fix: warnings related to chunk_length_s in ifw (#258)

Author: johannaSommer

src/pruna/algorithms/batching/ifw.py

@@ -128,6 +128,7 @@ def _apply(self, model: Any, smash_config: SmashConfigPrefixWrapper) -> Any:
                 else {"attn_implementation": "sdpa"}
             ),
             device=smash_config["device"],
+            ignore_warning=True,
         )
         return pipe
 

tests/fixtures.py

@@ -62,7 +62,6 @@ def whisper_tiny_random_model() -> tuple[Any, SmashConfig]:
     model = pipeline(
         "automatic-speech-recognition",
         model=model_id,
-        chunk_length_s=30,
         torch_dtype=torch.float16,
         device_map="cpu",
     )