A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 15, 14:45 UTC
[Misc] remove warning if triton>=3.2.0 (#16553)
Signed-off-by: DefTruth <qiustudent_r@163.com>
DefTruth committed
Apr 14, 2025 at 17:39 UTC
ce4ddd2d1a96faf393b0dd8c057cbf2774162297
1 file changed
+6
-5
vllm/attention/ops/triton_decode_attention.py
+6
-5
@@ -39,11 +39,12 @@ is_hip_ = current_platform.is_rocm()
39
40
logger = logging.getLogger(__name__)
41
42
-# TODO: Remove this when triton>=3.2.0. This issue will not affect performance
43
-# and accuracy.
44
-logger.warning(
45
- "The following error message 'operation scheduled before its operands' "
46
- "can be ignored.")
42
+# Only print the following warnings when triton version < 3.2.0.
43
+# The issue won't affect performance or accuracy.
44
+if triton.__version__ < '3.2.0':
45
+ logger.warning(
46
+ "The following error message 'operation scheduled before its operands' "
47
+ "can be ignored.")
48
49
50
@triton.jit