Upload processor

preprocessor_config.json

@@ -1,4 +1,7 @@
 {
+  "auto_map": {
+    "AutoProcessor": "processor.GIAProcessor"
+  },
   "crop_size": {
     "height": 224,
     "width": 224

processor.py

@@ -0,0 +1,43 @@
+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")

tokenizer_config.json

@@ -1,4 +1,7 @@
 {
+  "auto_map": {
+    "AutoProcessor": "processor.GIAProcessor"
+  },
   "clean_up_tokenization_spaces": true,
   "cls_token": "[CLS]",
   "do_lower_case": true,