EngineX-Hygon/sglang
5
0
Fork 0
Code Issues Pull Requests Actions 7 Projects Releases Wiki Activity

upgrade xgrammar to 0.1.19 (#6129)

Browse Source
This commit is contained in:
Yixin Dong
2025-05-08 17:42:02 -04:00
committed by GitHub
parent f6f96b0521
commit 911f3ba6f4
2 changed files with 12 additions and 20 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
python/pyproject.toml
View File

@@ -42,7 +42,7 @@ runtime_common = [
"transformers==4.51.1", "transformers==4.51.1",
"uvicorn", "uvicorn",
"uvloop", "uvloop",
"xgrammar==0.1.17", "xgrammar==0.1.19",
"blobfile==3.0.0" "blobfile==3.0.0"
] ]

30
python/sglang/srt/constrained/xgrammar_backend.py
View File

@@ -18,6 +18,7 @@ import logging
from typing import List, Optional, Tuple, Union from typing import List, Optional, Tuple, Union
import torch import torch
import xgrammar
from xgrammar import ( from xgrammar import (
CompiledGrammar, CompiledGrammar,
GrammarCompiler, GrammarCompiler,
@@ -58,17 +59,11 @@ class XGrammarGrammar(BaseGrammarObject):
self.override_stop_tokens = override_stop_tokens self.override_stop_tokens = override_stop_tokens
self.finished = False self.finished = False
# Fix (from vLLM team): postpone the import of apply_token_bitmask_inplace_kernels to the from xgrammar.kernels.apply_token_bitmask_inplace_cpu import (
# class init site to avoid re-initializing CUDA in forked subprocess. apply_token_bitmask_inplace_cpu,
from xgrammar.kernels import apply_token_bitmask_inplace_kernels )
self.use_token_bitmask_triton = get_bool_env_var( self.apply_vocab_mask_cpu = apply_token_bitmask_inplace_cpu
"SGLANG_TOKEN_BITMASK_TRITON", "false"
)
self.apply_vocab_mask_cuda = apply_token_bitmask_inplace_kernels.get(
"cuda", None
)
self.apply_vocab_mask_cpu = apply_token_bitmask_inplace_kernels.get("cpu", None)
def accept_token(self, token: int): def accept_token(self, token: int):
assert self.matcher.accept_token(token) assert self.matcher.accept_token(token)
@@ -113,15 +108,12 @@ class XGrammarGrammar(BaseGrammarObject):
return vocab_mask.to(device, non_blocking=True) return vocab_mask.to(device, non_blocking=True)
def apply_vocab_mask(self, logits: torch.Tensor, vocab_mask: torch.Tensor) -> None: def apply_vocab_mask(self, logits: torch.Tensor, vocab_mask: torch.Tensor) -> None:
if ( if logits.device.type == "cuda":
not self.use_token_bitmask_triton apply_token_bitmask_inplace_triton(logits, vocab_mask)
and logits.device.type == "cuda" elif logits.device.type == "cpu" and self.apply_vocab_mask_cpu:
and self.apply_vocab_mask_cuda self.apply_vocab_mask_cpu(logits, vocab_mask)
): else:
return self.apply_vocab_mask_cuda(logits, vocab_mask) raise RuntimeError(f"Unsupported device: {logits.device.type}")
if logits.device.type == "cpu" and self.apply_vocab_mask_cpu:
return self.apply_vocab_mask_cpu(logits, vocab_mask)
apply_token_bitmask_inplace_triton(logits, vocab_mask)
def copy(self): def copy(self):
matcher = GrammarMatcher( matcher = GrammarMatcher(