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 8, 05:00 UTC
[Bugfix] Fix Dynamo unexpected keyword argument (#34320)
Signed-off-by: Samu Tamminen <stammine@amd.com> Co-authored-by: Wentao Ye <44945378+yewentao256@users.noreply.github.com>
Samu Tamminen committed
Feb 16, 2026 at 11:32 UTC
a5ccc85c8c98115981a39b010b36f255e6446e77
1 file changed
+3
-3
vllm/model_executor/layers/quantization/input_quant_fp8.py
+3
-3
@@ -85,7 +85,7 @@ class QuantFP8(CustomOp):
85
x: torch.Tensor,
86
scale: torch.Tensor | None = None,
87
scale_ub: torch.Tensor | None = None,
88
- **kwargs,
88
+ use_triton: bool = False,
89
) -> tuple[torch.Tensor, torch.Tensor]:
90
from vllm.model_executor.layers.quantization.utils import fp8_utils
91
@@ -135,9 +135,8 @@ class QuantFP8(CustomOp):
135
x: torch.Tensor,
136
scale: torch.Tensor | None = None,
137
scale_ub: torch.Tensor | None = None,
138
- **kwargs,
138
+ use_triton: bool = False,
139
) -> tuple[torch.Tensor, torch.Tensor]:
140
- use_triton = kwargs.get("use_triton", False)
140
if self.is_group_quant and use_triton:
141
assert scale is None, "Dynamic group quantization does not use scale"
142
@@ -171,6 +170,7 @@ class QuantFP8(CustomOp):
170
x: torch.Tensor,
171
scale: torch.Tensor | None = None,
172
scale_ub: torch.Tensor | None = None,
173
+ use_triton: bool = False,
174
):
175
if self.is_group_quant and not self.static:
176
assert scale is None, "Dynamic group quantization does not use scale"