train.py

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

# Load a small pre-trained model and tokenizer
model_name = "distilgpt2"  # or choose another small model
model = AutoModelForCausalLM.from_pretrained(model_name)
tokenizer = AutoTokenizer.from_pretrained(model_name)

# Add a pad token (setting it to eos_token is one common approach for GPT-based models)
tokenizer.pad_token = tokenizer.eos_token  # Or you can choose to add a new pad token, e.g., '[PAD]'

# Load the dataset (Make sure data.json is in the correct location)
train_data = load_dataset("json", data_files={"train": "data.json"})

# Preprocess the dataset
def preprocess_function(examples):
    inputs = examples["input"]
    outputs = examples["output"]
    model_inputs = tokenizer(inputs, max_length=512, truncation=True, padding="max_length")
    labels = tokenizer(outputs, max_length=512, truncation=True, padding="max_length")
    model_inputs["labels"] = labels["input_ids"]
    return model_inputs

# Preprocess the train dataset using the map function
train_dataset = train_data["train"].map(preprocess_function, batched=True)

# Define training arguments
training_args = TrainingArguments(
    output_dir="./results",  
    num_train_epochs=3,  
    per_device_train_batch_size=4,  
    logging_dir="./logs",  
)

# Initialize Trainer
trainer = Trainer(
    model=model,
    args=training_args,
    train_dataset=train_dataset,
)

# Train the model
trainer.train()

# Save the fine-tuned model
model.save_pretrained("./lockin_model")
tokenizer.save_pretrained("./lockin_model")