chore: update transformers test dependency (#8752)

* update transformers test dependency

* add pad_token_id to the mock tokenizer

* fix HFLocal test + new test

Author: anakin87

Diff for: pyproject.toml

@@ -85,7 +85,7 @@ format-check = "ruff format --check {args}"
 extra-dependencies = [
   "numpy>=2", # Haystack is compatible both with numpy 1.x and 2.x, but we test with 2.x
 
-  "transformers[torch,sentencepiece]==4.44.2", # ExtractiveReader, TransformersSimilarityRanker, LocalWhisperTranscriber, HFGenerators...
+  "transformers[torch,sentencepiece]==4.47.1", # ExtractiveReader, TransformersSimilarityRanker, LocalWhisperTranscriber, HFGenerators...
   "huggingface_hub>=0.27.0",                   # Hugging Face API Generators and Embedders
   "sentence-transformers>=3.0.0",              # SentenceTransformersTextEmbedder and SentenceTransformersDocumentEmbedder
   "langdetect",                                # TextLanguageRouter and DocumentLanguageClassifier

Diff for: test/components/generators/chat/test_hugging_face_local.py

@@ -42,6 +42,7 @@ def mock_pipeline_tokenizer():
     # Mocking the tokenizer
     mock_tokenizer = Mock(spec=PreTrainedTokenizer)
     mock_tokenizer.encode.return_value = ["Berlin", "is", "cool"]
+    mock_tokenizer.pad_token_id = 100
     mock_pipeline.tokenizer = mock_tokenizer
 
     return mock_pipeline

Diff for: test/components/generators/test_hugging_face_local_generator.py

@@ -397,8 +397,12 @@ def test_stop_words_criteria_with_a_mocked_tokenizer(self):
         # "This is ambiguously, but is unrelated."
         input_ids_one = torch.LongTensor([[100, 19, 24621, 11937, 6, 68, 19, 73, 3897, 5]])
         input_ids_two = torch.LongTensor([[100, 19, 73, 24621, 11937]])  # "This is unambiguously"
-        stop_words_criteria = StopWordsCriteria(tokenizer=Mock(spec=PreTrainedTokenizerFast), stop_words=["mock data"])
+
+        mock_tokenizer = Mock(spec=PreTrainedTokenizerFast)
+        mock_tokenizer.pad_token = "<pad>"
+        stop_words_criteria = StopWordsCriteria(tokenizer=mock_tokenizer, stop_words=["mock data"])
         stop_words_criteria.stop_ids = stop_words_id
+
         assert not stop_words_criteria(input_ids_one, scores=None)
         assert stop_words_criteria(input_ids_two, scores=None)
 
@@ -459,3 +463,15 @@ def test_hf_pipeline_runs_with_our_criteria(self):
         results = generator.run(prompt="something that triggers something")
         assert results["replies"] != []
         assert generator.stopping_criteria_list is not None
+
+    @pytest.mark.integration
+    @pytest.mark.flaky(reruns=3, reruns_delay=10)
+    def test_live_run(self):
+        llm = HuggingFaceLocalGenerator(model="Qwen/Qwen2.5-0.5B-Instruct", generation_kwargs={"max_new_tokens": 50})
+        llm.warm_up()
+
+        result = llm.run(prompt="Please create a summary about the following topic: Climate change")
+
+        assert "replies" in result
+        assert isinstance(result["replies"][0], str)
+        assert "climate change" in result["replies"][0].lower()