from datasets import load_dataset
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM, Trainer, TrainingArguments

# Load dataset
dataset = load_dataset("Abdelkareem/wikihow-arabic-summarization")

# Load the model and tokenizer
model_name = "UBC-NLP/AraT5v2-base-1024"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForSeq2SeqLM.from_pretrained(model_name)

# Preprocessing function to tokenize the dataset
def preprocess_function(examples):
    inputs = examples["article"]
    targets = examples["summarize"]
    model_inputs = tokenizer(inputs, max_length=1024, truncation=True)
    labels = tokenizer(targets, max_length=150, truncation=True)
    model_inputs["labels"] = labels["input_ids"]
    return model_inputs

# Apply preprocessing to the dataset
tokenized_datasets = dataset.map(preprocess_function, batched=True)

# Define training arguments
training_args = TrainingArguments(
    output_dir="./results", 
    evaluation_strategy="epoch", 
    learning_rate=2e-5, 
    per_device_train_batch_size=4, 
    per_device_eval_batch_size=4, 
    num_train_epochs=3,
    weight_decay=0.01,
    logging_dir="./logs"
)

# Initialize the Trainer
trainer = Trainer(
    model=model, 
    args=training_args, 
    train_dataset=tokenized_datasets["train"], 
    eval_dataset=tokenized_datasets["validation"]
)

# Start the training process
trainer.train()