News for hackers. Stuff that matters.
Re:MegaTrain: Full Precision Training of 100B+ Parameter LLM... by vlovich123 on Wednesday April 08, 2026 @10:54AM (#47691098)
341 is two orders of magnitude faster than your 1 tok/s so it doesn’t seem like their stuff is all that obvious. I also have no baseline for training to know if 341tok/s is slow but it seems speedy for a 3090.
Re:MegaTrain: Full Precision Training of 100B+ Parameter LLM... by bastawhiz on Wednesday April 08, 2026 @10:56AM (#47691118)
OP said 1k, not 1
Re:MegaTrain: Full Precision Training of 100B+ Parameter LLM... by SubiculumCode on Wednesday April 08, 2026 @11:48AM (#47691841)
:) Coffee is good
Re:MegaTrain: Full Precision Training of 100B+ Parameter LLM... by thrawa8387336 on Wednesday April 08, 2026 @12:06PM (#47692119)
OOM is log10
Re:MegaTrain: Full Precision Training of 100B+ Parameter LLM... by rolandr on Wednesday April 08, 2026 @10:57AM (#47691144)
1k tok/s = 1000 tok/s...