app.py

''' Import Modules '''
from transformers import T5Tokenizer, T5ForConditionalGeneration

from fastapi import FastAPI
from pydantic import BaseModel

''' Model '''
save_path = "t5-base"
tokenizer = T5Tokenizer.from_pretrained(save_path)
model = T5ForConditionalGeneration.from_pretrained(save_path)

''' Server '''
class Request(BaseModel):
    text: str
    maxlen: int

app = FastAPI()
@app.post("/")
def main(request: Request):
    task_prefix = "summarize: "
    input_ids = tokenizer.encode(task_prefix + request.text, return_tensors="pt", max_length=512, truncation=True)

    output_ids = model.generate(
        input_ids,
        max_length=request.maxlen,
        num_beams=3,
        early_stopping=True,
        no_repeat_ngram_size=5
    )

    result = tokenizer.decode(output_ids[0], skip_special_tokens=True)
    return {"output": result}