search

open-compass / opencompass

OpenCompass is an LLM evaluation platform, supporting a wide range of models (Llama3, Mistral, InternLM2,GPT-4,LLaMa2, Qwen,GLM, Claude, etc) over 100+ datasets.
https://opencompass.org.cn/
Apache License 2.0
4.06k stars 429 forks source link

UnboundLocalError: local variable 'prompt_token_num' referenced before assignment and NO OUTPUTS #792

Open qy1026 opened 10 months ago

qy1026 commented 10 months ago

先决条件

问题类型

我正在使用官方支持的任务/模型/数据集进行评估。

环境

{'CUDA available': True,
 'CUDA_HOME': '/usr/local/cuda-12.2',
 'GCC': 'gcc (Ubuntu 9.4.0-1ubuntu1~20.04.3) 9.4.0',
 'GPU 0,1,2,3,4,5,6,7': 'NVIDIA L40',
 'MMEngine': '0.10.2',
 'NVCC': 'Cuda compilation tools, release 12.2, V12.2.91',
 'OpenCV': '4.9.0',
 'PyTorch': '2.1.2+cu121',
 'PyTorch compiling details': 'PyTorch built with:\n'
                              '  - GCC 9.3\n'
                              '  - C++ Version: 201703\n'
                              '  - Intel(R) oneAPI Math Kernel Library Version '
                              '2022.2-Product Build 20220804 for Intel(R) 64 '
                              'architecture applications\n'
                              '  - Intel(R) MKL-DNN v3.1.1 (Git Hash '
                              '64f6bcbcbab628e96f33a62c3e975f8535a7bde4)\n'
                              '  - OpenMP 201511 (a.k.a. OpenMP 4.5)\n'
                              '  - LAPACK is enabled (usually provided by '
                              'MKL)\n'
                              '  - NNPACK is enabled\n'
                              '  - CPU capability usage: AVX512\n'
                              '  - CUDA Runtime 12.1\n'
                              '  - NVCC architecture flags: '
                              '-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75;-gencode;arch=compute_80,code=sm_80;-gencode;arch=compute_86,code=sm_86;-gencode;arch=compute_90,code=sm_90\n'
                              '  - CuDNN 8.9.6  (built against CUDA 12.2)\n'
                              '    - Built with CuDNN 8.9.2\n'
                              '  - Magma 2.6.1\n'
                              '  - Build settings: BLAS_INFO=mkl, '
                              'BUILD_TYPE=Release, CUDA_VERSION=12.1, '
                              'CUDNN_VERSION=8.9.2, '
                              'CXX_COMPILER=/opt/rh/devtoolset-9/root/usr/bin/c++, '
                              'CXX_FLAGS= -D_GLIBCXX_USE_CXX11_ABI=0 '
                              '-fabi-version=11 -fvisibility-inlines-hidden '
                              '-DUSE_PTHREADPOOL -DNDEBUG -DUSE_KINETO '
                              '-DLIBKINETO_NOROCTRACER -DUSE_FBGEMM '
                              '-DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK '
                              '-DUSE_XNNPACK -DSYMBOLICATE_MOBILE_DEBUG_HANDLE '
                              '-O2 -fPIC -Wall -Wextra -Werror=return-type '
                              '-Werror=non-virtual-dtor -Werror=bool-operation '
                              '-Wnarrowing -Wno-missing-field-initializers '
                              '-Wno-type-limits -Wno-array-bounds '
                              '-Wno-unknown-pragmas -Wno-unused-parameter '
                              '-Wno-unused-function -Wno-unused-result '
                              '-Wno-strict-overflow -Wno-strict-aliasing '
                              '-Wno-stringop-overflow -Wno-psabi '
                              '-Wno-error=pedantic -Wno-error=old-style-cast '
                              '-Wno-invalid-partial-specialization '
                              '-Wno-unused-private-field '
                              '-Wno-aligned-allocation-unavailable '
                              '-Wno-missing-braces -fdiagnostics-color=always '
                              '-faligned-new -Wno-unused-but-set-variable '
                              '-Wno-maybe-uninitialized -fno-math-errno '
                              '-fno-trapping-math -Werror=format '
                              '-Werror=cast-function-type '
                              '-Wno-stringop-overflow, LAPACK_INFO=mkl, '
                              'PERF_WITH_AVX=1, PERF_WITH_AVX2=1, '
                              'PERF_WITH_AVX512=1, '
                              'TORCH_DISABLE_GPU_ASSERTS=ON, '
                              'TORCH_VERSION=2.1.2, USE_CUDA=ON, USE_CUDNN=ON, '
                              'USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, '
                              'USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, '
                              'USE_MPI=OFF, USE_NCCL=1, USE_NNPACK=ON, '
                              'USE_OPENMP=ON, USE_ROCM=OFF, \n',
 'Python': '3.10.13 (main, Sep 11 2023, 13:44:35) [GCC 11.2.0]',
 'TorchVision': '0.16.2+cu121',
 'numpy_random_seed': 2147483648,
 'opencompass': '0.2.1+',
 'sys.platform': 'linux'}

重现问题 - 代码/配置示例

skip

重现问题 - 命令或脚本

CUDA_VISIBLE_DEVICES="0,1,2,3" python run.py \
    --hf-path /data1/xxx/swift-qlora-adapter/yi-34b-v13-0102-r64-data3/Yi-34B-xxx-merged2/v0-20240102-091935/checkpoint-37195-merged/ \
    --tokenizer-path /data1/xxx/swift-qlora-adapter/yi-34b-v13-0102-r64-data3/Yi-34B-xxx-merged2/v0-20240102-091935/checkpoint-37195-merged/ \
    --datasets cmmlu_ppl \
    --model-kwargs device_map='auto' trust_remote_code=True \
    --tokenizer-kwargs padding_side='left' truncation='left' trust_remote_code=True \
    --batch-size 2 \
    --work-dir ./yi-34b-merged3/cmmlu_ppl/ \
    --num-gpus 4  \

重现问题 - 错误信息

In /home/xxx/llm_projects/opencompass-main/yi-34b-merged3/cmmlu_ppl/20240112_183510/infer, there are two .out files: infer in which, the error is

01/12 18:35:19 - OpenCompass - INFO - Task [opencompass.models.huggingface.HuggingFace_v0-20240102-091935_checkpoint-37195-merged/cmmlu-professional_law,opencompass.models.huggingface.HuggingFace_v0-20240102-091935_checkpoint-37195-merged/cmmlu-jurisprudence,opencompass.models.huggingface.HuggingFace_v0-20240102-091935_checkpoint-37195-merged/cmmlu-professional_medicine,opencompass.models.huggingface.HuggingFace_v0-20240102-091935_checkpoint-37195-merged/cmmlu-chinese_history,opencompass.models.huggingface.HuggingFace_v0-20240102-091935_checkpoint-37195-merged/cmmlu-college_medicine,opencompass.models.huggingface.HuggingFace_v0-20240102-091935_checkpoint-37195-merged/cmmlu-elementary_chinese,opencompass.models.huggingface.HuggingFace_v0-20240102-091935_checkpoint-37195-merged/cmmlu-elementary_information_and_technology,opencompass.models.huggingface.HuggingFace_v0-20240102-091935_checkpoint-37195-merged/cmmlu-clinical_knowledge,opencompass.models.huggingface.HuggingFace_v0-20240102-091935_checkpoint-37195-merged/cmmlu-professional_psychology,opencompass.models.huggingface.HuggingFace_v0-20240102-091935_checkpoint-37195-merged/cmmlu-elementary_mathematics,opencompass.models.huggingface.HuggingFace_v0-20240102-091935_checkpoint-37195-merged/cmmlu-sociology,opencompass.models.huggingface.HuggingFace_v0-20240102-091935_checkpoint-37195-merged/cmmlu-legal_and_moral_basis,opencompass.models.huggingface.HuggingFace_v0-20240102-091935_checkpoint-37195-merged/cmmlu-management,opencompass.models.huggingface.HuggingFace_v0-20240102-091935_checkpoint-37195-merged/cmmlu-business_ethics,opencompass.models.huggingface.HuggingFace_v0-20240102-091935_checkpoint-37195-merged/cmmlu-chinese_literature,opencompass.models.huggingface.HuggingFace_v0-20240102-091935_checkpoint-37195-merged/cmmlu-computer_science,opencompass.models.huggingface.HuggingFace_v0-20240102-091935_checkpoint-37195-merged/cmmlu-elementary_commonsense,opencompass.models.huggingface.HuggingFace_v0-20240102-091935_checkpoint-37195-merged/cmmlu-marxist_theory,opencompass.models.huggingface.HuggingFace_v0-20240102-091935_checkpoint-37195-merged/cmmlu-international_law,opencompass.models.huggingface.HuggingFace_v0-20240102-091935_checkpoint-37195-merged/cmmlu-traditional_chinese_medicine,opencompass.models.huggingface.HuggingFace_v0-20240102-091935_checkpoint-37195-merged/cmmlu-marketing,opencompass.models.huggingface.HuggingFace_v0-20240102-091935_checkpoint-37195-merged/cmmlu-chinese_teacher_qualification,opencompass.models.huggingface.HuggingFace_v0-20240102-091935_checkpoint-37195-merged/cmmlu-genetics,opencompass.models.huggingface.HuggingFace_v0-20240102-091935_checkpoint-37195-merged/cmmlu-professional_accounting,opencompass.models.huggingface.HuggingFace_v0-20240102-091935_checkpoint-37195-merged/cmmlu-public_relations,opencompass.models.huggingface.HuggingFace_v0-20240102-091935_checkpoint-37195-merged/cmmlu-electrical_engineering,opencompass.models.huggingface.HuggingFace_v0-20240102-091935_checkpoint-37195-merged/cmmlu-journalism,opencompass.models.huggingface.HuggingFace_v0-20240102-091935_checkpoint-37195-merged/cmmlu-computer_security,opencompass.models.huggingface.HuggingFace_v0-20240102-091935_checkpoint-37195-merged/cmmlu-agronomy,opencompass.models.huggingface.HuggingFace_v0-20240102-091935_checkpoint-37195-merged/cmmlu-high_school_biology,opencompass.models.huggingface.HuggingFace_v0-20240102-091935_checkpoint-37195-merged/cmmlu-virology,opencompass.models.huggingface.HuggingFace_v0-20240102-091935_checkpoint-37195-merged/cmmlu-astronomy,opencompass.models.huggingface.HuggingFace_v0-20240102-091935_checkpoint-37195-merged/cmmlu-sports_science,opencompass.models.huggingface.HuggingFace_v0-20240102-091935_checkpoint-37195-merged/cmmlu-ancient_chinese,opencompass.models.huggingface.HuggingFace_v0-20240102-091935_checkpoint-37195-merged/cmmlu-high_school_mathematics,opencompass.models.huggingface.HuggingFace_v0-20240102-091935_checkpoint-37195-merged/cmmlu-education,opencompass.models.huggingface.HuggingFace_v0-20240102-091935_checkpoint-37195-merged/cmmlu-world_history,opencompass.models.huggingface.HuggingFace_v0-20240102-091935_checkpoint-37195-merged/cmmlu-arts,opencompass.models.huggingface.HuggingFace_v0-20240102-091935_checkpoint-37195-merged/cmmlu-chinese_civil_service_exam]

Loading checkpoint shards:   0%|          | 0/15 [00:00<?, ?it/s]
Loading checkpoint shards:   7%|▋         | 1/15 [00:06<01:33,  6.71s/it]
Loading checkpoint shards:  13%|█▎        | 2/15 [00:12<01:23,  6.44s/it]
Loading checkpoint shards:  20%|██        | 3/15 [00:19<01:17,  6.45s/it]
Loading checkpoint shards:  27%|██▋       | 4/15 [00:25<01:09,  6.33s/it]
Loading checkpoint shards:  33%|███▎      | 5/15 [00:31<01:02,  6.25s/it]
Loading checkpoint shards:  40%|████      | 6/15 [00:38<00:56,  6.30s/it]
Loading checkpoint shards:  47%|████▋     | 7/15 [00:44<00:49,  6.22s/it]
Loading checkpoint shards:  53%|█████▎    | 8/15 [00:50<00:43,  6.17s/it]
Loading checkpoint shards:  60%|██████    | 9/15 [00:56<00:37,  6.23s/it]
Loading checkpoint shards:  67%|██████▋   | 10/15 [01:02<00:30,  6.17s/it]
Loading checkpoint shards:  73%|███████▎  | 11/15 [01:08<00:24,  6.14s/it]
Loading checkpoint shards:  80%|████████  | 12/15 [01:15<00:18,  6.21s/it]
Loading checkpoint shards:  87%|████████▋ | 13/15 [01:21<00:12,  6.15s/it]
Loading checkpoint shards:  93%|█████████▎| 14/15 [01:27<00:06,  6.12s/it]
Loading checkpoint shards: 100%|██████████| 15/15 [01:28<00:00,  4.78s/it]
Loading checkpoint shards: 100%|██████████| 15/15 [01:28<00:00,  5.92s/it]
01/12 18:36:51 - OpenCompass - INFO - Start inferencing [opencompass.models.huggingface.HuggingFace_v0-20240102-091935_checkpoint-37195-merged/cmmlu-professional_law]

  0%|          | 0/211 [00:00<?, ?it/s]
100%|██████████| 211/211 [00:00<00:00, 2920785.95it/s]
Traceback (most recent call last):
  File "/home/xxx/llm_projects/opencompass-main/opencompass/tasks/openicl_infer.py", line 150, in <module>
    inferencer.run()
  File "/home/xxx/llm_projects/opencompass-main/opencompass/tasks/openicl_infer.py", line 80, in run
    self._inference()
  File "/home/xxx/llm_projects/opencompass-main/opencompass/tasks/openicl_infer.py", line 128, in _inference
    inferencer.inference(retriever,
  File "/home/xxx/llm_projects/opencompass-main/opencompass/openicl/icl_inferencer/icl_ppl_inferencer.py", line 148, in inference
    token_num_list.append(prompt_token_num)
UnboundLocalError: local variable 'prompt_token_num' referenced before assignment
[2024-01-12 18:36:57,233] torch.distributed.elastic.multiprocessing.api: [ERROR] failed (exitcode: 1) local_rank: 0 (pid: 1867675) of binary: /home/xxx/anaconda3/envs/opencompass/bin/python
Traceback (most recent call last):
  File "/home/xxx/anaconda3/envs/opencompass/bin/torchrun", line 8, in <module>
    sys.exit(main())
  File "/home/xxx/anaconda3/envs/opencompass/lib/python3.10/site-packages/torch/distributed/elastic/multiprocessing/errors/__init__.py", line 346, in wrapper
    return f(*args, **kwargs)
  File "/home/xxx/anaconda3/envs/opencompass/lib/python3.10/site-packages/torch/distributed/run.py", line 806, in main
    run(args)
  File "/home/xxx/anaconda3/envs/opencompass/lib/python3.10/site-packages/torch/distributed/run.py", line 797, in run
    elastic_launch(
  File "/home/xxx/anaconda3/envs/opencompass/lib/python3.10/site-packages/torch/distributed/launcher/api.py", line 134, in __call__
    return launch_agent(self._config, self._entrypoint, list(args))
  File "/home/xxx/anaconda3/envs/opencompass/lib/python3.10/site-packages/torch/distributed/launcher/api.py", line 264, in launch_agent
    raise ChildFailedError(
torch.distributed.elastic.multiprocessing.errors.ChildFailedError: 
============================================================
/home/xxx/llm_projects/opencompass-main/opencompass/tasks/openicl_infer.py FAILED
------------------------------------------------------------
Failures:
  <NO_OTHER_FAILURES>
------------------------------------------------------------
Root Cause (first observed failure):
[0]:
  time      : 2024-01-12_18:36:57
  host      : vision12.local
  rank      : 0 (local_rank: 0)
  exitcode  : 1 (pid: 1867675)
  error_file: <N/A>
  traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html
============================================================

In the eval file, eval the error is

01/12 18:40:45 - OpenCompass - ERROR - /home/xxx/llm_projects/opencompass-main/opencompass/tasks/openicl_eval.py - _score - 236 - Task [opencompass.models.huggingface.HuggingFace_v0-20240102-091935_checkpoint-37195-merged/cmmlu-agronomy]: No predictions found.
01/12 18:40:45 - OpenCompass - INFO - time elapsed: 1.67s

and the outputs are all NONE, 

cmmlu-professional_psychology                -          -         -       -
cmmlu-public_relations                       -          -         -       -
cmmlu-security_study                         -          -         -       -
cmmlu-sociology                              -          -         -       -
cmmlu-sports_science                         -          -         -       -
cmmlu-traditional_chinese_medicine           -          -         -       -
cmmlu-virology                               -          -         -       -

其他信息

The model is my own finetuned Yi-34B

tonysy commented 9 months ago

How about using official Yi?

qy1026 commented 9 months ago

Strangely, this problem could be fixed by adding --max-out-len and --max-seq-len in the eval script. After setting up this two parameters, the UnboundLocalError can be solved and the outputs are properly generated.

BUT Why?