import gradio as gr
from transformers import pipeline, AutoTokenizer
import torch
import os
import spaces
@spaces.GPU
def load_model(model_name):
    return pipeline("text-generation", model=model_name, device_map="auto",  trust_remote_code=True)
@spaces.GPU()
def generate(
    model_name,
    user_input,
    temperature=0.7,
    top_p=0.95,
    top_k=50,
    max_new_tokens=512,
):
    pipe = load_model(model_name)

    # Set tokenize correctly. Otherwise ticking the box breaks it.
    if model_name == "LeroyDyer/Mixtral_AI_Cyber_1.0":
        prompt = user_input
    else:
        prompt = f"<|im_start|>user\n{user_input}<|im_end|>\n<|im_start|>assistant\n"
    outputs = pipe(prompt, max_new_tokens=max_new_tokens, do_sample=True,
                   temperature=temperature, top_k=top_k, top_p=top_p, repetition_penalty=1.1)
    return outputs[0]["generated_text"]

model_choices = ["LeroyDyer/Mixtral_Instruct","LeroyDyer/Mixtral_Chat","LeroyDyer/Mixtral_Chat_X","LeroyDyer/Mixtral_Chat_X_128k",
                 "LeroyDyer/Mixtral_AI_Instruct_X","LeroyDyer/Mixtral_AI_Thinker",
                 "LeroyDyer/Mixtral_AI_128k_BioMedical","Mixtral_BioMedical","LeroyDyer/Mixtral_Samantha",
                 "LeroyDyer/Mixtral_AI_Base_128k","LeroyDyer/Mixtral_Base","LeroyDyer/Mixtral_AI_1.0", 
                 "LeroyDyer/Mixtral_AI_Vision_128K","LeroyDyer/Mixtral_AI_Base","LeroyDyer/Mixtral_AI_Medic_Base",
                 "LeroyDyer/Mixtral_AI_Vision_128K_X","LeroyDyer/Mixtral_AI_Vision_V1_128",
                "LeroyDyer/Mixtral_AI_Cyber", "LeroyDyer/Mixtral_AI_Cyber_1.0","LeroyDyer/Mixtral_AI_Cyber_2.0",
                 ]
# What at the best options? 
g = gr.Interface(
    fn=generate,
    inputs=[
        gr.components.Dropdown(choices=model_choices, label="Model", value=model_choices[0], interactive=True),
        gr.components.Textbox(lines=2, label="Prompt", value="Write me a Python program that calculates the factorial of a given number."),
        gr.components.Slider(minimum=0, maximum=1, value=0.4, label="Temperature"),
        gr.components.Slider(minimum=0, maximum=1, value=0.95, label="Top p"),
        gr.components.Slider(minimum=0, maximum=100, step=1, value=50, label="Top k"),
        gr.components.Slider(minimum=1, maximum=2048, step=1, value=1024, label="Max tokens"),  
    ],
    outputs=[gr.Textbox(lines=10, label="Output")],
    title="SpydazWeb AI (LeroyDyer) Language Models",
    description="Each has thier own personality's ",
    concurrency_limit=1
)

g.launch(max_threads=4)