redlock

New deepseek paper: Natively Trainable Sparse Attention mechanism

5 points by redlock 4 months ago | 1 comment