TE
科技回声
首页24小时热榜最新最佳问答展示工作
GitHubTwitter
首页

科技回声

基于 Next.js 构建的科技新闻平台,提供全球科技新闻和讨论内容。

GitHubTwitter

首页

首页最新最佳问答展示工作

资源链接

HackerNews API原版 HackerNewsNext.js

© 2025 科技回声. 版权所有。

Mesh-TensorFlow: Model Parallelism for Supercomputers (TF Dev Summit ‘19)

3 点作者 legel大约 6 年前

1 comment

legel大约 6 年前
Mesh-TensorFlow (<a href="https:&#x2F;&#x2F;github.com&#x2F;tensorflow&#x2F;mesh" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;tensorflow&#x2F;mesh</a>) solves the problem of networks being too large to fit into a single GPU&#x27;s memory. I just trained a network with &gt; 1 billion parameters across 4 GPUs. Beside size of data set and compute power, representational capacity (size of embeddings) is a key blocker to learning, and so this library opens up new possibilities for many of us.