{"train/loss": 4.6222, "train/learning_rate": 0.00028865524291388006, "train/epoch": 1.06, "train/global_step": 290, "_timestamp": 1706811429.4937887, "_runtime": 1099.4098217487335, "_step": 30, "eval/loss": 4.737916946411133, "eval/runtime": 30.7851, "eval/samples_per_second": 501.248, "eval/steps_per_second": 1.981} |