from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
import gradio as gr
import spaces
import torch


tokenizer = AutoTokenizer.from_pretrained("merve/chatgpt-prompt-generator-v12")
model = AutoModelForSeq2SeqLM.from_pretrained("merve/chatgpt-prompt-generator-v12", from_tf=True)

device = "cuda:0" if torch.cuda.is_available() else "cpu"
model.to(device)

@spaces.GPU
def generate(prompt):
    batch = tokenizer(prompt, return_tensors="pt")
    batch.to(device)
    generated_ids = model.generate(batch["input_ids"], max_new_tokens=150)
    output = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)
    return output[0]

input_component = gr.Textbox(label = "Input a persona, e.g. photographer", value = "photographer")
output_component = gr.Textbox(label = "Prompt")
examples = [["photographer"], ["developer"]]
description = "This app generates ChatGPT prompts, it's based on a BART model trained on [this dataset](https://huggingface.co/datasets/fka/awesome-chatgpt-prompts). 📓 Simply enter a persona that you want the prompt to be generated based on. 🧙🏻🧑🏻‍🚀🧑🏻‍🎨🧑🏻‍🔬🧑🏻‍💻🧑🏼‍🏫🧑🏽‍🌾"
gr.Interface(generate, inputs = input_component, outputs=output_component, examples=examples, title = "👨🏻‍🎤 ChatGPT Prompt Generator v12 👨🏻‍🎤", description=description).launch()