llama.cpp
args: refactor mlock/mmap/directio into load-mode
#20834
Merged

Commits
  • args: overhaul mmap/mlock/dio into single arg
    taronaeo committed 84 days ago
  • docs: update docs with llama-gen-docs
    taronaeo committed 84 days ago
  • chore: satisfy code quality
    taronaeo committed 84 days ago
  • args: make the `+` sign an actual modifier now
    taronaeo committed 84 days ago
  • chore: general code clean up + comments
    taronaeo committed 84 days ago
  • arg: fix deprecated flags support
    taronaeo committed 84 days ago
  • arg: quick sanity check
    taronaeo committed 84 days ago
  • bench: sync llama-bench argument parsing
    taronaeo committed 84 days ago
  • fix: bugfix variable behaviour + llama-bench lm column size
    taronaeo committed 84 days ago
  • arg: inverse commands should do the opposite instead of doing nothing
    taronaeo committed 84 days ago
  • bench: fix incorrect dash
    taronaeo committed 84 days ago
  • bench: fix missing modifiers for deprecated flags
    taronaeo committed 84 days ago
  • llama: switch back to thread_local
    taronaeo committed 84 days ago
  • arg: switch back to single enum
    taronaeo committed 84 days ago
  • docs: update arg docs
    taronaeo committed 84 days ago
  • chore: fix missing `mlock` from llama_load_mode_from_str + cleanup llama-bench
    taronaeo committed 84 days ago
  • llama: fix mlock not activating
    taronaeo committed 84 days ago
  • arg: add deprecation warning when old and new flags are combined
    taronaeo committed 84 days ago
  • arg: cont add comment for todo in the future
    taronaeo committed 84 days ago
  • docs: sync with upstream
    taronaeo committed 84 days ago
  • docs: re-sync with upstream again
    taronaeo committed 84 days ago
Loading