llama.cpp
convert : use reflinks for faster conversion
#15727
Open
Overview
Commits
13
Changes
View On GitHub
Commits
  • convert : use reflinks for faster conversion
    compilade committed 143 days ago
  • convert : fix reflinks for stacked MoE tensors
    compilade committed 143 days ago
  • gguf-py : fix flake8 lint
    compilade committed 143 days ago
  • convert : detect filesystem block size for reflinks
    compilade committed 143 days ago
  • convert : use F32 operations on Mamba A_log
    compilade committed 143 days ago
  • convert : allow sharding reflinked models
    compilade committed 143 days ago
  • gguf-py : improve reflink size logging
    compilade committed 143 days ago
  • convert : more robust default ftype detection
    compilade committed 143 days ago
  • convert : remove unused field ModelTensorInfo.src_qtype
    compilade committed 143 days ago
  • gguf-py : allow previewing reflinked size on non-Linux platforms
    compilade committed 143 days ago
  • convert : better logging of partially reflinkable tensors
    compilade committed 143 days ago
  • gguf-py : handle cross-filesystem file range copies
    compilade committed 143 days ago
  • convert : for FP8, use scale type to decide auto type
    compilade committed 143 days ago
FAQ Terms Privacy Refunds Impressum
Loading