chenchenghao/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 11 Packages Projects Releases Wiki Activity

Revert "feat: reduce constrained-decoding overhead in TP" (#16845)

Browse Source
This commit is contained in:
Liangsheng Yin
2026-01-10 11:39:38 +08:00
committed by GitHub
parent fbc128a32e
commit e6d40bff81
7 changed files with 73 additions and 231 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
python/sglang/srt/sampling/sampling_batch_info.py
View File

@@ -201,11 +201,7 @@ class SamplingBatchInfo:
return
# Find a grammar from the list
first_grammar = next((grammar for grammar in self.grammars if grammar), None)
if first_grammar is None:
self.vocab_mask = None
self.apply_mask_func = None
return
first_grammar = next(grammar for grammar in self.grammars if grammar)
# TODO(lianmin): Maybe we can reuse the existing mask?
self.vocab_mask = first_grammar.allocate_vocab_mask(
@@ -236,7 +232,7 @@ class SamplingBatchInfo:
else:
self.acc_linear_penalties = None
def apply_logits_bias(self, logits: torch.Tensor, apply_vocab_mask: bool = True):
def apply_logits_bias(self, logits: torch.Tensor):
if self.acc_linear_penalties is not None:
# Used in the overlap mode
logits.add_(self.acc_linear_penalties)
@@ -245,7 +241,7 @@ class SamplingBatchInfo:
# Used in the non-overlap mode
self.penalizer_orchestrator.apply(logits)
if apply_vocab_mask and self.vocab_mask is not None:
if self.vocab_mask is not None:
self.apply_mask_func(logits=logits, vocab_mask=self.vocab_mask)
if self.logit_bias is not None: