A wrapper to simply load GPT-J and use it for generation. Uses DeepSpeed ons stage 2 or 3 for inference, as it reduce GPU memory usage.
Geek Repo:Geek Repo
Github PK Tool:Github PK Tool