UPDATE: Qwen 3.8 27B 140 tok/s on single RTX 3090 Megakernel: KL divergence 0.0009 vs llama.cpp

🌐 链接: https://www.reddit.com/r/LocalLLaMA/comments/1x37q7b/update_qwen_38_27b_140_toks_on_single_rtx_3090/
🏷️ 分组: Reddit｜AI·自托管·VPS
🕒 时间: 2026-10-11 21:04:42
Reddit
From the LocalLLaMA community on Reddit
From the LocalLLaMA community on Reddit
Explore this post and more from the LocalLLaMA community