[PERF] Decouple projections from GDN custom op. Attempt 2 (#28083)

Signed-off-by: Vadim Gimpelson <vadim.gimpelson@gmail.com>
2025-11-06 05:01:12 +04:00
parent 1767658559
commit b6a248bdd7
3 changed files with 205 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",
    ]