DeepSeek V2/V3 MLA implementation#12801

Merged

ggerganov merged 18 commits into

ggml-org:masterfrom

jukofyork:mla--ready-for-review

Apr 15, 2025

Commits on Apr 7, 2025

Merged using squash to remove all noise commit messages
jukofyork
committed
Force flash attention off for `LLM_ARCH_DEEPSEEK2` - embedding too large
jukofyork
committed

Commits on Apr 9, 2025

Merge branch 'ggml-org:master' into mla--ready-for-review
jukofyork
authored

Commits on Apr 11, 2025

Merge branch 'ggml-org:master' into mla--ready-for-review
jukofyork
authored

Commits on Apr 12, 2025