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

[Model] Pooling models default to using chunked prefill & prefix caching if supported. (#20930)

Browse Source 
Signed-off-by: wang.yuqi <noooop@126.com>
This commit is contained in:
wang.yuqi
2025-08-12 00:41:37 +08:00
committed by GitHub
parent 16fb668b61
commit 84cf78acee
31 changed files with 452 additions and 261 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
vllm/model_executor/models/internlm2.py
View File

@@ -31,7 +31,7 @@ from vllm.model_executor.model_loader.weight_utils import default_weight_loader
from vllm.model_executor.sampling_metadata import SamplingMetadata
from vllm.sequence import IntermediateTensors
from .interfaces import SupportsLoRA, SupportsPP
from .interfaces import SupportsLoRA, SupportsPP, default_pooling_type
from .utils import (is_pp_missing_parameter,
make_empty_intermediate_tensors_factory, make_layers,
maybe_prefix)
@@ -401,6 +401,7 @@ class InternLM2ForCausalLM(nn.Module, SupportsPP, SupportsLoRA):
return loaded_params
@default_pooling_type("ALL")
class InternLM2ForRewardModel(InternLM2ForCausalLM):
is_pooling_model = True