training_hyperparams.json

{
  "references": [
    "https://docs.unsloth.ai/get-started/fine-tuning-llms-guide/lora-hyperparameters-guide",
    "https://huggingface.co/blog/ImranzamanML/fine-tuning-1b-llama-32-a-comprehensive-article"
  ],
  "model_id": "unsloth/Llama-3.2-1B-Instruct",
  "model_source": "unsloth/Llama-3.2-1B-Instruct",
  "data_path": "/data/training/qa_sft_messages.jsonl",
  "hf_dataset": "NeuralVulture/ai-concepts-qa",
  "hf_split": "train",
  "hf_config": null,
  "lora_r": 16,
  "lora_alpha": 32,
  "lora_dropout": 0.05,
  "target_modules": [
    "q_proj",
    "k_proj",
    "v_proj",
    "o_proj",
    "gate_proj",
    "up_proj",
    "down_proj"
  ],
  "learning_rate": 0.0002,
  "lr_scheduler_type": "cosine",
  "warmup_steps": 0.05,
  "weight_decay": 0.01,
  "num_train_epochs": 2.0,
  "per_device_train_batch_size": 2,
  "gradient_accumulation_steps": 8,
  "effective_batch_size": 16,
  "max_length": 4096,
  "assistant_only_loss": false,
  "train_rows": 380,
  "eval_rows": 20,
  "data_file": null
}