DeepSpeed

Deep learning optimization library for distributed training and inference

Growing steadily this week

7d +0.05% +22 stars in 7d
30d +0.26% +113 stars in 30d
90d Not tracked yet — first tracked 2026-08-14
Aug 14, 2026: 42,931 starsAug 15, 2026: 42,933 starsAug 16, 2026: 42,941 starsAug 17, 2026: 42,945 starsAug 18, 2026: 42,953 starsAug 19, 2026: 42,957 starsAug 20, 2026: 42,964 starsAug 21, 2026: 42,971 starsAug 22, 2026: 42,975 starsAug 23, 2026: 42,978 starsAug 24, 2026: 42,982 starsAug 25, 2026: 42,992 starsAug 26, 2026: 43,000 starsAug 27, 2026: 43,009 starsAug 28, 2026: 43,018 starsAug 29, 2026: 43,020 starsAug 30, 2026: 43,026 starsAug 31, 2026: 43,037 starsSep 1, 2026: 43,044 starsSep 2, 2026: 43,057 starsSep 3, 2026: 43,062 starsSep 5, 2026: 43,065 starsSep 6, 2026: 43,065 starsSep 7, 2026: 43,071 starsSep 8, 2026: 43,074 starsSep 9, 2026: 43,085 starsSep 10, 2026: 43,088 starsSep 11, 2026: 43,095 starsSep 12, 2026: 43,097 starsSep 13, 2026: 43,101 starsSep 14, 2026: 43,105 starsSep 15, 2026: 43,115 starsSep 16, 2026: 43,123 starsSep 17, 2026: 43,127 starsSep 18, 2026: 43,132 starsSep 19, 2026: 43,138 starsSep 20, 2026: 43,145 starsSep 21, 2026: 43,145 starsSep 22, 2026: 43,152 starsSep 23, 2026: 43,156 starsSep 24, 2026: 43,158 starsSep 25, 2026: 43,159 starsSep 26, 2026: 43,161 starsSep 27, 2026: 43,160 starsSep 28, 2026: 43,161 starsSep 29, 2026: 43,165 starsSep 30, 2026: 43,167 starsOct 1, 2026: 43,171 starsOct 2, 2026: 43,173 starsOct 3, 2026: 43,172 starsOct 4, 2026: 43,178 starsOct 5, 2026: 43,197 starsOct 6, 2026: 43,201 starsOct 7, 2026: 43,206 starsOct 8, 2026: 43,206 starsOct 9, 2026: 43,214 starsOct 10, 2026: 43,218 starsOct 11, 2026: 43,218 starsOct 12, 2026: 43,221 starsOct 13, 2026: 43,229 starsOct 14, 2026: 43,228 starsOct 15, 2026: 43,231 starsOct 16, 2026: 43,236 stars 42,93143,08443,236 Aug 14Oct 16

42,931 → 43,236 stars since Aug 14

42,999 Stars 5,040 Forks 1,526 Open issues

Timeline

💬 1 Hacker News discussion

billion-parameterscompressiondata-parallelismdeep-learninggpuinferencemachine-learningmixture-of-expertsmodel-parallelismpipeline-parallelismpytorchtrillion-parameterszero