in data/datasets.py [0:0]
def __getitem__(self, idx):
item = self.dataset[idx]
# Handle images (should be a list)
images_data = item['images']
if not isinstance(images_data, list):
images_data = [images_data]
# Now process the images
processed_images = self._process_images(images_data)
messages = self._get_messages(item, len(processed_images))
input_ids, mask, attention_mask = self._prepare_inputs_and_loss_mask(messages)
labels = self._get_labels(input_ids, mask)
return {
"images": processed_images,
"input_ids": input_ids,
"attention_mask": attention_mask,
"labels": labels,
}