metrics.json

{
  "history": [
    {
      "epoch": 1,
      "train_loss": 0.953,
      "val_loss": 0.0244,
      "examples_per_sec": 1.3,
      "elapsed_sec": 3331.1
    },
    {
      "epoch": 2,
      "train_loss": 0.0203,
      "val_loss": 0.0177,
      "examples_per_sec": 1.3,
      "elapsed_sec": 3318.3
    },
    {
      "epoch": 3,
      "train_loss": 0.0151,
      "val_loss": 0.0151,
      "examples_per_sec": 1.3,
      "elapsed_sec": 3321.8
    },
    {
      "epoch": 4,
      "train_loss": 0.0125,
      "val_loss": 0.0147,
      "examples_per_sec": 1.3,
      "elapsed_sec": 3318.8
    },
    {
      "epoch": 5,
      "train_loss": 0.0109,
      "val_loss": 0.0146,
      "examples_per_sec": 1.3,
      "elapsed_sec": 3317.5
    }
  ],
  "best_val_loss": 0.0146,
  "config": {
    "seed": 42,
    "data": {
      "num_workers": 4
    },
    "model": {
      "name": "meta-llama/Llama-3.2-11B-Vision-Instruct"
    },
    "lora": {
      "rank": 16,
      "alpha": 32,
      "dropout": 0.05,
      "target_modules": [
        "q_proj",
        "v_proj",
        "k_proj",
        "o_proj",
        "gate_proj",
        "up_proj",
        "down_proj"
      ]
    },
    "train": {
      "epochs": 5,
      "batch_size": 4,
      "grad_accum_steps": 4,
      "lr": 2e-05,
      "weight_decay": 0.01,
      "betas": [
        0.9,
        0.999
      ],
      "warmup_ratio": 0.1,
      "grad_clip": 1.0,
      "max_length": 2048
    },
    "log": {
      "every_n_steps": 25
    }
  }
}