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

Use device_group for all_gather when disabling overlap scheduling (#8001)

Browse Source
This commit is contained in:
Qiaolin Yu
2025-07-15 19:38:58 -07:00
committed by GitHub
parent 3bc43c683e
commit 69f453e5a4
2 changed files with 15 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
python/sglang/bench_one_batch.py
View File

@@ -271,12 +271,13 @@ def _maybe_prepare_mlp_sync_batch(batch: ScheduleBatch, model_runner):
batch,
dp_size=model_runner.server_args.dp_size,
attn_tp_size=1,
tp_cpu_group=model_runner.tp_group.cpu_group,
tp_group=model_runner.tp_group,
get_idle_batch=None,
disable_cuda_graph=model_runner.server_args.disable_cuda_graph,
spec_algorithm=SpeculativeAlgorithm.NONE,
speculative_num_draft_tokens=None,
require_mlp_tp_gather=require_mlp_tp_gather(model_runner.server_args),
disable_overlap_schedule=model_runner.server_args.disable_overlap_schedule,
)