From 8305f66fc84fcfe96d9c01d41d3b9e8a7c58721c Mon Sep 17 00:00:00 2001 From: abing Date: Sat, 19 Sep 2026 02:32:27 -0700 Subject: [PATCH] fix bug (cherry picked from commit 60d31cb0c268c4b733ab290e133e3763b65f090e) --- python/sglang/srt/models/deepseek_v4.py | 9 ++++++--- 1 file changed, 6 insertions(+), 3 deletions(-) diff --git a/python/sglang/srt/models/deepseek_v4.py b/python/sglang/srt/models/deepseek_v4.py index 2cfe15220..8bcc19b02 100644 --- a/python/sglang/srt/models/deepseek_v4.py +++ b/python/sglang/srt/models/deepseek_v4.py @@ -4873,10 +4873,13 @@ class DeepseekV4ForCausalLM(nn.Module): self.wo_a_fp8 = wo_a_fp8_gemm_enabled(quant_config) self.determine_num_fused_shared_experts() self.vision = None - if config.model_type == "deepseek_v41" and config.vision_n_layers > 0: + if ( + config.model_type == "deepseek_v41" + and config.vision_n_layers > 0 + and not getattr(config, "language_model_only", False) + ): if ( - get_parallel().attn_cp_size != 1 - or get_parallel().pp_group.world_size != 1 + get_parallel().pp_group.world_size != 1 or not get_moe_a2a_backend().is_none() ): raise ValueError(