all_results.json

{
    "before_init_mem_cpu": 3866619904,
    "before_init_mem_gpu": 512,
    "epoch": 5.0,
    "eval_-_f1-score": 0.7665937272064187,
    "eval_-_precision": 0.7401408450704225,
    "eval_-_recall": 0.7950075642965204,
    "eval_-_support": 1322.0,
    "eval_<_f1-score": 0.9086178861788617,
    "eval_<_precision": 0.9121180464873335,
    "eval_<_recall": 0.9051444861993002,
    "eval_<_support": 7717.0,
    "eval_=_f1-score": 0.8044709845352932,
    "eval_=_precision": 0.7992090051718893,
    "eval_=_recall": 0.80980271270037,
    "eval_=_support": 3244.0,
    "eval_>_f1-score": 0.9064616988014591,
    "eval_>_precision": 0.9113294040602489,
    "eval_>_recall": 0.9016457172476351,
    "eval_>_support": 7717.0,
    "eval_accuracy": 0.88105,
    "eval_loss": 0.18984687328338623,
    "eval_macro_avg_f1-score": 0.8465360741805081,
    "eval_macro_avg_precision": 0.8406993251974735,
    "eval_macro_avg_recall": 0.8529001201109565,
    "eval_macro_avg_support": 20000.0,
    "eval_mem_cpu_alloc_delta": -269582336,
    "eval_mem_cpu_peaked_delta": 269582336,
    "eval_mem_gpu_alloc_delta": 0,
    "eval_mem_gpu_peaked_delta": 1242539008,
    "eval_runtime": 14.1119,
    "eval_samples": 20000,
    "eval_samples_per_second": 1417.242,
    "eval_steps_per_second": 5.598,
    "eval_weighted_avg_f1-score": 0.8815054969246255,
    "eval_weighted_avg_precision": 0.8821322092918199,
    "eval_weighted_avg_recall": 0.88105,
    "eval_weighted_avg_support": 20000.0,
    "init_mem_cpu_alloc_delta": 18591744,
    "init_mem_cpu_peaked_delta": 0,
    "init_mem_gpu_alloc_delta": 0,
    "init_mem_gpu_peaked_delta": 0,
    "total_flos": 2.2572496552911176e+18,
    "train_loss": 0.3141621667137878,
    "train_mem_cpu_alloc_delta": 827277312,
    "train_mem_cpu_peaked_delta": 311078912,
    "train_mem_gpu_alloc_delta": 757461504,
    "train_mem_gpu_peaked_delta": 19341542400,
    "train_runtime": 24803.7,
    "train_samples": 1386134,
    "train_samples_per_second": 1676.525,
    "train_steps_per_second": 3.274
}