specific method · filed under inference & serving
FlashAttention 3
The third version of FlashAttention, available as an attention backend.
Also called --attention-backend fa3, FlashAttention 3 (fa3).
- sources
- 2
- model
- 1
- lab adopt it
- 1
- strongest
- used
How sources treat it
One count per evidence span, weakest treatment to strongest.
optional 1used 1
Documented in
Evidence
2 spans quoted from the sources, strongest treatment first.
--attention-backend fa3 --mm-attention-backend fa3
usedinference servingin MiMo-V2.5Xiaomi
--attention-backend fa3
optionalinference servingin MiMo-V2.5Xiaomi
Filed alongside
Other methods under inference & serving :: inference kernel.
Batch-invariant deterministic kernelsCUDA GraphCUDA Graph capture size reductionKernel fusionMoE-side chunkingSynchronization-free static-shape MoE executionTensorRT-LLM multi-head attention backendAvoiding split-KDeepGEMM-based batch-invariant matrix multiplicationDistributed shared memory for cross-SM attention data exchangeDual-kernel batch-invariant attention decodingDynamic load balancingExp-free TopK kernelExpert-optimized Triton kernelsFA4 sheared-bias attention kernelFlashAttentionFlashAttention 4FlashKDAFP8 GEMMFused AttnRes merge and RMSNorm kernelFused latent down-projection and MoE-router GEMMFused mHC kernelsFused QSA kernelFused RoPE-attention-RoPE-cast kernel