학습하는데 시간은 얼마나 걸리셨나요?

제가 했을 때는 0.25밖에 안 나오는데 모델에 문제가 있는걸까요...이해가 안 가네요..
class 예측을 잘 못 하는데 아무것도 안 하고 단일 모델로 하신 거맞죠?

WARNING _ updating to 'imgsz=1280'. 'train' and 'val' imgsz must be an integer,이라는 에러가 훈련 imgsz에 리스트로 집어넣게되면 뜨는데 혹시 같은 메세지 받으신적있나요? 감사합니다.

삭제된 댓글입니다

valid set 은 빨리 수렴하던데 map 점수가 그럼 로스를 cls_loss를 0.08까진 학습을 시켜야 일반화가 된다는 말씀이시죠?

/usr/local/lib/python3.10/dist-packages/ultralytics/yolo/utils/loss.py in __call__(self, preds, batch)
    133         loss = torch.zeros(3, device=self.device)  # box, cls, dfl
    134         feats = preds[1] if isinstance(preds, tuple) else preds
--> 135         pred_distri, pred_scores = torch.cat([xi.view(feats[0].shape[0], self.no, -1) for xi in feats], 2).split(
    136             (self.reg_max * 4, self.nc), 1)
    137 

TypeError: 'NoneType' object is not iterable

모델 학습 과정에서 계속 이런 에러가 뜨는데 왜 그런걸까요ㅜ 

안녕하세요! 혹시 torch버전 어떤 걸로 사용하셨나요..?

안녕하세요 ! 코드보고 많이 배우고 있는 초심자입니다 ! 
다름이 아니라, 모델 Arguments 과정에서 다음과 같은 에러문구가 발생하여 해결방법을 여쭤봅니다.
Dataset 'data/yolo/custom.yaml' error ❌ 
Dataset 'data/yolo/custom.yaml' images not found ⚠️, missing paths ['/opt/homebrew/datasets/data/yolo/valid']
이라고 하는데 해결방법을 도무지 모르겠습니다... 고수님들 해결방법 부탁드립니다 ㅠ
참고로 data폴더와 open폴더는 동일한 위치에 있습니다. 

model = YOLO("yolov8x")
results = model.train 부분에서 

ValueError: Object arrays cannot be loaded when allow_pickle=False
에러 발생하는데 혹시 동일한 문제 해결한 분들 계실까요???

혹시 아래와 같은 오류가 발생 했을 때 어떻게 해결하는지 알  수 있을까요? ㅜㅠ

Downloading https://github.com/ultralytics/assets/releases/download/v0.0.0/yolov8x.pt to yolov8x.pt...
 Download failure, retrying 1/3 https://github.com/ultralytics/assets/releases/download/v0.0.0/yolov8x.pt...
Warning: Failed to open the file yolov8x.pt: No such file or directory        
                                                                           0.0%curl: (23) Failure writing output to destination

 Download failure, retrying 2/3 https://github.com/ultralytics/assets/releases/download/v0.0.0/yolov8x.pt...
Warning: Failed to open the file yolov8x.pt: No such file or directory        
                                                                           0.0%curl: (23) Failure writing output to destination

 Download failure, retrying 3/3 https://github.com/ultralytics/assets/releases/download/v0.0.0/yolov8x.pt...
Warning: Failed to open the file yolov8x.pt: No such file or directory        
                    

감사합니다.

혹시 학습한 best.pt 파일을 로드해서 계속  학습하려고 하는데 box_loss가 이전에 했던 loss 랑 별차이가 안나오는데요 맨처음 나온 loss값이 0.12정도 나온다하면 이후에 로드해서 학습시켜도 0.12 보다 조금더 높게 나오는데 로드해서 사ㅏ용하려면 따로 설정해야하는게 있나요? 아니면 여러번 로드해서 학습한다고 하면 유의미한 loss값 나오게하려면  epochs값을 올려야하나요?
코드는 model = ultralytics.YOLO('/content/drive/MyDrive/mAP_data/yolo/results/exp57/weights/best.pt')이렇게 불러와서 기존에 train 하던 코드그대로 진행합니다.

혹시 빠른 실시간 추론이 가능한 욜로말고 fps가 좀 낮은 모델로 테스트 해보셨는 지 궁금해요

혹시 전부 학습시키고 나면 csv파일에 몇개정도의 결과값이 쌓이나요??

안녕하세요 코드 정보 감사합니다.
객체탐지를 공부중인데 YOLO 학습을 위한 데이터 준비에서 궁금한점이 있습니다.
with open("../data/classes.txt", "r") as reader:
    lines = reader.readlines()
    classes = [line.strip().split(",")[1] for line in lines]

yaml_data = {
              "names": classes,
              "nc": len(classes),
              "path": "/Data/데이터 분석 대회/DACON/합성데이터 기반 객체 탐지 AI 경진대회/data/yolo/",
              "train": "train",
              "val": "valid",
              "test": "test"
            }

with open("../data/yolocustom.yaml", "w") as writer:
    yaml.dump(yaml_data, writer)

이부분이 YOLO를 위한 yaml을 만드는 작업인가요?
합성데이터 기반 객체 탐지 AI 경진대회의 경우 주어진 데이터에 classes.txt가 주어져서 위 코드와 같이 nc를 쉽게 정의할 수 있는데, 이러한 classes가 주어지지 않고 이미지와 labels(객체명, 좌표값) 정보만 주어진다면 어떻게 nc를 정할 수 있을까요?