Spaces:
Running
Running
acumplid
commited on
Commit
·
9803bf8
1
Parent(s):
88e6c95
Implemented new rerank
Browse files
rag.py
CHANGED
@@ -1,7 +1,8 @@
|
|
1 |
import logging
|
2 |
import os
|
3 |
import requests
|
4 |
-
|
|
|
5 |
|
6 |
|
7 |
from langchain_community.vectorstores import FAISS
|
@@ -15,11 +16,13 @@ class RAG:
|
|
15 |
#vectorstore = "vectorestore" # CA only
|
16 |
vectorstore = "index-BAAI_bge-m3-1500-200-recursive_splitter-CA_ES_UE"
|
17 |
|
18 |
-
def __init__(self, hf_token, embeddings_model, model_name):
|
19 |
|
20 |
|
21 |
self.model_name = model_name
|
22 |
self.hf_token = hf_token
|
|
|
|
|
23 |
|
24 |
# load vectore store
|
25 |
embeddings = HuggingFaceEmbeddings(model_name=embeddings_model, model_kwargs={'device': 'cpu'})
|
@@ -27,10 +30,50 @@ class RAG:
|
|
27 |
|
28 |
logging.info("RAG loaded!")
|
29 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
30 |
def get_context(self, instruction, number_of_contexts=2):
|
|
|
|
|
31 |
|
32 |
-
documentos = self.
|
33 |
|
|
|
34 |
return documentos
|
35 |
|
36 |
def predict(self, instruction, context, model_parameters):
|
|
|
1 |
import logging
|
2 |
import os
|
3 |
import requests
|
4 |
+
from transformers import AutoTokenizer, AutoModelForSequenceClassification
|
5 |
+
import torch
|
6 |
|
7 |
|
8 |
from langchain_community.vectorstores import FAISS
|
|
|
16 |
#vectorstore = "vectorestore" # CA only
|
17 |
vectorstore = "index-BAAI_bge-m3-1500-200-recursive_splitter-CA_ES_UE"
|
18 |
|
19 |
+
def __init__(self, hf_token, embeddings_model, model_name, rerank_model, rerank_number_contexts):
|
20 |
|
21 |
|
22 |
self.model_name = model_name
|
23 |
self.hf_token = hf_token
|
24 |
+
self.rerank_model = rerank_model
|
25 |
+
self.rerank_number_contexts = rerank_number_contexts
|
26 |
|
27 |
# load vectore store
|
28 |
embeddings = HuggingFaceEmbeddings(model_name=embeddings_model, model_kwargs={'device': 'cpu'})
|
|
|
30 |
|
31 |
logging.info("RAG loaded!")
|
32 |
|
33 |
+
def rerank_contexts(self, instruction, contexts, number_of_contexts=1):
|
34 |
+
"""
|
35 |
+
Rerank the contexts based on their relevance to the given instruction.
|
36 |
+
"""
|
37 |
+
|
38 |
+
rerank_model = self.rerank_model
|
39 |
+
|
40 |
+
|
41 |
+
tokenizer = AutoTokenizer.from_pretrained(rerank_model)
|
42 |
+
model = AutoModelForSequenceClassification.from_pretrained(rerank_model)
|
43 |
+
|
44 |
+
def get_score(query, passage):
|
45 |
+
"""Calculate the relevance score of a passage with respect to a query."""
|
46 |
+
|
47 |
+
|
48 |
+
inputs = tokenizer(query, passage, return_tensors='pt', truncation=True, padding=True, max_length=512)
|
49 |
+
|
50 |
+
|
51 |
+
with torch.no_grad():
|
52 |
+
outputs = model(**inputs)
|
53 |
+
|
54 |
+
|
55 |
+
logits = outputs.logits
|
56 |
+
|
57 |
+
|
58 |
+
score = logits.view(-1, ).float()
|
59 |
+
|
60 |
+
|
61 |
+
return score
|
62 |
+
|
63 |
+
scores = [get_score(instruction, c[0].page_content) for c in contexts]
|
64 |
+
combined = list(zip(contexts, scores))
|
65 |
+
sorted_combined = sorted(combined, key=lambda x: x[1], reverse=True)
|
66 |
+
sorted_texts, _ = zip(*sorted_combined)
|
67 |
+
|
68 |
+
return sorted_texts[:number_of_contexts]
|
69 |
+
|
70 |
def get_context(self, instruction, number_of_contexts=2):
|
71 |
+
"""Retrieve the most relevant contexts for a given instruction."""
|
72 |
+
documentos = self.vectore_store.similarity_search_with_score(instruction, k=self.rerank_number_contexts)
|
73 |
|
74 |
+
documentos = self.rerank_contexts(instruction, documentos, number_of_contexts=number_of_contexts)
|
75 |
|
76 |
+
print("Reranked documents")
|
77 |
return documentos
|
78 |
|
79 |
def predict(self, instruction, context, model_parameters):
|