Megatron-LM
For session 8 of the Eleuther AI ML Scalability & Performance reading group, I presented the Megatron-LM paper, which introduced tensor parallelism.
My annotated versions of these papers can be found be found on my Github here.
Papers:
Recording: