It’s a great paper but, at best, it just lets you fit some more context as far as I can tell. Recent hybrid models are so efficient cache-wise that this just feels like a marginal improvement. I never saw this much hype surrounding other quantization-related improvements. Meanwhile, I feel like ther
