model: unsloth/Qwen2.5-0.5B

data:
  dataset: tatsu-lab/alpaca
  format_type: auto

training:
  training_type: lora
  max_seq_length: 2048
  load_in_4bit: true
  output_dir: outputs
  num_epochs: 1
  learning_rate: 0.0002
  batch_size: 2
  gradient_accumulation_steps: 4
  warmup_steps: 5
  max_steps: 0
  save_steps: 0
  weight_decay: 0.01
  random_seed: 3407
  packing: false
  train_on_completions: false
  gradient_checkpointing: "unsloth"

lora:
  lora_r: 64
  lora_alpha: 16
  lora_dropout: 0.0
  target_modules: "q_proj,k_proj,v_proj,o_proj,gate_proj,up_proj,down_proj"
  vision_all_linear: false
  use_rslora: false
  use_loftq: false
  finetune_vision_layers: true
  finetune_language_layers: true
  finetune_attention_modules: true
  finetune_mlp_modules: true

logging:
  enable_wandb: false
  wandb_project: unsloth-training
  enable_tensorboard: false
  tensorboard_dir: runs