app.py

import gradio as gr
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer

# Load model and tokenizer - specify the correct model class
tokenizer = AutoTokenizer.from_pretrained("anthracite-org/magnum-v4-12b-gguf")
model = AutoModelForCausalLM.from_pretrained("anthracite-org/magnum-v4-12b-gguf", torch_dtype=torch.float16, device_map="auto")  # Adjust dtype and device_map as needed


def generate_text(prompt, max_length=50, num_return_sequences=1, temperature=0.7, top_p=0.95, top_k=50):
    try:
        inputs = tokenizer(prompt, return_tensors="pt").to(model.device) # Ensure inputs are on the same device as the model

        outputs = model.generate(
            **inputs,
            max_new_tokens=max_length,  # Use max_new_tokens instead of max_length with Transformers >= 4.30
            do_sample=True,
            temperature=temperature,
            top_p=top_p,
            top_k=top_k,
            num_return_sequences=num_return_sequences
        )

        generated_text = tokenizer.batch_decode(outputs, skip_special_tokens=True)
        return "\n\n".join(generated_text)

    except Exception as e:  # Handle potential errors
        return f"Error: {e}"


iface = gr.Interface(
    fn=generate_text,
    inputs=[
        gr.Textbox(lines=2, placeholder="Enter a prompt...", label="Prompt"),
        gr.Slider(50, 200, value=100, label="Max Length"),
        gr.Slider(0, 1, value=0.7, label="Temperature"),
        gr.Slider(0, 1, value=0.95, label="Top p"),
        gr.Slider(0, 50, value=50, step=1, label="Top k"),

    ],
    outputs="text",
    title="Text Generation with magnum-v4-12b-gguf",
    description="Generate text using the magnum model. Please be patient, as generation can take time."
)

iface.launch(share=True) # Consider adding share=True for others to try