newsHacker NewsTrust 72 · CommunityPublished 6d agoLive · 6d ago
DeepSeek open-sources inference optimizations with 60–85% faster generation [pdf]
450points146comments
Covers (incoming)
paperEvolution Fine-Tuning: Learning to Discover Across 371 Optimization TaskspaperB3O: Scalable Boltzmann Batch Bayesian OptimizationpaperCondensing Large-Scale Datasets Directly with Minimal Information Lossrepomlcommons/inferencerepoJuliaDecisionFocusedLearning/InferOpt.jlrepoKaden-Schutt/hipfire
Related across the graph
repomlcommons/inferencepaperEvolution Fine-Tuning: Learning to Discover Across 371 Optimization TaskspaperB3O: Scalable Boltzmann Batch Bayesian OptimizationrepoJuliaDecisionFocusedLearning/InferOpt.jlpaperCondensing Large-Scale Datasets Directly with Minimal Information LossrepoKaden-Schutt/hipfire