models/deepseek-doc-and-code/all_results.json

{
    "epoch": 4.0,
    "eval_loss": 0.3583337068557739,
    "eval_runtime": 1098.6863,
    "eval_samples": 29976,
    "eval_samples_per_second": 27.283,
    "eval_steps_per_second": 1.706,
    "train_loss": 0.4407963699874701,
    "train_runtime": 45928.5894,
    "train_samples": 29976,
    "train_samples_per_second": 2.611,
    "train_steps_per_second": 0.653
}