-
Notifications
You must be signed in to change notification settings - Fork 1
/
test_umap_variants10.py
150 lines (125 loc) · 6.13 KB
/
test_umap_variants10.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
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
import json
from umap.parametric_umap import load_ParametricUMAP
import os
import sqlite3
import numpy as np
from umap.parametric_umap import ParametricUMAP
from tqdm import tqdm
import gzip
import random
def gzip_file(input_file, output_file):
with open(input_file, 'rb') as f_in:
with gzip.open(output_file, 'wb') as f_out:
f_out.writelines(f_in)
def retrieve_embedding(db_name, key):
conn = sqlite3.connect(db_name)
c = conn.cursor()
c.execute("SELECT embedding FROM cache WHERE key=?", (key,))
result = c.fetchone()
conn.close()
if result:
return json.loads(result[0])
else:
return None
def process_english_data():
# Load existing JSON data
with open('static/english/wildchat_embeddings_all.json', 'r') as f:
wildchat_data = json.load(f)
with open('static/english/lmsyschat_embeddings_all.json', 'r') as f:
lmsyschat_data = json.load(f)
print(f"Loaded {len(wildchat_data)} WildChat and {len(lmsyschat_data)} LMSYS-Chat samples")
# Retrieve embeddings
wildchat_embed_db = 'wildchat_embeddings_cache.db'
lmsyschat_embed_db = 'lmsyschat_embeddings_cache.db'
embeddings = []
valid_samples = {'wildchat': [], 'lmsyschat': []}
for item in tqdm(wildchat_data, desc="Retrieving WildChat embeddings"):
embedding = retrieve_embedding(wildchat_embed_db, item['i'])
if embedding:
embeddings.append(embedding)
valid_samples['wildchat'].append(item)
for item in tqdm(lmsyschat_data, desc="Retrieving LMSYS-Chat embeddings"):
embedding = retrieve_embedding(lmsyschat_embed_db, item['i'])
if embedding:
embeddings.append(embedding)
valid_samples['lmsyschat'].append(item)
print(f"Valid samples: WildChat: {len(valid_samples['wildchat'])}, LMSYS-Chat: {len(valid_samples['lmsyschat'])}")
# Run UMAP with n_neighbors=15
#umap = ParametricUMAP(n_components=2, n_neighbors=30, min_dist=0.3, metric='cosine')
#umap = ParametricUMAP(n_components=2, n_neighbors=50, metric='cosine')
#umap = ParametricUMAP(n_components=2, n_neighbors=30, min_dist=0.6, metric='cosine')
#umap = ParametricUMAP(n_components=2, n_neighbors=30, metric='cosine')
#umap = ParametricUMAP(n_components=2, n_neighbors=30, min_dist=0.9, metric='cosine')
#umap = ParametricUMAP(n_components=2, n_neighbors=10, metric='cosine')
#umap = ParametricUMAP(n_components=2, n_neighbors=30, repulsion_strength=0.5, metric='cosine')
#umap = ParametricUMAP(n_components=2, n_neighbors=30, repulsion_strength=0.2, metric='cosine')
#umap = ParametricUMAP(n_components=2, n_neighbors=30, repulsion_strength=0.1, metric='cosine')
#umap = ParametricUMAP(n_components=2, n_neighbors=50, repulsion_strength=0.2, metric='cosine')
#umap = ParametricUMAP(n_components=2, n_neighbors=30, spread=0.5, metric='cosine')
umap = ParametricUMAP(n_components=2, n_neighbors=30, spread=0.2, metric='cosine')
#umap.n_training_epochs = 0.5
umap_embeddings = umap.fit_transform(embeddings)
# Save UMAP model
os.makedirs('umap_model/debugs02', exist_ok=True)
if hasattr(umap, "_raw_data"):
del umap._raw_data
if hasattr(umap, "knn_search_index") and hasattr(umap.knn_search_index, "_raw_data"):
del umap.knn_search_index._raw_data
umap.save('umap_model/debugs02')
# Create new UMAP cache databases
#for db_path in ['umap_debugmin0_wildchat_cache.db', 'umap_debugmin0_lmsyschat_cache.db']:
# conn = sqlite3.connect(db_path)
# c = conn.cursor()
# c.execute('''CREATE TABLE IF NOT EXISTS cache
# (key TEXT PRIMARY KEY, prompt TEXT, embedding TEXT)''')
# conn.commit()
# conn.close()
# Save new UMAP embeddings and create JSON files
os.makedirs('static/debugs02', exist_ok=True)
# Create sets of subsampled IDs
with open('static/english/wildchat_embeddings.json', 'r') as f:
wildchat_data_subsampled = json.load(f)
with open('static/english/lmsyschat_embeddings.json', 'r') as f:
lmsyschat_data_subsampled = json.load(f)
wildchat_subsampled_ids = set(item['i'] for item in wildchat_data_subsampled)
lmsyschat_subsampled_ids = set(item['i'] for item in lmsyschat_data_subsampled)
for dataset_name in ['wildchat', 'lmsyschat']:
if dataset_name == 'wildchat':
subsampled_ids = wildchat_subsampled_ids
else:
subsampled_ids = lmsyschat_subsampled_ids
json_data = []
json_data_subsampled = []
start_index = 0 if dataset_name == 'wildchat' else len(valid_samples['wildchat'])
for i, item in enumerate(valid_samples[dataset_name], start=start_index):
umap_embedding = umap_embeddings[i].tolist()
# Save to UMAP cache
#db_path = f'umap_debugmin0_{dataset_name}_cache.db'
#conn = sqlite3.connect(db_path)
#c = conn.cursor()
#c.execute('''INSERT OR REPLACE INTO cache
# (key, prompt, embedding) VALUES (?, ?, ?)''',
# (item['i'], '', json.dumps(umap_embedding)))
#conn.commit()
#conn.close()
new_item = {
'i': item['i'],
'e': [round(float(umap_embedding[0]), 4), round(float(umap_embedding[1]), 4)],
'c': item['c'],
'd': dataset_name
}
json_data.append(new_item)
if item['i'] in subsampled_ids:
json_data_subsampled.append(new_item)
# Save full JSON
with open(f'static/debugs02/{dataset_name}_embeddings_all.json', 'w') as f:
json.dump(json_data, f)
# Save and gzip subsampled JSON
#subsampled_json_data = random.sample(json_data, min(1500, len(json_data)))
subsampled_json_path = f'static/debugs02/{dataset_name}_embeddings.json'
with open(subsampled_json_path, 'w') as f:
json.dump(json_data_subsampled, f)
gzip_file(subsampled_json_path, f'{subsampled_json_path}.gz')
print("Processing complete!")
if __name__ == "__main__":
process_english_data()