model_gpt_train.py

# Install required packages before running:
# pip install torch transformers datasets evaluate safetensors

import torch
from datasets import load_dataset
from transformers import AutoTokenizer, AutoModelForSequenceClassification, TrainingArguments, Trainer
import evaluate

# 1️⃣ Load Dataset
dataset = load_dataset("imdb")

# 2️⃣ Tokenizer
model_name = "distilbert-base-uncased"
tokenizer = AutoTokenizer.from_pretrained(model_name)

def tokenize_fn(batch):
    return tokenizer(batch["text"], truncation=True, padding="max_length", max_length=256)

tokenized_datasets = dataset.map(tokenize_fn, batched=True)

# 3️⃣ Load Model
model = AutoModelForSequenceClassification.from_pretrained(model_name, num_labels=2)

# 4️⃣ Evaluation metric
accuracy = evaluate.load("accuracy")

def compute_metrics(eval_pred):
    logits, labels = eval_pred
    predictions = torch.argmax(torch.tensor(logits), dim=-1)
    return accuracy.compute(predictions=predictions, references=labels)

# 5️⃣ Training Arguments
training_args = TrainingArguments(
    output_dir="./results",
    evaluation_strategy="epoch",
    save_strategy="epoch",
    logging_dir="./logs",
    learning_rate=2e-5,
    per_device_train_batch_size=8,
    per_device_eval_batch_size=8,
    num_train_epochs=1,
    weight_decay=0.01,
    push_to_hub=False,
    save_safetensors=True  # 🔹 Save model in safetensors format
)

# 6️⃣ Trainer
trainer = Trainer(
    model=model,
    args=training_args,
    train_dataset=tokenized_datasets["train"].shuffle(seed=42).select(range(2000)),  # small subset
    eval_dataset=tokenized_datasets["test"].shuffle(seed=42).select(range(500)),     # small subset
    tokenizer=tokenizer,
    compute_metrics=compute_metrics
)

# 7️⃣ Train Model
trainer.train()

# 8️⃣ Save final model in safetensors
trainer.save_model("./final_safetensors_model")  # saves as model.safetensors
tokenizer.save_pretrained("./final_safetensors_model")

print("✅ Training complete. Model saved in safetensors format at './final_safetensors_model'")