▲ 1 MegaTrain: Full Precision Training of 100B+ Parameter LLMs on a Single GPU (arxiv.org) by chrsw | Apr 8, 2026 | 0 comments on HN Visit Link