download at runtime

.gitignore

@@ -1,4 +1,5 @@
 .venv
 .env
 __pycache__
-models
+models
+*.bin

app.py

@@ -1,75 +1,90 @@
+import requests
 import streamlit as st
-from langchain.prompts import PromptTemplate
 from langchain.llms import CTransformers
+from langchain.prompts import PromptTemplate
+import os
 
-# Load model directly
-# Load model directly
-from transformers import AutoModel
-
+def download_model() -> None:
+    """
+    Downloads the model from the provided URL and saves it to the current directory.
+    """
+    url = 'https://huggingface.co/TheBloke/Llama-2-7B-Chat-GGML/resolve/main/llama-2-7b-chat.ggmlv3.q8_0.bin'
+    file_name = url.split('/')[-1]
 
-#Function to get the response back
-def getLLMResponse(form_input,email_sender,email_recipient,email_style):
-    #llm = OpenAI(temperature=.9, model="text-davinci-003")
+    response = requests.get(url, stream=True)
 
-    # Wrapper for Llama-2-7B-Chat, Running Llama 2 on CPU
+    with open(file_name, 'wb') as file:
+        for chunk in response.iter_content(chunk_size=1024):
+            if chunk:
+                file.write(chunk)
 
-    #Quantization is reducing model precision by converting weights from 16-bit floats to 8-bit integers, 
-    #enabling efficient deployment on resource-limited devices, reducing model size, and maintaining performance.
+    print("File downloaded successfully!")
 
-    #C Transformers offers support for various open-source models, 
-    #among them popular ones like Llama, GPT4All-J, MPT, and Falcon.
 
+def getLLMResponse(form_input: str, email_sender: str, email_recipient: str, email_style: str) -> str:
+    """
+    Generates a response using the LLM model.
 
-    #C Transformers is the Python library that provides bindings for transformer models implemented in C/C++ using the GGML library
+    :param form_input: Email topic provided by the user.
+    :param email_sender: Sender name provided by the user.
+    :param email_recipient: Recipient name provided by the user.
+    :param email_style: Writing style provided by the user.
+    :return: Generated response.
+    """
+    llm = CTransformers(model='llama-2-7b-chat.ggmlv3.q8_0.bin',
+                        model_type='llama',
+                        config={'max_new_tokens': 256,
+                                'temperature': 0.01})
 
-    llm = CTransformers(model='models/llama-2-7b-chat.ggmlv3.q8_0.bin',     #https://huggingface.co/TheBloke/Llama-2-7B-Chat-GGML/tree/main
-                    model_type='llama',
-                    config={'max_new_tokens': 256,
-                            'temperature': 0.01})
-    
-    # llm = AutoModel.from_pretrained("TheBloke/Llama-2-7B-Chat-GGML")
-    #Template for building the PROMPT
     template = """
-    Write a email with {style} style and includes topic :{email_topic}.\n\nSender: {sender}\nRecipient: {recipient}
+    Write an email with {style} style and includes topic :{email_topic}.\n\nSender: {sender}\nRecipient: {recipient}
     \n\nEmail Text:
     
     """
 
-    #Creating the final PROMPT
     prompt = PromptTemplate(
-    input_variables=["style","email_topic","sender","recipient"],
+    input_variables=["style", "email_topic", "sender", "recipient"],
     template=template,)
 
-  
-    #Generating the response using LLM
-    response=llm(prompt.format(email_topic=form_input,sender=email_sender,recipient=email_recipient,style=email_style))
+    response = llm(prompt.format(email_topic=form_input, sender=email_sender, recipient=email_recipient, style=email_style))
     print(response)
 
     return response
 
 
 st.set_page_config(page_title="Generate Emails",
-                    page_icon='📧',
-                    layout='centered',
-                    initial_sidebar_state='collapsed')
+                   page_icon='📧',
+                   layout='centered',
+                   initial_sidebar_state='collapsed')
 st.header("Generate Emails 📧")
 
-form_input = st.text_area('Enter the email topic', height=275)
-
-#Creating columns for the UI - To receive inputs from user
-col1, col2, col3 = st.columns([10, 10, 5])
-with col1:
-    email_sender = st.text_input('Sender Name')
-with col2:
-    email_recipient = st.text_input('Recipient Name')
-with col3:
-    email_style = st.selectbox('Writing Style',
+model_loaded = st.session_state.get('model_loaded', False)
+
+if not model_loaded:
+    if st.button('Load Model'):
+        model_file = 'llama-2-7b-chat.ggmlv3.q8_0.bin'
+        if not os.path.isfile(model_file):
+            st.info('Loading the model, this could take ~5 minutes')
+            download_model()
+            st.session_state.model_loaded = True
+            st.info('Model loaded successfully')
+
+if st.session_state.get('model_loaded'):
+    form_input = st.text_area('Enter the email topic', height=275)
+
+    col1, col2, col3 = st.columns([10, 10, 5])
+    with col1:
+        email_sender = st.text_input('Sender Name')
+    with col2:
+        email_recipient = st.text_input('Recipient Name')
+    with col3:
+        email_style = st.selectbox('Writing Style',
                                     ('Formal', 'Appreciating', 'Not Satisfied', 'Neutral'),
-                                       index=0)
-
+                                   index=0)
 
-submit = st.button("Generate")
+    submit = st.button("Generate")
 
-#When 'Generate' button is clicked, execute the below code
-if submit:
-    st.write(getLLMResponse(form_input,email_sender,email_recipient,email_style))
+    if submit:
+        st.write(getLLMResponse(form_input, email_sender, email_recipient, email_style))
+else:
+    st.write("Please load the model to proceed.")