app.py

import gradio as gr
from transformers import GPTNeoForCausalLM, GPT2TokenizerFast

def generate_text(prompt, max_length, temperature, top_p, repetition_penalty):
    tokenizer = GPT2TokenizerFast.from_pretrained("JakeTurner616/Adonalsium-gpt-neo-1.3B")
    model = GPTNeoForCausalLM.from_pretrained("JakeTurner616/Adonalsium-gpt-neo-1.3B")

    # Set pad token
    if tokenizer.pad_token is None:
        tokenizer.pad_token = tokenizer.eos_token

    inputs = tokenizer(prompt, return_tensors="pt", padding=True, truncation=True)
    # Ensure that pad_token_id is set for the model
    model.config.pad_token_id = tokenizer.pad_token_id

    outputs = model.generate(
        input_ids=inputs["input_ids"],
        attention_mask=inputs["attention_mask"],  # Explicitly pass attention mask
        max_length=max_length,
        temperature=temperature,
        top_p=top_p,
        repetition_penalty=repetition_penalty,
        no_repeat_ngram_size=2,  # Correctly specify the no_repeat_ngram_size
        do_sample=True,  # Ensure sampling is enabled for temperature and top_p to take effect
        pad_token_id=tokenizer.pad_token_id
    )

    generated_texts = [tokenizer.decode(output, skip_special_tokens=True) for output in outputs]
    return generated_texts

# Define your interface with the correct output component
iface = gr.Interface(
    fn=generate_text,
    inputs=[
        gr.Textbox(lines=2, label="Input Prompt"),
        gr.Slider(minimum=10, maximum=300, step=10, value=40, label="Max Length"),
        gr.Slider(minimum=0.0, maximum=1.0, step=0.01, value=0.70, label="Temperature"),
        gr.Slider(minimum=0.0, maximum=1.0, step=0.01, value=0.5, label="Top P"),
        gr.Slider(minimum=1.0, maximum=2.0, step=0.1, value=1.2, label="Repetition Penalty")
    ],
    outputs="text",
    title="Cosmere Series Text Generator",
    description="Adjust the sliders to control text generation parameters."
)

iface.launch()