Samip Dahal

ZeroQuant quantization kernels and LKD

10

Hi, I was trying out the compression library for ZeroQuant quantization (for GPT-J model). While I was able to compress the model, I didn't see any throughput/latency gain from the...