File size: 6,788 Bytes
8869ff1
 
 
b52c479
c201c90
8869ff1
 
b52c479
 
 
 
 
 
 
 
 
5580b5a
 
 
 
b52c479
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
5580b5a
b52c479
 
5580b5a
 
b52c479
 
 
 
4463bb5
b52c479
 
 
 
4463bb5
 
 
 
b52c479
4463bb5
 
 
 
 
 
8869ff1
b52c479
 
 
 
 
 
 
 
8869ff1
b52c479
 
 
 
 
 
 
 
 
 
 
c201c90
b52c479
 
 
 
 
 
 
 
 
 
 
5580b5a
 
b52c479
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
import os
import streamlit as st
from langchain.llms import HuggingFaceHub
from llm import similarity
from file_manipulation import make_directory_if_not_exists

from models import return_models, return_text2text_generation_models, return_task_name, return_text_generation_models
class LLM_Langchain():
    def __init__(self):
        dummy_parent = "google"
        self.models_count = return_text2text_generation_models(dummy_parent, True) + return_text_generation_models(dummy_parent, True)
        st.warning("Warning: Some models may not work and some models may require GPU to run")
        st.text(f"As of now there are {self.models_count} model available")
        st.text("Made with Langchain, StreamLit, Hugging Face and 💖")
        st.header('🦜🔗 One stop for Open Source Models')

        # self.API_KEY = st.sidebar.text_input(
        #     'API Key',
        #     type='password',
        #     help="Type in your HuggingFace API key to use this app")

        self.task_name = st.sidebar.selectbox(
            label = "Choose the task you want to perform",
            options = return_task_name(),
            help="Choose your open source LLM to get started"
            )
        if self.task_name is None:
            model_parent_visibility = True
        else:
            model_parent_visibility = False
            
        model_parent_options = return_models(self.task_name)
        model_parent = st.sidebar.selectbox(
            label = "Choose your Source",
            options = model_parent_options,
            help="Choose your source of models",
            disabled=model_parent_visibility
            )

        if model_parent is None:
            model_name_visibility = True
        else:
            model_name_visibility = False
        if self.task_name == "text2text-generation":
            options = return_text2text_generation_models(model_parent)
        else:
            options = return_text_generation_models(model_parent)
        self.model_name = st.sidebar.selectbox(
            label = "Choose your Models",
            options = options,
            help="Choose your open source LLM to get started",
            disabled=model_name_visibility
            )

        self.temperature = st.sidebar.slider(
            label="Temperature",
            min_value=0.1,
            max_value=1.0,
            step=0.1,
            value=0.9,
            help="Set the temperature to get accurate results"
            )

        self.max_token_length = st.sidebar.slider(
            label="Token Length",
            min_value=32,
            max_value=1024,
            step=32,
            value=1024,
            help="Set the max tokens to get accurate results"
            )
        

        self.model_kwargs = {
            "temperature": self.temperature,
            "max_new_tokens": self.max_token_length
        }

        # os.environ['HUGGINGFACEHUB_API_TOKEN'] = self.API_KEY
        os.environ['HUGGINGFACEHUB_API_TOKEN'] = os.getenv("HF_KEY")


    def generate_response(self, input_text):
        
        template = "<|system|>\nYou are a intelligent chatbot.</s>\n<|user|>\n{input_text}.\n<|assistant|>"
        llm = HuggingFaceHub(
            repo_id = self.model_name,
            model_kwargs = self.model_kwargs
        )
        llm_chain = LLMChain(
                    prompt=template,
                    llm=llm,
                )

        result = llm_chain.run({
            "question": input_text,
            # "context": context
        })
        # return llm(input_text)
        return result
    

    def radio_button(self):
        options = ['FineTune', 'Inference']
        selected_option = st.radio(
            label="Choose your options",
            options=options
            )
        return selected_option
    

    def pdf_uploader(self):
        if self.selected_option == "Inference":
            self.uploader_visibility = True
        else:
            self.uploader_visibility = False

        self.file_upload_status = st.file_uploader(
            label="Upload PDF file",
            disabled=self.uploader_visibility
        )
        make_directory_if_not_exists('assets/')

        if self.file_upload_status is not None:
            self.pdf_file_path = f"assets/{self.file_upload_status.name}"
        
            with open(self.pdf_file_path, "wb") as f:
                f.write(self.file_upload_status.getbuffer())
            st.write("File Uploaded Successfully")

    def form_data(self):
        # with st.form('my_form'):
            try:
                # if not self.API_KEY.startswith('hf_'):
                #     st.warning('Please enter your API key!', icon='⚠')
                
                self.selected_option = self.radio_button()
                self.pdf_uploader()

                if self.selected_option == "FineTune":
                    if self.file_upload_status is None:
                        text_input_visibility = True
                    else:
                        text_input_visibility = False
                else:
                    text_input_visibility = False
                
                if "messages" not in st.session_state:
                    st.session_state.messages = []

                st.write(f"You are using {self.model_name} model")

                for message in st.session_state.messages:
                    with st.chat_message(message.get('role')):
                        st.write(message.get("content"))
                text = st.chat_input(disabled=text_input_visibility)
                
                if text:
                    st.session_state.messages.append(
                        {
                            "role":"user",
                            "content": text
                        }
                    )
                    with st.chat_message("user"):
                        st.write(text)
                    
                    if text.lower() == "clear":
                        del st.session_state.messages
                        return
                    if self.selected_option == 'FineTune':
                        result = similarity(self.pdf_file_path, self.model_name, self.model_kwargs, text)
                    else:
                        result = self.generate_response(text)
                    st.session_state.messages.append(
                        {
                            "role": "assistant",
                            "content": result
                        }
                    )
                    with st.chat_message('assistant'):
                        st.markdown(result)
                
            except Exception as e:
                st.error(e, icon="🚨")

model = LLM_Langchain()
model.form_data()