#Mixtral-8x7B

2 stories · 0 in 7 days · 0 in 30 days

DeepSeek-V2-Lite 2 MoE 2 MaskCoFT 1 Qwen3-30B-A3B 1 推理优化 1 长上下文 1

Stories, products and papers tagged "Mixtral-8x7B", newest first.

Sep 29

Papers Builders13:35

Chinese contentMaskCoFT：掩码协同微调降低 MoE 推理显存开销

跑 MoE 显存不够的人可以看看：微调路由加专家，Mixtral-8x7B 每层只缓存 4 个专家，取回次数降了 23.7%，速度更快精度还涨。

#MaskCoFT #Mixtral-8x7B #DeepSeek-V2-Lite #MoE

arXiv: DeepSeek@Junfeng Wu et al. (6)Source

Sep 15

Models BuildersSelected13:06

Physically Partitioned KVCache Format for MoE Inference

This research provides a method to improve the performance of large language models on GPUs by better managing memory usage, which is relevant for anyone working with or evaluating MoE models.

#ai-models #Mixture-of-Experts #MoE #inference

arXiv: DeepSeek@Enda Yu et al. (3)Source