model_name_or_path: "openai-community/gpt2-large"
dataset_name_or_path: "allenai/tulu-3-sft-olmo-2-mixture-0225"
project_name: "scaling-post-training" 
training_args:
  seed: 42
  num_train_epochs: 1
  per_device_train_batch_size: 2
  per_device_eval_batch_size: 2
  gradient_accumulation_steps: 8
  warmup_ratio: 0.05
  weight_decay: 0.01
  logging_steps: 10
  eval_strategy: "steps"
  eval_steps: 50
  report_to: "wandb"
  fp16: true
  learning_rate: 3.0e-5
  lr_scheduler_type: "cosine"
  run_name: "gpt2-large-sft"
  output_dir: "models/gpt2-large/sft"
  save_strategy: "best"
  metric_for_best_model: "eval_loss"
  load_best_model_at_end: true
  save_total_limit: 1
  hub_model_id: "gpt2-large-sft"