[PYTHON] retinanet 학습 중에 만난 오류들

dlerror: cudart64_110.dll not found

(project) D:\3rd_Deep\keras-retinanet-master>python keras_retinanet/bin/train.py --gpu=0 --epochs=10 --steps=1000 --workers=0 csv D:/3rd_Deep/reinanet_modeling/annotation.csv D:\3rd_Deep\reinanet_modeling/classmapping.csv
2021-04-07 16:20:13.646084: W tensorflow/stream_executor/platform/default/dso_loader.cc:60] Could not load dynamic library 'cudart64_110.dll'; dlerror: cudart64_110.dll not found
2021-04-07 16:20:13.650038: I tensorflow/stream_executor/cuda/cudart_stub.cc:29] Ignore above cudart dlerror if you do not have a GPU set up on your machine.
Traceback (most recent call last):
  File "keras_retinanet/bin/train.py", line 34, in <module>
    from .. import layers  # noqa: F401
  File "keras_retinanet/bin\..\..\keras_retinanet\layers\__init__.py", line 1, in <module>
    from ._misc import RegressBoxes, UpsampleLike, Anchors, ClipBoxes  # noqa: F401
  File "keras_retinanet/bin\..\..\keras_retinanet\layers\_misc.py", line 20, in <module>
    from ..utils import anchors as utils_anchors
  File "keras_retinanet/bin\..\..\keras_retinanet\utils\anchors.py", line 20, in <module>
    from ..utils.compute_overlap import compute_overlap
  File "keras_retinanet\utils\compute_overlap.pyx", line 1, in init keras_retinanet.utils.compute_overlap
    # --------------------------------------------------------
ValueError: numpy.ndarray size changed, may indicate binary incompatibility. Expected 88 from C header, got 80 from PyObject

retinanet을 포기하지 않고... 어떻게든 해보려고 열심히 라벨링하여 시도하는데 발생한 에러.

 

검색해보니 이걸 설치하면 된다는 문구 발견.

CUDA Toolkit 11.0 Download

설치 후 retinanet 학습시킬 코드 실행

python keras_retinanet/bin/train.py --gpu=0 --epochs=10 --steps=1000 --workers=0 csv D:/3rd_Deep/reinanet_modeling/annotation.csv D:\3rd_Deep\reinanet_modeling/classmapping.csv

 

여전히 같은 오류...

 

 

 

결국 가상환경을 새로 만들어서 retinanet을 재설치했다.

재설치 방법은 하단 링크 참고

[PYTHON] ANACONDA에서 RETINANET 설치

 

재설치 후 달라진 오류 내용.

(assult) D:\3rd_Deep\keras-retinanet-master>python keras_retinanet/bin/train.py --gpu=0 --epochs=10 --steps=1000 --workers=0 csv D:/3rd_Deep/reinanet_modeling/annotation.csv D:\3rd_Deep\reinanet_modeling/classmapping.csv
2021-04-07 19:22:12.247250: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library cudart64_110.dll
Traceback (most recent call last):
  File "keras_retinanet/bin/train.py", line 38, in <module>
    from ..callbacks.eval import Evaluate
  File "keras_retinanet/bin\..\..\keras_retinanet\callbacks\eval.py", line 18, in <module>
    from ..utils.eval import evaluate
  File "keras_retinanet/bin\..\..\keras_retinanet\utils\eval.py", line 18, in <module>
    from .visualization import draw_detections, draw_annotations
  File "keras_retinanet/bin\..\..\keras_retinanet\utils\visualization.py", line 17, in <module>
    import cv2
ModuleNotFoundError: No module named 'cv2'

그리고 매번 오류가 달라진다. 없다는 모듈들을 모두 재설치했더니 이번에 뜬 오류는...

Traceback (most recent call last):
  File "keras_retinanet\bin\train.py", line 553, in <module>
    main()
  File "keras_retinanet\bin\train.py", line 483, in main
    train_generator, validation_generator = create_generators(args, backbone.preprocess_image)
  File "keras_retinanet\bin\train.py", line 303, in create_generators
    **common_args
  File "keras_retinanet\bin\..\..\keras_retinanet\preprocessing\csv_generator.py", line 147, in __init__
    raise_from(ValueError('invalid CSV class file: {}: {}'.format(csv_class_file, e)), None)
  File "<string>", line 3, in raise_from
ValueError: invalid CSV class file: D:\3rd_Deep\reinanet_modeling\classmapping.csv: line 1: malformed class ID: invalid literal for int() with base 10: 'number'

이유가 무얼꼬... 생각보다 간단히 해결됐다. 컬럼명이 없어야한다. 컬럼명을 싹 다 지우고 다시 진행해보니,

드디어 진행되고 있다. steps를 1000으로 준 관계로 .... 오늘 내로는 안 끝날 것 같다. ^^;;;;