llama.cpp
[Mirror] llama: dynamic head_dim and n_rot for SWA
#89
Closed

Commits
  • llama: dynamic head_dim and n_rot for SWA
    ngxson committed 101 days ago
  • also add gguf_writer wrappers
    ngxson committed 101 days ago
  • fix build
    ngxson committed 101 days ago
  • build_rope_shift arg reorder
    ngxson committed 101 days ago
Loading