logo
WangZongwu
The article tags

竹杖芒鞋轻胜马，
一蓑烟雨任平生。

Scroll down
LLM Inference
H3_ Hybrid Architecture Using High Bandwidth Memory and High Bandwidth Flash for Cost-Efficient LLM Inference
  • 05/24
  • 13:45
🔐 本文已加密，请输入密码查看
SolidAttention_ Low-Latency SSD-based Serving on Memory-Constrained PCs
  • 05/19
  • 08:25
🔐 本文已加密，请输入密码查看
Adaptive Draft Sequence Length_ Enhancing Speculative Decoding Throughput on PIM-Enabled Systems
  • 05/11
  • 14:27
🔐 本文已加密，请输入密码查看
Fast On-device LLM Inference with NPUs
  • 04/06
  • 04:59
🔐 本文已加密，请输入密码查看
MoE-Lightning_ High-Throughput MoE Inference on Memory-constrained GPUs
  • 03/27
  • 12:19
🔐 本文已加密，请输入密码查看
T-MAC_ CPU Renaissance via Table Lookup for Low-Bit LLM Deployment on Edge
  • 03/24
  • 16:30
🔐 本文已加密，请输入密码查看
PD-Swap_ Prefill-Decode Logic Swapping for End-to-End LLM Inference on Edge FPGAs via Dynamic Partial Reconfiguration
  • 03/23
  • 17:20
🔐 本文已加密，请输入密码查看
LUT-LLM Efficient Large Language Model Inference with Memory-based Computations on FPGAs
  • 03/23
  • 17:13
🔐 本文已加密，请输入密码查看
Anda_ Unlocking Efficient LLM Inference with a Variable-Length Grouped Activation Data Format
  • 02/20
  • 11:04
🔐 本文已加密，请输入密码查看
Don't be so Stief! Learning KV Cache low-rank approximation over the Stiefel manifold
  • 02/20
  • 11:04
🔐 本文已加密，请输入密码查看
12
Please enter keywords to search
  • to close