<rss xmlns:atom="http://www.w3.org/2005/Atom" version="2.0"><channel><title>Sequence Packing - Tag - Naifan Li's Blog</title><link>https://blog.omagiclee.com/tags/sequence-packing/</link><description>Sequence Packing - Tag - Naifan Li's Blog</description><generator>Hugo -- gohugo.io</generator><language>en-us</language><lastBuildDate>Thu, 29 Jul 2021 12:16:16 +0800</lastBuildDate><atom:link href="https://blog.omagiclee.com/tags/sequence-packing/" rel="self" type="application/rss+xml"/><item><title>Summary: Sequence Packing</title><link>https://blog.omagiclee.com/posts/llms/sequence-packing/summary/</link><pubDate>Thu, 29 Jul 2021 12:16:16 +0800</pubDate><author>Naifan Li</author><guid>https://blog.omagiclee.com/posts/llms/sequence-packing/summary/</guid><description><![CDATA[<p>. Instead of padding, we concatenate multiple shorter sequences into a single, longer sequence. This minimizes wasted compute (through padding tokens)</p>
<h2 id="references">References</h2>
<ul>
<li><a href="https://arxiv.org/abs/2107.02027" target="_blank" rel="noopener noreffer ">Efficient Sequence Packing without Cross-contamination: Accelerating Large Language Models without Impacting Performance</a></li>
<li><a href="https://huggingface.co/blog/sirluk/llm-sequence-packing" target="_blank" rel="noopener noreffer ">https://huggingface.co/blog/sirluk/llm-sequence-packing</a></li>
<li><a href="https://docs.nvidia.com/nemo-framework/user-guide/24.12/nemotoolkit/features/optimizations/sequence_packing.html" target="_blank" rel="noopener noreffer ">https://docs.nvidia.com/nemo-framework/user-guide/24.12/nemotoolkit/features/optimizations/sequence_packing.html</a></li>
<li><a href="https://docs.pytorch.org/docs/stable/generated/torch.nn.utils.rnn.pack_sequence.html" target="_blank" rel="noopener noreffer ">https://docs.pytorch.org/docs/stable/generated/torch.nn.utils.rnn.pack_sequence.html</a></li>
</ul>]]></description></item></channel></rss>