app.py

# import streamlit as st
# from transformers import pipeline 

# # pipe=pipeline("sentiment-analysis")
# # col1, col2 = st.columns(2)

# # with col1:
# #     x=st.button("Sentiment Analysis")
# # with col2:    
# #     y=st.button("Text Summarization")

# # if x:
# #     t=st.text_input("Enter the Text")
# #     st.write(pipe(t))
# # if y:             
# t1=st.text_input("Enter the Text for Summarization")
# st.write(summarizer(t1))

#from transformers import AutoTokenizer, AutoModel
# import streamlit as st

#tokenizer = AutoTokenizer.from_pretrained("llmware/industry-bert-insurance-v0.1")

# #model = AutoModel.from_pretrained("llmware/industry-bert-insurance-v0.1")
# # Use a pipeline as a high-level helper
# from transformers import pipeline

# #pipe = pipeline("feature-extraction")
   
# t=st.text_input("Enter the Text")
# pipe = pipeline("summarization")
# st.write(pipe(t))


# import pandas as pd
# import numpy as np
# from ydata_synthetic.synthesizers.regular import RegularSynthesizer
# from ydata_synthetic.synthesizers import ModelParameters, TrainParameters
# import streamlit as st 
# from os import getcwd
# text_file=st.file_uploader("Upload the Data File")
# st.write("-------------------------")

# if text_file is not None:
#     df=pd.read_csv(text_file)
#     dd_list=df.columns
#     cat_cols=st.multiselect("Select the Categorical Columns",dd_list)
#     num_cols=st.multiselect("Select the Numerical Columns",dd_list)
#     Output_file=st.text_input('Enter Output File Name')
#     s=st.number_input('Enter the Sample Size',1000)
#     OP=Output_file + '.csv'
#     sub=st.button('Submit')
#     if sub:
#         batch_size = 50
#         epochs = 3
#         learning_rate = 2e-4
#         beta_1 = 0.5
#         beta_2 = 0.9

#         ctgan_args = ModelParameters(batch_size=batch_size,
#                                     lr=learning_rate,
#                                     betas=(beta_1, beta_2))

#         train_args = TrainParameters(epochs=epochs)
#         synth = RegularSynthesizer(modelname='ctgan', model_parameters=ctgan_args)
#         synth.fit(data=df, train_arguments=train_args, num_cols=num_cols, cat_cols=cat_cols)
#         df_syn = synth.sample(s)
#         df_syn.to_csv(OP)
#         c=getcwd()    
#         c=c + '/' + OP
#         with open(c,"rb") as file:
#             st.download_button(label=':blue[Download]',data=file,file_name=OP,mime="image/png") 
#         st.success("Thanks for using the app !!!")

# import torch
# import streamlit as st
# from transformers import AutoModelForCausalLM, AutoTokenizer

# #torch.set_default_device("cuda")

# model = AutoModelForCausalLM.from_pretrained("soulhq-ai/phi-2-insurance_qa-sft-lora", torch_dtype="auto", trust_remote_code=True)
# tokenizer = AutoTokenizer.from_pretrained("soulhq-ai/phi-2-insurance_qa-sft-lora", trust_remote_code=True)
# i=st.text_input('Prompt', 'Life of Brian')
# #inputs = tokenizer('''### Instruction: What Does Basic Homeowners Insurance Cover?\n### Response: ''', return_tensors="pt", return_attention_mask=False)
# inputs = tokenizer(i, return_tensors="pt", return_attention_mask=False)
# outputs = model.generate(**inputs, max_length=1024)
# text = tokenizer.batch_decode(outputs)[0]
# print(text)

# import torch
# import streamlit as st
# from transformers import AutoModelForSeq2SeqLM, AutoTokenizer

# model_name="facebook/blenderbot-400M-distill"

# model=AutoModelForSeq2SeqLM.from_pretrained(model_name)
# tokenizer = AutoTokenizer.from_pretrained(model_name)
# ch=[]
# def chat():
    
#     h_s="\n".join(ch)
#     i=st.text_input("enter")
#     i_s=tokenizer.encode_plus(h_s,i,return_tensors="pt")
#     outputs=model.generate(**i_s,max_length=60)
#     response=tokenizer.decode(outputs[0],skip_special_tokens=True).strip()
#     ch.append(i)
#     ch.append(response)
#     return response
# if __name__ == "__main__":
#     chat()

import streamlit as st
from streamlit_chat import message as st_message
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM


@st.experimental_singleton
def get_models():
    # Load the model and the tokenizer
    tokenizer = AutoTokenizer.from_pretrained("facebook/blenderbot_small-90M")

    model = AutoModelForSeq2SeqLM.from_pretrained(
        "facebook/blenderbot_small-90M")

    return tokenizer, model


if "history" not in st.session_state:
    st.session_state.history = []

st.title("Blenderbot")


def generate_answer():
    tokenizer, model = get_models()
    user_message = st.session_state.input_text
    inputs = tokenizer(st.session_state.input_text, return_tensors="pt")
    result = model.generate(**inputs)
    message_bot = tokenizer.decode(
        result[0], skip_special_tokens=True
    )   # decode the result to a string

    st.session_state.history.append({"message": user_message, "is_user": True})
    st.session_state.history.append({"message": message_bot, "is_user": False})


st.text_input("Tap to chat with the bot",
              key="input_text", on_change=generate_answer)

for chat in st.session_state.history:
    st_message(**chat)