$\infty$-former: Infinite Memory Transformer

Benchmark Model Rank Results
language-modelling-on-wikitext-103[?]-former (SM)#14Test perplexity: 16.61
language-modelling-on-wikitext-103-former (SM)#15Test perplexity: 16.61
language-modelling-on-wikitext-103โˆž-former (Sticky memories + initialized GPT-2 Small)#16Test perplexity: 16.61
language-modelling-on-wikitext-103โˆž-former (initialized GPT-2 Small)#17Test perplexity: 16.64
language-modelling-on-wikitext-103[?]-former (Sticky memories)#54Test perplexity: 24.22
language-modelling-on-wikitext-103\infty-former (Sticky memories)#55Test perplexity: 24.22
language-modelling-on-wikitext-103โˆž-former (Sticky memories)#56Test perplexity: 24.22