panyuyang
Results
2
comments of
panyuyang
Running with CUDA 12.0 ，torch 2.0.1 , can solve this problem
use: AutoTokenizer.from_pretrained(model_path, use_fast=True)