Fix pre-commit issues

- Fix QuarkW8A8Fp8 constructor args (uses weight_config/input_config)
- Fix line length in compressed_tensors_w8a8_int8.py
- Apply yapf and isort formatting

Signed-off-by: padg9912 <phone.and.desktop@gmail.com>

Author: certainly-param

vllm/model_executor/layers/quantization/compressed_tensors/schemes/compressed_tensors_w8a8_int8.py

@@ -83,7 +83,8 @@ def int8_to_fp8_weight_loader(param, loaded_weight, *args, **kwargs):
                     # This preserves the actual values rather than reinterpreting bits
                     fp8_dtype = current_platform.fp8_dtype()
                     # Simple cast - FP8 can represent the INT8 range
-                    loaded_weight = loaded_weight.to(torch.float32).to(fp8_dtype)
+                    loaded_weight = loaded_weight.to(
+                        torch.float32).to(fp8_dtype)
                 return original_loader(param, loaded_weight, *args, **kwargs)
 
             return self._fp8_scheme.create_weights(

vllm/model_executor/layers/quantization/quark/schemes/quark_w8a8_int8.py

@@ -47,9 +47,14 @@ def __init__(self, qscheme: str, is_static_input_scheme: Optional[bool],
             from vllm.model_executor.layers.quantization.quark.schemes.quark_w8a8_fp8 import (  # noqa: E501
                 QuarkW8A8Fp8)
 
+            # Create FP8 config matching INT8 setup
+            weight_config = {"qscheme": qscheme}
+            input_config = None
+            if is_static_input_scheme:
+                input_config = {"is_dynamic": False, "qscheme": qscheme}
+            
             self._fp8_scheme = QuarkW8A8Fp8(
-                qscheme=qscheme,
-                is_static_input_scheme=is_static_input_scheme)
+                weight_config=weight_config, input_config=input_config)
 
     @classmethod
     def get_min_capability(cls) -> int: