I built SpectralBERT an FFT-based alternative to Attention that's 14.5x faster at 65K tokens with better loss. Is this legit?

HF

Machine LearningHugging Face Blog10h

Machine LearningTowards Data ScienceAnanya Bhattacharyya10h

TD

Machine LearningThe DecoderMaximilian Schreiner11h

AS

Machine LearningarXiv stat.MLYuxi Chen, Hamza Golubovi…22h

AS

Machine LearningarXiv stat.MLMaria Oprea, Qin Li, Yuna…22h

RR

Machine LearningReddit r/computervision/u/50th-century1d

More in Machine Learning