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