[Feat]mtp aclgraph support (#3244)

### What this PR does / why we need it?
Currently, MTP Model in deepseek can not be capture in ACLGraph. This PR
is use to allow MTP to be captured in ACLGraph mode.

### Does this PR introduce _any_ user-facing change?

### How was this patch tested?


- vLLM version: v0.11.0rc3
- vLLM main: https://github.com/vllm-project/vllm/commit/v0.11.0

Signed-off-by: anon189Ty <Stari_Falcon@outlook.com>
This commit is contained in:
anon189Ty
2025-10-17 18:14:49 +08:00
committed by GitHub
parent 1b424fb7f1
commit 46e62efd44
6 changed files with 26 additions and 10 deletions

View File

@@ -2,7 +2,7 @@ import enum
from typing import Optional
import torch
from vllm.config import VllmConfig
from vllm.config import CUDAGraphMode, VllmConfig
from vllm.v1.core.sched.output import SchedulerOutput
from vllm.v1.sample.metadata import SamplingMetadata
from vllm.v1.spec_decode.metadata import SpecDecodeMetadata
@@ -33,7 +33,9 @@ class Proposer:
with_prefill: bool = False,
skip_attn: bool = False,
num_reqs: int = 0,
num_tokens_across_dp: Optional[torch.Tensor] = None):
num_tokens_across_dp: Optional[torch.Tensor] = None,
aclgraph_runtime_mode: CUDAGraphMode = CUDAGraphMode.NONE,
batch_descriptor=None):
"""Called by dummy_run in modle_runner"""
raise NotImplementedError