gia2-small / processor.py
ClementRomac's picture
ClementRomac HF staff
Upload processor
2e8ba46
raw
history blame
1.82 kB
from itertools import chain
from transformers import GitProcessor
class GIAProcessor(GitProcessor):
def __init__(self, image_processor, tokenizer):
super().__init__(image_processor, tokenizer)
self._block_size = 1024
def _group_texts(self, examples):
# Concatenate all texts.
concatenated_examples = {k: list(chain(*examples[k])) for k in examples.keys()}
total_length = len(concatenated_examples[list(examples.keys())[0]])
# We drop the small remainder, and if the total_length < block_size we exclude this batch and return an empty dict.
# We could add padding if the model supported it instead of this drop, you can customize this part to your needs.
total_length = (total_length // self._block_size) * self._block_size
# Split by chunks of max_len.
result = {
k: [t[i: i + self._block_size] for i in range(0, total_length, self._block_size)]
for k, t in concatenated_examples.items()
}
return result
def __call__(self, text=None, images=None, return_tensors=None, **kwargs):
if text is not None and images is None:
encoded_text = self.tokenizer(text, return_tensors=return_tensors)
encoding = self._group_texts(encoded_text)
elif text is not None and images is not None:
encoding = super().__call__(text, images, return_tensors, **kwargs)
return encoding
def batch_decode(self, *args, **kwargs):
return self.tokenizer.batch_decode(*args, **kwargs)
def decode(self, *args, **kwargs):
return self.tokenizer.decode(*args, **kwargs)
@property
def model_input_names(self):
return ["input_ids", "attention_mask", "pixel_values"]
GIAProcessor.register_for_auto_class("AutoProcessor")