training/metrics.jsonl

{"entropy": 0.45220437087118626, "epoch": 0.25396825396825395, "grad_norm": 2.71875, "learning_rate": 0.0, "loss": 1.9140263795852661, "mean_token_accuracy": 0.6275349594652653, "num_tokens": 31598.0, "step": 1}
{"entropy": 0.4340843269601464, "epoch": 0.5079365079365079, "grad_norm": 2.25, "learning_rate": 0.0002, "loss": 1.5281498432159424, "mean_token_accuracy": 0.6091829854995012, "num_tokens": 70107.0, "step": 2}
{"entropy": 0.5076298136264086, "epoch": 0.7619047619047619, "grad_norm": 2.1875, "learning_rate": 0.00015000000000000001, "loss": 1.7773090600967407, "mean_token_accuracy": 0.6582653727382421, "num_tokens": 110894.0, "step": 3}
{"entropy": 0.5993171056111654, "epoch": 1.0, "grad_norm": 1.609375, "learning_rate": 5.000000000000002e-05, "loss": 1.2829618453979492, "mean_token_accuracy": 0.6709515730539958, "num_tokens": 148838.0, "step": 4}
{"epoch": 1.0, "eval_entropy": 0.6153187089496188, "eval_loss": 1.7135308980941772, "eval_mean_token_accuracy": 0.678737986418936, "eval_num_tokens": 148838.0, "eval_runtime": 6.7623, "eval_samples_per_second": 1.331, "eval_steps_per_second": 1.331, "step": 4}
{"epoch": 1.0, "step": 4, "total_flos": 6836753973792768.0, "train_loss": 1.6256117820739746, "train_runtime": 250.3595, "train_samples_per_second": 0.256, "train_steps_per_second": 0.016}