AttributeError: 'NoneType' object has no attribute '_free_weak_ref' about yolov6 HOT 9 CLOSED

The problem happened in training myself data refer to https://github.com/meituan/YOLOv6/blob/main/docs/Train_custom_data.md

from yolov6.

I have the same problem, did you solve it?

from yolov6.

I changed my python version to 3.8 ,but also occur the error as follow:

Training start...

 Epoch  iou_loss   l1_loss  obj_loss  cls_loss

0%| | 0/82 [00:00<?, ?it/s] /home/jack/anaconda3/envs/pytorch_for_yolov6/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]
0/99 1.868 1.227 5.353 2.089: 100%|██████████| 82/82 [00:4
Inferencing model in val datasets.: 100%|███████| 18/18 [00:16<00:00, 1.08it/s]

Evaluating speed.

Evaluating mAP by pycocotools.
Saving ./runs/train/exp/predictions.json...
loading annotations into memory...
Done (t=0.12s)
creating index...
index created!
Loading and preparing results...
ERROR in evaluate and save model.
ERROR in training loop or eval/save model.

Training completed in 0.017 hours.
Traceback (most recent call last):
File "tools/train.py", line 87, in
main(args)
File "tools/train.py", line 77, in main
trainer.train()
File "/home/jack/data1/project/yolov6_0625/yolov6/core/engine.py", line 62, in train
self.train_in_loop()
File "/home/jack/data1/project/yolov6_0625/yolov6/core/engine.py", line 81, in train_in_loop
self.eval_and_save()
File "/home/jack/data1/project/yolov6_0625/yolov6/core/engine.py", line 107, in eval_and_save
self.eval_model()
File "/home/jack/data1/project/yolov6_0625/yolov6/core/engine.py", line 126, in eval_model
results = eval.run(self.data_dict,
File "/home/jack/anaconda3/envs/pytorch_for_yolov6/lib/python3.8/site-packages/torch/autograd/grad_mode.py", line 27, in decorate_context
return func(*args, **kwargs)
File "/home/jack/data1/project/yolov6_0625/tools/eval.py", line 76, in run
eval_result = val.eval_model(pred_result, model, dataloader, task)
File "/home/jack/data1/project/yolov6_0625/yolov6/core/evaler.py", line 129, in eval_model
pred = anno.loadRes(pred_json)
File "/home/jack/anaconda3/envs/pytorch_for_yolov6/lib/python3.8/site-packages/pycocotools/coco.py", line 327, in loadRes
assert set(annsImgIds) == (set(annsImgIds) & set(self.getImgIds())),
AssertionError: Results do not correspond to current coco set

from yolov6.

#76

from yolov6.

This is an erro happended due to torch1.11.0 version + torchvision0.12+cu11XX , just uninstall torch and torch visisonpip uninstall torch torchvision
then install pip install torch==1.10.0+cu113 torchvision==0.11.1+cu113 -f https://download.pytorch.org/whl/cu113/torch_stable.html
This I' ve tried fixed this problem.

from yolov6.

@mendoza-G ,
I modified the corresponding pytorch version according to your introduction, but the above error still occurs. For the data format, I refer to https://github.com/meituan/YOLOv6/blob/main/docs/Train_custom_data.md.

from yolov6.

@mendoza-G , I modified the corresponding pytorch version according to your introduction, but the above error still occurs. For the data format, I refer to https://github.com/meituan/YOLOv6/blob/main/docs/Train_custom_data.md.

Check your cuda version as well, this solution worked for me, it seems your data format is not righht, why not download COCO2017 to test your environment qualified and treat that as a reference.

from yolov6.

@mendoza-G
I believe that the coco data test is definitely no problem, but I currently need to train my own data set. Since the official manual is provided, it should be able to train normally, otherwise it is a bug, and if the official requirements.txt manual has pytorch and torchvision versions Specific requirements can specify the fixed version.

from yolov6.

Dear all,

I had solve the issue refer to #76

from yolov6.