Megatron-DeepSpeed
a34ca7f2 - [BNB] integrate `StableEmbeding` into `VocabParallelEmbedding` logic (#182)

Commit
4 years ago
[BNB] integrate `StableEmbeding` into `VocabParallelEmbedding` logic (#182) * move checks into args; undo StableEmbedding * integrate StableEmedding into Embedding * cleanup * ensure tp>1 is first for bnb
Author
Parents
Loading