Spaces:
Runtime error
Runtime error
AminFaraji
commited on
Update app.py
Browse files
app.py
CHANGED
@@ -1,9 +1,183 @@
|
|
1 |
-
import
|
2 |
-
import
|
3 |
-
import
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
4 |
|
5 |
-
|
6 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
7 |
|
8 |
-
|
9 |
-
demo.launch()
|
|
|
1 |
+
import argparse
|
2 |
+
# from dataclasses import dataclass
|
3 |
+
from langchain_community.vectorstores import Chroma
|
4 |
+
#from langchain_openai import OpenAIEmbeddings
|
5 |
+
#from langchain_openai import ChatOpenAI
|
6 |
+
from langchain.prompts import ChatPromptTemplate
|
7 |
+
|
8 |
+
|
9 |
+
# a template by which the bot will answer the quetion according the "context" of
|
10 |
+
# the text that will be imported as context later, determines the information that the question should be answered according to.
|
11 |
+
PROMPT_TEMPLATE = """
|
12 |
+
Answer the question based only on the following context:
|
13 |
+
|
14 |
+
{context}
|
15 |
+
|
16 |
+
---
|
17 |
+
|
18 |
+
Answer the question based on the above context: {question}
|
19 |
+
"""
|
20 |
+
# from langchain.document_loaders import DirectoryLoader
|
21 |
+
from langchain_community.document_loaders import DirectoryLoader
|
22 |
+
from langchain.text_splitter import RecursiveCharacterTextSplitter
|
23 |
+
from langchain.schema import Document
|
24 |
+
# from langchain.embeddings import OpenAIEmbeddings
|
25 |
+
#from langchain_openai import OpenAIEmbeddings
|
26 |
+
from langchain_community.vectorstores import Chroma
|
27 |
+
import openai
|
28 |
+
from dotenv import load_dotenv
|
29 |
+
import os
|
30 |
+
import shutil
|
31 |
+
# a custom embedding
|
32 |
+
from sentence_transformers import SentenceTransformer
|
33 |
+
from langchain_experimental.text_splitter import SemanticChunker
|
34 |
+
from typing import List
|
35 |
+
|
36 |
+
|
37 |
+
class MyEmbeddings:
|
38 |
+
def __init__(self):
|
39 |
+
self.model = SentenceTransformer("sentence-transformers/all-MiniLM-L6-v2")
|
40 |
+
|
41 |
+
def embed_documents(self, texts: List[str]) -> List[List[float]]:
|
42 |
+
return [self.model.encode(t).tolist() for t in texts]
|
43 |
+
def embed_query(self, query: str) -> List[float]:
|
44 |
+
return [self.model.encode([query])][0][0].tolist()
|
45 |
+
|
46 |
+
|
47 |
+
embeddings = MyEmbeddings()
|
48 |
+
|
49 |
+
splitter = SemanticChunker(embeddings)
|
50 |
+
CHROMA_PATH = "/content/drive/My Drive/chroma8"
|
51 |
+
# call the chroma generated in a directory
|
52 |
+
db = Chroma(persist_directory=CHROMA_PATH, embedding_function=embeddings)
|
53 |
+
|
54 |
+
|
55 |
+
from transformers import AutoTokenizer
|
56 |
+
import transformers
|
57 |
+
import torch
|
58 |
+
|
59 |
+
model = "tiiuae/falcon-7b-instruct" # meta-llama/Llama-2-7b-chat-hf
|
60 |
+
|
61 |
+
tokenizer = AutoTokenizer.from_pretrained(model, use_auth_token=True)
|
62 |
+
|
63 |
+
|
64 |
+
|
65 |
+
|
66 |
+
from transformers import pipeline
|
67 |
+
|
68 |
+
llama_pipeline = pipeline(
|
69 |
+
"text-generation", # LLM task
|
70 |
+
model=model,
|
71 |
+
torch_dtype=torch.float16,
|
72 |
+
device_map="auto",
|
73 |
+
)
|
74 |
+
|
75 |
+
|
76 |
+
|
77 |
+
def get_response(prompt: str) -> None:
|
78 |
+
"""
|
79 |
+
Generate a response from the Llama model.
|
80 |
+
|
81 |
+
Parameters:
|
82 |
+
prompt (str): The user's input/question for the model.
|
83 |
+
|
84 |
+
Returns:
|
85 |
+
None: Prints the model's response.
|
86 |
+
"""
|
87 |
+
sequences = llama_pipeline(
|
88 |
+
prompt,
|
89 |
+
do_sample=True,
|
90 |
+
top_k=10,
|
91 |
+
num_return_sequences=1,
|
92 |
+
eos_token_id=tokenizer.eos_token_id,
|
93 |
+
max_length=256,
|
94 |
+
)
|
95 |
+
print("Chatbot:", sequences[0]['generated_text'])
|
96 |
+
|
97 |
+
|
98 |
+
|
99 |
+
|
100 |
+
template = """Answer the query based only the provided context, and if the answer is not contained within the context below, say "I don't knowwwww"
|
101 |
|
102 |
+
Context:
|
103 |
+
{context}
|
104 |
+
|
105 |
+
{query}""".strip()
|
106 |
+
|
107 |
+
from langchain.prompts import PromptTemplate
|
108 |
+
|
109 |
+
prompt_template = PromptTemplate(
|
110 |
+
input_variables=["query", "context"],
|
111 |
+
template=template
|
112 |
+
)
|
113 |
+
|
114 |
+
|
115 |
+
|
116 |
+
|
117 |
+
|
118 |
+
|
119 |
+
|
120 |
+
|
121 |
+
|
122 |
+
|
123 |
+
|
124 |
+
# Generate a response from the Llama model
|
125 |
+
def get_llama_response(message: str, history: list) -> str:
|
126 |
+
"""
|
127 |
+
Generates a conversational response from the Llama model.
|
128 |
+
|
129 |
+
Parameters:
|
130 |
+
message (str): User's input message.
|
131 |
+
history (list): Past conversation history.
|
132 |
+
|
133 |
+
Returns:
|
134 |
+
str: Generated response from the Llama model.
|
135 |
+
"""
|
136 |
+
print('messageeeeeeeeeeeeeee:',message)
|
137 |
+
#query = format_message(message, history)
|
138 |
+
response = ""
|
139 |
+
|
140 |
+
query = """
|
141 |
+
Answer the question based only on the following context. Dont provide any information out of the context:
|
142 |
+
|
143 |
+
{context}
|
144 |
+
|
145 |
+
---
|
146 |
+
|
147 |
+
Answer the question based on the above context: {question}
|
148 |
+
"""
|
149 |
+
|
150 |
+
#message='how does alice meet the mad hatter?'
|
151 |
+
######################
|
152 |
+
# Search the DB for similar documents to the query.
|
153 |
+
results = db.similarity_search_with_relevance_scores(message, k=3)
|
154 |
+
if len(results) == 0 or results[0][1] < 0.5:
|
155 |
+
print(f"Unable to find matching results.")
|
156 |
+
|
157 |
+
|
158 |
+
context_text = "\n\n---\n\n".join([doc.page_content for doc, _score in results])
|
159 |
+
#context_text="amin is a math student."
|
160 |
+
####################3
|
161 |
+
|
162 |
+
query = prompt_template.format(query=message, context=context_text)
|
163 |
+
#query=query.format(context=context_text,question=message)
|
164 |
+
sequences = llama_pipeline(
|
165 |
+
query,
|
166 |
+
do_sample=True,
|
167 |
+
top_k=10,
|
168 |
+
num_return_sequences=1,
|
169 |
+
eos_token_id=tokenizer.eos_token_id,
|
170 |
+
max_length=1024,
|
171 |
+
)
|
172 |
+
|
173 |
+
generated_text = sequences[0]['generated_text']
|
174 |
+
response = generated_text[len(query):] # Remove the prompt from the output
|
175 |
+
|
176 |
+
print("Chatbot:", response.strip())
|
177 |
+
return response.strip()
|
178 |
+
|
179 |
+
|
180 |
+
|
181 |
+
import gradio as gr
|
182 |
|
183 |
+
gr.ChatInterface(get_llama_response).launch()
|
|