xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2026-07-09 15:27:16 +08:00
Code Issues Packages Projects Releases Wiki Activity

Upgrade xgrammar to 0.1.23 (#22988)

Browse Source 
Signed-off-by: Russell Bryant <rbryant@redhat.com>
This commit is contained in:
Russell Bryant 2025-09-02 22:32:59 -04:00 committed by GitHub
parent 42dc59dbac
commit e32a0e8678
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
2 changed files with 2 additions and 9 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
requirements/common.txt
View File

@ -25,7 +25,7 @@ outlines == 0.1.11 ; platform_machine == "s390x"
# required for outlines backend disk cache # required for outlines backend disk cache
diskcache == 5.6.3 diskcache == 5.6.3
lark == 1.2.2 lark == 1.2.2
xgrammar == 0.1.21; platform_machine == "x86_64" or platform_machine == "aarch64" or platform_machine == "arm64" xgrammar == 0.1.23; platform_machine == "x86_64" or platform_machine == "aarch64" or platform_machine == "arm64"
typing_extensions >= 4.10 typing_extensions >= 4.10
filelock >= 3.16.1 # need to contain https://github.com/tox-dev/filelock/pull/317 filelock >= 3.16.1 # need to contain https://github.com/tox-dev/filelock/pull/317
partial-json-parser # used for parsing partial JSON outputs partial-json-parser # used for parsing partial JSON outputs

9
vllm/v1/worker/gpu_model_runner.py
View File

@ -90,15 +90,11 @@ from .utils import (AttentionGroup, MultiModalBudget,
if TYPE_CHECKING: if TYPE_CHECKING:
import xgrammar as xgr import xgrammar as xgr
import xgrammar.kernels.apply_token_bitmask_inplace_torch_compile as xgr_torch_compile # noqa: E501
from vllm.model_executor.model_loader.tensorizer import TensorizerConfig from vllm.model_executor.model_loader.tensorizer import TensorizerConfig
from vllm.v1.core.sched.output import SchedulerOutput from vllm.v1.core.sched.output import SchedulerOutput
else: else:
xgr = LazyLoader("xgr", globals(), "xgrammar") xgr = LazyLoader("xgr", globals(), "xgrammar")
xgr_torch_compile = LazyLoader(
"xgr_torch_compile", globals(),
"xgrammar.kernels.apply_token_bitmask_inplace_torch_compile")
logger = init_logger(__name__) logger = init_logger(__name__)
@ -1333,10 +1329,7 @@ class GPUModelRunner(LoRAModelRunnerMixin, KVConnectorModelRunnerMixin):
# so we receive it in that format. # so we receive it in that format.
grammar_bitmask = torch.from_numpy(grammar_bitmask).contiguous() grammar_bitmask = torch.from_numpy(grammar_bitmask).contiguous()
# Force use of the torch.compile implementation from xgrammar to work xgr.apply_token_bitmask_inplace(
# around issues with the Triton kernel in concurrent structured output
# scenarios. See PR #19565 and issues #19493, #18376 for details.
xgr_torch_compile.apply_token_bitmask_inplace_torch_compile(
logits, logits,
grammar_bitmask.to(self.device, non_blocking=True), grammar_bitmask.to(self.device, non_blocking=True),
indices=out_indices if not skip_out_indices else None, indices=out_indices if not skip_out_indices else None,