Multihead Attention Fixes#209

Merged

milancurcic merged 9 commits intomodern-fortran:mainfrom

OneAdder:multihead_attention_optimization

Mar 26, 2025

Commits on Feb 26, 2025

multihead_attention_optimization: refactoring, split methods even more (will be needed for llama attention)
OneAdder
committed
multihead_attention_optimization: make attributes public
OneAdder
committed

Commits on Feb 27, 2025

multihead_attention_optimization: move heads separation out of sdpa backward
OneAdder
committed

Commits on Feb 28, 2025

multihead_attention_optimization: add attention mask to self_attention
OneAdder
committed