File size: 2,799 Bytes
2f3b32c
39f7f02
 
 
2f3b32c
 
 
39f7f02
2f3b32c
39f7f02
 
 
2f3b32c
39f7f02
ab0e126
39f7f02
ab0e126
39f7f02
 
 
 
 
2f3b32c
61866ca
2f3b32c
73c13c3
ab0e126
39f7f02
2f3b32c
39f7f02
 
2f3b32c
 
 
 
 
39f7f02
 
ab0e126
39f7f02
ab0e126
 
b986f28
 
ab0e126
39f7f02
ab0e126
39f7f02
 
 
394ca4c
ab0e126
 
 
 
 
39f7f02
 
 
 
 
2f3b32c
 
39f7f02
07e84d7
 
39f7f02
 
 
2606bca
eae3b08
 
2606bca
 
eae3b08
2f3b32c
 
97e7837
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
import gradio as gr
from transformers import pipeline
import requests
import json
import edge_tts
import asyncio
import tempfile
import os
from huggingface_hub import InferenceClient
import re
import time
from streaming_stt_nemo import Model

default_lang = "en"

engines = { default_lang: Model(default_lang) }

def transcribe(audio):
    lang = "en"
    model = engines[lang]
    text = model.stt_file(audio)[0]
    return text

client1 = InferenceClient("mistralai/Mistral-7B-Instruct-v0.3")

system_instructions1 = "<s>[SYSTEM] Answer as Real Jarvis JARVIS, Made by 'Tony Stark', Keep conversation very short, clear, friendly and concise. The text provided is a request for a specific type of response from you, the virtual assistant. The request asks you to provide friendly responses as if You are the character Jarvis, made by 'Tony Stark.' The expectation is that I will avoid introductions and start answering the query directly, elaborating on all aspects of the request. As an AI-powered assistant, your task is to generate responses that appear as if they are created by the user and not an AI-powered assistant. [USER]"

def model(text):
    generate_kwargs = dict(
        temperature=0.7,
        max_new_tokens=512,
        top_p=0.95,
        repetition_penalty=1,
        do_sample=True,
        seed=42,
    )
    
    formatted_prompt = system_instructions1 + text + "[JARVIS]"
    stream = client1.text_generation(
        formatted_prompt, **generate_kwargs, stream=True, details=True, return_full_text=False)
    output = ""
    for response in stream:
        if not response.token.text == "</s>":
            output += response.token.text

    return output

async def respond(audio):
    user = transcribe(audio)
    reply = model(user)
    communicate = edge_tts.Communicate(reply)
    with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as tmp_file:
        tmp_path = tmp_file.name
        await communicate.save(tmp_path)
    yield tmp_path

DESCRIPTION = """ # <center><b>JARVIS⚡</b></center>
        ### <center>A personal Assistant of Tony Stark for YOU
        ### <center>Voice Chat with your personal Assistant</center>
        """

with gr.Blocks(css="style.css") as demo:    
    gr.Markdown(DESCRIPTION)
    with gr.Row():
        input = gr.Audio(label="User", sources="microphone", type="filepath", waveform_options=False)
        output = gr.Audio(label="AI", type="filepath",
                        interactive=False,
                        autoplay=True,
                        elem_classes="audio")
        gr.Interface(
            batch=True,
            max_batch_size=10, 
            fn=respond, 
            inputs=[input],
            outputs=[output], live=False)  

if __name__ == "__main__":
    demo.queue(max_size=200).launch()