minke.yu/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 12 Packages Projects Releases Wiki Activity

[fix] Fix dynamic chunking profiling crash on GLM-5 models (#23060)

Browse Source 
Co-authored-by: liubaichuan <liubaichuan@infini-ai.com>
This commit is contained in:
Baichuan
2026-04-23 19:30:57 +08:00
committed by GitHub
co-authored by liubaichuan
parent 4868e367f8
commit 54e21bb3a5
1 changed files with 3 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files
python/sglang/srt/managers/scheduler_pp_mixin.py
+3
View File
@@ -20,6 +20,7 @@ from sglang.srt.layers.dp_attention import (
get_attention_dp_rank, get_attention_dp_rank,
get_attention_dp_size, get_attention_dp_size,
is_dp_attention_enabled, is_dp_attention_enabled,
set_is_extend_in_batch,
) )
from sglang.srt.managers.schedule_batch import Req, ScheduleBatch from sglang.srt.managers.schedule_batch import Req, ScheduleBatch
from sglang.srt.managers.utils import ( from sglang.srt.managers.utils import (
@@ -631,6 +632,8 @@ class SchedulerPPMixin:
model_worker_batch = batch.get_model_worker_batch() model_worker_batch = batch.get_model_worker_batch()
forward_batch = ForwardBatch.init_new(model_worker_batch, model_runner) forward_batch = ForwardBatch.init_new(model_worker_batch, model_runner)
set_is_extend_in_batch(batch.forward_mode.is_extend())
_ = model_runner.forward( _ = model_runner.forward(
forward_batch=forward_batch, pp_proxy_tensors=pp_proxy forward_batch=forward_batch, pp_proxy_tensors=pp_proxy
) )