Reimplementation of `Improving language models by retrieving from trillions of tokens`
A minimal reimplementation of DeepMind's RETRO with training code.
ClashLuke
Back