| { | |
| "epoch": 1.0, | |
| "eval_loss": 1.079127550125122, | |
| "eval_runtime": 187.0771, | |
| "eval_samples": 25000, | |
| "eval_samples_per_second": 133.635, | |
| "eval_steps_per_second": 16.704, | |
| "perplexity": 2.9421115861328713, | |
| "total_flos": 3.641341261824e+16, | |
| "train_loss": 1.3328714208984376, | |
| "train_runtime": 7223.6875, | |
| "train_samples": 200000, | |
| "train_samples_per_second": 27.687, | |
| "train_steps_per_second": 0.433 | |
| } |