wandb/run-20240412_072736-gyl3aena/files/wandb-summary.json

{"train/loss": 4.5875, "train/grad_norm": 226.922119140625, "train/learning_rate": 7.97387338950315e-06, "train/epoch": 0.0, "train/global_step": 40, "_timestamp": 1712907244.1121442, "_runtime": 387.48006534576416, "_step": 5, "train_runtime": 39.0078, "train_samples_per_second": 3.076, "train_steps_per_second": 1.538, "total_flos": 19812939669504.0, "train_loss": 4.6811678568522135}