Kevin FrançoisinneoxiaLLm infini-attention with linear complexityIntroducing Google’s Infini-attention to increase LLM attention windows and reduce quadratic complexityApr 26Apr 26
Kevin FrançoisinneoxiaDeep dive in embeddingsFull explanation of word and image embedding with the explanation of reference models such as Bert and ClipMar 21Mar 21
Kevin FrançoisinneoxiaMixtral 8x7B explainedThis review aims to expound upon the MoE framework and explore the added benefits that Mixtral 8x7B brings to these specific fieMar 21Mar 21
Kevin FrançoisinneoxiaProxy-Tuning: A Breakthrough in Customizing Large Language ModelsFine-tune LLM through next token probabilityMar 18Mar 18