Fix pos_mask application and update tests accordingly (#27892)

* Fix pos_mask application and update tests accordingly

* Fix style

* Adding comments

---------

Co-authored-by: Fernando Rodriguez <fernando.rodriguez@nielseniq.com>

src/transformers/models/flava/modeling_flava.py

@@ -1949,6 +1949,7 @@ class FlavaForPreTraining(FlavaPreTrainedModel):
 
                 if mim_labels is not None:
                     mim_labels = mim_labels[pos_mask]
+                    bool_masked_pos = bool_masked_pos[pos_mask]
 
         # MMM Image Loss
         if multimodal_masked_embeddings is not None and self.mmm_image_weight > 0:
@@ -1956,8 +1957,6 @@ class FlavaForPreTraining(FlavaPreTrainedModel):
             end_index = image_masked_embeddings.size(1) - 1
             sequence_for_image = sequence_for_image[:, 2 : 2 + end_index, :]
 
-            if pos_mask is not None:
-                sequence_for_image = sequence_for_image[pos_mask]
             if mim_labels is not None:
                 mim_labels = self._resize_to_2d(mim_labels)
                 bool_masked_pos = self._resize_to_2d(bool_masked_pos)
@@ -1979,8 +1978,6 @@ class FlavaForPreTraining(FlavaPreTrainedModel):
         if multimodal_masked_embeddings is not None and self.mmm_text_weight > 0:
             sequence_for_text = multimodal_masked_embeddings
             sequence_for_text = sequence_for_text[:, -text_masked_embeddings.size(1) :, :]
-            if pos_mask is not None:
-                sequence_for_text = sequence_for_text[pos_mask]
 
             if mlm_labels is not None:
                 mlm_labels = self._resize_to_2d(mlm_labels)

tests/models/flava/test_modeling_flava.py

@@ -1313,8 +1313,12 @@ class FlavaForPreTrainingIntegrationTest(unittest.TestCase):
             return_codebook_pixels=True,
             return_image_mask=True,
         )
+        # Create a clone of the input_ids tensor that will be its masked version
         inputs["input_ids_masked"] = inputs["input_ids"].clone()
+        # Mask the tokens "a" & "cat" from the "a photo of a cat" text using the special 103 value
         inputs["input_ids_masked"][0, 4:6] = 103
+        # MLM labels. It is a cloned version of input_ids where all values are -100 (i.e., ignored)
+        # except those that are masked, whose original values are stored
         inputs["mlm_labels"] = inputs["input_ids"].clone()
         inputs["mlm_labels"][:, :] = -100
         inputs["mlm_labels"][0, 4:6] = inputs["input_ids"][0, 4:6]
@@ -1338,3 +1342,54 @@ class FlavaForPreTrainingIntegrationTest(unittest.TestCase):
         self.assertAlmostEqual(outputs.loss_info.mmm_text.item(), 1.75533199, places=4)
         self.assertAlmostEqual(outputs.loss_info.mmm_image.item(), 7.0290069, places=4)
         self.assertAlmostEqual(outputs.loss.item(), 11.0626, places=4)
+
+    @slow
+    def test_inference_with_itm_labels(self):
+        model_name = "facebook/flava-full"
+        model = FlavaForPreTraining.from_pretrained(model_name).to(torch_device)
+        processor = FlavaProcessor.from_pretrained(model_name)
+        torch.manual_seed(1)
+        random.seed(1)
+
+        image = prepare_img()
+        inputs = processor(
+            text=["a photo of a cat", "a photo of a dog"],
+            images=[image, image],
+            padding="max_length",
+            max_length=77,
+            return_tensors="pt",
+            return_codebook_pixels=True,
+            return_image_mask=True,
+        )
+        # Create a clone of the input_ids tensor that will be its masked version
+        inputs["input_ids_masked"] = inputs["input_ids"].clone()
+        # Mask the tokens "a" & "cat" from the "a photo of a cat" text using the special 103 value
+        inputs["input_ids_masked"][0, 4:6] = 103
+        # MLM labels. It is a cloned version of input_ids where all values are -100 (i.e., ignored)
+        # except those that are masked, whose original values are stored
+        inputs["mlm_labels"] = inputs["input_ids"].clone()
+        inputs["mlm_labels"][:, :] = -100
+        inputs["mlm_labels"][0, 4:6] = inputs["input_ids"][0, 4:6]
+        # Manually create the itm_labels tensor that indicates if the image-text match.
+        # In this case, the firs pair matches and the second does not
+        inputs["itm_labels"] = torch.tensor([1, 0])
+        inputs = inputs.to(torch_device)
+        # forward pass
+        with torch.no_grad():
+            outputs = model(**inputs)
+
+        # verify the logits
+        self.assertEqual(
+            outputs.contrastive_logits_per_image.shape,
+            torch.Size((torch.count_nonzero(inputs["itm_labels"]).item(), inputs.input_ids.shape[0])),
+        )
+        self.assertEqual(
+            outputs.contrastive_logits_per_text.shape,
+            torch.Size((torch.count_nonzero(inputs["itm_labels"]).item(), inputs.pixel_values.shape[0])),
+        )
+
+        expected_logits = torch.tensor([[16.1291, 8.4033], [16.1291, 8.4033]], device=torch_device)
+        self.assertTrue(torch.allclose(outputs.contrastive_logits_per_image, expected_logits, atol=1e-3))
+        self.assertAlmostEqual(outputs.loss_info.mmm_text.item(), 1.75533199, places=4)
+        self.assertAlmostEqual(outputs.loss_info.mmm_image.item(), 6.89590501, places=4)
+        self.assertAlmostEqual(outputs.loss.item(), 9.1995, places=4)