torchrun --nproc-per-node 1 examples/llama/convert_hf_to_nanotron.py --checkpoint_path Meta-Llama-3.1-8B --save_path checkpoints/Nanotron-Llama-3.1-8B
# torchrun --nproc-per-node 1 examples/llama/convert_hf_to_nanotron.py --checkpoint_path Llama-3.2-3B --save_path checkpoints/Nanotron-Llama-3.2-3B