Add flashinfer mla backend for deepseek v3

sgl-project · Feb 22, 2025 · d9913bb · d9913bb
1 parent 9087694
commit d9913bb
Show file tree

Hide file tree

Showing 4 changed files with 561 additions and 19 deletions.
diff --git a/python/sglang/srt/configs/model_config.py b/python/sglang/srt/configs/model_config.py
@@ -103,6 +103,7 @@ def __init__(
             self.head_dim = 256
             self.attention_arch = AttentionArch.MLA
             self.kv_lora_rank = self.hf_config.kv_lora_rank
+            self.qk_nope_head_dim = self.hf_config.qk_nope_head_dim
             self.qk_rope_head_dim = self.hf_config.qk_rope_head_dim
         elif "MiniCPM3ForCausalLM" in self.hf_config.architectures:
             self.head_dim = 128