File size: 3,424 Bytes
6229408 2f3b32c 39f7f02 2f3b32c 39f7f02 207c35f c640ef1 2f3b32c 39f7f02 ab0e126 6229408 ab0e126 39f7f02 f4dad5f 39f7f02 2f3b32c 679f566 9379874 679f566 2f3b32c 390611c ab0e126 9379874 6229408 679f566 62e14ef 9379874 62e14ef 2f3b32c 5a78105 c68088d 9379874 39f7f02 679f566 39f7f02 ab0e126 b986f28 39f7f02 ab0e126 d49facf 39f7f02 679f566 394ca4c ab0e126 6229408 e5fdf4f 62e14ef e5fdf4f 6229408 9379874 6229408 9379874 6229408 9379874 6229408 |
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 |
import gradio as gr
import edge_tts
import asyncio
import tempfile
import os
from huggingface_hub import InferenceClient
import re
from streaming_stt_nemo import Model
import torch
import random
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
HF_TOKEN = os.environ.get("HF_TOKEN", None)
def client_fn(model):
if "Mixtral" in model:
return InferenceClient("mistralai/Mixtral-8x7B-Instruct-v0.1")
elif "Llama" in model:
return InferenceClient("meta-llama/Meta-Llama-3-8B-Instruct")
elif "Mistral" in model:
return InferenceClient("mistralai/Mistral-7B-Instruct-v0.2")
elif "Phi" in model:
return InferenceClient("microsoft/Phi-3-mini-4k-instruct")
else:
return InferenceClient("microsoft/Phi-3-mini-4k-instruct")
def randomize_seed_fn(seed: int) -> int:
seed = random.randint(0, 999999)
return seed
system_instructions1 = """
[SYSTEM] Answer as Real Jarvis JARVIS, Made by 'Tony Stark.'
Keep conversation friendly, short, clear, and concise.
Avoid unnecessary introductions and answer the user's questions directly.
Respond in a normal, conversational manner while being friendly and helpful.
[USER]
"""
def models(text, model="Mixtral 8x7B", seed=42):
seed = int(randomize_seed_fn(seed))
generator = torch.Generator().manual_seed(seed)
client = client_fn(model)
generate_kwargs = dict(
max_new_tokens=300,
seed=seed
)
formatted_prompt = system_instructions1 + text + "[JARVIS]"
stream = client.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, model, seed):
user = transcribe(audio)
reply = models(user, model, seed)
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():
select = gr.Dropdown([ 'Mixtral 8x7B',
'Llama 3 8B',
'Mistral 7B v0.3',
'Phi 3 mini',
],
value="Mistral 7B v0.3",
label="Model"
)
seed = gr.Slider(
label="Seed",
minimum=0,
maximum=999999,
step=1,
value=0,
visible=False
)
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, select, seed],
outputs=[output], live=True)
if __name__ == "__main__":
demo.queue(max_size=200).launch() |