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

Fix flashinfer backend with pcg (#20061)

Browse Source
This commit is contained in:
Qiaolin Yu
2026-03-06 20:01:43 -08:00
committed by GitHub
parent 8a411a9a2a
commit 925185f9ec
1 changed files with 4 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files
python/sglang/srt/layers/attention/flashinfer_backend.py
+4 -1
View File
@@ -16,6 +16,7 @@ from typing import TYPE_CHECKING, Callable, List, Optional, Union
import torch import torch
from sglang.srt.compilation.piecewise_context_manager import is_in_piecewise_cuda_graph
from sglang.srt.dllm.config import DllmConfig from sglang.srt.dllm.config import DllmConfig
from sglang.srt.environ import envs from sglang.srt.environ import envs
from sglang.srt.layers.attention.base_attn_backend import AttentionBackend from sglang.srt.layers.attention.base_attn_backend import AttentionBackend
@@ -479,7 +480,9 @@ class FlashInferAttnBackend(AttentionBackend):
use_ragged = False use_ragged = False
extend_no_prefix = False extend_no_prefix = False
else: else:
use_ragged = not self.enable_deterministic use_ragged = (
not self.enable_deterministic and not is_in_piecewise_cuda_graph()
)
extend_no_prefix = not any(forward_batch.extend_prefix_lens_cpu) extend_no_prefix = not any(forward_batch.extend_prefix_lens_cpu)
# Process multi-item scoring in attention backend instead of ForwardBatch # Process multi-item scoring in attention backend instead of ForwardBatch