main.py

from fastapi.staticfiles import StaticFiles
from fastapi.responses import FileResponse
from pydantic import BaseModel
from fastapi import FastAPI

import os

from transformers import GPT2LMHeadModel, GPT2Tokenizer
import torch

app = FastAPI()

class req(BaseModel):
  prompt: str
  length: int

@app.get("/")
def read_root():
  return FileResponse(path="templates/index.html", media_type="text/html")

@app.post("/api")
def read_root(data: req):
  name = "microsoft/DialoGPT-medium"
  # microsoft/DialoGPT-small
  # microsoft/DialoGPT-medium
  # microsoft/DialoGPT-large
  
  # Load the Hugging Face GPT-2 model and tokenizer
  model = GPT2LMHeadModel.from_pretrained(name)
  tokenizer = GPT2Tokenizer.from_pretrained(name)

  print("Prompt:", data.prompt)
  print("Length:", data.length)

  # Using CUDA for an optimal experience
  device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
  model = model.to(device)
  
  input_text = data.prompt
  
  # Tokenize the input text
  input_ids = tokenizer.encode(input_text, return_tensors="pt")
  
  # Generate output using the model
  output_ids = model.generate(input_ids, max_length=length, num_beams=5, no_repeat_ngram_size=2)
  generated_text = tokenizer.decode(output_ids[0], skip_special_tokens=True)
  
  answer_data = { "answer": generated_text }
  print("Answered with:", answer_data)
  
  return answer_data