PinnedTechnocratinCoderHack.comGlore: Memory-Efficient LLM Training by Gradient Low-Rank ProjectionGlore (Gradient Low-Rank Projection) is a memory-efficient technique for training large language models (LLMs) from scratch using a single…Mar 12Mar 12
PinnedTechnocratinCoderHack.comGitHub Actions for CI/CDContinuous integration (CI) and continuous delivery/deployment (CD) are practices that are intended to minimise mistakes when making…Sep 13, 2023Sep 13, 2023
TechnocratinCoderHack.comGrok model — first readRecently released grok1 at https://github.com/xai-org/grok-1/blob/main/model.pyMar 18Mar 18