Spaces:
Runtime error
Runtime error
MikeCraBash
commited on
Commit
•
7f049a0
1
Parent(s):
9ba91e5
app.py
CHANGED
@@ -1,14 +1,14 @@
|
|
1 |
-
#
|
|
|
2 |
# Date: 2024-5-16
|
3 |
|
4 |
# Basic Imports & Setup
|
5 |
import os
|
6 |
-
from
|
7 |
|
8 |
# Using Chainlit for our UI
|
9 |
import chainlit as cl
|
10 |
from chainlit.prompt import Prompt, PromptMessage
|
11 |
-
from chainlit.playground.providers import ChatOpenAI
|
12 |
|
13 |
# Getting the API key from the .env file
|
14 |
from dotenv import load_dotenv
|
@@ -27,7 +27,7 @@ docs = PyMuPDFLoader(direct_url).load()
|
|
27 |
|
28 |
import tiktoken
|
29 |
def tiktoken_len(text):
|
30 |
-
tokens = tiktoken.encoding_for_model("
|
31 |
text,
|
32 |
)
|
33 |
return len(tokens)
|
@@ -44,12 +44,12 @@ text_splitter = RecursiveCharacterTextSplitter(
|
|
44 |
split_chunks = text_splitter.split_documents(docs)
|
45 |
|
46 |
# Load the embeddings model
|
47 |
-
from
|
48 |
|
49 |
-
embedding_model =
|
50 |
|
51 |
# Load the vector store and retriever from Qdrant
|
52 |
-
from
|
53 |
|
54 |
qdrant_vectorstore = Qdrant.from_documents(
|
55 |
split_chunks,
|
@@ -60,10 +60,11 @@ qdrant_vectorstore = Qdrant.from_documents(
|
|
60 |
|
61 |
qdrant_retriever = qdrant_vectorstore.as_retriever()
|
62 |
|
63 |
-
|
64 |
-
|
|
|
65 |
|
66 |
-
from
|
67 |
|
68 |
RAG_PROMPT = """
|
69 |
SYSTEM:
|
@@ -120,14 +121,14 @@ from langchain.schema.runnable import RunnablePassthrough
|
|
120 |
retrieval_augmented_qa_chain = (
|
121 |
{"context": itemgetter("question") | qdrant_retriever, "question": itemgetter("question")}
|
122 |
| RunnablePassthrough.assign(context=itemgetter("context"))
|
123 |
-
| {"response": rag_prompt |
|
124 |
)
|
125 |
|
126 |
# Chainlit App
|
127 |
@cl.on_chat_start
|
128 |
async def start_chat():
|
129 |
settings = {
|
130 |
-
"model": "
|
131 |
"temperature": 0,
|
132 |
"max_tokens": 500,
|
133 |
"top_p": 1,
|
@@ -145,3 +146,5 @@ async def main(message: cl.Message):
|
|
145 |
|
146 |
msg = cl.Message(content=chainlit_answer)
|
147 |
await msg.send()
|
|
|
|
|
|
1 |
+
#
|
2 |
+
# HACK AI MAKERSPACE PREPR
|
3 |
# Date: 2024-5-16
|
4 |
|
5 |
# Basic Imports & Setup
|
6 |
import os
|
7 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer
|
8 |
|
9 |
# Using Chainlit for our UI
|
10 |
import chainlit as cl
|
11 |
from chainlit.prompt import Prompt, PromptMessage
|
|
|
12 |
|
13 |
# Getting the API key from the .env file
|
14 |
from dotenv import load_dotenv
|
|
|
27 |
|
28 |
import tiktoken
|
29 |
def tiktoken_len(text):
|
30 |
+
tokens = tiktoken.encoding_for_model("solar-10.7b").encode(
|
31 |
text,
|
32 |
)
|
33 |
return len(tokens)
|
|
|
44 |
split_chunks = text_splitter.split_documents(docs)
|
45 |
|
46 |
# Load the embeddings model
|
47 |
+
from langchain.embeddings import HuggingFaceEmbeddings
|
48 |
|
49 |
+
embedding_model = HuggingFaceEmbeddings(model_name="solar-10.7b")
|
50 |
|
51 |
# Load the vector store and retriever from Qdrant
|
52 |
+
from langchain.vectorstores import Qdrant
|
53 |
|
54 |
qdrant_vectorstore = Qdrant.from_documents(
|
55 |
split_chunks,
|
|
|
60 |
|
61 |
qdrant_retriever = qdrant_vectorstore.as_retriever()
|
62 |
|
63 |
+
# Load the Solar 10.7B model
|
64 |
+
tokenizer = AutoTokenizer.from_pretrained("Upstage/SOLAR-10.7B-v1.0")
|
65 |
+
model = AutoModelForCausalLM.from_pretrained("Upstage/SOLAR-10.7B-v1.0")
|
66 |
|
67 |
+
from langchain.prompts import ChatPromptTemplate
|
68 |
|
69 |
RAG_PROMPT = """
|
70 |
SYSTEM:
|
|
|
121 |
retrieval_augmented_qa_chain = (
|
122 |
{"context": itemgetter("question") | qdrant_retriever, "question": itemgetter("question")}
|
123 |
| RunnablePassthrough.assign(context=itemgetter("context"))
|
124 |
+
| {"response": rag_prompt | model, "context": itemgetter("context")}
|
125 |
)
|
126 |
|
127 |
# Chainlit App
|
128 |
@cl.on_chat_start
|
129 |
async def start_chat():
|
130 |
settings = {
|
131 |
+
"model": "solar-10.7b",
|
132 |
"temperature": 0,
|
133 |
"max_tokens": 500,
|
134 |
"top_p": 1,
|
|
|
146 |
|
147 |
msg = cl.Message(content=chainlit_answer)
|
148 |
await msg.send()
|
149 |
+
|
150 |
+
|