Handle image_embeds in ViltModel (#16696)

* update

* batch_size -> text_batch_size
Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>

src/transformers/models/vilt/modeling_vilt.py

@@ -704,7 +704,7 @@ VILT_IMAGES_AND_TEXT_CLASSIFICATION_INPUTS_DOCSTRING = r"""
             is useful if you want more control over how to convert `input_ids` indices into associated vectors than the
             model's internal embedding lookup matrix.
 
-        image_embeds (`torch.FloatTensor` of shape `(batch_size, num_patches, hidden_size)`, *optional*):
+        image_embeds (`torch.FloatTensor` of shape `(batch_size, num_images, num_patches, hidden_size)`, *optional*):
             Optionally, instead of passing `pixel_values`, you can choose to directly pass an embedded representation.
             This is useful if you want more control over how to convert `pixel_values` into patch embeddings.
 
@@ -805,18 +805,22 @@ class ViltModel(ViltPreTrainedModel):
         else:
             raise ValueError("You have to specify either input_ids or inputs_embeds")
 
-        batch_size, seq_length = input_shape
+        text_batch_size, seq_length = input_shape
         device = input_ids.device if input_ids is not None else inputs_embeds.device
 
         if attention_mask is None:
-            attention_mask = torch.ones(((batch_size, seq_length)), device=device)
+            attention_mask = torch.ones(((text_batch_size, seq_length)), device=device)
 
-        if pixel_values is None:
-            raise ValueError("You have to specify pixel_values")
+        if pixel_values is not None and image_embeds is not None:
+            raise ValueError("You cannot specify both pixel_values and image_embeds at the same time")
+        elif pixel_values is None and image_embeds is None:
+            raise ValueError("You have to specify either pixel_values or image_embeds")
 
-        batch_size, num_channels, height, width = pixel_values.shape
+        image_batch_size = pixel_values.shape[0] if pixel_values is not None else image_embeds.shape[0]
+        if image_batch_size != text_batch_size:
+            raise ValueError("The text inputs and image inputs need to have the same batch size")
         if pixel_mask is None:
-            pixel_mask = torch.ones(((batch_size, height, width)), device=device)
+            pixel_mask = torch.ones((image_batch_size, self.config.image_size, self.config.image_size), device=device)
 
         # Prepare head mask if needed
         # 1.0 in head_mask indicate we keep the head
@@ -1338,11 +1342,17 @@ class ViltForImagesAndTextClassification(ViltPreTrainedModel):
         )
         return_dict = return_dict if return_dict is not None else self.config.use_return_dict
 
-        if pixel_values.ndim == 4:
+        if pixel_values is not None and pixel_values.ndim == 4:
             # add dummy num_images dimension
             pixel_values = pixel_values.unsqueeze(1)
 
-        num_images = pixel_values.shape[1]
+        if image_embeds is not None and image_embeds.ndim == 3:
+            # add dummy num_images dimension
+            image_embeds = image_embeds.unsqueeze(1)
+
+        num_images = pixel_values.shape[1] if pixel_values is not None else None
+        if num_images is None:
+            num_images = image_embeds.shape[1] if image_embeds is not None else None
         if num_images != self.config.num_images:
             raise ValueError(
                 "Make sure to match the number of images in the model with the number of images in the input."
@@ -1356,11 +1366,11 @@ class ViltForImagesAndTextClassification(ViltPreTrainedModel):
                 input_ids,
                 attention_mask=attention_mask,
                 token_type_ids=token_type_ids,
-                pixel_values=pixel_values[:, i, :, :, :],
+                pixel_values=pixel_values[:, i, :, :, :] if pixel_values is not None else None,
                 pixel_mask=pixel_mask[:, i, :, :] if pixel_mask is not None else None,
                 head_mask=head_mask,
                 inputs_embeds=inputs_embeds,
-                image_embeds=image_embeds,
+                image_embeds=image_embeds[:, i, :, :] if image_embeds is not None else None,
                 image_token_type_idx=i + 1,
                 output_attentions=output_attentions,
                 output_hidden_states=output_hidden_states,