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
[Bugfix] Fix transform_config parsing in Compressed Tensors (#23945)
Signed-off-by: Kyle Sayers <kylesayrs@gmail.com>
Kyle Sayers committed
Sep 2, 2025 at 13:54 UTC
1c4131058465db9966ffc3a701a86a479216d0ea
1 file changed
+2
-2
vllm/model_executor/layers/quantization/compressed_tensors/compressed_tensors.py
+2
-2
@@ -63,7 +63,7 @@ class CompressedTensorsConfig(QuantizationConfig):
63
sparsity_ignore_list: list[str],
64
kv_cache_scheme: Optional[dict[str, Any]] = None,
65
config: Optional[dict[str, Any]] = None,
66
- transform_config: Optional[TransformConfig] = None,
66
+ transform_config: Optional[dict[str, Any]] = None,
67
):
68
super().__init__()
69
self.ignore = ignore
@@ -75,7 +75,7 @@ class CompressedTensorsConfig(QuantizationConfig):
75
self.sparsity_ignore_list = sparsity_ignore_list
76
self.config = config
77
78
- if transform_config is not None:
78
+ if transform_config:
79
self.transform_config = TransformConfig.model_validate(
80
transform_config)
81
else: