Spaces:
Runtime error
Runtime error
Commit
•
f0fa217
1
Parent(s):
012be20
Create pages/Phase1.py
Browse files- pages/Phase1.py +109 -0
pages/Phase1.py
ADDED
@@ -0,0 +1,109 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import streamlit as st
|
2 |
+
from huggingface_hub import InferenceClient
|
3 |
+
import os
|
4 |
+
import fitz # PyMuPDF
|
5 |
+
|
6 |
+
st.title("ChatGPT-like Chatbot")
|
7 |
+
|
8 |
+
base_url = "https://api-inference.huggingface.co/models/"
|
9 |
+
|
10 |
+
API_KEY = os.environ.get('HUGGINGFACE_API_KEY')
|
11 |
+
headers = {"Authorization": "Bearer " + str(API_KEY)}
|
12 |
+
|
13 |
+
model_links = {
|
14 |
+
"Mistral-7B": base_url + "mistralai/Mistral-7B-Instruct-v0.2",
|
15 |
+
"Mistral-22B": base_url + "mistral-community/Mixtral-8x22B-v0.1",
|
16 |
+
}
|
17 |
+
|
18 |
+
model_info = {
|
19 |
+
"Mistral-7B": {
|
20 |
+
'description': """The Mistral model is a **Large Language Model (LLM)** that's able to have question and answer interactions.\n \
|
21 |
+
\nIt was created by the [**Mistral AI**](https://mistral.ai/news/announcing-mistral-7b/) team as has over **7 billion parameters.** \n""",
|
22 |
+
'logo': 'https://mistral.ai/images/logo_hubc88c4ece131b91c7cb753f40e9e1cc5_2589_256x0_resize_q97_h2_lanczos_3.webp'},
|
23 |
+
"Zephyr-7B": {
|
24 |
+
'description': """The Zephyr model is a **Large Language Model (LLM)** that's able to have question and answer interactions.\n \
|
25 |
+
\nFrom Huggingface: \n\
|
26 |
+
Zephyr is a series of language models that are trained to act as helpful assistants. \
|
27 |
+
[Zephyr 7B Gemma](https://huggingface.co/HuggingFaceH4/zephyr-7b-gemma-v0.1)\
|
28 |
+
is the third model in the series, and is a fine-tuned version of google/gemma-7b \
|
29 |
+
that was trained on on a mix of publicly available, synthetic datasets using Direct Preference Optimization (DPO)\n""",
|
30 |
+
'logo': 'https://huggingface.co/HuggingFaceH4/zephyr-7b-gemma-v0.1/resolve/main/thumbnail.png'}
|
31 |
+
}
|
32 |
+
|
33 |
+
def format_prompt(message, custom_instructions=None):
|
34 |
+
prompt = ""
|
35 |
+
if custom_instructions:
|
36 |
+
prompt += f"[INST] {custom_instructions} [/INST]"
|
37 |
+
prompt += f"[INST] {message} [/INST]"
|
38 |
+
return prompt
|
39 |
+
|
40 |
+
def reset_conversation():
|
41 |
+
st.session_state.conversation = []
|
42 |
+
st.session_state.messages = []
|
43 |
+
return None
|
44 |
+
|
45 |
+
def read_pdf(file_path):
|
46 |
+
doc = fitz.open(file_path)
|
47 |
+
text = ""
|
48 |
+
for page in doc:
|
49 |
+
text += page.get_text()
|
50 |
+
return text
|
51 |
+
|
52 |
+
models = [key for key in model_links.keys()]
|
53 |
+
|
54 |
+
selected_model = st.sidebar.selectbox("Select Model", models)
|
55 |
+
temp_values = st.sidebar.slider('Select a temperature value', 0.0, 1.0, (0.5))
|
56 |
+
st.sidebar.button('Reset Chat', on_click=reset_conversation)
|
57 |
+
st.sidebar.write(f"You're now chatting with **{selected_model}**")
|
58 |
+
st.sidebar.markdown(model_info[selected_model]['description'])
|
59 |
+
st.sidebar.image(model_info[selected_model]['logo'])
|
60 |
+
st.sidebar.markdown("*Generated content may be inaccurate or false.*")
|
61 |
+
|
62 |
+
if "prev_option" not in st.session_state:
|
63 |
+
st.session_state.prev_option = selected_model
|
64 |
+
|
65 |
+
if st.session_state.prev_option != selected_model:
|
66 |
+
st.session_state.messages = []
|
67 |
+
st.session_state.prev_option = selected_model
|
68 |
+
reset_conversation()
|
69 |
+
|
70 |
+
repo_id = model_links[selected_model]
|
71 |
+
st.subheader(f'AI - {selected_model}')
|
72 |
+
|
73 |
+
if "messages" not in st.session_state:
|
74 |
+
st.session_state.messages = []
|
75 |
+
|
76 |
+
for message in st.session_state.messages:
|
77 |
+
with st.chat_message(message["role"]):
|
78 |
+
st.markdown(message["content"])
|
79 |
+
|
80 |
+
pdf_path = st.file_uploader("Upload a PDF file", type="pdf")
|
81 |
+
|
82 |
+
if pdf_path:
|
83 |
+
pdf_text = read_pdf(pdf_path)
|
84 |
+
|
85 |
+
if prompt := st.chat_input(f"Hi I'm {selected_model}, ask me a question"):
|
86 |
+
custom_instruction = "Act like a Human in conversation"
|
87 |
+
|
88 |
+
with st.chat_message("user"):
|
89 |
+
st.markdown(prompt st.session_state.messages.append({"role": "user", "content": prompt})
|
90 |
+
|
91 |
+
formatted_prompt = format_prompt(prompt, custom_instruction)
|
92 |
+
|
93 |
+
with st.chat_message("assistant"):
|
94 |
+
with st.spinner("Thinking..."):
|
95 |
+
client = InferenceClient(model=repo_id, token=API_KEY)
|
96 |
+
|
97 |
+
if pdf_path:
|
98 |
+
# If a PDF is uploaded, use its text for answering
|
99 |
+
formatted_prompt = f"{pdf_text}\n\n{formatted_prompt}"
|
100 |
+
|
101 |
+
response = client.text_generation(
|
102 |
+
formatted_prompt,
|
103 |
+
max_length=500,
|
104 |
+
temperature=temp_values
|
105 |
+
)
|
106 |
+
|
107 |
+
response_content = response["generated_text"]
|
108 |
+
st.markdown(response_content)
|
109 |
+
st.session_state.messages.append({"role": "assistant", "content": response_content})
|