{"id":520,"date":"2024-12-12T07:03:57","date_gmt":"2024-12-12T07:03:57","guid":{"rendered":"https:\/\/mailitics.com\/index.php\/2024\/12\/12\/2412-07902\/"},"modified":"2024-12-12T07:03:57","modified_gmt":"2024-12-12T07:03:57","slug":"2412-07902","status":"publish","type":"post","link":"https:\/\/mailitics.com\/index.php\/2024\/12\/12\/2412-07902\/","title":{"rendered":"Low-Rank Correction for Quantized LLMs"},"content":{"rendered":"

Low-Rank Correction for Quantized LLMs
\n \t

\n
<\/BR>
\n
\n \t

\n
<\/BR><\/p>\n

arXiv:2412.07902v1 Announce Type: new
\nAbstract: We consider the problem of model compression for Large Language Models (LLMs) at post-training time, where the task is to compress a well-trained model using only a small set of calibration input data. In this work, we introduce a new low-rank approach to correct for quantization errors of emph{activations} in LLMs: we propose to add low-rank weight matrices in full precision that act on the emph{unquantized} activations. We then solve a joint optimization problem over the quantized representation of the weights and additional low-rank weight matrices to quantize both weights and activations. We focus on the case of 4-bit weight-and-activation quantization (W4A4). Using ranks equivalent to 10% of the original weight matrix size, our approach reduces the accuracy gap with the original model by more than 50%. Using ranks equivalent to 30% of the original weight matrix, the accuracy gap is closed completely. We demonstrate our results on four recent LLMs, namely Llama-2, Llama-3, Phi-3 and Mixtral models.<\/div>\n

\t

\n
<\/BR>
\n Meyer Scetbon, James Hensman
\n \t

\n
<\/BR>
\nGo to original source<\/a>
\n \t

\n
<\/BR><\/p>\n","protected":false},"excerpt":{"rendered":"

Low-Rank Correction for Quantized LLMs arXiv:2412.07902v1 Announce Type: new Abstract: We consider the problem of model compression for Large Language Models (LLMs) at post-training time, where the task is to compress a well-trained model using only a small set of calibration input data. In this work, we introduce a new low-rank approach to correct for […]<\/p>\n","protected":false},"author":2,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[62,113,112],"tags":[318,588,589],"class_list":["post-520","post","type-post","status-publish","format-standard","hentry","category-aimldsaimlds","category-cs-lg","category-stat-ml","tag-llms","tag-low","tag-rank"],"_links":{"self":[{"href":"https:\/\/mailitics.com\/index.php\/wp-json\/wp\/v2\/posts\/520"}],"collection":[{"href":"https:\/\/mailitics.com\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/mailitics.com\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/mailitics.com\/index.php\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/mailitics.com\/index.php\/wp-json\/wp\/v2\/comments?post=520"}],"version-history":[{"count":0,"href":"https:\/\/mailitics.com\/index.php\/wp-json\/wp\/v2\/posts\/520\/revisions"}],"wp:attachment":[{"href":"https:\/\/mailitics.com\/index.php\/wp-json\/wp\/v2\/media?parent=520"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/mailitics.com\/index.php\/wp-json\/wp\/v2\/categories?post=520"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/mailitics.com\/index.php\/wp-json\/wp\/v2\/tags?post=520"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}