diff --git a/python/sglang/srt/models/olmo2.py b/python/sglang/srt/models/olmo2.py index de5087f31..97754f362 100644 --- a/python/sglang/srt/models/olmo2.py +++ b/python/sglang/srt/models/olmo2.py @@ -101,6 +101,7 @@ class Olmo2Attention(nn.Module): self.hidden_size, self.head_dim, self.total_num_heads, + total_num_kv_heads=self.total_num_kv_heads, bias=config.attention_bias, quant_config=quant_config, prefix=add_prefix("qkv_proj", prefix),