"The layer-by-layer streaming is clever. Curious how the throughput compares to running a quantized model on the same GPU — seems like quantization might stil

"The layer-by-layer streaming is clever. Curious how the throughput compares to running a quantized model on the same GPU — seems like quantization might still win on speed?"