DossierENTITY

MegaTrain

Coverage of MegaTrain in the Nexus archive.

Earliest in view: Apr 8 · 12:19 UTCMost recent: Apr 8 · 12:19 UTC

Co-mentioned in this coverage

Recent coverage

TECHNOLOGYApr 8 · 12:19 UTCHACKER NEWS
MegaTrain: Full Precision Training of 100B+ Parameter LLMs on a Single GPU
The article discusses a method called MegaTrain which enables full precision training of large language models with over 100 billion parameters on a single GPU.