Simplify the heuristics for setting --mem-fraction-static (#7054)

2025-06-10 19:01:39 -07:00
parent 0f218731e3
commit 6b12d6a8d5
3 changed files with 57 additions and 50 deletions
--- a/python/sglang/srt/layers/attention/triton_ops/decode_attention.py
+++ b/python/sglang/srt/layers/attention/triton_ops/decode_attention.py
@@ -31,11 +31,6 @@ _is_hip = is_hip()

 logger = logging.getLogger(__name__)

-# TODO: Remove this when triton>=3.2.0. This issue will not affect performance and accuracy.
-logger.warning(
-    "The following error message 'operation scheduled before its operands' can be ignored."
-)
-

 _MIN_BLOCK_KV = 32