arXiv cs.AI· Muhammed Fatih Bal{\i}n·· 4 小时前AI 评分38
CPTC:用两次乘法和一张 64 字节表实现近最优量化
Computed Permutation Trellis Coding: Near-Optimal Quantization from Two Multiplications and a 64-Byte Table
AI 导读
研究者提出低比特量化格式 CPTC(Computed Permutation Trellis Coding),解码器只需两次乘法加一张 64 字节表,在 4 bit 下距信息论理想仅 0.07 bit,并精确复现标准整数与浮点格式。
正文 · 原文
Abstract:We propose Computed Permutation Trellis Coding (CPTC), a low-bit quantization format for tensors written once and read many times. Formats that round each value on its own, such as INT4 and FP4, waste a large part of their bits against the information-theoretic ideal. Trellis codes recover this loss by encoding values jointly, but prior decoders either need a large lookup table for their best quality or compute values with a hash that restricts what they can decode. CPTC's decoder multiplies its state by two coefficients and adds the top bits of the products to index a 64-byte table. Because that sum is not uniformly distributed, the rarely used ends of a small table can hold values far out in the tails. One decoder serves every bit width from 1 to 6, reproduces the standard integer and floating-point formats exactly, and comes within 0.07 bits of the ideal at 4 bits. It matches a random-permutation trellis code whose table is thousands of times larger, and exceeds every QTIP codebook. In LLM weight quantization without fine-tuning, CPTC has the lowest KL divergence to the unquantized model at every tested bit width on Llama 3.1 8B and 70B Instruct against QTIP's codebooks (by up to 27%). The decoder has been built into the AWS Trainium 4 AI accelerator, where it runs at the throughput of the native FP4 and FP8 formats for a negligible fraction of the systolic array's area. Fused encoding kernels quantize an 8B model's linear layers in under two minutes on one GPU and, in cycle-level simulation, on one Trainium 4 chip.
| Comments: | 38 pages, 6 figures, 27 tables |
| Subjects: | Information Theory (cs.IT); Artificial Intelligence (cs.AI); Hardware Architecture (cs.AR); Machine Learning (cs.LG) |
| ACM classes: | E.4; I.2.6 |
| Cite as: | arXiv:2611.22019 [cs.IT] |
| (or arXiv:2611.22019v1 [cs.IT] for this version) | |
| https://doi.org/10.48550/arXiv.2611.22019 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Muhammed Fatih Balın [view email]
[v1]
Tue, 17 Nov 2026 00:17:01 UTC (419 KB)
来源:arXiv cs.AI · arxiv.org