| { |
| "aux_attn_accuracy": 0.9422222224871317, |
| "aux_loss": 0.2298412849009037, |
| "epoch": 1.0, |
| "eval_aux_attn_accuracy": 0.9623015873015873, |
| "eval_aux_loss": 0.17712821468474374, |
| "eval_lm_loss": 0.05031332116535244, |
| "eval_loss": 0.01336627546697855, |
| "eval_mean_token_accuracy": 0.9826168309128473, |
| "eval_runtime": 62.2964, |
| "eval_samples_per_second": 8.042, |
| "eval_steps_per_second": 2.023, |
| "eval_total_loss": 0.06802614390691152, |
| "lm_loss": 0.0855401718375894, |
| "mean_token_accuracy": 0.9700513021151225, |
| "total_flos": 1.066126245903519e+19, |
| "total_loss": 0.10852430010680109, |
| "train_loss": 0.027537114826725155, |
| "train_runtime": 28154.6326, |
| "train_samples_per_second": 1.758, |
| "train_steps_per_second": 0.11 |
| } |