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 11, 05:00 UTC
[Quantization] enable compressed-tensors marlin support for turing (#31000)
Signed-off-by: Jinzhen Lin <jinzhen.ljz@antgroup.com>
Jinzhen Lin committed
Dec 19, 2025 at 12:29 UTC
de08b8f61bd412f2bcc3a5f21da26871e39a9078
1 file changed
+1
-1
vllm/model_executor/layers/quantization/kernels/mixed_precision/marlin.py
+1
-1
@@ -30,7 +30,7 @@ from .MPLinearKernel import MPLinearKernel, MPLinearLayerConfig
30
class MarlinLinearKernel(MPLinearKernel):
31
@classmethod
32
def get_min_capability(cls) -> int:
33
- return 80
33
+ return 75
34
35
@classmethod
36
def can_implement(cls, c: MPLinearLayerConfig) -> tuple[bool, str | None]: