    "name": "高中政治",
+    "category": "Social Sciences"
+  },
+  "ideological_and_moral_cultivation": {
+    "name": "思想道德修养与法律基础",
+    "category": "Humanities"
+  },
+  "law": {
+    "name": "法学",
+    "category": "Humanities"
+  },
+  "legal_professional": {
+    "name": "法律职业资格",
+    "category": "Humanities"
+  },
+  "logic": {
+    "name": "逻辑学",
+    "category": "Humanities"
+  },
+  "mao_zedong_thought": {
+    "name": "毛泽东思想和中国特色社会主义理论体系概论",
+    "category": "Social Sciences"
+  },
+  "marxism": {
+    "name": "马克思主义基本原理",
+    "category": "Social Sciences"
+  },
+  "metrology_engineer": {
+    "name": "注册计量师",
+    "category": "STEM"
+  },
+  "middle_school_biology": {
+    "name": "初中生物",
+    "category": "STEM"
+  },
+  "middle_school_chemistry": {
+    "name": "初中化学",
+    "category": "STEM"
+  },
+  "middle_school_geography": {
+    "name": "初中地理",
+    "category": "Social Sciences"
+  },
+  "middle_school_history": {
+    "name": "初中历史",
+    "category": "Humanities"
+  },
+  "middle_school_mathematics": {
+    "name": "初中数学",
+    "category": "STEM"
+  },
+  "middle_school_physics": {
+    "name": "初中物理",
+    "category": "STEM"
+  },
+  "middle_school_politics": {
+    "name": "初中政治",
+    "category": "Social Sciences"
+  },
+  "modern_chinese_history": {
+    "name": "近代史纲要",
+    "category": "Humanities"
+  },
+  "operating_system": {
+    "name": "操作系统",
+    "category": "STEM"
+  },
+  "physician": {
+    "name": "医师资格",
+    "category": "Other"
+  },
+  "plant_protection": {
+    "name": "植物保护",
+    "category": "Other"
+  },
+  "probability_and_statistics": {
+    "name": "概率统计",
+    "category": "STEM"
+  },
+  "professional_tour_guide": {
+    "name": "导游资格",
+    "category": "Humanities"
+  },
+  "sports_science": {
+    "name": "体育学",
+    "category": "Other"
+  },
+  "tax_accountant": {
+    "name": "税务师",
+    "category": "Other"
+  },
+  "teacher_qualification": {
+    "name": "教师资格",
+    "category": "Social Sciences"
+  },
+  "urban_and_rural_planner": {
+    "name": "注册城乡规划师",
+    "category": "Other"
+  },
+  "veterinary_medicine": {
+    "name": "兽医学",
+    "category": "STEM"
+  }
+}

evaluation/cmmlu/cmmlu.py

@@ -0,0 +1,170 @@
+# Copyright 2025 the LlamaFactory team.
+# Copyright 2020 The HuggingFace Datasets Authors and the current dataset script contributor.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import os
+
+import datasets
+import pandas as pd
+
+
+_CITATION = """\
+@article{li2023cmmlu,
+  title={CMMLU: Measuring massive multitask language understanding in Chinese},
+  author={Haonan Li and Yixuan Zhang and Fajri Koto and Yifei Yang and others,
+  journal={arXiv preprint arXiv:2306.09212},
+  year={2023}
+}
+"""
+
+_DESCRIPTION = """\
+CMMLU is a comprehensive Chinese assessment suite specifically designed to evaluate the advanced knowledge
+and reasoning abilities of LLMs within the Chinese language and cultural context.
+"""
+
+_HOMEPAGE = "https://github.com/haonan-li/CMMLU"
+
+_LICENSE = "Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License"
+
+_URL = "cmmlu.zip"
+
+task_list = [
+    "agronomy",
+    "anatomy",
+    "ancient_chinese",
+    "arts",
+    "astronomy",
+    "business_ethics",
+    "chinese_civil_service_exam",
+    "chinese_driving_rule",
+    "chinese_food_culture",
+    "chinese_foreign_policy",
+    "chinese_history",
+    "chinese_literature",
+    "chinese_teacher_qualification",
+    "clinical_knowledge",
+    "college_actuarial_science",
+    "college_education",
+    "college_engineering_hydrology",
+    "college_law",
+    "college_mathematics",
+    "college_medical_statistics",
+    "college_medicine",
+    "computer_science",
+    "computer_security",
+    "conceptual_physics",
+    "construction_project_management",
+    "economics",
+    "education",
+    "electrical_engineering",
+    "elementary_chinese",
+    "elementary_commonsense",
+    "elementary_information_and_technology",
+    "elementary_mathematics",
+    "ethnology",
+    "food_science",
+    "genetics",
+    "global_facts",
+    "high_school_biology",
+    "high_school_chemistry",
+    "high_school_geography",
+    "high_school_mathematics",
+    "high_school_physics",
+    "high_school_politics",
+    "human_sexuality",
+    "international_law",
+    "journalism",
+    "jurisprudence",
+    "legal_and_moral_basis",
+    "logical",
+    "machine_learning",
+    "management",
+    "marketing",
+    "marxist_theory",
+    "modern_chinese",
+    "nutrition",
+    "philosophy",
+    "professional_accounting",
+    "professional_law",
+    "professional_medicine",
+    "professional_psychology",
+    "public_relations",
+    "security_study",
+    "sociology",
+    "sports_science",
+    "traditional_chinese_medicine",
+    "virology",
+    "world_history",
+    "world_religions",
+]
+
+
+class CMMLUConfig(datasets.BuilderConfig):
+    def __init__(self, **kwargs):
+        super().__init__(version=datasets.Version("1.0.1"), **kwargs)
+
+
+class CMMLU(datasets.GeneratorBasedBuilder):
+    BUILDER_CONFIGS = [
+        CMMLUConfig(
+            name=task_name,
+        )
+        for task_name in task_list
+    ]
+
+    def _info(self):
+        features = datasets.Features(
+            {
+                "question": datasets.Value("string"),
+                "A": datasets.Value("string"),
+                "B": datasets.Value("string"),
+                "C": datasets.Value("string"),
+                "D": datasets.Value("string"),
+                "answer": datasets.Value("string"),
+            }
+        )
+        return datasets.DatasetInfo(
+            description=_DESCRIPTION,
+            features=features,
+            homepage=_HOMEPAGE,
+            license=_LICENSE,
+            citation=_CITATION,
+        )
+
+    def _split_generators(self, dl_manager):
+        data_dir = dl_manager.download_and_extract(_URL)
+        task_name = self.config.name
+        return [
+            datasets.SplitGenerator(
+                name=datasets.Split.TEST,
+                gen_kwargs={
+                    "filepath": os.path.join(data_dir, f"test/{task_name}.csv"),
+                },
+            ),
+            datasets.SplitGenerator(
+                name=datasets.Split.TRAIN,
+                gen_kwargs={
+                    "filepath": os.path.join(data_dir, f"dev/{task_name}.csv"),
+                },
+            ),
+        ]
+
+    def _generate_examples(self, filepath):
+        df = pd.read_csv(filepath, header=0, index_col=0, encoding="utf-8")
+        for i, instance in enumerate(df.to_dict(orient="records")):
+            question = instance.pop("Question", "")
+            answer = instance.pop("Answer", "")
+            instance["question"] = question
+            instance["answer"] = answer
+            yield i, instance

evaluation/cmmlu/mapping.json

@@ -0,0 +1,270 @@
+{
+  "agronomy": {
+    "name": "农学",
+    "category": "Other"
+  },
+  "anatomy": {
+    "name": "解剖学",
+    "category": "STEM"
+  },
+  "ancient_chinese": {
+    "name": "古汉语",
+    "category": "Social Sciences"
+  },
+  "arts": {
+    "name": "艺术学",
+    "category": "Humanities"
+  },
+  "astronomy": {
+    "name": "天文学",
+    "category": "STEM"
+  },
+  "business_ethics": {
+    "name": "商业伦理",
+    "category": "Social Sciences"
+  },
+  "chinese_civil_service_exam": {
+    "name": "中国公务员考试",
+    "category": "Social Sciences"
+  },
+  "chinese_driving_rule": {
+    "name": "中国驾驶规则",
+    "category": "Other"
+  },
+  "chinese_food_culture": {
+    "name": "中国饮食文化",
+    "category": "Social Sciences"
+  },
+  "chinese_foreign_policy": {
+    "name": "中国外交政策",
+    "category": "Social Sciences"
+  },
+  "chinese_history": {
+    "name": "中国历史",
+    "category": "Humanities"
+  },
+  "chinese_literature": {
+    "name": "中国文学",
+    "category": "Humanities"
+  },
+  "chinese_teacher_qualification": {
+    "name": "中国教师资格",
+    "category": "Social Sciences"
+  },
+  "college_actuarial_science": {
+    "name": "大学精算学",
+    "category": "STEM"
+  },
+  "college_education": {
+    "name": "大学教育学",
+    "category": "Social Sciences"
+  },
+  "college_engineering_hydrology": {
+    "name": "大学工程水文学",
+    "category": "STEM"
+  },
+  "college_law": {
+    "name": "大学法律",
+    "category": "Humanities"
+  },
+  "college_mathematics": {
+    "name": "大学数学",
+    "category": "STEM"
+  },
+  "college_medical_statistics": {
+    "name": "大学医学统计",
+    "category": "STEM"
+  },
+  "clinical_knowledge": {
+    "name": "临床知识",
+    "category": "Other"
+  },
+  "college_medicine": {
+    "name": "大学医学",
+    "category": "Other"
+  },
+  "computer_science": {
+    "name": "计算机科学",
+    "category": "STEM"
+  },
+  "computer_security": {
+    "name": "计算机安全",
+    "category": "Other"
+  },
+  "conceptual_physics": {
+    "name": "概念物理学",
+    "category": "STEM"
+  },
+  "construction_project_management": {
+    "name": "建设工程管理",
+    "category": "Other"
+  },
+  "economics": {
+    "name": "经济学",
+    "category": "Social Sciences"
+  },
+  "education": {
+    "name": "教育学",
+    "category": "Social Sciences"
+  },
+  "elementary_chinese": {
+    "name": "小学语文",
+    "category": "Social Sciences"
+  },
+  "elementary_commonsense": {
+    "name": "小学常识",
+    "category": "Other"
+  },
+  "elementary_information_and_technology": {
+    "name": "小学信息技术",
+    "category": "Other"
+  },
+  "electrical_engineering": {
+    "name": "电气工程",
+    "category": "STEM"
+  },
+  "elementary_mathematics": {
+    "name": "初等数学",
+    "category": "STEM"
+  },
+  "ethnology": {
+    "name": "民族学",
+    "category": "Social Sciences"
+  },
+  "food_science": {
+    "name": "食品科学",
+    "category": "Other"
+  },
+  "genetics": {
+    "name": "遗传学",
+    "category": "STEM"
+  },
+  "global_facts": {
+    "name": "全球事实",
+    "category": "Humanities"
+  },
+  "high_school_biology": {
+    "name": "高中生物",
+    "category": "STEM"
+  },
+  "high_school_chemistry": {
+    "name": "高中化学",
+    "category": "STEM"
+  },
+  "high_school_geography": {
+    "name": "高中地理",
+    "category": "Social Sciences"
+  },
+  "high_school_mathematics": {
+    "name": "高中数学",
+    "category": "STEM"
+  },
+  "high_school_physics": {
+    "name": "高中物理学",
+    "category": "STEM"
+  },
+  "high_school_politics": {
+    "name": "高中政治",
+    "category": "Social Sciences"
+  },
+  "human_sexuality": {
+    "name": "人类性行为",
+    "category": "Other"
+  },
+  "international_law": {
+    "name": "国际法学",
+    "category": "Humanities"
+  },
+  "journalism": {
+    "name": "新闻学",
+    "category": "Social Sciences"
+  },
+  "jurisprudence": {
+    "name": "法理学",
+    "category": "Humanities"
+  },
+  "legal_and_moral_basis": {
+    "name": "法律与道德基础",
+    "category": "Other"
+  },
+  "logical": {
+    "name": "逻辑学",
+    "category": "Humanities"
+  },
+  "machine_learning": {
+    "name": "机器学习",
+    "category": "STEM"
+  },
+  "management": {
+    "name": "管理学",
+    "category": "Social Sciences"
+  },
+  "marketing": {
+    "name": "市场营销",
+    "category": "Social Sciences"
+  },
+  "marxist_theory": {
+    "name": "马克思主义理论",
+    "category": "Humanities"
+  },
+  "modern_chinese": {
+    "name": "现代汉语",
+    "category": "Social Sciences"
+  },
+  "nutrition": {
+    "name": "营养学",
+    "category": "Other"
+  },
+  "philosophy": {
+    "name": "哲学",
+    "category": "Humanities"
+  },
+  "professional_accounting": {
+    "name": "专业会计",
+    "category": "Social Sciences"
+  },
+  "professional_law": {
+    "name": "专业法学",
+    "category": "Humanities"
+  },
+  "professional_medicine": {
+    "name": "专业医学",
+    "category": "Other"
+  },
+  "professional_psychology": {
+    "name": "专业心理学",
+    "category": "Social Sciences"
+  },
+  "public_relations": {
+    "name": "公共关系",
+    "category": "Social Sciences"
+  },
+  "security_study": {
+    "name": "安全研究",
+    "category": "Social Sciences"
+  },
+  "sociology": {
+    "name": "社会学",
+    "category": "Social Sciences"
+  },
+  "sports_science": {
+    "name": "体育学",
+    "category": "Other"
+  },
+  "traditional_chinese_medicine": {
+    "name": "中医中药",
+    "category": "Other"
+  },
+  "virology": {
+    "name": "病毒学",
+    "category": "STEM"
+  },
+  "world_history": {
+    "name": "世界历史",
+    "category": "Humanities"
+  },
+  "world_religions": {
+    "name": "世界宗教",
+    "category": "Humanities"
+  }
+}

evaluation/mmlu/mapping.json

@@ -0,0 +1,230 @@
+{
+  "abstract_algebra": {
+    "name": "abstract algebra",
+    "category": "STEM"
+  },
+  "anatomy": {
+    "name": "anatomy",
+    "category": "Other"
+  },
+  "astronomy": {
+    "name": "astronomy",
+    "category": "STEM"
+  },
+  "business_ethics": {
+    "name": "business ethics",
+    "category": "Other"
+  },
+  "clinical_knowledge": {
+    "name": "clinical knowledge",
+    "category": "Other"
+  },
+  "college_biology": {
+    "name": "college biology",
+    "category": "STEM"
+  },
+  "college_chemistry": {
+    "name": "college chemistry",
+    "category": "STEM"
+  },
+  "college_computer_science": {
+    "name": "college computer science",
+    "category": "STEM"
+  },
+  "college_mathematics": {
+    "name": "college mathematics",
+    "category": "STEM"
+  },
+  "college_medicine": {
+    "name": "college medicine",
+    "category": "Other"
+  },
+  "college_physics": {
+    "name": "college physics",
+    "category": "STEM"
+  },
+  "computer_security": {
+    "name": "computer security",
+    "category": "STEM"
+  },
+  "conceptual_physics": {
+    "name": "conceptual physics",
+    "category": "STEM"
+  },
+  "econometrics": {
+    "name": "econometrics",
+    "category": "Social Sciences"
+  },
+  "electrical_engineering": {
+    "name": "electrical engineering",
+    "category": "STEM"
+  },
+  "elementary_mathematics": {
+    "name": "elementary mathematics",
+    "category": "STEM"
+  },
+  "formal_logic": {
+    "name": "formal logic",
+    "category": "Humanities"
+  },
+  "global_facts": {
+    "name": "global facts",
+    "category": "Other"
+  },
+  "high_school_biology": {
+    "name": "high school biology",
+    "category": "STEM"
+  },
+  "high_school_chemistry": {
+    "name": "high school chemistry",
+    "category": "STEM"
+  },
+  "high_school_computer_science": {
+    "name": "high school computer science",
+    "category": "STEM"
+  },
+  "high_school_european_history": {
+    "name": "high school european history",
+    "category": "Humanities"
+  },
+  "high_school_geography": {
+    "name": "high school geography",
+    "category": "Social Sciences"
+  },
+  "high_school_government_and_politics": {
+    "name": "high school government and politics",
+    "category": "Social Sciences"
+  },
+  "high_school_macroeconomics": {
+    "name": "high school macroeconomics",
+    "category": "Social Sciences"
+  },
+  "high_school_mathematics": {
+    "name": "high school mathematics",
+    "category": "STEM"
+  },
+  "high_school_microeconomics": {
+    "name": "high school microeconomics",
+    "category": "Social Sciences"
+  },
+  "high_school_physics": {
+    "name": "high school physics",
+    "category": "STEM"
+  },
+  "high_school_psychology": {
+    "name": "high school psychology",
+    "category": "Social Sciences"
+  },
+  "high_school_statistics": {
+    "name": "high school statistics",
+    "category": "STEM"
+  },
+  "high_school_us_history": {
+    "name": "high school us history",
+    "category": "Humanities"
+  },
+  "high_school_world_history": {
+    "name": "high school world history",
+    "category": "Humanities"
+  },
+  "human_aging": {
+    "name": "human aging",
+    "category": "Other"
+  },
+  "human_sexuality": {
+    "name": "human sexuality",
+    "category": "Social Sciences"
+  },
+  "international_law": {
+    "name": "international law",
+    "category": "Humanities"
+  },
+  "jurisprudence": {
+    "name": "jurisprudence",
+    "category": "Humanities"
+  },
+  "logical_fallacies": {
+    "name": "logical fallacies",
+    "category": "Humanities"
+  },
+  "machine_learning": {
+    "name": "machine learning",
+    "category": "STEM"
+  },
+  "management": {
+    "name": "management",
+    "category": "Other"
+  },
+  "marketing": {
+    "name": "marketing",
+    "category": "Other"
+  },
+  "medical_genetics": {
+    "name": "medical genetics",
+    "category": "Other"
+  },
+  "miscellaneous": {
+    "name": "miscellaneous",
+    "category": "Other"
+  },
+  "moral_disputes": {
+    "name": "moral disputes",
+    "category": "Humanities"
+  },
+  "moral_scenarios": {
+    "name": "moral scenarios",
+    "category": "Humanities"
+  },
+  "nutrition": {
+    "name": "nutrition",
+    "category": "Other"
+  },
+  "philosophy": {
+    "name": "philosophy",
+    "category": "Humanities"
+  },
+  "prehistory": {
+    "name": "prehistory",
+    "category": "Humanities"
+  },
+  "professional_accounting": {
+    "name": "professional accounting",
+    "category": "Other"
+  },
+  "professional_law": {
+    "name": "professional law",
+    "category": "Humanities"
+  },
+  "professional_medicine": {
+    "name": "professional medicine",
+    "category": "Other"
+  },
+  "professional_psychology": {
+    "name": "professional psychology",
+    "category": "Social Sciences"
+  },
+  "public_relations": {
+    "name": "public relations",
+    "category": "Social Sciences"
+  },
+  "security_studies": {
+    "name": "security studies",
+    "category": "Social Sciences"
+  },
+  "sociology": {
+    "name": "sociology",
+    "category": "Social Sciences"
+  },
+  "us_foreign_policy": {
+    "name": "us foreign policy",
+    "category": "Social Sciences"
+  },
+  "virology": {
+    "name": "virology",
+    "category": "Other"
+  },
+  "world_religions": {
+    "name": "world religions",
+    "category": "Humanities"
+  }
+}

evaluation/mmlu/mmlu.py

@@ -0,0 +1,163 @@
+# Copyright 2025 the LlamaFactory team.
+# Copyright 2020 The HuggingFace Datasets Authors and the current dataset script contributor.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import os
+
+import datasets
+import pandas as pd
+
+
+_CITATION = """\
+@article{hendryckstest2021,
+  title={Measuring Massive Multitask Language Understanding},
+  author={Dan Hendrycks and Collin Burns and others},
+  journal={Proceedings of the International Conference on Learning Representations (ICLR)},
+  year={2021}
+}
+"""
+
+_DESCRIPTION = """\
+Measuring Massive Multitask Language Understanding by Dan Hendrycks, Collin Burns, Steven Basart,
+Andy Zou, Mantas Mazeika, Dawn Song, and Jacob Steinhardt (ICLR 2021).
+"""
+
+_HOMEPAGE = "https://github.com/hendrycks/test"
+
+_LICENSE = "MIT"
+
+_URL = "mmlu.zip"
+
+task_list = [
+    "high_school_european_history",
+    "business_ethics",
+    "clinical_knowledge",
+    "medical_genetics",
+    "high_school_us_history",
+    "high_school_physics",
+    "high_school_world_history",
+    "virology",
+    "high_school_microeconomics",
+    "econometrics",
+    "college_computer_science",
+    "high_school_biology",
+    "abstract_algebra",
+    "professional_accounting",
+    "philosophy",
+    "professional_medicine",
+    "nutrition",
+    "global_facts",
+    "machine_learning",
+    "security_studies",
+    "public_relations",
+    "professional_psychology",
+    "prehistory",
+    "anatomy",
+    "human_sexuality",
+    "college_medicine",
+    "high_school_government_and_politics",
+    "college_chemistry",
+    "logical_fallacies",
+    "high_school_geography",
+    "elementary_mathematics",
+    "human_aging",
+    "college_mathematics",
+    "high_school_psychology",
+    "formal_logic",
+    "high_school_statistics",
+    "international_law",
+    "high_school_mathematics",
+    "high_school_computer_science",
+    "conceptual_physics",
+    "miscellaneous",
+    "high_school_chemistry",
+    "marketing",
+    "professional_law",
+    "management",
+    "college_physics",
+    "jurisprudence",
+    "world_religions",
+    "sociology",
+    "us_foreign_policy",
+    "high_school_macroeconomics",
+    "computer_security",
+    "moral_scenarios",
+    "moral_disputes",
+    "electrical_engineering",
+    "astronomy",
+    "college_biology",
+]
+
+
+class MMLUConfig(datasets.BuilderConfig):
+    def __init__(self, **kwargs):
+        super().__init__(version=datasets.Version("1.0.0"), **kwargs)
+
+
+class MMLU(datasets.GeneratorBasedBuilder):
+    BUILDER_CONFIGS = [
+        MMLUConfig(
+            name=task_name,
+        )
+        for task_name in task_list
+    ]
+
+    def _info(self):
+        features = datasets.Features(
+            {
+                "question": datasets.Value("string"),
+                "A": datasets.Value("string"),
+                "B": datasets.Value("string"),
+                "C": datasets.Value("string"),
+                "D": datasets.Value("string"),
+                "answer": datasets.Value("string"),
+            }
+        )
+        return datasets.DatasetInfo(
+            description=_DESCRIPTION,
+            features=features,
+            homepage=_HOMEPAGE,
+            license=_LICENSE,
+            citation=_CITATION,
+        )
+
+    def _split_generators(self, dl_manager):
+        data_dir = dl_manager.download_and_extract(_URL)
+        task_name = self.config.name
+        return [
+            datasets.SplitGenerator(
+                name=datasets.Split.TEST,
+                gen_kwargs={
+                    "filepath": os.path.join(data_dir, "data", "test", f"{task_name}_test.csv"),
+                },
+            ),
+            datasets.SplitGenerator(
+                name=datasets.Split.VALIDATION,
+                gen_kwargs={
+                    "filepath": os.path.join(data_dir, "data", "val", f"{task_name}_val.csv"),
+                },
+            ),
+            datasets.SplitGenerator(
+                name=datasets.Split.TRAIN,
+                gen_kwargs={
+                    "filepath": os.path.join(data_dir, "data", "dev", f"{task_name}_dev.csv"),
+                },
+            ),
+        ]
+
+    def _generate_examples(self, filepath):
+        df = pd.read_csv(filepath, header=None)
+        df.columns = ["question", "A", "B", "C", "D", "answer"]
+
+        yield from enumerate(df.to_dict(orient="records"))

src/llamafactory/cli.py

@@ -25,27 +25,64 @@ USAGE = (
     + "| Usage:                                                             |\n"
     + "|   llamafactory-cli api -h: launch an OpenAI-style API server       |\n"
     + "|   llamafactory-cli chat -h: launch a chat interface in CLI         |\n"
+    + "|   llamafactory-cli eval -h: evaluate models                        |\n"
     + "|   llamafactory-cli export -h: merge LoRA adapters and export model |\n"
     + "|   llamafactory-cli train -h: train models                          |\n"
     + "|   llamafactory-cli webchat -h: launch a chat interface in Web UI   |\n"
     + "|   llamafactory-cli webui: launch LlamaBoard                        |\n"
-    + "|   llamafactory-cli env: show environment info                      |\n"
     + "|   llamafactory-cli version: show version info                      |\n"
-    + "| Hint: You can use `lmf` as a shortcut for `llamafactory-cli`.      |\n"
     + "-" * 70
 )
 
 
+def _run_api():
+    from .api.app import run_api
+
+    return run_api()
+
+
+def _run_chat():
+    from .chat.chat_model import run_chat
+
+    return run_chat()
+
+
+def _run_eval():
+    from .eval.evaluator import run_eval
+
+    return run_eval()
+
+
+def _export_model():
+    from .train.tuner import export_model
+
+    return export_model()
+
+
+def _run_exp():
+    from .train.tuner import run_exp
+
+    return run_exp()
+
+
+def _run_web_demo():
+    from .webui.interface import run_web_demo
+
+    return run_web_demo()
+
+
+def _run_web_ui():
+    from .webui.interface import run_web_ui
+
+    return run_web_ui()
+
+
 def main():
+    from . import launcher
     from .extras import logging
     from .extras.env import VERSION, print_env
     from .extras.misc import find_available_port, get_device_count, is_env_enabled, use_ray
 
-    if is_env_enabled("USE_V1"):
-        from .v1 import launcher
-    else:
-        from . import launcher
-
     logger = logging.get_logger(__name__)
 
     WELCOME = (
@@ -61,14 +98,14 @@ def main():
     )
 
     COMMAND_MAP = {
-        "api": launcher.run_api,
-        "chat": launcher.run_chat,
+        "api": _run_api,
+        "chat": _run_chat,
         "env": print_env,
-        "eval": launcher.run_eval,
-        "export": launcher.export_model,
-        "train": launcher.run_exp,
-        "webchat": launcher.run_web_demo,
-        "webui": launcher.run_web_ui,
+        "eval": _run_eval,
+        "export": _export_model,
+        "train": _run_exp,
+        "webchat": _run_web_demo,
+        "webui": _run_web_ui,
         "version": partial(print, WELCOME),
         "help": partial(print, USAGE),
     }

src/llamafactory/extras/env.py

@@ -15,22 +15,22 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
+import os
+import platform
+
+import accelerate
+import datasets
+import peft
+import torch
+import transformers
+import trl
+from transformers.utils import is_torch_cuda_available, is_torch_npu_available
+
 
 VERSION = "0.9.4.dev0"
 
 
 def print_env() -> None:
-    import os
-    import platform
-
-    import accelerate
-    import datasets
-    import peft
-    import torch
-    import transformers
-    import trl
-    from transformers.utils import is_torch_cuda_available, is_torch_npu_available
-
     info = {
         "`llamafactory` version": VERSION,
         "Platform": platform.platform(),

src/llamafactory/launcher.py

@@ -12,46 +12,12 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-
-def run_api():
-    from llamafactory.api.app import run_api as _run_api
-
-    _run_api()
+from llamafactory.train.tuner import run_exp  # use absolute import
 
 
-def run_chat():
-    from llamafactory.chat.chat_model import run_chat as _run_chat
-
-    return _run_chat()
-
-
-def run_eval():
-    raise NotImplementedError("Evaluation will be deprecated in the future.")
-
-
-def export_model():
-    from llamafactory.train.tuner import export_model as _export_model
-
-    return _export_model()
-
-
-def run_exp():
-    from llamafactory.train.tuner import run_exp as _run_exp
-
-    return _run_exp()  # use absolute import
-
-
-def run_web_demo():
-    from llamafactory.webui.interface import run_web_demo as _run_web_demo
-
-    return _run_web_demo()
-
-
-def run_web_ui():
-    from llamafactory.webui.interface import run_web_ui as _run_web_ui
-
-    return _run_web_ui()
+def launch():
+    run_exp()
 
 
 if __name__ == "__main__":
-    run_exp()
+    launch()

src/llamafactory/v1/launcher.py

@@ -1,33 +0,0 @@
-# Copyright 2025 the LlamaFactory team.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-
-def run_train():
-    raise NotImplementedError("Please use `llamafactory-cli sft` or `llamafactory-cli rm`.")
-
-
-def run_chat():
-    from llamafactory.v1.core.chat_sampler import Sampler
-
-    Sampler().cli()
-
-
-def run_sft():
-    from llamafactory.v1.train.sft import SFTTrainer
-
-    SFTTrainer().run()
-
-
-if __name__ == "__main__":
-    run_train()