train_advanced.py

import logging
import wandb
from transformers import TrainerCallback

# Setup logging
logging.basicConfig(
    level=logging.INFO,
    format='%(asctime)s - %(name)s - %(levelname)s - %(message)s',
    handlers=[
        logging.FileHandler('training.log'),
        logging.StreamHandler()
    ]
)

class CustomCallback(TrainerCallback):
    def on_log(self, args, state, control, logs=None, **kwargs):
        if logs:
            logging.info(f"Step {state.global_step}: {logs}")
    
    def on_epoch_end(self, args, state, control, **kwargs):
        logging.info(f"Epoch {state.epoch} completed")

def setup_wandb():
    wandb.init(project="OrcaleSeek", entity="your-username")
    wandb.config = {
        "learning_rate": 2e-5,
        "architecture": "OrcaleSeek",
        "dataset": "Your-Dataset",
    }