Skip to content

GitLab

Commit ca625f43 authored by shihm's avatar shihm
Browse files

uodata

parent 7164651d
Expand all Show whitespace changes
Inline Side-by-side
Showing with 798 additions and 1306 deletions
data/dataset_info.json
View file @ ca625f43
...@@ -143,14 +143,6 @@ ...@@ -143,14 +143,6 @@
"hf_hub_url": "BelleGroup/school_math_0.25M", "hf_hub_url": "BelleGroup/school_math_0.25M",
"ms_hub_url": "AI-ModelScope/school_math_0.25M" "ms_hub_url": "AI-ModelScope/school_math_0.25M"
}, },
"belle_multiturn": {
"script_url": "belle_multiturn",
"formatting": "sharegpt"
},
"ultra_chat": {
"script_url": "ultra_chat",
"formatting": "sharegpt"
},
"open_platypus": { "open_platypus": {
"hf_hub_url": "garage-bAInd/Open-Platypus", "hf_hub_url": "garage-bAInd/Open-Platypus",
"ms_hub_url": "AI-ModelScope/Open-Platypus" "ms_hub_url": "AI-ModelScope/Open-Platypus"
...@@ -259,6 +251,10 @@ ...@@ -259,6 +251,10 @@
"assistant_tag": "assistant" "assistant_tag": "assistant"
} }
}, },
"infinity_instruct": {
"hf_hub_url": "BAAI/Infinity-Instruct",
"formatting": "sharegpt"
},
"agent_instruct": { "agent_instruct": {
"hf_hub_url": "THUDM/AgentInstruct", "hf_hub_url": "THUDM/AgentInstruct",
"ms_hub_url": "ZhipuAI/AgentInstruct", "ms_hub_url": "ZhipuAI/AgentInstruct",
...@@ -475,6 +471,14 @@ ...@@ -475,6 +471,14 @@
"ultrachat_de": { "ultrachat_de": {
"hf_hub_url": "mayflowergmbh/ultra-chat_de" "hf_hub_url": "mayflowergmbh/ultra-chat_de"
}, },
"dlr_web": {
"hf_hub_url": "Attention1115/DLR-Web",
"split": "full",
"columns": {
"prompt": "question",
"response": "response"
}
},
"dpo_en_demo": { "dpo_en_demo": {
"file_name": "dpo_en_demo.json", "file_name": "dpo_en_demo.json",
"ranking": true, "ranking": true,
...@@ -579,16 +583,6 @@ ...@@ -579,16 +583,6 @@
"system": "system" "system": "system"
} }
}, },
"hh_rlhf_en": {
"script_url": "hh_rlhf_en",
"ranking": true,
"columns": {
"prompt": "instruction",
"chosen": "chosen",
"rejected": "rejected",
"history": "history"
}
},
"nectar_rm": { "nectar_rm": {
"hf_hub_url": "AstraMindAI/RLAIF-Nectar", "hf_hub_url": "AstraMindAI/RLAIF-Nectar",
"ms_hub_url": "AI-ModelScope/RLAIF-Nectar", "ms_hub_url": "AI-ModelScope/RLAIF-Nectar",
...@@ -700,6 +694,36 @@ ...@@ -700,6 +694,36 @@
"prompt": "text" "prompt": "text"
} }
}, },
"cci3_hq": {
"hf_hub_url": "BAAI/CCI3-HQ",
"columns": {
"prompt": "text"
}
},
"cci3_data": {
"hf_hub_url": "BAAI/CCI3-Data",
"columns": {
"prompt": "text"
}
},
"cci4_base": {
"hf_hub_url": "BAAI/CCI4.0-M2-Base-v1",
"columns": {
"prompt": "text"
}
},
"cci4_cot": {
"hf_hub_url": "BAAI/CCI4.0-M2-CoT-v1",
"columns": {
"prompt": "text"
}
},
"cci4_extra": {
"hf_hub_url": "BAAI/CCI4.0-M2-Extra-v1",
"columns": {
"prompt": "text"
}
},
"the_stack": { "the_stack": {
"hf_hub_url": "bigcode/the-stack", "hf_hub_url": "bigcode/the-stack",
"ms_hub_url": "AI-ModelScope/the-stack", "ms_hub_url": "AI-ModelScope/the-stack",
......
data/reason_tool_use_demo_50.jsonl 0 → 100644
View file @ ca625f43
This source diff could not be displayed because it is too large. You can view the blob instead.
data/v1_dpo_demo.jsonl 0 → 100644
View file @ ca625f43
This diff is collapsed.
data/v1_dpo_demo.yaml 0 → 100644
View file @ ca625f43
dpo_zh_demo:
path: HuggingFaceH4/orca_dpo_pairs
split: train_prefs
converter: pair
data/v1_sft_demo.jsonl 0 → 100644
View file @ ca625f43
This diff is collapsed.
data/v1_sft_demo.yaml 0 → 100644
View file @ ca625f43
identity:
path: data/identity.json
source: local
converter: alpaca
alpaca_en_demo:
path: data/alpaca_en_demo.json
source: local
converter: alpaca
size: 500
evaluation/ceval/ceval.py deleted 100644 → 0
View file @ 7164651d
# 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 deleted 100644 → 0
View file @ 7164651d
{
"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 deleted 100644 → 0
View file @ 7164651d
# 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 deleted 100644 → 0
View file @ 7164651d
{
"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 deleted 100644 → 0
View file @ 7164651d
{
"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 deleted 100644 → 0
View file @ 7164651d
# 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"))
examples/README.md
View file @ ca625f43
...@@ -18,19 +18,19 @@ By default, LLaMA-Factory uses all visible computing devices. ...@@ -18,19 +18,19 @@ By default, LLaMA-Factory uses all visible computing devices.
Basic usage: Basic usage:
```bash ```bash
llamafactory-cli train examples/train_lora/llama3_lora_sft.yaml llamafactory-cli train examples/train_lora/qwen3_lora_sft.yaml
``` ```
Advanced usage: Advanced usage:
```bash ```bash
CUDA_VISIBLE_DEVICES=0,1 llamafactory-cli train examples/train_lora/llama3_lora_sft.yaml \ CUDA_VISIBLE_DEVICES=0,1 llamafactory-cli train examples/train_lora/qwen3_lora_sft.yaml \
learning_rate=1e-5 \ learning_rate=1e-5 \
logging_steps=1 logging_steps=1
``` ```
```bash ```bash
bash examples/train_lora/llama3_lora_sft.sh bash examples/train_lora/qwen3_lora_sft.sh
``` ```
## Examples ## Examples
...@@ -40,49 +40,43 @@ bash examples/train_lora/llama3_lora_sft.sh ...@@ -40,49 +40,43 @@ bash examples/train_lora/llama3_lora_sft.sh
#### (Continuous) Pre-Training #### (Continuous) Pre-Training
```bash ```bash
llamafactory-cli train examples/train_lora/llama3_lora_pretrain.yaml llamafactory-cli train examples/train_lora/qwen3_lora_pretrain.yaml
``` ```
#### Supervised Fine-Tuning #### Supervised Fine-Tuning
```bash ```bash
llamafactory-cli train examples/train_lora/llama3_lora_sft.yaml llamafactory-cli train examples/train_lora/qwen3_lora_sft.yaml
``` ```
#### Multimodal Supervised Fine-Tuning #### Multimodal Supervised Fine-Tuning
```bash ```bash
llamafactory-cli train examples/train_lora/qwen2_5vl_lora_sft.yaml llamafactory-cli train examples/train_lora/qwen3vl_lora_sft.yaml
``` ```
#### DPO/ORPO/SimPO Training #### DPO/ORPO/SimPO Training
```bash ```bash
llamafactory-cli train examples/train_lora/llama3_lora_dpo.yaml llamafactory-cli train examples/train_lora/qwen3_lora_dpo.yaml
``` ```
#### Multimodal DPO/ORPO/SimPO Training #### Multimodal DPO/ORPO/SimPO Training
```bash ```bash
llamafactory-cli train examples/train_lora/qwen2_5vl_lora_dpo.yaml llamafactory-cli train examples/train_lora/qwen3vl_lora_dpo.yaml
``` ```
#### Reward Modeling #### Reward Modeling
```bash ```bash
llamafactory-cli train examples/train_lora/llama3_lora_reward.yaml llamafactory-cli train examples/train_lora/qwen3_lora_reward.yaml
```
#### PPO Training
```bash
llamafactory-cli train examples/train_lora/llama3_lora_ppo.yaml
``` ```
#### KTO Training #### KTO Training
```bash ```bash
llamafactory-cli train examples/train_lora/llama3_lora_kto.yaml llamafactory-cli train examples/train_lora/qwen3_lora_kto.yaml
``` ```
#### Preprocess Dataset #### Preprocess Dataset
...@@ -90,32 +84,26 @@ llamafactory-cli train examples/train_lora/llama3_lora_kto.yaml ...@@ -90,32 +84,26 @@ llamafactory-cli train examples/train_lora/llama3_lora_kto.yaml
It is useful for large dataset, use `tokenized_path` in config to load the preprocessed dataset. It is useful for large dataset, use `tokenized_path` in config to load the preprocessed dataset.
```bash ```bash
llamafactory-cli train examples/train_lora/llama3_preprocess.yaml llamafactory-cli train examples/train_lora/qwen3_preprocess.yaml
```
#### Evaluating on MMLU/CMMLU/C-Eval Benchmarks
```bash
llamafactory-cli eval examples/train_lora/llama3_lora_eval.yaml
``` ```
#### Supervised Fine-Tuning on Multiple Nodes #### Supervised Fine-Tuning on Multiple Nodes
```bash ```bash
FORCE_TORCHRUN=1 NNODES=2 NODE_RANK=0 MASTER_ADDR=192.168.0.1 MASTER_PORT=29500 llamafactory-cli train examples/train_lora/llama3_lora_sft.yaml FORCE_TORCHRUN=1 NNODES=2 NODE_RANK=0 MASTER_ADDR=192.168.0.1 MASTER_PORT=29500 llamafactory-cli train examples/train_lora/qwen3_lora_sft.yaml
FORCE_TORCHRUN=1 NNODES=2 NODE_RANK=1 MASTER_ADDR=192.168.0.1 MASTER_PORT=29500 llamafactory-cli train examples/train_lora/llama3_lora_sft.yaml FORCE_TORCHRUN=1 NNODES=2 NODE_RANK=1 MASTER_ADDR=192.168.0.1 MASTER_PORT=29500 llamafactory-cli train examples/train_lora/qwen3_lora_sft.yaml
``` ```
#### Supervised Fine-Tuning with DeepSpeed ZeRO-3 (Weight Sharding) #### Supervised Fine-Tuning with DeepSpeed ZeRO-3 (Weight Sharding)
```bash ```bash
FORCE_TORCHRUN=1 llamafactory-cli train examples/train_lora/llama3_lora_sft_ds3.yaml FORCE_TORCHRUN=1 llamafactory-cli train examples/train_lora/qwen3_lora_sft_ds3.yaml
``` ```
#### Supervised Fine-Tuning with Ray on 4 GPUs #### Supervised Fine-Tuning with Ray on 4 GPUs
```bash ```bash
USE_RAY=1 llamafactory-cli train examples/train_lora/llama3_lora_sft_ray.yaml USE_RAY=1 llamafactory-cli train examples/train_lora/qwen3_lora_sft_ray.yaml
``` ```
### QLoRA Fine-Tuning ### QLoRA Fine-Tuning
...@@ -123,13 +111,13 @@ USE_RAY=1 llamafactory-cli train examples/train_lora/llama3_lora_sft_ray.yaml ...@@ -123,13 +111,13 @@ USE_RAY=1 llamafactory-cli train examples/train_lora/llama3_lora_sft_ray.yaml
#### Supervised Fine-Tuning with 4/8-bit Bitsandbytes/HQQ/EETQ Quantization (Recommended) #### Supervised Fine-Tuning with 4/8-bit Bitsandbytes/HQQ/EETQ Quantization (Recommended)
```bash ```bash
llamafactory-cli train examples/train_qlora/llama3_lora_sft_otfq.yaml llamafactory-cli train examples/train_qlora/qwen3_lora_sft_otfq.yaml
``` ```
#### Supervised Fine-Tuning with 4-bit Bitsandbytes Quantization on Ascend NPU #### Supervised Fine-Tuning with 4-bit Bitsandbytes Quantization on Ascend NPU
```bash ```bash
llamafactory-cli train examples/train_qlora/llama3_lora_sft_bnb_npu.yaml llamafactory-cli train examples/train_qlora/qwen3_lora_sft_bnb_npu.yaml
``` ```
#### Supervised Fine-Tuning with 4/8-bit GPTQ Quantization #### Supervised Fine-Tuning with 4/8-bit GPTQ Quantization
...@@ -155,14 +143,14 @@ llamafactory-cli train examples/train_qlora/llama3_lora_sft_aqlm.yaml ...@@ -155,14 +143,14 @@ llamafactory-cli train examples/train_qlora/llama3_lora_sft_aqlm.yaml
#### Supervised Fine-Tuning on Single Node #### Supervised Fine-Tuning on Single Node
```bash ```bash
FORCE_TORCHRUN=1 llamafactory-cli train examples/train_full/llama3_full_sft.yaml FORCE_TORCHRUN=1 llamafactory-cli train examples/train_full/qwen3_full_sft.yaml
``` ```
#### Supervised Fine-Tuning on Multiple Nodes #### Supervised Fine-Tuning on Multiple Nodes
```bash ```bash
FORCE_TORCHRUN=1 NNODES=2 NODE_RANK=0 MASTER_ADDR=192.168.0.1 MASTER_PORT=29500 llamafactory-cli train examples/train_full/llama3_full_sft.yaml FORCE_TORCHRUN=1 NNODES=2 NODE_RANK=0 MASTER_ADDR=192.168.0.1 MASTER_PORT=29500 llamafactory-cli train examples/train_full/qwen3_full_sft.yaml
FORCE_TORCHRUN=1 NNODES=2 NODE_RANK=1 MASTER_ADDR=192.168.0.1 MASTER_PORT=29500 llamafactory-cli train examples/train_full/llama3_full_sft.yaml FORCE_TORCHRUN=1 NNODES=2 NODE_RANK=1 MASTER_ADDR=192.168.0.1 MASTER_PORT=29500 llamafactory-cli train examples/train_full/qwen3_full_sft.yaml
``` ```
### Elastic and Fault-Tolerant Supervised Fine-Tuning on Multiple Nodes ### Elastic and Fault-Tolerant Supervised Fine-Tuning on Multiple Nodes
...@@ -170,13 +158,13 @@ FORCE_TORCHRUN=1 NNODES=2 NODE_RANK=1 MASTER_ADDR=192.168.0.1 MASTER_PORT=29500 ...@@ -170,13 +158,13 @@ FORCE_TORCHRUN=1 NNODES=2 NODE_RANK=1 MASTER_ADDR=192.168.0.1 MASTER_PORT=29500
To launch an elastic job with `MAX_RESTARTS` failures retries, run the following on at least `MIN_NNODES` nodes and at most `MAX_NNODES` nodes. `RDZV_ID` should be set as a unique job id (shared by all nodes participating in the job). See also [torchrun](https://docs.pytorch.org/docs/stable/elastic/run.html). To launch an elastic job with `MAX_RESTARTS` failures retries, run the following on at least `MIN_NNODES` nodes and at most `MAX_NNODES` nodes. `RDZV_ID` should be set as a unique job id (shared by all nodes participating in the job). See also [torchrun](https://docs.pytorch.org/docs/stable/elastic/run.html).
```bash ```bash
FORCE_TORCHRUN=1 MIN_NNODES=1 MAX_NNODES=3 MAX_RESTARTS=3 RDZV_ID=llamafactory MASTER_ADDR=192.168.0.1 MASTER_PORT=29500 llamafactory-cli train examples/train_full/llama3_full_sft.yaml FORCE_TORCHRUN=1 MIN_NNODES=1 MAX_NNODES=3 MAX_RESTARTS=3 RDZV_ID=llamafactory MASTER_ADDR=192.168.0.1 MASTER_PORT=29500 llamafactory-cli train examples/train_full/qwen3_full_sft.yaml
``` ```
#### Multimodal Supervised Fine-Tuning #### Multimodal Supervised Fine-Tuning
```bash ```bash
FORCE_TORCHRUN=1 llamafactory-cli train examples/train_full/qwen2_5vl_full_sft.yaml FORCE_TORCHRUN=1 llamafactory-cli train examples/train_full/qwen3vl_full_sft.yaml
``` ```
### Merging LoRA Adapters and Quantization ### Merging LoRA Adapters and Quantization
...@@ -186,19 +174,19 @@ FORCE_TORCHRUN=1 llamafactory-cli train examples/train_full/qwen2_5vl_full_sft.y ...@@ -186,19 +174,19 @@ FORCE_TORCHRUN=1 llamafactory-cli train examples/train_full/qwen2_5vl_full_sft.y
Note: DO NOT use quantized model or `quantization_bit` when merging LoRA adapters. Note: DO NOT use quantized model or `quantization_bit` when merging LoRA adapters.
```bash ```bash
llamafactory-cli export examples/merge_lora/llama3_lora_sft.yaml llamafactory-cli export examples/merge_lora/qwen3_lora_sft.yaml
``` ```
#### Quantizing Model using AutoGPTQ #### Quantizing Model using AutoGPTQ
```bash ```bash
llamafactory-cli export examples/merge_lora/llama3_gptq.yaml llamafactory-cli export examples/merge_lora/qwen3_gptq.yaml
``` ```
### Save Ollama modelfile ### Save Ollama modelfile
```bash ```bash
llamafactory-cli export examples/merge_lora/llama3_full_sft.yaml llamafactory-cli export examples/merge_lora/qwen3_full_sft.yaml
``` ```
### Inferring LoRA Fine-Tuned Models ### Inferring LoRA Fine-Tuned Models
...@@ -206,26 +194,26 @@ llamafactory-cli export examples/merge_lora/llama3_full_sft.yaml ...@@ -206,26 +194,26 @@ llamafactory-cli export examples/merge_lora/llama3_full_sft.yaml
#### Evaluation using vLLM's Multi-GPU Inference #### Evaluation using vLLM's Multi-GPU Inference
``` ```
python scripts/vllm_infer.py --model_name_or_path meta-llama/Meta-Llama-3-8B-Instruct --template llama3 --dataset alpaca_en_demo python scripts/vllm_infer.py --model_name_or_path Qwen/Qwen3-4B-Instruct-2507 --template qwen3_nothink --dataset alpaca_en_demo
python scripts/eval_bleu_rouge.py generated_predictions.jsonl python scripts/eval_bleu_rouge.py generated_predictions.jsonl
``` ```
#### Use CLI ChatBox #### Use CLI ChatBox
```bash ```bash
llamafactory-cli chat examples/inference/llama3_lora_sft.yaml llamafactory-cli chat examples/inference/qwen3_lora_sft.yaml
``` ```
#### Use Web UI ChatBox #### Use Web UI ChatBox
```bash ```bash
llamafactory-cli webchat examples/inference/llama3_lora_sft.yaml llamafactory-cli webchat examples/inference/qwen3_lora_sft.yaml
``` ```
#### Launch OpenAI-style API #### Launch OpenAI-style API
```bash ```bash
llamafactory-cli api examples/inference/llama3_lora_sft.yaml llamafactory-cli api examples/inference/qwen3_lora_sft.yaml
``` ```
### Extras ### Extras
...@@ -290,3 +278,15 @@ llamafactory-cli train examples/extras/llama_pro/llama3_freeze_sft.yaml ...@@ -290,3 +278,15 @@ llamafactory-cli train examples/extras/llama_pro/llama3_freeze_sft.yaml
```bash ```bash
bash examples/extras/fsdp_qlora/train.sh bash examples/extras/fsdp_qlora/train.sh
``` ```
#### OFT Fine-Tuning
```bash
llamafactory-cli train examples/extras/oft/llama3_oft_sft.yaml
```
#### QOFT Fine-Tuning
```bash
llamafactory-cli train examples/extras/qoft/llama3_oft_sft_bnb_npu.yaml
```
examples/README_zh.md
View file @ ca625f43
...@@ -18,19 +18,19 @@ LLaMA-Factory 默认使用所有可见的计算设备。 ...@@ -18,19 +18,19 @@ LLaMA-Factory 默认使用所有可见的计算设备。
基础用法： 基础用法：
```bash ```bash
llamafactory-cli train examples/train_lora/llama3_lora_sft.yaml llamafactory-cli train examples/train_lora/qwen3_lora_sft.yaml
``` ```
高级用法： 高级用法：
```bash ```bash
CUDA_VISIBLE_DEVICES=0,1 llamafactory-cli train examples/train_lora/llama3_lora_sft.yaml \ CUDA_VISIBLE_DEVICES=0,1 llamafactory-cli train examples/train_lora/qwen3_lora_sft.yaml \
learning_rate=1e-5 \ learning_rate=1e-5 \
logging_steps=1 logging_steps=1
``` ```
```bash ```bash
bash examples/train_lora/llama3_lora_sft.sh bash examples/train_lora/qwen3_lora_sft.sh
``` ```
## 示例 ## 示例
...@@ -40,49 +40,43 @@ bash examples/train_lora/llama3_lora_sft.sh ...@@ -40,49 +40,43 @@ bash examples/train_lora/llama3_lora_sft.sh
#### （增量）预训练 #### （增量）预训练
```bash ```bash
llamafactory-cli train examples/train_lora/llama3_lora_pretrain.yaml llamafactory-cli train examples/train_lora/qwen3_lora_pretrain.yaml
``` ```
#### 指令监督微调 #### 指令监督微调
```bash ```bash
llamafactory-cli train examples/train_lora/llama3_lora_sft.yaml llamafactory-cli train examples/train_lora/qwen3_lora_sft.yaml
``` ```
#### 多模态指令监督微调 #### 多模态指令监督微调
```bash ```bash
llamafactory-cli train examples/train_lora/qwen2_5vl_lora_sft.yaml llamafactory-cli train examples/train_lora/qwen3vl_lora_sft.yaml
``` ```
#### DPO/ORPO/SimPO 训练 #### DPO/ORPO/SimPO 训练
```bash ```bash
llamafactory-cli train examples/train_lora/llama3_lora_dpo.yaml llamafactory-cli train examples/train_lora/qwen3_lora_dpo.yaml
``` ```
#### 多模态 DPO/ORPO/SimPO 训练 #### 多模态 DPO/ORPO/SimPO 训练
```bash ```bash
llamafactory-cli train examples/train_lora/qwen2_5vl_lora_dpo.yaml llamafactory-cli train examples/train_lora/qwen3vl_lora_dpo.yaml
``` ```
#### 奖励模型训练 #### 奖励模型训练
```bash ```bash
llamafactory-cli train examples/train_lora/llama3_lora_reward.yaml llamafactory-cli train examples/train_lora/qwen3_lora_reward.yaml
```
#### PPO 训练
```bash
llamafactory-cli train examples/train_lora/llama3_lora_ppo.yaml
``` ```
#### KTO 训练 #### KTO 训练
```bash ```bash
llamafactory-cli train examples/train_lora/llama3_lora_kto.yaml llamafactory-cli train examples/train_lora/qwen3_lora_kto.yaml
``` ```
#### 预处理数据集 #### 预处理数据集
...@@ -90,20 +84,14 @@ llamafactory-cli train examples/train_lora/llama3_lora_kto.yaml ...@@ -90,20 +84,14 @@ llamafactory-cli train examples/train_lora/llama3_lora_kto.yaml
对于大数据集有帮助，在配置中使用 `tokenized_path` 以加载预处理后的数据集。 对于大数据集有帮助，在配置中使用 `tokenized_path` 以加载预处理后的数据集。
```bash ```bash
llamafactory-cli train examples/train_lora/llama3_preprocess.yaml llamafactory-cli train examples/train_lora/qwen3_preprocess.yaml
```
#### 在 MMLU/CMMLU/C-Eval 上评估
```bash
llamafactory-cli eval examples/train_lora/llama3_lora_eval.yaml
``` ```
#### 多机指令监督微调 #### 多机指令监督微调
```bash ```bash
FORCE_TORCHRUN=1 NNODES=2 NODE_RANK=0 MASTER_ADDR=192.168.0.1 MASTER_PORT=29500 llamafactory-cli train examples/train_lora/llama3_lora_sft.yaml FORCE_TORCHRUN=1 NNODES=2 NODE_RANK=0 MASTER_ADDR=192.168.0.1 MASTER_PORT=29500 llamafactory-cli train examples/train_lora/qwen3_lora_sft.yaml
FORCE_TORCHRUN=1 NNODES=2 NODE_RANK=1 MASTER_ADDR=192.168.0.1 MASTER_PORT=29500 llamafactory-cli train examples/train_lora/llama3_lora_sft.yaml FORCE_TORCHRUN=1 NNODES=2 NODE_RANK=1 MASTER_ADDR=192.168.0.1 MASTER_PORT=29500 llamafactory-cli train examples/train_lora/qwen3_lora_sft.yaml
``` ```
### 支持弹性和容错的多机指令监督微调 ### 支持弹性和容错的多机指令监督微调
...@@ -111,19 +99,19 @@ FORCE_TORCHRUN=1 NNODES=2 NODE_RANK=1 MASTER_ADDR=192.168.0.1 MASTER_PORT=29500 ...@@ -111,19 +99,19 @@ FORCE_TORCHRUN=1 NNODES=2 NODE_RANK=1 MASTER_ADDR=192.168.0.1 MASTER_PORT=29500
要启动一个支持弹性节点和容错的多机指令微调，在每个节点上执行以下命令。弹性节点数量范围为 `MIN_NNODES:MAX_NNODES`，每个节点最多允许因为错误重启 `MAX_RESTARTS` 次。`RDZV_ID` 应设置为一个唯一的作业 ID（由参与该作业的所有节点共享）。更多新可以参考官方文档 [torchrun](https://docs.pytorch.org/docs/stable/elastic/run.html)。 要启动一个支持弹性节点和容错的多机指令微调，在每个节点上执行以下命令。弹性节点数量范围为 `MIN_NNODES:MAX_NNODES`，每个节点最多允许因为错误重启 `MAX_RESTARTS` 次。`RDZV_ID` 应设置为一个唯一的作业 ID（由参与该作业的所有节点共享）。更多新可以参考官方文档 [torchrun](https://docs.pytorch.org/docs/stable/elastic/run.html)。
```bash ```bash
FORCE_TORCHRUN=1 MIN_NNODES=1 MAX_NNODES=3 MAX_RESTARTS=3 RDZV_ID=llamafactory MASTER_ADDR=192.168.0.1 MASTER_PORT=29500 llamafactory-cli train examples/train_full/llama3_full_sft.yaml FORCE_TORCHRUN=1 MIN_NNODES=1 MAX_NNODES=3 MAX_RESTARTS=3 RDZV_ID=llamafactory MASTER_ADDR=192.168.0.1 MASTER_PORT=29500 llamafactory-cli train examples/train_full/qwen3_full_sft.yaml
``` ```
#### 使用 DeepSpeed ZeRO-3 平均分配显存 #### 使用 DeepSpeed ZeRO-3 平均分配显存
```bash ```bash
FORCE_TORCHRUN=1 llamafactory-cli train examples/train_lora/llama3_lora_sft_ds3.yaml FORCE_TORCHRUN=1 llamafactory-cli train examples/train_lora/qwen3_lora_sft_ds3.yaml
``` ```
#### 使用 Ray 在 4 张 GPU 上微调 #### 使用 Ray 在 4 张 GPU 上微调
```bash ```bash
USE_RAY=1 llamafactory-cli train examples/train_lora/llama3_lora_sft_ray.yaml USE_RAY=1 llamafactory-cli train examples/train_lora/qwen3_lora_sft_ray.yaml
``` ```
### QLoRA 微调 ### QLoRA 微调
...@@ -131,13 +119,13 @@ USE_RAY=1 llamafactory-cli train examples/train_lora/llama3_lora_sft_ray.yaml ...@@ -131,13 +119,13 @@ USE_RAY=1 llamafactory-cli train examples/train_lora/llama3_lora_sft_ray.yaml
#### 基于 4/8 比特 Bitsandbytes/HQQ/EETQ 量化进行指令监督微调（推荐） #### 基于 4/8 比特 Bitsandbytes/HQQ/EETQ 量化进行指令监督微调（推荐）
```bash ```bash
llamafactory-cli train examples/train_qlora/llama3_lora_sft_otfq.yaml llamafactory-cli train examples/train_qlora/qwen3_lora_sft_otfq.yaml
``` ```
#### 在 NPU 上基于 4 比特 Bitsandbytes 量化进行指令监督微调 #### 在 NPU 上基于 4 比特 Bitsandbytes 量化进行指令监督微调
```bash ```bash
llamafactory-cli train examples/train_qlora/llama3_lora_sft_bnb_npu.yaml llamafactory-cli train examples/train_qlora/qwen3_lora_sft_bnb_npu.yaml
``` ```
#### 基于 4/8 比特 GPTQ 量化进行指令监督微调 #### 基于 4/8 比特 GPTQ 量化进行指令监督微调
...@@ -163,20 +151,20 @@ llamafactory-cli train examples/train_qlora/llama3_lora_sft_aqlm.yaml ...@@ -163,20 +151,20 @@ llamafactory-cli train examples/train_qlora/llama3_lora_sft_aqlm.yaml
#### 在单机上进行指令监督微调 #### 在单机上进行指令监督微调
```bash ```bash
FORCE_TORCHRUN=1 llamafactory-cli train examples/train_full/llama3_full_sft.yaml FORCE_TORCHRUN=1 llamafactory-cli train examples/train_full/qwen3_full_sft.yaml
``` ```
#### 在多机上进行指令监督微调 #### 在多机上进行指令监督微调
```bash ```bash
FORCE_TORCHRUN=1 NNODES=2 NODE_RANK=0 MASTER_ADDR=192.168.0.1 MASTER_PORT=29500 llamafactory-cli train examples/train_full/llama3_full_sft.yaml FORCE_TORCHRUN=1 NNODES=2 NODE_RANK=0 MASTER_ADDR=192.168.0.1 MASTER_PORT=29500 llamafactory-cli train examples/train_full/qwen3_full_sft.yaml
FORCE_TORCHRUN=1 NNODES=2 NODE_RANK=1 MASTER_ADDR=192.168.0.1 MASTER_PORT=29500 llamafactory-cli train examples/train_full/llama3_full_sft.yaml FORCE_TORCHRUN=1 NNODES=2 NODE_RANK=1 MASTER_ADDR=192.168.0.1 MASTER_PORT=29500 llamafactory-cli train examples/train_full/qwen3_full_sft.yaml
``` ```
#### 多模态指令监督微调 #### 多模态指令监督微调
```bash ```bash
FORCE_TORCHRUN=1 llamafactory-cli train examples/train_full/qwen2_5vl_full_sft.yaml FORCE_TORCHRUN=1 llamafactory-cli train examples/train_full/qwen3vl_full_sft.yaml
``` ```
### 合并 LoRA 适配器与模型量化 ### 合并 LoRA 适配器与模型量化
...@@ -186,19 +174,19 @@ FORCE_TORCHRUN=1 llamafactory-cli train examples/train_full/qwen2_5vl_full_sft.y ...@@ -186,19 +174,19 @@ FORCE_TORCHRUN=1 llamafactory-cli train examples/train_full/qwen2_5vl_full_sft.y
注：请勿使用量化后的模型或 `quantization_bit` 参数来合并 LoRA 适配器。 注：请勿使用量化后的模型或 `quantization_bit` 参数来合并 LoRA 适配器。
```bash ```bash
llamafactory-cli export examples/merge_lora/llama3_lora_sft.yaml llamafactory-cli export examples/merge_lora/qwen3_lora_sft.yaml
``` ```
#### 使用 AutoGPTQ 量化模型 #### 使用 AutoGPTQ 量化模型
```bash ```bash
llamafactory-cli export examples/merge_lora/llama3_gptq.yaml llamafactory-cli export examples/merge_lora/qwen3_gptq.yaml
``` ```
### 保存 Ollama 配置文件 ### 保存 Ollama 配置文件
```bash ```bash
llamafactory-cli export examples/merge_lora/llama3_full_sft.yaml llamafactory-cli export examples/merge_lora/qwen3_full_sft.yaml
``` ```
### 推理 LoRA 模型 ### 推理 LoRA 模型
...@@ -206,26 +194,26 @@ llamafactory-cli export examples/merge_lora/llama3_full_sft.yaml ...@@ -206,26 +194,26 @@ llamafactory-cli export examples/merge_lora/llama3_full_sft.yaml
#### 使用 vLLM 多卡推理评估 #### 使用 vLLM 多卡推理评估
``` ```
python scripts/vllm_infer.py --model_name_or_path meta-llama/Meta-Llama-3-8B-Instruct --template llama3 --dataset alpaca_en_demo python scripts/vllm_infer.py --model_name_or_path Qwen/Qwen3-4B-Instruct-2507 --template qwen3_nothink --dataset alpaca_en_demo
python scripts/eval_bleu_rouge.py generated_predictions.jsonl python scripts/eval_bleu_rouge.py generated_predictions.jsonl
``` ```
#### 使用命令行对话框 #### 使用命令行对话框
```bash ```bash
llamafactory-cli chat examples/inference/llama3_lora_sft.yaml llamafactory-cli chat examples/inference/qwen3_lora_sft.yaml
``` ```
#### 使用浏览器对话框 #### 使用浏览器对话框
```bash ```bash
llamafactory-cli webchat examples/inference/llama3_lora_sft.yaml llamafactory-cli webchat examples/inference/qwen3_lora_sft.yaml
``` ```
#### 启动 OpenAI 风格 API #### 启动 OpenAI 风格 API
```bash ```bash
llamafactory-cli api examples/inference/llama3_lora_sft.yaml llamafactory-cli api examples/inference/qwen3_lora_sft.yaml
``` ```
### 杂项 ### 杂项
...@@ -290,3 +278,15 @@ llamafactory-cli train examples/extras/llama_pro/llama3_freeze_sft.yaml ...@@ -290,3 +278,15 @@ llamafactory-cli train examples/extras/llama_pro/llama3_freeze_sft.yaml
```bash ```bash
bash examples/extras/fsdp_qlora/train.sh bash examples/extras/fsdp_qlora/train.sh
``` ```
#### OFT 微调
```bash
llamafactory-cli train examples/extras/oft/llama3_oft_sft.yaml
```
#### QOFT 微调
```bash
llamafactory-cli train examples/extras/qoft/llama3_oft_sft_bnb_npu.yaml
```
examples/accelerate/fsdp2_config.yaml 0 → 100644
View file @ ca625f43
compute_environment: LOCAL_MACHINE
debug: false
distributed_type: FSDP
downcast_bf16: 'no'
fsdp_config:
fsdp_auto_wrap_policy: TRANSFORMER_BASED_WRAP
fsdp_cpu_ram_efficient_loading: true
fsdp_offload_params: false
fsdp_reshard_after_forward: true
fsdp_state_dict_type: FULL_STATE_DICT
fsdp_version: 2
machine_rank: 0
main_training_function: main
mixed_precision: bf16 # or fp16
num_machines: 1 # the number of nodes
num_processes: 2 # the number of GPUs in all nodes
rdzv_backend: static
same_network: true
tpu_env: []
tpu_use_cluster: false
tpu_use_sudo: false
use_cpu: false
examples/accelerate/fsdp_config_multiple_nodes.yaml 0 → 100644
View file @ ca625f43
# If you want to run this example on multiple nodes, you need to set the following parameters:
# - num_machines: the number of nodes
# - num_processes: the number of GPUs in all nodes, num_machines * num_processes_per_machine
# - main_process_ip: the IP address of the main process, please keep it the same across all nodes
# - main_process_port: the port of all nodes, please keep it the same across all nodes
# - machine_rank: the rank of the current machine, starting from 0, and it should be 0 for main_process_ip
compute_environment: LOCAL_MACHINE
debug: false
distributed_type: FSDP
downcast_bf16: 'no'
fsdp_config:
fsdp_auto_wrap_policy: TRANSFORMER_BASED_WRAP
fsdp_backward_prefetch: BACKWARD_PRE
fsdp_forward_prefetch: false
fsdp_cpu_ram_efficient_loading: true
fsdp_offload_params: false
fsdp_sharding_strategy: FULL_SHARD
fsdp_state_dict_type: FULL_STATE_DICT
fsdp_sync_module_states: true
fsdp_use_orig_params: true
machine_rank: 0
main_training_function: main
mixed_precision: bf16 # or fp16
main_process_ip: 192.168.0.1
main_process_port: 29500
num_machines: 2 # the number of nodes
num_processes: 16 # the number of GPUs in all nodes, num_machines * num_processes_per_machine
rdzv_backend: static
same_network: true
tpu_env: []
tpu_use_cluster: false
tpu_use_sudo: false
use_cpu: false
examples/ascend/qwen3_full_sft_fsdp2.yaml 0 → 100644
View file @ ca625f43
# Start FSDP2 fine-tuning
# accelerate launch \
# --config_file examples/accelerate/fsdp2_config.yaml \
# src/train.py examples/ascend/qwen3_full_sft_fsdp2.yaml
# Change `num_processes` in fsdp2_config.yaml to 16 in A3
### model
model_name_or_path: Qwen/Qwen3-8B
trust_remote_code: true
use_v1_kernels: true
flash_attn: fa2
### method
stage: sft
do_train: true
finetuning_type: full
### dataset
dataset: alpaca_en_demo
template: qwen3
cutoff_len: 2048
max_samples: 1000
overwrite_cache: true
preprocessing_num_workers: 16
dataloader_num_workers: 4
### output
output_dir: saves/Qwen3-8B/full/sft
logging_steps: 1
save_steps: 500
max_steps: 500
plot_loss: true
overwrite_output_dir: true
save_only_model: false
report_to: none # choices: [none, wandb, tensorboard, swanlab, mlflow]
### train
per_device_train_batch_size: 8
gradient_accumulation_steps: 1
learning_rate: 1.0e-5
lr_scheduler_type: cosine
warmup_ratio: 0.1
bf16: true
ddp_timeout: 1800
resume_from_checkpoint: null
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment