[Quant] [Bugfix] Fix quantization config matching with hf_to_vllm_mapper (#20046)
This commit is contained in:
@@ -181,6 +181,7 @@ class ModelOptNvFp4Config(QuantizationConfig):
|
||||
exclude_modules: list[str],
|
||||
group_size: int = 16,
|
||||
) -> None:
|
||||
super().__init__()
|
||||
self.is_checkpoint_nvfp4_serialized = is_checkpoint_nvfp4_serialized
|
||||
if is_checkpoint_nvfp4_serialized:
|
||||
logger.warning(
|
||||
|
||||
Reference in New Issue
Block a user