[NPU]bugfix: fix for dsv3.2 and dsvl2 (#17007)

Co-authored-by: Hexq0210 <893781835@qq.com> Co-authored-by: liupeng374 <782420244@qq.com> Co-authored-by: cy <chenyang08056032@163.com>
2026-01-23 11:15:15 +08:00
parent 7ace64d1d8
commit c0b5a180fe
5 changed files with 127 additions and 44 deletions
@@ -1220,6 +1220,7 @@ class DeepseekV2AttentionMLA(nn.Module, DeepseekMHAForwardMixin):
                self.rotary_emb.forward = self.rotary_emb.forward_native
        else:
            self.rotary_emb = None
+        self.use_deepseek_yarn_rope = rope_scaling is not None

        self.attn_mqa = RadixAttention(
            self.num_local_heads,