fix: put HFTokenStreamingHandler in a lazy_import block (#7403)

* put HFTokenStreamingHandler in a lazy_import block

* fix pylint

haystack/components/generators/hugging_face_local.py

@@ -10,7 +10,7 @@
  deserialize_secrets_inplace,
  serialize_callable,
 )
-from haystack.utils.hf import HFTokenStreamingHandler, deserialize_hf_model_kwargs, serialize_hf_model_kwargs
+from haystack.utils.hf import deserialize_hf_model_kwargs, serialize_hf_model_kwargs
 
 logger = logging.getLogger(__name__)
 
@@ -19,7 +19,11 @@
 with LazyImport(message="Run 'pip install transformers[torch]'") as transformers_import:
  from transformers import StoppingCriteriaList, pipeline
 
- from haystack.utils.hf import StopWordsCriteria, resolve_hf_pipeline_kwargs # pylint: disable=ungrouped-imports
+ from haystack.utils.hf import ( # pylint: disable=ungrouped-imports
+ HFTokenStreamingHandler,
+ StopWordsCriteria,
+ resolve_hf_pipeline_kwargs,
+ )
 
 
 @component

haystack/utils/hf.py

@@ -222,8 +222,7 @@ def check_generation_params(kwargs: Optional[Dict[str, Any]], additional_accepte
 with LazyImport(message="Run 'pip install transformers[torch]'") as torch_and_transformers_import:
  from transformers import PreTrainedTokenizer, PreTrainedTokenizerFast, StoppingCriteria, TextStreamer
 
- transformers_import.check()
- torch_import.check()
+ torch_and_transformers_import.check()
 
  class StopWordsCriteria(StoppingCriteria):
  """