[Feature] Support MLA chunk-prefill & prefix cache for all MLA Architecture Models#7727

Merged

Jiang-Jia-Jun merged 10 commits intoPaddlePaddle:developfrom

chang-wenbin:CHUNK_MLA

May 9, 2026

Commits on May 7, 2026

support mla chunk-prefill & prefix_cache
chang-wenbin
committed
optimized mla chunk prefill
chang-wenbin
committed
optimized mla chunk prefill
chang-wenbin
committed
update mla chunk-prefill
chang-wenbin
committed

Commits on May 8, 2026

support mla chunk-prefill & prefix cache
chang-wenbin
committed
update mla chunk prefill prefix cache
chang-wenbin
committed
update tests
chang-wenbin
committed
update test
chang-wenbin
committed

Commits on May 9, 2026

optimize triton kernel
chang-wenbin
committed
update deepseek-v3 modeling
chang-wenbin
committed