发表机构
Singapore Management University; CSIRO; AIDX TECH PTE LTD; University of Alberta; Alberta Machine Intelligence Institute（新加坡管理大学; 联邦科学与工业研究组织; AIDX TECH PTE LTD; 阿尔伯塔大学; 阿尔伯塔机器智能研究所）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
Brevis将无损张量压缩转化为程序综合，通过带类型的DSL与有界A*搜索优化压缩，在10类模型检查点上实现了超通用及专用压缩器的存储缩减，且压缩解压速度可观。
AI 中文摘要
模型检查点的数量和规模不断增长，使得归档、迁移和部署的成本日益高昂。通用压缩器可降低存储需求，但会忽略张量结构；而现有的专用张量压缩器则依赖固定且格式特定的流程。我们提出Brevis，将无损张量压缩形式化为程序综合。我们设计了一种带类型的领域特定语言（DSL），通过一组可逆算子捕获重复区域、浮点字段等常见张量结构。给定一个张量，Brevis会合成一个独立的DSL程序，该程序可精确重建张量的每一位。从少量代表性张量样本中学习到的检查点特定生产先验，指导有界A*搜索以合成紧凑程序，这些程序后续可直接执行实现精确的位级解压。在涵盖语言、音频和图像生成模型的10个公开检查点上，Brevis将2.13 TB的检查点数据压缩至1.41 TB，实现了33.93%的存储缩减。其生成的归档文件比zstd、gzip等4种通用压缩器的归档文件小最多30.87%，也比专用张量压缩器ZipNN和DFloat11的归档文件更小。在实际并发配置下，Brevis实现了3.60 GB/s的压缩速度和6.61 GB/s的解压速度，同时保留了所有原始字节。
英文摘要
Model checkpoints are growing in both number and size, which makes archival, transfer, and deployment increasingly costly. General-purpose compressors can reduce storage requirements but ignore tensor structure, whereas existing tensor-specific compressors rely on fixed and format-specific pipelines. We present Brevis, which formulates lossless tensor compression as program synthesis. We design a typed domain-specific language (DSL) that captures recurring tensor structures, such as repeated regions and floating-point fields, through a set of reversible operators. Given a tensor, Brevis synthesizes a self-contained DSL program that reconstructs it bit-exactly. A checkpoint-specific production prior, learned from a small representative sample of tensors, guides a bounded A* search to synthesize compact programs, which can later be executed directly for bit-exact decompression. On 10 public checkpoints spanning language, audio, and image generation models, Brevis reduces 2.13 TB of checkpoint data to 1.41 TB, a 33.93% storage reduction. It produces archives up to 30.87% smaller than those of four general-purpose compressors, including zstd and gzip, and smaller archives than the tensor-specific compressors ZipNN and DFloat11. Under a practical concurrency configuration, Brevis achieves 3.60 GB/s compression and 6.61 GB/s decompression while preserving every source byte.
