Megatron-DeepSpeed
[WIP] dealing with multi-process noise
#193
Merged

Commits
  • deal with tokenizer logs
    stas00 committed 4 years ago
  • Trigger CI
    stas00 committed 4 years ago
  • fused_kernels.load noise
    stas00 committed 4 years ago
  • control ds replica noise
    stas00 committed 4 years ago
  • control ds replica noise
    stas00 committed 4 years ago
  • ds info log on rank0 only
    stas00 committed 4 years ago
  • only on rank 0
    stas00 committed 4 years ago
  • only if --deepspeed
    stas00 committed 4 years ago
  • fix
    stas00 committed 4 years ago
  • replica logging for transformers
    stas00 committed 4 years ago
Loading