AttributeError: 'NoneType' object has no attribute 'size'
Tranbaber opened this issue · 2 comments
When I train the coco dataset with mobilevitv2 on a computer with only one GPU, I have the following issue:
“2023-09-28 13:05:49 - LOGS - Using EMA
2023-09-28 13:05:49 - LOGS - No checkpoint found at 'results/train/training_checkpoint_last.pt'
2023-09-28 13:05:49 - INFO - Configuration file is stored here: results/train/config.yaml
2023-09-28 13:05:51 - DEBUG - Training epoch 0 with 117266 samples
2023-09-28 13:05:51 - LOGS - Exception occurred that interrupted the training:
Traceback (most recent call last):
File "D:\Study\pycharm_projects\MobileViTv2\ml-cvnets-main\engine\training_engine.py", line 606, in run
train_loss, train_ckpt_metric = self.train_epoch(epoch)
File "D:\Study\pycharm_projects\MobileViTv2\ml-cvnets-main\engine\training_engine.py", line 230, in train_epoch
for batch_id, batch in enumerate(self.train_loader):
File "C:\Users\72344.conda\envs\MobileViTv2\lib\site-packages\torch\utils\data\dataloader.py", line 681, in next
data = self._next_data()
File "C:\Users\72344.conda\envs\MobileViTv2\lib\site-packages\torch\utils\data\dataloader.py", line 721, in _next_data
data = self._dataset_fetcher.fetch(index) # may raise StopIteration
File "C:\Users\72344.conda\envs\MobileViTv2\lib\site-packages\torch\utils\data_utils\fetch.py", line 49, in fetch
data = [self.dataset[idx] for idx in possibly_batched_index]
File "C:\Users\72344.conda\envs\MobileViTv2\lib\site-packages\torch\utils\data_utils\fetch.py", line 49, in
data = [self.dataset[idx] for idx in possibly_batched_index]
File "D:\Study\pycharm_projects\MobileViTv2\ml-cvnets-main\data\datasets\detection\coco_ssd.py", line 157, in getitem
im_width, im_height = image.size
AttributeError: 'NoneType' object has no attribute 'size'
”
I would like to get everyone's help, thank you very much
@mchorton
Does. 'results/train/training_checkpoint_last.pt' exists?
Does your training dataset contains proper images? Error looks like your image hasnt been loaded. Try to create training/testing datasets from only one class and one file and try again- if that works it means some other files of your dataset are correpted, if that doesnt work probably configuration problem.