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

[bugfix] Fix missing args in bench one batch (#8877)

Browse Source
This commit is contained in:
Trevor Morris
2025-08-09 01:34:03 -07:00
committed by GitHub
parent 5018809222
commit 52e1f52f32
1 changed files with 6 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
python/sglang/bench_one_batch.py
View File

@@ -61,6 +61,7 @@ from sglang.srt.configs.model_config import ModelConfig
from sglang.srt.distributed.parallel_state import destroy_distributed_environment from sglang.srt.distributed.parallel_state import destroy_distributed_environment
from sglang.srt.entrypoints.engine import _set_envs_and_config from sglang.srt.entrypoints.engine import _set_envs_and_config
from sglang.srt.hf_transformers_utils import get_tokenizer from sglang.srt.hf_transformers_utils import get_tokenizer
from sglang.srt.layers.moe.utils import DeepEPMode, MoeA2ABackend
from sglang.srt.managers.schedule_batch import Req, ScheduleBatch from sglang.srt.managers.schedule_batch import Req, ScheduleBatch
from sglang.srt.managers.scheduler import Scheduler from sglang.srt.managers.scheduler import Scheduler
from sglang.srt.model_executor.forward_batch_info import ForwardBatch from sglang.srt.model_executor.forward_batch_info import ForwardBatch
@@ -300,6 +301,11 @@ def _maybe_prepare_mlp_sync_batch(batch: ScheduleBatch, model_runner):
disable_cuda_graph=model_runner.server_args.disable_cuda_graph, disable_cuda_graph=model_runner.server_args.disable_cuda_graph,
spec_algorithm=SpeculativeAlgorithm.NONE, spec_algorithm=SpeculativeAlgorithm.NONE,
speculative_num_draft_tokens=None, speculative_num_draft_tokens=None,
enable_two_batch_overlap=model_runner.server_args.enable_two_batch_overlap,
enable_deepep_moe=MoeA2ABackend(
model_runner.server_args.moe_a2a_backend
).is_deepep(),
deepep_mode=DeepEPMode(model_runner.server_args.deepep_mode),
require_mlp_tp_gather=require_mlp_tp_gather(model_runner.server_args), require_mlp_tp_gather=require_mlp_tp_gather(model_runner.server_args),
disable_overlap_schedule=model_runner.server_args.disable_overlap_schedule, disable_overlap_schedule=model_runner.server_args.disable_overlap_schedule,
) )