Search papers, labs, and topics across Lattice.
Imperial College London
1
0
2
Achieving up to 2.3x higher training throughput, Tangram transforms how we approach parallelization in heterogeneous GPU clusters.
