←新闻详情

X.PIN更新于 11月4日 23:25

消息称 DeepSeek 计划用华为昇腾芯片预训练下一代 V5 模型

DeepSeek plans to pre-train its next-gen V5 model on Huawei Ascend chips, per STAR Market Daily sources.

AI 导读

据 STAR Market Daily 消息,DeepSeek 计划在华为昇腾芯片上预训练下一代 V5 模型。此前昇腾只承担 DeepSeek 的推理和 V4-Pro 后训练任务。X.PIN 认为,DeepSeek 于 9 月 30 日开源其完整的昇腾计算与通信栈(DeepGEMM、FlashMLA、DeepEP)与此有关。

推荐理由

精选AI 评分 73

消息来自媒体信源,读者可据此了解 DeepSeek 训练算力从昇腾推理转向预训练的可能变化。

主题

标签

正文

DeepSeek plans to pre-train its next-gen V5 model on Huawei Ascend chips, per STAR Market Daily sources.

Until now, Ascend only handled DeepSeek inference and V4-Pro post-training. Suddenly, DeepSeek open-sourcing its full Ascend compute & comms stack (DeepGEMM, FlashMLA, DeepEP) on Sept 30 makes total sense.

来源

X.PIN