SpQR: A Sparse-Quantized Representation for Near-Lossless LLM Weight Compression · HackerPapers discussions

SpQR: A Sparse-Quantized Representation for Near-Lossless LLM Weight Compression — mentioned by AhtiK

Fri, 08 Mar 2024 08:18:45 GMT

Has anyone seen an implementation of 'SpQR: A Sparse-Quantized Representation,' published in June 2023 by Tim Dettmers et al.? https://arxiv.org/abs/2306.03078

Paper and discussion history. A mention is not an endorsement.

SpQR: A Sparse-Quantized Representation for Near-Lossless LLM Weight Compression — mentioned by raphlinus

Tue, 26 Sep 2023 19:38:21 GMT

ROCm makes me sad, as it reminds me of how much better GPUs could be than they are today. I've lately been exploring the idea of a "Good Parallel Computer," which combines most of the agility of a CPU with the efficient parallel throughput of a GPU. The central concept is that the decision to launch a workgroup is made by a programmable controller, rather than just being a cube of (x, y, z) or downstream of triangles. A particular workload it would likely excel at is sparse matrix multiplication, including multiple quantization levels like SpQR[1]. I'm hopeful that it could be an advance in…

Paper and discussion history. A mention is not an endorsement.

SpQR: A Sparse-Quantized Representation for Near-Lossless LLM Weight Compression — mentioned by GaggiX

Sat, 10 Jun 2023 19:58:56 GMT

https://arxiv.org/abs/2306.03078 From the same author and more recent.

Paper and discussion history. A mention is not an endorsement.

SpQR: A Sparse-Quantized Representation for Near-Lossless LLM Weight Compression — HN submission

Tue, 06 Jun 2023 17:23:13 GMT

SpQR: A Sparse-Quantized Representation for Near-Lossless LLM Weight Compression

8 points; 2 discussion replies on HN. Paper and source history. A submission is not an endorsement.