Search papers, labs, and topics across Lattice.
University of California
1
0
2
Sparse, quantized linear-attention models can outperform dense counterparts on neuromorphic hardware, achieving up to 37脳 higher throughput and 16脳 lower power consumption.
