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。

OpenLLaMA 7B Training Completed to 1T Tokens

58 点作者 jncraton将近 2 年前

2 条评论

mdaniel将近 2 年前
be sure to read the warning in their repo: <a href="https:&#x2F;&#x2F;github.com&#x2F;openlm-research&#x2F;open_llama#loading-the-weights-with-hugging-face-transformers">https:&#x2F;&#x2F;github.com&#x2F;openlm-research&#x2F;open_llama#loading-the-we...</a><p>&gt; Please note that it is advised to avoid using the Hugging Face fast tokenizer for now, as we’ve observed that the auto-converted fast tokenizer sometimes gives incorrect tokenization
fancyfredbot将近 2 年前
This is great. Based on the throughout of 2200 tokens&#x2F;sec and the 1,000,000,000,000 tokens used to train this was at least $183k worth of compute (that&#x27;s based on the three year committed use rate). And now we can have it for free!
评论 #36233829 未加载