diff --git a/backends/v3/src/backend.rs b/backends/v3/src/backend.rs index 0d18fbff..396287d3 100644 --- a/backends/v3/src/backend.rs +++ b/backends/v3/src/backend.rs @@ -47,7 +47,7 @@ impl BackendV3 { } else if prefix_caching { Attention::FlashInfer } else { - Attention::Paged + Attention::FlashDecoding }; let block_size = if attention == Attention::FlashDecoding { 256