[Migration] Migrate GGUF quantization support to plugin (#39612)

Signed-off-by: Isotr0py <mozf@mail2.sysu.edu.cn>
This commit is contained in:
Isotr0py
2026-06-12 12:02:21 -07:00
committed by GitHub
parent d6fd7ce8da
commit 6635279d8a
57 changed files with 72 additions and 9048 deletions
-3
View File
@@ -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,