from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
import gradio as gr

tokenizer = AutoTokenizer.from_pretrained("merve/chatgpt-prompt-generator-v12")
model = AutoModelForSeq2SeqLM.from_pretrained("merve/chatgpt-prompt-generator-v12", from_tf=True)
#
tokenizer2 = AutoTokenizer.from_pretrained("Kaludi/chatgpt-gpt4-prompts-bart-large-cnn-samsum")
model2 = AutoModelForSeq2SeqLM.from_pretrained("Kaludi/chatgpt-gpt4-prompts-bart-large-cnn-samsum", from_tf=True)

def generate(prompt):

    batch = tokenizer(prompt, return_tensors="pt")
    generated_ids = model.generate(batch["input_ids"], max_new_tokens=150)
    output = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)
    return output[0]
    
def generate2(prompt, max_new_tokens):
    batch = tokenizer2(prompt, return_tensors="pt")
    generated_ids = model2.generate(batch["input_ids"], max_new_tokens=150)
    output = tokenizer2.batch_decode(generated_ids, skip_special_tokens=True)
    return output[0]
def generate2_test(prompt):
    batch = tokenizer2(prompt, return_tensors="pt")
    generated_ids = model2.generate(batch["input_ids"], max_new_tokens=150)
    output = tokenizer2.batch_decode(generated_ids, skip_special_tokens=True)
    return output[0]
    
def generate_prompt(type, prompt, max_new_tokens):
    if type==1:
        return generate(prompt)
    elif type==2:
        return generate2(prompt, max_new_tokens)
#
input_component = gr.Textbox(label = "Input a persona, e.g. photographer", value = "photographer")
output_component = gr.Textbox(label = "Prompt")
examples = [["photographer"], ["developer"]]
description = ""
gr.Interface(generate2_test, inputs = input_component, outputs=output_component, examples=examples, title = "👨🏻‍🎤 ChatGPT Prompt Generator v12 👨🏻‍🎤", description=description).launch()