trn2.48xlarge

Latest news

1. 21 Oct
   i14 trains and serves Laplace 3B linear-attention model on AWS Trainium2

   i14 worked with AWS to train and serve its Laplace 3B linear-attention model using native PyTorch support via TorchNeuron Native on Trainium2 hardware. The model achieved about 120,000 training tokens/sec on a trn2.48xlarge instance and served 11,544 output tokens/sec across 1,280 concurrent sequences while keeping recurrent state size fixed at 29.5 MB per sequence.

   Models · 1 source

   

Articles

No articles about trn2.48xlarge yet.