MT-FlashMLA

MT-FlashMLA is an efficient MLA decoding kernel for MooreThreads GPU (Compute Capability 3.1), optimized for variable-length sequences serving.

Currently released:

• BF16, FP16
• Paged kvcache with block size of 64

Quick start

Install

python setup_musa.py install

Usage

from flash_mla import get_mla_metadata, flash_mla_with_kvcache

tile_scheduler_metadata, num_splits = get_mla_metadata(cache_seqlens, s_q * h_q // h_kv, h_kv)

for i in range(num_layers):
    ...
    o_i, lse_i = flash_mla_with_kvcache(
        q_i, kvcache_i, block_table, cache_seqlens, dv,
        tile_scheduler_metadata, num_splits, causal=True,
    )
    ...

Requirements

• MooreThreads GPU (Compute Capability 3.1)
• MUSA 4.0.0 and above
• torch_musa 2.5.0 and above

Acknowledgement

FlashMLA is inspired by FlashAttention 2&3 and cutlass projects.

Citation

@misc{flashmla2025,
      title={FlashMLA: Efficient MLA decoding kernel}, 
      author={Jiashi Li},
      year={2025},
      publisher = {GitHub},
      howpublished = {\url{https://github.com/deepseek-ai/FlashMLA}},
}