feat: TransformerSimilarityRanker add batching across Documents durin…

…g inference (#8344)

* First pass at adding batch support to TransformersSimilarityRanker

* Add test

* Add reno

haystack/components/rankers/transformers_similarity.py

@@ -16,6 +16,7 @@
 with LazyImport(message="Run 'pip install transformers[torch,sentencepiece]'") as torch_and_transformers_import:
     import accelerate  # pylint: disable=unused-import # the library is used but not directly referenced
     import torch
+    from torch.utils.data import DataLoader, Dataset
     from transformers import AutoModelForSequenceClassification, AutoTokenizer
 
 
@@ -42,7 +43,7 @@ class TransformersSimilarityRanker:
     ```
     """
 
-    def __init__(
+    def __init__(  # noqa: PLR0913
         self,
         model: Union[str, Path] = "cross-encoder/ms-marco-MiniLM-L-6-v2",
         device: Optional[ComponentDevice] = None,
@@ -57,6 +58,7 @@ def __init__(
         score_threshold: Optional[float] = None,
         model_kwargs: Optional[Dict[str, Any]] = None,
         tokenizer_kwargs: Optional[Dict[str, Any]] = None,
+        batch_size: int = 16,
     ):
         """
         Creates an instance of TransformersSimilarityRanker.
@@ -93,6 +95,9 @@ def __init__(
         :param tokenizer_kwargs:
             Additional keyword arguments for `AutoTokenizer.from_pretrained` when loading the tokenizer.
             Refer to specific model documentation for available kwargs.
+        :param batch_size:
+            The batch size to use for inference. The higher the batch size, the more memory is required.
+            If you run into memory issues, reduce the batch size.
 
         :raises ValueError:
             If `top_k` is not > 0.
@@ -117,6 +122,7 @@ def __init__(
         model_kwargs = resolve_hf_device_map(device=device, model_kwargs=model_kwargs)
         self.model_kwargs = model_kwargs
         self.tokenizer_kwargs = tokenizer_kwargs or {}
+        self.batch_size = batch_size
 
         # Parameter validation
         if self.scale_score and self.calibration_factor is None:
@@ -261,11 +267,28 @@ def run(
             text_to_embed = self.embedding_separator.join(meta_values_to_embed + [doc.content or ""])
             query_doc_pairs.append([self.query_prefix + query, self.document_prefix + text_to_embed])
 
-        features = self.tokenizer(query_doc_pairs, padding=True, truncation=True, return_tensors="pt").to(  # type: ignore
+        class _Dataset(Dataset):
+            def __init__(self, batch_encoding):
+                self.batch_encoding = batch_encoding
+
+            def __len__(self):
+                return len(self.batch_encoding["input_ids"])
+
+            def __getitem__(self, item):
+                return {key: self.batch_encoding.data[key][item] for key in self.batch_encoding.data.keys()}
+
+        batch_enc = self.tokenizer(query_doc_pairs, padding=True, truncation=True, return_tensors="pt").to(  # type: ignore
             self.device.first_device.to_torch()
         )
+        dataset = _Dataset(batch_enc)
+        inp_dataloader = DataLoader(dataset, batch_size=self.batch_size, shuffle=False)
+
+        similarity_scores = []
         with torch.inference_mode():
-            similarity_scores = self.model(**features).logits.squeeze(dim=1)  # type: ignore
+            for features in inp_dataloader:
+                model_preds = self.model(**features).logits.squeeze(dim=1)  # type: ignore
+                similarity_scores.extend(model_preds)
+        similarity_scores = torch.stack(similarity_scores)
 
         if scale_score:
             similarity_scores = torch.sigmoid(similarity_scores * calibration_factor)

releasenotes/notes/ranker-add-batching-during-inference-f077411ec389a63b.yaml

@@ -0,0 +1,4 @@
+---
+enhancements:
+  - |
+    We added batching during inference time to the TransformerSimilarityRanker to help prevent OOMs when ranking large amounts of Documents.

test/components/rankers/test_transformers_similarity.py

@@ -8,7 +8,7 @@
 import torch
 from transformers.modeling_outputs import SequenceClassifierOutput
 
-from haystack import ComponentError, Document
+from haystack import Document
 from haystack.components.rankers.transformers_similarity import TransformersSimilarityRanker
 from haystack.utils.auth import Secret
 from haystack.utils.device import ComponentDevice, DeviceMap
@@ -202,7 +202,9 @@ def test_from_dict_no_default_parameters(self):
 
     @patch("torch.sigmoid")
     @patch("torch.sort")
-    def test_embed_meta(self, mocked_sort, mocked_sigmoid):
+    @patch("torch.stack")
+    def test_embed_meta(self, mocked_stack, mocked_sort, mocked_sigmoid):
+        mocked_stack.return_value = torch.tensor([0])
         mocked_sort.return_value = (None, torch.tensor([0]))
         mocked_sigmoid.return_value = torch.tensor([0])
         embedder = TransformersSimilarityRanker(
@@ -232,7 +234,9 @@ def test_embed_meta(self, mocked_sort, mocked_sigmoid):
 
     @patch("torch.sigmoid")
     @patch("torch.sort")
-    def test_prefix(self, mocked_sort, mocked_sigmoid):
+    @patch("torch.stack")
+    def test_prefix(self, mocked_stack, mocked_sort, mocked_sigmoid):
+        mocked_stack.return_value = torch.tensor([0])
         mocked_sort.return_value = (None, torch.tensor([0]))
         mocked_sigmoid.return_value = torch.tensor([0])
         embedder = TransformersSimilarityRanker(
@@ -261,7 +265,9 @@ def test_prefix(self, mocked_sort, mocked_sigmoid):
         )
 
     @patch("torch.sort")
-    def test_scale_score_false(self, mocked_sort):
+    @patch("torch.stack")
+    def test_scale_score_false(self, mocked_stack, mocked_sort):
+        mocked_stack.return_value = torch.FloatTensor([-10.6859, -8.9874])
         mocked_sort.return_value = (None, torch.tensor([0, 1]))
         embedder = TransformersSimilarityRanker(model="model", scale_score=False)
         embedder.model = MagicMock()
@@ -277,7 +283,9 @@ def test_scale_score_false(self, mocked_sort):
         assert out["documents"][1].score == pytest.approx(-8.9874, abs=1e-4)
 
     @patch("torch.sort")
-    def test_score_threshold(self, mocked_sort):
+    @patch("torch.stack")
+    def test_score_threshold(self, mocked_stack, mocked_sort):
+        mocked_stack.return_value = torch.FloatTensor([0.955, 0.001])
         mocked_sort.return_value = (None, torch.tensor([0, 1]))
         embedder = TransformersSimilarityRanker(model="model", scale_score=False, score_threshold=0.1)
         embedder.model = MagicMock()
@@ -359,6 +367,48 @@ def test_run(self, query, docs_before_texts, expected_first_text, scores):
         assert docs_after[1].score == pytest.approx(sorted_scores[1], abs=1e-6)
         assert docs_after[2].score == pytest.approx(sorted_scores[2], abs=1e-6)
 
+    @pytest.mark.integration
+    @pytest.mark.parametrize(
+        "query,docs_before_texts,expected_first_text,scores",
+        [
+            (
+                "City in Bosnia and Herzegovina",
+                ["Berlin", "Belgrade", "Sarajevo"],
+                "Sarajevo",
+                [2.2864143829792738e-05, 0.00012495707778725773, 0.009869757108390331],
+            ),
+            (
+                "Machine learning",
+                ["Python", "Bakery in Paris", "Tesla Giga Berlin"],
+                "Python",
+                [1.9063229046878405e-05, 1.434577916370472e-05, 1.3049247172602918e-05],
+            ),
+            (
+                "Cubist movement",
+                ["Nirvana", "Pablo Picasso", "Coffee"],
+                "Pablo Picasso",
+                [1.3313065210240893e-05, 9.90335684036836e-05, 1.3518535524781328e-05],
+            ),
+        ],
+    )
+    def test_run_small_batch_size(self, query, docs_before_texts, expected_first_text, scores):
+        """
+        Test if the component ranks documents correctly.
+        """
+        ranker = TransformersSimilarityRanker(model="cross-encoder/ms-marco-MiniLM-L-6-v2", batch_size=2)
+        ranker.warm_up()
+        docs_before = [Document(content=text) for text in docs_before_texts]
+        output = ranker.run(query=query, documents=docs_before)
+        docs_after = output["documents"]
+
+        assert len(docs_after) == 3
+        assert docs_after[0].content == expected_first_text
+
+        sorted_scores = sorted(scores, reverse=True)
+        assert docs_after[0].score == pytest.approx(sorted_scores[0], abs=1e-6)
+        assert docs_after[1].score == pytest.approx(sorted_scores[1], abs=1e-6)
+        assert docs_after[2].score == pytest.approx(sorted_scores[2], abs=1e-6)
+
     #  Returns an empty list if no documents are provided
     @pytest.mark.integration
     def test_returns_empty_list_if_no_documents_are_provided(self):