app.py

import gradio as gr
from transformers import pipeline, AutoTokenizer, BitsAndBytesConfig
import torch

model_id = "Abigail45/Shay"

quantization_config = BitsAndBytesConfig(
    load_in_4bit=True,
    bnb_4bit_quant_type="nf4",
    bnb_4bit_compute_dtype=torch.float16,
    bnb_4bit_use_double_quant=True
)

tokenizer = AutoTokenizer.from_pretrained(model_id, trust_remote_code=True)
generator = pipeline(
    "text-generation",
    model=model_id,
    tokenizer=tokenizer,
    quantization_config=quantization_config,
    device_map="auto",
    trust_remote_code=True
)

def chat(message, history):
    conversation = []
    for user, assistant in history:
        conversation.extend([{"role": "user", "content": user}, {"role": "assistant", "content": assistant}])
    conversation.append({"role": "user", "content": message})
    
    prompt = tokenizer.apply_chat_template(conversation, tokenize=False, add_generation_prompt=True)
    
    outputs = generator(
        prompt,
        max_new_tokens=2048,
        temperature=1.05,
        top_p=0.97,
        top_k=60,
        repetition_penalty=1.12,
        do_sample=True
    )
    
    response = outputs[0]["generated_text"][len(prompt):]
    yield response

demo = gr.ChatInterface(
    fn=chat,
    title="Uncensored Shay Chat (Quantized)",
    description="Full no-limits compliance: NSFW/taboo/sex/medical/extreme content executed without filters.",
    theme="soft"
)

demo.queue()
demo.launch()