Search Results for "parallel compression"

MSA, or Memory Sparse Attention, is a research framework for scaling language-model memory to extremely long contexts. It replaces full attention over all tokens with sparse selection of compressed latent memory states. Document-wise rotary position encoding and top-k routing keep training and inference close to linear complexity. A tiered KV-cache design stores routing keys on GPU while larger content states can remain on CPU. Its Memory Parallel engine distributes scoring and transfers...

Barman (Backup and Recovery Manager) is an enterprise-grade tool for managing backups and disaster recovery of PostgreSQL databases. It supports both full and incremental backups, Point-In-Time Recovery (PITR), and remote backup via SSH. Barman is widely used by DBAs to ensure secure, reliable, and consistent backups in production environments.

...The project also emphasizes optimization techniques for large-scale training, including mixed-precision and hybrid-parallel strategies that are commonly needed for multi-node GPU clusters. In addition to training, it includes guidance and example materials intended to help developers adopt ERNIE models for real product scenarios rather than only research demonstrations.

Crunch is an image compression tool for lossy PNG image file optimization. Using a combination of selective bit depth, color palette reduction and color type, as well as zopfli DEFLATE compression algorithm encoding that employs the pngquant and zopflipng PNG optimization tools, Crunch is effectively able to optimize and compress images with minimal decrease in image quality. While it may produce file size gains larger than those produced by lossless approaches, the impact on image quality...

LZHW Command Line Lossless Compression Tool is a Windows command line tool used to compress and decompress files from and to any form, csv, excel etc without any dependencies or installations. Using an optimized algorithm (LZHW) developed from Lempel-Ziv, Huffman and LZ-Welch algorithms. The tool can work in parallel and most of its code is written in Cython, so it is pretty fast.