xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2026-05-23 06:43:30 +08:00
Code Issues Packages Projects Releases Wiki Activity

[BugFix] Pass in correct VLLM config in FlashInfer backend (#13207) (#16973)

Browse Source 
Signed-off-by: 苏政渊 <suzhengyuan@moonshot.cn>
Co-authored-by: 苏政渊 <suzhengyuan@moonshot.cn>
This commit is contained in:
Zhengyuan Su (苏政渊) 2025-04-22 21:44:10 +08:00 committed by GitHub
parent d059110498
commit f961d7f6ef
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 3 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
vllm/attention/backends/flashinfer.py
View File

@ -37,7 +37,7 @@ from vllm.attention.backends.utils import (PAD_SLOT_ID, compute_slot_mapping,
is_block_tables_empty) is_block_tables_empty)
from vllm.attention.layer import Attention from vllm.attention.layer import Attention
from vllm.attention.ops.paged_attn import PagedAttention from vllm.attention.ops.paged_attn import PagedAttention
from vllm.config import VllmConfig, get_current_vllm_config from vllm.config import VllmConfig
from vllm.logger import init_logger from vllm.logger import init_logger
from vllm.utils import (async_tensor_h2d, get_kv_cache_torch_dtype, from vllm.utils import (async_tensor_h2d, get_kv_cache_torch_dtype,
make_tensor_with_pad) make_tensor_with_pad)
@ -187,7 +187,7 @@ class FlashInferState(AttentionState):
# Global hyperparameters shared by all attention layers # Global hyperparameters shared by all attention layers
self.global_hyperparameters: Optional[PerLayerParameters] = None self.global_hyperparameters: Optional[PerLayerParameters] = None
self.vllm_config = get_current_vllm_config() self.vllm_config = self.runner.vllm_config
def _get_workspace_buffer(self): def _get_workspace_buffer(self):
if self._workspace_buffer is None: if self._workspace_buffer is None:
@ -613,7 +613,7 @@ class FlashInferMetadataBuilder(AttentionMetadataBuilder[FlashInferMetadata]):
# Global hyperparameters shared by all attention layers # Global hyperparameters shared by all attention layers
self.global_hyperparameters: Optional[PerLayerParameters] = None self.global_hyperparameters: Optional[PerLayerParameters] = None
self.vllm_config = get_current_vllm_config() self.vllm_config = self.runner.vllm_config
def prepare(self): def prepare(self):
self.slot_mapping: List[int] = [] self.slot_mapping: List[int] = []