diff --git a/python/sglang/srt/models/deepseek_v4.py b/python/sglang/srt/models/deepseek_v4.py index 68b4d1b08..7a564232f 100644 --- a/python/sglang/srt/models/deepseek_v4.py +++ b/python/sglang/srt/models/deepseek_v4.py @@ -1699,7 +1699,7 @@ class MQALayer(MqaAttentionBase): layer=self.attn_mqa, forward_batch=forward_batch, compress_ratio=self.compress_ratio, - attn_sink=self.attn_sink, + attn_sink=attn_sink[: self.n_local_heads], save_kv_cache=kv is not None, ) else: