Skip to content

NVIDIA Megatron Core 0.9.0

Latest
Compare
Choose a tag to compare
@ko3n1g ko3n1g released this 24 Oct 10:30
· 10 commits to core_r0.9.0 since this release
  • Uneven pipeline parallelism
    • Enable pipeline parallelism where first and last ranks have fewer transformer layers than the intermediate ranks
  • Per layer CUDAGraph support for GPT training with Transformer Engine modules
  • Enable different TP sizes for the vision encoder
  • Enable pipeline parallelism for T5 & Llava models
  • Support multi-tile multi-image input in Llava models
  • MoE
    • FP8 support
    • Runtime upcycling support
    • Dispatcher implementation optimizations
    • Shared expert support with overlapping optimizations
      • Qwen Model support
  • Known Issues
    • When using sequence parallel, during the transformer block forward pass, dropout is not using the appropriate rng context.