all_results.json

{
    "epoch": 5.0,
    "eval_modality_acc_video": 85.7857882976532,
    "eval_modality_accuracy": 85.7857882976532,
    "eval_modality_f1_macro": 46.17456793785095,
    "eval_modality_f1_video": 92.3491358757019,
    "eval_modality_precision_macro": 50.0,
    "eval_modality_precision_video": 100.0,
    "eval_modality_pred_dist_asr": 14.214214214214213,
    "eval_modality_pred_dist_video": 85.78578578578579,
    "eval_modality_recall_macro": 42.8928941488266,
    "eval_modality_recall_video": 85.7857882976532,
    "eval_model_preparation_time": 0.0117,
    "eval_ndcg@10": 62.39320635795593,
    "eval_r@1": 46.0999995470047,
    "eval_r@10": 79.79999780654907,
    "eval_r@5": 71.29999995231628,
    "eval_runtime": 134.251,
    "eval_samples": 1000,
    "eval_samples_per_second": 7.449,
    "eval_steps_per_second": 0.469,
    "total_flos": 0.0,
    "train_loss": 0.37849533693159876,
    "train_runtime": 13899.499,
    "train_samples": 7695,
    "train_samples_per_second": 2.768,
    "train_steps_per_second": 0.346
}