From 6bc3062894652e5fe8a4357110fa1d18ce360a2e Mon Sep 17 00:00:00 2001 From: Vincent Zhong <207368749+vincentzed@users.noreply.github.com> Date: Thu, 20 Nov 2025 15:04:30 -0500 Subject: [PATCH] Fix launch of `Olmo3` (#13666) Signed-off-by: Vincent Zhong <207368749+vincentzed@users.noreply.github.com> --- python/sglang/srt/models/olmo2.py | 1 + 1 file changed, 1 insertion(+) 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),