[–] fgfm 2y ago ↗ The Anyscale team shared how you can achieve considerable speedups for model loading in production with examples on the Llama 2 variants.
1 comment
[ 3.7 ms ] story [ 12.7 ms ] thread