from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
from typing import List
 
tokenizer = AutoTokenizer.from_pretrained("yeye776/OndeviceAI-T5-base")
model = AutoModelForSeq2SeqLM.from_pretrained("yeye776/OndeviceAI-T5-base")
 
prompt = "분류 및 인식해줘 :"
def prepare_input(question: str):
    inputs = f"{prompt} {question}"
    input_ids = tokenizer(inputs, max_length=700, return_tensors="pt").input_ids
    return input_ids
 
def inference(question: str) -> str:
    input_data = prepare_input(question=question)
    input_data = input_data.to(model.device)
    outputs = model.generate(inputs=input_data, num_beams=10, top_k=10, max_length=1024)
 
    result = tokenizer.decode(token_ids=outputs[0], skip_special_tokens=True)
 
    return result
 
inference("안방 조명 켜줘")

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 0.0007
train_batch_size: 4
eval_batch_size: 4
seed: 42
gradient_accumulation_steps: 8
total_train_batch_size: 32
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: cosine
lr_scheduler_warmup_ratio: 0.06
num_epochs: 10

Training results

Framework versions

Transformers 4.36.2
Pytorch 2.1.2+cu121
Datasets 2.16.1
Tokenizers 0.15.0