biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

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

Browse Source 
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
Download Patch File Download Diff File Expand all files Collapse all files

7
vllm/model_executor/models/mamba2.py
View File

@@ -30,7 +30,6 @@ from vllm.model_executor.models.interfaces import (HasInnerState,
IsAttentionFree)
from vllm.model_executor.models.mamba_cache import (MambaCacheManager,
MambaCacheParams)
from vllm.model_executor.sampling_metadata import SamplingMetadata
from vllm.sequence import IntermediateTensors
from vllm.utils import LayerBlockType
@@ -335,10 +334,8 @@ class Mamba2ForCausalLM(nn.Module, HasInnerState, IsAttentionFree):
def get_seqlen_agnostic_capture_inputs(self, batch_size: int):
return self.mamba_cache.get_seqlen_agnostic_capture_inputs(batch_size)
def compute_logits(self, hidden_states: torch.Tensor,
sampling_metadata: SamplingMetadata) -> torch.Tensor:
logits = self.logits_processor(self.lm_head, hidden_states,
sampling_metadata)
def compute_logits(self, hidden_states: torch.Tensor) -> torch.Tensor:
logits = self.logits_processor(self.lm_head, hidden_states)
return logits
def load_weights(self, weights: Iterable[tuple[str,