DeepSpeed
Mixtral FastGen Support
#4828
Merged

Commits
  • Kernel changes, cleanup still necessary
    cmikeh2 committed 2 years ago
  • Add explicit theta to rotary embeddings
    cmikeh2 committed 2 years ago
  • Mixtral model implementation
    cmikeh2 committed 2 years ago
  • Commit the unsaved files
    cmikeh2 committed 2 years ago
  • Minor fixes
    cmikeh2 committed 2 years ago
  • Missing engine factor, rope theta fixes
    cmikeh2 committed 2 years ago
  • MoE type mismatches
    cmikeh2 committed 2 years ago
  • Misnamed mapping
    cmikeh2 committed 2 years ago
  • Clear output
    cmikeh2 committed 2 years ago
  • Fix unit test
    cmikeh2 committed 2 years ago
  • Fix unit tests
    cmikeh2 committed 2 years ago
  • Clean up top_k support in the C++ code
    cmikeh2 committed 2 years ago
  • Merge branch 'master' into cholmes/mixtral-fastgen-support
    mrwyattii committed 2 years ago
Loading