What's Changed

• Post-release setup for 0.5.3.dev1 by @rasbt in #1799
• Add Phi 3 128k model by @deveworld in #1800
• Add token counts to compute performance by @rasbt in #1801
• Fixed the issue that precision is always "32-true". by @jianpingw in #1802
• Add Nvidia Llama 3.1 70B Nemotron weights by @rasbt in #1803
• Choose evaluation example from test set by @rasbt in #1804
• Pretrain tok sec by @rasbt in #1805
• typo in convert_to_litgpt command by @wasifferoze in #1807
• Move distributed all_reduce import into a function by @IvanYashchuk in #1810
• Remove hardcoded 32-precision conversion by @rasbt in #1814

New Contributors

• @deveworld made their first contribution in #1800
• @jianpingw made their first contribution in #1802
• @wasifferoze made their first contribution in #1807
• @IvanYashchuk made their first contribution in #1810

Full Changelog: v0.5.2...v0.5.3