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 14, 05:30 UTC
[Kernel] [Mamba] Remove BLOCK_H=1 from list of tuneable configurations for `_chunk_cumsum_fwd_kernel` (#25197)
Signed-off-by: Thomas Parnell <tpa@zurich.ibm.com> Co-authored-by: Chih-Chieh-Yang <chih.chieh.yang@ibm.com>
Thomas Parnell committed
Sep 24, 2025 at 01:23 UTC
5e25b122366a639f6da9fa568b532d6932c45aff
1 file changed
-1
vllm/model_executor/layers/mamba/ops/ssd_chunk_state.py
-1
@@ -17,7 +17,6 @@ from .mamba_ssm import softplus
17
18
@triton.autotune(
19
configs=[
20
- triton.Config({'BLOCK_SIZE_H': 1}),
20
triton.Config({'BLOCK_SIZE_H': 2}),
21
triton.Config({'BLOCK_SIZE_H': 4}),
22
triton.Config({'BLOCK_SIZE_H': 8}),