-
Notifications
You must be signed in to change notification settings - Fork 2
/
run_sbert_similarity.py
61 lines (48 loc) · 2.28 KB
/
run_sbert_similarity.py
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
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
import os
import json
import torch
import argparse
import numpy as np
import pandas as pd
from tqdm import tqdm
from torch.nn import functional as F
from sentence_transformers import SentenceTransformer
import sys
sys.path.append('../')
from utils import fix_seeds
def define_arguments():
parser = argparse.ArgumentParser()
parser.add_argument("--pre_ft_text_path", type=str, required=True)
parser.add_argument("--target_text_path", type=str, required=True)
parser.add_argument("--output_path", type=str, required=True)
parser.add_argument("--seed", type=int, default=0)
return parser.parse_args()
if __name__ == '__main__':
args = define_arguments()
sbert_model = SentenceTransformer('sentence-transformers/all-MiniLM-L6-v2')
sbert_model.eval()
sbert_model.requires_grad_(False)
fix_seeds(args)
with torch.no_grad():
with open(args.pre_ft_text_path, 'r') as file:
pre_ft_generated_text = json.load(file)
pre_ft_generated_text = list(pre_ft_generated_text.values())
pre_ft_embeddings = sbert_model.encode(pre_ft_generated_text, convert_to_tensor=True)
with open(args.target_text_path, 'r') as file:
target_generated_text = json.load(file)
target_generated_text = list(target_generated_text.values())
target_embeddings = sbert_model.encode(target_generated_text, convert_to_tensor=True)
assert pre_ft_embeddings.shape == target_embeddings.shape
sbert_dists = 1.0 - F.cosine_similarity(pre_ft_embeddings, target_embeddings).clip(-1.0, 1.0)
results = []
for prompt_idx in tqdm(range(len(sbert_dists))):
results.append({
"sbert_dist": sbert_dists[prompt_idx].item(),
"sbert_dist_log10": np.log10(sbert_dists[prompt_idx].item() + np.finfo(np.float64).eps),
"prompt_idx": prompt_idx,
})
results_df = pd.DataFrame(results)
os.makedirs(args.output_path, exist_ok=True)
results_df.to_csv(os.path.join(args.output_path, f"sbert.csv"), index=False)
print(f"Log10 SBERT Mean={results_df['sbert_dist_log10'].mean()}, Log10 SBERT STD={results_df['sbert_dist_log10'].std()}")
print(f"SBERT Mean={results_df['sbert_dist'].mean()}, SBERT STD={results_df['sbert_dist'].std()}")