Skip to content

Version 1.0.1

Choose a tag to compare

@mvds00 mvds00 released this 21 Aug 08:39
· 169 commits to main since this release

This release contains may small and some large improvements:

  • sliced Llama, Phi and Phi3 evaluation, unlocking virtually unlimited model size
  • 30% speedup of model evaluation
  • period re-evaluation of all models
  • sync model/metagraph rewrite
  • subprocess logging improvement
  • improved error handling and reporting
  • reduction/prevention of excessive errors/logging
  • logging of more statistics
  • even more runtime configurable parameters
  • improved competitions.json format
  • fixes for tokenizer hacks
  • allow for a limited number of tokenization issues (true fix under investigation)
  • various clarifications and simplifications