cartridges-long-context

Train reusable pre-computed KV cache representations of large text corpora for efficient retrieval, achieving 38.6x memory reduction and 26.4x throughput improvement.

Independent GitHub snapshot · Automated classification · Original documentation governs