{
  "fine_tuned": {
    "model": "opus-mt-en-vi-finetuned (1 epoch)",
    "samples": 17272,
    "bleu": 29.29,
    "chrfpp": 49.13,
    "runtime_minutes": 64.8,
    "speed_samples_per_second": 4.44,
    "batch_size": 16,
    "num_beams": 4,
    "max_length": 256,
    "max_new_tokens": 256
  },
  "baseline": {
    "model": "Helsinki-NLP/opus-mt-en-vi",
    "samples": 17272,
    "bleu": 3.73,
    "chrfpp": 20.87,
    "runtime_minutes": 28.8,
    "speed_samples_per_second": 10.0,
    "batch_size": 256,
    "num_beams": 4,
    "fp16": true
  },
  "sanity_check_256": {
    "baseline_fp32_bs16": {
      "bleu": 3.53,
      "chrfpp": 20.5,
      "speed_sps": 4.72
    },
    "baseline_fp16_bs256": {
      "bleu": 3.57,
      "chrfpp": 20.56,
      "speed_sps": 10.01
    },
    "fine_tuned_fp16_bs256": {
      "bleu": 29.01,
      "chrfpp": 49.07,
      "speed_sps": 9.79
    }
  },
  "training": {
    "base_model": "Helsinki-NLP/opus-mt-en-vi",
    "epochs": 1,
    "train_samples_approx": "328K",
    "val_samples_full": 17272,
    "batch_size": 16,
    "eval_samples_during_training": 128,
    "train_loss": 1.959,
    "eval_loss": 1.496,
    "eval_bleu_during_training": 34.29,
    "eval_chrf_during_training": 54.54,
    "train_runtime_hours": 3.26,
    "gpu": "NVIDIA L4",
    "colab_ram_gb": 53,
    "colab_gpu_ram_gb": 22.5
  }
}