YOLOv6
YOLOv6 copied to clipboard
meituan
ERROR in evaluate and save model. ERROR in training loop or eval/save model.
Training start...
Epoch iou_loss l1_loss obj_loss cls_loss
0%| | 0/691 [00:00<?, ?it/s] /home/acus/anaconda3/envs/mengshan/lib/python3.8/site-packages/torch/functional.py:568: UserWarning: torch.meshgrid: in an upcoming release, it will be required to pass the indexing argument. (Triggered internally at ../aten/src/ATen/native/TensorShape.cpp:2228.)
return _VF.meshgrid(tensors, **kwargs) # type: ignore[attr-defined]
0/399 2.178 1.209 3.51 1.606: 100%|██████████| 691/691 [02:23<00:00,
Inferencing model in val datasets.: 100%|███████████████████| 87/87 [00:37<00:00, 2.32it/s]
Evaluating speed.
Evaluating mAP by pycocotools. Saving runs/train/exp6/predictions.json... loading annotations into memory... Done (t=0.04s) creating index... index created! Loading and preparing results... DONE (t=4.52s) creating index... index created! ERROR in evaluate and save model. ERROR in training loop or eval/save model.
Training completed in 0.053 hours.
Traceback (most recent call last):
File "tools/train.py", line 112, in
After the training is completed, before the evaluation results, I am getting the following error. I would really appreciate it, to get some insights on how to go about this issue as soon as possible.
I have an identical problem. On google colab it works perfectly, but on the local machine it throws this error.
I could fix this issue by configuring dataset.yaml instead of coco.yaml. I have considered my dataset as a custom dataset, because I am considering only two classes among the coco dataset.
I solved the problem myself. First, I used my own dataset.I found that the training didn't make an error by presenting only part of the data, and then I gradually checked and found that some photos named pure numbers caused the error.The problem arises presumably because pure numbers are judged to be int rather than str.I solved the problem by deleting the images with purely numeric names, and changing the names should work as well
I solved the problem myself. First, I used my own dataset.I found that the training didn't make an error by presenting only part of the data, and then I gradually checked and found that some photos named pure numbers caused the error.The problem arises presumably because pure numbers are judged to be int rather than str.I solved the problem by deleting the images with purely numeric names, and changing the names should work as well
I also confirm that if the name of the image consists only of numbers like for example '12345311232.jpg' this problem occurs and the images must be removed from the dataset so that the algorithm can be trained and evaluated.
我有这个错误
ERROR in training loop or eval/save model.
Training completed in 18.118 hours.
Traceback (most recent call last):
File "tools/train_face.py", line 86, in
I solved the problem myself. First, I used my own dataset.I found that the training didn't make an error by presenting only part of the data, and then I gradually checked and found that some photos named pure numbers caused the error.The problem arises presumably because pure numbers are judged to be int rather than str.I solved the problem by deleting the images with purely numeric names, and changing the names should work as well
u are right, i modify cocoeval.py's 80th line: self.params.imgIds = sorted([str(i) for i in cocoGt.getImgIds()]) ,and then ,it's work.
@dylanlb could you provide me with the modification you did I have the same problem.
yeah, I have solve the problem pycocotools unable to run : '<' not supported between instances of 'str' and 'int' that's been bothering me for two days, thank you very much.
I solved the problem myself. First, I used my own dataset.I found that the training didn't make an error by presenting only part of the data, and then I gradually checked and found that some photos named pure numbers caused the error.The problem arises presumably because pure numbers are judged to be int rather than str.I solved the problem by deleting the images with purely numeric names, and changing the names should work as well
u are right, i modify cocoeval.py's 80th line: self.params.imgIds = sorted([str(i) for i in cocoGt.getImgIds()]) ,and then ,it's work. yeah, I have solve the problem pycocotools unable to run : '<' not supported between instances of 'str' and 'int' that's been bothering me for two days, thank you very much.
@dylanlb Hello, can you tell me what kind of file cocoeval.py is and where exactly did you find this line?
