diff --git a/python/sglang/srt/models/deepseek_v2.py b/python/sglang/srt/models/deepseek_v2.py index 57bae2913561..9715d147f535 100644 --- a/python/sglang/srt/models/deepseek_v2.py +++ b/python/sglang/srt/models/deepseek_v2.py @@ -1217,10 +1217,7 @@ def __init__( ) if rope_scaling and rope_scaling.get("apply_yarn_scaling", True): - mscale_all_dim = rope_scaling.get("mscale_all_dim", False) - scaling_factor = rope_scaling["factor"] - mscale = yarn_get_mscale(scaling_factor, float(mscale_all_dim)) - self.scaling = self.scaling * mscale * mscale + self.scaling = compute_mla_mscale_scaling(rope_scaling, self.scaling) else: self.rotary_emb = None self.use_deepseek_yarn_rope = rope_scaling is not None