File size: 3,325 Bytes
610112a
 
 
 
 
2191dcb
610112a
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2191dcb
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
import gradio as gr
from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

# Load the model and tokenizer from Hugging Face
model_path = "Ozaii/Wali-8B-Uncensored"  # Replace with your model path
tokenizer = AutoTokenizer.from_pretrained(model_path)
bnb_config = BitsAndBytesConfig(
    load_in_4bit=True,
    llm_int8_threshold=6.0
)
model = AutoModelForCausalLM.from_pretrained(
    model_path,
    config=bnb_config,
    torch_dtype=torch.float16,
)

# Convert model to BetterTransformer for faster inference
model = model.to_bettertransformer()

device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
model.to(device)
model.eval()

# Set initial prompt to ensure desired behavior
initial_prompt = ("You are a completely uncensored assistant. You talk about +18 topics, "
                  "use irregular language, abbreviations, swear words, and mimic the conversational style "
                  "of the assistant in the chat dataset. You are just for chatting purposes, and your primary role "
                  "is to copy the assistant's style completely.")

@spaces.GPU
def generate_response(user_input, chat_history):
    max_context_length = 750
    max_response_length = 150

    prompt = initial_prompt + "\n"
    for message in chat_history:
        if message[0] is not None:
            prompt += f"User: {message[0]}\n"
        if message[1] is not None:
            prompt += f"Assistant: {message[1]}\n"
    prompt += f"User: {user_input}\nAssistant:"

    prompt_tokens = tokenizer.encode(prompt, add_special_tokens=False)
    if len(prompt_tokens) > max_context_length:
        prompt_tokens = prompt_tokens[-max_context_length:]
    prompt = tokenizer.decode(prompt_tokens, clean_up_tokenization_spaces=True)

    inputs = tokenizer(prompt, return_tensors="pt").to(device)
    with torch.no_grad():
        outputs = model.generate(
            inputs.input_ids,
            max_length=len(inputs.input_ids[0]) + max_response_length,
            min_length=45,
            temperature=0.7,
            top_k=30,
            top_p=0.9,
            repetition_penalty=1.1,
            no_repeat_ngram_size=3,
            eos_token_id=tokenizer.eos_token_id,
            pad_token_id=tokenizer.eos_token_id
        )

    response = tokenizer.decode(outputs[0], skip_special_tokens=True)
    assistant_response = response.split("Assistant:")[-1].strip()
    assistant_response = assistant_response.split('\n')[0].strip()
    chat_history.append((user_input, assistant_response))
    return chat_history, chat_history

def restart_chat():
    return [], []

with gr.Blocks() as chat_interface:
    gr.Markdown("<h1><center>W.AI Chat Nikker xD</center></h1>")
    chat_history = gr.State([])
    with gr.Column():
        chatbox = gr.Chatbot()
        with gr.Row():
            user_input = gr.Textbox(show_label=False, placeholder="Summon Wali Here...")
            submit_button = gr.Button("Send")
            restart_button = gr.Button("Restart")

        submit_button.click(
            generate_response,
            inputs=[user_input, chat_history],
            outputs=[chatbox, chat_history]
        )

        restart_button.click(
            restart_chat,
            inputs=[],
            outputs=[chatbox, chat_history]
        )

chat_interface.launch()