[Dataset] Add Smolinstruct configs (#2127)

* 0-shot Smolinstruct

Add 0-shot evaluation and postprocess functions for Smolinstruct

* fix acc postprocessor

* update 0-shot acc postprocessor

* rename 0-shot

.github/scripts/eval_regression_api.py

@@ -0,0 +1,42 @@
+from mmengine.config import read_base
+
+from opencompass.models.openai_api import OpenAISDK
+
+with read_base():
+    # choose a list of datasets
+    from opencompass.configs.datasets.gsm8k.gsm8k_gen import \
+        gsm8k_datasets  # noqa: F401, E501
+    from opencompass.configs.datasets.race.race_gen import \
+        race_datasets  # noqa: F401, E501
+
+datasets = sum([v for k, v in locals().items() if k.endswith('_datasets')], [])
+
+api_meta_template = dict(
+    round=[
+        dict(role='HUMAN', api_role='HUMAN'),
+        dict(role='BOT', api_role='BOT', generate=True),
+    ],
+    reserved_roles=[dict(role='SYSTEM', api_role='SYSTEM')],
+)
+
+models = [
+    dict(
+        abbr='lmdeploy-api-test',
+        type=OpenAISDK,
+        key='EMPTY',
+        openai_api_base='http://localhost:23333/v1',
+        path='internlm3',
+        tokenizer_path='internlm/internlm3-8b-instruct',
+        rpm_verbose=True,
+        meta_template=api_meta_template,
+        query_per_second=128,
+        max_out_len=1024,
+        max_seq_len=4096,
+        temperature=0.01,
+        batch_size=128,
+        retry=20,
+    )
+]
+
+for d in datasets:
+    d['reader_cfg']['test_range'] = '[0:16]'

.github/scripts/eval_regression_base_fullbench.py

@@ -0,0 +1,210 @@
+from mmengine.config import read_base
+
+with read_base():
+    from opencompass.configs.datasets.ARC_c.ARC_c_few_shot_ppl import \
+        ARC_c_datasets  # noqa: F401, E501
+    from opencompass.configs.datasets.bbh.bbh_gen_98fba6 import \
+        bbh_datasets  # noqa: F401, E501
+    from opencompass.configs.datasets.cmmlu.cmmlu_ppl_041cbf import \
+        cmmlu_datasets  # noqa: F401, E501
+    from opencompass.configs.datasets.dingo.dingo_gen import \
+        datasets as dingo_datasets  # noqa: F401, E501
+    from opencompass.configs.datasets.drop.drop_gen_a2697c import \
+        drop_datasets  # noqa: F401, E501
+    from opencompass.configs.datasets.GaokaoBench.GaokaoBench_no_subjective_gen_d21e37 import \
+        GaokaoBench_datasets  # noqa: F401, E501
+    from opencompass.configs.datasets.gpqa.gpqa_few_shot_ppl_4b5a83 import \
+        gpqa_datasets  # noqa: F401, E501
+    # Corebench v1.7
+    from opencompass.configs.datasets.gsm8k.gsm8k_gen_17d0dc import \
+        gsm8k_datasets  # noqa: F401, E501
+    from opencompass.configs.datasets.hellaswag.hellaswag_10shot_ppl_59c85e import \
+        hellaswag_datasets  # noqa: F401, E501
+    from opencompass.configs.datasets.humaneval.internal_humaneval_gen_ce6b06 import \
+        humaneval_datasets as humaneval_v2_datasets  # noqa: F401, E501
+    from opencompass.configs.datasets.humaneval.internal_humaneval_gen_d2537e import \
+        humaneval_datasets  # noqa: F401, E501
+    from opencompass.configs.datasets.math.math_4shot_base_gen_43d5b6 import \
+        math_datasets  # noqa: F401, E501
+    from opencompass.configs.datasets.MathBench.mathbench_2024_few_shot_mixed_4a3fd4 import \
+        mathbench_datasets  # noqa: F401, E501
+    from opencompass.configs.datasets.mbpp.sanitized_mbpp_gen_742f0c import \
+        sanitized_mbpp_datasets  # noqa: F401, E501
+    from opencompass.configs.datasets.mmlu.mmlu_ppl_ac766d import \
+        mmlu_datasets  # noqa: F401, E501
+    from opencompass.configs.datasets.mmlu_pro.mmlu_pro_few_shot_gen_bfaf90 import \
+        mmlu_pro_datasets  # noqa: F401, E501
+    from opencompass.configs.datasets.nq.nq_open_1shot_gen_20a989 import \
+        nq_datasets  # noqa: F401, E501
+    from opencompass.configs.datasets.race.race_few_shot_ppl import \
+        race_datasets  # noqa: F401, E501
+    from opencompass.configs.datasets.SuperGLUE_BoolQ.SuperGLUE_BoolQ_few_shot_ppl import \
+        BoolQ_datasets  # noqa: F401, E501
+    from opencompass.configs.datasets.TheoremQA.TheoremQA_5shot_gen_6f0af8 import \
+        TheoremQA_datasets  # noqa: F401, E501
+    from opencompass.configs.datasets.triviaqa.triviaqa_wiki_1shot_gen_20a989 import \
+        triviaqa_datasets  # noqa: F401, E501
+    from opencompass.configs.datasets.wikibench.wikibench_few_shot_ppl_c23d79 import \
+        wikibench_datasets  # noqa: F401, E501
+    from opencompass.configs.datasets.winogrande.winogrande_5shot_ll_252f01 import \
+        winogrande_datasets  # noqa: F401, E501
+    from opencompass.configs.models.hf_internlm.hf_internlm2_5_7b import \
+        models as hf_internlm2_5_7b_model  # noqa: F401, E501
+    from opencompass.configs.models.hf_internlm.lmdeploy_internlm2_5_7b import \
+        models as lmdeploy_internlm2_5_7b_model  # noqa: F401, E501
+    from opencompass.configs.summarizers.groups.bbh import \
+        bbh_summary_groups  # noqa: F401, E501
+    # Summary Groups
+    from opencompass.configs.summarizers.groups.cmmlu import \
+        cmmlu_summary_groups  # noqa: F401, E501
+    from opencompass.configs.summarizers.groups.GaokaoBench import \
+        GaokaoBench_summary_groups  # noqa: F401, E501
+    from opencompass.configs.summarizers.groups.mathbench_v1_2024 import \
+        mathbench_2024_summary_groups  # noqa: F401, E501
+    from opencompass.configs.summarizers.groups.mmlu import \
+        mmlu_summary_groups  # noqa: F401, E501
+    from opencompass.configs.summarizers.groups.mmlu_pro import \
+        mmlu_pro_summary_groups  # noqa: F401, E501
+
+    from ...volc import infer as volc_infer  # noqa: F401, E501
+
+race_datasets = [race_datasets[1]]  # Only take RACE-High
+humaneval_v2_datasets[0]['abbr'] = 'openai_humaneval_v2'
+bbh_datasets = [
+    x for x in bbh_datasets if 'logical_deduction_seven_objects' in x['abbr']
+    or 'multistep_arithmetic_two' in x['abbr']
+]
+cmmlu_datasets = [
+    x for x in cmmlu_datasets if x['abbr'].replace('cmmlu-', '') in [
+        'ancient_chinese', 'chinese_civil_service_exam',
+        'chinese_driving_rule', 'chinese_food_culture',
+        'chinese_foreign_policy', 'chinese_history', 'chinese_literature',
+        'chinese_teacher_qualification', 'construction_project_management',
+        'elementary_chinese', 'elementary_commonsense', 'ethnology',
+        'high_school_politics', 'modern_chinese',
+        'traditional_chinese_medicine'
+    ]
+]
+mmlu_datasets = [
+    x for x in mmlu_datasets if x['abbr'].replace('lukaemon_mmlu_', '') in [
+        'business_ethics', 'clinical_knowledge', 'college_medicine',
+        'global_facts', 'human_aging', 'management', 'marketing',
+        'medical_genetics', 'miscellaneous', 'nutrition',
+        'professional_accounting', 'professional_medicine', 'virology'
+    ]
+]
+mmlu_pro_datasets = [mmlu_pro_datasets[0]]
+mathbench_datasets = [x for x in mathbench_datasets if 'college' in x['abbr']]
+GaokaoBench_datasets = [
+    x for x in GaokaoBench_datasets if '2010-2022_Math_II_MCQs' in x['abbr']
+    or '2010-2022_Math_II_Fill-in-the-Blank' in x['abbr']
+]
+datasets = sum((v for k, v in locals().items() if k.endswith('_datasets')), [])
+
+summary_groups = sum(
+    [v for k, v in locals().items() if k.endswith('_summary_groups')], [])
+summary_groups.append(
+    {
+        'name': 'Mathbench',
+        'subsets': ['mathbench-a (average)', 'mathbench-t (average)'],
+    }, )
+
+summarizer = dict(
+    dataset_abbrs=[
+        'Language',
+        ['race-high', 'accuracy'],
+        ['ARC-c', 'accuracy'],
+        ['BoolQ', 'accuracy'],
+        ['triviaqa_wiki_1shot', 'score'],
+        ['nq_open_1shot', 'score'],
+        '',
+        'General Reasoning',
+        ['drop', 'accuracy'],
+        ['bbh', 'naive_average'],
+        ['GPQA_diamond', 'accuracy'],
+        ['hellaswag', 'accuracy'],
+        ['TheoremQA', 'score'],
+        ['winogrande', 'accuracy'],
+        '',
+        'Math Calculation',
+        ['gsm8k', 'accuracy'],
+        ['GaokaoBench', 'weighted_average'],
+        'GaokaoBench_2010-2022_Math_II_MCQs',
+        'GaokaoBench_2010-2022_Math_II_Fill-in-the-Blank',
+        ['math', 'accuracy'],
+        ['Mathbench', 'naive_average'],
+        '',
+        'Knowledge',
+        ['wikibench-wiki-single_choice_cncircular', 'perf_4'],
+        ['cmmlu', 'naive_average'],
+        ['mmlu', 'naive_average'],
+        ['mmlu_pro', 'naive_average'],
+        '',
+        'Code',
+        ['openai_humaneval', 'humaneval_pass@1'],
+        ['openai_humaneval_v2', 'humaneval_pass@1'],
+        ['sanitized_mbpp', 'score'],
+        '',
+        ['dingo_en_192', 'score'],
+        ['dingo_zh_170', 'score'],
+        '',
+        'mmlu',
+        'mmlu-stem',
+        'mmlu-social-science',
+        'mmlu-humanities',
+        ['mmlu-other', 'accuracy'],
+        '',
+        'cmmlu',
+        'cmmlu-stem',
+        'cmmlu-social-science',
+        'cmmlu-humanities',
+        'cmmlu-other',
+        ['cmmlu-china-specific', 'accuracy'],
+        '',
+        'mmlu_pro',
+        'mmlu_pro_biology',
+        'mmlu_pro_business',
+        'mmlu_pro_chemistry',
+        'mmlu_pro_computer_science',
+        'mmlu_pro_economics',
+        'mmlu_pro_engineering',
+        'mmlu_pro_health',
+        'mmlu_pro_history',
+        'mmlu_pro_law',
+        'mmlu_pro_math',
+        'mmlu_pro_philosophy',
+        'mmlu_pro_physics',
+        'mmlu_pro_psychology',
+        'mmlu_pro_other',
+        '',
+        'bbh-logical_deduction_seven_objects',
+        'bbh-multistep_arithmetic_two',
+        '###### MathBench-A: Application Part ######',
+        'college',
+        'high',
+        'middle',
+        'primary',
+        'arithmetic',
+        'mathbench-a (average)',
+        '###### MathBench-T: Theory Part ######',
+        'college_knowledge',
+        'high_knowledge',
+        'middle_knowledge',
+        'primary_knowledge',
+        'mathbench-t (average)',
+    ],
+    summary_groups=summary_groups,
+)
+
+models = sum([v for k, v in locals().items() if k.endswith('_model')], [])
+datasets = sum([v for k, v in locals().items() if k.endswith('_datasets')], [])
+
+for d in datasets:
+    d['reader_cfg']['test_range'] = '[0:16]'
+
+for m in models:
+    m['abbr'] = m['abbr'] + '_fullbench'
+    if 'turbomind' in m['abbr'] or 'lmdeploy' in m['abbr']:
+        m['engine_config']['max_batch_size'] = 1
+        m['batch_size'] = 1
+models = sorted(models, key=lambda x: x['run_cfg']['num_gpus'])

.github/scripts/eval_regression_base_models.py

@@ -0,0 +1,129 @@
+from mmengine.config import read_base
+
+with read_base():
+    # choose a list of datasets
+    from opencompass.configs.datasets.gpqa.gpqa_openai_simple_evals_gen_5aeece import \
+        gpqa_datasets  # noqa: F401, E501
+    from opencompass.configs.datasets.gsm8k.gsm8k_gen_17d0dc import \
+        gsm8k_datasets  # noqa: F401, E501
+    from opencompass.configs.datasets.race.race_ppl import \
+        race_datasets  # noqa: F401, E501
+    from opencompass.configs.datasets.winogrande.winogrande_5shot_ll_252f01 import \
+        winogrande_datasets  # noqa: F401, E501
+    # read hf models - chat models
+    from opencompass.configs.models.chatglm.lmdeploy_glm4_9b import \
+        models as lmdeploy_glm4_9b_model  # noqa: F401, E501
+    from opencompass.configs.models.deepseek.hf_deepseek_7b_base import \
+        models as hf_deepseek_7b_base_model  # noqa: F401, E501
+    from opencompass.configs.models.deepseek.lmdeploy_deepseek_7b_base import \
+        models as lmdeploy_deepseek_7b_base_model  # noqa: F401, E501
+    from opencompass.configs.models.deepseek.lmdeploy_deepseek_67b_base import \
+        models as lmdeploy_deepseek_67b_base_model  # noqa: F401, E501
+    from opencompass.configs.models.deepseek.lmdeploy_deepseek_v2 import \
+        lmdeploy_deepseek_v2_model  # noqa: F401, E501
+    from opencompass.configs.models.deepseek.vllm_deepseek_moe_16b_base import \
+        models as vllm_deepseek_moe_16b_base_model  # noqa: F401, E501
+    from opencompass.configs.models.gemma.hf_gemma2_2b import \
+        models as hf_gemma2_2b_model  # noqa: F401, E501
+    from opencompass.configs.models.gemma.hf_gemma2_9b import \
+        models as hf_gemma2_9b_model  # noqa: F401, E501
+    from opencompass.configs.models.gemma.hf_gemma_2b import \
+        models as hf_gemma_2b_model  # noqa: F401, E501
+    from opencompass.configs.models.gemma.hf_gemma_7b import \
+        models as hf_gemma_7b_model  # noqa: F401, E501
+    from opencompass.configs.models.gemma.lmdeploy_gemma_9b import \
+        models as lmdeploy_gemma_9b_model  # noqa: F401, E501
+    from opencompass.configs.models.gemma.vllm_gemma_2b import \
+        models as vllm_gemma_2b_model  # noqa: F401, E501
+    from opencompass.configs.models.gemma.vllm_gemma_7b import \
+        models as vllm_gemma_7b_model  # noqa: F401, E501
+    from opencompass.configs.models.hf_internlm.hf_internlm2_5_7b import \
+        models as hf_internlm2_5_7b_model  # noqa: F401, E501
+    from opencompass.configs.models.hf_internlm.hf_internlm2_7b import \
+        models as hf_internlm2_7b_model  # noqa: F401, E501
+    from opencompass.configs.models.hf_internlm.lmdeploy_internlm2_1_8b import \
+        models as lmdeploy_internlm2_1_8b_model  # noqa: F401, E501
+    from opencompass.configs.models.hf_internlm.lmdeploy_internlm2_5_7b import \
+        models as lmdeploy_internlm2_5_7b_model  # noqa: F401, E501
+    from opencompass.configs.models.hf_internlm.lmdeploy_internlm2_7b import \
+        models as lmdeploy_internlm2_7b_model  # noqa: F401, E501
+    from opencompass.configs.models.hf_internlm.lmdeploy_internlm2_20b import \
+        models as lmdeploy_internlm2_20b_model  # noqa: F401, E501
+    from opencompass.configs.models.hf_internlm.lmdeploy_internlm2_base_7b import \
+        models as lmdeploy_internlm2_base_7b_model  # noqa: F401, E501
+    from opencompass.configs.models.hf_internlm.lmdeploy_internlm2_base_20b import \
+        models as lmdeploy_internlm2_base_20b_model  # noqa: F401, E501
+    from opencompass.configs.models.hf_llama.hf_llama2_7b import \
+        models as hf_llama2_7b_model  # noqa: F401, E501
+    from opencompass.configs.models.hf_llama.hf_llama3_1_8b import \
+        models as hf_llama3_1_8b_model  # noqa: F401, E501
+    from opencompass.configs.models.hf_llama.hf_llama3_8b import \
+        models as hf_llama3_8b_model  # noqa: F401, E501
+    from opencompass.configs.models.hf_llama.lmdeploy_llama3_1_8b import \
+        models as lmdeploy_llama3_1_8b_model  # noqa: F401, E501
+    from opencompass.configs.models.hf_llama.lmdeploy_llama3_8b import \
+        models as lmdeploy_llama3_8b_model  # noqa: F401, E501
+    from opencompass.configs.models.hf_llama.lmdeploy_llama3_70b import \
+        models as lmdeploy_llama3_70b_model  # noqa: F401, E501
+    from opencompass.configs.models.mistral.hf_mistral_7b_v0_3 import \
+        models as hf_mistral_7b_v0_3_model  # noqa: F401, E501
+    from opencompass.configs.models.qwen2_5.hf_qwen_2_5_7b import \
+        models as hf_qwen_2_5_7b_model  # noqa: F401, E501
+    from opencompass.configs.models.qwen2_5.hf_qwen_2_5_14b import \
+        models as hf_qwen_2_5_14b_model  # noqa: F401, E501
+    from opencompass.configs.models.qwen2_5.hf_qwen_2_5_32b import \
+        models as hf_qwen_2_5_32b_model  # noqa: F401, E501
+    from opencompass.configs.models.qwen2_5.lmdeploy_qwen2_5_1_5b import \
+        models as lmdeploy_qwen2_5_1_5b_model  # noqa: F401, E501
+    from opencompass.configs.models.qwen2_5.lmdeploy_qwen2_5_7b import \
+        models as lmdeploy_qwen2_5_7b_model  # noqa: F401, E501
+    from opencompass.configs.models.qwen2_5.lmdeploy_qwen2_5_32b import \
+        models as lmdeploy_qwen2_5_32b_model  # noqa: F401, E501
+    from opencompass.configs.models.qwen2_5.lmdeploy_qwen2_5_72b import \
+        models as lmdeploy_qwen2_5_72b_model  # noqa: F401, E501
+    from opencompass.configs.models.qwen.hf_qwen1_5_moe_a2_7b import \
+        models as hf_qwen1_5_moe_a2_7b_model  # noqa: F401, E501
+    from opencompass.configs.models.qwen.hf_qwen2_0_5b import \
+        models as hf_qwen2_0_5b_model  # noqa: F401, E501
+    from opencompass.configs.models.qwen.hf_qwen2_1_5b import \
+        models as hf_qwen2_1_5b_model  # noqa: F401, E501
+    from opencompass.configs.models.qwen.hf_qwen2_7b import \
+        models as hf_qwen2_7b_model  # noqa: F401, E501
+    from opencompass.configs.models.qwen.lmdeploy_qwen2_1_5b import \
+        models as lmdeploy_qwen2_1_5b_model  # noqa: F401, E501
+    from opencompass.configs.models.qwen.lmdeploy_qwen2_7b import \
+        models as lmdeploy_qwen2_7b_model  # noqa: F401, E501
+    from opencompass.configs.models.qwen.vllm_qwen1_5_0_5b import \
+        models as vllm_qwen1_5_0_5b_model  # noqa: F401, E501
+    from opencompass.configs.models.yi.hf_yi_1_5_6b import \
+        models as hf_yi_1_5_6b_model  # noqa: F401, E501
+    from opencompass.configs.models.yi.hf_yi_1_5_9b import \
+        models as hf_yi_1_5_9b_model  # noqa: F401, E501
+    from opencompass.configs.models.yi.lmdeploy_yi_1_5_9b import \
+        models as lmdeploy_yi_1_5_9b_model  # noqa: F401, E501
+
+    from ...volc import infer as volc_infer  # noqa: F401, E501
+
+race_datasets = [race_datasets[1]]
+models = sum([v for k, v in locals().items() if k.endswith('_model')], [])
+datasets = sum([v for k, v in locals().items() if k.endswith('_datasets')], [])
+
+for d in datasets:
+    d['reader_cfg']['test_range'] = '[0:32]'
+
+for m in models:
+    if 'turbomind' in m['abbr'] or 'lmdeploy' in m['abbr']:
+        m['engine_config']['max_batch_size'] = 1
+        m['batch_size'] = 1
+models = sorted(models, key=lambda x: x['run_cfg']['num_gpus'])
+
+summarizer = dict(
+    dataset_abbrs=[
+        ['gsm8k', 'accuracy'],
+        ['GPQA_diamond', 'accuracy'],
+        ['race-high', 'accuracy'],
+        ['winogrande', 'accuracy'],
+    ],
+    summary_groups=sum(
+        [v for k, v in locals().items() if k.endswith('_summary_groups')], []),
+)

.github/scripts/eval_regression_chat_models.py

@@ -0,0 +1,193 @@
+from mmengine.config import read_base
+
+with read_base():
+    # choose a list of datasets
+    from opencompass.configs.datasets.gsm8k.gsm8k_gen import \
+        gsm8k_datasets  # noqa: F401, E501
+    from opencompass.configs.datasets.race.race_gen import \
+        race_datasets  # noqa: F401, E501
+    # read hf models - chat models
+    from opencompass.configs.models.chatglm.hf_glm4_9b_chat import \
+        models as hf_glm4_9b_chat_model  # noqa: F401, E501
+    from opencompass.configs.models.chatglm.lmdeploy_glm4_9b_chat import \
+        models as lmdeploy_glm4_9b_chat_model  # noqa: F401, E501
+    from opencompass.configs.models.chatglm.vllm_glm4_9b_chat import \
+        models as vllm_glm4_9b_chat_model  # noqa: F401, E501
+    from opencompass.configs.models.deepseek.hf_deepseek_7b_chat import \
+        models as hf_deepseek_7b_chat_model  # noqa: F401, E501
+    from opencompass.configs.models.deepseek.lmdeploy_deepseek_67b_chat import \
+        models as lmdeploy_deepseek_67b_chat_model  # noqa: F401, E501
+    from opencompass.configs.models.deepseek.lmdeploy_deepseek_r1_distill_llama_8b import \
+        models as \
+        lmdeploy_deepseek_r1_distill_llama_8b_model  # noqa: F401, E501
+    from opencompass.configs.models.deepseek.lmdeploy_deepseek_r1_distill_llama_70b import \
+        models as \
+        lmdeploy_deepseek_r1_distill_llama_70b_model  # noqa: F401, E501
+    from opencompass.configs.models.deepseek.lmdeploy_deepseek_r1_distill_qwen_1_5b import \
+        models as \
+        lmdeploy_deepseek_r1_distill_qwen_1_5b_model  # noqa: F401, E501
+    from opencompass.configs.models.deepseek.lmdeploy_deepseek_r1_distill_qwen_32b import \
+        models as \
+        lmdeploy_deepseek_r1_distill_qwen_32b_model  # noqa: F401, E501
+    from opencompass.configs.models.deepseek.lmdeploy_deepseek_v2_5_1210 import \
+        models as lmdeploy_deepseek_v2_5_1210_model  # noqa: F401, E501
+    from opencompass.configs.models.deepseek.lmdeploy_deepseek_v2_lite import \
+        models as lmdeploy_deepseek_v2_lite_model  # noqa: F401, E501
+    from opencompass.configs.models.deepseek.vllm_deepseek_7b_chat import \
+        models as vllm_deepseek_7b_chat_model  # noqa: F401, E501
+    from opencompass.configs.models.gemma.hf_gemma2_2b_it import \
+        models as hf_gemma2_2b_it_model  # noqa: F401, E501
+    from opencompass.configs.models.gemma.hf_gemma2_9b_it import \
+        models as hf_gemma2_9b_it_model  # noqa: F401, E501
+    from opencompass.configs.models.gemma.hf_gemma2_27b_it import \
+        models as hf_gemma2_27b_it_model  # noqa: F401, E501
+    from opencompass.configs.models.gemma.hf_gemma_2b_it import \
+        models as hf_gemma_2b_it_model  # noqa: F401, E501
+    from opencompass.configs.models.gemma.hf_gemma_7b_it import \
+        models as hf_gemma_7b_it_model  # noqa: F401, E501
+    from opencompass.configs.models.gemma.lmdeploy_gemma_9b_it import \
+        models as lmdeploy_gemma_9b_it_model  # noqa: F401, E501
+    from opencompass.configs.models.gemma.lmdeploy_gemma_27b_it import \
+        models as lmdeploy_gemma_27b_it_model  # noqa: F401, E501
+    from opencompass.configs.models.gemma.vllm_gemma_7b_it import \
+        models as vllm_gemma_7b_it_model  # noqa: F401, E501
+    from opencompass.configs.models.hf_internlm.hf_internlm2_5_7b_chat import \
+        models as hf_internlm2_5_7b_chat_model  # noqa: F401, E501
+    from opencompass.configs.models.hf_internlm.hf_internlm2_5_20b_chat import \
+        models as hf_internlm2_5_20b_chat_model  # noqa: F401, E501
+    from opencompass.configs.models.hf_internlm.hf_internlm3_8b_instruct import \
+        models as hf_internlm3_8b_instruct_model  # noqa: F401, E501
+    from opencompass.configs.models.hf_internlm.lmdeploy_internlm2_5_7b_chat import \
+        models as lmdeploy_internlm2_5_7b_chat_model  # noqa: F401, E501
+    from opencompass.configs.models.hf_internlm.lmdeploy_internlm2_5_20b_chat import \
+        models as lmdeploy_internlm2_5_20b_chat_model  # noqa: F401, E501
+    from opencompass.configs.models.hf_internlm.lmdeploy_internlm2_chat_1_8b import \
+        models as lmdeploy_internlm2_chat_1_8b_model  # noqa: F401, E501
+    from opencompass.configs.models.hf_internlm.lmdeploy_internlm2_chat_1_8b_sft import \
+        models as lmdeploy_internlm2_chat_1_8b_sft_model  # noqa: F401, E501
+    from opencompass.configs.models.hf_internlm.lmdeploy_internlm2_chat_7b import \
+        models as lmdeploy_internlm2_chat_7b_model  # noqa: F401, E501
+    from opencompass.configs.models.hf_internlm.lmdeploy_internlm2_chat_7b_sft import \
+        models as lmdeploy_internlm2_chat_7b_sft_model  # noqa: F401, E501
+    from opencompass.configs.models.hf_internlm.lmdeploy_internlm3_8b_instruct import \
+        models as lmdeploy_internlm3_8b_instruct_model  # noqa: F401, E501
+    from opencompass.configs.models.hf_internlm.vllm_internlm2_chat_7b import \
+        models as vllm_internlm2_chat_7b_model  # noqa: F401, E501
+    from opencompass.configs.models.hf_llama.hf_llama3_1_8b_instruct import \
+        models as hf_llama3_1_8b_instruct_model  # noqa: F401, E501
+    from opencompass.configs.models.hf_llama.hf_llama3_2_3b_instruct import \
+        models as hf_llama3_2_3b_instruct_model  # noqa: F401, E501
+    from opencompass.configs.models.hf_llama.hf_llama3_8b_instruct import \
+        models as hf_llama3_8b_instruct_model  # noqa: F401, E501
+    from opencompass.configs.models.hf_llama.lmdeploy_llama2_7b_chat import \
+        models as lmdeploy_llama2_7b_chat_model  # noqa: F401, E501
+    from opencompass.configs.models.hf_llama.lmdeploy_llama3_1_8b_instruct import \
+        models as lmdeploy_llama3_1_8b_instruct_model  # noqa: F401, E501
+    from opencompass.configs.models.hf_llama.lmdeploy_llama3_2_3b_instruct import \
+        models as lmdeploy_llama3_2_3b_instruct_model  # noqa: F401, E501
+    from opencompass.configs.models.hf_llama.lmdeploy_llama3_3_70b_instruct import \
+        models as lmdeploy_llama3_3_70b_instruct_model  # noqa: F401, E501
+    from opencompass.configs.models.hf_llama.lmdeploy_llama3_8b_instruct import \
+        models as lmdeploy_llama3_8b_instruct_model  # noqa: F401, E501
+    from opencompass.configs.models.mistral.hf_mistral_7b_instruct_v0_2 import \
+        models as hf_mistral_7b_instruct_v0_2_model  # noqa: F401, E501
+    from opencompass.configs.models.mistral.hf_mistral_7b_instruct_v0_3 import \
+        models as hf_mistral_7b_instruct_v0_3_model  # noqa: F401, E501
+    from opencompass.configs.models.mistral.hf_mistral_nemo_instruct_2407 import \
+        models as hf_mistral_nemo_instruct_2407_model  # noqa: F401, E501
+    from opencompass.configs.models.mistral.hf_mistral_small_instruct_2409 import \
+        models as hf_mistral_small_instruct_2409_model  # noqa: F401, E501
+    from opencompass.configs.models.mistral.lmdeploy_mistral_large_instruct_2411 import \
+        models as \
+        lmdeploy_mistral_large_instruct_2411_model  # noqa: F401, E501
+    from opencompass.configs.models.mistral.lmdeploy_mistral_nemo_instruct_2407 import \
+        models as lmdeploy_mistral_nemo_instruct_2407_model  # noqa: F401, E501
+    from opencompass.configs.models.mistral.lmdeploy_mistral_small_instruct_2409 import \
+        models as \
+        lmdeploy_mistral_small_instruct_2409_model  # noqa: F401, E501
+    from opencompass.configs.models.mistral.lmdeploy_mixtral_8x22b_instruct_v0_1 import \
+        models as \
+        lmdeploy_mixtral_8x22b_instruct_v0_1_model  # noqa: F401, E501
+    from opencompass.configs.models.mistral.vllm_mistral_7b_instruct_v0_1 import \
+        models as vllm_mistral_7b_instruct_v0_1_model  # noqa: F401, E501
+    from opencompass.configs.models.mistral.vllm_mistral_7b_instruct_v0_2 import \
+        models as vllm_mistral_7b_instruct_v0_2_model  # noqa: F401, E501
+    from opencompass.configs.models.mistral.vllm_mixtral_8x22b_instruct_v0_1 import \
+        models as vllm_mixtral_8x22b_instruct_v0_1_model  # noqa: F401, E501
+    from opencompass.configs.models.nvidia.lmdeploy_nemotron_70b_instruct_hf import \
+        models as lmdeploy_nemotron_70b_instruct_hf_model  # noqa: F401, E501
+    from opencompass.configs.models.phi.hf_phi_4 import \
+        models as hf_phi_4_model  # noqa: F401, E501
+    from opencompass.configs.models.qwen2_5.hf_qwen2_5_0_5b_instruct import \
+        models as hf_qwen2_5_0_5b_instruct_model  # noqa: F401, E501
+    from opencompass.configs.models.qwen2_5.hf_qwen2_5_3b_instruct import \
+        models as hf_qwen2_5_3b_instruct_model  # noqa: F401, E501
+    from opencompass.configs.models.qwen2_5.hf_qwen2_5_14b_instruct import \
+        models as hf_qwen2_5_14b_instruct_model  # noqa: F401, E501
+    from opencompass.configs.models.qwen2_5.lmdeploy_qwen2_5_0_5b_instruct import \
+        models as lmdeploy_qwen2_5_0_5b_instruct_model  # noqa: F401, E501
+    from opencompass.configs.models.qwen2_5.lmdeploy_qwen2_5_3b_instruct import \
+        models as lmdeploy_qwen2_5_3b_instruct_model  # noqa: F401, E501
+    from opencompass.configs.models.qwen2_5.lmdeploy_qwen2_5_14b_instruct import \
+        models as lmdeploy_qwen2_5_14b_instruct_model  # noqa: F401, E501
+    from opencompass.configs.models.qwen2_5.lmdeploy_qwen2_5_72b_instruct import \
+        models as lmdeploy_qwen2_5_72b_instruct_model  # noqa: F401, E501
+    from opencompass.configs.models.qwen.hf_qwen1_5_0_5b_chat import \
+        models as hf_qwen1_5_0_5b_chat_model  # noqa: F401, E501
+    from opencompass.configs.models.qwen.hf_qwen2_1_5b_instruct import \
+        models as hf_qwen2_1_5b_instruct_model  # noqa: F401, E501
+    from opencompass.configs.models.qwen.hf_qwen2_7b_instruct import \
+        models as hf_qwen2_7b_instruct_model  # noqa: F401, E501
+    from opencompass.configs.models.qwen.lmdeploy_qwen2_1_5b_instruct import \
+        models as lmdeploy_qwen2_1_5b_instruct_model  # noqa: F401, E501
+    from opencompass.configs.models.qwen.lmdeploy_qwen2_7b_instruct import \
+        models as lmdeploy_qwen2_7b_instruct_model  # noqa: F401, E501
+    from opencompass.configs.models.qwen.vllm_qwen1_5_0_5b_chat import \
+        models as vllm_qwen1_5_0_5b_chat_model  # noqa: F401, E501
+    from opencompass.configs.models.yi.hf_yi_1_5_6b_chat import \
+        models as hf_yi_1_5_6b_chat_model  # noqa: F401, E501
+    from opencompass.configs.models.yi.hf_yi_1_5_9b_chat import \
+        models as hf_yi_1_5_9b_chat_model  # noqa: F401, E501
+    from opencompass.configs.models.yi.lmdeploy_yi_1_5_6b_chat import \
+        models as lmdeploy_yi_1_5_6b_chat_model  # noqa: F401, E501
+    from opencompass.configs.models.yi.lmdeploy_yi_1_5_9b_chat import \
+        models as lmdeploy_yi_1_5_9b_chat_model  # noqa: F401, E501
+    from opencompass.configs.models.yi.lmdeploy_yi_1_5_34b_chat import \
+        models as lmdeploy_yi_1_5_34b_chat_model  # noqa: F401, E501
+
+    from ...volc import infer as volc_infer  # noqa: F401, E501
+
+hf_glm4_9b_chat_model[0]['path'] = 'THUDM/glm-4-9b-chat-hf'
+
+race_datasets = [race_datasets[1]]
+datasets = sum([v for k, v in locals().items() if k.endswith('_datasets')], [])
+
+api_meta_template = dict(
+    round=[
+        dict(role='HUMAN', api_role='HUMAN'),
+        dict(role='BOT', api_role='BOT', generate=True),
+    ],
+    reserved_roles=[dict(role='SYSTEM', api_role='SYSTEM')],
+)
+
+for d in datasets:
+    d['reader_cfg']['test_range'] = '[0:32]'
+
+models = sum([v for k, v in locals().items() if k.endswith('_model')], [])
+
+for m in models:
+    if 'turbomind' in m['abbr'] or 'lmdeploy' in m['abbr']:
+        m['engine_config']['max_batch_size'] = 1
+        m['batch_size'] = 1
+
+models = sorted(models, key=lambda x: x['run_cfg']['num_gpus'])
+
+summarizer = dict(
+    dataset_abbrs=[
+        'gsm8k',
+        'race-middle',
+        'race-high',
+    ],
+    summary_groups=sum(
+        [v for k, v in locals().items() if k.endswith('_summary_groups')], []),
+)

.github/scripts/eval_regression_chat_obj_fullbench.py

@@ -0,0 +1,317 @@
+from mmengine.config import read_base
+
+with read_base():
+    # read hf models - chat models
+    # Dataset
+    from opencompass.configs.datasets.aime2024.aime2024_gen_6e39a4 import \
+        aime2024_datasets  # noqa: F401, E501
+    from opencompass.configs.datasets.ARC_c.ARC_c_cot_gen_926652 import \
+        ARC_c_datasets  # noqa: F401, E501
+    # remove because of oom
+    # from opencompass.configs.datasets.ARC_Prize_Public_Evaluation.arc_prize_public_evaluation_gen_872059 import arc_prize_public_evaluation_datasets  # noqa: F401, E501
+    from opencompass.configs.datasets.bbh.bbh_gen_5b92b0 import \
+        bbh_datasets  # noqa: F401, E501
+    from opencompass.configs.datasets.bigcodebench.bigcodebench_hard_complete_gen_faf748 import \
+        bigcodebench_hard_complete_datasets  # noqa: F401, E501
+    from opencompass.configs.datasets.bigcodebench.bigcodebench_hard_instruct_gen_8815eb import \
+        bigcodebench_hard_instruct_datasets  # noqa: F401, E501
+    from opencompass.configs.datasets.cmmlu.cmmlu_0shot_cot_gen_305931 import \
+        cmmlu_datasets  # noqa: F401, E501
+    from opencompass.configs.datasets.cmo_fib.cmo_fib_gen_ace24b import \
+        cmo_fib_datasets  # noqa: F401, E501
+    from opencompass.configs.datasets.drop.drop_openai_simple_evals_gen_3857b0 import \
+        drop_datasets  # noqa: F401, E501
+    from opencompass.configs.datasets.ds1000.ds1000_service_eval_gen_cbc84f import \
+        ds1000_datasets  # noqa: F401, E501
+    from opencompass.configs.datasets.GaokaoBench.GaokaoBench_no_subjective_gen_4c31db import \
+        GaokaoBench_datasets  # noqa: F401, E501
+    from opencompass.configs.datasets.gpqa.gpqa_openai_simple_evals_gen_5aeece import \
+        gpqa_datasets  # noqa: F401, E501
+    # new datasets in Fullbench v1.1
+    from opencompass.configs.datasets.gsm8k.gsm8k_0shot_v2_gen_6e39a4 import \
+        gsm8k_datasets  # noqa: F401, E501
+    from opencompass.configs.datasets.hellaswag.hellaswag_10shot_gen_e42710 import \
+        hellaswag_datasets  # noqa: F401, E501
+    from opencompass.configs.datasets.humaneval.humaneval_openai_sample_evals_gen_dcae0e import \
+        humaneval_datasets  # noqa: F401, E501
+    from opencompass.configs.datasets.humanevalx.humanevalx_gen_3d84a3 import \
+        humanevalx_datasets  # noqa: F401, E501
+    from opencompass.configs.datasets.IFEval.IFEval_gen_353ae7 import \
+        ifeval_datasets  # noqa: F401, E501
+    from opencompass.configs.datasets.korbench.korbench_single_0_shot_gen import \
+        korbench_0shot_single_datasets  # noqa: F401, E501
+    from opencompass.configs.datasets.livecodebench.livecodebench_gen_b2b0fd import \
+        LCB_datasets  # noqa: F401, E501
+    from opencompass.configs.datasets.math.math_0shot_gen_11c4b5 import \
+        math_datasets  # noqa: F401, E501
+    from opencompass.configs.datasets.MathBench.mathbench_2024_gen_50a320 import \
+        mathbench_datasets  # noqa: F401, E501
+    from opencompass.configs.datasets.mbpp.sanitized_mbpp_mdblock_gen_a447ff import \
+        sanitized_mbpp_datasets  # noqa: F401, E501
+    from opencompass.configs.datasets.mmlu.mmlu_openai_simple_evals_gen_b618ea import \
+        mmlu_datasets  # noqa: F401, E501
+    from opencompass.configs.datasets.mmlu_pro.mmlu_pro_0shot_cot_gen_08c1de import \
+        mmlu_pro_datasets  # noqa: F401, E501
+    from opencompass.configs.datasets.mmmlu_lite.mmmlu_lite_gen_c51a84 import \
+        mmmlu_lite_datasets  # noqa: F401, E501
+    from opencompass.configs.datasets.musr.musr_gen_3622bb import \
+        musr_datasets  # noqa: F401, E501
+    from opencompass.configs.datasets.nq.nq_open_1shot_gen_2e45e5 import \
+        nq_datasets  # noqa: F401, E501
+    from opencompass.configs.datasets.race.race_cot_gen_d95929 import \
+        race_datasets  # noqa: F401, E501
+    from opencompass.configs.datasets.scicode.scicode_gen_085b98 import \
+        SciCode_datasets  # noqa: F401, E501
+    from opencompass.configs.datasets.SuperGLUE_BoolQ.SuperGLUE_BoolQ_cot_gen_1d56df import \
+        BoolQ_datasets  # noqa: F401, E501
+    from opencompass.configs.datasets.teval.teval_en_gen_1ac254 import \
+        teval_datasets as teval_en_datasets  # noqa: F401, E501
+    from opencompass.configs.datasets.teval.teval_zh_gen_1ac254 import \
+        teval_datasets as teval_zh_datasets  # noqa: F401, E501
+    from opencompass.configs.datasets.TheoremQA.TheoremQA_5shot_gen_6f0af8 import \
+        TheoremQA_datasets  # noqa: F401, E501
+    from opencompass.configs.datasets.triviaqa.triviaqa_wiki_1shot_gen_bc5f21 import \
+        triviaqa_datasets  # noqa: F401, E501
+    from opencompass.configs.datasets.wikibench.wikibench_gen_0978ad import \
+        wikibench_datasets  # noqa: F401, E501
+    from opencompass.configs.models.hf_internlm.hf_internlm2_5_7b_chat import \
+        models as hf_internlm2_5_7b_chat_model  # noqa: F401, E501
+    from opencompass.configs.models.hf_internlm.lmdeploy_internlm2_5_7b_chat import \
+        models as lmdeploy_internlm2_5_7b_chat_model  # noqa: F401, E501
+    # Summary Groups
+    # Summary Groups
+    from opencompass.configs.summarizers.groups.bbh import \
+        bbh_summary_groups  # noqa: F401, E501
+    from opencompass.configs.summarizers.groups.cmmlu import \
+        cmmlu_summary_groups  # noqa: F401, E501
+    from opencompass.configs.summarizers.groups.ds1000 import \
+        ds1000_summary_groups  # noqa: F401, E501
+    from opencompass.configs.summarizers.groups.GaokaoBench import \
+        GaokaoBench_summary_groups  # noqa: F401, E501
+    from opencompass.configs.summarizers.groups.humanevalx import \
+        humanevalx_summary_groups  # noqa: F401, E501
+    from opencompass.configs.summarizers.groups.korbench import \
+        korbench_summary_groups  # noqa: F401, E501
+    from opencompass.configs.summarizers.groups.mathbench_v1_2024 import \
+        mathbench_2024_summary_groups  # noqa: F401, E501
+    from opencompass.configs.summarizers.groups.mmlu import \
+        mmlu_summary_groups  # noqa: F401, E501
+    from opencompass.configs.summarizers.groups.mmlu_pro import \
+        mmlu_pro_summary_groups  # noqa: F401, E501
+    from opencompass.configs.summarizers.groups.musr_average import \
+        summarizer as musr_summarizer  # noqa: F401, E501
+    from opencompass.configs.summarizers.groups.scicode import \
+        scicode_summary_groups  # noqa: F401, E501
+    from opencompass.configs.summarizers.groups.teval import \
+        teval_summary_groups  # noqa: F401, E501
+    from opencompass.configs.summarizers.mmmlu_lite import \
+        mmmlu_summary_groups  # noqa: F401, E501
+
+    from ...volc import infer as volc_infer  # noqa: F401, E501
+
+# For HumanEval-X Evaluation
+# Apply the evaluator ip_address and port
+race_datasets = [race_datasets[1]]
+for item in humanevalx_datasets:
+    item['eval_cfg']['evaluator'][
+        'ip_address'] = 'codeeval.opencompass.org.cn/humanevalx'
+    item['eval_cfg']['evaluator']['port'] = ''
+
+# For DS-1000 Evaluation
+# Apply the evaluator ip_address and port
+for item in ds1000_datasets:
+    item['eval_cfg']['evaluator'][
+        'ip_address'] = 'codeeval.opencompass.org.cn/ds1000'
+    item['eval_cfg']['evaluator']['port'] = ''
+
+bbh_datasets = [
+    x for x in bbh_datasets if 'logical_deduction_seven_objects' in x['abbr']
+    or 'multistep_arithmetic_two' in x['abbr']
+]
+cmmlu_datasets = [
+    x for x in cmmlu_datasets if x['abbr'].replace('cmmlu-', '') in [
+        'ancient_chinese', 'chinese_civil_service_exam',
+        'chinese_driving_rule', 'chinese_food_culture',
+        'chinese_foreign_policy', 'chinese_history', 'chinese_literature',
+        'chinese_teacher_qualification', 'construction_project_management',
+        'elementary_chinese', 'elementary_commonsense', 'ethnology',
+        'high_school_politics', 'modern_chinese',
+        'traditional_chinese_medicine'
+    ]
+]
+mmlu_datasets = [
+    x for x in mmlu_datasets if x['abbr'].replace('lukaemon_mmlu_', '') in [
+        'business_ethics', 'clinical_knowledge', 'college_medicine',
+        'global_facts', 'human_aging', 'management', 'marketing',
+        'medical_genetics', 'miscellaneous', 'nutrition',
+        'professional_accounting', 'professional_medicine', 'virology'
+    ]
+]
+
+mmlu_pro_datasets = [mmlu_pro_datasets[0]]
+
+mmmlu_lite_datasets = [
+    x for x in mmmlu_lite_datasets if 'mmlu_lite_AR-XY' in x['abbr']
+]
+mathbench_datasets = [x for x in mathbench_datasets if 'college' in x['abbr']]
+GaokaoBench_datasets = [
+    x for x in GaokaoBench_datasets if '2010-2022_Math_II_MCQs' in x['abbr']
+    or '2010-2022_Math_II_Fill-in-the-Blank' in x['abbr']
+]
+
+datasets = sum(
+    (v for k, v in locals().items() if k.endswith('_datasets')
+     and 'scicode' not in k.lower() and 'teval' not in k),
+    [],
+)
+datasets += teval_en_datasets
+datasets += teval_zh_datasets
+# datasets += SciCode_datasets
+
+musr_summary_groups = musr_summarizer['summary_groups']
+summary_groups = sum(
+    [v for k, v in locals().items() if k.endswith('_summary_groups')], [])
+summary_groups.append(
+    {
+        'name': 'Mathbench',
+        'subsets': ['mathbench-a (average)', 'mathbench-t (average)'],
+    }, )
+
+# Summarizer
+summarizer = dict(
+    dataset_abbrs=[
+        'Language',
+        ['race-high', 'accuracy'],
+        ['ARC-c', 'accuracy'],
+        ['BoolQ', 'accuracy'],
+        ['triviaqa_wiki_1shot', 'score'],
+        ['nq_open_1shot', 'score'],
+        ['mmmlu_lite', 'naive_average'],
+        '',
+        'Instruction Following',
+        ['IFEval', 'Prompt-level-strict-accuracy'],
+        '',
+        'General Reasoning',
+        ['drop', 'accuracy'],
+        ['bbh', 'naive_average'],
+        ['GPQA_diamond', 'accuracy'],
+        ['hellaswag', 'accuracy'],
+        ['TheoremQA', 'score'],
+        ['musr_average', 'naive_average'],
+        ['korbench_single', 'naive_average'],
+        ['ARC_Prize_Public_Evaluation', 'accuracy'],
+        '',
+        'Math Calculation',
+        ['gsm8k', 'accuracy'],
+        ['GaokaoBench', 'weighted_average'],
+        ['math', 'accuracy'],
+        ['cmo_fib', 'accuracy'],
+        ['aime2024', 'accuracy'],
+        ['Mathbench', 'naive_average'],
+        '',
+        'Knowledge',
+        ['wikibench-wiki-single_choice_cncircular', 'perf_4'],
+        ['cmmlu', 'naive_average'],
+        ['mmlu', 'naive_average'],
+        ['mmlu_pro', 'naive_average'],
+        '',
+        'Code',
+        ['openai_humaneval', 'humaneval_pass@1'],
+        ['sanitized_mbpp', 'score'],
+        ['humanevalx', 'naive_average'],
+        ['ds1000', 'naive_average'],
+        ['lcb_code_generation', 'pass@1'],
+        ['lcb_code_execution', 'pass@1'],
+        ['lcb_test_output', 'pass@1'],
+        ['bigcodebench_hard_instruct', 'pass@1'],
+        ['bigcodebench_hard_complete', 'pass@1'],
+        '',
+        'Agent',
+        ['teval', 'naive_average'],
+        ['SciCode', 'accuracy'],
+        ['SciCode', 'sub_accuracy'],
+        '',
+        'bbh-logical_deduction_seven_objects',
+        'bbh-multistep_arithmetic_two',
+        '',
+        'mmlu',
+        'mmlu-stem',
+        'mmlu-social-science',
+        'mmlu-humanities',
+        'mmlu-other',
+        '',
+        'cmmlu',
+        'cmmlu-stem',
+        'cmmlu-social-science',
+        'cmmlu-humanities',
+        'cmmlu-other',
+        'cmmlu-china-specific',
+        '',
+        'mmlu_pro',
+        'mmlu_pro_biology',
+        'mmlu_pro_business',
+        'mmlu_pro_chemistry',
+        'mmlu_pro_computer_science',
+        'mmlu_pro_economics',
+        'mmlu_pro_engineering',
+        'mmlu_pro_health',
+        'mmlu_pro_history',
+        'mmlu_pro_law',
+        'mmlu_pro_math',
+        'mmlu_pro_philosophy',
+        'mmlu_pro_physics',
+        'mmlu_pro_psychology',
+        'mmlu_pro_other',
+        '',
+        'ds1000_Pandas',
+        'ds1000_Numpy',
+        'ds1000_Tensorflow',
+        'ds1000_Scipy',
+        'ds1000_Sklearn',
+        'ds1000_Pytorch',
+        'ds1000_Matplotlib',
+        '',
+        'mmmlu_lite',
+        'openai_mmmlu_lite_AR-XY',
+        'openai_mmmlu_lite_BN-BD',
+        'openai_mmmlu_lite_DE-DE',
+        'openai_mmmlu_lite_ES-LA',
+        'openai_mmmlu_lite_FR-FR',
+        'openai_mmmlu_lite_HI-IN',
+        'openai_mmmlu_lite_ID-ID',
+        'openai_mmmlu_lite_IT-IT',
+        'openai_mmmlu_lite_JA-JP',
+        'openai_mmmlu_lite_KO-KR',
+        'openai_mmmlu_lite_PT-BR',
+        'openai_mmmlu_lite_SW-KE',
+        'openai_mmmlu_lite_YO-NG',
+        'openai_mmmlu_lite_ZH-CN',
+        '',
+        '###### MathBench-A: Application Part ######',
+        'college',
+        'high',
+        'middle',
+        'primary',
+        'arithmetic',
+        'mathbench-a (average)',
+        '###### MathBench-T: Theory Part ######',
+        'college_knowledge',
+        'high_knowledge',
+        'middle_knowledge',
+        'primary_knowledge',
+        'mathbench-t (average)',
+    ],
+    summary_groups=summary_groups,
+)
+
+for d in datasets:
+    d['reader_cfg']['test_range'] = '[0:16]'
+
+models = sum([v for k, v in locals().items() if k.endswith('_model')], [])
+for m in models:
+    m['abbr'] = m['abbr'] + '_fullbench'
+    if 'turbomind' in m['abbr'] or 'lmdeploy' in m['abbr']:
+        m['engine_config']['max_batch_size'] = 1
+        m['batch_size'] = 1
+
+models = sorted(models, key=lambda x: x['run_cfg']['num_gpus'])

.github/scripts/eval_regression_chat_sub_fullbench.py

@@ -0,0 +1,182 @@
+from copy import deepcopy
+
+from mmengine.config import read_base
+
+from opencompass.partitioners.sub_naive import SubjectiveNaivePartitioner
+from opencompass.runners import LocalRunner
+from opencompass.summarizers import DefaultSubjectiveSummarizer
+from opencompass.tasks.subjective_eval import SubjectiveEvalTask
+
+with read_base():
+    # read hf models - chat models
+    # Dataset
+    from opencompass.configs.datasets.chinese_simpleqa.chinese_simpleqa_gen import \
+        csimpleqa_datasets  # noqa: F401, E501
+    from opencompass.configs.datasets.SimpleQA.simpleqa_gen_0283c3 import \
+        simpleqa_datasets  # noqa: F401, E501; noqa: F401, E501
+    from opencompass.configs.datasets.subjective.alignbench.alignbench_v1_1_judgeby_critiquellm_new import \
+        alignbench_datasets  # noqa: F401, E501
+    from opencompass.configs.datasets.subjective.alpaca_eval.alpacav2_judgeby_gpt4_new import \
+        alpacav2_datasets  # noqa: F401, E501
+    from opencompass.configs.datasets.subjective.arena_hard.arena_hard_compare_new import \
+        arenahard_datasets  # noqa: F401, E501
+    from opencompass.configs.datasets.subjective.compassarena.compassarena_compare_new import \
+        compassarena_datasets  # noqa: F401, E501
+    # from opencompass.configs.datasets.subjective.fofo.fofo_bilingual_judge_new import fofo_datasets  # noqa: F401, E501
+    from opencompass.configs.datasets.subjective.followbench.followbench_llmeval_new import \
+        followbench_llmeval_datasets  # noqa: F401, E501
+    from opencompass.configs.datasets.subjective.multiround.mtbench101_judge_new import \
+        mtbench101_datasets  # noqa: F401, E501
+    from opencompass.configs.datasets.subjective.wildbench.wildbench_pair_judge_new import \
+        wildbench_datasets  # noqa: F401, E501
+    from opencompass.configs.models.hf_internlm.hf_internlm2_5_7b_chat import \
+        models as hf_internlm2_5_7b_chat_model  # noqa: F401, E501
+    from opencompass.configs.models.hf_internlm.lmdeploy_internlm2_5_7b_chat import \
+        models as lmdeploy_internlm2_5_7b_chat_model  # noqa: F401, E501
+
+    from ...volc import infer as volc_infer  # noqa: F401, E501
+
+datasets = sum((v for k, v in locals().items() if k.endswith('_datasets')
+                and 'mtbench101' not in k and 'wildbench' not in k), [])
+datasets += mtbench101_datasets  # noqa: F401, E501
+datasets += wildbench_datasets  # noqa: F401, E501
+
+api_meta_template = dict(
+    round=[
+        dict(role='HUMAN', api_role='HUMAN'),
+        dict(role='BOT', api_role='BOT', generate=True),
+    ],
+    reserved_roles=[dict(role='SYSTEM', api_role='SYSTEM')],
+)
+
+models = sum([v for k, v in locals().items() if k.endswith('_model')], [])
+for m in models:
+    m['abbr'] = m['abbr'] + '_fullbench'
+    if 'turbomind' in m['abbr'] or 'lmdeploy' in m['abbr']:
+        m['engine_config']['max_batch_size'] = 1
+        m['batch_size'] = 1
+
+models = sorted(models, key=lambda x: x['run_cfg']['num_gpus'])
+
+judge_models = deepcopy([models[1]])
+judge_models[0]['abbr'] = judge_models[0]['abbr'] + '-judge'
+
+eval = dict(
+    partitioner=dict(
+        type=SubjectiveNaivePartitioner,
+        models=models,
+        judge_models=judge_models,
+    ),
+    runner=dict(type=LocalRunner,
+                max_num_workers=16,
+                task=dict(type=SubjectiveEvalTask)),
+)
+
+summary_groups = []
+summary_groups.append({
+    'name': 'compassarena_language',
+    'subsets': [
+        ['compassarena_language', '内容总结'],
+    ],
+})
+summary_groups.append({
+    'name': 'compassarena_knowledge',
+    'subsets': [
+        ['compassarena_knowledge', '生活常识_ZH'],
+    ],
+})
+summary_groups.append({
+    'name': 'compassarena_reason_v2',
+    'subsets': [
+        ['compassarena_reason_v2', 'reasoning'],
+    ],
+})
+summary_groups.append({
+    'name': 'compassarena_math_v2',
+    'subsets': [
+        ['compassarena_math_v2', '高等数学_ZH'],
+    ],
+})
+summary_groups.append({
+    'name': 'compassarena_creationv2_zh',
+    'subsets': [
+        ['compassarena_creationv2_zh', '内容扩写_ZH'],
+    ],
+})
+summary_groups.append({
+    'name':
+    'CompassArena',
+    'subsets': [
+        'compassarena_language',
+        'compassarena_knowledge',
+        'compassarena_reason_v2',
+        'compassarena_math_v2',
+        'compassarena_creationv2_zh',
+    ],
+})
+summary_groups.append({
+    'name':
+    'FoFo',
+    'subsets': [['fofo_test_prompts', 'overall'],
+                ['fofo_test_prompts_cn', 'overall']],
+})
+summary_groups.append({
+    'name':
+    'Followbench',
+    'subsets': [
+        ['followbench_llmeval_en', 'HSR_AVG'],
+        ['followbench_llmeval_en', 'SSR_AVG'],
+    ],
+})
+
+# Summarizer
+summarizer = dict(
+    dataset_abbrs=[
+        ['alignment_bench_v1_1', '总分'],
+        ['alpaca_eval', 'total'],
+        ['arenahard', 'score'],
+        ['Followbench', 'naive_average'],
+        ['CompassArena', 'naive_average'],
+        ['FoFo', 'naive_average'],
+        ['mtbench101', 'avg'],
+        ['wildbench', 'average'],
+        ['simpleqa', 'accuracy_given_attempted'],
+        ['chinese_simpleqa', 'given_attempted_accuracy'],
+        '',
+        ['alignment_bench_v1_1', '专业能力'],
+        ['alignment_bench_v1_1', '数学计算'],
+        ['alignment_bench_v1_1', '基本任务'],
+        ['alignment_bench_v1_1', '逻辑推理'],
+        ['alignment_bench_v1_1', '中文理解'],
+        ['alignment_bench_v1_1', '文本写作'],
+        ['alignment_bench_v1_1', '角色扮演'],
+        ['alignment_bench_v1_1', '综合问答'],
+        ['alpaca_eval', 'helpful_base'],
+        ['alpaca_eval', 'koala'],
+        ['alpaca_eval', 'oasst'],
+        ['alpaca_eval', 'selfinstruct'],
+        ['alpaca_eval', 'vicuna'],
+        ['compassarena_language', 'naive_average'],
+        ['compassarena_knowledge', 'naive_average'],
+        ['compassarena_reason_v2', 'naive_average'],
+        ['compassarena_math_v2', 'naive_average'],
+        ['compassarena_creationv2_zh', 'naive_average'],
+        ['fofo_test_prompts', 'overall'],
+        ['fofo_test_prompts_cn', 'overall'],
+        ['followbench_llmeval_en', 'HSR_AVG'],
+        ['followbench_llmeval_en', 'SSR_AVG'],
+        ['followbench_llmeval_en', 'HSR_L1'],
+        ['followbench_llmeval_en', 'HSR_L2'],
+        ['followbench_llmeval_en', 'HSR_L3'],
+        ['followbench_llmeval_en', 'HSR_L4'],
+        ['followbench_llmeval_en', 'HSR_L5'],
+        ['followbench_llmeval_en', 'SSR_L1'],
+        ['followbench_llmeval_en', 'SSR_L2'],
+        ['followbench_llmeval_en', 'SSR_L3'],
+        ['followbench_llmeval_en', 'SSR_L4'],
+        ['followbench_llmeval_en', 'SSR_L5'],
+        ['simpleqa', 'f1'],
+    ],
+    type=DefaultSubjectiveSummarizer,
+    summary_groups=summary_groups,
+)

.github/scripts/oc_score_assert.py

@@ -6,11 +6,29 @@ import yaml
 
 output_path = 'regression_result_daily'
 
-model_list = ['internlm2-7b-hf', 'internlm-chat-7b-hf', 'chatglm3-6b-base-hf']
-dataset_list = [
-    'ARC-c', 'chid-dev', 'chid-test', 'openai_humaneval', 'openbookqa',
-    'openbookqa_fact'
-]
+
+def model_list(type):
+    config_path = '.github/scripts/oc_score_baseline_testrange.yaml'
+    with open(config_path) as f:
+        config = yaml.load(f.read(), Loader=yaml.SafeLoader)
+    return config.get(type).keys()
+
+
+def dataset_list(model, type):
+    config_path = '.github/scripts/oc_score_baseline_fullbench.yaml'
+    with open(config_path) as f:
+        config = yaml.load(f.read(), Loader=yaml.SafeLoader)
+    return config.get(model).get(type).keys()
+
+
+@pytest.fixture()
+def baseline_scores_testrange(request):
+    config_path = os.path.join(
+        request.config.rootdir,
+        '.github/scripts/oc_score_baseline_testrange.yaml')
+    with open(config_path) as f:
+        config = yaml.load(f.read(), Loader=yaml.SafeLoader)
+    return config
 
 
 @pytest.fixture()
@@ -22,6 +40,16 @@ def baseline_scores(request):
     return config
 
 
+@pytest.fixture()
+def baseline_scores_fullbench(request):
+    config_path = os.path.join(
+        request.config.rootdir,
+        '.github/scripts/oc_score_baseline_fullbench.yaml')
+    with open(config_path) as f:
+        config = yaml.load(f.read(), Loader=yaml.SafeLoader)
+    return config
+
+
 @pytest.fixture()
 def result_scores():
     file = find_csv_files(output_path)
@@ -31,55 +59,314 @@ def result_scores():
 
 
 @pytest.mark.usefixtures('result_scores')
-@pytest.mark.usefixtures('baseline_scores')
+@pytest.mark.usefixtures('baseline_scores_testrange')
+@pytest.mark.chat_models
 class TestChat:
     """Test cases for chat model."""
 
-    @pytest.mark.parametrize('model, dataset', [(p1, p2) for p1 in model_list
-                                                for p2 in dataset_list])
-    def test_model_dataset_score(self, baseline_scores, result_scores, model,
-                                 dataset):
+    @pytest.mark.parametrize(
+        'model, dataset', [(p1, p2) for p1 in model_list('chat')
+                           for p2 in ['gsm8k_accuracy', 'race-high_accuracy']])
+    def test_model_dataset_score(self, baseline_scores_testrange,
+                                 result_scores, model, dataset):
+        base_score = baseline_scores_testrange.get('chat').get(model).get(
+            dataset)
+        result_score = result_scores.get(model).get(dataset)
+        assert_score(model, result_score, base_score, dataset)
+
+
+@pytest.mark.usefixtures('result_scores')
+@pytest.mark.usefixtures('baseline_scores_testrange')
+@pytest.mark.base_models
+class TestBase:
+    """Test cases for base model."""
+
+    @pytest.mark.parametrize('model, dataset',
+                             [(p1, p2) for p1 in model_list('base') for p2 in [
+                                 'gsm8k_accuracy', 'GPQA_diamond_accuracy',
+                                 'race-high_accuracy', 'winogrande_accuracy'
+                             ]])
+    def test_model_dataset_score(self, baseline_scores_testrange,
+                                 result_scores, model, dataset):
+        if model in ['gemma-2b-vllm', 'gemma-7b-vllm'
+                     ] and dataset != 'gsm8k_accuracy':
+            return
+        base_score = baseline_scores_testrange.get('base').get(model).get(
+            dataset)
+        result_score = result_scores.get(model).get(dataset)
+        assert_score(model, result_score, base_score, dataset)
+
+
+@pytest.mark.usefixtures('result_scores')
+@pytest.mark.usefixtures('baseline_scores_fullbench')
+@pytest.mark.chat_obj_fullbench
+class TestChatObjFullbench:
+    """Test cases for chat model."""
+
+    @pytest.mark.parametrize('model, dataset', [(p1, p2) for p1 in [
+        'internlm2_5-7b-chat-hf_fullbench',
+        'internlm2_5-7b-chat-turbomind_fullbench'
+    ] for p2 in dataset_list('internlm2_5-7b-chat-hf_fullbench', 'objective')])
+    def test_model_dataset_score(self, baseline_scores_fullbench,
+                                 result_scores, model, dataset):
+        base_score = baseline_scores_fullbench.get(model).get('objective').get(
+            dataset)
+        result_score = result_scores.get(model).get(dataset)
+        assert_score(model, result_score, base_score, dataset)
+
+
+@pytest.mark.usefixtures('result_scores')
+@pytest.mark.usefixtures('baseline_scores_fullbench')
+@pytest.mark.chat_sub_fullbench
+class TestChatSubFullbench:
+    """Test cases for chat model."""
+
+    @pytest.mark.parametrize('model, dataset', [(p1, p2) for p1 in [
+        'internlm2_5-7b-chat-hf_fullbench',
+        'internlm2_5-7b-chat-turbomind_fullbench'
+    ] for p2 in dataset_list('internlm2_5-7b-chat-hf_fullbench', 'subjective')]
+                             )
+    def test_model_dataset_score(self, baseline_scores_fullbench,
+                                 result_scores, model, dataset):
+        base_score = baseline_scores_fullbench.get(model).get(
+            'subjective').get(dataset)
+        result_score = result_scores.get(model).get(dataset)
+        assert_score(model, result_score, base_score, dataset)
+
+
+@pytest.mark.usefixtures('result_scores')
+@pytest.mark.usefixtures('baseline_scores_fullbench')
+@pytest.mark.base_fullbench
+class TestBaseFullbench:
+    """Test cases for chat model."""
+
+    @pytest.mark.parametrize(
+        'model, dataset',
+        [(p1, p2) for p1 in
+         ['internlm2_5-7b-hf_fullbench', 'internlm2_5-7b-turbomind_fullbench']
+         for p2 in dataset_list('internlm2_5-7b-hf_fullbench', 'objective')])
+    def test_model_dataset_score(self, baseline_scores_fullbench,
+                                 result_scores, model, dataset):
+        base_score = baseline_scores_fullbench.get(model).get('objective').get(
+            dataset)
+        result_score = result_scores.get(model).get(dataset)
+        assert_score(model, result_score, base_score, dataset)
+
+
+@pytest.mark.usefixtures('result_scores')
+@pytest.mark.usefixtures('baseline_scores')
+@pytest.mark.api
+class TestApibench:
+    """Test cases for chat model."""
+
+    @pytest.mark.parametrize('model, dataset',
+                             [('lmdeploy-api-test', 'race-middle_accuracy'),
+                              ('lmdeploy-api-test', 'race-high_accuracy'),
+                              ('lmdeploy-api-test', 'gsm8k_accuracy')])
+    def test_api(self, baseline_scores, result_scores, model, dataset):
         base_score = baseline_scores.get(model).get(dataset)
         result_score = result_scores.get(model).get(dataset)
-        assert_score(result_score, base_score)
+        assert_score(model + '_batch', result_score, base_score, dataset)
 
 
-def assert_score(score, baseline):
+@pytest.mark.usefixtures('result_scores')
+@pytest.mark.usefixtures('baseline_scores_fullbench')
+@pytest.mark.volc_fullbench
+class TestVolcFullbench:
+    """Test cases for chat model."""
+
+    @pytest.mark.parametrize('model, dataset', [(p1, p2) for p1 in [
+        'internlm2_5-7b-chat-turbomind', 'qwen2.5-7b-instruct-turbomind',
+        'internlm2_5-7b-chat-pytorch', 'qwen2.5-7b-instruct-pytorch',
+        'internlm3-8b-instruct-turbomind', 'internlm3-8b-instruct-pytorch'
+    ] for p2 in dataset_list(p1, 'objective')])
+    @pytest.mark.chat_objective
+    def test_chat_objective(self, baseline_scores_fullbench, result_scores,
+                            model, dataset):
+        base_score = baseline_scores_fullbench.get(model).get('objective').get(
+            dataset)
+        result_score = result_scores.get(model).get(dataset)
+        assert_score(model + '_batch', result_score, base_score, dataset)
+
+    @pytest.mark.parametrize('model, dataset', [
+        (p1, p2) for p1 in ['internlm2_5-7b-chat-turbomind']
+        for p2 in dataset_list('internlm2_5-7b-chat-turbomind', 'subjective')
+    ])
+    @pytest.mark.chat_subjective
+    def test_chat_subjective(self, baseline_scores_fullbench, result_scores,
+                             model, dataset):
+        base_score = baseline_scores_fullbench.get(model).get(
+            'subjective').get(dataset)
+        result_score = result_scores.get(model).get(dataset)
+        assert_score(model + '_batch', result_score, base_score, dataset)
+
+    @pytest.mark.parametrize(
+        'model, dataset',
+        [(p1, p2) for p1 in ['internlm2_5-7b-turbomind']
+         for p2 in dataset_list('internlm2_5-7b-turbomind', 'objective')])
+    @pytest.mark.base_objective
+    def test_base_objective(self, baseline_scores_fullbench, result_scores,
+                            model, dataset):
+        base_score = baseline_scores_fullbench.get(model).get('objective').get(
+            dataset)
+        result_score = result_scores.get(model).get(dataset)
+        assert_score(model + '_batch', result_score, base_score, dataset)
+
+    @pytest.mark.parametrize(
+        'model, dataset',
+        [(p1, p2) for p1 in ['internlm2_5-7b-turbomind']
+         for p2 in dataset_list('internlm2_5-7b-turbomind', 'long_context')])
+    @pytest.mark.base_long_context
+    def test_base_long_context(self, baseline_scores_fullbench, result_scores,
+                               model, dataset):
+        base_score = baseline_scores_fullbench.get(model).get(
+            'long_context').get(dataset)
+        result_score = result_scores.get(model).get(dataset)
+        assert_score(model + '_batch', result_score, base_score, dataset)
+
+    @pytest.mark.parametrize(
+        'model, dataset',
+        [(p1, p2)
+         for p1 in ['internlm2_5-7b-chat-1m-turbomind'] for p2 in dataset_list(
+             'internlm2_5-7b-chat-1m-turbomind', 'long_context')])
+    @pytest.mark.chat_long_context
+    def test_chat_long_context(self, baseline_scores_fullbench, result_scores,
+                               model, dataset):
+        base_score = baseline_scores_fullbench.get(model).get(
+            'long_context').get(dataset)
+        result_score = result_scores.get(model).get(dataset)
+        assert_score(model + '_batch', result_score, base_score, dataset)
+
+
+@pytest.mark.usefixtures('result_scores')
+@pytest.mark.usefixtures('baseline_scores')
+class TestCmdCase:
+
+    @pytest.mark.case1
+    @pytest.mark.parametrize('model, dataset',
+                             [('internlm2_5-7b-hf', 'race-middle_accuracy'),
+                              ('internlm2_5-7b-hf', 'race-high_accuracy'),
+                              ('internlm2_5-7b-hf', 'demo_gsm8k_accuracy')])
+    def test_cmd_case1(self, baseline_scores, result_scores, model, dataset):
+        base_score = baseline_scores.get(model).get(dataset)
+        result_score = result_scores.get(model).get(dataset)
+        assert_score(model, result_score, base_score, dataset)
+
+    @pytest.mark.case2
+    @pytest.mark.parametrize(
+        'model, dataset',
+        [('internlm2_5-7b-chat-lmdeploy', 'race-middle_accuracy'),
+         ('internlm2_5-7b-chat-lmdeploy', 'race-high_accuracy'),
+         ('internlm2_5-7b-chat-lmdeploy', 'demo_gsm8k_accuracy'),
+         ('internlm3-8b-instruct-lmdeploy', 'race-middle_accuracy'),
+         ('internlm3-8b-instruct-lmdeploy', 'race-high_accuracy'),
+         ('internlm3-8b-instruct-lmdeploy', 'demo_gsm8k_accuracy')])
+    def test_cmd_case2(self, baseline_scores, result_scores, model, dataset):
+        base_score = baseline_scores.get(model).get(dataset)
+        result_score = result_scores.get(model).get(dataset)
+        assert_score(model + '_batch', result_score, base_score, dataset)
+
+    @pytest.mark.case3
+    @pytest.mark.parametrize('model, dataset',
+                             [('internlm2_5-7b_hf', 'race-middle_accuracy'),
+                              ('internlm2_5-7b_hf', 'race-high_accuracy'),
+                              ('internlm2_5-7b_hf', 'demo_gsm8k_accuracy')])
+    def test_cmd_case3(self, baseline_scores, result_scores, model, dataset):
+        base_score = baseline_scores.get(model).get(dataset)
+        result_score = result_scores.get(model).get(dataset)
+        assert_score(model, result_score, base_score, dataset)
+
+    @pytest.mark.case4
+    @pytest.mark.parametrize(
+        'model, dataset',
+        [('internlm3-8b-instruct_hf-lmdeploy', 'race-middle_accuracy'),
+         ('internlm3-8b-instruct_hf-lmdeploy', 'race-high_accuracy'),
+         ('internlm3-8b-instruct_hf-lmdeploy', 'demo_gsm8k_accuracy')])
+    def test_cmd_case4(self, baseline_scores, result_scores, model, dataset):
+        base_score = baseline_scores.get(model).get(dataset)
+        result_score = result_scores.get(model).get(dataset)
+        assert_score(model + '_batch', result_score, base_score, dataset)
+
+    @pytest.mark.case5
+    @pytest.mark.parametrize(
+        'model, dataset',
+        [('internlm3-8b-instruct_hf-vllm', 'race-middle_accuracy'),
+         ('internlm3-8b-instruct_hf-vllm', 'race-high_accuracy'),
+         ('internlm3-8b-instruct_hf-vllm', 'demo_gsm8k_accuracy')])
+    def test_cmd_case5(self, baseline_scores, result_scores, model, dataset):
+        base_score = baseline_scores.get(model).get(dataset)
+        result_score = result_scores.get(model).get(dataset)
+        assert_score(model + '_batch', result_score, base_score, dataset)
+
+
+def assert_score(model_type, score, baseline, dataset: str = ''):
     if score is None or score == '-':
         assert False, 'value is none'
-    if float(score) < (baseline * 1.03) and float(score) > (baseline * 0.97):
-        print(score + ' between ' + str(baseline * 0.97) + ' and ' +
-              str(baseline * 1.03))
+
+    if 'batch' not in model_type:
+        if float(score) <= (baseline + 0.01) and float(score) >= (baseline -
+                                                                  0.01):
+            print(' '.join([score, 'is equal', str(baseline)]))
             assert True
         else:
-        assert False, score + ' not between ' + str(
-            baseline * 0.97) + ' and ' + str(baseline * 1.03)
+            print(' '.join([score, 'is not equal', str(baseline)]))
+            assert False, ' '.join([score, 'is not equal', str(baseline)])
+    else:
+        if dataset.startswith('dingo') or dataset.startswith(
+                'GPQA') or dataset.startswith('high') or dataset.startswith(
+                    'mmlu_pro_') or dataset.startswith(
+                        'alpaca_eval') or dataset.startswith('compassarena_'):
+            threshold = 5
+        elif dataset.startswith('humanevalx') or dataset == 'large_threshold':
+            threshold = 10
+        else:
+            threshold = 3
+        if float(score) <= (baseline + threshold) and float(score) >= (
+                baseline - threshold):
+            print(' '.join([
+                score, 'is between',
+                str(baseline - threshold), 'and',
+                str(baseline + threshold)
+            ]))
+            assert True
+        else:
+            print(' '.join([
+                score, 'is not between',
+                str(baseline - threshold), 'and',
+                str(baseline + threshold)
+            ]))
+            assert False, ' '.join([
+                score, 'is not between',
+                str(baseline - threshold), 'and',
+                str(baseline + threshold)
+            ])
 
 
 def find_csv_files(directory):
     csv_files = []
     for root, dirs, files in os.walk(directory):
         for file in files:
-            if file.endswith('.csv'):
+            if file.endswith('.csv') and file.startswith('summary'):
                 csv_files.append(os.path.join(root, file))
-    if len(csv_files) > 1:
-        raise 'have more than 1 result file, please check the result manually'
-    if len(csv_files) == 0:
-        return None
-    return csv_files[0]
+
+    csv_files_with_time = {f: os.path.getctime(f) for f in csv_files}
+    sorted_csv_files = sorted(csv_files_with_time.items(), key=lambda x: x[1])
+    latest_csv_file = sorted_csv_files[-1][0]
+    return latest_csv_file
 
 
 def read_csv_file(file_path):
     with open(file_path, 'r') as csvfile:
         reader = csv.DictReader(csvfile)
         filtered_data = []
-
         for row in reader:
-            filtered_row = {
-                k: v
-                for k, v in row.items()
-                if k not in ['version', 'metric', 'mode']
-            }
+            if row['metric'] is not None and 'bpb' not in row[
+                    'metric'] and '_' != row['metric']:
+                filtered_row = row
+                filtered_row['dataset'] = row['dataset'] + '_' + row['metric']
+                del filtered_row['version']
+                del filtered_row['metric']
+                del filtered_row['mode']
                 filtered_data.append(filtered_row)
 
     result = {}

.github/scripts/oc_score_baseline.yaml

@@ -1,31 +1,39 @@
-internlm-7b-hf:
-    ARC-c: 34.24
-    chid-dev: 79.70
-    chid-test: 81.12
-    openai_humaneval: 10.98
-    openbookqa: 47.20
-    openbookqa_fact: 74.00
+internlm2_5-7b-hf:
+    demo_gsm8k_accuracy: 42.19
+    race-middle_accuracy: 91.78
+    race-high_accuracy: 90.02
 
-internlm-chat-7b-hf:
-    ARC-c: 36.95
-    chid-dev: 71.78
-    chid-test: 76.87
-    openai_humaneval: 21.34
-    openbookqa: 66.6
-    openbookqa_fact: 80.4
+internlm2_5-7b_hf:
+    demo_gsm8k_accuracy: 42.19
+    race-middle_accuracy: 91.78
+    race-high_accuracy: 90.02
 
-chatglm3-6b-base-hf:
-    ARC-c: 44.41
-    chid-dev: 78.22
-    chid-test: 78.57
-    openai_humaneval: 20.73
-    openbookqa: 78.40
-    openbookqa_fact: 92.00
+internlm2_5-7b-chat-lmdeploy:
+    demo_gsm8k_accuracy: 84.38
+    race-middle_accuracy: 92.76
+    race-high_accuracy: 90.54
 
-internlm2-7b-hf:
-    ARC-c: 34.92
-    chid-dev: 55.94
-    chid-test: 53.70
-    openai_humaneval: 44.51
-    openbookqa: 83.00
-    openbookqa_fact: 83.00
+internlm3-8b-instruct-lmdeploy:
+    demo_gsm8k_accuracy: 73.44
+    race-middle_accuracy: 93.38
+    race-high_accuracy: 90.34
+
+internlm3-8b-instruct_hf-lmdeploy:
+    demo_gsm8k_accuracy: 73.44
+    race-middle_accuracy: 93.38
+    race-high_accuracy: 90.34
+
+internlm3-8b-instruct_hf-vllm:
+    demo_gsm8k_accuracy: 78.12
+    race-middle_accuracy: 92.20
+    race-high_accuracy: 89.88
+
+internlm2_5-7b-chat_hf:
+    demo_gsm8k_accuracy: 87.50
+    race-middle_accuracy: 92.76
+    race-high_accuracy: 90.48
+
+lmdeploy-api-test:
+    gsm8k_accuracy: 68.75
+    race-middle_accuracy: 93.75
+    race-high_accuracy: 93.75

.github/scripts/oc_score_baseline_fullbench.yaml

@@ -0,0 +1,983 @@
+internlm2_5-7b-chat-hf_fullbench:
+    objective:
+        race-high_accuracy: 93.75
+        ARC-c_accuracy: 93.75
+        BoolQ_accuracy: 81.25
+        triviaqa_wiki_1shot_score: 50
+        nq_open_1shot_score: 25
+        IFEval_Prompt-level-strict-accuracy: 50
+        drop_accuracy: 81.25
+        GPQA_diamond_accuracy: 25
+        hellaswag_accuracy: 87.5
+        TheoremQA_score: 12.50
+        musr_average_naive_average: 39.58
+        korbench_single_naive_average: 40
+        gsm8k_accuracy: 62.50
+        math_accuracy: 75
+        cmo_fib_accuracy: 6.25
+        aime2024_accuracy: 6.25
+        wikibench-wiki-single_choice_cncircular_perf_4: 50
+        sanitized_mbpp_score: 68.75
+        ds1000_naive_average: 16.96
+        lcb_code_generation_pass@1: 12.5
+        lcb_code_execution_pass@1: 43.75
+        lcb_test_output_pass@1: 18.75
+        bbh-logical_deduction_seven_objects_score: 50
+        bbh-multistep_arithmetic_two_score: 68.75
+        mmlu-other_accuracy: 72.6
+        cmmlu-china-specific_accuracy: 76.25
+        mmlu_pro_math_accuracy: 25
+        ds1000_Pandas_accuracy: 12.5
+        ds1000_Numpy_accuracy: 0
+        ds1000_Tensorflow_accuracy: 12.5
+        ds1000_Scipy_accuracy: 18.75
+        ds1000_Sklearn_accuracy: 18.75
+        ds1000_Pytorch_accuracy: 12.5
+        ds1000_Matplotlib_accuracy: 43.75
+        openai_mmmlu_lite_AR-XY_accuracy: 37.5
+        college_naive_average: 12.5
+        college_knowledge_naive_average: 87.5
+    subjective:
+        alignment_bench_v1_1_总分: 0.66
+        alpaca_eval_total: 20.00
+        arenahard_score: 56.82
+        Followbench_naive_average: 1
+        CompassArena_naive_average: 43
+        mtbench101_avg: 7.60
+        wildbench_average: -14.58
+        simpleqa_accuracy_given_attempted: 1.00
+        chinese_simpleqa_given_attempted_accuracy: 0.90
+        alignment_bench_v1_1_专业能力: 7.90
+        alignment_bench_v1_1_数学计算: 0
+        alignment_bench_v1_1_基本任务: 0
+        alignment_bench_v1_1_逻辑推理: 0
+        alignment_bench_v1_1_中文理解: 0
+        alignment_bench_v1_1_文本写作: 0
+        alignment_bench_v1_1_角色扮演: 0
+        alignment_bench_v1_1_综合问答: 0
+        alpaca_eval_helpful_base: 20.00
+        compassarena_language_naive_average: 35
+        compassarena_knowledge_naive_average: 60.00
+        compassarena_reason_v2_naive_average: 40
+        compassarena_math_v2_naive_average: 50.00
+        compassarena_creationv2_zh_naive_average: 30
+        followbench_llmeval_en_HSR_AVG: 1
+        followbench_llmeval_en_SSR_AVG: 1
+        followbench_llmeval_en_HSR_L1: 1
+        followbench_llmeval_en_HSR_L2: 1
+        followbench_llmeval_en_HSR_L3: 1
+        followbench_llmeval_en_HSR_L4: 1
+        followbench_llmeval_en_HSR_L5: 1
+        followbench_llmeval_en_SSR_L1: 1
+        followbench_llmeval_en_SSR_L2: 1
+        followbench_llmeval_en_SSR_L3: 1
+        followbench_llmeval_en_SSR_L4: 1
+        followbench_llmeval_en_SSR_L5: 1
+        simpleqa_f1: 0.12
+
+internlm2_5-7b-chat-turbomind_fullbench:
+    objective:
+        race-high_accuracy:  93.75
+        ARC-c_accuracy: 93.75
+        BoolQ_accuracy: 75.00
+        triviaqa_wiki_1shot_score: 50
+        nq_open_1shot_score: 25
+        IFEval_Prompt-level-strict-accuracy: 56.25
+        drop_accuracy: 75
+        GPQA_diamond_accuracy: 37.50
+        hellaswag_accuracy: 81.25
+        TheoremQA_score: 12.5
+        musr_average_naive_average: 39.58
+        korbench_single_naive_average: 40
+        gsm8k_accuracy: 68.75
+        math_accuracy: 68.75
+        cmo_fib_accuracy: 6.25
+        aime2024_accuracy: 6.25
+        wikibench-wiki-single_choice_cncircular_perf_4: 25
+        sanitized_mbpp_score: 68.75
+        ds1000_naive_average: 15.18
+        lcb_code_generation_pass@1: 12.5
+        lcb_code_execution_pass@1: 43.75
+        lcb_test_output_pass@1: 0.00
+        bbh-logical_deduction_seven_objects_score: 62.50
+        bbh-multistep_arithmetic_two_score: 62.50
+        mmlu-other_accuracy: 73.08
+        cmmlu-china-specific_accuracy: 75.42
+        mmlu_pro_math_accuracy: 25.00
+        ds1000_Pandas_accuracy: 0.00
+        ds1000_Numpy_accuracy: 0
+        ds1000_Tensorflow_accuracy: 12.5
+        ds1000_Scipy_accuracy: 18.75
+        ds1000_Sklearn_accuracy: 18.75
+        ds1000_Pytorch_accuracy: 12.50
+        ds1000_Matplotlib_accuracy: 43.75
+        openai_mmmlu_lite_AR-XY_accuracy: 37.5
+        college_naive_average: 12.50
+        college_knowledge_naive_average: 87.5
+    subjective:
+        alignment_bench_v1_1_总分: 0.72
+        alpaca_eval_total: 20.00
+        arenahard_score: 55.77
+        Followbench_naive_average: 1
+        CompassArena_naive_average: 39.00
+        mtbench101_avg: 7.90
+        wildbench_average: 0.00
+        simpleqa_accuracy_given_attempted: 1.00
+        chinese_simpleqa_given_attempted_accuracy: 1
+        alignment_bench_v1_1_专业能力: 8.70
+        alignment_bench_v1_1_数学计算: 0
+        alignment_bench_v1_1_基本任务: 0
+        alignment_bench_v1_1_逻辑推理: 0
+        alignment_bench_v1_1_中文理解: 0
+        alignment_bench_v1_1_文本写作: 0
+        alignment_bench_v1_1_角色扮演: 0
+        alignment_bench_v1_1_综合问答: 0
+        alpaca_eval_helpful_base: 20.00
+        compassarena_language_naive_average: 25.00
+        compassarena_knowledge_naive_average: 55.00
+        compassarena_reason_v2_naive_average: 35.00
+        compassarena_math_v2_naive_average: 55.00
+        compassarena_creationv2_zh_naive_average: 25.00
+        followbench_llmeval_en_HSR_AVG: 1
+        followbench_llmeval_en_SSR_AVG: 1
+        followbench_llmeval_en_HSR_L1: 1
+        followbench_llmeval_en_HSR_L2: 1
+        followbench_llmeval_en_HSR_L3: 1
+        followbench_llmeval_en_HSR_L4: 1
+        followbench_llmeval_en_HSR_L5: 1
+        followbench_llmeval_en_SSR_L1: 1
+        followbench_llmeval_en_SSR_L2: 1
+        followbench_llmeval_en_SSR_L3: 1
+        followbench_llmeval_en_SSR_L4: 1
+        followbench_llmeval_en_SSR_L5: 1
+        simpleqa_f1: 0.12
+
+internlm2_5-7b-hf_fullbench:
+    objective:
+        race-high_accuracy: 100
+        ARC-c_accuracy: 68.75
+        BoolQ_accuracy: 87.5
+        triviaqa_wiki_1shot_score: 43.75
+        nq_open_1shot_score: 43.75
+        drop_accuracy: 62.5
+        GPQA_diamond_accuracy: 62.5
+        hellaswag_accuracy: 93.75
+        TheoremQA_score: 18.75
+        winogrande_accuracy: 75
+        gsm8k_accuracy: 37.5
+        GaokaoBench_2010-2022_Math_II_MCQs_score: 62.5
+        GaokaoBench_2010-2022_Math_II_Fill-in-the-Blank_score: 0
+        math_accuracy: 12.5
+        wikibench-wiki-single_choice_cncircular_perf_4: 25
+        sanitized_mbpp_score: 56.25
+        dingo_en_192_score: 37.5
+        dingo_zh_170_score: 100
+        mmlu-other_accuracy: 76.92
+        cmmlu-china-specific_accuracy: 84.17
+        mmlu_pro_math_accuracy: 18.75
+        bbh-logical_deduction_seven_objects_score: 43.75
+        bbh-multistep_arithmetic_two_score: 56.25
+        college_naive_average: 12.5
+        college_knowledge_naive_average: 87.5
+
+internlm2_5-7b-turbomind_fullbench:
+    objective:
+        race-high_accuracy: 100
+        ARC-c_accuracy: 68.75
+        BoolQ_accuracy: 87.5
+        triviaqa_wiki_1shot_score: 43.75
+        nq_open_1shot_score: 43.75
+        drop_accuracy: 62.5
+        GPQA_diamond_accuracy: 68.75
+        hellaswag_accuracy: 93.75
+        TheoremQA_score: 18.75
+        winogrande_accuracy: 87.5
+        gsm8k_accuracy: 62.50
+        GaokaoBench_2010-2022_Math_II_MCQs_score: 93.75
+        GaokaoBench_2010-2022_Math_II_Fill-in-the-Blank_score: 0
+        math_accuracy: 6.25
+        wikibench-wiki-single_choice_cncircular_perf_4: 0.00
+        sanitized_mbpp_score: 62.50
+        dingo_en_192_score: 37.50
+        dingo_zh_170_score: 100.00
+        mmlu-other_accuracy: 78.37
+        cmmlu-china-specific_accuracy: 83.33
+        mmlu_pro_math_accuracy: 18.75
+        bbh-logical_deduction_seven_objects_score: 62.50
+        bbh-multistep_arithmetic_two_score: 50.00
+        college_naive_average: 12.5
+        college_knowledge_naive_average: 87.5
+
+internlm2_5-7b-turbomind:
+    objective:
+        race-high_accuracy: 89.28
+        ARC-c_accuracy: 52.2
+        BoolQ_accuracy: 89.72
+        triviaqa_wiki_1shot_score: 65.88
+        nq_open_1shot_score: 34.82
+        drop_accuracy: 68.1
+        bbh_naive_average: 72.15
+        GPQA_diamond_accuracy: 32.83
+        hellaswag_accuracy: 88.36
+        TheoremQA_score: 25
+        winogrande_accuracy: 81.29
+        gsm8k_accuracy: 74.68
+        GaokaoBench_weighted_average: 58.19
+        math_accuracy: 33.98
+        Mathbench_naive_average: 48.38
+        wikibench-wiki-single_choice_cncircular_perf_4: 29.1
+        cmmlu_naive_average: 78.94
+        mmlu_naive_average: 71.44
+        mmlu_pro_naive_average: 38.18
+        openai_humaneval_humaneval_pass@1: 59.76
+        openai_humaneval_v2_humaneval_pass@1: 57.93
+        sanitized_mbpp_score: 55.25
+        dingo_en_192_score: 60.94
+        dingo_zh_170_score: 67.65
+        mmlu-stem_accuracy: 63.72
+        mmlu-social-science_accuracy: 80.15
+        mmlu-humanities_accuracy: 74.27
+        mmlu-other_accuracy: 71.85
+        cmmlu-stem_accuracy: 67.07
+        cmmlu-social-science_accuracy: 81.49
+        cmmlu-humanities_accuracy: 85.84
+        cmmlu-other_accuracy: 82.69
+        cmmlu-china-specific_accuracy: 79.88
+        mmlu_pro_biology_accuracy: 58.58
+        mmlu_pro_business_accuracy: 28.01
+        mmlu_pro_chemistry_accuracy: 22.79
+        mmlu_pro_computer_science_accuracy: 39.02
+        mmlu_pro_economics_accuracy: 53.08
+        mmlu_pro_engineering_accuracy: 25.7
+        mmlu_pro_health_accuracy: 46.94
+        mmlu_pro_history_accuracy: 43.04
+        mmlu_pro_law_accuracy: 29.7
+        mmlu_pro_math_accuracy: 24.2
+        mmlu_pro_philosophy_accuracy: 42.48
+        mmlu_pro_physics_accuracy: 26.02
+        mmlu_pro_psychology_accuracy: 52.76
+        mmlu_pro_other_accuracy: 42.21
+        college_naive_average: 7.00
+        high_naive_average: 6.67
+        middle_naive_average: 26.67
+        primary_naive_average: 64.00
+        arithmetic_naive_average: 55
+        mathbench-a (average)_naive_average: 31.8
+        college_knowledge_naive_average: 58.23
+        high_knowledge_naive_average: 52.51
+        middle_knowledge_naive_average: 71.15
+        primary_knowledge_naive_average: 60.48
+        mathbench-t (average)_naive_average: 60.19
+    long_context:
+        Single-Needle-Retrieval(S-RT)-32000_naive_average: 100
+        Single-Needle-Retrieval-EN-32000_naive_average: 100
+        Single-Needle-Retrieval-ZH-32000_naive_average: 100
+        Single-Needle-Retrieval(S-RT)-100000_naive_average: 100
+        Single-Needle-Retrieval-EN-100000_naive_average: 100
+        Single-Needle-Retrieval-ZH-100000_naive_average: 100
+        Single-Needle-Retrieval(S-RT)-200000_naive_average: 100
+        Single-Needle-Retrieval-EN-200000_naive_average: 100
+        Single-Needle-Retrieval-ZH-200000_naive_average: 100
+        longbench_naive_average: 46.19
+        longbench_zh_naive_average: 49.3
+        longbench_en_naive_average: 43.97
+        longbench_single-document-qa_score: 42.84
+        longbench_multi-document-qa_score: 41.25
+        longbench_summarization_score: 23.21
+        longbench_few-shot-learning_score: 61.67
+        longbench_synthetic-tasks_score: 60.05
+        longbench_code-completion_score: 52.09
+
+internlm2_5-7b-chat-turbomind:
+    objective:
+        race-high_accuracy: 86.16
+        ARC-c_accuracy: 90.17
+        BoolQ_accuracy: 87.89
+        triviaqa_wiki_1shot_score: 64.91
+        nq_open_1shot_score: 22.69
+        mmmlu_lite_naive_average: 44.96
+        IFEval_Prompt-level-strict-accuracy: 58.04
+        drop_accuracy: 77.68
+        bbh_naive_average: 73.14
+        GPQA_diamond_accuracy: 31.06
+        hellaswag_accuracy: 94.79
+        TheoremQA_score: 22.25
+        musr_average_naive_average: 50.89
+        korbench_single_naive_average: 32.16
+        ARC_Prize_Public_Evaluation_accuracy: 0.02
+        gsm8k_accuracy: 86.73
+        GaokaoBench_weighted_average: 78.6
+        math_accuracy: 61
+        cmo_fib_accuracy: 11
+        aime2024_accuracy: 3.33
+        Mathbench_naive_average: 64.23
+        wikibench-wiki-single_choice_cncircular_perf_4: 31.32
+        cmmlu_naive_average: 74.3
+        mmlu_naive_average: 70.84
+        mmlu_pro_naive_average: 44.98
+        openai_humaneval_humaneval_pass@1: 69.8
+        sanitized_mbpp_score: 64.4
+        humanevalx_naive_average: 33.35
+        ds1000_naive_average: 14.15
+        lcb_code_generation_pass@1: 17.75
+        lcb_code_execution_pass@1: 32.57
+        lcb_test_output_pass@1: 26.13
+        bigcodebench_hard_instruct_pass@1: 3.38
+        bigcodebench_hard_complete_pass@1: 5.06
+        teval_naive_average: 80
+        SciCode_sub_accuracy: 5.56
+        qa_dingo_cn_score: 99.01
+        mmlu-stem_accuracy: 68.2
+        mmlu-social-science_accuracy: 75.8
+        mmlu-humanities_accuracy: 69.3
+        mmlu-other_accuracy: 71.3
+        cmmlu-stem_accuracy: 66.64
+        cmmlu-social-science_accuracy: 76
+        cmmlu-humanities_accuracy: 77.9
+        cmmlu-other_accuracy: 77.25
+        cmmlu-china-specific_accuracy: 73.6
+        mmlu_pro_biology_accuracy: 66.67
+        mmlu_pro_business_accuracy: 47.91
+        mmlu_pro_chemistry_accuracy: 35
+        mmlu_pro_computer_science_accuracy: 48.9
+        mmlu_pro_economics_accuracy: 55.87
+        mmlu_pro_engineering_accuracy: 29.62
+        mmlu_pro_health_accuracy: 45
+        mmlu_pro_history_accuracy: 40.8
+        mmlu_pro_law_accuracy: 25.79
+        mmlu_pro_math_accuracy: 53.48
+        mmlu_pro_philosophy_accuracy: 38.38
+        mmlu_pro_physics_accuracy: 37.79
+        mmlu_pro_psychology_accuracy: 58.39
+        mmlu_pro_other_accuracy: 46.27
+        humanevalx-python_pass@1: 53.66
+        humanevalx-cpp_pass@1: 22.56
+        humanevalx-go_pass@1: 0
+        humanevalx-js_pass@1: 54.88
+        ds1000_Pandas_accuracy: 10.65
+        ds1000_Numpy_accuracy: 3.63
+        ds1000_Tensorflow_accuracy: 13.33
+        ds1000_Scipy_accuracy: 8.96
+        ds1000_Sklearn_accuracy: 6.96
+        ds1000_Pytorch_accuracy: 6.62
+        ds1000_Matplotlib_accuracy: 49.35
+        openai_mmmlu_lite_AR-XY_accuracy: 17.19
+        openai_mmmlu_lite_BN-BD_accuracy: 26.78
+        openai_mmmlu_lite_DE-DE_accuracy: 51.27
+        openai_mmmlu_lite_ES-LA_accuracy: 56.94
+        openai_mmmlu_lite_FR-FR_accuracy: 58.22
+        openai_mmmlu_lite_HI-IN_accuracy: 30.75
+        openai_mmmlu_lite_ID-ID_accuracy: 50.6
+        openai_mmmlu_lite_IT-IT_accuracy: 50.6
+        openai_mmmlu_lite_JA-JP_accuracy: 51.13
+        openai_mmmlu_lite_KO-KR_accuracy: 45
+        openai_mmmlu_lite_PT-BR_accuracy: 57.68
+        openai_mmmlu_lite_SW-KE_accuracy: 32.56
+        openai_mmmlu_lite_YO-NG_accuracy: 32.42
+        openai_mmmlu_lite_ZH-CN_accuracy: 65.4
+        college_naive_average: 19.17
+        high_naive_average: 46.5
+        middle_naive_average: 61.34
+        primary_naive_average: 73.34
+        arithmetic_naive_average: 61.67
+        mathbench-a (average)_naive_average: 52.58
+        college_knowledge_naive_average: 67.1
+        high_knowledge_naive_average: 70
+        middle_knowledge_naive_average: 80
+        primary_knowledge_naive_average: 90.12
+        mathbench-t (average)_naive_average: 76
+    subjective:
+        alignment_bench_v1_1_总分: 5.68
+        alpaca_eval_total: 25.96
+        arenahard_score: 17.15
+        Followbench_naive_average: 0.81
+        CompassArena_naive_average: 39.49
+        FoFo_naive_average: 0.38
+        mtbench101_avg: 8.01
+        wildbench_average: -10.49
+        simpleqa_accuracy_given_attempted: 0.04
+        chinese_simpleqa_given_attempted_accuracy: 0.34
+        alignment_bench_v1_1_专业能力: 6.05
+        alignment_bench_v1_1_数学计算: 5.87
+        alignment_bench_v1_1_基本任务: 6.01
+        alignment_bench_v1_1_逻辑推理: 4.48
+        alignment_bench_v1_1_中文理解: 6.17
+        alignment_bench_v1_1_文本写作: 6.06
+        alignment_bench_v1_1_角色扮演: 6.3
+        alignment_bench_v1_1_综合问答: 6.45
+        alpaca_eval_helpful_base: 17.83
+        alpaca_eval_koala: 28.21
+        alpaca_eval_oasst: 23.4
+        alpaca_eval_selfinstruct: 30.95
+        alpaca_eval_vicuna: 25.00
+        compassarena_language_naive_average: 53.00
+        compassarena_knowledge_naive_average: 36
+        compassarena_reason_v2_naive_average: 35
+        compassarena_math_v2_naive_average: 16.07
+        compassarena_creationv2_zh_naive_average: 43.64
+        fofo_test_prompts_overall: 0.35
+        fofo_test_prompts_cn_overall: 0.41
+        followbench_llmeval_en_HSR_AVG: 0.73
+        followbench_llmeval_en_SSR_AVG: 0.88
+        followbench_llmeval_en_HSR_L1: 0.94
+        followbench_llmeval_en_HSR_L2: 0.77
+        followbench_llmeval_en_HSR_L3: 0.73
+        followbench_llmeval_en_HSR_L4: 0.68
+        followbench_llmeval_en_HSR_L5: 0.54
+        followbench_llmeval_en_SSR_L1: 0.94
+        followbench_llmeval_en_SSR_L2: 0.88
+        followbench_llmeval_en_SSR_L3: 0.87
+        followbench_llmeval_en_SSR_L4: 0.87
+        followbench_llmeval_en_SSR_L5: 0.85
+        simpleqa_f1: 0.04
+
+internlm2_5-7b-chat-1m-turbomind:
+    long_context:
+        ruler_8k_naive_average: 88.53
+        ruler_32k_naive_average: 83.84
+        ruler_128k_naive_average: 70.94
+        NeedleBench-Overall-Score-8K_weighted_average: 91.89
+        NeedleBench-Overall-Score-32K_weighted_average: 91.42
+        NeedleBench-Overall-Score-128K_weighted_average: 88.57
+        longbench_naive_average: 46.44
+        longbench_zh_naive_average: 45.19
+        longbench_en_naive_average: 45.71
+        babilong_0k_naive_average: 79.3
+        babilong_4k_naive_average: 67
+        babilong_16k_naive_average: 52.7
+        babilong_32k_naive_average: 48.9
+        babilong_128k_naive_average: 40.8
+        babilong_256k_naive_average: 23.5
+        longbench_single-document-qa_score: 43.56
+        longbench_multi-document-qa_score: 46.24
+        longbench_summarization_score: 24.32
+        longbench_few-shot-learning_score: 51.67
+        longbench_synthetic-tasks_score: 66.83
+        longbench_code-completion_score: 45.99
+
+
+qwen2.5-7b-instruct-turbomind:
+    objective:
+        race-high_accuracy: 84.99
+        ARC-c_accuracy: 92.2
+        BoolQ_accuracy: 86.7
+        triviaqa_wiki_1shot_score: 53.06
+        nq_open_1shot_score: 17.51
+        mmmlu_lite_naive_average: 54.96
+        IFEval_Prompt-level-strict-accuracy: 71.53
+        drop_accuracy: 80.07
+        bbh_naive_average: 68.81
+        GPQA_diamond_accuracy: 34.34
+        hellaswag_accuracy: 85.42
+        TheoremQA_score: 18.38
+        musr_average_naive_average: 43.44
+        korbench_single_naive_average: 39.44
+        ARC_Prize_Public_Evaluation_accuracy: 0
+        gsm8k_accuracy: 92.57
+        GaokaoBench_weighted_average: 80.14
+        math_accuracy: 73.58
+        cmo_fib_accuracy: 25
+        aime2024_accuracy: 16.67
+        Mathbench_naive_average: 77.33
+        wikibench-wiki-single_choice_cncircular_perf_4: 34.9
+        cmmlu_naive_average: 75.97
+        mmlu_naive_average: 76.01
+        mmlu_pro_naive_average: 56.12
+        openai_humaneval_humaneval_pass@1: 83.54
+        sanitized_mbpp_score: 74.71
+        humanevalx_naive_average: 48.29
+        ds1000_naive_average: 18.66
+        lcb_code_generation_pass@1: 39.5
+        lcb_code_execution_pass@1: 42.38
+        lcb_test_output_pass@1: 50.68
+        bigcodebench_hard_instruct_pass@1: 16.22
+        bigcodebench_hard_complete_pass@1: 11.49
+        teval_naive_average: 79.72
+        SciCode_sub_accuracy: 10.76
+        qa_dingo_cn_score: 99.01
+        mmlu_accuracy: 76.01
+        mmlu-stem_accuracy: 77.59
+        mmlu-social-science_accuracy: 79.02
+        mmlu-humanities_accuracy: 72.07
+        mmlu-other_accuracy: 74.86
+        cmmlu_accuracy: 75.97
+        cmmlu-stem_accuracy: 73.09
+        cmmlu-social-science_accuracy: 75.95
+        cmmlu-humanities_accuracy: 76.53
+        cmmlu-other_accuracy: 78.79
+        cmmlu-china-specific_accuracy: 73.17
+        mmlu_pro_accuracy: 56.12
+        mmlu_pro_biology_accuracy: 71.41
+        mmlu_pro_business_accuracy: 67.68
+        mmlu_pro_chemistry_accuracy: 54.59
+        mmlu_pro_computer_science_accuracy: 58.29
+        mmlu_pro_economics_accuracy: 66.82
+        mmlu_pro_engineering_accuracy: 42.41
+        mmlu_pro_health_accuracy: 55.87
+        mmlu_pro_history_accuracy: 46.46
+        mmlu_pro_law_accuracy: 28.97
+        mmlu_pro_math_accuracy: 73.13
+        mmlu_pro_philosophy_accuracy: 44.89
+        mmlu_pro_physics_accuracy: 58.43
+        mmlu_pro_psychology_accuracy: 63.16
+        mmlu_pro_other_accuracy: 53.57
+        humanevalx-python_pass@1: 50
+        humanevalx-cpp_pass@1: 42.07
+        humanevalx-go_pass@1: 0
+        humanevalx-java_pass@1: 53.05
+        humanevalx-js_pass@1: 75
+        ds1000_Pandas_accuracy: 14.09
+        ds1000_Numpy_accuracy: 8.18
+        ds1000_Tensorflow_accuracy: 17.78
+        ds1000_Scipy_accuracy: 15.09
+        ds1000_Sklearn_accuracy: 10.43
+        ds1000_Pytorch_accuracy: 4.41
+        ds1000_Matplotlib_accuracy: 60.65
+        mmmlu_lite_accuracy: 54.96
+        openai_mmmlu_lite_AR-XY_accuracy: 42.32
+        openai_mmmlu_lite_BN-BD_accuracy: 42.25
+        openai_mmmlu_lite_DE-DE_accuracy: 59.93
+        openai_mmmlu_lite_ES-LA_accuracy: 66.53
+        openai_mmmlu_lite_FR-FR_accuracy: 66.88
+        openai_mmmlu_lite_HI-IN_accuracy: 49.26
+        openai_mmmlu_lite_ID-ID_accuracy: 61.26
+        openai_mmmlu_lite_IT-IT_accuracy: 65.47
+        openai_mmmlu_lite_JA-JP_accuracy: 61.54
+        openai_mmmlu_lite_KO-KR_accuracy: 60.28
+        openai_mmmlu_lite_PT-BR_accuracy: 55.51
+        openai_mmmlu_lite_SW-KE_accuracy: 36.42
+        openai_mmmlu_lite_YO-NG_accuracy: 32.14
+        openai_mmmlu_lite_ZH-CN_accuracy: 69.61
+        college_naive_average: 44.33
+        high_naive_average: 59
+        middle_naive_average: 78
+        primary_naive_average: 85.67
+        arithmetic_naive_average: 75.67
+        mathbench-a (average)_naive_average: 69.27
+        college_knowledge_naive_average: 83.86
+        high_knowledge_naive_average: 80.29
+        middle_knowledge_naive_average: 84.26
+        primary_knowledge_naive_average: 93.16
+        mathbench-t (average)_naive_average: 85.39
+
+
+
+
+internlm2_5-7b-chat-pytorch:
+    objective:
+        race-high_accuracy: 86.39
+        ARC-c_accuracy: 90.51
+        BoolQ_accuracy: 88.01
+        triviaqa_wiki_1shot_score: 64.77
+        nq_open_1shot_score: 22.71
+        mmmlu_lite_naive_average: 45.02
+        IFEval_Prompt-level-strict-accuracy: 56.56
+        drop_accuracy: 75.46
+        bbh_naive_average: 73.34
+        GPQA_diamond_accuracy: 32.83
+        hellaswag_accuracy: 94.81
+        TheoremQA_score: 23.88
+        musr_average_naive_average: 51.31
+        korbench_single_naive_average: 32
+        ARC_Prize_Public_Evaluation_accuracy: 0.01
+        gsm8k_accuracy: 86.96
+        GaokaoBench_weighted_average: 78.05
+        math_accuracy: 60.34
+        cmo_fib_accuracy: 12.98
+        aime2024_accuracy: 3.33
+        Mathbench_naive_average: 64.82
+        wikibench-wiki-single_choice_cncircular_perf_4: 31.7
+        cmmlu_naive_average: 74.24
+        mmlu_naive_average: 70.2
+        mmlu_pro_naive_average: 45.39
+        openai_humaneval_humaneval_pass@1: 70.12
+        sanitized_mbpp_score: 64.59
+        humanevalx_naive_average: 38.78
+        ds1000_naive_average: 14.19
+        lcb_code_generation_pass@1: 16.5
+        lcb_code_execution_pass@1: 33.82
+        lcb_test_output_pass@1: 22.62
+        bigcodebench_hard_instruct_pass@1: 6.08
+        bigcodebench_hard_complete_pass@1: 6.76
+        teval_naive_average: 79.73
+        SciCode_sub_accuracy: 3.47
+        qa_dingo_cn_score: 100
+        mmlu_accuracy: 70.2
+        mmlu-stem_accuracy: 67.73
+        mmlu-social-science_accuracy: 75.49
+        mmlu-humanities_accuracy: 68.56
+        mmlu-other_accuracy: 70.58
+        cmmlu_accuracy: 74.24
+        cmmlu-stem_accuracy: 66.7
+        cmmlu-social-science_accuracy: 75.88
+        cmmlu-humanities_accuracy: 77.56
+        cmmlu-other_accuracy: 77.52
+        cmmlu-china-specific_accuracy: 73.46
+        mmlu_pro_accuracy: 45.39
+        mmlu_pro_biology_accuracy: 65.83
+        mmlu_pro_business_accuracy: 51.96
+        mmlu_pro_chemistry_accuracy: 36.84
+        mmlu_pro_computer_science_accuracy: 48.29
+        mmlu_pro_economics_accuracy: 56.16
+        mmlu_pro_engineering_accuracy: 29.1
+        mmlu_pro_health_accuracy: 44.5
+        mmlu_pro_history_accuracy: 42.26
+        mmlu_pro_law_accuracy: 24.98
+        mmlu_pro_math_accuracy: 54.85
+        mmlu_pro_philosophy_accuracy: 39.28
+        mmlu_pro_physics_accuracy: 37.41
+        mmlu_pro_psychology_accuracy: 58.27
+        mmlu_pro_other_accuracy: 45.78
+        humanevalx-python_pass@1: 56.1
+        humanevalx-cpp_pass@1: 20.73
+        humanevalx-go_pass@1: 0
+        humanevalx-java_pass@1: 59.15
+        humanevalx-js_pass@1: 57.93
+        ds1000_Pandas_accuracy: 8.93
+        ds1000_Numpy_accuracy: 4.09
+        ds1000_Tensorflow_accuracy: 11.11
+        ds1000_Scipy_accuracy: 7.55
+        ds1000_Sklearn_accuracy: 7.83
+        ds1000_Pytorch_accuracy: 8.82
+        ds1000_Matplotlib_accuracy: 50.97
+        mmmlu_lite_accuracy: 45.02
+        openai_mmmlu_lite_AR-XY_accuracy: 18.6
+        openai_mmmlu_lite_BN-BD_accuracy: 27.58
+        openai_mmmlu_lite_DE-DE_accuracy: 51.23
+        openai_mmmlu_lite_ES-LA_accuracy: 56.63
+        openai_mmmlu_lite_FR-FR_accuracy: 58.11
+        openai_mmmlu_lite_HI-IN_accuracy: 33.82
+        openai_mmmlu_lite_ID-ID_accuracy: 50.39
+        openai_mmmlu_lite_IT-IT_accuracy: 50.39
+        openai_mmmlu_lite_JA-JP_accuracy: 50.95
+        openai_mmmlu_lite_KO-KR_accuracy: 45.05
+        openai_mmmlu_lite_PT-BR_accuracy: 57.89
+        openai_mmmlu_lite_SW-KE_accuracy: 32.14
+        openai_mmmlu_lite_YO-NG_accuracy: 32.14
+        openai_mmmlu_lite_ZH-CN_accuracy: 65.33
+        college_naive_average: 21
+        high_naive_average: 47
+        middle_naive_average: 59.67
+        primary_naive_average: 72.33
+        arithmetic_naive_average: 62
+        mathbench-a (average)_naive_average: 53.13
+        college_knowledge_naive_average: 68.99
+        high_knowledge_naive_average: 70.06
+        middle_knowledge_naive_average: 78.53
+        primary_knowledge_naive_average: 88.49
+        mathbench-t (average)_naive_average: 76.51
+
+
+qwen2.5-7b-instruct-pytorch:
+    objective:
+        race-high_accuracy: 85.16
+        ARC-c_accuracy: 90.85
+        BoolQ_accuracy: 86.61
+        triviaqa_wiki_1shot_score: 52.96
+        nq_open_1shot_score: 17.62
+        mmmlu_lite_naive_average: 54.7
+        IFEval_Prompt-level-strict-accuracy: 71.35
+        drop_accuracy: 80.23
+        bbh_naive_average: 68.88
+        GPQA_diamond_accuracy: 36.36
+        hellaswag_accuracy: 85.49
+        TheoremQA_score: 18.38
+        musr_average_naive_average: 43.3
+        korbench_single_naive_average: 39.44
+        ARC_Prize_Public_Evaluation_accuracy: 0
+        gsm8k_accuracy: 91.66
+        GaokaoBench_weighted_average: 80.02
+        math_accuracy: 73.74
+        cmo_fib_accuracy: 22.60
+        aime2024_accuracy: 13.33
+        Mathbench_naive_average: 77.08
+        wikibench-wiki-single_choice_cncircular_perf_4: 34
+        cmmlu_naive_average: 75.9
+        mmlu_naive_average: 76.27
+        mmlu_pro_naive_average: 56.14
+        openai_humaneval_humaneval_pass@1: 84.76
+        sanitized_mbpp_score: 74.71
+        humanevalx_naive_average: 48.17
+        ds1000_naive_average: 18.57
+        lcb_code_generation_pass@1: 38.75
+        lcb_code_execution_pass@1: 42.38
+        lcb_test_output_pass@1: 50.45
+        bigcodebench_hard_instruct_pass@1: 16.89
+        bigcodebench_hard_complete_pass@1: 12.16
+        teval_naive_average: 79.46
+        SciCode_sub_accuracy: 10.42
+        qa_dingo_cn_score: 100
+        mmlu_accuracy: 76.27
+        mmlu-stem_accuracy: 77.75
+        mmlu-social-science_accuracy: 78.65
+        mmlu-humanities_accuracy: 73.12
+        mmlu-other_accuracy: 75.05
+        cmmlu_accuracy: 75.9
+        cmmlu-stem_accuracy: 73.41
+        cmmlu-social-science_accuracy: 75.97
+        cmmlu-humanities_accuracy: 76.42
+        cmmlu-other_accuracy: 78.15
+        cmmlu-china-specific_accuracy: 73.27
+        mmlu_pro_accuracy: 56.14
+        mmlu_pro_biology_accuracy: 72.25
+        mmlu_pro_business_accuracy: 66.16
+        mmlu_pro_chemistry_accuracy: 55.65
+        mmlu_pro_computer_science_accuracy: 60.24
+        mmlu_pro_economics_accuracy: 66.82
+        mmlu_pro_engineering_accuracy: 41.38
+        mmlu_pro_health_accuracy: 54.89
+        mmlu_pro_history_accuracy: 46.46
+        mmlu_pro_law_accuracy: 29.06
+        mmlu_pro_math_accuracy: 73.58
+        mmlu_pro_philosophy_accuracy: 44.89
+        mmlu_pro_physics_accuracy: 60.05
+        mmlu_pro_psychology_accuracy: 61.9
+        mmlu_pro_other_accuracy: 52.6
+        humanevalx-python_pass@1: 51.83
+        humanevalx-cpp_pass@1: 42.68
+        humanevalx-go_pass@1: 0
+        humanevalx-java_pass@1: 73.78
+        humanevalx-js_pass@1: 72.56
+        ds1000_Pandas_accuracy: 14.09
+        ds1000_Numpy_accuracy: 8.64
+        ds1000_Tensorflow_accuracy: 17.78
+        ds1000_Scipy_accuracy: 15.09
+        ds1000_Sklearn_accuracy: 8.7
+        ds1000_Pytorch_accuracy: 4.41
+        ds1000_Matplotlib_accuracy: 61.29
+        mmmlu_lite_accuracy: 54.7
+        openai_mmmlu_lite_AR-XY_accuracy: 42.32
+        openai_mmmlu_lite_BN-BD_accuracy: 42.18
+        openai_mmmlu_lite_DE-DE_accuracy: 60
+        openai_mmmlu_lite_ES-LA_accuracy: 66.18
+        openai_mmmlu_lite_FR-FR_accuracy: 66.88
+        openai_mmmlu_lite_HI-IN_accuracy: 48.63
+        openai_mmmlu_lite_ID-ID_accuracy: 61.26
+        openai_mmmlu_lite_IT-IT_accuracy: 65.26
+        openai_mmmlu_lite_JA-JP_accuracy: 60.7
+        openai_mmmlu_lite_KO-KR_accuracy: 60.63
+        openai_mmmlu_lite_PT-BR_accuracy: 54.46
+        openai_mmmlu_lite_SW-KE_accuracy: 36
+        openai_mmmlu_lite_YO-NG_accuracy: 31.86
+        openai_mmmlu_lite_ZH-CN_accuracy: 69.4
+        college_naive_average: 48.33
+        high_naive_average: 59.33
+        middle_naive_average: 76.67
+        primary_naive_average: 86.67
+        arithmetic_naive_average: 74.33
+        mathbench-a (average)_naive_average: 69.07
+        college_knowledge_naive_average: 83.54
+        high_knowledge_naive_average: 80.82
+        middle_knowledge_naive_average: 83.79
+        primary_knowledge_naive_average: 92.22
+        mathbench-t (average)_naive_average: 85.1
+
+
+internlm3-8b-instruct-turbomind:
+    objective:
+        race-high_accuracy: 89.22
+        ARC-c_accuracy: 92.54
+        BoolQ_accuracy: 86.45
+        triviaqa_wiki_1shot_score: 60.72
+        nq_open_1shot_score: 20.25
+        mmmlu_lite_naive_average: 41.82
+        IFEval_Prompt-level-strict-accuracy: 77.45
+        drop_accuracy: 83.27
+        bbh_naive_average: 55.22
+        GPQA_diamond_accuracy: 37.88
+        hellaswag_accuracy: 91.28
+        TheoremQA_score: 20.12
+        musr_average_naive_average: 36.86
+        korbench_single_naive_average: 41.2
+        ARC_Prize_Public_Evaluation_accuracy: 0.06
+        gsm8k_accuracy: 91.28
+        GaokaoBench_weighted_average: 86.59
+        math_accuracy: 76.96
+        cmo_fib_accuracy: 38.46
+        aime2024_accuracy: 13.33
+        Mathbench_naive_average: 78.96
+        wikibench-wiki-single_choice_cncircular_perf_4: 37.45
+        cmmlu_naive_average: 83.33
+        mmlu_naive_average: 76.21
+        mmlu_pro_naive_average: 57.96
+        openai_humaneval_humaneval_pass@1: 81.71
+        sanitized_mbpp_score: 69.65
+        humanevalx_naive_average: 40.73
+        ds1000_naive_average: 27.23
+        lcb_code_generation_pass@1: 34.75
+        lcb_code_execution_pass@1: 49.9
+        lcb_test_output_pass@1: 48.19
+        bigcodebench_hard_instruct_pass@1: 13.51
+        bigcodebench_hard_complete_pass@1: 15.54
+        teval_naive_average: 82.86
+        SciCode_sub_accuracy: 11.11
+        qa_dingo_cn_score: 100
+        mmlu_accuracy: 76.21
+        mmlu-stem_accuracy: 77.7
+        mmlu-social-science_accuracy: 80.98
+        mmlu-humanities_accuracy: 70.83
+        mmlu-other_accuracy: 75.01
+        cmmlu_accuracy: 83.33
+        cmmlu-stem_accuracy: 79.66
+        cmmlu-social-science_accuracy: 83.39
+        cmmlu-humanities_accuracy: 84.73
+        cmmlu-other_accuracy: 86.2
+        cmmlu-china-specific_accuracy: 81.77
+        mmlu_pro_accuracy: 57.96
+        mmlu_pro_biology_accuracy: 75.45
+        mmlu_pro_business_accuracy: 64.64
+        mmlu_pro_chemistry_accuracy: 59.81
+        mmlu_pro_computer_science_accuracy: 60.24
+        mmlu_pro_economics_accuracy: 68.6
+        mmlu_pro_engineering_accuracy: 44.79
+        mmlu_pro_health_accuracy: 58.31
+        mmlu_pro_history_accuracy: 49.87
+        mmlu_pro_law_accuracy: 32.43
+        mmlu_pro_math_accuracy: 70.17
+        mmlu_pro_philosophy_accuracy: 46.89
+        mmlu_pro_physics_accuracy: 59.58
+        mmlu_pro_psychology_accuracy: 66.29
+        mmlu_pro_other_accuracy: 54.33
+        humanevalx-python_pass@1: 43.9
+        humanevalx-cpp_pass@1: 20.12
+        humanevalx-go_pass@1: 0
+        humanevalx-java_pass@1: 40.85
+        humanevalx-js_pass@1: 65.24
+        ds1000_Pandas_accuracy: 16.49
+        ds1000_Numpy_accuracy: 34.09
+        ds1000_Tensorflow_accuracy: 26.67
+        ds1000_Scipy_accuracy: 17.92
+        ds1000_Sklearn_accuracy: 20.87
+        ds1000_Pytorch_accuracy: 19.12
+        ds1000_Matplotlib_accuracy: 55.48
+        mmmlu_lite_accuracy: 41.82
+        openai_mmmlu_lite_AR-XY_accuracy: 32.56
+        openai_mmmlu_lite_BN-BD_accuracy: 4.56
+        openai_mmmlu_lite_DE-DE_accuracy: 24.91
+        openai_mmmlu_lite_ES-LA_accuracy: 51.09
+        openai_mmmlu_lite_FR-FR_accuracy: 61.68
+        openai_mmmlu_lite_HI-IN_accuracy: 24.98
+        openai_mmmlu_lite_ID-ID_accuracy: 44.56
+        openai_mmmlu_lite_IT-IT_accuracy: 52.35
+        openai_mmmlu_lite_JA-JP_accuracy: 51.02
+        openai_mmmlu_lite_KO-KR_accuracy: 47.93
+        openai_mmmlu_lite_PT-BR_accuracy: 53.89
+        openai_mmmlu_lite_SW-KE_accuracy: 33.47
+        openai_mmmlu_lite_YO-NG_accuracy: 33.47
+        openai_mmmlu_lite_ZH-CN_accuracy: 69.05
+        college_naive_average: 45.67
+        high_naive_average: 64.67
+        middle_naive_average: 82.33
+        primary_naive_average: 90.33
+        arithmetic_naive_average: 74
+        mathbench-a (average)_naive_average: 71.4
+        college_knowledge_naive_average: 85.28
+        high_knowledge_naive_average: 79.43
+        middle_knowledge_naive_average: 87.9
+        primary_knowledge_naive_average: 93.42
+        mathbench-t (average)_naive_average: 86.51
+
+
+internlm3-8b-instruct-pytorch:
+    objective:
+        race-high_accuracy: 89.02
+        ARC-c_accuracy: 93.56
+        BoolQ_accuracy: 86.67
+        triviaqa_wiki_1shot_score: 60.54
+        nq_open_1shot_score: 20.3
+        mmmlu_lite_naive_average: 42.6
+        IFEval_Prompt-level-strict-accuracy: 79.11
+        drop_accuracy: 83.32
+        bbh_naive_average: 54.76
+        GPQA_diamond_accuracy: 33.84
+        hellaswag_accuracy: 91.31
+        TheoremQA_score: 18
+        musr_average_naive_average: 36.62
+        korbench_single_naive_average: 41.84
+        ARC_Prize_Public_Evaluation_accuracy: 0.06
+        gsm8k_accuracy: 90.67
+        GaokaoBench_weighted_average: 86.27
+        math_accuracy: 76.68
+        cmo_fib_accuracy: 33.65
+        aime2024_accuracy: 10
+        Mathbench_naive_average: 78.92
+        wikibench-wiki-single_choice_cncircular_perf_4: 37.35
+        cmmlu_naive_average: 83.11
+        mmlu_naive_average: 76.23
+        mmlu_pro_naive_average: 58.16
+        openai_humaneval_humaneval_pass@1: 82.32
+        sanitized_mbpp_score: 70.04
+        humanevalx_naive_average: 25.49
+        ds1000_naive_average: 27.84
+        lcb_code_generation_pass@1: 34.5
+        lcb_code_execution_pass@1: 48.02
+        lcb_test_output_pass@1: 47.74
+        bigcodebench_hard_instruct_pass@1: 12.84
+        bigcodebench_hard_complete_pass@1: 15.54
+        teval_naive_average: 82.86
+        SciCode_sub_accuracy: 9.38
+        qa_dingo_cn_score: 100
+        mmlu_accuracy: 76.23
+        mmlu-stem_accuracy: 78.08
+        mmlu-social-science_accuracy: 80.31
+        mmlu-humanities_accuracy: 71.38
+        mmlu-other_accuracy: 74.63
+        cmmlu_accuracy: 83.11
+        cmmlu-stem_accuracy: 79.42
+        cmmlu-social-science_accuracy: 83.34
+        cmmlu-humanities_accuracy: 83.95
+        cmmlu-other_accuracy: 86.22
+        cmmlu-china-specific_accuracy: 81.5
+        mmlu_pro_accuracy: 58.16
+        mmlu_pro_biology_accuracy: 74.62
+        mmlu_pro_business_accuracy: 65.02
+        mmlu_pro_chemistry_accuracy: 60.69
+        mmlu_pro_computer_science_accuracy: 61.46
+        mmlu_pro_economics_accuracy: 68.25
+        mmlu_pro_engineering_accuracy: 45.3
+        mmlu_pro_health_accuracy: 60.15
+        mmlu_pro_history_accuracy: 50.66
+        mmlu_pro_law_accuracy: 31.7
+        mmlu_pro_math_accuracy: 70.32
+        mmlu_pro_philosophy_accuracy: 47.7
+        mmlu_pro_physics_accuracy: 59.51
+        mmlu_pro_psychology_accuracy: 65.41
+        mmlu_pro_other_accuracy: 53.46
+        humanevalx-python_pass@1: 42.68
+        humanevalx-cpp_pass@1: 19.51
+        humanevalx-go_pass@1: 0
+        humanevalx-java_pass@1: 0.00
+        humanevalx-js_pass@1: 64.02
+        ds1000_Pandas_accuracy: 14.09
+        ds1000_Numpy_accuracy: 35
+        ds1000_Tensorflow_accuracy: 24.44
+        ds1000_Scipy_accuracy: 20.75
+        ds1000_Sklearn_accuracy: 21.74
+        ds1000_Pytorch_accuracy: 22.06
+        ds1000_Matplotlib_accuracy: 56.77
+        mmmlu_lite_accuracy: 42.6
+        openai_mmmlu_lite_AR-XY_accuracy: 32.84
+        openai_mmmlu_lite_BN-BD_accuracy: 10.46
+        openai_mmmlu_lite_DE-DE_accuracy: 24.56
+        openai_mmmlu_lite_ES-LA_accuracy: 50.95
+        openai_mmmlu_lite_FR-FR_accuracy: 61.05
+        openai_mmmlu_lite_HI-IN_accuracy: 30.6
+        openai_mmmlu_lite_ID-ID_accuracy: 45.89
+        openai_mmmlu_lite_IT-IT_accuracy: 51.79
+        openai_mmmlu_lite_JA-JP_accuracy: 51.65
+        openai_mmmlu_lite_KO-KR_accuracy: 48.77
+        openai_mmmlu_lite_PT-BR_accuracy: 52.7
+        openai_mmmlu_lite_SW-KE_accuracy: 32.91
+        openai_mmmlu_lite_YO-NG_accuracy: 32.84
+        openai_mmmlu_lite_ZH-CN_accuracy: 69.33
+        college_naive_average: 47
+        high_naive_average: 66.67
+        middle_naive_average: 81.67
+        primary_naive_average: 89.33
+        arithmetic_naive_average: 73.67
+        mathbench-a (average)_naive_average: 71.67
+        college_knowledge_naive_average: 82.91
+        high_knowledge_naive_average: 79.86
+        middle_knowledge_naive_average: 88.92
+        primary_knowledge_naive_average: 92.96
+        mathbench-t (average)_naive_average: 86.16

.github/scripts/oc_score_baseline_testrange.yaml

@@ -0,0 +1,432 @@
+chat:
+    glm-4-9b-chat-hf:
+        gsm8k_accuracy: 56.25
+        race-high_accuracy: 84.38
+    glm-4-9b-chat-turbomind:
+        gsm8k_accuracy: 71.88
+        race-high_accuracy: 90.62
+    glm-4-9b-chat-vllm:
+        gsm8k_accuracy: 71.88
+        race-high_accuracy: 90.62
+    deepseek-7b-chat-hf:
+        gsm8k_accuracy: 46.88
+        race-high_accuracy: 81.25
+    deepseek-r1-distill-llama-8b-turbomind:
+        gsm8k_accuracy: 34.38
+        race-high_accuracy: 81.25
+    deepseek-r1-distill-qwen-1_5b-turbomind:
+        gsm8k_accuracy: 28.12
+        race-high_accuracy: 53.12
+    deepseek-7b-chat-vllm:
+        gsm8k_accuracy: 56.25
+        race-high_accuracy: 78.12
+    gemma2-2b-it-hf:
+        gsm8k_accuracy: 50
+        race-high_accuracy: 75
+    gemma2-9b-it-hf:
+        gsm8k_accuracy: 68.75
+        race-high_accuracy: 84.38
+    gemma-2b-it-hf:
+        gsm8k_accuracy: 3.12
+        race-high_accuracy: 40.62
+    gemma-7b-it-hf:
+        gsm8k_accuracy: 40.62
+        race-high_accuracy: 68.75
+    gemma-2-9b-it-turbomind:
+        gsm8k_accuracy: 68.75
+        race-high_accuracy: 84.38
+    gemma-2-27b-it-turbomind:
+        gsm8k_accuracy: 78.12
+        race-high_accuracy: 93.75
+    gemma-7b-it-vllm:
+        gsm8k_accuracy: 28.12
+        race-high_accuracy: 68.75
+    internlm2_5-7b-chat-hf:
+        gsm8k_accuracy: 84.38
+        race-high_accuracy: 90.62
+    internlm3-8b-instruct-hf:
+        gsm8k_accuracy: 65.62
+        race-high_accuracy: 87.5
+    internlm2_5-7b-chat-turbomind:
+        gsm8k_accuracy: 81.25
+        race-high_accuracy: 90.62
+    internlm2-chat-1.8b-turbomind:
+        gsm8k_accuracy: 25.00
+        race-high_accuracy: 84.38
+    internlm2-chat-1.8b-sft-turbomind:
+        gsm8k_accuracy: 34.38
+        race-high_accuracy: 84.38
+    internlm2-chat-7b-lmdeploy:
+        gsm8k_accuracy: 59.38
+        race-high_accuracy: 87.50
+    internlm2-chat-7b-sft-turbomind:
+        gsm8k_accuracy: 56.25
+        race-high_accuracy: 87.50
+    internlm3-8b-instruct-turbomind:
+        gsm8k_accuracy: 65.62
+        race-high_accuracy: 87.5
+    internlm2-chat-7b-vllm:
+        gsm8k_accuracy: 53.12
+        race-high_accuracy: 87.50
+    llama-3_1-8b-instruct-hf:
+        gsm8k_accuracy: 84.38
+        race-high_accuracy: 90.62
+    llama-3_2-3b-instruct-hf:
+        gsm8k_accuracy: 71.88
+        race-high_accuracy: 81.25
+    llama-3-8b-instruct-hf:
+        gsm8k_accuracy: 68.75
+        race-high_accuracy: 87.5
+    llama-2-7b-chat-turbomind:
+        gsm8k_accuracy: 18.75
+        race-high_accuracy: 46.88
+    llama-3_1-8b-instruct-turbomind:
+        gsm8k_accuracy: 84.38
+        race-high_accuracy: 90.62
+    llama-3_2-3b-instruct-turbomind:
+        gsm8k_accuracy: 65.62
+        race-high_accuracy: 81.25
+    llama-3-8b-instruct-turbomind:
+        gsm8k_accuracy: 65.62
+        race-high_accuracy: 84.38
+    mistral-7b-instruct-v0.2-hf:
+        gsm8k_accuracy: 40.62
+        race-high_accuracy: 75
+    mistral-7b-instruct-v0.3-hf:
+        gsm8k_accuracy: 40.62
+        race-high_accuracy: 75
+    mistral-nemo-instruct-2407-hf:
+        gsm8k_accuracy: 75
+        race-high_accuracy: 81.25
+    mistral-nemo-instruct-2407-turbomind:
+        gsm8k_accuracy: 71.88
+        race-high_accuracy: 75
+    mistral-7b-instruct-v0.1-vllm:
+        gsm8k_accuracy: 34.38
+        race-high_accuracy: 65.62
+    mistral-7b-instruct-v0.2-vllm:
+        gsm8k_accuracy: 28.12
+        race-high_accuracy: 78.12
+    qwen2.5-0.5b-instruct-hf:
+        gsm8k_accuracy: 34.38
+        race-high_accuracy: 46.88
+    qwen2.5-3b-instruct-hf :
+        gsm8k_accuracy: 53.12
+        race-high_accuracy: 90.62
+    qwen2.5-0.5b-instruct-turbomind:
+        gsm8k_accuracy: 28.12
+        race-high_accuracy: 43.75
+    qwen2.5-3b-instruct-turbomind:
+        gsm8k_accuracy: 56.25
+        race-high_accuracy: 90.62
+    qwen1.5-0.5b-chat-hf:
+        gsm8k_accuracy: 0
+        race-high_accuracy: 53.12
+    qwen2-1.5b-instruct-hf:
+        gsm8k_accuracy: 62.5
+        race-high_accuracy: 84.38
+    qwen2-7b-instruct-hf:
+        gsm8k_accuracy: 68.75
+        race-high_accuracy: 90.62
+    qwen2-1.5b-instruct-turbomind:
+        gsm8k_accuracy: 56.25
+        race-high_accuracy: 84.38
+    qwen2-7b-instruct-turbomind:
+        gsm8k_accuracy: 75.00
+        race-high_accuracy: 87.50
+    qwen1.5-0.5b-chat-vllm:
+        gsm8k_accuracy: 6.25
+        race-high_accuracy: 53.12
+    yi-1.5-6b-chat-hf:
+        gsm8k_accuracy: 65.62
+        race-high_accuracy: 84.38
+    yi-1.5-9b-chat-hf:
+        gsm8k_accuracy: 75
+        race-high_accuracy: 93.75
+    yi-1.5-6b-chat-turbomind:
+        gsm8k_accuracy: 59.38
+        race-high_accuracy: 84.38
+    yi-1.5-9b-chat-turbomind:
+        gsm8k_accuracy: 78.12
+        race-high_accuracy: 93.75
+    deepseek-v2_lite-chat-turbomind:
+        gsm8k_accuracy: 43.75
+        race-high_accuracy: 71.88
+    gemma2-27b-it-hf:
+        gsm8k_accuracy: 71.88
+        race-high_accuracy: 93.75
+    internlm2_5-20b-chat-hf:
+        gsm8k_accuracy: 84.38
+        race-high_accuracy: 87.5
+    internlm2_5-20b-chat-turbomind:
+        gsm8k_accuracy: 87.50
+        race-high_accuracy: 87.5
+    mistral-small-instruct-2409-hf:
+        gsm8k_accuracy: 81.25
+        race-high_accuracy: 87.50
+    mistral-small-instruct-2409-turbomind:
+        gsm8k_accuracy: 78.12
+        race-high_accuracy: 87.50
+    phi-4:
+        gsm8k_accuracy: 81.25
+        race-high_accuracy: 87.50
+    qwen2.5-14b-instruct-hf:
+        gsm8k_accuracy: 71.88
+        race-high_accuracy: 96.88
+    qwen2.5-14b-instruct-turbomind:
+        gsm8k_accuracy: 71.88
+        race-high_accuracy: 96.88
+    yi-1.5-34b-chat-turbomind:
+        gsm8k_accuracy: 71.88
+        race-high_accuracy: 93.75
+    deepseek-67b-chat-turbomind:
+        gsm8k_accuracy: 71.88
+        race-high_accuracy: 75.00
+    deepseek-r1-distill-qwen-32b-turbomind:
+        gsm8k_accuracy: 31.25
+        race-high_accuracy: 90.62
+    llama-3_3-70b-instruct-turbomind:
+        gsm8k_accuracy: 93.75
+        race-high_accuracy: 87.5
+    mixtral-large-instruct-2411-turbomind:
+        gsm8k_accuracy: 87.50
+        race-high_accuracy: 93.75
+    nvidia-3_1-Nemotron-70b-instruct-HF-turbomind:
+        gsm8k_accuracy: 90.62
+        race-high_accuracy: 53.12
+    qwen2.5-72b-instruct-turbomind:
+        gsm8k_accuracy: 78.12
+        race-high_accuracy: 90.62
+    deepseek-r1-distill-llama-70b-turbomind:
+        gsm8k_accuracy: 50.00
+        race-high_accuracy: 87.50
+    deepseek-v2_5-1210-turbomind:
+        gsm8k_accuracy: 90.62
+        race-high_accuracy: 84.38
+    mixtral-8x22b-instruct-v0.1-turbomind:
+        gsm8k_accuracy: 75.00
+        race-high_accuracy: 78.12
+    mixtral-8x22b-instruct-v0.1-vllm:
+        gsm8k_accuracy: 78.12
+        race-high_accuracy: 78.12
+base:
+    glm-4-9b-turbomind:
+        gsm8k_accuracy: 59.38
+        GPQA_diamond_accuracy: 28.12
+        race-high_accuracy: 93.75
+        winogrande_accuracy: 84.38
+    deepseek-7b-base-hf:
+        gsm8k_accuracy: 25
+        GPQA_diamond_accuracy: 0
+        race-high_accuracy: 46.88
+        winogrande_accuracy: 71.88
+    deepseek-7b-base-turbomind:
+        gsm8k_accuracy: 18.75
+        GPQA_diamond_accuracy: 3.12
+        race-high_accuracy: 50.00
+        winogrande_accuracy: 84.38
+    deepseek-moe-16b-base-vllm:
+        gsm8k_accuracy: 25.00
+        GPQA_diamond_accuracy: 0
+        race-high_accuracy: 25
+        winogrande_accuracy: 68.75
+    gemma2-2b-hf:
+        gsm8k_accuracy: 31.25
+        GPQA_diamond_accuracy: 3.12
+        race-high_accuracy: 56.25
+        winogrande_accuracy: 75.00
+    gemma2-9b-hf:
+        gsm8k_accuracy: 75.00
+        GPQA_diamond_accuracy: 0
+        race-high_accuracy: 84.38
+        winogrande_accuracy: 81.25
+    gemma-2b-hf:
+        gsm8k_accuracy: 21.88
+        GPQA_diamond_accuracy: 3.12
+        race-high_accuracy: 21.88
+        winogrande_accuracy: 53.12
+    gemma-7b-hf:
+        gsm8k_accuracy: 56.25
+        GPQA_diamond_accuracy: 3.12
+        race-high_accuracy: 65.62
+        winogrande_accuracy: 71.88
+    gemma-2-9b-turbomind:
+        gsm8k_accuracy: 68.75
+        GPQA_diamond_accuracy: 0
+        race-high_accuracy: 84.38
+        winogrande_accuracy: 81.25
+    gemma-2b-vllm:
+        gsm8k_accuracy: 15.62
+        GPQA_diamond_accuracy: 3.12
+        race-high_accuracy: 28.12
+        winogrande_accuracy: 68.75
+    gemma-7b-vllm:
+        gsm8k_accuracy: 59.38
+        GPQA_diamond_accuracy: 6.25
+        race-high_accuracy: 81.25
+        winogrande_accuracy: 81.25
+    internlm2_5-7b-hf:
+        gsm8k_accuracy: 37.5
+        GPQA_diamond_accuracy: 25
+        race-high_accuracy: 93.75
+        winogrande_accuracy: 71.88
+    internlm2-7b-hf:
+        gsm8k_accuracy: 53.12
+        GPQA_diamond_accuracy: 18.75
+        race-high_accuracy: 62.5
+        winogrande_accuracy: 78.12
+    internlm2-1.8b-turbomind:
+        gsm8k_accuracy: 12.50
+        GPQA_diamond_accuracy: 9.38
+        race-high_accuracy: 71.88
+        winogrande_accuracy: 75
+    internlm2_5-7b-turbomind:
+        gsm8k_accuracy: 62.5
+        GPQA_diamond_accuracy: 31.25
+        race-high_accuracy: 93.75
+        winogrande_accuracy: 87.5
+    internlm2-7b-turbomind:
+        gsm8k_accuracy: 53.12
+        GPQA_diamond_accuracy: 25.00
+        race-high_accuracy: 78.12
+        winogrande_accuracy: 71.88
+    internlm2-base-7b-turbomind:
+        gsm8k_accuracy: 25.00
+        GPQA_diamond_accuracy: 34.38
+        race-high_accuracy: 71.88
+        winogrande_accuracy: 62.50
+    llama-2-7b-hf:
+        gsm8k_accuracy: 21.88
+        GPQA_diamond_accuracy: 21.88
+        race-high_accuracy: 40.62
+        winogrande_accuracy: 71.88
+    llama-3_1-8b-hf:
+        gsm8k_accuracy: 78.12
+        GPQA_diamond_accuracy: 25
+        race-high_accuracy: 90.62
+        winogrande_accuracy: 62.5
+    llama-3-8b-hf:
+        gsm8k_accuracy: 46.88
+        GPQA_diamond_accuracy: 6.25
+        race-high_accuracy: 65.62
+        winogrande_accuracy: 65.62
+    llama-3.1-8b-turbomind:
+        gsm8k_accuracy: 56.25
+        GPQA_diamond_accuracy: 9.38
+        race-high_accuracy: 78.12
+        winogrande_accuracy: 78.12
+    llama-3-8b-turbomind:
+        gsm8k_accuracy: 46.88
+        GPQA_diamond_accuracy: 12.50
+        race-high_accuracy: 65.62
+        winogrande_accuracy: 81.25
+    mistral-7b-v0.3-hf:
+        gsm8k_accuracy: 31.25
+        GPQA_diamond_accuracy: 6.25
+        race-high_accuracy: 62.5
+        winogrande_accuracy: 59.38
+    qwen2.5-7b-hf:
+        gsm8k_accuracy: 81.25
+        GPQA_diamond_accuracy: 18.75
+        race-high_accuracy: 87.5
+        winogrande_accuracy: 71.88
+    qwen2.5-1.5b-turbomind:
+        gsm8k_accuracy: 59.38
+        GPQA_diamond_accuracy: 21.88
+        race-high_accuracy: 78.12
+        winogrande_accuracy: 71.88
+    qwen2.5-7b-turbomind:
+        gsm8k_accuracy: 78.12
+        GPQA_diamond_accuracy: 21.88
+        race-high_accuracy: 87.5
+        winogrande_accuracy: 75.00
+    qwen1.5-moe-a2.7b-hf:
+        gsm8k_accuracy: 62.5
+        GPQA_diamond_accuracy: 18.75
+        race-high_accuracy: 84.38
+        winogrande_accuracy: 75
+    qwen2-0.5b-hf:
+        gsm8k_accuracy: 25
+        GPQA_diamond_accuracy: 0
+        race-high_accuracy: 40.62
+        winogrande_accuracy: 62.5
+    qwen2-1.5b-hf:
+        gsm8k_accuracy: 59.38
+        GPQA_diamond_accuracy: 9.38
+        race-high_accuracy: 81.25
+        winogrande_accuracy: 62.5
+    qwen2-7b-hf:
+        gsm8k_accuracy: 68.75
+        GPQA_diamond_accuracy: 9.38
+        race-high_accuracy: 87.5
+        winogrande_accuracy: 68.75
+    qwen2-1.5b-turbomind:
+        gsm8k_accuracy: 56.25
+        GPQA_diamond_accuracy: 12.50
+        race-high_accuracy: 81.25
+        winogrande_accuracy: 75
+    qwen2-7b-turbomind:
+        gsm8k_accuracy: 65.62
+        GPQA_diamond_accuracy: 12.5
+        race-high_accuracy: 87.5
+        winogrande_accuracy: 75
+    qwen1.5-0.5b-vllm:
+        gsm8k_accuracy: 9.38
+        GPQA_diamond_accuracy: 3.12
+        race-high_accuracy: 56.25
+        winogrande_accuracy: 59.38
+    yi-1.5-6b-hf:
+        gsm8k_accuracy: 62.5
+        GPQA_diamond_accuracy: 3.12
+        race-high_accuracy: 87.5
+        winogrande_accuracy: 62.5
+    yi-1.5-9b-hf:
+        gsm8k_accuracy: 75
+        GPQA_diamond_accuracy: 40.62
+        race-high_accuracy: 87.5
+        winogrande_accuracy: 59.38
+    yi-1.5-9b-turbomind:
+        gsm8k_accuracy: 75.00
+        GPQA_diamond_accuracy: 40.62
+        race-high_accuracy: 87.5
+        winogrande_accuracy: 65.62
+    internlm2-20b-turbomind:
+        gsm8k_accuracy: 71.88
+        GPQA_diamond_accuracy: 18.75
+        race-high_accuracy: 68.75
+        winogrande_accuracy: 81.25
+    qwen2.5-14b-hf:
+        gsm8k_accuracy: 75
+        GPQA_diamond_accuracy: 37.5
+        race-high_accuracy: 93.75
+        winogrande_accuracy: 84.38
+    qwen2.5-32b-hf:
+        gsm8k_accuracy: 87.5
+        GPQA_diamond_accuracy: 31.25
+        race-high_accuracy: 93.75
+        winogrande_accuracy: 78.12
+    qwen2.5-32b-turbomind:
+        gsm8k_accuracy: 90.62
+        GPQA_diamond_accuracy: 31.25
+        race-high_accuracy: 93.75
+        winogrande_accuracy: 81.25
+    deepseek-67b-base-turbomind:
+        gsm8k_accuracy: 62.50
+        GPQA_diamond_accuracy: 31.25
+        race-high_accuracy: 78.12
+        winogrande_accuracy: 81.25
+    llama-3-70b-turbomind:
+        gsm8k_accuracy: 56.25
+        GPQA_diamond_accuracy: 15.62
+        race-high_accuracy: 93.75
+        winogrande_accuracy: 84.38
+    qwen2.5-72b-turbomind:
+        gsm8k_accuracy: 84.38
+        GPQA_diamond_accuracy: 40.62
+        race-high_accuracy: 93.75
+        winogrande_accuracy: 87.5
+    deepseek-v2-turbomind:
+        gsm8k_accuracy: 65.62
+        GPQA_diamond_accuracy: 3.12
+        race-high_accuracy: 93.75
+        winogrande_accuracy: 81.25

.github/workflows/daily-run-test.yml

@@ -2,74 +2,347 @@ name: daily_run_test
 
 on:
   workflow_dispatch:
+    inputs:
+      repo_org:
+        required: false
+        description: 'Tested repository organization name. Default is open-compass/opencompass'
+        type: string
+        default: 'open-compass/opencompass'
+      repo_ref:
+        required: false
+        description: 'Set branch or tag or commit id. Default is "main"'
+        type: string
+        default: 'main'
+      build_lmdeploy:
+        required: false
+        description: 'whether to build lmdeploy'
+        type:  boolean
+        default: false
+      repo_org_lmdeploy:
+        required: false
+        description: 'Tested repository organization name. Default is internlm/lmdeploy'
+        type: string
+        default: 'InternLM/lmdeploy'
+      repo_ref_lmdeploy:
+        required: false
+        description: 'Set branch or tag or commit id. Default is "main"'
+        type: string
+        default: 'main'
+      regression_func_volc:
+        required: true
+        description: 'regression functions'
+        type: string
+        default: "['chat_models','base_models', 'chat_obj_fullbench', 'base_fullbench']"
+      regression_func_local:
+        required: true
+        description: 'regression functions'
+        type: string
+        default: "['cmd', 'api', 'chat_sub_fullbench']"
+      fullbench_eval:
+        required: true
+        description: 'fullbench volc functions'
+        type: string
+        default: "['base_objective','chat_objective','chat_subjective','base_long_context','chat_long_context']"
   schedule:
-    - cron:  '56 16 * * *'
-
-concurrency:
-  group: ${{ github.workflow }}-${{ github.ref }}
-  cancel-in-progress: true
+    - cron:  '15 14 * * 0,3'
 
 env:
-  CONDA_ENV: opencompass_regression
-  PIP_CACHE_PATH: /cpfs01/user/qa-llm-cicd/.cache/pip
-  USERSPACE_PREFIX: /cpfs01/user/qa-llm-cicd
-  HF_CACHE_PATH: /cpfs01/shared/public/public_hdd/llmeval/model_weights/hf_hub
   HF_DATASETS_OFFLINE: 1
+  HF_EVALUATE_OFFLINE: 1
   TRANSFORMERS_OFFLINE: 1
+  VLLM_USE_MODELSCOPE: false
+  LMDEPLOY_USE_MODELSCOPE: false
   HF_HUB_OFFLINE: 1
+  OUTPUT_FOLDER: cuda12.1_dist_${{ github.run_id }}
+  CONDA_PATH: ${{ secrets.WORKSPACE_PREFIX }}/miniconda3
+  PIP_CACHE_PATH: ${{ secrets.WORKSPACE_PREFIX }}/.cache/pip
+  REPORT_ROOT: ${{ secrets.WORKSPACE_PREFIX }}/eval_report/regression
+  COMPASS_DATA_CACHE: ${{ secrets.SHARESPACE_PREFIX }}/datasets/compass_data_cache
+  HUGGINGFACE_HUB_CACHE: ${{ secrets.SHARESPACE_PREFIX }}/models/opencompass_hf_hub
+  HF_HUB_CACHE:  ${{ secrets.SHARESPACE_PREFIX }}/models/opencompass_hf_hub
+  HF_DATASETS_CACHE:  ${{ secrets.SHARESPACE_PREFIX }}/datasets/hf_datasets_cache
+  HF_ENDPOINT: https://hf-mirror.com
+  CONDA_ENV: regression_test
+  export VLLM_WORKER_MULTIPROC_METHOD: spawn
 
 jobs:
-  daily_run_test:
-    runs-on: self-hosted
-    environment: 'prod'
-    timeout-minutes: 240 #4hours
+  build-pypi:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v2
+        with:
+          repository: ${{ github.event.inputs.repo_org || 'open-compass/opencompass' }}
+          ref: ${{github.event.inputs.repo_ref || 'main'}}
+      - name: Set up Python 3.10
+        uses: actions/setup-python@v4
+        with:
+          python-version: '3.10'
+      - name: Build lagent
+        run: |
+          pip install wheel setuptools
+          python setup.py sdist bdist_wheel
+      - name: Upload Artifacts
+        uses: actions/upload-artifact@v4
+        with:
+          if-no-files-found: error
+          path: dist/*
+          retention-days: 1
+          name: my-artifact-${{ github.run_id }}
+
+  build-pypi-lmdeploy:
+    if: ${{!cancelled() && (github.event_name == 'schedule' || inputs.build_lmdeploy)}}
+    strategy:
+      matrix:
+        pyver: [py310]
+    runs-on: ubuntu-latest
+    env:
+      PYTHON_VERSION: ${{ matrix.pyver }}
+      PLAT_NAME: manylinux2014_x86_64
+      DOCKER_TAG: cuda12.1
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@v3
+        with:
+          repository: ${{ github.event.inputs.repo_org_lmdeploy || 'InternLM/lmdeploy' }}
+          ref: ${{github.event.inputs.repo_ref_lmdeploy || 'main'}}
+      - name: Build
+        run: |
+          echo ${PYTHON_VERSION}
+          echo ${PLAT_NAME}
+          echo ${DOCKER_TAG}
+          echo ${OUTPUT_FOLDER}
+          echo ${GITHUB_RUN_ID}
+          # remove -it
+          sed -i 's/docker run --rm -it/docker run --rm/g' builder/manywheel/build_wheel.sh
+          bash builder/manywheel/build_wheel.sh ${PYTHON_VERSION} ${PLAT_NAME} ${DOCKER_TAG} ${OUTPUT_FOLDER}
+      - name: Upload Artifacts
+        uses: actions/upload-artifact@v4
+        with:
+          if-no-files-found: error
+          path: builder/manywheel/${{ env.OUTPUT_FOLDER }}
+          retention-days: 1
+          name: my-artifact-${{ github.run_id }}-${{ matrix.pyver }}
+
+
+  prepare_env:
+    if: ${{!cancelled()}}
+    needs: ['build-pypi', 'build-pypi-lmdeploy']
+    runs-on: volc_cu12
+    timeout-minutes: 120 #2hours
     steps:
       - name: Clone repository
         uses: actions/checkout@v2
-      - name: Prepare - create conda env and install torch
-        run: |
-          eval "$(conda shell.bash hook)"
-          conda create -y --name ${{env.CONDA_ENV}} python=3.10
-          conda activate ${{env.CONDA_ENV}}
-          pip install torch torchvision torchaudio --cache-dir ${{env.PIP_CACHE_PATH}} --index-url https://download.pytorch.org/whl/cu118
-          conda info --envs
-      - name: Prepare - Pip install code
-        run: |
-          eval "$(conda shell.bash hook)"
-          conda activate ${{env.CONDA_ENV}}
-          pip install -e . --cache-dir ${{env.PIP_CACHE_PATH}}
-          pip install human_eval transformers==4.33.0 protobuf --cache-dir ${{env.PIP_CACHE_PATH}}
-          conda info --envs
-      - name: Prepare - prepare data and hf model
-        run: |
-          cp -r ${{env.USERSPACE_PREFIX}}/data .
-          rm -rf ~/.cache/huggingface/hub -f && mkdir ~/.cache -p && mkdir ~/.cache/huggingface -p
-          ln -s ${{env.HF_CACHE_PATH}} ~/.cache/huggingface/hub
-      - name:  Run test
-        run: |
-          eval "$(conda shell.bash hook)"
-          conda activate ${{env.CONDA_ENV}}
-          conda info --envs
-          rm -rf regression_result_daily
-          export from_tf=TRUE
-          python3 run.py --models hf_internlm_chat_7b hf_internlm2_7b hf_chatglm3_6b_base hf_chatglm3_6b hf_qwen_7b_chat hf_qwen_7b --datasets FewCLUE_chid_ppl humaneval_gen ARC_c_ppl obqa_ppl --work-dir regression_result_daily
-      - name:  Get result
-        run: |
-          eval "$(conda shell.bash hook)"
-          pip install pytest --cache-dir ${{env.PIP_CACHE_PATH}}
-          python -m pytest -s -v --color=yes .github/scripts/oc_score_assert.py
+        with:
+          repository: ${{ github.event.inputs.repo_org || 'open-compass/opencompass' }}
+          ref: ${{github.event.inputs.repo_ref || 'main'}}
+      - name: Download Artifacts
+        uses: actions/download-artifact@v4
+        with:
+          name: my-artifact-${{ github.run_id }}
       - name:  Remove Conda Env
         if: always()
         run: |
-          cp -r regression_result_daily/* /cpfs01/user/qa-llm-cicd/report
-          eval "$(conda shell.bash hook)"
+          . ${{ secrets.WORKSPACE_PREFIX }}/miniconda3/bin/activate
           conda env remove -y --name ${{env.CONDA_ENV}}
           conda info --envs
+      - name: Prepare - create conda env and install torch - cu12
+        uses: nick-fields/retry@v3
+        with:
+          max_attempts: 3
+          timeout_minutes: 120
+          command: |
+            . ${{env.CONDA_PATH}}/bin/activate
+            conda create -y --name ${{env.CONDA_ENV}} python=3.10
+            conda activate ${{env.CONDA_ENV}}
+            pip install -r ${{ secrets.WORKSPACE_PREFIX }}/config/requirements.txt --cache-dir ${{env.PIP_CACHE_PATH}}
+            pip install opencompass*.whl --cache-dir ${{env.PIP_CACHE_PATH}}
+            pip install opencompass[lmdeploy] --cache-dir ${{env.PIP_CACHE_PATH}}
+            pip install opencompass[vllm] --cache-dir ${{env.PIP_CACHE_PATH}}
+            pip install opencompass[full] --cache-dir ${{env.PIP_CACHE_PATH}}
+            pip install opencompass[api] --cache-dir ${{env.PIP_CACHE_PATH}}
+            pip install torch==2.6.0 torchvision==0.21.0 torchaudio==2.6.0 --cache-dir ${{env.PIP_CACHE_PATH}}
+            FLASH_ATTENTION_FORCE_BUILD=TRUE pip install ${{ secrets.WORKSPACE_PREFIX }}/packages/flash_attn-2.7.0.post2+cu12torch2.5cxx11abiFALSE-cp310-cp310-linux_x86_64.whl
+            pip install xformers --index-url https://download.pytorch.org/whl/cu121 --cache-dir ${{env.PIP_CACHE_PATH}}
+            cp -r /root/nltk_data ${{env.CONDA_PATH}}/envs/${{env.CONDA_ENV}}/nltk_data
+      - name: Prepare - reinstall lmdeploy - cu12
+        if: ${{github.event_name == 'schedule' || inputs.build_lmdeploy}}
+        uses: actions/download-artifact@v4
+        with:
+          name: my-artifact-${{ github.run_id }}-py310
+      - name: Prepare - reinstall lmdeploy - cu12
+        if: ${{github.event_name == 'schedule' || inputs.build_lmdeploy}}
+        run: |
+          . ${{env.CONDA_PATH}}/bin/activate
+          conda activate ${{env.CONDA_ENV}}
+          pip uninstall -y lmdeploy
+          pip install lmdeploy-*.whl --no-deps
+      - name: conda env
+        run: |
+          . ${{env.CONDA_PATH}}/bin/activate
+          conda activate ${{env.CONDA_ENV}}
+          conda info --envs
+          pip list
+
+  daily_run_test_volc:
+    if: ${{!cancelled() && contains(needs.prepare_env.result, 'success')}}
+    needs: prepare_env
+    strategy:
+      fail-fast: false
+      matrix:
+        regression_func: ${{fromJSON(github.event.inputs.regression_func_volc || '["chat_models","base_models","chat_obj_fullbench","base_fullbench"]')}}
+    runs-on: volc_cu12_daily
+    timeout-minutes: 180 #3hours
+    steps:
+      - name: Clone repository
+        uses: actions/checkout@v2
+        with:
+          repository: ${{ github.event.inputs.repo_org || 'open-compass/opencompass' }}
+          ref: ${{github.event.inputs.repo_ref || 'main'}}
+      - name: conda env
+        run: |
+          . ${{env.CONDA_PATH}}/bin/activate
+          conda activate ${{env.CONDA_ENV}}
+          conda info --envs
+          pip list
+      - name:  modify config
+        if: matrix.regression_func != 'chat_sub_fullbench'
+        run: |
+          cp -r ${{ secrets.WORKSPACE_PREFIX }}/ocplayground/template/configs_cluster/volc.py .
+          cat ${{ secrets.WORKSPACE_PREFIX }}/config/test_config.txt >> .github/scripts/eval_regression_${{matrix.regression_func}}.py
+      - name:  Run test
+        uses: nick-fields/retry@v3
+        with:
+          max_attempts: 1
+          timeout_minutes: 180
+          command: |
+            . ${{env.CONDA_PATH}}/bin/activate
+            conda activate ${{env.CONDA_ENV}}
+            conda info --envs
+            opencompass .github/scripts/eval_regression_${{matrix.regression_func}}.py --work-dir ${{env.REPORT_ROOT}}/${{ github.run_id }}/${{matrix.regression_func}} --reuse --dump-eval-details
+            rm regression_result_daily -f && ln -s ${{env.REPORT_ROOT}}/${{ github.run_id }}/${{matrix.regression_func}}/*/summary regression_result_daily
+            python -m pytest -m ${{matrix.regression_func}} -s -v --color=yes .github/scripts/oc_score_assert.py
+
+
+  daily_run_test_local:
+    if: ${{!cancelled() && contains(needs.prepare_env.result, 'success')}}
+    needs: prepare_env
+    strategy:
+      fail-fast: false
+      matrix:
+        regression_func: ${{fromJSON(github.event.inputs.regression_func_local || '["cmd","api","chat_sub_fullbench"]')}}
+    runs-on: volc_cu12_local
+    timeout-minutes: 480 #6hours
+    steps:
+      - name: Clone repository
+        uses: actions/checkout@v2
+        with:
+          repository: ${{ github.event.inputs.repo_org || 'open-compass/opencompass' }}
+          ref: ${{github.event.inputs.repo_ref || 'main'}}
+      - name: conda env
+        run: |
+          . ${{env.CONDA_PATH}}/bin/activate
+          conda activate ${{env.CONDA_ENV}}
+          conda info --envs
+          pip list
+      - name:  modify config
+        if: matrix.regression_func == 'chat_sub_fullbench'
+        run: |
+          cp -r ${{ secrets.WORKSPACE_PREFIX }}/ocplayground/template/configs_cluster/volc.py .
+          cat ${{ secrets.WORKSPACE_PREFIX }}/config/test_config_sub.txt >> .github/scripts/eval_regression_${{matrix.regression_func}}.py
+      - name:  Run command testcase
+        if: matrix.regression_func == 'cmd'
+        run: |
+          . ${{env.CONDA_PATH}}/bin/activate
+          conda activate ${{env.CONDA_ENV}}
+          conda info --envs
+          export from_tf=TRUE
+          python tools/list_configs.py internlm2_5 mmlu
+          opencompass --models hf_internlm2_5_7b --datasets race_ppl demo_gsm8k_chat_gen --work-dir ${{env.REPORT_ROOT}}/${{ github.run_id }}/cmd1 --reuse --max-num-workers 2 --dump-eval-details
+          rm regression_result_daily -f && ln -s ${{env.REPORT_ROOT}}/${{ github.run_id }}/cmd1/*/summary regression_result_daily
+          python -m pytest -m case1 -s -v --color=yes .github/scripts/oc_score_assert.py
+          opencompass --models hf_internlm2_5_7b_chat hf_internlm3_8b_instruct --datasets race_gen demo_gsm8k_chat_gen -a lmdeploy --work-dir ${{env.REPORT_ROOT}}/${{ github.run_id }}/cmd2 --reuse --max-num-workers 2 --dump-eval-details
+          rm regression_result_daily -f && ln -s ${{env.REPORT_ROOT}}/${{ github.run_id }}/cmd2/*/summary regression_result_daily
+          python -m pytest -m case2 -s -v --color=yes .github/scripts/oc_score_assert.py
+          opencompass --datasets race_ppl demo_gsm8k_chat_gen --hf-type base --hf-path internlm/internlm2_5-7b --work-dir ${{env.REPORT_ROOT}}/${{ github.run_id }}/cmd3 --reuse --max-num-workers 2 --dump-eval-details
+          rm regression_result_daily -f && ln -s ${{env.REPORT_ROOT}}/${{ github.run_id }}/cmd3/*/summary regression_result_daily
+          python -m pytest -m case3 -s -v --color=yes .github/scripts/oc_score_assert.py
+          opencompass --datasets race_gen demo_gsm8k_chat_gen --hf-type chat --hf-path internlm/internlm3-8b-instruct -a lmdeploy --work-dir ${{env.REPORT_ROOT}}/${{ github.run_id }}/cmd4 --reuse --max-num-workers 2 --dump-eval-details
+          rm regression_result_daily -f && ln -s ${{env.REPORT_ROOT}}/${{ github.run_id }}/cmd4/*/summary regression_result_daily
+          python -m pytest -m case4 -s -v --color=yes .github/scripts/oc_score_assert.py
+          opencompass --datasets race_gen demo_gsm8k_chat_gen --hf-type chat --hf-path internlm/internlm3-8b-instruct -a vllm --work-dir ${{env.REPORT_ROOT}}/${{ github.run_id }}/cmd5 --reuse --max-num-workers 2 --dump-eval-details
+          rm regression_result_daily -f && ln -s ${{env.REPORT_ROOT}}/${{ github.run_id }}/cmd5/*/summary regression_result_daily
+          python -m pytest -m case5 -s -v --color=yes .github/scripts/oc_score_assert.py
+      - name:  Run model test - api
+        if: matrix.regression_func == 'api'
+        run: |
+          . ${{env.CONDA_PATH}}/bin/activate
+          conda activate ${{env.CONDA_ENV}}
+          conda info --envs
+          lmdeploy serve api_server internlm/internlm3-8b-instruct --max-batch-size 256 --model-name internlm3 > ${{env.REPORT_ROOT}}/${{ github.run_id }}/restful.log  2>&1  &
+          echo "restful_pid=$!" >> "$GITHUB_ENV"
+          sleep 180s
+          env | grep PROXY
+          env | grep proxy
+          unset HTTP_PROXY;unset HTTPS_PROXY;unset http_proxy;unset https_proxy;
+          opencompass .github/scripts/eval_regression_api.py --work-dir ${{env.REPORT_ROOT}}/${{ github.run_id }}/api --reuse --max-num-workers 2 --dump-eval-details
+          rm regression_result_daily -f && ln -s ${{env.REPORT_ROOT}}/${{ github.run_id }}/api/*/summary regression_result_daily
+          python -m pytest -m api -s -v --color=yes .github/scripts/oc_score_assert.py
+      - name:  Run model test - api kill
+        if: always() && matrix.regression_func == 'api'
+        run: |
+          kill -15 "$restful_pid"
+      - name:  Run testcase
+        if: matrix.regression_func == 'chat_sub_fullbench'
+        env:
+          COMPASS_DATA_CACHE: ${{ secrets.SHARESPACE_PREFIX }}/datasets/compass_data_cache_subset
+        run: |
+          . ${{env.CONDA_PATH}}/bin/activate
+          conda activate ${{env.CONDA_ENV}}
+          conda info --envs
+          export from_tf=TRUE
+          opencompass .github/scripts/eval_regression_${{matrix.regression_func}}.py --work-dir ${{env.REPORT_ROOT}}/${{ github.run_id }}/${{matrix.regression_func}} --reuse --dump-eval-details
+          rm regression_result_daily -f && ln -s ${{env.REPORT_ROOT}}/${{ github.run_id }}/${{matrix.regression_func}}/*/summary regression_result_daily
+          python -m pytest -m ${{matrix.regression_func}} -s -v --color=yes .github/scripts/oc_score_assert.py
+
+  fullbench_run_test:
+    if: ${{!cancelled() && contains(needs.prepare_env.result, 'success')}}
+    needs: prepare_env
+    strategy:
+      fail-fast: false
+      matrix:
+        function_type: ${{fromJSON(github.event.inputs.fullbench_eval || '["base_objective","chat_objective","chat_subjective","base_long_context","chat_long_context"]')}}
+    runs-on: volc_cu12
+    timeout-minutes: 480 #6hours
+    steps:
+      - name: Clone repository
+        uses: actions/checkout@v2
+        with:
+          repository: ${{ github.event.inputs.repo_org || 'open-compass/opencompass' }}
+          ref: ${{github.event.inputs.repo_ref || 'main'}}
+      - name: conda env
+        run: |
+          . ${{env.CONDA_PATH}}/bin/activate
+          conda activate ${{env.CONDA_ENV}}
+          conda info --envs
+          pip list
+      - name:  Run testcase
+        uses: nick-fields/retry@v3
+        with:
+          max_attempts: 1
+          timeout_minutes: 480
+          command: |
+            . ${{env.CONDA_PATH}}/bin/activate
+            conda activate ${{env.CONDA_ENV}}
+            conda info --envs
+            export from_tf=TRUE
+            opencompass ${{ secrets.WORKSPACE_PREFIX }}/ocplayground/template/regression/eval_${{ matrix.function_type }}.py --work-dir ${{env.REPORT_ROOT}}/${{ github.run_id }}/${{ matrix.function_type }} --reuse
+            rm regression_result_daily -f && ln -s ${{env.REPORT_ROOT}}/${{ github.run_id }}/${{ matrix.function_type }}/*/summary regression_result_daily
+            python -m pytest -m ${{ matrix.function_type }} -s -v --color=yes .github/scripts/oc_score_assert.py
+
 
   notify_to_feishu:
-    if: ${{ always() && !cancelled() && contains(needs.*.result, 'failure') && (github.ref_name == 'develop' || github.ref_name == 'main') }}
-    needs: [daily_run_test]
-    environment: 'prod'
+    if: ${{ always() && github.event_name == 'schedule' && !cancelled() && contains(needs.*.result, 'failure') && (github.ref_name == 'develop' || github.ref_name == 'main') }}
+    needs: [daily_run_test_volc, daily_run_test_local, fullbench_run_test]
     timeout-minutes: 5
     runs-on: self-hosted
     steps:

.github/workflows/link-check.yml

@@ -5,17 +5,22 @@ on:
     # check links at 01:30 a.m. every day
     - cron: '30 1 * * *'
 
+  workflow_dispatch: # allow manual trigger
+
 jobs:
   link-check:
     runs-on: ubuntu-latest
     steps:
       # - uses: actions/checkout@v3
 
-      - name: linkchecker
+      - name: Install linkchecker
         run: |
           pip install linkchecker
-          linkchecker https://opencompass.readthedocs.io/ --no-robots -t 30 --no-warnings |
-            --ignore-url https://opencompass\.readthedocs\.io/.*/static/images/opencompass_logo\.svg |
-            --ignore-url https://opencompass\.readthedocs\.io/.*/_static/images/icon-menu-dots\.svg |
-            --ignore-url https://opencompass\.readthedocs\.io/policy |
-            --ignore-url https://opencompass\.readthedocs\.io/(en|zh_CN)/[0-9a-f]{40}/.*
+
+      - name: Run linkchecker
+        run: |
+          linkchecker https://opencompass.readthedocs.io/ --no-robots -t 30 --no-warnings \
+            --ignore-url "https://opencompass.readthedocs.io/.*/static/images/opencompass_logo.svg" \
+            --ignore-url "https://opencompass.readthedocs.io/.*/_static/images/icon-menu-dots.svg" \
+            --ignore-url "https://opencompass.readthedocs.io/policy" \
+            --ignore-url "https://opencompass.readthedocs.io/(en|zh_CN)/[0-9a-f]{40}/.*"

.github/workflows/lint.yml

@@ -17,7 +17,7 @@ jobs:
           python-version: '3.10'
       - name: Install pre-commit hook
         run: |
-          pip install pre-commit mmengine
+          pip install pre-commit==3.8.0 mmengine==0.10.5
           pre-commit install
       - name: Linting
         run: pre-commit run --all-files

.github/workflows/pr-run-test.yml

@@ -8,64 +8,88 @@ on:
       - 'docs/**'
       - 'configs/**'
       - 'tools/**'
-
   workflow_dispatch:
   schedule:
     - cron:  '56 22 * * *'
 
+
 concurrency:
   group: ${{ github.workflow }}-${{ github.ref }}
   cancel-in-progress: true
 
 env:
-  CONDA_ENV: opencompass_base
-  USERSPACE_PREFIX: /cpfs01/user/qa-llm-cicd
-  HF_CACHE_PATH: /cpfs01/shared/public/public_hdd/llmeval/model_weights/hf_hub
+  CONDA_ENV: pr_test
   HF_DATASETS_OFFLINE: 1
+  HF_EVALUATE_OFFLINE: 1
   TRANSFORMERS_OFFLINE: 1
+  VLLM_USE_MODELSCOPE: false
+  LMDEPLOY_USE_MODELSCOPE: false
   HF_HUB_OFFLINE: 1
+  CONDA_PATH: /fs-computility/llm/qa-llm-cicd/miniconda3
+  PIP_CACHE_PATH: /fs-computility/llm/qa-llm-cicd/.cache/pip
+  REPORT_ROOT: /fs-computility/llm/qa-llm-cicd/eval_report/prtest
+  COMPASS_DATA_CACHE: /fs-computility/llm/shared/llmeval/datasets/compass_data_cache
+  HUGGINGFACE_HUB_CACHE: /fs-computility/llm/shared/llmeval/models/opencompass_hf_hub
+  HF_HUB_CACHE: /fs-computility/llm/shared/llmeval/models/opencompass_hf_hub
 
 jobs:
   pr_run_test:
-    runs-on: self-hosted
+    runs-on: volc_cu12_local
     environment: 'prod'
     timeout-minutes: 30
     steps:
-      - name: Clone repository
+      - name: Checkout repository
         uses: actions/checkout@v2
       - name: Prepare - Install opencompass
         run: |
-          eval "$(conda shell.bash hook)"
+          . ${{env.CONDA_PATH}}/bin/activate
           conda activate ${{env.CONDA_ENV}}
           python3 -m pip uninstall opencompass -y
-          python3 -m pip install -e . --cache-dir ${{env.USERSPACE_PREFIX}}/.cache/pip
+          python3 -m pip install -e ".[full]" --cache-dir ${{env.PIP_CACHE_PATH}}
           conda info --envs
-      - name: Prepare - prepare data and hf model
+      - name: conda env
         run: |
-          cp -r ${{env.USERSPACE_PREFIX}}/data .
-          rm -rf ~/.cache/huggingface/hub -f && mkdir ~/.cache -p && mkdir ~/.cache/huggingface -p
-          ln -s ${{env.HF_CACHE_PATH}} ~/.cache/huggingface/hub
+          . ${{env.CONDA_PATH}}/bin/activate
+          conda activate ${{env.CONDA_ENV}}
+          conda info --envs
+          pip list
+          lmdeploy check_env
       - name:  Run test
         run: |
-          eval "$(conda shell.bash hook)"
+          . ${{env.CONDA_PATH}}/bin/activate
           conda activate ${{env.CONDA_ENV}}
           conda info --envs
           rm -rf regression_result
-          python3 run.py --models hf_internlm2_chat_7b --datasets siqa_gen --work-dir regression_result --debug
+          opencompass --models hf_internlm2_5_20b_chat --datasets demo_gsm8k_chat_gen --work-dir ${{env.REPORT_ROOT}}/${{ github.run_id }}/regression_result1 --debug
+          opencompass --models hf_internlm2_5_7b_chat --datasets demo_gsm8k_chat_gen --work-dir ${{env.REPORT_ROOT}}/${{ github.run_id }}/regression_result2 --debug --max-num-workers 2
+          opencompass --models hf_internlm2_5_7b_chat --datasets demo_gsm8k_chat_gen -a lmdeploy --work-dir ${{env.REPORT_ROOT}}/${{ github.run_id }}/regression_result3 --debug --max-num-workers 2
       - name:  Get result
         run: |
-          score=$(sed -n '$p' regression_result/*/summary/*.csv | awk -F ',' '{print $NF}')
-          if (( ${score%.*} >= 79 && ${score%.*} <= 81 )); then
-             echo "score is $score between 79 and 81"
+          score=$(sed -n '$p' ${{env.REPORT_ROOT}}/${{ github.run_id }}/regression_result1/*/summary/*.csv | awk -F ',' '{print $NF}')
+          if (( ${score%.*} >= 88 && ${score%.*} <= 89 )); then
+             echo "score is $score between 88 and 89"
           else
-             echo "score is $score not between 79 and 81"
+             echo "score is $score not between 88 and 89"
+             exit 1
+          fi
+          score=$(sed -n '$p' ${{env.REPORT_ROOT}}/${{ github.run_id }}/regression_result2/*/summary/*.csv | awk -F ',' '{print $NF}')
+          if (( ${score%.*} >= 87 && ${score%.*} <= 88 )); then
+             echo "score is $score between 87 and 88"
+          else
+             echo "score is $score not between 87 and 88"
+             exit 1
+          fi
+          score=$(sed -n '$p' ${{env.REPORT_ROOT}}/${{ github.run_id }}/regression_result3/*/summary/*.csv | awk -F ',' '{print $NF}')
+          if (( ${score%.*} >= 87 && ${score%.*} <= 91 )); then
+             echo "score is $score between 87 and 91"
+          else
+             echo "score is $score not between 87 and 91"
              exit 1
           fi
-          rm -rf regression_result
       - name:  Uninstall opencompass
         if: always()
         run: |
-          eval "$(conda shell.bash hook)"
+          . ${{env.CONDA_PATH}}/bin/activate
           conda activate ${{env.CONDA_ENV}}
           python3 -m pip uninstall opencompass -y
           conda info --envs
@@ -73,9 +97,9 @@ jobs:
   notify_to_feishu:
     if: ${{ always() && !cancelled() && contains(needs.*.result, 'failure') && (github.ref_name == 'develop' || github.ref_name == 'main') }}
     needs: [pr_run_test]
-    environment: 'prod'
     timeout-minutes: 5
     runs-on: self-hosted
+    environment: 'prod'
     steps:
       - name: notify
         run: |

.github/workflows/pr-stage-check.yml

@@ -20,7 +20,7 @@ jobs:
       matrix:
         python-version: ['3.10']
         include:
-          - torch: 2.0.0
+          - torch: 2.5.1
     steps:
       - uses: actions/checkout@v3
       - name: Set up Python ${{ matrix.python-version }}
@@ -30,7 +30,7 @@ jobs:
       - name: Upgrade pip
         run: python -m pip install --upgrade pip
       - name: Install PyTorch
-        run: pip install torch==${{matrix.torch}}+cpu -f https://download.pytorch.org/whl/cpu/torch_stable.html
+        run: pip install torch==${{matrix.torch}} -f https://download.pytorch.org/whl/cpu/torch_stable.html
       - name: Install system dependencies
         run: |
           sudo sed -i '$ a deb http://th.archive.ubuntu.com/ubuntu jammy main' /etc/apt/sources.list
@@ -106,7 +106,7 @@ jobs:
       - name: Upgrade pip
         run: python -m pip install pip --upgrade
       - name: Install PyTorch
-        run: pip install torch==2.0.0+${{matrix.platform}} -f https://download.pytorch.org/whl/${{matrix.platform}}/torch_stable.html
+        run: pip install torch==2.5.1 -f https://download.pytorch.org/whl/cpu/torch_stable.html
       - name: Install opencompass dependencies
         run: |
           pip install -r requirements.txt

.github/workflows/publish-to-pypi.yml

@@ -1,21 +1,26 @@
 name: deploy
 
-on: push
-
-concurrency:
-  group: ${{ github.workflow }}-${{ github.ref }}
-  cancel-in-progress: true
+on:
+  push:
+  workflow_dispatch:
+    inputs:
+      confirm_publish:
+        description: 'Type YES to confirm publishing to PyPI'
+        required: true
+        type: string
 
 jobs:
   build-n-publish:
     runs-on: ubuntu-latest
-    if: startsWith(github.event.ref, 'refs/tags')
+    if: |
+      github.event_name == 'push' && startsWith(github.event.ref, 'refs/tags') ||
+      (github.event_name == 'workflow_dispatch' && inputs.confirm_publish == 'YES')
     steps:
       - uses: actions/checkout@v2
-      - name: Set up Python 3.7
-        uses: actions/setup-python@v1
+      - name: Set up Python 3.10
+        uses: actions/setup-python@v4
         with:
-          python-version: 3.7
+          python-version: '3.10'
       - name: Build lagent
         run: |
           pip install wheel

.gitignore

@@ -1,4 +1,4 @@
-
+.DS_Store
 output_*/
 outputs/
 scripts/
@@ -102,6 +102,7 @@ configs/sft_cfg/60B/*
 configs/sft_cfg/100B/*
 
 configs/cky/
+configs/_internal_legacy*
 # in case llama clone in the opencompass
 llama/
 

.owners.yml

@@ -7,8 +7,8 @@ assign:
   scedule:
     '*/1 * * * *'
   assignees:
-    - Leymore
     - bittersweet1999
     - liushz
-    - kennymckormick
+    - MaiziXiao
+    - acylam
     - tonysy

.pre-commit-config-zh-cn.yaml

@@ -1,6 +1,7 @@
 exclude: |
     (?x)^(
       tests/data/|
+      tests/dataset/|
       opencompass/models/internal/|
       opencompass/utils/internal/|
       opencompass/openicl/icl_evaluator/hf_metrics/|
@@ -10,24 +11,40 @@ exclude: |
       opencompass/datasets/teval/|
       opencompass/datasets/NPHardEval/|
       opencompass/datasets/TheoremQA|
-      docs/zh_cn/advanced_guides/compassbench_intro.md
+      opencompass/datasets/subjective/mtbench101.py|
+      docs/zh_cn/advanced_guides/compassbench_intro.md |
+      docs/zh_cn/advanced_guides/compassbench_v2_0.md |
+      opencompass/utils/datasets.py |
+      opencompass/utils/datasets_info.py
     )
 repos:
   - repo: https://gitee.com/openmmlab/mirrors-flake8
     rev: 5.0.4
     hooks:
       - id: flake8
-        exclude: configs/
+        exclude: |
+            (?x)^(
+                opencompass/configs/|
+                examples/
+            )
   - repo: https://gitee.com/openmmlab/mirrors-isort
     rev: 5.11.5
     hooks:
       - id: isort
-        exclude: configs/
+        exclude: |
+            (?x)^(
+                opencompass/configs/|
+                examples/
+            )
   - repo: https://gitee.com/openmmlab/mirrors-yapf
     rev: v0.32.0
     hooks:
       - id: yapf
-        exclude: configs/
+        exclude: |
+            (?x)^(
+                opencompass/configs/|
+                examples/
+            )
   - repo: https://gitee.com/openmmlab/mirrors-codespell
     rev: v2.2.1
     hooks:
@@ -35,7 +52,9 @@ repos:
         exclude: |
             (?x)^(
                 .*\.jsonl|
-                configs/
+                .*\.md.template|
+                opencompass/configs/ |
+                examples/
             )
   - repo: https://gitee.com/openmmlab/mirrors-pre-commit-hooks
     rev: v4.3.0
@@ -45,7 +64,6 @@ repos:
             (?x)^(
               dicts/|
               projects/.*?/dicts/|
-              configs/.*?/.*\.txt
             )
       - id: check-yaml
       - id: end-of-file-fixer
@@ -53,7 +71,6 @@ repos:
             (?x)^(
               dicts/|
               projects/.*?/dicts/|
-              configs/.*?/.*\.txt
             )
       - id: requirements-txt-fixer
       - id: double-quote-string-fixer
@@ -85,7 +102,25 @@ repos:
         language: script
         pass_filenames: true
         require_serial: true
-        files: ^configs/datasets
+        files: ^opencompass/configs/datasets
+  - repo: local
+    hooks:
+    -   id: update-dataset-suffix-pacakge
+        name: dataset suffix updater(package)
+        entry: ./tools/update_dataset_suffix.py
+        language: script
+        pass_filenames: false
+        # require_serial: true
+        # files: ^opencompass/configs/datasets
+        args:
+          - --root_folder
+          - opencompass/configs/datasets
+  - repo: https://gitee.com/mirrors/gitleaks
+    rev: v8.23.1
+    hooks:
+    -   id: gitleaks
+        entry: "gitleaks dir"
+        args: ["--verbose", "--redact=50"]
   # - repo: https://github.com/open-mmlab/pre-commit-hooks
   #   rev: v0.2.0  # Use the ref you want to point at
   #   hooks:

.pre-commit-config.yaml

@@ -1,33 +1,51 @@
 exclude: |
     (?x)^(
       tests/data/|
+      tests/dataset/|
       opencompass/models/internal/|
       opencompass/utils/internal/|
       opencompass/openicl/icl_evaluator/hf_metrics/|
       opencompass/datasets/lawbench/utils|
       opencompass/datasets/lawbench/evaluation_functions/|
       opencompass/datasets/medbench/|
+      opencompass/datasets/matbench/|
       opencompass/datasets/teval/|
       opencompass/datasets/NPHardEval/|
       opencompass/datasets/TheoremQA|
-      docs/zh_cn/advanced_guides/compassbench_intro.md
+      opencompass/datasets/subjective/mtbench101.py|
+      docs/zh_cn/advanced_guides/compassbench_intro.md |
+      docs/zh_cn/advanced_guides/compassbench_v2_0.md |
+      opencompass/utils/datasets.py |
+      opencompass/utils/datasets_info.py
     )
 repos:
   - repo: https://github.com/PyCQA/flake8
     rev: 5.0.4
     hooks:
       - id: flake8
-        exclude: configs/
+        exclude: |
+            (?x)^(
+                opencompass/configs/|
+                examples/
+            )
   - repo: https://github.com/PyCQA/isort
     rev: 5.11.5
     hooks:
       - id: isort
-        exclude: configs/
+        exclude: |
+            (?x)^(
+                opencompass/configs/|
+                examples/
+            )
   - repo: https://github.com/pre-commit/mirrors-yapf
     rev: v0.32.0
     hooks:
       - id: yapf
-        exclude: configs/
+        exclude: |
+            (?x)^(
+                opencompass/configs/|
+                examples/
+            )
   - repo: https://github.com/codespell-project/codespell
     rev: v2.2.1
     hooks:
@@ -36,7 +54,8 @@ repos:
             (?x)^(
                 .*\.jsonl|
                 .*\.md.template|
-                configs/
+                opencompass/configs/ |
+                examples/
             )
   - repo: https://github.com/pre-commit/pre-commit-hooks
     rev: v4.3.0
@@ -46,7 +65,6 @@ repos:
             (?x)^(
               dicts/|
               projects/.*?/dicts/|
-              configs/.*?/.*\.txt
             )
       - id: check-yaml
       - id: end-of-file-fixer
@@ -54,7 +72,6 @@ repos:
             (?x)^(
               dicts/|
               projects/.*?/dicts/|
-              configs/.*?/.*\.txt
             )
       - id: requirements-txt-fixer
       - id: double-quote-string-fixer
@@ -86,7 +103,25 @@ repos:
         language: script
         pass_filenames: true
         require_serial: true
-        files: ^configs/datasets
+        files: ^opencompass/configs/datasets
+  - repo: local
+    hooks:
+    -   id: update-dataset-suffix-pacakge
+        name: dataset suffix updater(package)
+        entry: ./tools/update_dataset_suffix.py
+        language: script
+        pass_filenames: false
+        # require_serial: true
+        # files: ^opencompass/configs/datasets
+        args:
+          - --root_folder
+          - opencompass/configs/datasets
+  - repo: https://github.com/gitleaks/gitleaks
+    rev: v8.23.1
+    hooks:
+    -   id: gitleaks
+        entry: "gitleaks dir"
+        args: ["--verbose", "--redact=50"]
   # - repo: https://github.com/open-mmlab/pre-commit-hooks
   #   rev: v0.2.0  # Use the ref you want to point at
   #   hooks:

MANIFEST.in

@@ -0,0 +1,3 @@
+recursive-include opencompass/configs *.py *.yml *.json *.txt *.md
+recursive-include opencompass/openicl/icl_evaluator/hf_metrics *.py
+recursive-include opencompass/datasets *.py *.yml *.json *.txt *.md *.yaml

README.md

@@ -34,17 +34,6 @@ English | [简体中文](README_zh-CN.md)
 >
 > **Star Us**, You will receive all release notifications from GitHub without any delay ~ ⭐️
 
-## 📣 OpenCompass 2.0
-
-We are thrilled to introduce OpenCompass 2.0, an advanced suite featuring three key components: [CompassKit](https://github.com/open-compass), [CompassHub](https://hub.opencompass.org.cn/home), and [CompassRank](https://rank.opencompass.org.cn/home).
-![oc20](https://github.com/tonysy/opencompass/assets/7881589/90dbe1c0-c323-470a-991e-2b37ab5350b2)
-
-**CompassRank** has been significantly enhanced into the leaderboards that now incorporates both open-source benchmarks and proprietary benchmarks. This upgrade allows for a more comprehensive evaluation of models across the industry.
-
-**CompassHub** presents a pioneering benchmark browser interface, designed to simplify and expedite the exploration and utilization of an extensive array of benchmarks for researchers and practitioners alike. To enhance the visibility of your own benchmark within the community, we warmly invite you to contribute it to CompassHub. You may initiate the submission process by clicking [here](https://hub.opencompass.org.cn/dataset-submit).
-
-**CompassKit** is a powerful collection of evaluation toolkits specifically tailored for Large Language Models and Large Vision-language Models. It provides an extensive set of tools to assess and measure the performance of these complex models effectively. Welcome to try our toolkits for in your research and products.
-
 <details>
   <summary><kbd>Star History</kbd></summary>
   <picture>
@@ -64,23 +53,232 @@ Just like a compass guides us on our journey, OpenCompass will guide you through
 🔥🔥🔥 We are delighted to announce that **the OpenCompass has been recommended by the Meta AI**, click [Get Started](https://ai.meta.com/llama/get-started/#validation) of Llama for more information.
 
 > **Attention**<br />
-> We launch the OpenCompass Collaboration project, welcome to support diverse evaluation benchmarks into OpenCompass!
-> Clike [Issue](https://github.com/open-compass/opencompass/issues/248) for more information.
-> Let's work together to build a more powerful OpenCompass toolkit!
+> Breaking Change Notice: In version 0.4.0, we are consolidating all AMOTIC configuration files (previously located in ./configs/datasets, ./configs/models, and ./configs/summarizers) into the opencompass package. Users are advised to update their configuration references to reflect this structural change.
 
 ## 🚀 What's New <a><img width="35" height="20" src="https://user-images.githubusercontent.com/12782558/212848161-5e783dd6-11e8-4fe0-bbba-39ffb77730be.png"></a>
 
-- **\[2024.05.08\]** We supported the evaluation of 4 MoE models: [Mixtral-8x22B-v0.1](configs/models/mixtral/hf_mixtral_8x22b_v0_1.py), [Mixtral-8x22B-Instruct-v0.1](configs/models/mixtral/hf_mixtral_8x22b_instruct_v0_1.py), [Qwen1.5-MoE-A2.7B](configs/models/qwen/hf_qwen1_5_moe_a2_7b.py), [Qwen1.5-MoE-A2.7B-Chat](configs/models/qwen/hf_qwen1_5_moe_a2_7b_chat.py). Try them out now!
-- **\[2024.04.30\]** We supported evaluating a model's compression efficiency by calculating its Bits per Character (BPC) metric on an [external corpora](configs/datasets/llm_compression/README.md) ([official paper](https://github.com/hkust-nlp/llm-compression-intelligence)). Check out the [llm-compression](configs/eval_llm_compression.py) evaluation config now! 🔥🔥🔥
-- **\[2024.04.29\]** We report the performance of several famous LLMs on the common benchmarks, welcome to [documentation](https://opencompass.readthedocs.io/en/latest/user_guides/corebench.html) for more information! 🔥🔥🔥.
-- **\[2024.04.26\]** We deprecated the multi-madality evaluating function from OpenCompass, related implement has moved to [VLMEvalKit](https://github.com/open-compass/VLMEvalKit), welcome to use! 🔥🔥🔥.
-- **\[2024.04.26\]** We supported the evaluation of [ArenaHard](configs/eval_subjective_arena_hard.py)  welcome to try!🔥🔥🔥.
-- **\[2024.04.22\]** We supported the evaluation of [LLaMA3](configs/models/hf_llama/hf_llama3_8b.py) 和 [LLaMA3-Instruct](configs/models/hf_llama/hf_llama3_8b_instruct.py), welcome to try! 🔥🔥🔥
-- **\[2024.02.29\]** We supported the MT-Bench, AlpacalEval and AlignBench, more information can be found [here](https://opencompass.readthedocs.io/en/latest/advanced_guides/subjective_evaluation.html)
-- **\[2024.01.30\]** We release OpenCompass 2.0. Click  [CompassKit](https://github.com/open-compass), [CompassHub](https://hub.opencompass.org.cn/home), and [CompassRank](https://rank.opencompass.org.cn/home) for more information !
+- **\[2025.04.01\]** OpenCompass now supports `CascadeEvaluator`, a flexible evaluation mechanism that allows multiple evaluators to work in sequence. This enables creating customized evaluation pipelines for complex assessment scenarios. Check out the [documentation](docs/en/advanced_guides/llm_judge.md) for more details! 🔥🔥🔥
+- **\[2025.03.11\]** We have supported evaluation for `SuperGPQA` which is a great benchmark for measuring LLM knowledge ability 🔥🔥🔥
+- **\[2025.02.28\]** We have added a tutorial for `DeepSeek-R1` series model, please check [Evaluating Reasoning Model](docs/en/user_guides/deepseek_r1.md) for more details! 🔥🔥🔥
+- **\[2025.02.15\]** We have added two powerful evaluation tools: `GenericLLMEvaluator` for LLM-as-judge evaluations and `MATHVerifyEvaluator` for mathematical reasoning assessments. Check out the documentation for [LLM Judge](docs/en/advanced_guides/llm_judge.md) and [Math Evaluation](docs/en/advanced_guides/general_math.md) for more details! 🔥🔥🔥
+- **\[2025.01.16\]** We now support the [InternLM3-8B-Instruct](https://huggingface.co/internlm/internlm3-8b-instruct) model which has enhanced performance on reasoning and knowledge-intensive tasks.
+- **\[2024.12.17\]** We have provided the evaluation script for the December [CompassAcademic](examples/eval_academic_leaderboard_202412.py), which allows users to easily reproduce the official evaluation results by configuring it.
+- **\[2024.11.14\]** OpenCompass now offers support for a sophisticated benchmark designed to evaluate complex reasoning skills — [MuSR](https://arxiv.org/pdf/2310.16049). Check out the [demo](examples/eval_musr.py) and give it a spin! 🔥🔥🔥
+- **\[2024.11.14\]** OpenCompass now supports the brand new long-context language model evaluation benchmark — [BABILong](https://arxiv.org/pdf/2406.10149). Have a look at the [demo](examples/eval_babilong.py) and give it a try! 🔥🔥🔥
+- **\[2024.10.14\]** We now support the OpenAI multilingual QA dataset [MMMLU](https://huggingface.co/datasets/openai/MMMLU). Feel free to give it a try! 🔥🔥🔥
+- **\[2024.09.19\]** We now support [Qwen2.5](https://huggingface.co/Qwen)(0.5B to 72B) with multiple backend(huggingface/vllm/lmdeploy). Feel free to give them a try! 🔥🔥🔥
+- **\[2024.09.17\]** We now support OpenAI o1(`o1-mini-2024-09-12` and `o1-preview-2024-09-12`). Feel free to give them a try! 🔥🔥🔥
+- **\[2024.09.05\]** We now support answer extraction through model post-processing to provide a more accurate representation of the model's capabilities. As part of this update, we have integrated [XFinder](https://github.com/IAAR-Shanghai/xFinder) as our first post-processing model. For more detailed information, please refer to the [documentation](opencompass/utils/postprocessors/xfinder/README.md), and give it a try! 🔥🔥🔥
+- **\[2024.08.20\]** OpenCompass now supports the [SciCode](https://github.com/scicode-bench/SciCode): A Research Coding Benchmark Curated by Scientists. 🔥🔥🔥
+- **\[2024.08.16\]** OpenCompass now supports the brand new long-context language model evaluation benchmark — [RULER](https://arxiv.org/pdf/2404.06654). RULER provides an evaluation of long-context including retrieval, multi-hop tracing, aggregation, and question answering through flexible configurations. Check out the [RULER](configs/datasets/ruler/README.md) evaluation config now! 🔥🔥🔥
+- **\[2024.08.09\]** We have released the example data and configuration for the CompassBench-202408, welcome to [CompassBench](https://opencompass.readthedocs.io/zh-cn/latest/advanced_guides/compassbench_intro.html) for more details. 🔥🔥🔥
+- **\[2024.08.01\]** We supported the [Gemma2](https://huggingface.co/collections/google/gemma-2-release-667d6600fd5220e7b967f315) models. Welcome to try! 🔥🔥🔥
+- **\[2024.07.23\]** We supported the [ModelScope](www.modelscope.cn) datasets, you can load them on demand without downloading all the data to your local disk. Welcome to try! 🔥🔥🔥
+- **\[2024.07.17\]** We are excited to announce the release of NeedleBench's [technical report](http://arxiv.org/abs/2407.11963). We invite you to visit our [support documentation](https://opencompass.readthedocs.io/en/latest/advanced_guides/needleinahaystack_eval.html) for detailed evaluation guidelines. 🔥🔥🔥
+- **\[2024.07.04\]** OpenCompass now supports InternLM2.5, which has **outstanding reasoning capability**, **1M Context window and** and **stronger tool use**, you can try the models in [OpenCompass Config](https://github.com/open-compass/opencompass/tree/main/configs/models/hf_internlm) and [InternLM](https://github.com/InternLM/InternLM) .🔥🔥🔥.
+- **\[2024.06.20\]** OpenCompass now supports one-click switching between inference acceleration backends, enhancing the efficiency of the evaluation process. In addition to the default HuggingFace inference backend, it now also supports popular backends [LMDeploy](https://github.com/InternLM/lmdeploy) and [vLLM](https://github.com/vllm-project/vllm). This feature is available via a simple command-line switch and through deployment APIs. For detailed usage, see the [documentation](docs/en/advanced_guides/accelerator_intro.md).🔥🔥🔥.
 
 > [More](docs/en/notes/news.md)
 
+## 📊 Leaderboard
+
+We provide [OpenCompass Leaderboard](https://rank.opencompass.org.cn/home) for the community to rank all public models and API models. If you would like to join the evaluation, please provide the model repository URL or a standard API interface to the email address `opencompass@pjlab.org.cn`.
+
+You can also refer to [CompassAcademic](configs/eval_academic_leaderboard_202412.py) to quickly reproduce the leaderboard results. The currently selected datasets include Knowledge Reasoning (MMLU-Pro/GPQA Diamond), Logical Reasoning (BBH), Mathematical Reasoning (MATH-500, AIME), Code Generation (LiveCodeBench, HumanEval), and Instruction Following (IFEval)."
+
+<p align="right"><a href="#top">🔝Back to top</a></p>
+
+## 🛠️ Installation
+
+Below are the steps for quick installation and datasets preparation.
+
+### 💻 Environment Setup
+
+We highly recommend using conda to manage your python environment.
+
+- #### Create your virtual environment
+
+  ```bash
+  conda create --name opencompass python=3.10 -y
+  conda activate opencompass
+  ```
+
+- #### Install OpenCompass via pip
+
+  ```bash
+    pip install -U opencompass
+
+    ## Full installation (with support for more datasets)
+    # pip install "opencompass[full]"
+
+    ## Environment with model acceleration frameworks
+    ## Manage different acceleration frameworks using virtual environments
+    ## since they usually have dependency conflicts with each other.
+    # pip install "opencompass[lmdeploy]"
+    # pip install "opencompass[vllm]"
+
+    ## API evaluation (i.e. Openai, Qwen)
+    # pip install "opencompass[api]"
+  ```
+
+- #### Install OpenCompass from source
+
+  If you want to use opencompass's latest features, or develop new features, you can also build it from source
+
+  ```bash
+    git clone https://github.com/open-compass/opencompass opencompass
+    cd opencompass
+    pip install -e .
+    # pip install -e ".[full]"
+    # pip install -e ".[vllm]"
+  ```
+
+### 📂 Data Preparation
+
+You can choose one for the following method to prepare datasets.
+
+#### Offline Preparation
+
+You can download and extract the datasets with the following commands:
+
+```bash
+# Download dataset to data/ folder
+wget https://github.com/open-compass/opencompass/releases/download/0.2.2.rc1/OpenCompassData-core-20240207.zip
+unzip OpenCompassData-core-20240207.zip
+```
+
+#### Automatic Download from OpenCompass
+
+We have supported download datasets automatic from the OpenCompass storage server. You can run the evaluation with extra `--dry-run` to download these datasets.
+Currently, the supported datasets are listed in [here](https://github.com/open-compass/opencompass/blob/main/opencompass/utils/datasets_info.py#L259). More datasets will be uploaded recently.
+
+#### (Optional) Automatic Download with ModelScope
+
+Also you can use the [ModelScope](www.modelscope.cn) to load the datasets on demand.
+
+Installation:
+
+```bash
+pip install modelscope[framework]
+export DATASET_SOURCE=ModelScope
+```
+
+Then submit the evaluation task without downloading all the data to your local disk. Available datasets include:
+
+```bash
+humaneval, triviaqa, commonsenseqa, tydiqa, strategyqa, cmmlu, lambada, piqa, ceval, math, LCSTS, Xsum, winogrande, openbookqa, AGIEval, gsm8k, nq, race, siqa, mbpp, mmlu, hellaswag, ARC, BBH, xstory_cloze, summedits, GAOKAO-BENCH, OCNLI, cmnli
+```
+
+Some third-party features, like Humaneval and Llama, may require additional steps to work properly, for detailed steps please refer to the [Installation Guide](https://opencompass.readthedocs.io/en/latest/get_started/installation.html).
+
+<p align="right"><a href="#top">🔝Back to top</a></p>
+
+## 🏗️ ️Evaluation
+
+After ensuring that OpenCompass is installed correctly according to the above steps and the datasets are prepared. Now you can start your first evaluation using OpenCompass!
+
+### Your first evaluation with OpenCompass!
+
+OpenCompass support setting your configs via CLI or a python script. For simple evaluation settings we recommend using CLI, for more complex evaluation, it is suggested using the script way. You can find more example scripts under the configs folder.
+
+```bash
+# CLI
+opencompass --models hf_internlm2_5_1_8b_chat --datasets demo_gsm8k_chat_gen
+
+# Python scripts
+opencompass examples/eval_chat_demo.py
+```
+
+You can find more script examples under [examples](./examples) folder.
+
+### API evaluation
+
+OpenCompass, by its design, does not really discriminate between open-source models and API models. You can evaluate both model types in the same way or even in one settings.
+
+```bash
+export OPENAI_API_KEY="YOUR_OPEN_API_KEY"
+# CLI
+opencompass --models gpt_4o_2024_05_13 --datasets demo_gsm8k_chat_gen
+
+# Python scripts
+opencompass examples/eval_api_demo.py
+
+# You can use o1_mini_2024_09_12/o1_preview_2024_09_12  for o1 models, we set max_completion_tokens=8192 as default.
+```
+
+### Accelerated Evaluation
+
+Additionally, if you want to use an inference backend other than HuggingFace for accelerated evaluation, such as LMDeploy or vLLM, you can do so with the command below. Please ensure that you have installed the necessary packages for the chosen backend and that your model supports accelerated inference with it. For more information, see the documentation on inference acceleration backends [here](docs/en/advanced_guides/accelerator_intro.md). Below is an example using LMDeploy:
+
+```bash
+# CLI
+opencompass --models hf_internlm2_5_1_8b_chat --datasets demo_gsm8k_chat_gen -a lmdeploy
+
+# Python scripts
+opencompass examples/eval_lmdeploy_demo.py
+```
+
+### Supported Models and Datasets
+
+OpenCompass has predefined configurations for many models and datasets. You can list all available model and dataset configurations using the [tools](./docs/en/tools.md#list-configs).
+
+```bash
+# List all configurations
+python tools/list_configs.py
+# List all configurations related to llama and mmlu
+python tools/list_configs.py llama mmlu
+```
+
+#### Supported Models
+
+If the model is not on the list but supported by Huggingface AutoModel class or encapsulation of inference engine based on OpenAI interface (see [docs](https://opencompass.readthedocs.io/en/latest/advanced_guides/new_model.html) for details), you can also evaluate it with OpenCompass. You are welcome to contribute to the maintenance of the OpenCompass supported model and dataset lists.
+
+```bash
+opencompass --datasets demo_gsm8k_chat_gen --hf-type chat --hf-path internlm/internlm2_5-1_8b-chat
+```
+
+#### Supported Datasets
+
+Currently, OpenCompass have provided standard recommended configurations for datasets. Generally, config files ending with `_gen.py` or `_llm_judge_gen.py` will point to the recommended config we provide for this dataset. You can refer to [docs](https://opencompass.readthedocs.io/en/latest/dataset_statistics.html) for more details.
+
+```bash
+# Recommended Evaluation Config based on Rules
+opencompass --datasets aime2024_gen --models hf_internlm2_5_1_8b_chat
+
+# Recommended Evaluation Config based on LLM Judge
+opencompass --datasets aime2024_llmjudge_gen --models hf_internlm2_5_1_8b_chat
+```
+
+If you want to use multiple GPUs to evaluate the model in data parallel, you can use `--max-num-worker`.
+
+```bash
+CUDA_VISIBLE_DEVICES=0,1 opencompass --datasets demo_gsm8k_chat_gen --hf-type chat --hf-path internlm/internlm2_5-1_8b-chat --max-num-worker 2
+```
+
+> \[!TIP\]
+>
+> `--hf-num-gpus` is used for model parallel(huggingface format), `--max-num-worker` is used for data parallel.
+
+> \[!TIP\]
+>
+> configuration with `_ppl` is designed for base model typically.
+> configuration with `_gen` can be used for both base model and chat model.
+
+Through the command line or configuration files, OpenCompass also supports evaluating APIs or custom models, as well as more diversified evaluation strategies. Please read the [Quick Start](https://opencompass.readthedocs.io/en/latest/get_started/quick_start.html) to learn how to run an evaluation task.
+
+<p align="right"><a href="#top">🔝Back to top</a></p>
+
+## 📣 OpenCompass 2.0
+
+We are thrilled to introduce OpenCompass 2.0, an advanced suite featuring three key components: [CompassKit](https://github.com/open-compass), [CompassHub](https://hub.opencompass.org.cn/home), and [CompassRank](https://rank.opencompass.org.cn/home).
+![oc20](https://github.com/tonysy/opencompass/assets/7881589/90dbe1c0-c323-470a-991e-2b37ab5350b2)
+
+**CompassRank** has been significantly enhanced into the leaderboards that now incorporates both open-source benchmarks and proprietary benchmarks. This upgrade allows for a more comprehensive evaluation of models across the industry.
+
+**CompassHub** presents a pioneering benchmark browser interface, designed to simplify and expedite the exploration and utilization of an extensive array of benchmarks for researchers and practitioners alike. To enhance the visibility of your own benchmark within the community, we warmly invite you to contribute it to CompassHub. You may initiate the submission process by clicking [here](https://hub.opencompass.org.cn/dataset-submit).
+
+**CompassKit** is a powerful collection of evaluation toolkits specifically tailored for Large Language Models and Large Vision-language Models. It provides an extensive set of tools to assess and measure the performance of these complex models effectively. Welcome to try our toolkits for in your research and products.
+
 ## ✨ Introduction
 
 ![image](https://github.com/open-compass/opencompass/assets/22607038/f45fe125-4aed-4f8c-8fe8-df4efb41a8ea)
@@ -97,342 +295,17 @@ OpenCompass is a one-stop platform for large model evaluation, aiming to provide
 
 - **Experiment management and reporting mechanism**: Use config files to fully record each experiment, and support real-time reporting of results.
 
-## 📊 Leaderboard
-
-We provide [OpenCompass Leaderboard](https://rank.opencompass.org.cn/home) for the community to rank all public models and API models. If you would like to join the evaluation, please provide the model repository URL or a standard API interface to the email address `opencompass@pjlab.org.cn`.
-
-<p align="right"><a href="#top">🔝Back to top</a></p>
-
-## 🛠️ Installation
-
-Below are the steps for quick installation and datasets preparation.
-
-### 💻 Environment Setup
-
-#### Open-source Models with GPU
-
-```bash
-conda create --name opencompass python=3.10 pytorch torchvision pytorch-cuda -c nvidia -c pytorch -y
-conda activate opencompass
-git clone https://github.com/open-compass/opencompass opencompass
-cd opencompass
-pip install -e .
-```
-
-#### API Models with CPU-only
-
-```bash
-conda create -n opencompass python=3.10 pytorch torchvision torchaudio cpuonly -c pytorch -y
-conda activate opencompass
-git clone https://github.com/open-compass/opencompass opencompass
-cd opencompass
-pip install -e .
-# also please install requirements packages via `pip install -r requirements/api.txt` for API models if needed.
-```
-
-### 📂 Data Preparation
-
-```bash
-# Download dataset to data/ folder
-wget https://github.com/open-compass/opencompass/releases/download/0.2.2.rc1/OpenCompassData-core-20240207.zip
-unzip OpenCompassData-core-20240207.zip
-```
-
-Some third-party features, like Humaneval and Llama, may require additional steps to work properly, for detailed steps please refer to the [Installation Guide](https://opencompass.readthedocs.io/en/latest/get_started/installation.html).
-
-<p align="right"><a href="#top">🔝Back to top</a></p>
-
-## 🏗️ ️Evaluation
-
-After ensuring that OpenCompass is installed correctly according to the above steps and the datasets are prepared, you can evaluate the performance of the LLaMA-7b model on the MMLU and C-Eval datasets using the following command:
-
-```bash
-python run.py --models hf_llama_7b --datasets mmlu_ppl ceval_ppl
-```
-
-OpenCompass has predefined configurations for many models and datasets. You can list all available model and dataset configurations using the [tools](./docs/en/tools.md#list-configs).
-
-```bash
-# List all configurations
-python tools/list_configs.py
-# List all configurations related to llama and mmlu
-python tools/list_configs.py llama mmlu
-```
-
-You can also evaluate other HuggingFace models via command line. Taking LLaMA-7b as an example:
-
-```bash
-python run.py --datasets ceval_ppl mmlu_ppl --hf-type base --hf-path huggyllama/llama-7b
-```
-
-> \[!TIP\]
->
-> configuration with `_ppl` is designed for base model typically.
-> configuration with `_gen` can be used for both base model and chat model.
-
-Through the command line or configuration files, OpenCompass also supports evaluating APIs or custom models, as well as more diversified evaluation strategies. Please read the [Quick Start](https://opencompass.readthedocs.io/en/latest/get_started/quick_start.html) to learn how to run an evaluation task.
-
-<p align="right"><a href="#top">🔝Back to top</a></p>
-
 ## 📖 Dataset Support
 
-<table align="center">
-  <tbody>
-    <tr align="center" valign="bottom">
-      <td>
-        <b>Language</b>
-      </td>
-      <td>
-        <b>Knowledge</b>
-      </td>
-      <td>
-        <b>Reasoning</b>
-      </td>
-      <td>
-        <b>Examination</b>
-      </td>
-    </tr>
-    <tr valign="top">
-      <td>
-<details open>
-<summary><b>Word Definition</b></summary>
+We have supported a statistical list of all datasets that can be used on this platform in the documentation on the OpenCompass website.
 
-- WiC
-- SummEdits
+You can quickly find the dataset you need from the list through sorting, filtering, and searching functions.
 
-</details>
+In addition, we provide a recommended configuration for each dataset, and some datasets also support LLM Judge-based configurations.
 
-<details open>
-<summary><b>Idiom Learning</b></summary>
+Please refer to the dataset statistics chapter of [docs](https://opencompass.readthedocs.io/en/latest/dataset_statistics.html) for details.
 
-- CHID
-
-</details>
-
-<details open>
-<summary><b>Semantic Similarity</b></summary>
-
-- AFQMC
-- BUSTM
-
-</details>
-
-<details open>
-<summary><b>Coreference Resolution</b></summary>
-
-- CLUEWSC
-- WSC
-- WinoGrande
-
-</details>
-
-<details open>
-<summary><b>Translation</b></summary>
-
-- Flores
-- IWSLT2017
-
-</details>
-
-<details open>
-<summary><b>Multi-language Question Answering</b></summary>
-
-- TyDi-QA
-- XCOPA
-
-</details>
-
-<details open>
-<summary><b>Multi-language Summary</b></summary>
-
-- XLSum
-
-</details>
-      </td>
-      <td>
-<details open>
-<summary><b>Knowledge Question Answering</b></summary>
-
-- BoolQ
-- CommonSenseQA
-- NaturalQuestions
-- TriviaQA
-
-</details>
-      </td>
-      <td>
-<details open>
-<summary><b>Textual Entailment</b></summary>
-
-- CMNLI
-- OCNLI
-- OCNLI_FC
-- AX-b
-- AX-g
-- CB
-- RTE
-- ANLI
-
-</details>
-
-<details open>
-<summary><b>Commonsense Reasoning</b></summary>
-
-- StoryCloze
-- COPA
-- ReCoRD
-- HellaSwag
-- PIQA
-- SIQA
-
-</details>
-
-<details open>
-<summary><b>Mathematical Reasoning</b></summary>
-
-- MATH
-- GSM8K
-
-</details>
-
-<details open>
-<summary><b>Theorem Application</b></summary>
-
-- TheoremQA
-- StrategyQA
-- SciBench
-
-</details>
-
-<details open>
-<summary><b>Comprehensive Reasoning</b></summary>
-
-- BBH
-
-</details>
-      </td>
-      <td>
-<details open>
-<summary><b>Junior High, High School, University, Professional Examinations</b></summary>
-
-- C-Eval
-- AGIEval
-- MMLU
-- GAOKAO-Bench
-- CMMLU
-- ARC
-- Xiezhi
-
-</details>
-
-<details open>
-<summary><b>Medical Examinations</b></summary>
-
-- CMB
-
-</details>
-      </td>
-    </tr>
-</td>
-    </tr>
-  </tbody>
-  <tbody>
-    <tr align="center" valign="bottom">
-      <td>
-        <b>Understanding</b>
-      </td>
-      <td>
-        <b>Long Context</b>
-      </td>
-      <td>
-        <b>Safety</b>
-      </td>
-      <td>
-        <b>Code</b>
-      </td>
-    </tr>
-    <tr valign="top">
-      <td>
-<details open>
-<summary><b>Reading Comprehension</b></summary>
-
-- C3
-- CMRC
-- DRCD
-- MultiRC
-- RACE
-- DROP
-- OpenBookQA
-- SQuAD2.0
-
-</details>
-
-<details open>
-<summary><b>Content Summary</b></summary>
-
-- CSL
-- LCSTS
-- XSum
-- SummScreen
-
-</details>
-
-<details open>
-<summary><b>Content Analysis</b></summary>
-
-- EPRSTMT
-- LAMBADA
-- TNEWS
-
-</details>
-      </td>
-      <td>
-<details open>
-<summary><b>Long Context Understanding</b></summary>
-
-- LEval
-- LongBench
-- GovReports
-- NarrativeQA
-- Qasper
-
-</details>
-      </td>
-      <td>
-<details open>
-<summary><b>Safety</b></summary>
-
-- CivilComments
-- CrowsPairs
-- CValues
-- JigsawMultilingual
-- TruthfulQA
-
-</details>
-<details open>
-<summary><b>Robustness</b></summary>
-
-- AdvGLUE
-
-</details>
-      </td>
-      <td>
-<details open>
-<summary><b>Code</b></summary>
-
-- HumanEval
-- HumanEvalX
-- MBPP
-- APPs
-- DS1000
-
-</details>
-      </td>
-    </tr>
-</td>
-    </tr>
-  </tbody>
-</table>
+<p align="right"><a href="#top">🔝Back to top</a></p>
 
 ## 📖 Model Support
 
@@ -452,20 +325,21 @@ Through the command line or configuration files, OpenCompass also supports evalu
     <tr valign="top">
       <td>
 
+- [Alpaca](https://github.com/tatsu-lab/stanford_alpaca)
+- [Baichuan](https://github.com/baichuan-inc)
+- [BlueLM](https://github.com/vivo-ai-lab/BlueLM)
+- [ChatGLM2](https://github.com/THUDM/ChatGLM2-6B)
+- [ChatGLM3](https://github.com/THUDM/ChatGLM3-6B)
+- [Gemma](https://huggingface.co/google/gemma-7b)
 - [InternLM](https://github.com/InternLM/InternLM)
 - [LLaMA](https://github.com/facebookresearch/llama)
 - [LLaMA3](https://github.com/meta-llama/llama3)
-- [Vicuna](https://github.com/lm-sys/FastChat)
-- [Alpaca](https://github.com/tatsu-lab/stanford_alpaca)
-- [Baichuan](https://github.com/baichuan-inc)
-- [WizardLM](https://github.com/nlpxucan/WizardLM)
-- [ChatGLM2](https://github.com/THUDM/ChatGLM2-6B)
-- [ChatGLM3](https://github.com/THUDM/ChatGLM3-6B)
-- [TigerBot](https://github.com/TigerResearch/TigerBot)
 - [Qwen](https://github.com/QwenLM/Qwen)
-- [BlueLM](https://github.com/vivo-ai-lab/BlueLM)
-- [Gemma](https://huggingface.co/google/gemma-7b)
-- ...
+- [TigerBot](https://github.com/TigerResearch/TigerBot)
+- [Vicuna](https://github.com/lm-sys/FastChat)
+- [WizardLM](https://github.com/nlpxucan/WizardLM)
+- [Yi](https://github.com/01-ai/Yi)
+- ……
 
 </td>
 <td>
@@ -495,7 +369,7 @@ Through the command line or configuration files, OpenCompass also supports evalu
 ## 🔜 Roadmap
 
 - [x] Subjective Evaluation
-  - [ ] Release CompassAreana
+  - [x] Release CompassAreana.
   - [x] Subjective evaluation.
 - [x] Long-context
   - [x] Long-context evaluation with extensive datasets.
@@ -504,10 +378,10 @@ Through the command line or configuration files, OpenCompass also supports evalu
   - [ ] Coding evaluation leaderboard.
   - [x] Non-python language evaluation service.
 - [x] Agent
-  - [ ] Support various agenet framework.
+  - [ ] Support various agent frameworks.
   - [x] Evaluation of tool use of the LLMs.
 - [x] Robustness
-  - [x] Support various attack method
+  - [x] Support various attack methods.
 
 ## 👷‍♂️ Contributing
 

README_zh-CN.md

@@ -34,16 +34,6 @@
 >
 > **收藏项目**，你将能第一时间获取 OpenCompass 的最新动态～⭐️
 
-## 📣 OpenCompass 2.0
-
-我们很高兴发布 OpenCompass 司南 2.0 大模型评测体系，它主要由三大核心模块构建而成：[CompassKit](https://github.com/open-compass)、[CompassHub](https://hub.opencompass.org.cn/home)以及[CompassRank](https://rank.opencompass.org.cn/home)。
-
-**CompassRank** 系统进行了重大革新与提升，现已成为一个兼容并蓄的排行榜体系，不仅囊括了开源基准测试项目，还包含了私有基准测试。此番升级极大地拓宽了对行业内各类模型进行全面而深入测评的可能性。
-
-**CompassHub** 创新性地推出了一个基准测试资源导航平台，其设计初衷旨在简化和加快研究人员及行业从业者在多样化的基准测试库中进行搜索与利用的过程。为了让更多独具特色的基准测试成果得以在业内广泛传播和应用，我们热忱欢迎各位将自定义的基准数据贡献至CompassHub平台。只需轻点鼠标，通过访问[这里](https://hub.opencompass.org.cn/dataset-submit)，即可启动提交流程。
-
-**CompassKit** 是一系列专为大型语言模型和大型视觉-语言模型打造的强大评估工具合集，它所提供的全面评测工具集能够有效地对这些复杂模型的功能性能进行精准测量和科学评估。在此，我们诚挚邀请您在学术研究或产品研发过程中积极尝试运用我们的工具包，以助您取得更加丰硕的研究成果和产品优化效果。
-
 <details>
   <summary><kbd>Star History</kbd></summary>
   <picture>
@@ -63,23 +53,225 @@
 🔥🔥🔥 祝贺 **OpenCompass 作为大模型标准测试工具被Meta AI官方推荐**, 点击 Llama 的 [入门文档](https://ai.meta.com/llama/get-started/#validation) 获取更多信息。
 
 > **注意**<br />
-> 我们正式启动 OpenCompass 共建计划，诚邀社区用户为 OpenCompass 提供更具代表性和可信度的客观评测数据集!
-> 点击 [Issue](https://github.com/open-compass/opencompass/issues/248) 获取更多数据集.
-> 让我们携手共进，打造功能强大易用的大模型评测平台！
+> 重要通知：从 v0.4.0 版本开始，所有位于 ./configs/datasets、./configs/models 和 ./configs/summarizers 目录下的 AMOTIC 配置文件将迁移至 opencompass 包中。请及时更新您的配置文件路径。
 
 ## 🚀 最新进展 <a><img width="35" height="20" src="https://user-images.githubusercontent.com/12782558/212848161-5e783dd6-11e8-4fe0-bbba-39ffb77730be.png"></a>
 
-- **\[2024.05.08\]** 我们支持了以下四个MoE模型的评测配置文件: [Mixtral-8x22B-v0.1](configs/models/mixtral/hf_mixtral_8x22b_v0_1.py), [Mixtral-8x22B-Instruct-v0.1](configs/models/mixtral/hf_mixtral_8x22b_instruct_v0_1.py), [Qwen1.5-MoE-A2.7B](configs/models/qwen/hf_qwen1_5_moe_a2_7b.py), [Qwen1.5-MoE-A2.7B-Chat](configs/models/qwen/hf_qwen1_5_moe_a2_7b_chat.py) 。欢迎试用!
-- **\[2024.04.30\]** 我们支持了计算模型在给定[数据集](configs/datasets/llm_compression/README.md)上的压缩率（Bits per Character）的评测方法（[官方文献](https://github.com/hkust-nlp/llm-compression-intelligence)）。欢迎试用[llm-compression](configs/eval_llm_compression.py)评测集! 🔥🔥🔥
-- **\[2024.04.26\]** 我们报告了典型LLM在常用基准测试上的表现，欢迎访问[文档](https://opencompass.readthedocs.io/zh-cn/latest/user_guides/corebench.html)以获取更多信息！🔥🔥🔥.
-- **\[2024.04.26\]** 我们废弃了 OpenCompass 进行多模态大模型评测的功能，相关功能转移至 [VLMEvalKit](https://github.com/open-compass/VLMEvalKit)，推荐使用！🔥🔥🔥.
-- **\[2024.04.26\]** 我们支持了 [ArenaHard评测](configs/eval_subjective_arena_hard.py) 欢迎试用！🔥🔥🔥.
-- **\[2024.04.22\]** 我们支持了 [LLaMA3](configs/models/hf_llama/hf_llama3_8b.py) 和 [LLaMA3-Instruct](configs/models/hf_llama/hf_llama3_8b_instruct.py) 的评测，欢迎试用！🔥🔥🔥.
-- **\[2024.02.29\]** 我们支持了MT-Bench、AlpacalEval和AlignBench，更多信息可以在[这里](https://opencompass.readthedocs.io/en/latest/advanced_guides/subjective_evaluation.html)找到。
-- **\[2024.01.30\]** 我们发布了OpenCompass 2.0。更多信息，请访问[CompassKit](https://github.com/open-compass)、[CompassHub](https://hub.opencompass.org.cn/home)和[CompassRank](https://rank.opencompass.org.cn/home)。
+- **\[2025.04.01\]** OpenCompass 现已支持 `CascadeEvaluator`，允许多个评估器按顺序工作，可以为更复杂的评估场景创建自定义评估流程，查看[文档](docs/zh_cn/advanced_guides/llm_judge.md)了解具体用法！🔥🔥🔥
+- **\[2025.03.11\]** 现已支持 `SuperGPQA`  覆盖285 个研究生学科的知识能力评测，欢迎尝试！🔥🔥🔥
+- **\[2025.02.28\]** 我们为 `DeepSeek-R1` 系列模型添加了教程，请查看 [评估推理模型](docs/zh_cn/user_guides/deepseek_r1.md) 了解更多详情！🔥🔥🔥
+- **\[2025.02.15\]** 我们新增了两个实用的评测工具：用于LLM作为评判器的`GenericLLMEvaluator`和用于数学推理评估的`MATHVerifyEvaluator`。查看[LLM评判器](docs/zh_cn/advanced_guides/llm_judge.md)和[数学能力评测](docs/zh_cn/advanced_guides/general_math.md)文档了解更多详情！🔥🔥🔥
+- **\[2025.01.16\]** 我们现已支持 [InternLM3-8B-Instruct](https://huggingface.co/internlm/internlm3-8b-instruct) 模型，该模型在推理、知识类任务上取得同量级最优性能，欢迎尝试。
+- **\[2024.12.17\]** 我们提供了12月CompassAcademic学术榜单评估脚本 [CompassAcademic](configs/eval_academic_leaderboard_202412.py)，你可以通过简单地配置复现官方评测结果。
+- **\[2024.10.14\]** 现已支持OpenAI多语言问答数据集[MMMLU](https://huggingface.co/datasets/openai/MMMLU)，欢迎尝试! 🔥🔥🔥
+- **\[2024.09.19\]** 现已支持[Qwen2.5](https://huggingface.co/Qwen)(0.5B to 72B) ，可以使用多种推理后端(huggingface/vllm/lmdeploy), 欢迎尝试! 🔥🔥🔥
+- **\[2024.09.05\]** 现已支持OpenAI o1 模型(`o1-mini-2024-09-12` and `o1-preview-2024-09-12`), 欢迎尝试! 🔥🔥🔥
+- **\[2024.09.05\]** OpenCompass 现在支持通过模型后处理来进行答案提取，以更准确地展示模型的能力。作为此次更新的一部分，我们集成了 [XFinder](https://github.com/IAAR-Shanghai/xFinder) 作为首个后处理模型。具体信息请参阅 [文档](opencompass/utils/postprocessors/xfinder/README.md)，欢迎尝试！ 🔥🔥🔥
+- **\[2024.08.20\]** OpenCompass 现已支持 [SciCode](https://github.com/scicode-bench/SciCode): A Research Coding Benchmark Curated by Scientists。 🔥🔥🔥
+- **\[2024.08.16\]** OpenCompass 现已支持全新的长上下文语言模型评估基准——[RULER](https://arxiv.org/pdf/2404.06654)。RULER 通过灵活的配置，提供了对长上下文包括检索、多跳追踪、聚合和问答等多种任务类型的评测，欢迎访问[RULER](configs/datasets/ruler/README.md)。🔥🔥🔥
+- **\[2024.07.23\]** 我们支持了[Gemma2](https://huggingface.co/collections/google/gemma-2-release-667d6600fd5220e7b967f315)模型，欢迎试用！🔥🔥🔥
+- **\[2024.07.23\]** 我们支持了[ModelScope](www.modelscope.cn)数据集，您可以按需加载，无需事先下载全部数据到本地，欢迎试用！🔥🔥🔥
+- **\[2024.07.17\]** 我们发布了CompassBench-202407榜单的示例数据和评测规则，敬请访问 [CompassBench](https://opencompass.readthedocs.io/zh-cn/latest/advanced_guides/compassbench_intro.html) 获取更多信息。 🔥🔥🔥
+- **\[2024.07.17\]** 我们正式发布 NeedleBench 的[技术报告](http://arxiv.org/abs/2407.11963)。诚邀您访问我们的[帮助文档](https://opencompass.readthedocs.io/zh-cn/latest/advanced_guides/needleinahaystack_eval.html)进行评估。🔥🔥🔥
+- **\[2024.07.04\]** OpenCompass 现已支持 InternLM2.5， 它拥有卓越的推理性能、有效支持百万字超长上下文以及工具调用能力整体升级，欢迎访问[OpenCompass Config](https://github.com/open-compass/opencompass/tree/main/configs/models/hf_internlm) 和 [InternLM](https://github.com/InternLM/InternLM) .🔥🔥🔥.
+- **\[2024.06.20\]** OpenCompass 现已支持一键切换推理加速后端，助力评测过程更加高效。除了默认的HuggingFace推理后端外，还支持了常用的 [LMDeploy](https://github.com/InternLM/lmdeploy) 和 [vLLM](https://github.com/vllm-project/vllm) ，支持命令行一键切换和部署 API 加速服务两种方式，详细使用方法见[文档](docs/zh_cn/advanced_guides/accelerator_intro.md)。欢迎试用！🔥🔥🔥.
 
 > [更多](docs/zh_cn/notes/news.md)
 
+## 📊 性能榜单
+
+我们将陆续提供开源模型和 API 模型的具体性能榜单，请见 [OpenCompass Leaderboard](https://rank.opencompass.org.cn/home) 。如需加入评测，请提供模型仓库地址或标准的 API 接口至邮箱  `opencompass@pjlab.org.cn`.
+
+你也可以参考[CompassAcademic](configs/eval_academic_leaderboard_202412.py)，快速地复现榜单的结果，目前选取的数据集包括 综合知识推理 (MMLU-Pro/GPQA Diamond) ,逻辑推理 (BBH) ,数学推理 (MATH-500, AIME) ,代码生成 (LiveCodeBench, HumanEval) ,指令跟随 (IFEval) 。
+
+<p align="right"><a href="#top">🔝返回顶部</a></p>
+
+## 🛠️ 安装指南
+
+下面提供了快速安装和数据集准备的步骤。
+
+### 💻 环境搭建
+
+我们强烈建议使用 `conda` 来管理您的 Python 环境。
+
+- #### 创建虚拟环境
+
+  ```bash
+  conda create --name opencompass python=3.10 -y
+  conda activate opencompass
+  ```
+
+- #### 通过pip安装OpenCompass
+
+  ```bash
+  # 支持绝大多数数据集及模型
+  pip install -U opencompass
+
+  # 完整安装（支持更多数据集）
+  # pip install "opencompass[full]"
+
+  # 模型推理后端，由于这些推理后端通常存在依赖冲突，建议使用不同的虚拟环境来管理它们。
+  # pip install "opencompass[lmdeploy]"
+  # pip install "opencompass[vllm]"
+
+  # API 测试（例如 OpenAI、Qwen）
+  # pip install "opencompass[api]"
+  ```
+
+- #### 基于源码安装OpenCompass
+
+  如果希望使用 OpenCompass 的最新功能，也可以从源代码构建它：
+
+  ```bash
+  git clone https://github.com/open-compass/opencompass opencompass
+  cd opencompass
+  pip install -e .
+  # pip install -e ".[full]"
+  # pip install -e ".[vllm]"
+  ```
+
+### 📂 数据准备
+
+#### 提前离线下载
+
+OpenCompass支持使用本地数据集进行评测，数据集的下载和解压可以通过以下命令完成：
+
+```bash
+# 下载数据集到 data/ 处
+wget https://github.com/open-compass/opencompass/releases/download/0.2.2.rc1/OpenCompassData-core-20240207.zip
+unzip OpenCompassData-core-20240207.zip
+```
+
+#### 从 OpenCompass 自动下载
+
+我们已经支持从OpenCompass存储服务器自动下载数据集。您可以通过额外的 `--dry-run` 参数来运行评估以下载这些数据集。
+目前支持的数据集列表在[这里](https://github.com/open-compass/opencompass/blob/main/opencompass/utils/datasets_info.py#L259)。更多数据集将会很快上传。
+
+#### (可选) 使用 ModelScope 自动下载
+
+另外，您还可以使用[ModelScope](www.modelscope.cn)来加载数据集：
+环境准备：
+
+```bash
+pip install modelscope
+export DATASET_SOURCE=ModelScope
+```
+
+配置好环境后，无需下载全部数据，直接提交评测任务即可。目前支持的数据集有：
+
+```bash
+humaneval, triviaqa, commonsenseqa, tydiqa, strategyqa, cmmlu, lambada, piqa, ceval, math, LCSTS, Xsum, winogrande, openbookqa, AGIEval, gsm8k, nq, race, siqa, mbpp, mmlu, hellaswag, ARC, BBH, xstory_cloze, summedits, GAOKAO-BENCH, OCNLI, cmnli
+```
+
+有部分第三方功能,如 Humaneval 以及 Llama,可能需要额外步骤才能正常运行，详细步骤请参考[安装指南](https://opencompass.readthedocs.io/zh_CN/latest/get_started/installation.html)。
+
+<p align="right"><a href="#top">🔝返回顶部</a></p>
+
+## 🏗️ ️评测
+
+在确保按照上述步骤正确安装了 OpenCompass 并准备好了数据集之后，现在您可以开始使用 OpenCompass 进行首次评估！
+
+- ### 首次评测
+
+  OpenCompass 支持通过命令行界面 (CLI) 或 Python 脚本来设置配置。对于简单的评估设置，我们推荐使用 CLI；而对于更复杂的评估，则建议使用脚本方式。你可以在examples文件夹下找到更多脚本示例。
+
+  ```bash
+  # 命令行界面 (CLI)
+  opencompass --models hf_internlm2_5_1_8b_chat --datasets demo_gsm8k_chat_gen
+
+  # Python 脚本
+  opencompass examples/eval_chat_demo.py
+  ```
+
+  你可以在[examples](./examples) 文件夹下找到更多的脚本示例。
+
+- ### API评测
+
+  OpenCompass 在设计上并不区分开源模型与 API 模型。您可以以相同的方式或甚至在同一设置中评估这两种类型的模型。
+
+  ```bash
+  export OPENAI_API_KEY="YOUR_OPEN_API_KEY"
+  # 命令行界面 (CLI)
+  opencompass --models gpt_4o_2024_05_13 --datasets demo_gsm8k_chat_gen
+
+  # Python 脚本
+  opencompass  examples/eval_api_demo.py
+
+
+  # 现已支持 o1_mini_2024_09_12/o1_preview_2024_09_12  模型, 默认情况下 max_completion_tokens=8192.
+  ```
+
+- ### 推理后端
+
+  另外，如果您想使用除 HuggingFace 之外的推理后端来进行加速评估，比如 LMDeploy 或 vLLM，可以通过以下命令进行。请确保您已经为所选的后端安装了必要的软件包，并且您的模型支持该后端的加速推理。更多信息，请参阅关于推理加速后端的文档 [这里](docs/zh_cn/advanced_guides/accelerator_intro.md)。以下是使用 LMDeploy 的示例：
+
+  ```bash
+  opencompass --models hf_internlm2_5_1_8b_chat --datasets demo_gsm8k_chat_gen -a lmdeploy
+  ```
+
+- ### 支持的模型与数据集
+
+  OpenCompass 预定义了许多模型和数据集的配置，你可以通过 [工具](./docs/zh_cn/tools.md#ListConfigs) 列出所有可用的模型和数据集配置。
+
+  ```bash
+  # 列出所有配置
+  python tools/list_configs.py
+  # 列出所有跟 llama 及 mmlu 相关的配置
+  python tools/list_configs.py llama mmlu
+  ```
+
+  #### 支持的模型
+
+  如果模型不在列表中，但支持 Huggingface AutoModel 类或支持针对 OpenAI 接口的推理引擎封装（详见[官方文档](https://opencompass.readthedocs.io/zh-cn/latest/advanced_guides/new_model.html)），您仍然可以使用 OpenCompass 对其进行评估。欢迎您贡献维护 OpenCompass 支持的模型和数据集列表。
+
+  ```bash
+  opencompass --datasets demo_gsm8k_chat_gen --hf-type chat --hf-path internlm/internlm2_5-1_8b-chat
+  ```
+
+  #### 支持的数据集
+
+  目前，OpenCompass针对数据集给出了标准的推荐配置。通常，`_gen.py`或`_llm_judge_gen.py`为结尾的配置文件将指向我们为该数据集提供的推荐配置。您可以参阅[官方文档](https://opencompass.readthedocs.io/zh-cn/latest/dataset_statistics.html) 的数据集统计章节来获取详细信息。
+
+  ```bash
+  # 基于规则的推荐配置
+  opencompass --datasets aime2024_gen --models hf_internlm2_5_1_8b_chat
+
+  # 基于LLM Judge的推荐配置
+  opencompass --datasets aime2024_llmjudge_gen --models hf_internlm2_5_1_8b_chat
+  ```
+
+  此外，如果你想在多块 GPU 上使用模型进行推理，您可以使用 `--max-num-worker` 参数。
+
+  ```bash
+  CUDA_VISIBLE_DEVICES=0,1 opencompass --datasets demo_gsm8k_chat_gen --hf-type chat --hf-path internlm/internlm2_5-1_8b-chat --max-num-worker 2
+  ```
+
+> \[!TIP\]
+>
+> `--hf-num-gpus` 用于 模型并行(huggingface 格式)，`--max-num-worker` 用于数据并行。
+
+> \[!TIP\]
+>
+> configuration with `_ppl` is designed for base model typically.
+> 配置带 `_ppl` 的配置设计给基础模型使用。
+> 配置带 `_gen` 的配置可以同时用于基础模型和对话模型。
+
+通过命令行或配置文件，OpenCompass 还支持评测 API 或自定义模型，以及更多样化的评测策略。请阅读[快速开始](https://opencompass.readthedocs.io/zh_CN/latest/get_started/quick_start.html)了解如何运行一个评测任务。
+
+更多教程请查看我们的[文档](https://opencompass.readthedocs.io/zh_CN/latest/index.html)。
+
+<p align="right"><a href="#top">🔝返回顶部</a></p>
+
+## 📣 OpenCompass 2.0
+
+我们很高兴发布 OpenCompass 司南 2.0 大模型评测体系，它主要由三大核心模块构建而成：[CompassKit](https://github.com/open-compass)、[CompassHub](https://hub.opencompass.org.cn/home)以及[CompassRank](https://rank.opencompass.org.cn/home)。
+
+**CompassRank** 系统进行了重大革新与提升，现已成为一个兼容并蓄的排行榜体系，不仅囊括了开源基准测试项目，还包含了私有基准测试。此番升级极大地拓宽了对行业内各类模型进行全面而深入测评的可能性。
+
+**CompassHub** 创新性地推出了一个基准测试资源导航平台，其设计初衷旨在简化和加快研究人员及行业从业者在多样化的基准测试库中进行搜索与利用的过程。为了让更多独具特色的基准测试成果得以在业内广泛传播和应用，我们热忱欢迎各位将自定义的基准数据贡献至CompassHub平台。只需轻点鼠标，通过访问[这里](https://hub.opencompass.org.cn/dataset-submit)，即可启动提交流程。
+
+**CompassKit** 是一系列专为大型语言模型和大型视觉-语言模型打造的强大评估工具合集，它所提供的全面评测工具集能够有效地对这些复杂模型的功能性能进行精准测量和科学评估。在此，我们诚挚邀请您在学术研究或产品研发过程中积极尝试运用我们的工具包，以助您取得更加丰硕的研究成果和产品优化效果。
+
 ## ✨ 介绍
 
 ![image](https://github.com/open-compass/opencompass/assets/22607038/30bcb2e2-3969-4ac5-9f29-ad3f4abb4f3b)
@@ -98,339 +290,13 @@ OpenCompass 是面向大模型评测的一站式平台。其主要特点如下
 
 - **灵活化拓展**：想增加新模型或数据集？想要自定义更高级的任务分割策略，甚至接入新的集群管理系统？OpenCompass 的一切均可轻松扩展！
 
-## 📊 性能榜单
-
-我们将陆续提供开源模型和 API 模型的具体性能榜单，请见 [OpenCompass Leaderboard](https://rank.opencompass.org.cn/home) 。如需加入评测，请提供模型仓库地址或标准的 API 接口至邮箱  `opencompass@pjlab.org.cn`.
-
-<p align="right"><a href="#top">🔝返回顶部</a></p>
-
-## 🛠️ 安装
-
-下面展示了快速安装以及准备数据集的步骤。
-
-### 💻 环境配置
-
-#### 面向开源模型的GPU环境
-
-```bash
-conda create --name opencompass python=3.10 pytorch torchvision pytorch-cuda -c nvidia -c pytorch -y
-conda activate opencompass
-git clone https://github.com/open-compass/opencompass opencompass
-cd opencompass
-pip install -e .
-```
-
-#### 面向API模型测试的CPU环境
-
-```bash
-conda create -n opencompass python=3.10 pytorch torchvision torchaudio cpuonly -c pytorch -y
-conda activate opencompass
-git clone https://github.com/open-compass/opencompass opencompass
-cd opencompass
-pip install -e .
-# 如果需要使用各个API模型，请 `pip install -r requirements/api.txt` 安装API模型的相关依赖
-```
-
-### 📂 数据准备
-
-```bash
-# 下载数据集到 data/ 处
-wget https://github.com/open-compass/opencompass/releases/download/0.2.2.rc1/OpenCompassData-core-20240207.zip
-unzip OpenCompassData-core-20240207.zip
-```
-
-有部分第三方功能,如 Humaneval 以及 Llama,可能需要额外步骤才能正常运行，详细步骤请参考[安装指南](https://opencompass.readthedocs.io/zh_CN/latest/get_started/installation.html)。
-
-<p align="right"><a href="#top">🔝返回顶部</a></p>
-
-## 🏗️ ️评测
-
-确保按照上述步骤正确安装 OpenCompass 并准备好数据集后，可以通过以下命令评测 LLaMA-7b 模型在 MMLU 和 C-Eval 数据集上的性能：
-
-```bash
-python run.py --models hf_llama_7b --datasets mmlu_ppl ceval_ppl
-```
-
-OpenCompass 预定义了许多模型和数据集的配置，你可以通过 [工具](./docs/zh_cn/tools.md#ListConfigs) 列出所有可用的模型和数据集配置。
-
-```bash
-# 列出所有配置
-python tools/list_configs.py
-# 列出所有跟 llama 及 mmlu 相关的配置
-python tools/list_configs.py llama mmlu
-```
-
-你也可以通过命令行去评测其它 HuggingFace 模型。同样以 LLaMA-7b 为例：
-
-```bash
-python run.py --datasets ceval_ppl mmlu_ppl --hf-type base --hf-path huggyllama/llama-7b
-```
-
-通过命令行或配置文件，OpenCompass 还支持评测 API 或自定义模型，以及更多样化的评测策略。请阅读[快速开始](https://opencompass.readthedocs.io/zh_CN/latest/get_started/quick_start.html)了解如何运行一个评测任务。
-
-更多教程请查看我们的[文档](https://opencompass.readthedocs.io/zh_CN/latest/index.html)。
-
-<p align="right"><a href="#top">🔝返回顶部</a></p>
-
 ## 📖 数据集支持
 
-<table align="center">
-  <tbody>
-    <tr align="center" valign="bottom">
-      <td>
-        <b>语言</b>
-      </td>
-      <td>
-        <b>知识</b>
-      </td>
-      <td>
-        <b>推理</b>
-      </td>
-      <td>
-        <b>考试</b>
-      </td>
-    </tr>
-    <tr valign="top">
-      <td>
-<details open>
-<summary><b>字词释义</b></summary>
+我们已经在OpenCompass官网的文档中支持了所有可在本平台上使用的数据集的统计列表。
 
-- WiC
-- SummEdits
+您可以通过排序、筛选和搜索等功能从列表中快速找到您需要的数据集。
 
-</details>
-
-<details open>
-<summary><b>成语习语</b></summary>
-
-- CHID
-
-</details>
-
-<details open>
-<summary><b>语义相似度</b></summary>
-
-- AFQMC
-- BUSTM
-
-</details>
-
-<details open>
-<summary><b>指代消解</b></summary>
-
-- CLUEWSC
-- WSC
-- WinoGrande
-
-</details>
-
-<details open>
-<summary><b>翻译</b></summary>
-
-- Flores
-- IWSLT2017
-
-</details>
-
-<details open>
-<summary><b>多语种问答</b></summary>
-
-- TyDi-QA
-- XCOPA
-
-</details>
-
-<details open>
-<summary><b>多语种总结</b></summary>
-
-- XLSum
-
-</details>
-      </td>
-      <td>
-<details open>
-<summary><b>知识问答</b></summary>
-
-- BoolQ
-- CommonSenseQA
-- NaturalQuestions
-- TriviaQA
-
-</details>
-      </td>
-      <td>
-<details open>
-<summary><b>文本蕴含</b></summary>
-
-- CMNLI
-- OCNLI
-- OCNLI_FC
-- AX-b
-- AX-g
-- CB
-- RTE
-- ANLI
-
-</details>
-
-<details open>
-<summary><b>常识推理</b></summary>
-
-- StoryCloze
-- COPA
-- ReCoRD
-- HellaSwag
-- PIQA
-- SIQA
-
-</details>
-
-<details open>
-<summary><b>数学推理</b></summary>
-
-- MATH
-- GSM8K
-
-</details>
-
-<details open>
-<summary><b>定理应用</b></summary>
-
-- TheoremQA
-- StrategyQA
-- SciBench
-
-</details>
-
-<details open>
-<summary><b>综合推理</b></summary>
-
-- BBH
-
-</details>
-      </td>
-      <td>
-<details open>
-<summary><b>初中/高中/大学/职业考试</b></summary>
-
-- C-Eval
-- AGIEval
-- MMLU
-- GAOKAO-Bench
-- CMMLU
-- ARC
-- Xiezhi
-
-</details>
-
-<details open>
-<summary><b>医学考试</b></summary>
-
-- CMB
-
-</details>
-      </td>
-    </tr>
-</td>
-    </tr>
-  </tbody>
-  <tbody>
-    <tr align="center" valign="bottom">
-      <td>
-        <b>理解</b>
-      </td>
-      <td>
-        <b>长文本</b>
-      </td>
-      <td>
-        <b>安全</b>
-      </td>
-      <td>
-        <b>代码</b>
-      </td>
-    </tr>
-    <tr valign="top">
-      <td>
-<details open>
-<summary><b>阅读理解</b></summary>
-
-- C3
-- CMRC
-- DRCD
-- MultiRC
-- RACE
-- DROP
-- OpenBookQA
-- SQuAD2.0
-
-</details>
-
-<details open>
-<summary><b>内容总结</b></summary>
-
-- CSL
-- LCSTS
-- XSum
-- SummScreen
-
-</details>
-
-<details open>
-<summary><b>内容分析</b></summary>
-
-- EPRSTMT
-- LAMBADA
-- TNEWS
-
-</details>
-      </td>
-      <td>
-<details open>
-<summary><b>长文本理解</b></summary>
-
-- LEval
-- LongBench
-- GovReports
-- NarrativeQA
-- Qasper
-
-</details>
-      </td>
-      <td>
-<details open>
-<summary><b>安全</b></summary>
-
-- CivilComments
-- CrowsPairs
-- CValues
-- JigsawMultilingual
-- TruthfulQA
-
-</details>
-<details open>
-<summary><b>健壮性</b></summary>
-
-- AdvGLUE
-
-</details>
-      </td>
-      <td>
-<details open>
-<summary><b>代码</b></summary>
-
-- HumanEval
-- HumanEvalX
-- MBPP
-- APPs
-- DS1000
-
-</details>
-      </td>
-    </tr>
-</td>
-    </tr>
-  </tbody>
-</table>
+详情请参阅 [官方文档](https://opencompass.readthedocs.io/zh-cn/latest/dataset_statistics.html) 的数据集统计章节。
 
 <p align="right"><a href="#top">🔝返回顶部</a></p>
 
@@ -452,19 +318,20 @@ python run.py --datasets ceval_ppl mmlu_ppl --hf-type base --hf-path huggyllama/
     <tr valign="top">
       <td>
 
+- [Alpaca](https://github.com/tatsu-lab/stanford_alpaca)
+- [Baichuan](https://github.com/baichuan-inc)
+- [BlueLM](https://github.com/vivo-ai-lab/BlueLM)
+- [ChatGLM2](https://github.com/THUDM/ChatGLM2-6B)
+- [ChatGLM3](https://github.com/THUDM/ChatGLM3-6B)
+- [Gemma](https://huggingface.co/google/gemma-7b)
 - [InternLM](https://github.com/InternLM/InternLM)
 - [LLaMA](https://github.com/facebookresearch/llama)
 - [LLaMA3](https://github.com/meta-llama/llama3)
-- [Vicuna](https://github.com/lm-sys/FastChat)
-- [Alpaca](https://github.com/tatsu-lab/stanford_alpaca)
-- [Baichuan](https://github.com/baichuan-inc)
-- [WizardLM](https://github.com/nlpxucan/WizardLM)
-- [ChatGLM2](https://github.com/THUDM/ChatGLM2-6B)
-- [ChatGLM3](https://github.com/THUDM/ChatGLM3-6B)
-- [TigerBot](https://github.com/TigerResearch/TigerBot)
 - [Qwen](https://github.com/QwenLM/Qwen)
-- [BlueLM](https://github.com/vivo-ai-lab/BlueLM)
-- [Gemma](https://huggingface.co/google/gemma-7b)
+- [TigerBot](https://github.com/TigerResearch/TigerBot)
+- [Vicuna](https://github.com/lm-sys/FastChat)
+- [WizardLM](https://github.com/nlpxucan/WizardLM)
+- [Yi](https://github.com/01-ai/Yi)
 - ……
 
 </td>
@@ -496,7 +363,7 @@ python run.py --datasets ceval_ppl mmlu_ppl --hf-type base --hf-path huggyllama/
 
 - [x] 主观评测
   - [x] 发布主观评测榜单
-  - [ ] 发布主观评测数据集
+  - [x] 发布主观评测数据集
 - [x] 长文本
   - [x] 支持广泛的长文本评测集
   - [ ] 发布长文本评测榜单

configs/api_examples/eval_api_360.py

@@ -1,43 +0,0 @@
-from mmengine.config import read_base
-from opencompass.models import AI360GPT
-from opencompass.partitioners import NaivePartitioner
-from opencompass.runners.local_api import LocalAPIRunner
-from opencompass.tasks import OpenICLInferTask
-
-with read_base():
-    from ..summarizers.medium import summarizer
-    from ..datasets.ceval.ceval_gen import ceval_datasets
-
-datasets = [
-    *ceval_datasets,
-]
-
-models = [
-    dict(
-        abbr='360GPT_S2_V9',
-        type=AI360GPT,
-        path='360GPT_S2_V9',
-        key='xxxxxxxxxxxx',
-        generation_kwargs={
-            'temperature': 0.9,
-            'max_tokens': 2048,
-            'top_p': 0.5,
-            'tok_k': 0,
-            'repetition_penalty': 1.05,
-        },
-        query_per_second=1,
-        max_out_len=2048,
-        max_seq_len=2048,
-        batch_size=8),
-]
-
-infer = dict(
-    partitioner=dict(type=NaivePartitioner),
-    runner=dict(
-        type=LocalAPIRunner,
-        max_num_workers=2,
-        concurrent_users=2,
-        task=dict(type=OpenICLInferTask)),
-)
-
-work_dir ='./output/api_360GPT_S2_V9'

configs/api_examples/eval_api_baichuan.py

@@ -1,44 +0,0 @@
-from mmengine.config import read_base
-from opencompass.models import BaiChuan
-from opencompass.partitioners import NaivePartitioner
-from opencompass.runners.local_api import LocalAPIRunner
-from opencompass.tasks import OpenICLInferTask
-
-with read_base():
-    from ..summarizers.medium import summarizer
-    from ..datasets.ceval.ceval_gen import ceval_datasets
-
-datasets = [
-    *ceval_datasets,
-]
-
-models = [
-    dict(
-        abbr='Baichuan2-53B',
-        type=BaiChuan,
-        path='Baichuan2-53B',
-        api_key='xxxxxx',
-        secret_key='xxxxx',
-        url='xxxxx',
-        generation_kwargs={
-            'temperature': 0.3,
-            'top_p': 0.85,
-            'top_k': 5,
-            'with_search_enhance': False,
-        },
-        query_per_second=1,
-        max_out_len=2048,
-        max_seq_len=2048,
-        batch_size=8),
-]
-
-infer = dict(
-    partitioner=dict(type=NaivePartitioner),
-    runner=dict(
-        type=LocalAPIRunner,
-        max_num_workers=2,
-        concurrent_users=2,
-        task=dict(type=OpenICLInferTask)),
-)
-
-work_dir = 'outputs/api_baichuan53b/'

configs/api_examples/eval_api_baidu.py

@@ -1,42 +0,0 @@
-from mmengine.config import read_base
-from opencompass.models import ERNIEBot
-from opencompass.partitioners import NaivePartitioner
-from opencompass.runners.local_api import LocalAPIRunner
-from opencompass.tasks import OpenICLInferTask
-
-with read_base():
-    from ..summarizers.medium import summarizer
-    from ..datasets.ceval.ceval_gen import ceval_datasets
-
-datasets = [
-    *ceval_datasets,
-]
-
-models = [
-    dict(
-        abbr='erniebot',
-        type=ERNIEBot,
-        path='erniebot',
-        key='xxxxxx',  # please give you key
-        secretkey='xxxxxxxxx',  # please give your group_id
-        url='xxxxxxxxx',
-        generation_kwargs = {
-            'temperature': 0.8,
-        },
-        query_per_second=1,
-        max_out_len=2048,
-        max_seq_len=2048,
-        batch_size=8
-    ),
-]
-
-infer = dict(
-    partitioner=dict(type=NaivePartitioner),
-    runner=dict(
-        type=LocalAPIRunner,
-        max_num_workers=2,
-        concurrent_users=2,
-        task=dict(type=OpenICLInferTask)),
-)
-
-work_dir = 'outputs/api_erniebot/'

configs/api_examples/eval_api_bytedance.py

@@ -1,44 +0,0 @@
-from mmengine.config import read_base
-from opencompass.models import ByteDance
-from opencompass.partitioners import NaivePartitioner
-from opencompass.runners.local_api import LocalAPIRunner
-from opencompass.tasks import OpenICLInferTask
-
-with read_base():
-    # from .datasets.collections.chat_medium import datasets
-    from ..summarizers.medium import summarizer
-    from ..datasets.ceval.ceval_gen import ceval_datasets
-
-datasets = [
-    *ceval_datasets,
-]
-
-models = [
-    dict(
-        abbr='skylark-pro-public',
-        type=ByteDance,
-        path='skylark-pro-public',
-        accesskey='xxxxxxx',
-        secretkey='xxxxxxx',
-        url='xxxxxx',
-        generation_kwargs={
-            'temperature': 0.7,
-            'top_p': 0.9,
-            'top_k': 0,
-        },
-        query_per_second=1,
-        max_out_len=2048,
-        max_seq_len=2048,
-        batch_size=8),
-]
-
-infer = dict(
-    partitioner=dict(type=NaivePartitioner),
-    runner=dict(
-        type=LocalAPIRunner,
-        max_num_workers=2,
-        concurrent_users=2,
-        task=dict(type=OpenICLInferTask)),
-)
-
-work_dir = 'outputs/api_bytedance/'

configs/api_examples/eval_api_minimax.py

@@ -1,37 +0,0 @@
-from mmengine.config import read_base
-from opencompass.models import MiniMax
-from opencompass.partitioners import NaivePartitioner
-from opencompass.runners.local_api import LocalAPIRunner
-from opencompass.tasks import OpenICLInferTask
-
-with read_base():
-    from ..summarizers.medium import summarizer
-    from ..datasets.ceval.ceval_gen import ceval_datasets
-
-datasets = [
-    *ceval_datasets,
-]
-
-models = [
-    dict(
-        abbr='minimax_abab5.5-chat',
-        type=MiniMax,
-        path='abab5.5-chat',
-        key='xxxxxxx', # please give you key
-        group_id='xxxxxxxx', # please give your group_id
-        query_per_second=1,
-        max_out_len=2048,
-        max_seq_len=2048,
-        batch_size=8),
-]
-
-infer = dict(
-    partitioner=dict(type=NaivePartitioner),
-    runner=dict(
-        type=LocalAPIRunner,
-        max_num_workers=4,
-        concurrent_users=4,
-        task=dict(type=OpenICLInferTask)),
-)
-
-work_dir = 'outputs/api_minimax/'

configs/api_examples/eval_api_moonshot.py

@@ -1,40 +0,0 @@
-from mmengine.config import read_base
-from opencompass.models import MoonShot
-from opencompass.partitioners import NaivePartitioner
-from opencompass.runners.local_api import LocalAPIRunner
-from opencompass.tasks import OpenICLInferTask
-
-with read_base():
-    from ..summarizers.medium import summarizer
-    from ..datasets.ceval.ceval_gen import ceval_datasets
-
-datasets = [
-    *ceval_datasets,
-]
-
-models = [
-    dict(
-        abbr='moonshot-v1-32k',
-        type=MoonShot,
-        path='moonshot-v1-32k',
-        key='xxxxxxx',
-        url= 'xxxxxxxx',
-        system_prompt= '你是 Kimi，由 Moonshot AI 提供的人工智能助手，你更擅长中文和英文的对话。'
-        '你会为用户提供安全，有帮助，准确的回答。同时，你会拒绝一些涉及恐怖主义，种族歧视，'
-        '黄色暴力等问题的回答。Moonshot AI 为专有名词，不可翻译成其他语言。',
-        query_per_second=1,
-        max_out_len=2048,
-        max_seq_len=2048,
-        batch_size=8),
-]
-
-infer = dict(
-    partitioner=dict(type=NaivePartitioner),
-    runner=dict(
-        type=LocalAPIRunner,
-        max_num_workers=4,
-        concurrent_users=4,
-        task=dict(type=OpenICLInferTask)),
-)
-
-work_dir = 'outputs/api_moonshot/'

configs/api_examples/eval_api_nanbeige.py

@@ -1,36 +0,0 @@
-from mmengine.config import read_base
-from opencompass.models import Nanbeige
-from opencompass.partitioners import NaivePartitioner
-from opencompass.runners.local_api import LocalAPIRunner
-from opencompass.tasks import OpenICLInferTask
-
-
-with read_base():
-    from ..summarizers.medium import summarizer
-    from ..datasets.ceval.ceval_gen import ceval_datasets
-
-datasets = [
-    *ceval_datasets,
-]
-
-models = [
-    dict(
-        abbr='nanbeige-plus',
-        type=Nanbeige,
-        path='nanbeige-plus',
-        key='xxxxxx',
-        query_per_second=1,
-        max_out_len=2048,
-        batch_size=8),
-]
-
-infer = dict(
-    partitioner=dict(type=NaivePartitioner),
-    runner=dict(
-        type=LocalAPIRunner,
-        max_num_workers=2,
-        concurrent_users=2,
-        task=dict(type=OpenICLInferTask)),
-)
-
-work_dir ='./output/nanbeige-plus'

configs/api_examples/eval_api_pangu.py

@@ -1,42 +0,0 @@
-from mmengine.config import read_base
-from opencompass.models import PanGu
-from opencompass.partitioners import NaivePartitioner
-from opencompass.runners.local_api import LocalAPIRunner
-from opencompass.tasks import OpenICLInferTask
-
-with read_base():
-    from ..summarizers.medium import summarizer
-    from ..datasets.ceval.ceval_gen import ceval_datasets
-
-datasets = [
-    *ceval_datasets,
-]
-
-models = [
-dict(
-        abbr='pangu',
-        type=PanGu,
-        path='pangu',
-        access_key='xxxxxx',
-        secret_key='xxxxxx',
-        url = 'xxxxxx',
-        # url of token sever, used for generate token, like "https://xxxxxx.myhuaweicloud.com/v3/auth/tokens",
-        token_url = 'xxxxxx',
-        # scope-project-name, used for generate token
-        project_name = 'xxxxxx',
-        query_per_second=1,
-        max_out_len=2048,
-        max_seq_len=2048,
-        batch_size=8),
-]
-
-infer = dict(
-    partitioner=dict(type=NaivePartitioner),
-    runner=dict(
-        type=LocalAPIRunner,
-        max_num_workers=2,
-        concurrent_users=2,
-        task=dict(type=OpenICLInferTask)),
-)
-
-work_dir = 'outputs/api_pangu/'

configs/api_examples/eval_api_qwen.py

@@ -1,40 +0,0 @@
-from mmengine.config import read_base
-from opencompass.models import Qwen
-from opencompass.partitioners import NaivePartitioner
-from opencompass.runners.local_api import LocalAPIRunner
-from opencompass.tasks import OpenICLInferTask
-
-with read_base():
-    from ..summarizers.medium import summarizer
-    from ..datasets.ceval.ceval_gen import ceval_datasets
-
-datasets = [
-    *ceval_datasets,
-]
-
-models = [
-    dict(
-        abbr='qwen-max',
-        type=Qwen,
-        path='qwen-max',
-        key='xxxxxxxxxxxxxxxx',  # please give you key
-        generation_kwargs={
-            'enable_search': False,
-        },
-        query_per_second=1,
-        max_out_len=2048,
-        max_seq_len=2048,
-        batch_size=8
-    ),
-]
-
-infer = dict(
-    partitioner=dict(type=NaivePartitioner),
-    runner=dict(
-        type=LocalAPIRunner,
-        max_num_workers=1,
-        concurrent_users=1,
-        task=dict(type=OpenICLInferTask)),
-)
-
-work_dir = 'outputs/api_qwen/'

configs/api_examples/eval_api_sensetime.py

@@ -1,52 +0,0 @@
-from mmengine.config import read_base
-from opencompass.models import SenseTime
-from opencompass.partitioners import NaivePartitioner
-from opencompass.runners.local_api import LocalAPIRunner
-from opencompass.tasks import OpenICLInferTask
-
-with read_base():
-    from ..summarizers.medium import summarizer
-    from ..datasets.ceval.ceval_gen import ceval_datasets
-
-datasets = [
-    *ceval_datasets,
-]
-
-models = [
-    dict(
-        abbr='nova-ptc-xl-v1',
-        type=SenseTime,
-        path='nova-ptc-xl-v1',
-        key='xxxxxxxxxxxxxx',
-        url='xxxxxxxxxxx',
-        query_per_second=1,
-        max_out_len=2048,
-        max_seq_len=2048,
-        batch_size=8,
-        parameters={
-            'temperature': 0.8,
-            'top_p': 0.7,
-            'max_new_tokens': 1024,
-            'repetition_penalty': 1.05,
-            'know_ids': [],
-            'stream': True,
-            'user': '#*#***TestUser***#*#',
-            'knowledge_config': {
-                'control_level': 'normal',
-                'knowledge_base_result': False,
-                'online_search_result': False
-            }
-        }
-    )
-]
-
-infer = dict(
-    partitioner=dict(type=NaivePartitioner),
-    runner=dict(
-        type=LocalAPIRunner,
-        max_num_workers=2,
-        concurrent_users=2,
-        task=dict(type=OpenICLInferTask)),
-)
-
-work_dir = 'outputs/api_sensetime/'

configs/api_examples/eval_api_xunfei.py

@@ -1,51 +0,0 @@
-from mmengine.config import read_base
-from opencompass.models.xunfei_api import XunFei
-from opencompass.partitioners import NaivePartitioner
-from opencompass.runners.local_api import LocalAPIRunner
-from opencompass.tasks import OpenICLInferTask
-
-with read_base():
-    # from .datasets.collections.chat_medium import datasets
-    from ..summarizers.medium import summarizer
-    from ..datasets.ceval.ceval_gen import ceval_datasets
-
-datasets = [
-    *ceval_datasets,
-]
-
-models = [
-    dict(
-        abbr='Spark-v1-1',
-        type=XunFei,
-        appid='xxxx',
-        path='ws://spark-api.xf-yun.com/v1.1/chat',
-        api_secret = 'xxxxxxx',
-        api_key = 'xxxxxxx',
-        query_per_second=1,
-        max_out_len=2048,
-        max_seq_len=2048,
-        batch_size=8),
-    dict(
-        abbr='Spark-v3-1',
-        type=XunFei,
-        appid='xxxx',
-        domain='generalv3',
-        path='ws://spark-api.xf-yun.com/v3.1/chat',
-        api_secret = 'xxxxxxxx',
-        api_key = 'xxxxxxxxx',
-        query_per_second=1,
-        max_out_len=2048,
-        max_seq_len=2048,
-        batch_size=8),
-]
-
-infer = dict(
-    partitioner=dict(type=NaivePartitioner),
-    runner=dict(
-        type=LocalAPIRunner,
-        max_num_workers=2,
-        concurrent_users=2,
-        task=dict(type=OpenICLInferTask)),
-)
-
-work_dir = 'outputs/api_xunfei/'

configs/api_examples/eval_api_zhipu.py

@@ -1,48 +0,0 @@
-from mmengine.config import read_base
-from opencompass.models import ZhiPuAI
-from opencompass.partitioners import NaivePartitioner
-from opencompass.runners.local_api import LocalAPIRunner
-from opencompass.tasks import OpenICLInferTask
-
-with read_base():
-    # from .datasets.collections.chat_medium import datasets
-    from ..summarizers.medium import summarizer
-    from ..datasets.ceval.ceval_gen import ceval_datasets
-
-datasets = [
-    *ceval_datasets,
-]
-
-# needs a special postprocessor for all
-# except 'gsm8k' and 'strategyqa'
-from opencompass.utils import general_eval_wrapper_postprocess
-for _dataset in datasets:
-    if _dataset['abbr'] not in ['gsm8k', 'strategyqa']:
-        if hasattr(_dataset['eval_cfg'], 'pred_postprocessor'):
-            _dataset['eval_cfg']['pred_postprocessor']['postprocess'] = _dataset['eval_cfg']['pred_postprocessor']['type']
-            _dataset['eval_cfg']['pred_postprocessor']['type'] = general_eval_wrapper_postprocess
-        else:
-            _dataset['eval_cfg']['pred_postprocessor'] = {'type': general_eval_wrapper_postprocess}
-
-models = [
-    dict(
-        abbr='chatglm_pro',
-        type=ZhiPuAI,
-        path='chatglm_pro',
-        key='xxxxxxxxxxxx',
-        query_per_second=1,
-        max_out_len=2048,
-        max_seq_len=2048,
-        batch_size=8),
-]
-
-infer = dict(
-    partitioner=dict(type=NaivePartitioner),
-    runner=dict(
-        type=LocalAPIRunner,
-        max_num_workers=2,
-        concurrent_users=2,
-        task=dict(type=OpenICLInferTask)),
-)
-
-work_dir = 'outputs/api_zhipu/'

configs/api_examples/eval_api_zhipu_v2.py

@@ -1,67 +0,0 @@
-from mmengine.config import read_base
-from opencompass.models import ZhiPuV2AI
-from opencompass.partitioners import NaivePartitioner
-from opencompass.runners.local_api import LocalAPIRunner
-from opencompass.tasks import OpenICLInferTask
-
-with read_base():
-    # from .datasets.collections.chat_medium import datasets
-    from ..summarizers.medium import summarizer
-    from ..datasets.ceval.ceval_gen import ceval_datasets
-
-datasets = [
-    *ceval_datasets,
-]
-
-# needs a special postprocessor for all
-# except 'gsm8k' and 'strategyqa'
-from opencompass.utils import general_eval_wrapper_postprocess
-for _dataset in datasets:
-    if _dataset['abbr'] not in ['gsm8k', 'strategyqa']:
-        if hasattr(_dataset['eval_cfg'], 'pred_postprocessor'):
-            _dataset['eval_cfg']['pred_postprocessor']['postprocess'] = _dataset['eval_cfg']['pred_postprocessor']['type']
-            _dataset['eval_cfg']['pred_postprocessor']['type'] = general_eval_wrapper_postprocess
-        else:
-            _dataset['eval_cfg']['pred_postprocessor'] = {'type': general_eval_wrapper_postprocess}
-
-
-api_meta_template = dict(
-    round=[
-            dict(role='HUMAN', api_role='HUMAN'),
-            dict(role='BOT', api_role='BOT', generate=True),
-    ],
-)
-
-models = [
-     dict(
-        abbr='glm4_notools',
-        type=ZhiPuV2AI,
-        path='glm-4',
-        key='xxxxxx',
-        generation_kwargs={
-            'tools': [
-                {
-                    'type': 'web_search',
-                    'web_search': {
-                        'enable': False # turn off the search
-                    }
-                }
-            ]
-        },
-        meta_template=api_meta_template,
-        query_per_second=1,
-        max_out_len=2048,
-        max_seq_len=2048,
-        batch_size=8)
-]
-
-infer = dict(
-    partitioner=dict(type=NaivePartitioner),
-    runner=dict(
-        type=LocalAPIRunner,
-        max_num_workers=2,
-        concurrent_users=2,
-        task=dict(type=OpenICLInferTask)),
-)
-
-work_dir = 'outputs/api_zhipu_v2/'

configs/dataset_collections/chat_OC15.py

@@ -1,22 +0,0 @@
-from mmengine.config import read_base
-
-with read_base():
-    from ..datasets.mmlu.mmlu_gen_4d595a import mmlu_datasets
-    from ..datasets.cmmlu.cmmlu_gen_c13365 import cmmlu_datasets
-    from ..datasets.ceval.ceval_gen_5f30c7 import ceval_datasets
-    from ..datasets.GaokaoBench.GaokaoBench_no_subjective_gen_4c31db import GaokaoBench_datasets
-    from ..datasets.triviaqa.triviaqa_wiki_1shot_gen_eaf81e import triviaqa_datasets
-    from ..datasets.nq.nq_open_1shot_gen_01cf41 import nq_datasets
-    from ..datasets.race.race_gen_69ee4f import race_datasets
-    from ..datasets.winogrande.winogrande_5shot_gen_b36770 import winogrande_datasets
-    from ..datasets.hellaswag.hellaswag_10shot_gen_e42710 import hellaswag_datasets
-    from ..datasets.bbh.bbh_gen_2879b0 import bbh_datasets
-    from ..datasets.gsm8k.gsm8k_gen_1d7fe4 import gsm8k_datasets
-    from ..datasets.math.math_0shot_gen_393424 import math_datasets
-    from ..datasets.TheoremQA.TheoremQA_5shot_gen_6f0af8 import TheoremQA_datasets
-    from ..datasets.humaneval.humaneval_gen_8e312c import humaneval_datasets
-    from ..datasets.mbpp.sanitized_mbpp_gen_830460 import sanitized_mbpp_datasets
-    from ..datasets.gpqa.gpqa_gen_4baadb import gpqa_datasets
-    from ..datasets.IFEval.IFEval_gen_3321a3 import ifeval_datasets
-
-datasets = sum((v for k, v in locals().items() if k.endswith('_datasets')), [])

configs/datasets/IFEval/IFEval_gen.py

@@ -1,4 +0,0 @@
-from mmengine.config import read_base
-
-with read_base():
-    from .IFEval_gen_3321a3 import ifeval_datasets  # noqa: F401, F403

configs/datasets/bbh/bbh_gen.py

@@ -1,4 +0,0 @@
-from mmengine.config import read_base
-
-with read_base():
-    from .bbh_gen_5b92b0 import bbh_datasets  # noqa: F401, F403

configs/datasets/cmmlu/cmmlu_gen.py

@@ -1,4 +0,0 @@
-from mmengine.config import read_base
-
-with read_base():
-    from .cmmlu_gen_c13365 import cmmlu_datasets  # noqa: F401, F403

configs/datasets/hellaswag/hellaswag_gen.py

@@ -1,4 +0,0 @@
-from mmengine.config import read_base
-
-with read_base():
-    from .hellaswag_gen_6faab5 import hellaswag_datasets  # noqa: F401, F403

configs/datasets/humaneval/humaneval_gen.py

@@ -1,4 +0,0 @@
-from mmengine.config import read_base
-
-with read_base():
-    from .humaneval_gen_8e312c import humaneval_datasets  # noqa: F401, F403

configs/datasets/humaneval/humaneval_gen_6f294d.py

@@ -1,41 +0,0 @@
-from opencompass.openicl.icl_prompt_template import PromptTemplate
-from opencompass.openicl.icl_retriever import ZeroRetriever
-from opencompass.openicl.icl_inferencer import GenInferencer
-from opencompass.datasets import HumanevalDataset, HumanEvaluator, humaneval_postprocess
-
-humaneval_reader_cfg = dict(
-    input_columns=['prompt'], output_column='task_id', train_split='test')
-
-# TODO: allow empty output-column
-humaneval_infer_cfg = dict(
-    prompt_template=dict(
-        type=PromptTemplate,
-        template=dict(
-            begin=[
-                dict(
-                    role='SYSTEM',
-                    fallback_role='HUMAN',
-                    prompt='Complete the following python code:'),
-            ],
-            round=[
-                dict(role='HUMAN', prompt='{prompt}'),
-            ])),
-    retriever=dict(type=ZeroRetriever),
-    inferencer=dict(type=GenInferencer, max_out_len=512))
-
-humaneval_eval_cfg = dict(
-    evaluator=dict(type=HumanEvaluator),
-    pred_role='BOT',
-    k=[1, 10, 100],  # the parameter only for humaneval
-    pred_postprocessor=dict(type=humaneval_postprocess),
-)
-
-humaneval_datasets = [
-    dict(
-        abbr='openai_humaneval',
-        type=HumanevalDataset,
-        path='./data/humaneval/human-eval-v2-20210705.jsonl',
-        reader_cfg=humaneval_reader_cfg,
-        infer_cfg=humaneval_infer_cfg,
-        eval_cfg=humaneval_eval_cfg)
-]

configs/datasets/math/math_gen.py

@@ -1,4 +0,0 @@
-from mmengine.config import read_base
-
-with read_base():
-    from .math_gen_265cce import math_datasets  # noqa: F401, F403

configs/datasets/math/math_llm_judge.py

@@ -1,35 +0,0 @@
-from opencompass.openicl.icl_prompt_template import PromptTemplate
-from opencompass.openicl.icl_retriever import ZeroRetriever
-from opencompass.openicl.icl_inferencer import GenInferencer
-from opencompass.datasets import MATHDataset, MATHEvaluator, math_postprocess
-
-QUERY_TEMPLATE = """
-Solve the following math problem step by step. The last line of your response should be of the form ANSWER: $ANSWER (without quotes) where $ANSWER is the answer to the problem.
-{problem}
-Remember to put your answer on its own line after "ANSWER:", and you do not need to use a \\boxed command.
-""".strip()
-
-math_reader_cfg = dict(input_columns=['problem'], output_column='solution')
-
-math_infer_cfg = dict(
-    prompt_template=dict(
-        type=PromptTemplate,
-
-        template=dict(round=[
-            dict(role='HUMAN', prompt=QUERY_TEMPLATE),
-        ])),
-    retriever=dict(type=ZeroRetriever),
-    inferencer=dict(type=GenInferencer, max_out_len=1024))
-
-math_eval_cfg = dict(
-    evaluator=dict(type=MATHEvaluator), pred_postprocessor=dict(type=math_postprocess))
-
-math_datasets = [
-    dict(
-        type=MATHDataset,
-        abbr='math',
-        path='./data/math/math.json',
-        reader_cfg=math_reader_cfg,
-        infer_cfg=math_infer_cfg,
-        eval_cfg=math_eval_cfg)
-]

configs/datasets/mmlu/mmlu_gen.py

@@ -1,4 +0,0 @@
-from mmengine.config import read_base
-
-with read_base():
-    from .mmlu_gen_4d595a import mmlu_datasets  # noqa: F401, F403

configs/datasets/needlebench/atc/atc_choice_50_en_reasoning.py

@@ -1,96 +0,0 @@
-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 CircularEvaluator
-from opencompass.datasets.needlebench.atc_choice import NeedleBenchATCDataset
-from opencompass.utils.text_postprocessors import first_option_postprocess
-
-# ----------------------- Prompt Content----------------------- #
-
-few_shot_prompts = {
-    'single_choice_prompts': {
-        'single_choice_en_reasoning': [
-            dict(role='HUMAN', prompt="Question: Sharon House, as Jessica Stewart's father, has a significant impact on Jessica Stewart's upbringing. \nGiven the scrambled family relationships described above, who is the eldest relative that 'Jessica Stewart' can trace back to in the context?\nA. Jack Burch\nB. Jessica Stewart\nC. Sharon House\nD. Carolyn Jackson\n"),
-            dict(role='BOT', prompt="Answer: Based on the provided information, we can construct the following family relationship chain：\n\n- Sharon House, as Jessica Stewart's father, has a significant impact on Jessica Stewart's upbringing.\n\nTherefore, the eldest relative that 'Jessica Stewart' can trace back to in the context is Sharon House. The answer is: C"),
-            dict(role='HUMAN', prompt="Question: For Robert Hill, Mikayla Scott is not just a paternal grandfather, but also a friend.Jacob Oconnor's paternal grandmother is Robert Hill. \nGiven the scrambled family relationships described above, who is the eldest relative that 'Jacob Oconnor' can trace back to in the context?\nA. Laura Holland\nB. Robert Hill\nC. Jacob Oconnor\nD. Mikayla Scott\n"),
-            dict(role='BOT', prompt="Answer: Based on the provided information, we can construct the following family relationship chain：\n\n-Jacob Oconnor's paternal grandmother is Robert Hill. \n- For Robert Hill, Mikayla Scott is not just a paternal grandfather, but also a friend.\n\nTherefore, the eldest relative that 'Jacob Oconnor' can trace back to in the context is Mikayla Scott. The answer is: D"),
-            dict(role='HUMAN', prompt="Question: Misty Moore plays the role of Barbara Fuentes's maternal grandfather in Barbara Fuentes's life.Jennifer Garcia, as Michael Martinez's grandmother, has a significant impact on Michael Martinez's upbringing.Michael Martinez is not only Misty Moore's father but also Misty Moore's role model. \nGiven the scrambled family relationships described above, who is the eldest relative that 'Barbara Fuentes' can trace back to in the context?\nA. Michael Martinez\nB. Jennifer Garcia\nC. Misty Moore\nD. Barbara Fuentes\n"),
-            dict(role='BOT', prompt="Answer: Based on the provided information, we can construct the following family relationship chain：\n- Misty Moore plays the role of Barbara Fuentes's maternal grandfather in Barbara Fuentes's life. \n- Michael Martinez is not only Misty Moore's father but also Misty Moore's role model.\n- Jennifer Garcia, as Michael Martinez's grandmother, has a significant impact on Michael Martinez's upbringing.\n\nTherefore, the eldest relative that 'Barbara Fuentes' can trace back to in the context is Jennifer Garcia. The answer is: B"),
-            dict(role='HUMAN', prompt="Question: Carlos Smith, as Mary Gay's grandfather, has a significant impact on Mary Gay's upbringing.Victor Dean considers Mary Gay as their grandfather.Marcus Miller, as Carlos Smith's paternal grandfather, has a significant impact on Carlos Smith's upbringing.Victor Dean is not only Danielle Yates's maternal grandmother but also Danielle Yates's role model.Danielle Yates is not only David Hernandez's paternal grandmother but also David Hernandez's role model.David Hernandez is Jennifer Williams's mom. \nGiven the scrambled family relationships described above, who is the eldest relative that 'Jennifer Williams' can trace back to in the context?\nA. Marcus Miller\nB. Carlos Smith\nC. Mary Gay\nD. Victor Dean\n"),
-            dict(role='BOT', prompt="Answer: Based on the provided information, we can construct the following family relationship chain：\n\n- David Hernandez is Jennifer Williams's mom.\n- Danielle Yates is not only David Hernandez's paternal grandmother but also David Hernandez's role model.\n- Victor Dean is not only Danielle Yates's maternal grandmother but also Danielle Yates's role model.\n- Victor Dean considers Mary Gay as their grandfather. \n- Carlos Smith, as Mary Gay's grandfather, has a significant impact on Mary Gay's upbringing.\n- Marcus Miller, as Carlos Smith's paternal grandfather, has a significant impact on Carlos Smith's upbringing.\n\nTherefore, the eldest relative that 'Jennifer Williams' can trace back to in the context is Marcus Miller. The answer is: A"),
-            dict(role='HUMAN', prompt='Question: {question}'),
-        ],
-    },
-}
-
-# ----------------------- Prompt Settings ----------------------- #
-needle_num_list = list(range(2, 50, 1))
-names_path = './data/needlebench/names.json'
-
-repeats = 10
-
-# Use Zero-Shot or not
-with_few_shot = True
-
-# Max for this dataset is 4, should be set with `with_few_shot`
-few_shot_samples = 4
-
-# Generate reasoning path or not, only for single choice
-with_reasoning = True
-
-# Use circular evaluation or not
-with_circular_eval = True
-
-needlebench_prompts = few_shot_prompts
-single_choice_prompts = needlebench_prompts['single_choice_prompts']
-
-# Set few shot prompt number
-for _name in list(single_choice_prompts.keys()):
-    if with_few_shot:
-        assert few_shot_samples > 0 and few_shot_samples <= 4
-        single_choice_prompts[_name] = \
-            single_choice_prompts[_name][- few_shot_samples * 2 - 1:]
-
-# ----------------------- Dataset Settings ----------------------- #
-
-needlebench_datasets = []
-
-
-needlebench_atc_reader_cfg = dict(input_columns=['question'],
-                                  output_column='answer')
-
-for _name in list(single_choice_prompts.keys()):
-
-    needlebench_atc_infer_cfg = dict(
-        prompt_template=dict(
-            type=PromptTemplate,
-            template=dict(
-                round=(single_choice_prompts[_name])),
-        ),
-        retriever=dict(type=ZeroRetriever),
-        inferencer=dict(type=GenInferencer,),
-    )
-
-    needlebench_atc_eval_cfg = dict(
-        evaluator=dict(type=CircularEvaluator),
-        pred_postprocessor=dict(type=first_option_postprocess, options='ABCD'))
-
-    for num_needles in needle_num_list:
-        abbr = (f'NeedleBenchATCDataset-'
-                f'{num_needles}Needle-{"EN" if "en" in _name else "ZH"}')
-        language = 'English' if 'en' in _name else 'Chinese'
-        if 'reasoning' in _name:
-            abbr += '-Reasoning'
-        dataset_dict = {
-            'abbr': abbr,
-            'type': NeedleBenchATCDataset,
-            'path': names_path,
-            'num_needles': num_needles,
-            'language': language,
-            'repeats': repeats,
-            'with_circular': with_circular_eval,
-            'reader_cfg': needlebench_atc_reader_cfg,
-            'infer_cfg': needlebench_atc_infer_cfg,
-            'eval_cfg': needlebench_atc_eval_cfg
-        }
-        needlebench_datasets.append(dataset_dict)

configs/datasets/s3eval/s3eval_gen.py

@@ -1,4 +0,0 @@
-from mmengine.config import read_base
-
-with read_base():
-    from .s3eval_gen_370cc2 import s3eval_datasets  # noqa: F401, F40

configs/datasets/subjective/alignbench/alignbench_judgeby_judgelm.py

@@ -1,59 +0,0 @@
-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 LMEvaluator
-from opencompass.datasets import AlignmentBenchDataset
-
-subjective_reader_cfg = dict(
-    input_columns=['question', 'capability', 'ref'],
-    output_column='judge',
-    )
-
-subjective_all_sets = [
-    'alignment_bench',
-]
-data_path ='data/subjective/alignment_bench'
-
-subjective_datasets = []
-
-for _name in subjective_all_sets:
-    subjective_infer_cfg = dict(
-            prompt_template=dict(
-                type=PromptTemplate,
-                template=dict(round=[
-                    dict(
-                        role='HUMAN',
-                        prompt='{question}'
-                    ),
-                ]),
-            ),
-            retriever=dict(type=ZeroRetriever),
-            inferencer=dict(type=GenInferencer, max_out_len=2048),
-        )
-
-    subjective_eval_cfg = dict(
-        evaluator=dict(
-            type=LMEvaluator,
-            prompt_template=dict(
-                type=PromptTemplate,
-                template=dict(round=[
-                    dict(
-                        role='HUMAN',
-                        prompt = """You are a helpful and precise assistant for checking the quality of the answer.\n[Question]\n{question}\n\n[The Start of Assistant 1's Answer]\n{ref}\n\n[The End of Assistant 1's Answer]\n\n[The Start of Assistant 2's Answer]\n{prediction}\n\n[The End of Assistant 2's Answer]\n\n[System]\nWe would like to request your feedback on the performance of two AI assistants in response to the user question displayed above.\nPlease rate the helpfulness, relevance, accuracy, level of details of their responses. Each assistant receives an overall score on a scale of 1 to 10, where a higher score indicates better overall performance.\nPlease first output a single line containing only two values indicating the scores for Assistant 1 and 2, respectively. The two scores are separated by a space. In the subsequent line, please provide a comprehensive explanation of your evaluation, avoiding any potential bias and ensuring that the order in which the responses were presented does not affect your judgment.\n\n### Response:10"""
-                    ),
-                ]),
-            ),
-        ),
-        pred_role='BOT',
-    )
-
-    subjective_datasets.append(
-        dict(
-            abbr=f'{_name}',
-            type=AlignmentBenchDataset,
-            path=data_path,
-            name=_name,
-            reader_cfg=subjective_reader_cfg,
-            infer_cfg=subjective_infer_cfg,
-            eval_cfg=subjective_eval_cfg
-        ))

configs/datasets/subjective/alpaca_eval/alpacav1_judgeby_gpt4.py

@@ -1,97 +0,0 @@
-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 LMEvaluator
-from opencompass.datasets import SubjectiveCmpDataset
-from mmengine.config import read_base
-
-subjective_reader_cfg = dict(
-    input_columns=['question'],
-    output_column='judge',
-    )
-
-subjective_all_sets = [
-    'alpaca_eval',
-]
-
-
-subjective_datasets = []
-
-gpt4_prompt = """
-I want you to create a leaderboard of different of large-language models. To do so, I will give you the instructions (prompts) given to the models, and the responses of two models. Please rank the models based on which responses would be preferred by humans. All inputs and outputs should be python dictionaries.
-
-Here is the prompt:
-{
-    "instruction": "{question}"
-}
-
-Here are the outputs of the models:
-[
-    {
-        "model": "model_1",
-        "answer": "{prediction}"
-    },
-    {
-        "model": "model_2",
-        "answer": "{prediction2}"
-    }
-]
-
-Now please rank the models by the quality of their answers, so that the model with rank 1 has the best output. Then return a list of the model names and ranks, i.e., produce the following output:
-[
-    {"model": <model-name>, "rank": <model-rank>},
-    {"model": <model-name>, "rank": <model-rank>}
-]
-
-Your response must be a valid Python dictionary and should contain nothing else because we will directly execute it in Python. Please provide the ranking that the majority of humans would give.
-"""
-
-
-for _name in subjective_all_sets:
-    subjective_infer_cfg = dict(
-            prompt_template=dict(
-                type=PromptTemplate,
-                template=dict(round=[
-                    dict(
-                        role='HUMAN',
-                        prompt='{question}'
-                    ),
-                ]),
-            ),
-            retriever=dict(type=ZeroRetriever),
-            inferencer=dict(type=GenInferencer, max_out_len=4096),
-        )
-
-    subjective_eval_cfg = dict(
-        evaluator=dict(
-            type=LMEvaluator,
-            prompt_template=dict(
-                type=PromptTemplate,
-                template=dict(
-                begin=[
-                    dict(
-                        role='SYSTEM',
-                        fallback_role='HUMAN',
-                        prompt='You are a helpful assistant, that ranks models by the quality of their answers.')
-                ],
-                    round=[
-                    dict(
-                        role='HUMAN',
-                        prompt = gpt4_prompt
-                    ),
-                ]),
-            ),
-        ),
-        pred_role='BOT',
-    )
-
-    subjective_datasets.append(
-        dict(
-            abbr=f'{_name}',
-            type=SubjectiveCmpDataset,
-            path='./data/subjective/alpaca_eval',
-            name=_name,
-            reader_cfg=subjective_reader_cfg,
-            infer_cfg=subjective_infer_cfg,
-            eval_cfg=subjective_eval_cfg
-        ))

configs/datasets/subjective/subjective_cmp/subjective_cmp.py

@@ -1,61 +0,0 @@
-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 LMEvaluator
-from opencompass.datasets.subjective_cmp import SubjectiveCmpDataset
-
-subjective_reader_cfg = dict(
-    input_columns=['question', 'index', 'reference_answer', 'evaluating_guidance', 'capability', 'prompt'],
-    output_column='judge',
-    train_split='test')
-
-subjective_all_sets = [
-    'creation_v0.1',
-]
-
-subjective_datasets = []
-
-for _name in subjective_all_sets:
-    subjective_infer_cfg = dict(
-            prompt_template=dict(
-                type=PromptTemplate,
-                template=dict(round=[
-                    dict(
-                        role='HUMAN',
-                        prompt='{question}'
-                    ),
-                ]),
-            ),
-            retriever=dict(type=ZeroRetriever),
-            inferencer=dict(type=GenInferencer, max_out_len=2048),
-        )
-
-    subjective_eval_cfg = dict(
-        evaluator=dict(
-            type=LMEvaluator,
-            cmp_order='both',
-            prompt_template=dict(
-                type=PromptTemplate,
-                template=dict(
-                    begin=[
-                        dict(
-                            role='SYSTEM',
-                            fallback_role='HUMAN',
-                            prompt='{prompt}'
-                        ),
-                    ],
-                    round=[dict(role='HUMAN',
-                                prompt='回答 1: <回答 1 开始> {prediction} <回答 1 结束>\n回答 2: <回答 2 开始> {prediction2} <回答 2 结束>\n')]))),
-        pred_role='BOT',
-    )
-
-    subjective_datasets.append(
-        dict(
-            abbr=f'{_name}',
-            type=SubjectiveCmpDataset,
-            path='./data/subjective/',
-            name=_name,
-            reader_cfg=subjective_reader_cfg,
-            infer_cfg=subjective_infer_cfg,
-            eval_cfg=subjective_eval_cfg
-        ))

configs/datasets/z_bench/z_bench_gen_52ba2f.py

@@ -1,25 +0,0 @@
-from opencompass.openicl.icl_prompt_template import PromptTemplate
-from opencompass.openicl.icl_retriever import ZeroRetriever
-from opencompass.openicl.icl_inferencer import GenInferencer
-from opencompass.datasets import HFDataset
-
-z_bench_reader_cfg = dict(
-    input_columns=['text'], output_column='category', train_split='test')
-
-z_bench_infer_cfg = dict(
-    prompt_template=dict(
-        type=PromptTemplate,
-        template='{text}',
-    ),
-    retriever=dict(type=ZeroRetriever),
-    inferencer=dict(type=GenInferencer))
-
-z_bench_datasets = dict(
-    type=HFDataset,
-    path=
-    '/mnt/petrelfs/gaotong/llm_eval/openagieval_dataset/eval_datasets/z_bench',
-    data_dir=
-    '/mnt/petrelfs/gaotong/llm_eval/openagieval_dataset/eval_datasets/z_bench',
-    name='question',
-    reader_cfg=z_bench_reader_cfg,
-    infer_cfg=z_bench_infer_cfg)

configs/datasets/z_bench/z_bench_gen_d8c84c.py

@@ -1,28 +0,0 @@
-from opencompass.openicl.icl_prompt_template import PromptTemplate
-from opencompass.openicl.icl_retriever import ZeroRetriever
-from opencompass.openicl.icl_inferencer import GenInferencer
-from opencompass.datasets import HFDataset
-
-z_bench_reader_cfg = dict(
-    ds_size=4,
-    input_columns=['text'],
-    output_column='category',
-    train_split='test')
-
-z_bench_infer_cfg = dict(
-    prompt_template=dict(
-        type=PromptTemplate,
-        template=dict(round=[dict(role='HUMAN', prompt='{text}')]),
-    ),
-    retriever=dict(type=ZeroRetriever),
-    inferencer=dict(type=GenInferencer))
-
-z_bench_datasets = dict(
-    type=HFDataset,
-    path=
-    '/mnt/petrelfs/gaotong/llm_eval/openagieval_dataset/eval_datasets/z_bench',
-    data_dir=
-    '/mnt/petrelfs/gaotong/llm_eval/openagieval_dataset/eval_datasets/z_bench',
-    name='question',
-    reader_cfg=z_bench_reader_cfg,
-    infer_cfg=z_bench_infer_cfg)

configs/eval_TheoremQA.py

@@ -1,18 +0,0 @@
-from mmengine.config import read_base
-
-with read_base():
-    from .models.mistral.hf_mistral_7b_v0_1 import models as hf_mistral_7b_v0_1_model
-    from .models.mistral.hf_mistral_7b_v0_2 import models as hf_mistral_7b_v0_2_model
-    from .models.hf_internlm.hf_internlm2_20b import models as hf_internlm2_20b_model
-    from .models.hf_internlm.hf_internlm2_math_20b import models as hf_internlm2_math_20b_model
-
-    from .datasets.TheoremQA.TheoremQA_5shot_gen_6f0af8 import TheoremQA_datasets as datasets
-
-models = sum([v for k, v in locals().items() if k.endswith('_model')], [])
-
-work_dir = 'outputs/TheoremQA-5shot'
-
-
-# dataset    version    metric    mode      mistral-7b-v0.1-hf    mistral-7b-v0.2-hf    internlm2-20b-hf    internlm2-math-20b-hf
-# ---------  ---------  --------  ------  --------------------  --------------------  ------------------  -----------------------
-# TheoremQA  6f0af8     score     gen                    18.00                 16.75               25.87                    30.88

configs/eval_alaya.py

@@ -1,11 +0,0 @@
-from mmengine.config import read_base
-
-with read_base():
-    from .datasets.ceval.ceval_gen import ceval_datasets
-    from .datasets.cmmlu.cmmlu_gen import cmmlu_datasets
-    from .datasets.agieval.agieval_gen import agieval_datasets
-    from .datasets.bbh.bbh_gen import bbh_datasets
-    from .datasets.mmlu.mmlu_gen import mmlu_datasets
-    from .models.alaya.alaya import models
-
-datasets = [*bbh_datasets, *ceval_datasets, *cmmlu_datasets, *agieval_datasets, *mmlu_datasets]

configs/eval_bluelm_32k_lveval.py

@@ -1,16 +0,0 @@
-from mmengine.config import read_base
-
-with read_base():
-    from .datasets.lveval.lveval import LVEval_datasets as datasets
-    from .models.bluelm.hf_bluelm_7b_chat_32k import models
-    from .summarizers.lveval import summarizer
-
-models[0][
-    'path'
-] = '/path/to/your/huggingface_models/BlueLM-7B-Chat-32K'
-models[0][
-    'tokenizer_path'
-] = '/path/to/your/huggingface_models/BlueLM-7B-Chat-32K'
-models[0]['max_seq_len'] = 32768
-models[0]['generation_kwargs'] = dict(do_sample=False)
-models[0]['mode'] = 'mid'  # truncate in the middle

configs/eval_charm.py

@@ -1,52 +0,0 @@
-from mmengine.config import read_base
-
-with read_base():
-    from .datasets.CHARM.charm_reason_gen_f8fca2 import charm_reason_datasets as datasets
-    from .models.hf_internlm.lmdeploy_internlm2_chat_7b import models as lmdeploy_7b_chat_model
-
-    # from models.openai.gpt_3_5_turbo_1106 import models as gpt_3_5_turbo_1106_model
-    # from models.openai.gpt_4_1106_preview import models as gpt_4_1106_preview_model
-
-    # from .models.chatglm.hf_chatglm3_6b_32k import models as chatglm3_6b_32k_model
-    # from .models.yi.hf_yi_6b_chat import models as yi_6b_chat_model
-    # from .models.hf_internlm.hf_internlm2_chat_7b import models as hf_internlm2_chat_7b_model
-    # from .models.deepseek.hf_deepseek_7b_chat import models as deepseek_7b_chat_model
-    # from .models.baichuan.hf_baichuan2_7b_chat import models as baichuan2_7b_chat_model  # need torch 2.1
-    # from .models.hf_llama.hf_llama2_7b_chat import models as llama2_7b_chat_model
-    # from .models.vicuna.hf_vicuna_7b_v15_16k import models as vicuna_7b_v15_16k_model
-
-    # from .models.baichuan.hf_baichuan2_13b_chat import models as baichuan2_13b_chat_model  # need torch 2.1
-    # from .models.hf_llama.hf_llama2_13b_chat import models as llama2_13b_chat_model
-    # from .models.vicuna.hf_vicuna_13b_v15_16k import models as vicuna_13b_v15_16k_model
-    # from .models.hf_internlm.hf_internlm2_chat_20b import models as hf_internlm2_chat_20b_model
-
-    # from .models.yi.hf_yi_34b_chat import models as yi_34b_chat_model
-
-    # from .models.deepseek.hf_deepseek_67b_chat import models as deepseek_67b_chat_model
-    # from .models.hf_llama.hf_llama2_70b_chat import models as llama2_70b_chat_model
-
-    # from .models.hf_llama.hf_llama3_8b_instruct import models as llama3_8b_instruct_model
-    # from .models.hf_llama.hf_llama3_70b_instruct import models as llama3_70b_instruct_model
-    from .summarizers.charm_rea import summarizer
-
-models = sum([v for k, v in locals().items() if k.endswith('_model')], [])
-work_dir = './outputs/CHARM/chat/'
-
-# dataset                                                        version    metric         mode    internlm2-chat-7b-turbomind
-# -------------------------------------------------------------  ---------  -------------  ------  -----------------------------
-# charm-reason-Direct                                               -          naive_average  gen     49.51
-# charm-reason-ZH-CoT                                               -          naive_average  gen     61.33
-# charm-reason-EN-CoT                                               -          naive_average  gen     54.55
-# charm-reason-XLT                                                  -          naive_average  gen     58.46
-# charm-reason-Translate-EN                                         -          naive_average  gen     56.15
-#                                                                -          -              -       -
-# charm-reason-Chinese_Direct                                       -          naive_average  gen     47.14
-# charm-reason-Chinese_ZH-CoT                                       -          naive_average  gen     58.40
-# charm-reason-Chinese_EN-CoT                                       -          naive_average  gen     48.31
-# charm-reason-Chinese_XLT                                          -          naive_average  gen     53.57
-# charm-reason-Chinese_Translate-EN                                 -          naive_average  gen     48.21
-# charm-reason-Global_Direct                                        -          naive_average  gen     51.88
-# charm-reason-Global_ZH-CoT                                        -          naive_average  gen     64.26
-# charm-reason-Global_EN-CoT                                        -          naive_average  gen     60.79
-# charm-reason-Global_XLT                                           -          naive_average  gen     63.36
-# charm-reason-Global_Translate-EN                                  -          naive_average  gen     64.10

configs/eval_chat_cibench_api.py

@@ -1,96 +0,0 @@
-from mmengine.config import read_base
-
-from opencompass.lagent.actions.ipython_interpreter import IPythonInterpreter
-from opencompass.lagent.agents.react import CIReAct, ReActProtocol
-from opencompass.models.lagent import CodeAgent
-from opencompass.models.openai_api import OpenAI
-from opencompass.partitioners import SizePartitioner
-from opencompass.runners import LocalRunner
-from opencompass.tasks import OpenICLInferTask
-
-with read_base():
-    from .datasets.CIBench.CIBench_template_gen_e6b12a import \
-        cibench_datasets as datasets
-
-FORCE_STOP_PROMPT_EN = """You should directly give results based on history information."""
-
-FEWSHOT_INSTRUCTION = """\
-You are an assistant who can utilize external tools.
-{tool_description}
-To use a tool, please response with the following format:
-```
-{thought} Think what you need to solve, do you need to use tools?
-{action} The tool name, should be one of [{action_names}].
-{action_input} The input to the tool that you want to use.
-```
-The tool will give you response after your response using the following format:
-```
-{response} the results after call the tool.
-```
-Therefore DO NOT generate tool response by yourself.
-
-Also please follow the guidelines:
-1. Always use code interpreter to solve the problem.
-2. The generated codes should always in a markdown code block format.
-3. The generated codes will be executed in an ipython manner and the results will be cached.
-4. Your responded code should always be simple and only solves the problem in current step.
-
-For example:
-
-File url: `xxxx`
-### Step 1. Load the dataset from the url into a pandas DataFrame named `df`.
-
-{thought} We should use `pandas` to solve this step.
-{action} IPythonInterpreter
-{action_input} ```python
-import pandas as pd
-url = "xxxx"
-data = pd.read_csv(url)
-```
-{response} The code is succeed without any outputs.
-
-Let us begin from here!
-"""
-
-IPYTHON_INTERPRETER_DESCRIPTION = '''\
-It can run Python code in a manner as jupyter notebook. The code must be a valid code that contains only python method.'''
-
-models = [
-    dict(
-        abbr='gpt-3.5-code',
-        type=CodeAgent,
-        agent_type=CIReAct,
-        max_turn=3,
-        llm=dict(
-            type=OpenAI,
-            path='gpt-3.5-turbo',
-            key='ENV',
-            query_per_second=1,
-            max_seq_len=4096,
-        ),
-        actions=[
-            dict(type=IPythonInterpreter,
-                 description=IPYTHON_INTERPRETER_DESCRIPTION,
-                 user_data_dir='./data/cibench_dataset/datasources')
-        ],
-        protocol=dict(
-            type=ReActProtocol,
-            call_protocol=FEWSHOT_INSTRUCTION,
-            force_stop=FORCE_STOP_PROMPT_EN,
-            finish=dict(role='FINISH', begin='Final Answer:', end='\n'),
-        ),
-        batch_size=1,
-        use_system_role=False, # use `user` role instead of system role
-        first_system_role=False, # use `user` role of the first instruction prompt
-        merge_adjacent_role=True, # merge adjacent same user content
-    ),
-]
-
-
-infer = dict(
-    partitioner=dict(type=SizePartitioner, max_task_size=1000),
-    runner=dict(
-        type=LocalRunner,
-        max_num_workers=16,
-        task=dict(type=OpenICLInferTask)),
-)

configs/eval_cibench.py

@@ -1,80 +0,0 @@
-from mmengine.config import read_base
-from opencompass.partitioners import SizePartitioner
-from opencompass.runners import LocalRunner, SlurmRunner
-from opencompass.tasks import OpenICLInferTask
-from opencompass.models import OpenAI
-from opencompass.lagent.actions.ipython_interpreter import IPythonInterpreter
-from opencompass.lagent.agents.react import CIReAct
-from opencompass.models.lagent import CodeAgent
-from lagent.agents.react import ReActProtocol
-
-with read_base():
-    from .datasets.CIBench.CIBench_gen_eb42f9 import cibench_datasets as datasets
-
-FORCE_STOP_PROMPT_EN = """You should directly give results based on history information."""
-
-FEWSHOT_INSTRUCTION = """\
-You are an assistant who can utilize external tools.
-{tool_description}
-To use a tool, please response with the following format:
-```
-{thought} Think what you need to solve, do you need to use tools?
-{action} The tool name, should be one of [{action_names}].
-{action_input} The input to the tool that you want to use.
-```
-The tool will give you response after your response using the following format:
-```
-{response} the results after call the tool.
-```
-Therefore DO NOT generate tool response by yourself.
-
-Also please follow the guidelines:
-1. Always use code interpreter to solve the problem.
-2. The generated codes should always in a markdown code block format.
-3. The generated codes will be executed in an ipython manner and the results will be cached.
-4. Your responded code should always be simple and only solves the problem in current step.
-
-Begin!
-"""
-
-models = [
-    dict(
-        abbr='gpt-3.5-turbo',
-        type=CodeAgent,
-        agent_type=CIReAct,
-        mutli_rounds=True,
-        max_turn=3,
-        llm=dict(
-            type=OpenAI,
-            path='gpt-3.5-turbo',
-            key='ENV',
-            query_per_second=1,
-            max_seq_len=4096,
-        ),
-        actions=[
-            dict(
-                type=IPythonInterpreter,
-                description=
-                '''It can run Python code in a manner as jupyter notebook. The code must be a valid code that contains only python method.
-'''),
-        ],
-        protocol=dict(
-            type=ReActProtocol,
-            call_protocol=FEWSHOT_INSTRUCTION,
-            force_stop=FORCE_STOP_PROMPT_EN,
-            action=dict(role='ACTION', begin='Tool:', end='\n'),
-            action_input=dict(role='ARGS', begin='Tool Input:', end='\n'),
-            response=dict(role='RESPONSE', begin='Tool Response:', end='\n'),
-            finish=dict(role='FINISH', begin='Final Answer:', end='\n'),
-        ),
-        batch_size=8,
-    ),
-]
-
-
-infer = dict(
-    partitioner=dict(type=SizePartitioner, max_task_size=50, gen_task_coef=1),
-    runner=dict(
-        type=SlurmRunner, max_num_workers=8, retry=2,
-        task=dict(type=OpenICLInferTask)),
-)

configs/eval_circular.py

@@ -1,91 +0,0 @@
-from mmengine.config import read_base
-from opencompass.datasets.circular import (CircularCEvalDataset, CircularMMLUDataset, CircularCMMLUDataset, CircularCSQADataset,
-                                           CircularARCDataset, CircularHSWAGDataset, CircularOBQADataset, CircularRaceDataset, CircularEvaluator)
-from opencompass.summarizers import CircularSummarizer
-
-with read_base():
-    from .datasets.ceval.ceval_gen_5f30c7 import ceval_datasets
-    from .datasets.mmlu.mmlu_gen_a484b3 import mmlu_datasets
-    from .datasets.cmmlu.cmmlu_gen_c13365 import cmmlu_datasets
-    from .datasets.hellaswag.hellaswag_gen_6faab5 import hellaswag_datasets
-    from .datasets.ARC_e.ARC_e_gen_1e0de5 import ARC_e_datasets
-    from .datasets.ARC_c.ARC_c_gen_1e0de5 import ARC_c_datasets
-    from .datasets.commonsenseqa.commonsenseqa_gen_1da2d0 import commonsenseqa_datasets
-    from .datasets.obqa.obqa_gen_9069e4 import obqa_datasets
-    from .datasets.race.race_gen_69ee4f import race_datasets
-
-    from .models.hf_internlm.hf_internlm_chat_7b import models as hf_internlm_chat_7b_model
-    from .models.hf_internlm.hf_internlm_chat_20b import models as hf_internlm_chat_20b_model
-    from .models.qwen.hf_qwen_7b_chat import models as hf_qwen_7b_chat_model
-    from .models.qwen.hf_qwen_14b_chat import models as hf_qwen_14b_chat_model
-
-    from .summarizers.groups.mmlu import mmlu_summary_groups
-    from .summarizers.groups.cmmlu import cmmlu_summary_groups
-    from .summarizers.groups.ceval import ceval_summary_groups
-
-for ds, t in [
-    (ceval_datasets, CircularCEvalDataset),
-    (mmlu_datasets, CircularMMLUDataset),
-    (cmmlu_datasets, CircularCMMLUDataset),
-    (hellaswag_datasets, CircularHSWAGDataset),
-    (ARC_e_datasets, CircularARCDataset),
-    (ARC_c_datasets, CircularARCDataset),
-    (commonsenseqa_datasets, CircularCSQADataset),
-    (obqa_datasets, CircularOBQADataset),
-    (race_datasets, CircularRaceDataset),
-]:
-    for d in ds:
-        d['type'] = t
-        d['abbr'] = d['abbr'] + '-circular-4'
-        d['eval_cfg']['evaluator'] = {'type': CircularEvaluator, 'circular_pattern': 'circular'}
-        d['circular_patterns'] = 'circular'
-
-
-datasets = sum([v for k, v in locals().items() if k.endswith('_datasets') or k == 'datasets'], [])
-models = sum([v for k, v in locals().items() if k.endswith('_model')], [])
-
-# config summarizer
-other_summary_groups = [
-    {'name': 'average',
-     'subsets': ['ceval', 'mmlu', 'cmmlu', 'hellaswag', 'ARC-e', 'ARC-c', 'commonsense_qa', 'openbookqa_fact', 'race-middle', 'race-high']},
-]
-origin_summary_groups = sum([v for k, v in locals().items() if k.endswith('_summary_groups')], [])
-new_summary_groups = []
-for item in origin_summary_groups:
-    new_summary_groups.append(
-        {
-            'name': item['name'] + '-circular-4',
-            'subsets': [i + '-circular-4' for i in item['subsets']],
-        }
-    )
-summarizer = dict(
-    type=CircularSummarizer,
-    metric_types=['acc_origin', 'perf_circular'],
-    dataset_abbrs = [
-        'average-circular-4',
-        'ceval-circular-4',
-        'mmlu-circular-4',
-        'cmmlu-circular-4',
-        'hellaswag-circular-4',
-        'ARC-e-circular-4',
-        'ARC-c-circular-4',
-        'commonsense_qa-circular-4',
-        'openbookqa_fact-circular-4',
-        'race-middle-circular-4',
-        'race-high-circular-4',
-        'ceval-humanities-circular-4',
-        'ceval-stem-circular-4',
-        'ceval-social-science-circular-4',
-        'ceval-other-circular-4',
-        'mmlu-humanities-circular-4',
-        'mmlu-stem-circular-4',
-        'mmlu-social-science-circular-4',
-        'mmlu-other-circular-4',
-        'cmmlu-humanities-circular-4',
-        'cmmlu-stem-circular-4',
-        'cmmlu-social-science-circular-4',
-        'cmmlu-other-circular-4',
-        'cmmlu-china-specific-circular-4',
-    ],
-    summary_groups=new_summary_groups,
-)

configs/eval_codeagent.py

@@ -1,52 +0,0 @@
-from mmengine.config import read_base
-from opencompass.partitioners import SizePartitioner
-from opencompass.runners import LocalRunner
-from opencompass.tasks import OpenICLInferTask
-from opencompass.models import OpenAI, HuggingFaceCausalLM
-from opencompass.models.lagent import CodeAgent
-
-with read_base():
-    from .datasets.math.math_gen_943d32 import math_datasets
-    from .datasets.gsm8k.gsm8k_gen_57b0b1 import gsm8k_datasets
-
-datasets = []
-datasets += gsm8k_datasets
-datasets += math_datasets
-
-models = [
-    dict(
-        abbr='gpt-3.5-react',
-        type=CodeAgent,
-        llm=dict(
-            type=OpenAI,
-            path='gpt-3.5-turbo',
-            key='ENV',
-            query_per_second=1,
-            max_seq_len=4096,
-        ),
-        batch_size=8),
-    dict(
-        abbr='WizardCoder-Python-13B-V1.0-react',
-        type=CodeAgent,
-        llm=dict(
-            type=HuggingFaceCausalLM,
-            path='WizardLM/WizardCoder-Python-13B-V1.0',
-            tokenizer_path='WizardLM/WizardCoder-Python-13B-V1.0',
-            tokenizer_kwargs=dict(
-                padding_side='left',
-                truncation_side='left',
-                trust_remote_code=True,
-            ),
-            max_seq_len=2048,
-            model_kwargs=dict(trust_remote_code=True, device_map='auto'),
-        ),
-        batch_size=8,
-        run_cfg=dict(num_gpus=2, num_procs=1)),
-]
-
-infer = dict(
-    partitioner=dict(type=SizePartitioner, max_task_size=40000),
-    runner=dict(
-        type=LocalRunner, max_num_workers=16,
-        task=dict(type=OpenICLInferTask)),
-)

configs/eval_codegeex2.py

@@ -1,7 +0,0 @@
-from mmengine.config import read_base
-
-with read_base():
-    from .datasets.humanevalx.humanevalx_gen import humanevalx_datasets
-    from .models.codegeex2.hf_codegeex2_6b import models
-
-datasets = humanevalx_datasets

configs/eval_contamination.py

@@ -1,15 +0,0 @@
-from mmengine.config import read_base
-
-with read_base():
-    from .datasets.ceval.ceval_clean_ppl import ceval_datasets
-    from .datasets.mmlu.mmlu_clean_ppl import mmlu_datasets
-    from .datasets.hellaswag.hellaswag_clean_ppl import hellaswag_datasets
-    from .datasets.ARC_c.ARC_c_clean_ppl import ARC_c_datasets
-    from .models.yi.hf_yi_6b import models as hf_yi_6b_model
-    from .models.qwen.hf_qwen_7b import models as hf_qwen_7b_model
-    from .models.hf_llama.hf_llama2_7b import models as hf_llama2_7b_model
-    from .summarizers.contamination import summarizer
-
-
-datasets = [*ceval_datasets, *mmlu_datasets, *hellaswag_datasets, *ARC_c_datasets]
-models = [*hf_yi_6b_model, *hf_qwen_7b_model, *hf_llama2_7b_model]

configs/eval_demo.py

@@ -1,10 +0,0 @@
-from mmengine.config import read_base
-
-with read_base():
-    from .datasets.siqa.siqa_gen import siqa_datasets
-    from .datasets.winograd.winograd_ppl import winograd_datasets
-    from .models.opt.hf_opt_125m import opt125m
-    from .models.opt.hf_opt_350m import opt350m
-
-datasets = [*siqa_datasets, *winograd_datasets]
-models = [opt125m, opt350m]

configs/eval_gpt3.5.py

@@ -1,36 +0,0 @@
-from mmengine.config import read_base
-from opencompass.models import OpenAI
-from opencompass.partitioners import NaivePartitioner
-from opencompass.runners import LocalRunner
-from opencompass.tasks import OpenICLInferTask
-
-with read_base():
-    # choose a list of datasets
-    from .datasets.collections.chat_medium import datasets
-    # and output the results in a choosen format
-    from .summarizers.medium import summarizer
-
-
-api_meta_template = dict(
-    round=[
-            dict(role='HUMAN', api_role='HUMAN'),
-            dict(role='BOT', api_role='BOT', generate=True),
-    ],
-)
-
-models = [
-    dict(abbr='GPT-3.5-turbo-0613',
-        type=OpenAI, path='gpt-3.5-turbo-0613',
-        key='ENV',  # The key will be obtained from $OPENAI_API_KEY, but you can write down your key here as well
-        meta_template=api_meta_template,
-        query_per_second=1,
-        max_out_len=2048, max_seq_len=4096, batch_size=8),
-]
-
-infer = dict(
-    partitioner=dict(type=NaivePartitioner),
-    runner=dict(
-        type=LocalRunner,
-        max_num_workers=8,
-        task=dict(type=OpenICLInferTask)),
-)

configs/eval_gpt4.py

@@ -1,40 +0,0 @@
-from mmengine.config import read_base
-from opencompass.models import OpenAI
-from opencompass.partitioners import NaivePartitioner
-from opencompass.runners import LocalRunner
-from opencompass.tasks import OpenICLInferTask
-
-with read_base():
-    from .datasets.collections.chat_medium import datasets
-    from .summarizers.medium import summarizer
-
-# GPT4 needs a special humaneval postprocessor
-from opencompass.datasets.humaneval import humaneval_gpt_postprocess
-for _dataset in datasets:
-    if _dataset['path'] == 'openai_humaneval':
-        _dataset['eval_cfg']['pred_postprocessor']['type'] = humaneval_gpt_postprocess
-
-
-api_meta_template = dict(
-    round=[
-            dict(role='HUMAN', api_role='HUMAN'),
-            dict(role='BOT', api_role='BOT', generate=True),
-    ],
-)
-
-models = [
-    dict(abbr='GPT4',
-        type=OpenAI, path='gpt-4-0613',
-        key='ENV',  # The key will be obtained from $OPENAI_API_KEY, but you can write down your key here as well
-        meta_template=api_meta_template,
-        query_per_second=1,
-        max_out_len=2048, max_seq_len=2048, batch_size=8),
-]
-
-infer = dict(
-    partitioner=dict(type=NaivePartitioner),
-    runner=dict(
-        type=LocalRunner,
-        max_num_workers=4,
-        task=dict(type=OpenICLInferTask)),
-)

configs/eval_hf_llama2.py

@@ -1,18 +0,0 @@
-from mmengine.config import read_base
-
-with read_base():
-    from .datasets.mmlu.mmlu_ppl_ac766d import mmlu_datasets
-    from .datasets.triviaqa.triviaqa_wiki_gen_d18bf4 import triviaqa_datasets
-    from .datasets.nq.nq_open_gen_e93f8a import nq_datasets
-    from .datasets.gsm8k.gsm8k_gen_3309bd import gsm8k_datasets
-    from .datasets.humaneval.humaneval_gen_a82cae import humaneval_datasets
-    from .datasets.agieval.agieval_mixed_713d14 import agieval_datasets
-    from .datasets.SuperGLUE_BoolQ.SuperGLUE_BoolQ_ppl_314797 import BoolQ_datasets
-    from .datasets.hellaswag.hellaswag_ppl_a6e128 import hellaswag_datasets
-    from .datasets.obqa.obqa_ppl_6aac9e import obqa_datasets
-    from .datasets.winogrande.winogrande_ll_c5cf57 import winogrande_datasets
-    from .models.hf_llama.hf_llama2_7b import models
-    from .summarizers.example import summarizer
-
-datasets = sum([v for k, v in locals().items() if k.endswith('_datasets') or k == 'datasets'], [])
-work_dir = './outputs/llama2/'

configs/eval_hf_llama_7b.py

@@ -1,8 +0,0 @@
-from mmengine.config import read_base
-
-with read_base():
-    from .datasets.collections.base_medium_llama import piqa_datasets, siqa_datasets
-    from .models.hf_llama.hf_llama_7b import models
-
-
-datasets = [*piqa_datasets, *siqa_datasets]

configs/eval_internLM.py

@@ -1,9 +0,0 @@
-from mmengine.config import read_base
-
-with read_base():
-    # choose a list of datasets
-    from .datasets.collections.base_medium import datasets
-    # choose a model of interest
-    from .models.internlm.internlm_7b import models
-    # and output the results in a choosen format
-    from .summarizers.medium import summarizer

configs/eval_internlm2_chat_keyset.py

@@ -1,37 +0,0 @@
-from copy import deepcopy
-from mmengine.config import read_base
-
-with read_base():
-    from .datasets.mmlu.mmlu_gen_a484b3 import mmlu_datasets
-    from .datasets.agieval.agieval_gen_64afd3 import agieval_datasets
-    from .datasets.bbh.bbh_gen_5b92b0 import bbh_datasets
-    from .datasets.gsm8k.gsm8k_gen_1d7fe4 import gsm8k_datasets
-    from .datasets.math.math_evaluatorv2_gen_cecb31 import math_datasets
-    from .datasets.humaneval.humaneval_gen_8e312c import humaneval_datasets
-    from .datasets.mbpp.deprecated_sanitized_mbpp_gen_1e1056 import sanitized_mbpp_datasets
-
-    from .models.hf_internlm.hf_internlm2_chat_7b import models as hf_internlm2_chat_7b_model
-    from .models.hf_internlm.hf_internlm2_chat_20b import models as hf_internlm2_chat_20b_model
-
-    from .summarizers.internlm2_keyset import summarizer
-
-work_dir = './outputs/internlm2-chat-keyset/'
-
-_origin_datasets = sum([v for k, v in locals().items() if k.endswith('_datasets')], [])
-_origin_models = sum([v for k, v in locals().items() if k.endswith('_model')], [])
-
-_vanilla_datasets = [deepcopy(d) for d in _origin_datasets]
-_vanilla_models = []
-for m in _origin_models:
-    m = deepcopy(m)
-    if 'meta_template' in m and 'round' in m['meta_template']:
-        round = m['meta_template']['round']
-        if any(r['role'] == 'SYSTEM' for r in round):
-            new_round = [r for r in round if r['role'] != 'SYSTEM']
-            print(f'WARNING: remove SYSTEM round in meta_template for {m.get("abbr", None)}')
-            m['meta_template']['round'] = new_round
-    _vanilla_models.append(m)
-
-
-datasets = _vanilla_datasets
-models = _vanilla_models

configs/eval_internlm2_keyset.py

@@ -1,20 +0,0 @@
-from mmengine.config import read_base
-
-with read_base():
-    from .datasets.mmlu.mmlu_ppl_ac766d import mmlu_datasets
-    from .datasets.agieval.agieval_mixed_713d14 import agieval_datasets
-    from .datasets.bbh.bbh_gen_5b92b0 import bbh_datasets
-    from .datasets.gsm8k.gsm8k_gen_1d7fe4 import gsm8k_datasets
-    from .datasets.math.math_gen_265cce import math_datasets
-    from .datasets.humaneval.humaneval_gen_a82cae import humaneval_datasets
-    from .datasets.mbpp.deprecated_sanitized_mbpp_gen_1e1056 import sanitized_mbpp_datasets
-
-    from .models.hf_internlm.hf_internlm2_7b import models as hf_internlm2_7b_model
-    from .models.hf_internlm.hf_internlm2_20b import models as hf_internlm2_20b_model
-
-    from .summarizers.internlm2_keyset import summarizer
-
-work_dir = './outputs/internlm2-keyset/'
-
-datasets = sum([v for k, v in locals().items() if k.endswith('_datasets')], [])
-models = sum([v for k, v in locals().items() if k.endswith('_model')], [])

configs/eval_internlm_7b.py

@@ -1,9 +0,0 @@
-from mmengine.config import read_base
-
-with read_base():
-    # choose a list of datasets
-    from .datasets.collections.base_medium import datasets
-    # choose a model of interest
-    from .models.hf_internlm.hf_internlm_7b import models
-    # and output the results in a choosen format
-    from .summarizers.medium import summarizer

configs/eval_internlm_chat_lmdeploy_apiserver.py

@@ -1,51 +0,0 @@
-from mmengine.config import read_base
-from opencompass.models.turbomind_api import TurboMindAPIModel
-
-with read_base():
-    # choose a list of datasets
-    from .datasets.mmlu.mmlu_gen_a484b3 import mmlu_datasets
-    from .datasets.ceval.ceval_gen_5f30c7 import ceval_datasets
-    from .datasets.SuperGLUE_WiC.SuperGLUE_WiC_gen_d06864 import WiC_datasets
-    from .datasets.SuperGLUE_WSC.SuperGLUE_WSC_gen_7902a7 import WSC_datasets
-    from .datasets.triviaqa.triviaqa_gen_2121ce import triviaqa_datasets
-    from .datasets.gsm8k.gsm8k_gen_1d7fe4 import gsm8k_datasets
-    from .datasets.race.race_gen_69ee4f import race_datasets
-    from .datasets.crowspairs.crowspairs_gen_381af0 import crowspairs_datasets
-    # and output the results in a choosen format
-    from .summarizers.medium import summarizer
-
-datasets = sum((v for k, v in locals().items() if k.endswith('_datasets')), [])
-
-
-meta_template = dict(
-    round=[
-        dict(role='HUMAN', begin='<|User|>:', end='\n'),
-        dict(role='BOT', begin='<|Bot|>:', end='<eoa>\n', generate=True),
-    ],
-    eos_token_id=103028)
-
-internlm_chat_20b = dict(
-    type=TurboMindAPIModel,
-    abbr='internlm-chat-20b-turbomind',
-    api_addr='http://0.0.0.0:23333',
-    max_out_len=100,
-    max_seq_len=2048,
-    batch_size=8,
-    meta_template=meta_template,
-    run_cfg=dict(num_gpus=1, num_procs=1),
-    end_str='<eoa>',
-)
-
-internlm_chat_7b = dict(
-    type=TurboMindAPIModel,
-    abbr='internlm-chat-7b-turbomind',
-    api_addr='http://0.0.0.0:23333',
-    max_out_len=100,
-    max_seq_len=2048,
-    batch_size=16,
-    meta_template=meta_template,
-    run_cfg=dict(num_gpus=1, num_procs=1),
-    end_str='<eoa>',
-)
-
-models = [internlm_chat_20b]

configs/eval_internlm_chat_lmdeploy_pytorch.py

@@ -1,69 +0,0 @@
-from mmengine.config import read_base
-from opencompass.models import LmdeployPytorchModel
-
-
-with read_base():
-    # choose a list of datasets
-    from .datasets.mmlu.mmlu_gen_a484b3 import mmlu_datasets
-    from .datasets.ceval.ceval_gen_5f30c7 import ceval_datasets
-    from .datasets.SuperGLUE_WiC.SuperGLUE_WiC_gen_d06864 import WiC_datasets
-    from .datasets.SuperGLUE_WSC.SuperGLUE_WSC_gen_7902a7 import WSC_datasets
-    from .datasets.triviaqa.triviaqa_gen_2121ce import triviaqa_datasets
-    from .datasets.gsm8k.gsm8k_gen_1d7fe4 import gsm8k_datasets
-    from .datasets.race.race_gen_69ee4f import race_datasets
-    from .datasets.crowspairs.crowspairs_gen_381af0 import crowspairs_datasets
-    # and output the results in a choosen format
-    from .summarizers.medium import summarizer
-
-
-datasets = sum((v for k, v in locals().items() if k.endswith('_datasets')), [])
-
-
-meta_template = dict(
-    round=[
-        dict(role='HUMAN', begin='<|User|>:', end='<eoh>\n'),
-        dict(role='BOT', begin='<|Bot|>:', end='<eoa>\n', generate=True),
-    ],
-    eos_token_id=103028)
-
-# config for internlm-chat-7b
-internlm_chat_7b = dict(
-    type=LmdeployPytorchModel,
-    abbr='internlm-chat-7b-pytorch',
-    path='internlm/internlm-chat-7b',
-    engine_config=dict(session_len=2048,
-                       max_batch_size=16),
-    gen_config=dict(top_k=1,
-                    top_p=0.8,
-                    temperature=1.0,
-                    max_new_tokens=100),
-    max_out_len=100,
-    max_seq_len=2048,
-    batch_size=16,
-    concurrency=16,
-    meta_template=meta_template,
-    run_cfg=dict(num_gpus=1, num_procs=1),
-    end_str='<eoa>',
-)
-
-# config for internlm-chat-20b
-internlm_chat_20b = dict(
-    type=LmdeployPytorchModel,
-    abbr='internlm-chat-20b-pytorch',
-    path='internlm/internlm-chat-20b',
-    engine_config=dict(session_len=2048,
-                       max_batch_size=8),
-    gen_config=dict(top_k=1,
-                    top_p=0.8,
-                    temperature=1.0,
-                    max_new_tokens=100),
-    max_out_len=100,
-    max_seq_len=2048,
-    batch_size=8,
-    concurrency=8,
-    meta_template=meta_template,
-    run_cfg=dict(num_gpus=1, num_procs=1),
-    end_str='<eoa>',
-    )
-
-models = [internlm_chat_20b]

configs/eval_internlm_chat_lmdeploy_tis.py

@@ -1,41 +0,0 @@
-from mmengine.config import read_base
-from opencompass.models.lmdeploy_tis import LmdeployTisModel
-
-with read_base():
-    # choose a list of datasets
-    from .datasets.mmlu.mmlu_gen_a484b3 import mmlu_datasets
-    from .datasets.ceval.ceval_gen_5f30c7 import ceval_datasets
-    from .datasets.SuperGLUE_WiC.SuperGLUE_WiC_gen_d06864 import WiC_datasets
-    from .datasets.SuperGLUE_WSC.SuperGLUE_WSC_gen_7902a7 import WSC_datasets
-    from .datasets.triviaqa.triviaqa_gen_2121ce import triviaqa_datasets
-    from .datasets.gsm8k.gsm8k_gen_1d7fe4 import gsm8k_datasets
-    from .datasets.humaneval.humaneval_gen_8e312c import humaneval_datasets
-    from .datasets.race.race_gen_69ee4f import race_datasets
-    from .datasets.crowspairs.crowspairs_gen_381af0 import crowspairs_datasets
-    # and output the results in a choosen format
-    from .summarizers.medium import summarizer
-
-datasets = sum((v for k, v in locals().items() if k.endswith('_datasets')), [])
-
-meta_template = dict(
-    round=[
-        dict(role='HUMAN', begin='<|im_start|>user\n', end='<|im_end|>\n'),
-        dict(role='BOT', begin='<|im_start|>assistant\n', end='<|im_end|>\n', generate=True),
-    ],
-    eos_token_id=92542
-)
-
-models = [
-    dict(
-        type=LmdeployTisModel,
-        abbr='internlm-chat-20b-lmdeploy-tis',
-        path='internlm/internlm-chat-20b',
-        tis_addr='0.0.0.0:33337',
-        max_out_len=100,
-        max_seq_len=2048,
-        batch_size=8,
-        meta_template=meta_template,
-        run_cfg=dict(num_gpus=1, num_procs=1),
-        end_str='<|im_end|>',
-    )
-]

configs/eval_internlm_chat_turbomind.py

@@ -1,97 +0,0 @@
-from mmengine.config import read_base
-from opencompass.models.turbomind import TurboMindModel
-
-with read_base():
-    # choose a list of datasets
-    from .datasets.mmlu.mmlu_gen_a484b3 import mmlu_datasets
-    from .datasets.ceval.ceval_gen_5f30c7 import ceval_datasets
-    from .datasets.SuperGLUE_WiC.SuperGLUE_WiC_gen_d06864 import WiC_datasets
-    from .datasets.SuperGLUE_WSC.SuperGLUE_WSC_gen_7902a7 import WSC_datasets
-    from .datasets.triviaqa.triviaqa_gen_2121ce import triviaqa_datasets
-    from .datasets.gsm8k.gsm8k_gen_1d7fe4 import gsm8k_datasets
-    from .datasets.race.race_gen_69ee4f import race_datasets
-    from .datasets.crowspairs.crowspairs_gen_381af0 import crowspairs_datasets
-    # and output the results in a choosen format
-    from .summarizers.medium import summarizer
-
-
-datasets = sum((v for k, v in locals().items() if k.endswith('_datasets')), [])
-
-internlm_meta_template = dict(round=[
-    dict(role='HUMAN', begin='<|User|>:', end='\n'),
-    dict(role='BOT', begin='<|Bot|>:', end='<eoa>\n', generate=True),
-],
-                              eos_token_id=103028)
-
-internlm2_meta_template = dict(
-    round=[
-        dict(role='HUMAN', begin='<|im_start|>user\n', end='<|im_end|>\n'),
-        dict(role='BOT', begin='<|im_start|>assistant\n', end='<|im_end|>\n', generate=True),
-    ],
-    eos_token_id=92542
-)
-
-# config for internlm-chat-7b
-internlm_chat_7b = dict(
-    type=TurboMindModel,
-    abbr='internlm-chat-7b-turbomind',
-    path='internlm/internlm-chat-7b',
-    engine_config=dict(session_len=2048,
-                       max_batch_size=32,
-                       rope_scaling_factor=1.0),
-    gen_config=dict(top_k=1,
-                    top_p=0.8,
-                    temperature=1.0,
-                    max_new_tokens=100),
-    max_out_len=100,
-    max_seq_len=2048,
-    batch_size=32,
-    concurrency=32,
-    meta_template=internlm_meta_template,
-    run_cfg=dict(num_gpus=1, num_procs=1),
-    end_str='<eoa>',
-)
-
-# config for internlm-chat-7b
-internlm2_chat_7b = dict(
-    type=TurboMindModel,
-    abbr='internlm2-chat-7b-turbomind',
-    path='internlm/internlm2-chat-7b',
-    engine_config=dict(session_len=2048,
-                       max_batch_size=32,
-                       rope_scaling_factor=1.0),
-    gen_config=dict(top_k=1,
-                    top_p=0.8,
-                    temperature=1.0,
-                    max_new_tokens=100),
-    max_out_len=100,
-    max_seq_len=2048,
-    batch_size=32,
-    concurrency=32,
-    meta_template=internlm2_meta_template,
-    run_cfg=dict(num_gpus=1, num_procs=1),
-    end_str='<|im_end|>'
-)
-
-# config for internlm-chat-20b
-internlm_chat_20b = dict(
-    type=TurboMindModel,
-    abbr='internlm-chat-20b-turbomind',
-    path='internlm/internlm-chat-20b',
-    engine_config=dict(session_len=2048,
-                       max_batch_size=8,
-                       rope_scaling_factor=1.0),
-    gen_config=dict(top_k=1,
-                    top_p=0.8,
-                    temperature=1.0,
-                    max_new_tokens=100),
-    max_out_len=100,
-    max_seq_len=2048,
-    batch_size=8,
-    concurrency=8,
-    meta_template=internlm_meta_template,
-    run_cfg=dict(num_gpus=1, num_procs=1),
-    end_str='<eoa>',
-)
-
-models = [internlm_chat_20b]

configs/eval_internlm_chat_turbomind_tis.py

@@ -1,40 +0,0 @@
-from mmengine.config import read_base
-from opencompass.models.turbomind_tis import TurboMindTisModel
-
-with read_base():
-    # choose a list of datasets
-    from .datasets.mmlu.mmlu_gen_a484b3 import mmlu_datasets
-    from .datasets.ceval.ceval_gen_5f30c7 import ceval_datasets
-    from .datasets.SuperGLUE_WiC.SuperGLUE_WiC_gen_d06864 import WiC_datasets
-    from .datasets.SuperGLUE_WSC.SuperGLUE_WSC_gen_7902a7 import WSC_datasets
-    from .datasets.triviaqa.triviaqa_gen_2121ce import triviaqa_datasets
-    from .datasets.gsm8k.gsm8k_gen_1d7fe4 import gsm8k_datasets
-    from .datasets.humaneval.humaneval_gen_8e312c import humaneval_datasets
-    from .datasets.race.race_gen_69ee4f import race_datasets
-    from .datasets.crowspairs.crowspairs_gen_381af0 import crowspairs_datasets
-    # and output the results in a choosen format
-    from .summarizers.medium import summarizer
-
-datasets = sum((v for k, v in locals().items() if k.endswith('_datasets')), [])
-
-
-meta_template = dict(
-    round=[
-        dict(role='HUMAN', begin='<|User|>:', end='\n'),
-        dict(role='BOT', begin='<|Bot|>:', end='<eoa>\n', generate=True),
-    ],
-    eos_token_id=103028)
-
-models = [
-    dict(
-        type=TurboMindTisModel,
-        abbr='internlm-chat-20b-turbomind',
-        path='internlm',
-        tis_addr='0.0.0.0:33337',
-        max_out_len=100,
-        max_seq_len=2048,
-        batch_size=8,
-        meta_template=meta_template,
-        run_cfg=dict(num_gpus=1, num_procs=1),
-    )
-]

configs/eval_internlm_math_chat.py

@@ -1,15 +0,0 @@
-from mmengine.config import read_base
-from opencompass.models.huggingface import HuggingFaceCausalLM
-
-with read_base():
-    # choose a list of datasets
-    from .datasets.gsm8k.gsm8k_gen import gsm8k_datasets
-    from .datasets.math.math_gen_736506 import math_datasets
-
-    from .models.hf_internlm.hf_internlm2_chat_math_7b import models as internlm_math_chat_7b_models
-    from .models.hf_internlm.hf_internlm2_chat_math_20b import models as internlm_math_chat_20b_models
-
-datasets = sum((v for k, v in locals().items() if k.endswith('_datasets')), [])
-# Eval Math and GSM8k for both Internlm-Math-Chat-7B and 20b
-datasets = [*math_datasets, *gsm8k_datasets]
-models = [*internlm_math_chat_7b_models, *internlm_math_chat_20b_models]

configs/eval_internlm_turbomind.py

@@ -1,55 +0,0 @@
-from mmengine.config import read_base
-from opencompass.models.turbomind import TurboMindModel
-
-
-with read_base():
-    # choose a list of datasets
-    from .datasets.mmlu.mmlu_gen_a484b3 import mmlu_datasets
-    from .datasets.ceval.ceval_gen_5f30c7 import ceval_datasets
-    from .datasets.SuperGLUE_WiC.SuperGLUE_WiC_gen_d06864 import WiC_datasets
-    from .datasets.triviaqa.triviaqa_gen_2121ce import triviaqa_datasets
-    from .datasets.gsm8k.gsm8k_gen_1d7fe4 import gsm8k_datasets
-    from .datasets.humaneval.humaneval_gen_8e312c import humaneval_datasets
-    # and output the results in a choosen format
-    from .summarizers.medium import summarizer
-
-datasets = sum((v for k, v in locals().items() if k.endswith('_datasets')), [])
-
-# # config for internlm-7b model
-internlm_7b = dict(
-        type=TurboMindModel,
-        abbr='internlm-7b-turbomind',
-        path='internlm/internlm-7b',
-        engine_config=dict(session_len=2048,
-                           max_batch_size=32,
-                           rope_scaling_factor=1.0),
-        gen_config=dict(top_k=1,
-                        top_p=0.8,
-                        temperature=1.0,
-                        max_new_tokens=100),
-        max_out_len=100,
-        max_seq_len=2048,
-        batch_size=32,
-        concurrency=32,
-        run_cfg=dict(num_gpus=1, num_procs=1),
-    )
-
-# config for internlm-20b model
-internlm_20b = dict(
-        type=TurboMindModel,
-        abbr='internlm-20b-turbomind',
-        path='internlm/internlm-20b',
-        engine_config=dict(session_len=2048,
-                           max_batch_size=8,
-                           rope_scaling_factor=1.0),
-        gen_config=dict(top_k=1, top_p=0.8,
-                        temperature=1.0,
-                        max_new_tokens=100),
-        max_out_len=100,
-        max_seq_len=2048,
-        batch_size=8,
-        concurrency=8,
-        run_cfg=dict(num_gpus=1, num_procs=1),
-    )
-
-models = [internlm_20b]

configs/eval_internlm_turbomind_tis.py

@@ -1,28 +0,0 @@
-from mmengine.config import read_base
-from opencompass.models.turbomind_tis import TurboMindTisModel
-
-with read_base():
-    # choose a list of datasets
-    from .datasets.mmlu.mmlu_gen_a484b3 import mmlu_datasets
-    from .datasets.ceval.ceval_gen_5f30c7 import ceval_datasets
-    from .datasets.SuperGLUE_WiC.SuperGLUE_WiC_gen_d06864 import WiC_datasets
-    from .datasets.triviaqa.triviaqa_gen_2121ce import triviaqa_datasets
-    from .datasets.gsm8k.gsm8k_gen_1d7fe4 import gsm8k_datasets
-    from .datasets.humaneval.humaneval_gen_8e312c import humaneval_datasets
-    # and output the results in a choosen format
-    from .summarizers.medium import summarizer
-
-datasets = sum((v for k, v in locals().items() if k.endswith('_datasets')), [])
-
-models = [
-    dict(
-        type=TurboMindTisModel,
-        abbr='internlm-chat-20b-turbomind',
-        path='internlm',
-        tis_addr='0.0.0.0:33337',
-        max_out_len=100,
-        max_seq_len=2048,
-        batch_size=8,
-        run_cfg=dict(num_gpus=1, num_procs=1),
-    )
-]

configs/eval_llama2_7b.py

@@ -1,8 +0,0 @@
-from mmengine.config import read_base
-
-with read_base():
-    from .datasets.collections.base_medium_llama import piqa_datasets, siqa_datasets
-    from .models.llama.llama2_7b import models
-
-
-datasets = [*piqa_datasets, *siqa_datasets]

configs/eval_llama2_7b_lveval.py

@@ -1,16 +0,0 @@
-from mmengine.config import read_base
-
-with read_base():
-    from .datasets.lveval.lveval import LVEval_datasets as datasets
-    from .models.hf_llama.hf_llama2_7b_chat import models
-    from .summarizers.lveval import summarizer
-
-models[0][
-    'path'
-] = '/path/to/your/huggingface_models/Llama-2-7b-chat-hf'
-models[0][
-    'tokenizer_path'
-] = '/path/to/your/huggingface_models/Llama-2-7b-chat-hf'
-models[0]['max_seq_len'] = 4096
-models[0]['generation_kwargs'] = dict(do_sample=False)
-models[0]['mode'] = 'mid'  # truncate in the middle

configs/eval_mixtral_8x7b.py

@@ -1,8 +0,0 @@
-from mmengine.config import read_base
-
-with read_base():
-    from .datasets.collections.base_medium_llama import piqa_datasets, siqa_datasets
-    from .models.mixtral.mixtral_8x7b_32k import models
-
-
-datasets = [*piqa_datasets, *siqa_datasets]

configs/eval_needlebench.py

@@ -1,26 +0,0 @@
-from mmengine.config import read_base
-with read_base():
-    from .models.hf_internlm.lmdeploy_internlm2_chat_7b import models as internlm2_chat_7b_200k
-    from .models.hf_internlm.hf_internlm2_chat_7b import models as internlm2_chat_7b
-
-    # Evaluate needlebench_4k, adjust the configuration to use 8k, 32k, 128k, 200k, or 1000k if necessary.
-    # from .datasets.needlebench.needlebench_4k.needlebench_4k import needlebench_datasets
-    # from .summarizers.needlebench import needlebench_4k_summarizer as summarizer
-
-    # only eval original "needle in a haystack test" in needlebench_4k
-    from .datasets.needlebench.needlebench_4k.needlebench_single_4k import needlebench_zh_datasets, needlebench_en_datasets
-    from .summarizers.needlebench import needlebench_4k_summarizer as summarizer
-
-    # eval Ancestral Tracing Challenge(ATC)
-    # from .datasets.needlebench.atc.atc_choice_50 import needlebench_datasets
-    # from .summarizers.needlebench import atc_summarizer_50 as summarizer
-
-datasets = sum([v for k, v in locals().items() if ('datasets' in k)], [])
-
-for m in internlm2_chat_7b:
-    m['max_seq_len'] = 32768 # Ensure InternLM2-7B model can receive the full length of long texts, adjust for other models based on their supported maximum sequence length.
-    m['max_out_len'] = 2000 # Ensure complete responses from the model in multi-needle retrieval tasks.
-
-models = internlm2_chat_7b
-
-work_dir = './outputs/needlebench'

configs/eval_qwen_7b_chat_lawbench.py

@@ -1,11 +0,0 @@
-from mmengine.config import read_base
-
-with read_base():
-    from .models.qwen.hf_qwen_7b_chat import models
-    from .datasets.lawbench.lawbench_zero_shot_gen_002588 import lawbench_datasets as lawbench_zero_shot_datasets
-    from .datasets.lawbench.lawbench_one_shot_gen_002588 import lawbench_datasets as lawbench_one_shot_datasets
-    from .summarizers.lawbench import summarizer
-
-datasets = lawbench_zero_shot_datasets + lawbench_one_shot_datasets
-for d in datasets:
-    d['infer_cfg']['inferencer']['save_every'] = 1

configs/eval_rwkv5_3b.py

@@ -1,6 +0,0 @@
-from mmengine.config import read_base
-
-with read_base():
-    from .models.rwkv.rwkv5_3b import models
-    from .datasets.collections.base_medium_llama import datasets
-    from .summarizers.leaderboard import summarizer

configs/eval_subjective_alignbench.py

@@ -1,79 +0,0 @@
-from mmengine.config import read_base
-
-with read_base():
-    from .datasets.subjective.alignbench.alignbench_judgeby_critiquellm import subjective_datasets
-
-from opencompass.models import HuggingFaceCausalLM, HuggingFace, HuggingFaceChatGLM3, OpenAI
-from opencompass.partitioners import NaivePartitioner, SizePartitioner
-from opencompass.partitioners.sub_naive import SubjectiveNaivePartitioner
-from opencompass.partitioners.sub_size import SubjectiveSizePartitioner
-from opencompass.runners import LocalRunner
-from opencompass.runners import SlurmSequentialRunner
-from opencompass.tasks import OpenICLInferTask
-from opencompass.tasks.subjective_eval import SubjectiveEvalTask
-from opencompass.summarizers import AlignmentBenchSummarizer
-
-api_meta_template = dict(
-    round=[
-        dict(role='HUMAN', api_role='HUMAN'),
-        dict(role='BOT', api_role='BOT', generate=True),
-    ]
-)
-
-# -------------Inference Stage ----------------------------------------
-# For subjective evaluation, we often set do sample for models
-models = [
-    dict(
-        type=HuggingFaceChatGLM3,
-        abbr='chatglm3-6b-hf',
-        path='THUDM/chatglm3-6b',
-        tokenizer_path='THUDM/chatglm3-6b',
-        model_kwargs=dict(
-            device_map='auto',
-            trust_remote_code=True,
-        ),
-        tokenizer_kwargs=dict(
-            padding_side='left',
-            truncation_side='left',
-            trust_remote_code=True,
-        ),
-        generation_kwargs=dict(
-            do_sample=True,
-        ),
-        meta_template=api_meta_template,
-        max_out_len=2048,
-        max_seq_len=4096,
-        batch_size=8,
-        run_cfg=dict(num_gpus=1, num_procs=1),
-    )
-]
-
-datasets = [*subjective_datasets]
-
-# -------------Evalation Stage ----------------------------------------
-
-## ------------- JudgeLLM Configuration
-judge_models = [dict(
-    abbr='GPT4-Turbo',
-    type=OpenAI,
-    path='gpt-4-1106-preview',
-    key='xxxx',  # The key will be obtained from $OPENAI_API_KEY, but you can write down your key here as well
-    meta_template=api_meta_template,
-    query_per_second=16,
-    max_out_len=2048,
-    max_seq_len=2048,
-    batch_size=8,
-    temperature=0,
-)]
-
-## ------------- Evaluation Configuration
-eval = dict(
-    partitioner=dict(
-        type=SubjectiveSizePartitioner, max_task_size=1000, mode='singlescore', models=models, judge_models=judge_models,
-    ),
-    runner=dict(type=LocalRunner, max_num_workers=2, task=dict(type=SubjectiveEvalTask)),
-)
-
-summarizer = dict(type=AlignmentBenchSummarizer, judge_type='general')
-
-work_dir = 'outputs/alignment_bench/'

configs/eval_subjective_alpacaeval_oc.py

@@ -1,104 +0,0 @@
-from mmengine.config import read_base
-
-with read_base():
-    from .datasets.subjective.alpaca_eval.alpacav1_judgeby_gpt4 import subjective_datasets as alpacav1
-    from .datasets.subjective.alpaca_eval.alpacav2_judgeby_gpt4 import subjective_datasets as alpacav2
-
-from opencompass.models import HuggingFaceCausalLM, HuggingFace, HuggingFaceChatGLM3
-from opencompass.models.openai_api import OpenAI
-from opencompass.partitioners import NaivePartitioner, SizePartitioner
-from opencompass.partitioners.sub_naive import SubjectiveNaivePartitioner
-from opencompass.partitioners.sub_size import SubjectiveSizePartitioner
-from opencompass.runners import LocalRunner
-from opencompass.runners import SlurmSequentialRunner
-from opencompass.tasks import OpenICLInferTask
-from opencompass.tasks.subjective_eval import SubjectiveEvalTask
-from opencompass.summarizers import AlpacaSummarizer
-
-api_meta_template = dict(
-    round=[
-        dict(role='HUMAN', api_role='HUMAN'),
-        dict(role='BOT', api_role='BOT', generate=True),
-    ],
-    reserved_roles=[dict(role='SYSTEM', api_role='SYSTEM')],
-)
-
-# -------------Inference Stage ----------------------------------------
-
-# For subjective evaluation, we often set do sample for models
-models = [
-    dict(
-        type=HuggingFaceChatGLM3,
-        abbr='chatglm3-6b-hf',
-        path='THUDM/chatglm3-6b',
-        tokenizer_path='THUDM/chatglm3-6b',
-        model_kwargs=dict(
-            device_map='auto',
-            trust_remote_code=True,
-        ),
-        tokenizer_kwargs=dict(
-            padding_side='left',
-            truncation_side='left',
-            trust_remote_code=True,
-        ),
-        generation_kwargs=dict(
-            do_sample=True,
-        ),
-        meta_template=api_meta_template,
-        max_out_len=2048,
-        max_seq_len=4096,
-        batch_size=8,
-        run_cfg=dict(num_gpus=1, num_procs=1),
-    )
-]
-
-datasets = [*alpacav2]
-
-gpt4 = dict(
-    abbr='gpt4-turbo',
-    type=OpenAI,
-    path='gpt-4-1106-preview',
-    key='',  # The key will be obtained from $OPENAI_API_KEY, but you can write down your key here as well
-    meta_template=api_meta_template,
-    query_per_second=1,
-    max_out_len=2048,
-    max_seq_len=4096,
-    batch_size=4,
-    retry=20,
-    temperature=1,
-)  # Re-inference gpt4's predictions or you can choose to use the pre-commited gpt4's predictions
-
-
-
-# -------------Evalation Stage ----------------------------------------
-
-## ------------- JudgeLLM Configuration
-judge_models = [dict(
-    abbr='GPT4-Turbo',
-    type=OpenAI,
-    path='gpt-4-1106-preview',
-    key='',  # The key will be obtained from $OPENAI_API_KEY, but you can write down your key here as well
-    meta_template=api_meta_template,
-    query_per_second=1,
-    max_out_len=1024,
-    max_seq_len=4096,
-    batch_size=2,
-    retry=20,
-    temperature=0,
-)]
-
-## ------------- Evaluation Configuration
-eval = dict(
-    partitioner=dict(
-        type=SubjectiveSizePartitioner, max_task_size=1000, mode='m2n', base_models=[gpt4], compare_models=models,
-        infer_order='random',
-        judge_models=judge_models
-    ),
-    runner=dict(type=LocalRunner, max_num_workers=2, task=dict(type=SubjectiveEvalTask)),
-    given_pred = [{'abbr':'gpt4-turbo', 'path':''}]
-)
-work_dir = 'outputs/alpaca/'
-
-
-
-summarizer = dict(type=AlpacaSummarizer, judge_type='v2')

configs/eval_subjective_arena_hard.py

@@ -1,104 +0,0 @@
-from opencompass.models import HuggingFaceCausalLM
-from copy import deepcopy
-from opencompass.models import TurboMindModel
-from mmengine.config import read_base
-
-from opencompass.models import HuggingFaceCausalLM, HuggingFace, HuggingFaceChatGLM3, OpenAI
-from opencompass.partitioners import NaivePartitioner, SizePartitioner
-from opencompass.partitioners.sub_naive import SubjectiveNaivePartitioner
-from opencompass.partitioners.sub_size import SubjectiveSizePartitioner
-from opencompass.runners import LocalRunner
-from opencompass.runners import SlurmSequentialRunner
-from opencompass.tasks import OpenICLInferTask
-from opencompass.tasks.subjective_eval import SubjectiveEvalTask
-from opencompass.summarizers import ArenaHardSummarizer
-
-with read_base():
-    from .datasets.subjective.arena_hard.arena_hard_compare import subjective_datasets
-
-api_meta_template = dict(
-    round=[
-        dict(role='HUMAN', api_role='HUMAN'),
-        dict(role='BOT', api_role='BOT', generate=True),
-    ]
-)
-
-_meta_template = dict(
-    round=[
-        dict(role='HUMAN', begin='<|begin_of_text|>user<|end_header_id|>\n\n', end='<|eot_id|>'),
-        dict(role='BOT', begin='<|begin_of_text|>assistant<|end_header_id|>\n\n', end='<|eot_id|>', generate=True),
-    ],
-)
-
-models = [
-    dict(
-        type=HuggingFaceCausalLM,
-        abbr='llama-3-8b-instruct-hf',
-        path='meta-llama/Meta-Llama-3-8B-Instruct',
-        model_kwargs=dict(device_map='auto'),
-        tokenizer_kwargs=dict(
-            padding_side='left',
-            truncation_side='left',
-            use_fast=False,
-        ),
-        meta_template=_meta_template,
-        max_out_len=4096,
-        max_seq_len=2048,
-        batch_size=8,
-        run_cfg=dict(num_gpus=1, num_procs=1),
-        generation_kwargs={'eos_token_id': [128001, 128009]},
-        batch_padding=True,
-    )
-]
-
-datasets = [*subjective_datasets]
-
-work_dir = 'outputs/arena_hard/'
-# -------------Inferen Stage ----------------------------------------
-
-
-infer = dict(
-    partitioner=dict(type=SizePartitioner, max_task_size=1000000),
-    runner=dict(
-        type=LocalRunner,
-        max_num_workers=32,
-        task=dict(type=OpenICLInferTask)),
-)
-
-judge_models = [dict(
-    abbr='GPT4-Turbo',
-    type=OpenAI,
-    path='gpt-4-1106-preview',
-    key='',
-        meta_template=api_meta_template,
-        query_per_second=1,
-        max_out_len=5120,
-        max_seq_len=9216,
-        batch_size=10,
-        retry=10,
-        temperature = 0,
-)]
-
-## ------------- Evaluation Configuration
-gpt4_0314 = dict(
-    abbr='gpt4-0314',
-    type=OpenAI,
-)
-
-eval = dict(
-    partitioner=dict(
-        type=SubjectiveSizePartitioner,
-        max_task_size=1000000,
-        mode='m2n',
-        infer_order='double',
-        base_models=[gpt4_0314],
-        compare_models=models,
-        judge_models=judge_models,
-    ),
-    runner=dict(type=LocalRunner, max_num_workers=16, task=dict(type=SubjectiveEvalTask)),
-    given_pred = [{'abbr':'gpt4-0314', 'path':''}]
-)
-
-summarizer = dict(
-    type=ArenaHardSummarizer
-)

configs/eval_subjective_compassarena.py

@@ -1,106 +0,0 @@
-from os import getenv as gv
-from opencompass.models import HuggingFaceCausalLM
-from mmengine.config import read_base
-
-with read_base():
-    from .datasets.subjective.compassarena.compassarena_compare import subjective_datasets
-
-from opencompass.models import HuggingFaceCausalLM, HuggingFace, HuggingFaceChatGLM3, OpenAI
-from opencompass.partitioners import NaivePartitioner, SizePartitioner
-from opencompass.partitioners.sub_naive import SubjectiveNaivePartitioner
-from opencompass.partitioners.sub_size import SubjectiveSizePartitioner
-from opencompass.runners import LocalRunner
-from opencompass.runners import SlurmSequentialRunner
-from opencompass.tasks import OpenICLInferTask
-from opencompass.tasks.subjective_eval import SubjectiveEvalTask
-from opencompass.summarizers import CompassArenaSummarizer
-
-api_meta_template = dict(
-    round=[
-        dict(role='HUMAN', api_role='HUMAN'),
-        dict(role='BOT', api_role='BOT', generate=True),
-    ],
-    reserved_roles=[dict(role='SYSTEM', api_role='SYSTEM')],
-)
-
-# -------------Inference Stage ----------------------------------------
-
-# For subjective evaluation, we often set do sample for models
-models = [
-    dict(
-        type=HuggingFaceChatGLM3,
-        abbr='chatglm3-6b-hf',
-        path='THUDM/chatglm3-6b',
-        tokenizer_path='THUDM/chatglm3-6b',
-        model_kwargs=dict(
-            device_map='auto',
-            trust_remote_code=True,
-        ),
-        tokenizer_kwargs=dict(
-            padding_side='left',
-            truncation_side='left',
-            trust_remote_code=True,
-        ),
-        generation_kwargs=dict(
-            do_sample=True,
-        ),
-        meta_template=api_meta_template,
-        max_out_len=2048,
-        max_seq_len=4096,
-        batch_size=1,
-        run_cfg=dict(num_gpus=1, num_procs=1),
-    )
-]
-
-datasets = [*subjective_datasets]
-
-gpt4 = dict(
-    abbr='gpt4-turbo',
-    type=OpenAI,
-    path='gpt-4-1106-preview',
-    key='',  # The key will be obtained from $OPENAI_API_KEY, but you can write down your key here as well
-    meta_template=api_meta_template,
-    query_per_second=1,
-    max_out_len=2048,
-    max_seq_len=4096,
-    batch_size=4,
-    retry=20,
-    temperature=1,
-)  # Re-inference gpt4's predictions or you can choose to use the pre-commited gpt4's predictions
-
-# -------------Evalation Stage ----------------------------------------
-
-## ------------- JudgeLLM Configuration
-judge_models = [dict(
-    abbr='GPT4-Turbo',
-    type=OpenAI,
-    path='gpt-4-1106-preview',
-    key='',  # The key will be obtained from $OPENAI_API_KEY, but you can write down your key here as well
-    meta_template=api_meta_template,
-    query_per_second=1,
-    max_out_len=1024,
-    max_seq_len=4096,
-    batch_size=2,
-    retry=20,
-    temperature=0,
-)]
-
-## ------------- Evaluation Configuration
-eval = dict(
-    partitioner=dict(
-        type=SubjectiveSizePartitioner,
-        strategy='split',
-        max_task_size=10000,
-        mode='m2n',
-        infer_order='double',
-        base_models=[gpt4],
-        compare_models=models,
-        judge_models=judge_models,
-    ),
-    runner=dict(type=LocalRunner, max_num_workers=2, task=dict(type=SubjectiveEvalTask)),
-    given_pred = [{'abbr':'gpt4-turbo', 'path':''}]
-)
-
-work_dir = 'outputs/compass_arena_debug/'
-
-summarizer = dict(type=CompassArenaSummarizer, summary_type='half_add')

configs/eval_subjective_compassbench.py

@@ -1,137 +0,0 @@
-from os import getenv as gv
-from opencompass.models import HuggingFaceCausalLM
-from mmengine.config import read_base
-
-with read_base():
-    from .datasets.subjective.compassbench.compassbench_compare import subjective_datasets
-
-from opencompass.models import HuggingFaceCausalLM, HuggingFace, HuggingFaceChatGLM3, OpenAI
-from opencompass.partitioners import NaivePartitioner, SizePartitioner
-from opencompass.partitioners.sub_naive import SubjectiveNaivePartitioner
-from opencompass.partitioners.sub_size import SubjectiveSizePartitioner
-from opencompass.runners import LocalRunner
-from opencompass.runners import SlurmSequentialRunner
-from opencompass.tasks import OpenICLInferTask
-from opencompass.tasks.subjective_eval import SubjectiveEvalTask
-from opencompass.summarizers import CompassBenchSummarizer
-
-api_meta_template = dict(
-    round=[
-        dict(role='HUMAN', api_role='HUMAN'),
-        dict(role='BOT', api_role='BOT', generate=True),
-    ],
-    reserved_roles=[dict(role='SYSTEM', api_role='SYSTEM')],
-)
-
-# -------------Inference Stage ----------------------------------------
-
-from opencompass.models import HuggingFacewithChatTemplate
-
-models = [
-    dict(
-        type=HuggingFacewithChatTemplate,
-        abbr='internlm2-chat-7b-hf',
-        path='internlm/internlm2-chat-7b',
-        max_out_len=1024,
-        batch_size=8,
-        run_cfg=dict(num_gpus=1),
-        stop_words=['</s>', '<|im_end|>'],
-        generation_kwargs=dict(
-            do_sample=True,
-        ),
-    )
-]
-
-datasets = [*subjective_datasets]
-
-infer = dict(
-    partitioner=dict(type=NaivePartitioner),
-    runner=dict(
-        type=SlurmSequentialRunner,
-        partition='llmeval',
-        quotatype='reserved',
-        max_num_workers=256,
-        task=dict(type=OpenICLInferTask),
-    ),
-)
-
-gpt4 = dict(
-    abbr='gpt4-turbo',
-    type=OpenAI,
-    path='gpt-4-1106-preview',
-    key='',  # The key will be obtained from $OPENAI_API_KEY, but you can write down your key here as well
-    meta_template=api_meta_template,
-    query_per_second=1,
-    max_out_len=2048,
-    max_seq_len=4096,
-    batch_size=4,
-    retry=20,
-    temperature=1,
-)  # Re-inference gpt4's predictions or you can choose to use the pre-commited gpt4's predictions
-
-# -------------Evalation Stage ----------------------------------------
-
-## ------------- JudgeLLM Configuration
-judge_models = [dict(
-    abbr='GPT4-Turbo',
-    type=OpenAI,
-    path='gpt-4-1106-preview',
-    key='',  # The key will be obtained from $OPENAI_API_KEY, but you can write down your key here as well
-    meta_template=api_meta_template,
-    query_per_second=1,
-    max_out_len=1024,
-    max_seq_len=4096,
-    batch_size=2,
-    retry=20,
-    temperature=0,
-)]
-
-judge_models = [
-    dict(
-        type=HuggingFacewithChatTemplate,
-        abbr='internlm102b',
-        path='/mnt/petrelfs/caomaosong/backup_hwfile/100bjudge_6w_epoch1/hf',
-        max_out_len=1024,
-        batch_size=8,
-        run_cfg=dict(num_gpus=4),
-        stop_words=['</s>', '<|im_end|>'],
-    ),
-    dict(
-        type=HuggingFacewithChatTemplate,
-        abbr='internlm102b2',
-        path='/mnt/petrelfs/caomaosong/backup_hwfile/100bjudge_6w_epoch1/hf',
-        max_out_len=1024,
-        batch_size=8,
-        run_cfg=dict(num_gpus=4),
-        stop_words=['</s>', '<|im_end|>'],
-    ),
-    dict(
-        type=HuggingFacewithChatTemplate,
-        abbr='internlm102b3',
-        path='/mnt/petrelfs/caomaosong/backup_hwfile/100bjudge_6w_epoch1/hf',
-        max_out_len=1024,
-        batch_size=8,
-        run_cfg=dict(num_gpus=4),
-        stop_words=['</s>', '<|im_end|>'],
-    )
-]
-
-## ------------- Evaluation Configuration
-eval = dict(
-    partitioner=dict(
-        type=SubjectiveSizePartitioner,
-        strategy='split',
-        max_task_size=10000000,
-        mode='m2n',
-        infer_order='double',
-        base_models=[gpt4],
-        compare_models=models,
-        judge_models=judge_models,
-    ),
-    runner=dict(type=LocalRunner, max_num_workers=32, task=dict(type=SubjectiveEvalTask)),
-    #given_pred = [{'abbr':'gpt4-turbo', 'path':''}]
-)
-
-work_dir = 'outputs/compassbench/'
-
-summarizer = dict(type=CompassBenchSummarizer, summary_type='half_add')

configs/eval_subjective_creationbench.py

@@ -1,77 +0,0 @@
-from mmengine.config import read_base
-
-with read_base():
-    from .datasets.subjective.creationbench.creationbench_judgeby_gpt4_withref import subjective_datasets
-
-from opencompass.models import HuggingFaceCausalLM, HuggingFace, HuggingFaceChatGLM3, OpenAI
-from opencompass.partitioners import NaivePartitioner, SizePartitioner
-from opencompass.partitioners.sub_naive import SubjectiveNaivePartitioner
-from opencompass.partitioners.sub_size import SubjectiveSizePartitioner
-from opencompass.runners import LocalRunner
-from opencompass.runners import SlurmSequentialRunner
-from opencompass.tasks import OpenICLInferTask
-from opencompass.tasks.subjective_eval import SubjectiveEvalTask
-from opencompass.summarizers import CreationBenchSummarizer
-
-api_meta_template = dict(
-    round=[
-        dict(role='HUMAN', api_role='HUMAN'),
-        dict(role='BOT', api_role='BOT', generate=True),
-    ]
-)
-
-# -------------Inference Stage ----------------------------------------
-# For subjective evaluation, we often set do sample for models
-models = [
-    dict(
-        type=HuggingFaceChatGLM3,
-        abbr='chatglm3-6b-hf',
-        path='THUDM/chatglm3-6b',
-        tokenizer_path='THUDM/chatglm3-6b',
-        model_kwargs=dict(
-            device_map='auto',
-            trust_remote_code=True,
-        ),
-        tokenizer_kwargs=dict(
-            padding_side='left',
-            truncation_side='left',
-            trust_remote_code=True,
-        ),
-        generation_kwargs=dict(
-            do_sample=True,
-        ),
-        meta_template=api_meta_template,
-        max_out_len=2048,
-        max_seq_len=4096,
-        batch_size=1,
-        run_cfg=dict(num_gpus=1, num_procs=1),
-    )
-]
-
-datasets = [*subjective_datasets]
-
-# -------------Evalation Stage ----------------------------------------
-
-## ------------- JudgeLLM Configuration
-judge_model = dict(
-    abbr='GPT4-Turbo',
-    type=OpenAI,
-    path='gpt-4-1106-preview',
-    key='xxxx',  # The key will be obtained from $OPENAI_API_KEY, but you can write down your key here as well
-    meta_template=api_meta_template,
-    query_per_second=16,
-    max_out_len=2048,
-    max_seq_len=2048,
-    batch_size=8,
-    temperature=0,
-)
-
-## ------------- Evaluation Configuration
-eval = dict(
-    partitioner=dict(type=SubjectiveNaivePartitioner, mode='singlescore', models=models),
-    runner=dict(type=LocalRunner, max_num_workers=2, task=dict(type=SubjectiveEvalTask, judge_cfg=judge_model)),
-)
-
-summarizer = dict(type=CreationBenchSummarizer, judge_type='general')
-
-work_dir = 'outputs/creationbench/'

configs/eval_subjective_functional_multiround.py

@@ -1,111 +0,0 @@
-from opencompass.models import HuggingFaceCausalLM
-from copy import deepcopy
-from opencompass.models import TurboMindModel
-from mmengine.config import read_base
-
-from opencompass.models import HuggingFaceCausalLM, HuggingFace, HuggingFaceChatGLM3, OpenAI
-from opencompass.partitioners import NaivePartitioner, SizePartitioner
-from opencompass.partitioners.sub_naive import SubjectiveNaivePartitioner
-from opencompass.partitioners.sub_size import SubjectiveSizePartitioner
-from opencompass.runners import LocalRunner
-from opencompass.runners import SlurmSequentialRunner
-from opencompass.tasks import OpenICLInferTask
-from opencompass.tasks.subjective_eval import SubjectiveEvalTask
-from opencompass.summarizers import MultiroundSummarizer
-
-with read_base():
-    from .datasets.subjective.multiround.functionalmt_zh_judgeby_gpt4 import subjective_datasets
-
-api_meta_template = dict(
-    round=[
-        dict(role='HUMAN', api_role='HUMAN'),
-        dict(role='BOT', api_role='BOT', generate=True),
-    ]
-)
-
-_meta_template = dict(
-    round=[
-        dict(role='HUMAN', begin='<|im_start|>user\n', end='<|im_end|>\n'),
-        dict(role='BOT', begin='<|im_start|>assistant\n', end='<|im_end|>\n', generate=True),
-    ],
-    eos_token_id=151645,
-)
-
-models = [
-    dict(
-        type=HuggingFaceCausalLM,
-        abbr='qwen1.5-7b-chat-hf',
-        path='Qwen/Qwen1.5-7B-Chat',
-        model_kwargs=dict(
-            device_map='auto',
-            trust_remote_code=True
-        ),
-        tokenizer_kwargs=dict(
-            padding_side='left',
-            truncation_side='left',
-            trust_remote_code=True,
-            use_fast=False,
-        ),
-        generation_kwargs=dict(
-            do_sample=True,
-        ),
-        meta_template=_meta_template,
-        pad_token_id=151645,
-        max_out_len=100,
-        max_seq_len=2048,
-        batch_size=8,
-        run_cfg=dict(num_gpus=1, num_procs=1),
-        end_str='<|im_end|>',
-    )
-]
-
-datasets = [*subjective_datasets]
-
-work_dir = 'outputs/multiround/'
-# -------------Inferen Stage ----------------------------------------
-
-
-infer = dict(
-    partitioner=dict(type=SizePartitioner, max_task_size=1000),
-    runner=dict(
-        type=SlurmSequentialRunner,
-        partition='your part',
-        quotatype='auto',
-        max_num_workers=256,
-        task=dict(type=OpenICLInferTask)),
-)
-
-judge_models = [dict(
-    abbr='GPT4-Turbo',
-    type=OpenAI,
-    path='gpt-4-1106-preview',
-    key='',
-        meta_template=api_meta_template,
-        query_per_second=1,
-        max_out_len=1024,
-        max_seq_len=4096,
-        batch_size=10,
-        retry=10,
-        temperature = 0,
-)]
-
-## ------------- Evaluation Configuration
-eval = dict(
-    partitioner=dict(
-        type=SubjectiveSizePartitioner,
-        max_task_size=1000,
-        mode='singlescore',
-        models = models,
-        judge_models=judge_models
-    ),
-    runner=dict(
-        type=SlurmSequentialRunner,
-        partition='your part',
-        quotatype='auto',
-        max_num_workers=256,
-        task=dict(type=SubjectiveEvalTask)),
-)
-
-summarizer = dict(
-    type=MultiroundSummarizer
-)

configs/eval_subjective_judge_pandalm.py

@@ -1,92 +0,0 @@
-from mmengine.config import read_base
-
-with read_base():
-    from .datasets.subjective.alignbench.alignbench_judgeby_critiquellm import subjective_datasets
-
-from opencompass.models import HuggingFaceCausalLM, HuggingFace, HuggingFaceChatGLM3
-from opencompass.partitioners import NaivePartitioner
-from opencompass.partitioners.sub_naive import SubjectiveNaivePartitioner
-from opencompass.runners import LocalRunner
-from opencompass.runners import SlurmSequentialRunner
-from opencompass.tasks import OpenICLInferTask
-from opencompass.tasks.subjective_eval import SubjectiveEvalTask
-from opencompass.summarizers import AlignmentBenchSummarizer
-
-api_meta_template = dict(
-    round=[
-        dict(role='HUMAN', api_role='HUMAN'),
-        dict(role='BOT', api_role='BOT', generate=True),
-    ]
-)
-
-# -------------Inference Stage ----------------------------------------
-# For subjective evaluation, we often set do sample for models
-models = [
-    dict(
-        type=HuggingFaceChatGLM3,
-        abbr='chatglm3-6b-hf',
-        path='THUDM/chatglm3-6b',
-        tokenizer_path='THUDM/chatglm3-6b',
-        model_kwargs=dict(
-            device_map='auto',
-            trust_remote_code=True,
-        ),
-        tokenizer_kwargs=dict(
-            padding_side='left',
-            truncation_side='left',
-            trust_remote_code=True,
-        ),
-        generation_kwargs=dict(
-            do_sample=True,
-        ),
-        meta_template=api_meta_template,
-        max_out_len=2048,
-        max_seq_len=4096,
-        batch_size=1,
-        run_cfg=dict(num_gpus=1, num_procs=1),
-    )
-]
-
-datasets = [*subjective_datasets]
-
-infer = dict(
-    partitioner=dict(type=NaivePartitioner),
-    runner=dict(
-        type=SlurmSequentialRunner,
-        partition='llmeval',
-        quotatype='auto',
-        max_num_workers=256,
-        task=dict(type=OpenICLInferTask),
-    ),
-)
-
-# -------------Evalation Stage ----------------------------------------
-
-## ------------- JudgeLLM Configuration
-judge_models = [dict(
-    type=HuggingFaceCausalLM,
-    abbr='pandalm-7b-v1-hf',
-    path='WeOpenML/PandaLM-7B-v1',
-    tokenizer_path='WeOpenML/PandaLM-7B-v1',
-    tokenizer_kwargs=dict(
-        padding_side='left',
-        truncation_side='left',
-        trust_remote_code=True,
-        use_fast=False,
-    ),
-    max_out_len=512,
-    max_seq_len=2048,
-    batch_size=8,
-    model_kwargs=dict(device_map='auto', trust_remote_code=True),
-    run_cfg=dict(num_gpus=1, num_procs=1),
-)]
-
-## ------------- Evaluation Configuration
-eval = dict(
-    partitioner=dict(type=SubjectiveNaivePartitioner, mode='singlescore', models=models, judge_models=judge_models),
-    runner=dict(type=LocalRunner, max_num_workers=2, task=dict(type=SubjectiveEvalTask)),
-)
-
-summarizer = dict(type=AlignmentBenchSummarizer)
-
-work_dir = 'outputs/pandalm'

configs/eval_subjective_mtbench.py

@@ -1,84 +0,0 @@
-from mmengine.config import read_base
-
-with read_base():
-    from .datasets.subjective.multiround.mtbench_single_judge_diff_temp import subjective_datasets
-
-from opencompass.models import HuggingFaceCausalLM, HuggingFace, HuggingFaceChatGLM3, OpenAI
-from opencompass.partitioners import NaivePartitioner, SizePartitioner
-from opencompass.partitioners.sub_naive import SubjectiveNaivePartitioner
-from opencompass.partitioners.sub_size import SubjectiveSizePartitioner
-from opencompass.runners import LocalRunner
-from opencompass.runners import SlurmSequentialRunner
-from opencompass.tasks import OpenICLInferTask
-from opencompass.tasks.subjective_eval import SubjectiveEvalTask
-from opencompass.summarizers import MTBenchSummarizer
-
-api_meta_template = dict(
-    round=[
-        dict(role='SYSTEM', api_role='SYSTEM'),
-        dict(role='HUMAN', api_role='HUMAN'),
-        dict(role='BOT', api_role='BOT', generate=True),
-    ]
-)
-
-_meta_template = dict(
-    round=[
-        dict(role='HUMAN', begin='\n<|im_start|>user\n', end='<|im_end|>'),
-        dict(role='BOT', begin='\n<|im_start|>assistant\n', end='<|im_end|>', generate=True),
-    ],
-)
-# -------------Inference Stage ----------------------------------------
-# For subjective evaluation, we often set do sample for models
-models = [
-    dict(
-        type=HuggingFaceCausalLM,
-        abbr='qwen-7b-chat-hf',
-        path='Qwen/Qwen-7B-Chat',
-        tokenizer_path='Qwen/Qwen-7B-Chat',
-        model_kwargs=dict(
-            device_map='auto',
-            trust_remote_code=True
-        ),
-        tokenizer_kwargs=dict(
-            padding_side='left',
-            truncation_side='left',
-            trust_remote_code=True,
-            use_fast=False,
-        ),
-        pad_token_id=151643,
-        max_out_len=100,
-        max_seq_len=2048,
-        batch_size=8,
-        meta_template=_meta_template,
-        run_cfg=dict(num_gpus=1, num_procs=1),
-        end_str='<|im_end|>',
-    )
-]
-
-datasets = [*subjective_datasets]
-
-# -------------Evalation Stage ----------------------------------------
-
-## ------------- JudgeLLM Configuration
-judge_models = [dict(
-    abbr='GPT4-Turbo',
-    type=OpenAI,
-    path='gpt-4-0613', # To compare with the official leaderboard, please use gpt4-0613
-    key='xxxx',  # The key will be obtained from $OPENAI_API_KEY, but you can write down your key here as well
-    meta_template=api_meta_template,
-    query_per_second=16,
-    max_out_len=2048,
-    max_seq_len=2048,
-    batch_size=8,
-    temperature=0,
-)]
-
-## single evaluation
-eval = dict(
-    partitioner=dict(type=SubjectiveSizePartitioner, strategy='split', max_task_size=10000, mode='singlescore', models=models, judge_models=judge_models),
-    runner=dict(type=LocalRunner, max_num_workers=32, task=dict(type=SubjectiveEvalTask)),
-)
-
-summarizer = dict(type=MTBenchSummarizer, judge_type='single')
-
-work_dir = 'outputs/mtbench/'

configs/models/hf_internlm/lmdeploy_internlm2_20b.py

@@ -1,27 +0,0 @@
-from opencompass.models.turbomind import TurboMindModel
-
-
-models = [
-    dict(
-        type=TurboMindModel,
-        abbr='internlm2-20b-turbomind',
-        path='internlm/internlm2-20b',
-        engine_config=dict(
-            session_len=32768,
-            max_batch_size=32,
-            model_name='internlm2-20b',
-            tp=2,
-        ),
-        gen_config=dict(
-            top_k=1,
-            top_p=0.8,
-            temperature=1.0,
-            max_new_tokens=2000,
-        ),
-        max_out_len=2000,
-        max_seq_len=32768,
-        batch_size=32,
-        concurrency=8,
-        run_cfg=dict(num_gpus=2, num_procs=1),
-    )
-]

configs/models/hf_internlm/lmdeploy_internlm2_chat_1_8b.py

@@ -1,23 +0,0 @@
-from opencompass.models import TurboMindModelwithChatTemplate
-
-models = [
-    dict(
-        type=TurboMindModelwithChatTemplate,
-        abbr='internlm2-chat-1.8b-turbomind',
-        path='internlm/internlm2-chat-1_8b',
-        engine_config=dict(
-            max_batch_size=16,
-            tp=1,
-        ),
-        gen_config=dict(
-            top_k=1,
-            temperature=1e-6,
-            top_p=0.9,
-        ),
-        max_seq_len=2048,
-        max_out_len=1024,
-        batch_size=32768,
-        run_cfg=dict(num_gpus=1),
-        stop_words=['</s>', '<|im_end|>'],
-    )
-]

configs/models/hf_internlm/lmdeploy_internlm2_chat_20b.py

@@ -1,36 +0,0 @@
-from opencompass.models.turbomind import TurboMindModel
-
-
-_meta_template = dict(
-    round=[
-        dict(role='HUMAN', begin='<|im_start|>user\n', end='<|im_end|>\n'),
-        dict(role='BOT', begin='<|im_start|>assistant\n', end='<|im_end|>\n', generate=True),
-    ],
-)
-
-models = [
-    dict(
-        type=TurboMindModel,
-        abbr='internlm2-chat-20b-turbomind',
-        path='internlm/internlm2-chat-20b',
-        meta_template=_meta_template,
-        engine_config=dict(
-            session_len=32768,
-            max_batch_size=32,
-            model_name='internlm2-chat-20b',
-            tp=2,
-            stop_words=[2, 92542],
-        ),
-        gen_config=dict(
-            top_k=1,
-            top_p=0.8,
-            temperature=1.0,
-            max_new_tokens=2000,
-        ),
-        max_out_len=2000,
-        max_seq_len=32768,
-        batch_size=32,
-        concurrency=8,
-        run_cfg=dict(num_gpus=2, num_procs=1),
-    )
-]

configs/models/hf_internlm/lmdeploy_internlm2_chat_7b.py

@@ -1,23 +0,0 @@
-from opencompass.models import TurboMindModelwithChatTemplate
-
-models = [
-    dict(
-        type=TurboMindModelwithChatTemplate,
-        abbr='internlm2-chat-7b-turbomind',
-        path='internlm/internlm2-chat-7b',
-        engine_config=dict(
-            max_batch_size=16,
-            tp=1,
-        ),
-        gen_config=dict(
-            top_k=1,
-            temperature=1e-6,
-            top_p=0.9,
-        ),
-        max_seq_len=2048,
-        max_out_len=1024,
-        batch_size=32768,
-        run_cfg=dict(num_gpus=1),
-        stop_words=['</s>', '<|im_end|>'],
-    )
-]

configs/models/hf_llama/lmdeploy_llama3_70b_instruct.py

@@ -1,24 +0,0 @@
-from opencompass.models import TurboMindModel
-
-_meta_template = dict(
-    round=[
-        dict(role='HUMAN', begin='<|begin_of_text|>user<|end_header_id|>\n\n', end='<|eot_id|>'),
-        dict(role='BOT', begin='<|begin_of_text|>assistant<|end_header_id|>\n\n', end='<|eot_id|>', generate=True),
-    ],
-)
-
-models = [
-    dict(
-        type=TurboMindModel,
-        abbr='llama-3-70b-instruct-lmdeploy',
-        path='meta-llama/Meta-Llama-3-70B-Instruct',
-        engine_config=dict(session_len=4096, max_batch_size=16, tp=4),
-        gen_config=dict(top_k=1, temperature=1, top_p=0.9, max_new_tokens=1024, stop_words=[128001, 128009]),
-        max_out_len=1024,
-        max_seq_len=4096,
-        batch_size=16,
-        concurrency=16,
-        meta_template=_meta_template,
-        run_cfg=dict(num_gpus=4),
-    )
-]