FA3 support qwen3 (#5441)

2026-04-23 00:17:25 +08:00 · 2025-12-09 16:16:16 +08:00
parent 83ea9646f9
commit 31410415db
6 changed files with 242 additions and 44 deletions
@@ -190,6 +190,7 @@ std::vector<paddle::Tensor> GQARopeWriteCacheKernel(
    const int kv_token_num,
    const int max_seq_len,
    const float rms_norm_eps,
+    const bool use_neox_rotary_style,
    const std::string& cache_quant_type,
    const bool rope_3d);