dataset_path: data/lora_dataset_20250910_145055.jsonl
lora_config:
  lora_alpha: 32
  lora_dropout: 0.1
  r: 16
  target_modules:
  - q_proj
  - v_proj
  - k_proj
  - o_proj
  - gate_proj
  - up_proj
  - down_proj
max_length: 2048
model_name: meta-llama/Llama-3.1-8B-Instruct
model_path: ./models/llama-3.1-8b-instruct
repetition_penalty: 1.1
temperature: 0.7
top_k: 40
top_p: 0.9
training_config:
  batch_size: 1
  eval_steps: 50
  gradient_accumulation_steps: 16
  learning_rate: 0.0001
  num_epochs: 2
  save_steps: 50
  warmup_steps: 25
  logging_steps: 5
  save_total_limit: 2
  prediction_loss_only: true
  remove_unused_columns: false
  push_to_hub: false
  report_to: null