community[patch]: Fix generation_config not setting properly for Deep…

…Sparse (#15036)

- **Description:** Tiny but important bugfix to use a more stable
interface for specifying generation_config parameters for DeepSparse LLM

libs/community/langchain_community/llms/deepsparse.py

@@ -63,7 +63,7 @@ def validate_environment(cls, values: Dict) -> Dict:
         except ImportError:
             raise ImportError(
                 "Could not import `deepsparse` package. "
-                "Please install it with `pip install deepsparse`"
+                "Please install it with `pip install deepsparse[llm]`"
             )
 
         model_config = values["model_config"] or {}
@@ -103,9 +103,7 @@ def _call(
             text = combined_output
         else:
             text = (
-                self.pipeline(
-                    sequences=prompt, generation_config=self.generation_config
-                )
+                self.pipeline(sequences=prompt, **self.generation_config)
                 .generations[0]
                 .text
             )
@@ -143,9 +141,7 @@ async def _acall(
             text = combined_output
         else:
             text = (
-                self.pipeline(
-                    sequences=prompt, generation_config=self.generation_config
-                )
+                self.pipeline(sequences=prompt, **self.generation_config)
                 .generations[0]
                 .text
             )
@@ -184,7 +180,7 @@ def _stream(
                     print(chunk, end='', flush=True)
         """
         inference = self.pipeline(
-            sequences=prompt, generation_config=self.generation_config, streaming=True
+            sequences=prompt, streaming=True, **self.generation_config
         )
         for token in inference:
             chunk = GenerationChunk(text=token.generations[0].text)
@@ -222,7 +218,7 @@ async def _astream(
                     print(chunk, end='', flush=True)
         """
         inference = self.pipeline(
-            sequences=prompt, generation_config=self.generation_config, streaming=True
+            sequences=prompt, streaming=True, **self.generation_config
         )
         for token in inference:
             chunk = GenerationChunk(text=token.generations[0].text)