training_config.json

{
    "model_name_or_path": "LiquidAI/LFM2-2.6B",
    "use_lora": true,
    "lora_r": 64,
    "lora_alpha": 128,
    "lora_dropout": 0.05,
    "data_path": "./kokoro_processed_data",
    "max_seq_length": 2048,
    "response_template": "### Response:",
    "output_dir": "./lfm_trl_finetuned",
    "num_train_epochs": 3,
    "per_device_train_batch_size": 4,
    "per_device_eval_batch_size": 4,
    "gradient_accumulation_steps": 4,
    "learning_rate": 2e-4,
    "warmup_ratio": 0.1,
    "logging_steps": 10,
    "save_steps": 100,
    "eval_steps": 100,
    "bf16": true,
    "tf32": true,
    "seed": 42
}