[V0 Deprecation] Remove V0 sampling metadata (#25345)

Signed-off-by: Woosuk Kwon <woosuk@thinkingmachines.ai>
This commit is contained in:
Woosuk Kwon
2025-09-21 10:37:11 -07:00
committed by GitHub
parent c438b2951c
commit 1c3ffdbecc
141 changed files with 172 additions and 583 deletions
-3
View File
@@ -29,7 +29,6 @@ from vllm.model_executor.models.interfaces import (HasInnerState, IsHybrid,
SupportsV0Only)
from vllm.model_executor.models.mamba_cache import (MambaCacheManager,
MambaCacheParams)
from vllm.model_executor.sampling_metadata import SamplingMetadata
from vllm.sequence import IntermediateTensors
from .utils import make_layers, maybe_prefix
@@ -695,12 +694,10 @@ class Phi4FlashForCausalLM(nn.Module, HasInnerState, IsHybrid, SupportsV0Only):
def compute_logits(
self,
hidden_states: torch.Tensor,
sampling_metadata: SamplingMetadata,
) -> Optional[torch.Tensor]:
processed_logits = self.logits_processor(
self.lm_head,
hidden_states,
sampling_metadata,
self.embedding_bias,
)
return processed_logits