Repository navigation
Expand file tree
/
Copy pathEmbeddingModels.py
More file actions
107 lines (93 loc) · 4.2 KB
/
Copy pathEmbeddingModels.py
File metadata and controls
107 lines (93 loc) · 4.2 KB
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
import logging
from abc import ABC, abstractmethod
from typing import Dict
import time
from openai import OpenAI, AzureOpenAI
from azure.core.credentials import AzureKeyCredential
from sentence_transformers import SentenceTransformer
from transformers import DPRContextEncoder, DPRContextEncoderTokenizer, DPRQuestionEncoder, DPRQuestionEncoderTokenizer
from dotenv import load_dotenv
import os
from tenacity import retry, stop_after_attempt, wait_random_exponential
class BaseEmbeddingModel(ABC):
@abstractmethod
def create_embedding(self, text):
'''
Embed the text.
Return: stats info
- stats (Dict):
- embedding (list)
- time (float)
'''
pass
class SBertEmbeddingModel(BaseEmbeddingModel):
def __init__(self, model_name="sentence-transformers/multi-qa-mpnet-base-cos-v1"):
self.model = SentenceTransformer(model_name)
def create_embedding(self, text) -> Dict:
logging.info(f"SBert is creating embedding for text: {text[:min(len(text), 10)]}...")
start = time.time()
embedding = self.model.encode(text).tolist()
end = time.time()
return {
"embedding": embedding,
"time": end - start,
}
class DPRContextEmbeddingModel(BaseEmbeddingModel):
def __init__(self, model_name="facebook/dpr-ctx_encoder-multiset-base"):
self.tokenizer = DPRContextEncoderTokenizer.from_pretrained(model_name)
self.model = DPRContextEncoder.from_pretrained(model_name)
def create_embedding(self, text) -> Dict:
logging.info(f"DPRContextEncoder is creating embedding for text: {text[:min(len(text), 10)]}...")
start = time.time()
input_ids = self.tokenizer(text, return_tensors="pt", max_length=512)["input_ids"]
embedding = self.model(input_ids).pooler_output.detach().numpy().flatten().tolist()
end = time.time()
return {
"embedding": embedding,
"time": end - start,
}
class DPRQueryEmbeddingModel(BaseEmbeddingModel):
def __init__(self, model_name="facebook/dpr-question_encoder-multiset-base"):
self.tokenizer = DPRQuestionEncoderTokenizer.from_pretrained(model_name)
self.model = DPRQuestionEncoder.from_pretrained(model_name)
def create_embedding(self, text) -> Dict:
logging.info(f"DPRQueryEncoder is creating embedding for text: {text[:min(len(text), 10)]}...")
start = time.time()
input_ids = self.tokenizer(text, return_tensors="pt", max_length=512)["input_ids"]
embedding = self.model(input_ids).pooler_output.detach().numpy().flatten().tolist()
end = time.time()
return {
"embedding": embedding,
"time": end - start,
}
class TextEmbedding3SmallModel(BaseEmbeddingModel):
def __init__(self, openai_key_path: str, model_name="text-embedding-3-small"):
###############OPENAI API#################
# self.model_name = model_name
# self.client = None
# load_dotenv(openai_key_path)
# self.client = OpenAI(api_key=os.getenv("API_KEY"))
##########################################
###############AZURE OPENAI API#################
self.model_name = model_name
self.client = None
openai_key_path = os.path.join(os.path.dirname(os.path.abspath(__file__)), ".env")
assert os.path.exists(openai_key_path)
load_dotenv(openai_key_path)
self.client = AzureOpenAI(
azure_endpoint=os.getenv("AZURE_ENDPOINT"),
api_version=os.getenv("AZURE_API_VERSION"),
api_key=os.getenv("AZURE_API_KEY"),
)
############################################
@retry(wait=wait_random_exponential(min=1, max=20), stop=stop_after_attempt(6))
def create_embedding(self, text) -> Dict:
assert self.client is not None
logging.info(f"text-embedding-3-small is creating embedding for text: {text[:min(len(text), 10)]}...")
start = time.time()
embedding = self.client.embeddings.create(input=[text], model=self.model_name).data[0].embedding
end = time.time()
return {
"embedding": embedding,
"time": end - start,
}