[PERF] Decouple projections from GDN custom op (#27512)

Signed-off-by: Vadim Gimpelson <vadim.gimpelson@gmail.com>
2025-11-04 20:11:41 +04:00
parent 97e3dda84b
commit 5fd8f02ea9
3 changed files with 204 additions and 53 deletions
--- a/vllm/config/compilation.py
+++ b/vllm/config/compilation.py
@@ -462,7 +462,7 @@ class CompilationConfig:
        "vllm::short_conv",
        "vllm::linear_attention",
        "vllm::plamo2_mamba_mixer",
-        "vllm::gdn_attention",
+        "vllm::gdn_attention_core",
        "vllm::kda_attention",
        "vllm::sparse_attn_indexer",
    ]