- 발표 자료 : https://github.com/jiphyeonjeon/season2/tree/main/advanced
★ 영상에서 다룬 내용들
- GPT 1, 2, 3
- BERT
- T5
- Switch Transformers
- Message Passing
- MPI, NCCL, DP
- Ring All-reduce
- Horovod
- DDP
- Mesh-tensorflow
- Megatron-LM
- GPipe, PipeDream, Interleaved Scheduling
- 3D Parallelism
- Mixed Precision
- ZeRO, ZeRO-offload, ZeRO-infinity
- Deep Speed
- 1-Bit Adam
- Progressive Layer Dropping