TrnRT is an open-source inference engine for running large language models on AWS Trainium chips, giving developers a self-hosted alternative to managed serving on the custom silicon.
Read it at Show HN →
← Front page