Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
@@ -1,57 +1,71 @@
|
|
1 |
import gradio as gr
|
2 |
-
from langchain_community.chat_models import ChatOllama
|
3 |
-
from langchain_community.embeddings import GPT4AllEmbeddings
|
|
|
|
|
|
|
|
|
|
|
4 |
|
5 |
-
from langchain.prompts import ChatPromptTemplate
|
6 |
-
from langchain.schema.runnable import RunnablePassthrough
|
7 |
-
from langchain_community.vectorstores import FAISS
|
8 |
# import os
|
9 |
# os.system("ollama pull alen_ox/llama_3_fin")
|
10 |
|
11 |
vector_db_path = "vectorstores/db_faiss"
|
12 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
13 |
# embeddings = OllamaEmbeddings(model="nomic-embed-text", show_progress=False)
|
14 |
-
embeddings = GPT4AllEmbeddings(model_name = "all-MiniLM-L6-v2.gguf2.f16.gguf", gpt4all_kwargs = {'allow_download': 'True'})
|
15 |
|
16 |
-
db = FAISS.load_local(vector_db_path, embeddings, allow_dangerous_deserialization=True)
|
17 |
|
18 |
-
# # Create retriever
|
19 |
-
retriever = db.as_retriever(
|
20 |
-
|
21 |
-
|
22 |
-
)
|
23 |
-
local_llm = 'llama3.1'
|
24 |
|
25 |
-
llm = ChatOllama(model=local_llm,
|
26 |
-
|
27 |
-
|
28 |
-
|
29 |
|
30 |
# Create prompt template
|
31 |
|
32 |
def respond(message, history, system_message, path_document):
|
33 |
-
print(message, history, system_message, path_document)
|
34 |
-
respon = ''
|
35 |
-
print("Answer:\n\n", end=" ", flush=True)
|
36 |
-
template = """Bạn là trợ lý ảo vì vậy bạn hãy sử dụng dữ liệu dưới đây để trả lời câu hỏi,
|
37 |
-
|
38 |
-
|
39 |
-
|
40 |
-
|
41 |
-
|
42 |
-
|
43 |
-
prompt = ChatPromptTemplate.from_template(template)
|
44 |
|
45 |
-
rag_chain = (
|
46 |
-
|
47 |
-
|
48 |
-
|
49 |
-
)
|
50 |
-
for chunk in rag_chain.stream(message):
|
51 |
-
|
|
|
|
|
|
|
|
|
|
|
52 |
print(chunk.content, end="", flush=True)
|
53 |
yield respon
|
54 |
-
|
55 |
demo = gr.ChatInterface(
|
56 |
respond,
|
57 |
additional_inputs=[
|
|
|
1 |
import gradio as gr
|
2 |
+
# from langchain_community.chat_models import ChatOllama
|
3 |
+
# from langchain_community.embeddings import GPT4AllEmbeddings
|
4 |
+
|
5 |
+
# from langchain.prompts import ChatPromptTemplate
|
6 |
+
# from langchain.schema.runnable import RunnablePassthrough
|
7 |
+
# from langchain_community.vectorstores import FAISS
|
8 |
+
from langchain_community.llms import LlamaCpp
|
9 |
|
|
|
|
|
|
|
10 |
# import os
|
11 |
# os.system("ollama pull alen_ox/llama_3_fin")
|
12 |
|
13 |
vector_db_path = "vectorstores/db_faiss"
|
14 |
|
15 |
+
llm = LlamaCpp(
|
16 |
+
model_path="Qwen2-7B-Instruct.Q5_K_M.gguf",
|
17 |
+
temperature=0.75,
|
18 |
+
max_tokens=2000,
|
19 |
+
top_p=1,
|
20 |
+
# callback_manager=callback_manager,
|
21 |
+
verbose=True, # Verbose is required to pass to the callback manager
|
22 |
+
)
|
23 |
# embeddings = OllamaEmbeddings(model="nomic-embed-text", show_progress=False)
|
24 |
+
# embeddings = GPT4AllEmbeddings(model_name = "all-MiniLM-L6-v2.gguf2.f16.gguf", gpt4all_kwargs = {'allow_download': 'True'})
|
25 |
|
26 |
+
# db = FAISS.load_local(vector_db_path, embeddings, allow_dangerous_deserialization=True)
|
27 |
|
28 |
+
# # # Create retriever
|
29 |
+
# retriever = db.as_retriever(
|
30 |
+
# search_type="similarity",
|
31 |
+
# search_kwargs= {"k": 3}
|
32 |
+
# )
|
33 |
+
# local_llm = 'llama3.1'
|
34 |
|
35 |
+
# llm = ChatOllama(model=local_llm,
|
36 |
+
# keep_alive="3h",
|
37 |
+
# max_tokens=512,
|
38 |
+
# temperature=0)
|
39 |
|
40 |
# Create prompt template
|
41 |
|
42 |
def respond(message, history, system_message, path_document):
|
43 |
+
# print(message, history, system_message, path_document)
|
44 |
+
# respon = ''
|
45 |
+
# print("Answer:\n\n", end=" ", flush=True)
|
46 |
+
# template = """Bạn là trợ lý ảo vì vậy bạn hãy sử dụng dữ liệu dưới đây để trả lời câu hỏi,
|
47 |
+
# nếu không có thông tin hãy đưa ra câu trả lời sát nhất với câu hỏi từ các thông tin tìm được
|
48 |
+
# Content: {content}
|
49 |
+
# Question: {question}
|
50 |
+
# Chỉ đưa ra các câu trả lời hữu ích.
|
51 |
+
# Helpful answer:
|
52 |
+
# """
|
53 |
+
# prompt = ChatPromptTemplate.from_template(template)
|
54 |
|
55 |
+
# rag_chain = (
|
56 |
+
# {"content": retriever, "question": RunnablePassthrough()}
|
57 |
+
# | prompt
|
58 |
+
# | llm
|
59 |
+
# )
|
60 |
+
# for chunk in rag_chain.stream(message):
|
61 |
+
# respon += chunk.content
|
62 |
+
# print(chunk.content, end="", flush=True)
|
63 |
+
# yield respon
|
64 |
+
|
65 |
+
for chunk in llm.stream(message):
|
66 |
+
respon += chunk
|
67 |
print(chunk.content, end="", flush=True)
|
68 |
yield respon
|
|
|
69 |
demo = gr.ChatInterface(
|
70 |
respond,
|
71 |
additional_inputs=[
|