TeraPipe: Token-Level Pipeline Parallelism for Training Large-Scale Language Models.
Zhuohan LiSiyuan ZhuangShiyuan GuoDanyang ZhuoHao ZhangDawn SongIon StoicaPublished in: ICML (2021)
Keyphrases
- document length
- language model
- language modeling
- smoothing methods
- information retrieval
- n gram
- probabilistic model
- document retrieval
- speech recognition
- document level
- query expansion
- language modelling
- retrieval model
- training set
- context sensitive
- test collection
- statistical language models
- document ranking
- pseudo relevance feedback
- translation model
- vector space model
- document collections
- ad hoc information retrieval
- relevance model
- query terms