-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathrun.sh
36 lines (27 loc) · 3.37 KB
/
run.sh
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
# set the dataset pair that will be used
data='xsum'
data2='squad'
# set the machine
# model='meta-llama/Llama-2-7b-hf'
# model='lmsys/vicuna-v1.5'
# model='/n/holylabs/LABS/barak_lab/Lab/models/llama_models/7B'
# model='/n/holylabs/LABS/barak_lab/Lab/models/llama_models/alpaca-7B'
model='/n/holylabs/LABS/barak_lab/Lab/models/llama_models/vicuna-7b-v1.5'
# echo "runing test"
echo $data
echo $data2
echo $model
# evaluating all the zero-shot detectors with the same Human-authored sources and machine generated texts
python run_ood.py --output_name main --base_model_name $model --mask_filling_model_name t5-3b --baselines_only --n_perturbation_list 1,10 --n_samples 500 --pct_words_masked 0.3 --span_length 2 --dataset $data --dataset2 $data > vicuna-7b-baselines-t5-3b-$data-$data.out
python run_ood.py --output_name main --base_model_name $model --mask_filling_model_name t5-3b --baselines_only --n_perturbation_list 1,10 --n_samples 500 --pct_words_masked 0.3 --span_length 2 --dataset $data2 --dataset2 $data2 > vicuna-7b-baselines-t5-3b-$data2-$data2.out
# evaluating all the zero-shot detectors with different Human-authored sources and machine generated texts
echo "transfer from data 1 to all"
python run_ood.py --output_name main --base_model_name $model --mask_filling_model_name t5-3b --baselines_only --n_perturbation_list 1,10 --n_samples 500 --pct_words_masked 0.3 --span_length 2 --dataset $data --dataset2 'squad' > vicuna-7b-baselines-t5-3b-$data-squad.out
python run_ood.py --output_name main --base_model_name $model --mask_filling_model_name t5-3b --baselines_only --n_perturbation_list 1,10 --n_samples 500 --pct_words_masked 0.3 --span_length 2 --dataset $data --dataset2 'code' > vicuna-7b-baselines-t5-3b-$data-code.out
python run_ood.py --output_name main --base_model_name $model --mask_filling_model_name t5-3b --baselines_only --n_perturbation_list 1,10 --n_samples 500 --pct_words_masked 0.3 --span_length 2 --dataset $data --dataset2 'xsum' > vicuna-7b-baselines-t5-3b-$data-xsum.out
python run_ood.py --output_name main --base_model_name $model --mask_filling_model_name t5-3b --baselines_only --n_perturbation_list 1,10 --n_samples 500 --pct_words_masked 0.3 --span_length 2 --dataset $data --dataset2 'writing' > vicuna-7b-baselines-t5-3b-$data-writing.out
echo "transfer from other datasets to data 1"
python run_ood.py --output_name main --base_model_name $model --mask_filling_model_name t5-3b --baselines_only --n_perturbation_list 1,10 --n_samples 500 --pct_words_masked 0.3 --span_length 2 --dataset $data --dataset2 'squad' > vicuna-7b-baselines-t5-3b-$data-squad.out
python run_ood.py --output_name main --base_model_name $model --mask_filling_model_name t5-3b --baselines_only --n_perturbation_list 1,10 --n_samples 500 --pct_words_masked 0.3 --span_length 2 --dataset $data --dataset2 'code' > vicuna-7b-baselines-t5-3b-$data-code.out
python run_ood.py --output_name main --base_model_name $model --mask_filling_model_name t5-3b --baselines_only --n_perturbation_list 1,10 --n_samples 500 --pct_words_masked 0.3 --span_length 2 --dataset $data --dataset2 'xsum' > vicuna-7b-baselines-t5-3b-$data-xsum.out
python run_ood.py --output_name main --base_model_name $model --mask_filling_model_name t5-3b --baselines_only --n_perturbation_list 1,10 --n_samples 500 --pct_words_masked 0.3 --span_length 2 --dataset $data --dataset2 'writing' > vicuna-7b-baselines-t5-3b-$data-writing.out