tien314 commited on
Commit
c2e612d
·
verified ·
1 Parent(s): 16558da

Upload main.py

Browse files
Files changed (1) hide show
  1. main.py +108 -0
main.py ADDED
@@ -0,0 +1,108 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ import streamlit as st
2
+ from langchain_community.retrievers import BM25Retriever
3
+ import pandas as pd
4
+ from langchain.docstore.document import Document
5
+ from langchain.text_splitter import CharacterTextSplitter
6
+ from operator import itemgetter
7
+ from langchain_core.prompts import PromptTemplate
8
+ from langchain_groq import ChatGroq
9
+ from langchain.chains.question_answering import load_qa_chain
10
+ import os
11
+
12
+
13
+ @st.cache_data
14
+ def load_data():
15
+ df = pd.read_csv("C:\\Users\\JVC Store\\Downloads\\data\\data 6\\train.csv")
16
+ df = df.drop(columns = ['Unnamed: 0','hs_code_2','hs_code_4'])
17
+ documents = []
18
+
19
+ for index, row in df.iterrows():
20
+ text = row['full_description']
21
+ hs_code = row['hs_code']
22
+ documents.append(Document(page_content=text, metadata={'hs_code': hs_code}))
23
+
24
+ splitter = CharacterTextSplitter(
25
+ chunk_size=100,
26
+ chunk_overlap=0,
27
+ separator = ' '
28
+ )
29
+
30
+ split_documents = []
31
+ for doc in documents:
32
+ chunks = splitter.split_text(doc.page_content)
33
+ #remove chunk split word
34
+ word_chunks = []
35
+ current_chunk = []
36
+
37
+ for chunk in chunks:
38
+ words = chunk.split()
39
+ for word in words:
40
+ if len(' '.join(current_chunk + [word])) <=100:
41
+ current_chunk.append(word)
42
+ else:
43
+ word_chunks.append(' '.join(current_chunk))
44
+ current_chunk = [word]
45
+ if current_chunk:
46
+ word_chunks.append(' '.join(current_chunk))
47
+
48
+ split_documents.append(Document(page_content=word_chunks[0], metadata=doc.metadata))
49
+
50
+
51
+ docs = []
52
+ for doc in split_documents:
53
+ metadata = doc.metadata
54
+ metadata_str = str(metadata).strip('{}')
55
+ page = doc.page_content
56
+ docs.append([metadata_str + " " + page])
57
+
58
+
59
+ cleaned_list = [item.replace('"','').replace("'",'') for items in docs for item in items]
60
+ retriever = BM25Retriever.from_texts(cleaned_list)
61
+ retriever.k = 5
62
+ return retriever
63
+
64
+
65
+ def load_llm():
66
+
67
+ api_key2 = "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"
68
+
69
+ llm2 = ChatGroq(model = "llama-3.1-70b-versatile", temperature = 0,api_key = api_key2)
70
+ return llm2
71
+
72
+
73
+ def predict(sentence,retriever,llm2):
74
+ sentence = sentence.lower()
75
+ context = retriever.get_relevant_documents(sentence)
76
+ #print("context:",context)
77
+ template2 = """
78
+ You are an expert in HS Code classification.
79
+ Based on the provided product description, accurately determine and return only one 6-digit HS Code that best matches the description.
80
+ Always return the HS Code as a 6-digit number only.
81
+ example: 123456
82
+ Context:\n {context} \n
83
+ Description:\n {description} \n
84
+ Answer:
85
+ """
86
+ prompt2 = PromptTemplate(template=template2, input_variables=['context','description'])
87
+ chain = load_qa_chain(llm2, chain_type = 'stuff', prompt = prompt2)
88
+ response = chain.invoke({'input_documents': context, 'description':sentence})
89
+ answer = response.get("output_text")
90
+ return answer
91
+
92
+
93
+ if 'retriever' not in st.session_state:
94
+ st.session_state.retriever = None
95
+ if 'llm' not in st.session_state:
96
+ st.session_state.llm = None
97
+
98
+ if st.session_state.retriever is None:
99
+ st.session_state.retriever = load_data()
100
+
101
+ if st.session_state.llm is None:
102
+ st.session_state.llm = load_llm()
103
+
104
+ sentence = st.text_input("please enter description:")
105
+
106
+ if sentence !='':
107
+ answer = predict(sentence,st.session_state.retriever,st.session_state.llm )
108
+ st.write("answer:",answer)