zshsh98/mmd-mp

This is the source code for Detecting Machine-Generated Texts by Multi-Population Aware Optimization for Maximum Mean Discrepancy (ICLR2024).

mmd-mp's Issues

The default of the "text_generated_model_name"

The default of the original parser's of the argument "text_generated_model_name" is "gpt2", and it would have error when we try to traversal it, like key 'g' is not exist. So I assume it should be as an element, like alter to : default = ['gpt2'] ?

OSError: No such device (os error 19) When Loading Model with Transformers

Token indices sequence length is longer than the specified maximum sequence length for this model (749 > 512). Running this sequence through the model will result in indexing errors
Traceback (most recent call last):
File "run_meta_mmd_trans.py", line 925, in
base_model, base_tokenizer = load_base_model_and_tokenizer(args.base_model_name)
File "run_meta_mmd_trans.py", line 531, in load_base_model_and_tokenizer
base_model = transformers.AutoModelForCausalLM.from_pretrained(model_path_dit[name])
File "/root/miniconda3/envs/detectGPT/lib/python3.7/site-packages/transformers/models/auto/auto_factory.py", line 485, in from_pretrained
pretrained_model_name_or_path, *model_args, config=config, **hub_kwargs, **kwargs
File "/root/miniconda3/envs/detectGPT/lib/python3.7/site-packages/transformers/modeling_utils.py", line 2604, in from_pretrained
state_dict = load_state_dict(resolved_archive_file)
File "/root/miniconda3/envs/detectGPT/lib/python3.7/site-packages/transformers/modeling_utils.py", line 450, in load_state_dict
with safe_open(checkpoint_file, framework="pt") as f:
OSError: No such device (os error 19)