incomplete captions generated!

[Monikshah]

I ran your test.py file, and printed the gen (generated caption), there are so many captions that are incomplete sentences such as:

• "a desk with two computer monitors and a laptop on a"
• "a traffic light in front of a building with a"

I am not sure if this is the case. Cam you please let me know why I am getting many incomplete sentences as generated caption using this model?

Thank you!

[thalapandi]

Type of dataset attribute: <class 'main.PairedDataset'>
collate_fn is callable.
Length of train dataloader: 35403
Length of validation dataloader: 1563
Length of test dataloader: 313
Is collate_fn callable? True

TypeError Traceback (most recent call last)
in <cell line: 74>()
72 print("Is collate_fn callable?", callable(dataloader_train.dataset.collate_fn))
73
---> 74 for it, (images, captions) in enumerate(dataloader_train):
75 try:
76 print("Iteration:", it)

2 frames
/usr/local/lib/python3.10/dist-packages/torch/utils/data/dataloader.py in next(self)
629 # TODO(pytorch/pytorch#76750)
630 self._reset() # type: ignore[call-arg]
--> 631 data = self._next_data()
632 self._num_yielded += 1
633 if self._dataset_kind == _DatasetKind.Iterable and \

/usr/local/lib/python3.10/dist-packages/torch/utils/data/dataloader.py in _next_data(self)
673 def _next_data(self):
674 index = self._next_index() # may raise StopIteration
--> 675 data = self._dataset_fetcher.fetch(index) # may raise StopIteration
676 if self._pin_memory:
677 data = _utils.pin_memory.pin_memory(data, self._pin_memory_device)

/usr/local/lib/python3.10/dist-packages/torch/utils/data/_utils/fetch.py in fetch(self, possibly_batched_index)
47 if self.auto_collation:
48 if hasattr(self.dataset, "getitems") and self.dataset.getitems:
---> 49 data = self.dataset.getitems(possibly_batched_index)
50 else:
51 data = [self.dataset[idx] for idx in possibly_batched_index]

TypeError: 'generator' object is not callable

[Peodidjdj]

我遇到和你一样的问题..如果我后面解决了会再评论的