transformers
Continuous batching refactor
#40426
Merged

Commits
  • Rework of the CB example
    remi-or committed 120 days ago
  • Further rework of CB example
    remi-or committed 120 days ago
  • Refactor PA cache, slice on tokens, add debug prints -- WIP
    remi-or committed 120 days ago
  • Slice cache -- WIP
    remi-or committed 120 days ago
  • Added a mechanism to check batched outputs in CB script
    remi-or committed 120 days ago
  • Less logging, debug flag for slice, !better reset! -- WIP
    remi-or committed 120 days ago
  • QOL and safety margins
    remi-or committed 120 days ago
  • Refactor and style
    remi-or committed 120 days ago
  • Better saving of cb example
    remi-or committed 120 days ago
  • Fix
    remi-or committed 120 days ago
  • Fixes and QOL
    remi-or committed 120 days ago
  • Mor einformations about metrics
    remi-or committed 120 days ago
  • Further logging
    remi-or committed 120 days ago
  • Style
    remi-or committed 120 days ago
  • Licenses
    remi-or committed 120 days ago
  • Removed some comments
    remi-or committed 120 days ago
  • Add a slice input flag
    remi-or committed 120 days ago
  • Fix in example
    remi-or committed 120 days ago
  • Added back some open-telemetry deps
    remi-or committed 120 days ago
  • Removed some aux function
    remi-or committed 120 days ago
  • Added FA2 option to example script
    remi-or committed 120 days ago
  • Fixed math (all of it)
    remi-or committed 120 days ago
  • Added a simple example
    remi-or committed 120 days ago
  • Renamed core to classes
    remi-or committed 120 days ago
  • Made allocation of attention mask optionnal
    remi-or committed 120 days ago
  • Style
    remi-or committed 120 days ago
  • Merge branch 'main' into conbat
    remi-or committed 120 days ago
Loading