GateGPT: 56k tokens per second Transformer (KV cache) on FPGA at 80 MHz

GateGPT: 56k tokens per second Transformer (KV cache) on FPGA at 80 MHz (twitter.com)

28 points

bylaxmena

2 hours ago |