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

# Load the dataset
dataset = load_dataset("json", data_files="dataset.jsonl")

# Load the model and tokenizer
model_name = "Salesforce/codegen-2B-multi"
model = AutoModelForCausalLM.from_pretrained(model_name)
tokenizer = AutoTokenizer.from_pretrained(model_name)

# Tokenize the dataset
def tokenize_function(examples):
    return tokenizer(examples["input"], text_target=examples["output"], truncation=True)

tokenized_dataset = dataset.map(tokenize_function, batched=True)

# Define training arguments
training_args = TrainingArguments(
    output_dir="./results",
    overwrite_output_dir=True,
    evaluation_strategy="epoch",
    learning_rate=5e-5,
    per_device_train_batch_size=4,
    num_train_epochs=3,
    save_strategy="epoch",
    logging_dir="./logs",
)

# Train the model
trainer = Trainer(
    model=model,
    args=training_args,
    train_dataset=tokenized_dataset["train"],
    eval_dataset=tokenized_dataset["train"],
)

trainer.train()
trainer.save_model("./fine_tuned_codegen")
tokenizer.save_pretrained("./fine_tuned_codegen")
print("Training complete. Model saved.")