[Quant] [Bugfix] Fix quantization config matching with hf_to_vllm_mapper (#20046)
This commit is contained in:
@@ -467,6 +467,7 @@ class TransformersForCausalLM(nn.Module, SupportsQuant, SupportsLoRA,
|
||||
# FIXME(Isotr0py): Don't use any weights mapper for Transformers backend,
|
||||
# this makes thing complicated. We need to remove this mapper after refactor
|
||||
# `TransformersModel` in the future.
|
||||
# NOTE: `SupportsQuant` can be updated after property decorator is removed
|
||||
@property
|
||||
def hf_to_vllm_mapper(self):
|
||||
prefix_mapper = {
|
||||
|
||||
Reference in New Issue
Block a user