Steven Gonsalvez

Software Engineer

MoonshotAI/FlashKDA: FlashKDA: high-performance Kimi Delta Attention kernels

Why CEREBRO kept it

High-perf LLM attention kernels, token optimization

The text below is an automated extraction of the article at https://github.com/MoonshotAI/FlashKDA, stored verbatim in the public cerebro-vault repository. Copyright remains with the original publisher (github.com).

FlashKDA: Flash Kimi Delta Attention — high-performance KDA kernels built on CUTLASS - 2026-04-22 — Deep-Dive Blog: the design decisions behind FlashKDA v1, read it here. - SM90 and above - CUDA 12.9 and above - PyTorch 2.4 and above git clone https://github.com/MoonshotAI/FlashKDA.git flash-kda cd flash-kda git submodule update --init --recursive pip install -v --no-build-isolation .By default, the build detects the current CUDA device and compiles for that architecture. For wheel or CI builds, compile all supported architectures explicitly: FLASH_KDA_CUDA_ARCHS=all pip install -v --no-build-

Backlinks

Appeared in 1 briefing

Related

Shares tags: ai/llm-mechanics

Also from github.com