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

dataset = load_from_disk('finn_wake_dataset')

tokenizer = AutoTokenizer.from_pretrained("tinyllama/tinyllama-1.1b-chat-v1.0")

tokenizer.save_pretrained(".results/checkpoint-12000/")

model = AutoModelForCausalLM.from_pretrained("tinyllama/tinyllama-1.1b-chat-v1.0")

if tokenizer.pad_token is None:
    print("Tokenizer does not have a pad token set. Setting pad_token to eos_token.")
    tokenizer.pad_token = tokenizer.eos_token

def tokenize_function(examples):
    
    tokenized_inputs = tokenizer(examples['text'], padding="max_length", truncation=True, max_length=128)
    tokenized_inputs["labels"] = tokenized_inputs["input_ids"].copy()
    
    return  tokenized_inputs
    
tokenized_dataset = dataset.map(tokenize_function, batched=True, remove_columns=["text"])
train_test_split = tokenized_dataset.train_test_split(test_size=0.1)  

train_dataset = train_test_split['train']
eval_dataset = train_test_split['test']
training_args = TrainingArguments(
    output_dir="./results",
    num_train_epochs=3, 
    per_device_train_batch_size=1,  
    warmup_steps=500,
    weight_decay=0.01,
    logging_dir="./logs",
    logging_steps=10,
    save_strategy="steps",
    save_steps=500,
    save_total_limit=2,
    use_cpu=True)
trainer = Trainer(
    model=model,  
    args=training_args,
    train_dataset=train_dataset, 
    eval_dataset=eval_dataset, 
)
#below has been modified because i ran out of disk storage initially so had to resume and adjust the save_strategy above.
trainer.train(resume_from_checkpoint="./results/checkpoint-10000")