adapter_config.json

{
    "adapter_path": "adapters/disarm_ew_llama3_lora_memory_optimized",
    "batch_size": 1,
    "config": "scripts/mlx_finetune_config_memory_optimized.yaml",
    "data": "Data/training",
    "eval_interval": 50,
    "eval_iters": 10,
    "fine_tune_type": "lora",
    "grad_checkpoint": false,
    "gradient_accumulation_steps": 4,
    "iters": 600,
    "learning_rate": 0.0003,
    "log_interval": 10,
    "lora_alpha": 16,
    "lora_dropout": 0.0,
    "lora_parameters": {
        "rank": 8,
        "dropout": 0.0,
        "scale": 20.0
    },
    "lora_rank": 16,
    "lr_schedule": null,
    "lr_scheduler": "cosine",
    "mask_prompt": false,
    "max_seq_length": 2048,
    "model": "ArapCheruiyot/disarm_ew-llama3",
    "num_layers": 4,
    "optimizer": "adam",
    "optimizer_config": {
        "adam": {},
        "adamw": {},
        "muon": {},
        "sgd": {},
        "adafactor": {}
    },
    "output_dir": "outputs/mlx_finetune_memory_optimized",
    "quantize": false,
    "resume_adapter_file": null,
    "save_every": 100,
    "save_interval": 100,
    "seed": 0,
    "steps_per_eval": 200,
    "steps_per_report": 10,
    "test": false,
    "test_batches": 500,
    "train": true,
    "use_metal": true,
    "use_metal_float16": true,
    "use_wired_memory": true,
    "val_batches": 25,
    "wandb": null,
    "warmup_steps": 50,
    "weight_decay": 0.01,
    "wired_memory_limit_mb": 16384
}