denisdsh / vizwiz-vqa-pytorch Goto Github PK

Hi,
Thanks a lot for your implementation. I tried running it on the VizWiz data with the default hyperparameters in the repo. Unfortunately, the model only achieves around 36% accuracy on the test set, with around 49% on the train set, and hence suffers from a lot of overfitting. Can you please tell me if there are some other things to finetune or why that might be happening?
I'm using the data from here: https://vizwiz.org/tasks-and-datasets/vqa/

Model logs will be saved in logs/vizwiz/2022-03-03_15:53:09
train.py
train E000: 0% 0/156 [00:00<?, ?it/s]i = 0
item['question']] = tensor([18, 4, 41, 5, 2, 37, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0])
__getitem() 19905
self.img_names[i] = VizWiz_train_00000010.jpg
i = 128
item['question']] = tensor([ 3, 4, 6, 5, 880, 118, 9, 29, 63, 185, 610, 24,
373, 24, 2126, 127, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0])
__getitem() 19905
self.img_names[i] = VizWiz_train_00000010.jpg
train E000: 0% 0/156 [00:00<?, ?it/s]
Traceback (most recent call last):
File "train.py", line 204, in
main()
File "train.py", line 167, in main
train(model, train_loader, optimizer, tracker, epoch=i, split=config['training']['train_split'])
File "train.py", line 26, in train
for item in tq:
File "/usr/local/lib/python3.7/dist-packages/tqdm/std.py", line 1180, in iter
for obj in iterable:
File "/usr/local/lib/python3.7/dist-packages/torch/utils/data/dataloader.py", line 521, in next
data = self._next_data()
File "/usr/local/lib/python3.7/dist-packages/torch/utils/data/dataloader.py", line 1203, in _next_data
return self._process_data(data)
File "/usr/local/lib/python3.7/dist-packages/torch/utils/data/dataloader.py", line 1229, in _process_data
data.reraise()
File "/usr/local/lib/python3.7/dist-packages/torch/_utils.py", line 434, in reraise
raise exception
KeyError: Caught KeyError in DataLoader worker process 0.
Original Traceback (most recent call last):
File "/usr/local/lib/python3.7/dist-packages/torch/utils/data/_utils/worker.py", line 287, in _worker_loop
data = fetcher.fetch(index)
File "/usr/local/lib/python3.7/dist-packages/torch/utils/data/_utils/fetch.py", line 49, in fetch
data = [self.dataset[idx] for idx in possibly_batched_index]
File "/usr/local/lib/python3.7/dist-packages/torch/utils/data/_utils/fetch.py", line 49, in
data = [self.dataset[idx] for idx in possibly_batched_index]
File "/content/drive/MyDrive/VizWiz-VQA-PyTorch-master/datasets/vqa_dataset.py", line 118, in getitem
feature_id = self.name_to_id[img_name[10]]
KeyError: 'VizWiz_train_00000010.jpg'