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

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

Browse Source 
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
This commit is contained in:
Cyrus Leung
2026-01-12 17:19:30 +08:00
committed by GitHub
parent 60446cd684
commit 600aaab8d6
6 changed files with 6 additions and 15 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
vllm/model_executor/models/qwen3_next_mtp.py
View File

@@ -27,7 +27,6 @@ from vllm.model_executor.models.qwen3_next import (
from vllm.sequence import IntermediateTensors
from vllm.transformers_utils.configs import Qwen3NextConfig
from .interfaces import SupportsPP
from .utils import (
AutoWeightsLoader,
is_pp_missing_parameter,
@@ -221,7 +220,7 @@ class Qwen3NextMultiTokenPredictor(nn.Module):
@support_torch_compile
class Qwen3NextMTP(nn.Module, SupportsPP, QwenNextMixtureOfExperts):
class Qwen3NextMTP(nn.Module, QwenNextMixtureOfExperts):
packed_modules_mapping = {
"qkv_proj": [
"q_proj",
@@ -253,9 +252,6 @@ class Qwen3NextMTP(nn.Module, SupportsPP, QwenNextMixtureOfExperts):
prefix=maybe_prefix(prefix, "lm_head"),
)
self.logits_processor = LogitsProcessor(config.vocab_size)
self.make_empty_intermediate_tensors = (
self.model.make_empty_intermediate_tensors
)
self.set_moe_parameters()
def embed_input_ids(self, input_ids: torch.Tensor) -> torch.Tensor: