Search papers, labs, and topics across Lattice.
1
0
2
Achieve 5-16x input compression for neural networks without significant accuracy loss by learning a task-specific quantization scheme.
