[Migration] Migrate GGUF quantization support to plugin (#39612)
Signed-off-by: Isotr0py <mozf@mail2.sysu.edu.cn>
This commit is contained in:
@@ -228,9 +228,6 @@ class ApertusAttention(nn.Module):
|
||||
quant_config: QuantizationConfig | None,
|
||||
) -> None:
|
||||
is_neox_style = True
|
||||
is_gguf = quant_config and quant_config.get_name() == "gguf"
|
||||
if is_gguf and config.model_type == "apertus":
|
||||
is_neox_style = False
|
||||
|
||||
self.rotary_emb = get_rope(
|
||||
self.head_dim,
|
||||
|
||||
Reference in New Issue
Block a user