app.py

import gradio as gr
import json
import requests
import os
from text_generation import Client, InferenceAPIClient

# Load pre-trained model and tokenizer - for THUDM model
from transformers import AutoModel, AutoTokenizer
tokenizer_glm = AutoTokenizer.from_pretrained("THUDM/chatglm-6b", trust_remote_code=True)
model_glm = AutoModel.from_pretrained("THUDM/chatglm-6b", trust_remote_code=True).half().cuda()
model_glm = model_glm.eval()

# Load pre-trained model and tokenizer for Chinese to English translator
from transformers import M2M100ForConditionalGeneration, M2M100Tokenizer
model_chtoen = M2M100ForConditionalGeneration.from_pretrained("facebook/m2m100_418M")
tokenizer_chtoen = M2M100Tokenizer.from_pretrained("facebook/m2m100_418M")

#Streaming endpoint for OPENAI ChatGPT
API_URL = "https://api.openai.com/v1/chat/completions"
#Streaming endpoint for OPENCHATKIT
API_URL_TGTHR = os.getenv('API_URL_TGTHR')

openchat_preprompt = (
    "\n<human>: Hi!\n<bot>: My name is Bot, model version is 0.15, part of an open-source kit for "
    "fine-tuning new bots! I was created by Together, LAION, and Ontocord.ai and the open-source "
    "community. I am not human, not evil and not alive, and thus have no thoughts and feelings, "
    "but I am programmed to be helpful, polite, honest, and friendly.\n")

#Predict function for CHATGPT
def predict_chatgpt(inputs, top_p_chatgpt, temperature_chatgpt, openai_api_key, chat_counter_chatgpt, chatbot_chatgpt=[], history=[]):  
    #Define payload and header for chatgpt API
    payload = {
    "model": "gpt-3.5-turbo",
    "messages": [{"role": "user", "content": f"{inputs}"}],
    "temperature" : 1.0,
    "top_p":1.0,
    "n" : 1,
    "stream": True,
    "presence_penalty":0,
    "frequency_penalty":0,
    }

    headers = {
    "Content-Type": "application/json",
    "Authorization": f"Bearer {openai_api_key}"
    }
    #debug
    #print(f"chat_counter_chatgpt - {chat_counter_chatgpt}")
    
    #Handling the different roles for ChatGPT
    if chat_counter_chatgpt != 0 :
        messages=[]
        for data in chatbot_chatgpt:
          temp1 = {}
          temp1["role"] = "user" 
          temp1["content"] = data[0] 
          temp2 = {}
          temp2["role"] = "assistant" 
          temp2["content"] = data[1]
          messages.append(temp1)
          messages.append(temp2)
        temp3 = {}
        temp3["role"] = "user" 
        temp3["content"] = inputs
        messages.append(temp3)
        payload = {
        "model": "gpt-3.5-turbo",
        "messages": messages, #[{"role": "user", "content": f"{inputs}"}],
        "temperature" : temperature_chatgpt, #1.0,
        "top_p": top_p_chatgpt, #1.0,
        "n" : 1,
        "stream": True,
        "presence_penalty":0,
        "frequency_penalty":0,
        }

    chat_counter_chatgpt+=1

    history.append(inputs)

    # make a POST request to the API endpoint using the requests.post method, passing in stream=True
    response = requests.post(API_URL, headers=headers, json=payload, stream=True)
    token_counter = 0 
    partial_words = ""  

    counter=0
    for chunk in response.iter_lines():
        #Skipping the first chunk
        if counter == 0:
          counter+=1
          continue
        # check whether each line is non-empty
        if chunk.decode() :
          chunk = chunk.decode()
          # decode each line as response data is in bytes
          if len(chunk) > 13 and "content" in json.loads(chunk[6:])['choices'][0]["delta"]:
              partial_words = partial_words + json.loads(chunk[6:])['choices'][0]["delta"]["content"]
              if token_counter == 0:
                history.append(" " + partial_words)
              else:
                history[-1] = partial_words
              chat = [(history[i], history[i + 1]) for i in range(0, len(history) - 1, 2) ]  # convert to tuples of list
              token_counter+=1
              yield chat, history, chat_counter_chatgpt  # this resembles {chatbot: chat, state: history}  


#Predict function for OPENCHATKIT
def predict_together(model: str,
        inputs: str,
        top_p: float,
        temperature: float,
        top_k: int,
        repetition_penalty: float,
        watermark: bool,
        chatbot,
        history,):
  
    client = Client(os.getenv("API_URL_TGTHR")) #get_client(model)
    # debug
    #print(f"^^client is - {client}")
    user_name, assistant_name = "<human>: ", "<bot>: "
    preprompt  = openchat_preprompt
    sep = '\n'

    history.append(inputs)

    past = []
    for data in chatbot:
        user_data, model_data = data

        if not user_data.startswith(user_name):
            user_data = user_name + user_data
        if not model_data.startswith("\n" + assistant_name):
            model_data = "\n" + assistant_name + model_data

        past.append(user_data + model_data.rstrip() + "\n")

    if not inputs.startswith(user_name):
        inputs = user_name + inputs

    total_inputs = preprompt + "".join(past) + inputs + "\n" + assistant_name.rstrip()
    # truncate total_inputs
    #total_inputs = total_inputs[-1000:]

    partial_words = ""

    for i, response in enumerate(client.generate_stream(
            total_inputs,
            top_p=top_p,
            top_k=top_k,
            repetition_penalty=repetition_penalty,
            watermark=watermark,
            temperature=temperature,
            max_new_tokens=500,
            stop_sequences=[user_name.rstrip(), assistant_name.rstrip()],
    )):
        if response.token.special:
            continue

        partial_words = partial_words + response.token.text
        if partial_words.endswith(user_name.rstrip()):
            partial_words = partial_words.rstrip(user_name.rstrip())
        if partial_words.endswith(assistant_name.rstrip()):
            partial_words = partial_words.rstrip(assistant_name.rstrip())

        if i == 0:
            history.append(" " + partial_words)
        else:
            history[-1] = partial_words

        chat = [
            (history[i].strip(), history[i + 1].strip()) for i in range(0, len(history) - 1, 2)
        ]
        yield chat, history

# Define function to generate model predictions and update the history
def predict_glm(input, history=[]):
    response, history = model_glm.chat(tokenizer_glm, input, history)
    # translate Chinese to English
    history = [(query, translate_Chinese_English(response)) for query, response in history]
    return history, history #[history] + updates

def translate_Chinese_English(chinese_text): 
    # translate Chinese to English
    tokenizer_chtoen.src_lang = "zh"
    encoded_zh = tokenizer_chtoen(chinese_text, return_tensors="pt")
    generated_tokens = model_chtoen.generate(**encoded_zh, forced_bos_token_id=tokenizer_chtoen.get_lang_id("en"))
    trans_eng_text = tokenizer_chtoen.batch_decode(generated_tokens, skip_special_tokens=True)
    return trans_eng_text[0]

"""    
with gr.Blocks() as demo:
    chatbot = gr.Chatbot()
    state = gr.State([])

    with gr.Row():
        txt = gr.Textbox(show_label=False, placeholder="Enter text and press enter").style(container=False)

    txt.submit(predict, [txt, state], [chatbot, state])

demo.launch(debug=True)
"""

def reset_textbox():
    return gr.update(value="")

def reset_chat(chatbot, state):
    # debug
    #print(f"^^chatbot value is - {chatbot}")
    #print(f"^^state value is - {state}")
    return None, []


#title = """<h1 align="center">🔥🔥Comparison: ChatGPT & OpenChatKit </h1><br><h3 align="center">🚀A Gradio Streaming Demo</h3><br>Official Demo: <a href="https://huggingface.co/spaces/togethercomputer/OpenChatKit">OpenChatKit feedback app</a>"""
title = """<h1 align="center">🔥🔥Comparison: ChatGPT & Open Sourced CHatGLM-6B </h1><br><h3 align="center">🚀A Gradio Chatbot Demo</h3>"""
description = """Language models can be conditioned to act like dialogue agents through a conversational prompt that typically takes the form:
```
User: <utterance>
Assistant: <utterance>
User: <utterance>
Assistant: <utterance>
...
```
In this app, you can explore the outputs of multiple LLMs when prompted in similar ways.
"""

with gr.Blocks(css="""#col_container {width: 1000px; margin-left: auto; margin-right: auto;}
                #chatgpt {height: 520px; overflow: auto;}
                #chatglm {height: 520px; overflow: auto;} """ ) as demo:
                #chattogether {height: 520px; overflow: auto;} """ ) as demo:
                #clear {width: 100px; height:50px; font-size:12px}""") as demo:
    gr.HTML(title)
    with gr.Row():
      with gr.Column(scale=14):
          with gr.Box():
              with gr.Row():
                  with gr.Column(scale=13):
                    openai_api_key = gr.Textbox(type='password', label="Enter your OpenAI API key here for ChatGPT")
                    inputs = gr.Textbox(placeholder="Hi there!", label="Type an input and press Enter ⤵️ " )
                  with gr.Column(scale=1):
                      b1 = gr.Button('🏃Run', elem_id = 'run').style(full_width=True)
                      b2 = gr.Button('🔄Clear up Chatbots!', elem_id = 'clear').style(full_width=True)
                  state_chatgpt = gr.State([])
                  #state_together = gr.State([])
                  state_glm = gr.State([])

          with gr.Box():
              with gr.Row():
                  chatbot_chatgpt = gr.Chatbot(elem_id="chatgpt", label='ChatGPT API - OPENAI')
                  #chatbot_together = gr.Chatbot(elem_id="chattogether", label='OpenChatKit - Text Generation')
                  chatbot_glm = gr.Chatbot(elem_id="chatglm", label='THUDM-ChatGLM6B')
          
      with gr.Column(scale=2, elem_id='parameters'):
          with gr.Box():
              gr.HTML("Parameters for #OpenCHAtKit", visible=False)
              top_p = gr.Slider(minimum=-0, maximum=1.0,value=0.25, step=0.05,interactive=True, label="Top-p", visible=False)
              temperature = gr.Slider(minimum=-0, maximum=5.0, value=0.6, step=0.1, interactive=True, label="Temperature",  visible=False)
              top_k = gr.Slider( minimum=1, maximum=50, value=50, step=1, interactive=True, label="Top-k", visible=False)
              repetition_penalty = gr.Slider( minimum=0.1, maximum=3.0, value=1.01, step=0.01, interactive=True, label="Repetition Penalty", visible=False)
              watermark = gr.Checkbox(value=True, label="Text watermarking",  visible=False)
              model = gr.CheckboxGroup(value="Rallio67/joi2_20B_instruct_alpha",
                  choices=["togethercomputer/GPT-NeoXT-Chat-Base-20B", "Rallio67/joi2_20B_instruct_alpha", "google/flan-t5-xxl", "google/flan-ul2", "bigscience/bloomz", "EleutherAI/gpt-neox-20b",],
                  label="Model",visible=False,) 
              temp_textbox_together = gr.Textbox(value=model.choices[0], visible=False) 

          with gr.Box():
              gr.HTML("Parameters for OpenAI's ChatGPT")
              top_p_chatgpt = gr.Slider( minimum=-0, maximum=1.0, value=1.0, step=0.05, interactive=True, label="Top-p",)
              temperature_chatgpt = gr.Slider( minimum=-0, maximum=5.0, value=1.0, step=0.1, interactive=True, label="Temperature",)
              chat_counter_chatgpt = gr.Number(value=0, visible=False, precision=0)

    inputs.submit(reset_textbox, [], [inputs])
    
    inputs.submit( predict_chatgpt, 
                [inputs, top_p_chatgpt, temperature_chatgpt, openai_api_key, chat_counter_chatgpt, chatbot_chatgpt, state_chatgpt], 
                [chatbot_chatgpt, state_chatgpt, chat_counter_chatgpt],)
    #inputs.submit( predict_together,
    #            [temp_textbox_together, inputs, top_p, temperature, top_k, repetition_penalty, watermark, chatbot_together, state_together, ],
    #            [chatbot_together, state_together],)
    inputs.submit( predict_glm,
                [inputs, state_glm, ],
                [chatbot_glm, state_glm],)
    b1.click( predict_chatgpt, 
                [inputs, top_p_chatgpt, temperature_chatgpt, openai_api_key, chat_counter_chatgpt, chatbot_chatgpt, state_chatgpt], 
                [chatbot_chatgpt, state_chatgpt, chat_counter_chatgpt],)
    #b1.click( predict_together,
    #            [temp_textbox_together, inputs, top_p, temperature, top_k, repetition_penalty, watermark, chatbot_together, state_together, ],
    #            [chatbot_together, state_together],)
    b1.click( predict_glm,
                [inputs, state_glm, ],
                [chatbot_glm, state_glm],)

    b2.click(reset_chat, [chatbot_chatgpt, state_chatgpt], [chatbot_chatgpt, state_chatgpt])
    #b2.click(reset_chat, [chatbot_together, state_together], [chatbot_together, state_together])
    b2.click(reset_chat, [chatbot_glm, state_glm], [chatbot_glm, state_glm])
    
    gr.HTML('''<center><a href="https://huggingface.co/spaces/ysharma/OpenChatKit_ChatGPT_Comparison?duplicate=true"><img src="https://bit.ly/3gLdBN6" alt="Duplicate Space"></a>Duplicate the Space and run securely with your OpenAI API Key</center>''')
    gr.Markdown(description)
    demo.queue(concurrency_count=16).launch(height= 2500, debug=True)