[Quant] [Bugfix] Fix quantization config matching with hf_to_vllm_mapper (#20046)
This commit is contained in:
@@ -63,6 +63,7 @@ class BitBLASConfig(QuantizationConfig):
|
||||
# (since we have only one group per output channel)
|
||||
desc_act = False
|
||||
|
||||
super().__init__()
|
||||
self.weight_bits = weight_bits
|
||||
self.group_size = group_size
|
||||
self.desc_act = desc_act
|
||||
|
||||
Reference in New Issue
Block a user