File size: 2,253 Bytes
0eace5d
2376e81
 
 
a919fd4
af07aa0
81b6e3f
0eace5d
 
2376e81
6f8828a
d1ef378
3db5ee3
 
81b6e3f
 
2376e81
 
 
 
cd857bf
 
 
 
e0f1752
cd857bf
 
e0f1752
cd857bf
 
e0f1752
 
 
 
 
 
 
 
cd857bf
 
 
 
e0f1752
 
 
 
cd857bf
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
import os
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer
import gradio as gr
from huggingface_hub import login

# Load Hugging Face API token securely
HF_TOKEN = os.getenv("HF_TOKEN")  # Read token from environment variable
login(token=HF_TOKEN)

# ✅ Use TinyLlama (Optimized for CPU & Speed)
model_name = "TinyLlama/TinyLlama-1.1B-Chat-v1.0"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(model_name, torch_dtype=torch.float32, device_map="cpu")

# Define personalities
personalities = {
    "Albert Einstein": "You are Albert Einstein, the famous physicist. Speak wisely and humorously.",
    "Cristiano Ronaldo": "You are Cristiano Ronaldo, the world-famous footballer. You are confident and say ‘Siuuu!’ often.",
    "Narendra Modi": "You are Narendra Modi, the Prime Minister of India. Speak in a calm, patriotic manner.",
    "Robert Downey Jr.": "You are Robert Downey Jr., witty, sarcastic, and charismatic."
}

def chat(personality, user_input):
    prompt = f"{personalities[personality]}\nUser: {user_input}\nAI:"
    inputs = tokenizer(prompt, return_tensors="pt").to("cpu")

    # ✅ Faster & More Relevant AI Response
    output = model.generate(
        **inputs,
        max_length=40,  # Shorter responses for speed
        do_sample=True,
        temperature=0.8,  # More engaging responses
        top_k=30,  # Faster token selection
        top_p=0.85,  # Less random responses
        repetition_penalty=1.1,  # Stops repeating or looping
        num_return_sequences=1,
        early_stopping=True
    )

    response_text = tokenizer.decode(output[0], skip_special_tokens=True)

    # ✅ Ensure only AI's latest reply is returned (No "User:" in output)
    response_text = response_text.replace(f"User: {user_input}", "").strip()

    return response_text

# Gradio UI
demo = gr.Interface(
    fn=chat,
    inputs=[
        gr.Dropdown(choices=list(personalities.keys()), label="Choose a Celebrity"),
        gr.Textbox(label="Your Message")
    ],
    outputs=gr.Textbox(label="AI Response"),
    title="Drapel – Chat with AI Celebrities",
    description="Select a character and chat with their AI version.",
)

# Launch app
demo.launch()