Views
No views yet
meta-llama/Llama-3.2-1B-Instruct1import torch
2from transformers import AutoModelForCausalLM, AutoTokenizer
3from peft import PeftModel
4
5base_model = "meta-llama/Llama-3.2-1B-Instruct"
6adapter_id = "cindermond/world-model-plancraft-llama3-2-1b-instruct-filtered"
7
8tokenizer = AutoTokenizer.from_pretrained(base_model, use_fast=True)
9model = AutoModelForCausalLM.from_pretrained(
10 base_model,
11 device_map="auto",
12 torch_dtype=torch.bfloat16 if torch.cuda.is_available() else torch.float32,
13)
14model = PeftModel.from_pretrained(model, adapter_id)
15model.eval()1@article{zhao2026nesys,
2 title = {Neuro-Symbolic Synergy for Interactive World Modeling},
3 author = {Zhao, Hongyu and Zhou, Siyu and Yang, Haolin and Qin, Zengyi and Zhou, Tianyi},
4 journal = {arXiv preprint arXiv:2602.10480},
5 year = {2026}
6}