[Model] Remove incorrect SupportsPP from MTP models (#32150)

Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>

vllm/model_executor/models/longcat_flash_mtp.py

@@ -24,7 +24,6 @@ from vllm.model_executor.models.longcat_flash import FlashConfig
 from vllm.sequence import IntermediateTensors
 
 from .deepseek_v2 import DeepseekV2DecoderLayer
-from .interfaces import SupportsPP
 from .utils import maybe_prefix
 
 
@@ -124,7 +123,7 @@ class LongCatMultiTokenPredictor(nn.Module):
         )
 
 
-class LongCatFlashMTP(nn.Module, SupportsPP):
+class LongCatFlashMTP(nn.Module):
     def __init__(self, *, vllm_config: VllmConfig, prefix: str = ""):
         super().__init__()
         # LongCat MTP without MoE layers