chenchenghao/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 13 Packages Projects Releases Wiki Activity

Deprecate global_server_args_dict (#11331)

Browse Source
This commit is contained in:
Liangsheng Yin
2025-10-13 01:20:47 +08:00
committed by GitHub
parent 2157d12ae8
commit 1083e7e3df
54 changed files with 240 additions and 321 deletions
Download Patch File Download Diff File Expand all files Collapse all files
python/sglang/srt/layers/attention/flashinfer_mla_backend.py
+5 -5
View File
@@ -28,8 +28,8 @@ from sglang.srt.layers.attention.flashinfer_backend import (
create_flashinfer_kv_indices_triton,
)
from sglang.srt.layers.dp_attention import get_attention_tp_size
from sglang.srt.managers.schedule_batch import global_server_args_dict
from sglang.srt.model_executor.forward_batch_info import ForwardBatch, ForwardMode
from sglang.srt.server_args import get_global_server_args
from sglang.srt.speculative.spec_info import SpecInput
from sglang.srt.utils import (
is_flashinfer_available,
@@ -193,9 +193,9 @@ class FlashInferMLAAttnBackend(AttentionBackend):
self.skip_prefill = skip_prefill
self.enable_chunk_kv = (
not skip_prefill
and global_server_args_dict["disaggregation_mode"] != "decode"
and not global_server_args_dict["disable_chunked_prefix_cache"]
and not global_server_args_dict["flashinfer_mla_disable_ragged"]
and get_global_server_args().disaggregation_mode != "decode"
and not get_global_server_args().disable_chunked_prefix_cache
and not get_global_server_args().flashinfer_mla_disable_ragged
)
self.page_size = model_runner.page_size
@@ -306,7 +306,7 @@ class FlashInferMLAAttnBackend(AttentionBackend):
prefix_lens = forward_batch.extend_prefix_lens
extend_no_prefix = not any(forward_batch.extend_prefix_lens_cpu)
use_ragged = (
not global_server_args_dict["flashinfer_mla_disable_ragged"]
not get_global_server_args().flashinfer_mla_disable_ragged
and extend_no_prefix
)