[Paper] Specter: Fast MoE Inference via Quantization-Enabled Multi-Step Speculative Offloading
🌐 链接: https://www.reddit.com/r/LocalLLaMA/comments/1xbj6an/paper_specter_fast_moe_inference_via/
🏷️ 分组: Reddit｜AI·自托管·VPS
🕒 时间: 2026-10-21 18:09:24![From the LocalLLaMA community on Reddit: [Paper] Specter: Fast MoE Inference via Quantization-Enabled Multi-Step Speculative Offloading](/static/https://cdn4.telesco.pe/file/AX4ilPJAv4os09ZqiW0Ltbsftm6CptE6BWpva1126CxfiGwyQNPJsertuwwFtLk6KlXlt3sUJII129LYhzbSTexeOeaktFfTO27UbqdW5EthYjZYJsSDo6pA0oXTLTkKYnWI5xmwIqq94tskR2Vxnm4j4ubTZnTsYk9kttNBXSopnQmcrjgglZOW5soMJZiMQSqv2BFL_gCckXXIW0bJCq64GBPft3DJQghbM-fI05yD1Rh5f19C_CUchcwjpf8696Hc16SbxoW-oIPgtAyWmz2nh6EI66OxUKD-_sPKd_4ZoTvge-nIOGXP-n5RmNj-TZzmjAxdei5wNS2TyGhz1A.jpg)
🌐 链接: https://www.reddit.com/r/LocalLLaMA/comments/1xbj6an/paper_specter_fast_moe_inference_via/
🏷️ 分组: Reddit｜AI·自托管·VPS
🕒 时间: 2026-10-21 18:09:24
Reddit
From the LocalLLaMA community on Reddit: [Paper] Specter: Fast MoE Inference via Quantization-Enabled Multi-Step Speculative Offloading
Explore this post and more from the LocalLLaMA community
