| {"step": 100, "eval_loss": 2.701767250008932, "eval_perplexity": 14.906051180575668, "timestamp": 1762195499.101227} |
| {"step": 200, "eval_loss": 2.589470992610711, "eval_perplexity": 13.322721920351709, "timestamp": 1762198241.6253998} |
| {"step": 1000, "eval_loss": 2.2758695520716348, "eval_perplexity": 9.736381623497158, "timestamp": 1762217015.4859388} |
| {"step": 2000, "eval_loss": 2.2758695520716348, "eval_perplexity": 9.736381623497158, "timestamp": 1762231144.4016669} |
| {"step": 0, "eval_loss": 2.820284757720524, "eval_perplexity": 16.78162869015132, "timestamp": 1762239554.4713366, "note": "baseline_before_training"} |
| {"step": 0, "eval_loss": 2.820284757720524, "eval_perplexity": 16.78162869015132, "timestamp": 1762244725.5739768, "note": "baseline_before_training"} |
| {"step": 1000, "eval_loss": 2.056745729505061, "eval_perplexity": 7.820478402725101, "timestamp": 1762258895.763943} |
| {"step": 2000, "eval_loss": 1.3733391432044708, "eval_perplexity": 3.948513358781165, "timestamp": 1762273019.888762} |
| {"step": 3000, "eval_loss": 1.2104632968351552, "eval_perplexity": 3.3550386713334865, "timestamp": 1762287081.1852064} |
| {"step": 4000, "eval_loss": 1.1494164200887527, "eval_perplexity": 3.1563503894330243, "timestamp": 1762346381.3138683} |
| {"step": 5000, "eval_loss": 1.0639458149671555, "eval_perplexity": 2.897782574313701, "timestamp": 1762360080.333399} |
| {"step": 6000, "eval_loss": 0.9279604775477328, "eval_perplexity": 2.529345257000655, "timestamp": 1762373797.1349308} |
| {"step": 7000, "eval_loss": 0.8813955082091368, "eval_perplexity": 2.4142664853863813, "timestamp": 1762396459.8795457} |
|
|
