Spaces:
Runtime error
Runtime error
first commit
Browse files
app.py
ADDED
@@ -0,0 +1,111 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# Demonstrates the basic usage
|
2 |
+
# 1. Select model
|
3 |
+
# 2. Provide a query
|
4 |
+
# 3. Invoke the model
|
5 |
+
|
6 |
+
import streamlit as st
|
7 |
+
from dotenv import load_dotenv
|
8 |
+
import os
|
9 |
+
import time
|
10 |
+
|
11 |
+
from langchain_community.llms import HuggingFaceHub
|
12 |
+
from langchain_community.llms import HuggingFaceEndpoint
|
13 |
+
|
14 |
+
|
15 |
+
# Load the API keys, if running locally
|
16 |
+
# CHANGE the path to the env file
|
17 |
+
|
18 |
+
# If HF space is used then set the env var HUGGINGFACEHUB_API_TOKEN in the settings
|
19 |
+
try:
|
20 |
+
load_dotenv('C:\\Users\\raj\\.jupyter\\.env')
|
21 |
+
except:
|
22 |
+
print("Environment file not found !! MUST find the env var HUGGINGFACEHUB_API_TOKEN to work.")
|
23 |
+
|
24 |
+
|
25 |
+
|
26 |
+
# Title
|
27 |
+
st.title('Try out the model')
|
28 |
+
|
29 |
+
# Models select box
|
30 |
+
models = [
|
31 |
+
'mistralai/Mistral-7B-Instruct-v0.2',
|
32 |
+
'google/flan-t5-xxl',
|
33 |
+
'tiiuae/falcon-40b-instruct',
|
34 |
+
'microsoft/phi-2'
|
35 |
+
]
|
36 |
+
|
37 |
+
model_id = st.sidebar.selectbox(
|
38 |
+
'Select model',
|
39 |
+
options=tuple(models)
|
40 |
+
)
|
41 |
+
|
42 |
+
# Read the API key from environment - switch key for different providers
|
43 |
+
api_token = os.environ.get('HUGGINGFACEHUB_API_TOKEN')
|
44 |
+
|
45 |
+
if 'model-response' not in st.session_state:
|
46 |
+
st.session_state['model-response'] = '<provide query & click on invoke>'
|
47 |
+
|
48 |
+
# draw the box for model response
|
49 |
+
st.text_area('Response', value = st.session_state['model-response'], height=400)
|
50 |
+
|
51 |
+
# draw the box for query
|
52 |
+
query = st.text_area('Query', placeholder='provide query & invoke', value='who was the president of the USA in 2023?')
|
53 |
+
|
54 |
+
# Model parameter controls
|
55 |
+
# https://api.python.langchain.com/en/latest/llms/langchain_community.llms.huggingface_endpoint.HuggingFaceEndpoint.html
|
56 |
+
|
57 |
+
# Temperature
|
58 |
+
temperature = st.sidebar.slider(
|
59 |
+
label='Temperature',
|
60 |
+
min_value=0.01,
|
61 |
+
max_value=1.0
|
62 |
+
)
|
63 |
+
|
64 |
+
# Top p
|
65 |
+
top_p = st.sidebar.slider(
|
66 |
+
label='Top p',
|
67 |
+
min_value=0.01,
|
68 |
+
max_value=1.0,
|
69 |
+
value=0.01
|
70 |
+
)
|
71 |
+
|
72 |
+
# Top k
|
73 |
+
top_k = st.sidebar.slider(
|
74 |
+
label='Top k',
|
75 |
+
min_value=1,
|
76 |
+
max_value=50,
|
77 |
+
value=10
|
78 |
+
)
|
79 |
+
|
80 |
+
repetition_penalty = st.sidebar.slider(
|
81 |
+
label='Repeatition penalty',
|
82 |
+
min_value=0.0,
|
83 |
+
max_value=5.0,
|
84 |
+
value=1.0
|
85 |
+
)
|
86 |
+
|
87 |
+
# Maximum token
|
88 |
+
max_tokens = st.sidebar.number_input(
|
89 |
+
label='Max tokens',
|
90 |
+
value=50
|
91 |
+
)
|
92 |
+
|
93 |
+
# invoke the LLM
|
94 |
+
model_kwargs={ "temperature": "0.1" }
|
95 |
+
def invoke():
|
96 |
+
llm_hf = HuggingFaceEndpoint(
|
97 |
+
repo_id=model_id,
|
98 |
+
temperature=temperature,
|
99 |
+
top_k = top_k,
|
100 |
+
top_p = top_p,
|
101 |
+
repetition_penalty = repetition_penalty,
|
102 |
+
max_new_tokens=max_tokens
|
103 |
+
)
|
104 |
+
|
105 |
+
# Show spinner, while we are waiting for the response
|
106 |
+
with st.spinner('Invoking LLM ... '):
|
107 |
+
time.sleep(5)
|
108 |
+
st.session_state['model-response'] = llm_hf.invoke(query)
|
109 |
+
print(query)
|
110 |
+
|
111 |
+
st.button("Invoke", on_click=invoke)
|