from datasets import load_dataset
from transformers import T5Tokenizer, T5ForConditionalGeneration, Trainer, TrainingArguments

# Initialize the tokenizer
model_name = "t5-small"
tokenizer = T5Tokenizer.from_pretrained(model_name)

# Load the dataset with the specific configuration
dataset = load_dataset("wiki_atomic_edits", "english_insertions", trust_remote_code=True)

# Inspect the dataset splits
print(dataset.keys())  # Print available dataset splits

# Preprocessing Function
def preprocess_function(examples):
    inputs = examples["base_sentence"]
    targets = examples["edited_sentence"]
    model_inputs = tokenizer(inputs, max_length=128, truncation=True, padding="max_length")
    labels = tokenizer(targets, max_length=128, truncation=True, padding="max_length")
    labels["input_ids"] = [
        [(label if label != tokenizer.pad_token_id else -100) for label in labels_example]
        for labels_example in labels["input_ids"]
    ]
    model_inputs["labels"] = labels["input_ids"]
    return model_inputs

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

# Initialize the model
model = T5ForConditionalGeneration.from_pretrained(model_name)

# Set up training arguments
training_args = TrainingArguments(
    output_dir="./results",
    evaluation_strategy="epoch",  # Updated from eval_strategy to evaluation_strategy
    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 Trainer
trainer = Trainer(
    model=model,
    args=training_args,
    train_dataset=tokenized_datasets["train"],
    eval_dataset=tokenized_datasets.get("validation")  # Use .get() to avoid KeyError
)

# Start training
trainer.train()