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 6, 20:30 UTC
[Kernel] [Helion] [14/N] Set autotune_ignore_errors=True during autotuning (#36683)
Signed-off-by: Yanan Cao <gmagogsfm@gmail.com> Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
Yanan Cao committed
Mar 11, 2026 at 14:25 UTC
0ce21c46a055c4dc89d58b38f3ff62759011801b
1 file changed
+4
-1
vllm/kernels/helion/register.py
+4
-1
@@ -395,7 +395,10 @@ class HelionKernelWrapper:
395
autotune_effort: str = "quick",
396
) -> Config:
397
"""Run autotuning for a single input configuration."""
398
- extra_kwargs = {"autotune_effort": autotune_effort}
398
+ extra_kwargs = {
399
+ "autotune_effort": autotune_effort,
400
+ "autotune_ignore_errors": True,
401
+ }
402
autotune_kernel = create_helion_decorated_kernel(
403
self.raw_kernel_func, self.helion_settings, extra_kwargs
404
)