import torch
import pandas as pd
from transformers import pipeline, AutoTokenizer, AutoModelForCausalLM, AutoModelForSequenceClassification, TrainingArguments 
import gradio as gr
from gradio.mix import Parallel, Series
#import torch.nn.functional as F

from datasets import load_dataset
dataset = load_dataset("bananabot/engMollywoodSummaries")
dataset

device = "cuda" if torch.cuda.is_available() else "cpu"

model_name = "EleutherAI/gpt-neo-125M"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(model_name).to(device)

max_length=123
input_txt = "This malayalam movie is about"
n_steps = 8

input_ids = tokenizer(input_txt, return_tensors="pt")["input_ids"].to(device)
output = model.generate(input_ids,  max_length=max_length, num_beams=5, do_sample=True, no_repeat_ngram_size=2, temperature=1.37, top_k=69, top_p=0.96)
print(tokenizer.decode(output[0]))

generator = gr.Interface.load("models/EleutherAI/gpt-neo-125M")
translator = gr.Interface.load("models/Helsinki-NLP/opus-mt-en-ml")
gr.Series(generator, translator, inputs=gr.inputs.Textbox(lines=13, label="Input Text")).launch()  # this demo generates text, then translates it to Malayalam, and outputs the final result.