MMODELYST
People/Mike Lewis
PPL

Mike Lewis

Papers (2)
Train Short, Test Long: Attention with Linear Biases Enables Input Length Extrapolation8-bit Optimizers via Block-wise Quantization