{ "contains_mixed_model_types": false, "metric_recommendations": { "cloze_en_contains": { "checkpoint_name": "step_14700", "checkpoint_path": "/mnt/apps/llm-nanochat/checkpoints/20260629_resume-gpt2medium-gpt2preln-k20-wsddecayonly-rerunmissing-lr3p5294e5-anchor20k-final2e5-webwiki-step14200-to14850/step_14700.pt", "direction": "max", "value": 0.16 }, "cloze_en_exact": { "checkpoint_name": "step_14700", "checkpoint_path": "/mnt/apps/llm-nanochat/checkpoints/20260629_resume-gpt2medium-gpt2preln-k20-wsddecayonly-rerunmissing-lr3p5294e5-anchor20k-final2e5-webwiki-step14200-to14850/step_14700.pt", "direction": "max", "value": 0.0 }, "cloze_it_contains": { "checkpoint_name": "step_14700", "checkpoint_path": "/mnt/apps/llm-nanochat/checkpoints/20260629_resume-gpt2medium-gpt2preln-k20-wsddecayonly-rerunmissing-lr3p5294e5-anchor20k-final2e5-webwiki-step14200-to14850/step_14700.pt", "direction": "max", "value": 0.24 }, "cloze_it_exact": { "checkpoint_name": "step_14700", "checkpoint_path": "/mnt/apps/llm-nanochat/checkpoints/20260629_resume-gpt2medium-gpt2preln-k20-wsddecayonly-rerunmissing-lr3p5294e5-anchor20k-final2e5-webwiki-step14200-to14850/step_14700.pt", "direction": "max", "value": 0.0 }, "distinct_1": { "checkpoint_name": "step_14700", "checkpoint_path": "/mnt/apps/llm-nanochat/checkpoints/20260629_resume-gpt2medium-gpt2preln-k20-wsddecayonly-rerunmissing-lr3p5294e5-anchor20k-final2e5-webwiki-step14200-to14850/step_14700.pt", "direction": "max", "value": 0.27804410354745923 }, "distinct_2": { "checkpoint_name": "step_14700", "checkpoint_path": "/mnt/apps/llm-nanochat/checkpoints/20260629_resume-gpt2medium-gpt2preln-k20-wsddecayonly-rerunmissing-lr3p5294e5-anchor20k-final2e5-webwiki-step14200-to14850/step_14700.pt", "direction": "max", "value": 0.5643070787637089 }, "language_consistency_en": { "checkpoint_name": "step_14700", "checkpoint_path": "/mnt/apps/llm-nanochat/checkpoints/20260629_resume-gpt2medium-gpt2preln-k20-wsddecayonly-rerunmissing-lr3p5294e5-anchor20k-final2e5-webwiki-step14200-to14850/step_14700.pt", "direction": "max", "value": 1.0 }, "language_consistency_it": { "checkpoint_name": "step_14700", "checkpoint_path": "/mnt/apps/llm-nanochat/checkpoints/20260629_resume-gpt2medium-gpt2preln-k20-wsddecayonly-rerunmissing-lr3p5294e5-anchor20k-final2e5-webwiki-step14200-to14850/step_14700.pt", "direction": "max", "value": 0.85 }, "language_switch_rate_en": { "checkpoint_name": "step_14700", "checkpoint_path": "/mnt/apps/llm-nanochat/checkpoints/20260629_resume-gpt2medium-gpt2preln-k20-wsddecayonly-rerunmissing-lr3p5294e5-anchor20k-final2e5-webwiki-step14200-to14850/step_14700.pt", "direction": "min", "value": 0.0 }, "language_switch_rate_it": { "checkpoint_name": "step_14700", "checkpoint_path": "/mnt/apps/llm-nanochat/checkpoints/20260629_resume-gpt2medium-gpt2preln-k20-wsddecayonly-rerunmissing-lr3p5294e5-anchor20k-final2e5-webwiki-step14200-to14850/step_14700.pt", "direction": "min", "value": 0.0 }, "loop_rate": { "checkpoint_name": "step_14700", "checkpoint_path": "/mnt/apps/llm-nanochat/checkpoints/20260629_resume-gpt2medium-gpt2preln-k20-wsddecayonly-rerunmissing-lr3p5294e5-anchor20k-final2e5-webwiki-step14200-to14850/step_14700.pt", "direction": "min", "value": 0.375 }, "ppl_en": { "checkpoint_name": "step_14700", "checkpoint_path": "/mnt/apps/llm-nanochat/checkpoints/20260629_resume-gpt2medium-gpt2preln-k20-wsddecayonly-rerunmissing-lr3p5294e5-anchor20k-final2e5-webwiki-step14200-to14850/step_14700.pt", "direction": "min", "value": 80.87098332928396 }, "ppl_it": { "checkpoint_name": "step_14700", "checkpoint_path": "/mnt/apps/llm-nanochat/checkpoints/20260629_resume-gpt2medium-gpt2preln-k20-wsddecayonly-rerunmissing-lr3p5294e5-anchor20k-final2e5-webwiki-step14200-to14850/step_14700.pt", "direction": "min", "value": 35.982162207895115 }, "ppl_mixed": { "checkpoint_name": "step_14700", "checkpoint_path": "/mnt/apps/llm-nanochat/checkpoints/20260629_resume-gpt2medium-gpt2preln-k20-wsddecayonly-rerunmissing-lr3p5294e5-anchor20k-final2e5-webwiki-step14200-to14850/step_14700.pt", "direction": "min", "value": 85.07813449956652 }, "repeated_4gram_rate": { "checkpoint_name": "step_14700", "checkpoint_path": "/mnt/apps/llm-nanochat/checkpoints/20260629_resume-gpt2medium-gpt2preln-k20-wsddecayonly-rerunmissing-lr3p5294e5-anchor20k-final2e5-webwiki-step14200-to14850/step_14700.pt", "direction": "min", "value": 0.75 }, "source_loss_books_en": { "checkpoint_name": "step_14700", "checkpoint_path": "/mnt/apps/llm-nanochat/checkpoints/20260629_resume-gpt2medium-gpt2preln-k20-wsddecayonly-rerunmissing-lr3p5294e5-anchor20k-final2e5-webwiki-step14200-to14850/step_14700.pt", "direction": "min", "value": 4.411039806547619 }, "source_loss_books_it": { "checkpoint_name": "step_14700", "checkpoint_path": "/mnt/apps/llm-nanochat/checkpoints/20260629_resume-gpt2medium-gpt2preln-k20-wsddecayonly-rerunmissing-lr3p5294e5-anchor20k-final2e5-webwiki-step14200-to14850/step_14700.pt", "direction": "min", "value": 4.39037105015346 }, "source_loss_code": { "checkpoint_name": "step_14700", "checkpoint_path": "/mnt/apps/llm-nanochat/checkpoints/20260629_resume-gpt2medium-gpt2preln-k20-wsddecayonly-rerunmissing-lr3p5294e5-anchor20k-final2e5-webwiki-step14200-to14850/step_14700.pt", "direction": "min", "value": 7.720833333333333 }, "source_loss_web_en": { "checkpoint_name": "step_14700", "checkpoint_path": "/mnt/apps/llm-nanochat/checkpoints/20260629_resume-gpt2medium-gpt2preln-k20-wsddecayonly-rerunmissing-lr3p5294e5-anchor20k-final2e5-webwiki-step14200-to14850/step_14700.pt", "direction": "min", "value": 5.489309210526316 }, "source_loss_web_it": { "checkpoint_name": "step_14700", "checkpoint_path": "/mnt/apps/llm-nanochat/checkpoints/20260629_resume-gpt2medium-gpt2preln-k20-wsddecayonly-rerunmissing-lr3p5294e5-anchor20k-final2e5-webwiki-step14200-to14850/step_14700.pt", "direction": "min", "value": 5.4087171052631575 }, "source_loss_wiki_en": { "checkpoint_name": "step_14700", "checkpoint_path": "/mnt/apps/llm-nanochat/checkpoints/20260629_resume-gpt2medium-gpt2preln-k20-wsddecayonly-rerunmissing-lr3p5294e5-anchor20k-final2e5-webwiki-step14200-to14850/step_14700.pt", "direction": "min", "value": 2.9984019886363638 }, "source_loss_wiki_it": { "checkpoint_name": "step_14700", "checkpoint_path": "/mnt/apps/llm-nanochat/checkpoints/20260629_resume-gpt2medium-gpt2preln-k20-wsddecayonly-rerunmissing-lr3p5294e5-anchor20k-final2e5-webwiki-step14200-to14850/step_14700.pt", "direction": "min", "value": 2.92021484375 }, "val_loss_en": { "checkpoint_name": "step_14700", "checkpoint_path": "/mnt/apps/llm-nanochat/checkpoints/20260629_resume-gpt2medium-gpt2preln-k20-wsddecayonly-rerunmissing-lr3p5294e5-anchor20k-final2e5-webwiki-step14200-to14850/step_14700.pt", "direction": "min", "value": 4.392855086416568 }, "val_loss_it": { "checkpoint_name": "step_14700", "checkpoint_path": "/mnt/apps/llm-nanochat/checkpoints/20260629_resume-gpt2medium-gpt2preln-k20-wsddecayonly-rerunmissing-lr3p5294e5-anchor20k-final2e5-webwiki-step14200-to14850/step_14700.pt", "direction": "min", "value": 3.5830233214331453 }, "val_loss_mixed": { "checkpoint_name": "step_14700", "checkpoint_path": "/mnt/apps/llm-nanochat/checkpoints/20260629_resume-gpt2medium-gpt2preln-k20-wsddecayonly-rerunmissing-lr3p5294e5-anchor20k-final2e5-webwiki-step14200-to14850/step_14700.pt", "direction": "min", "value": 4.443570063664363 } }, "recommended_checkpoint": { "checkpoint_name": "step_14700", "checkpoint_path": "/mnt/apps/llm-nanochat/checkpoints/20260629_resume-gpt2medium-gpt2preln-k20-wsddecayonly-rerunmissing-lr3p5294e5-anchor20k-final2e5-webwiki-step14200-to14850/step_14700.pt", "direction": "min", "value": 4.443570063664363 }, "recommended_metric": "val_loss_mixed", "rows": [ { "aggregate_dataset_count": 0, "aggregate_validation_loss_mean": null, "aggregate_validation_perplexity_mean": null, "checkpoint_name": "step_14700", "checkpoint_path": "/mnt/apps/llm-nanochat/checkpoints/20260629_resume-gpt2medium-gpt2preln-k20-wsddecayonly-rerunmissing-lr3p5294e5-anchor20k-final2e5-webwiki-step14200-to14850/step_14700.pt", "checkpoint_selector": "manual", "checkpoint_step": 14700, "cloze_en_contains": 0.16, "cloze_en_exact": 0.0, "cloze_it_contains": 0.24, "cloze_it_exact": 0.0, "delta_vs_previous_generation_pass_rate": null, "delta_vs_previous_validation_loss_mean": null, "distinct_1": 0.27804410354745923, "distinct_2": 0.5643070787637089, "generation_pass_rate": null, "generation_pass_rate_regression_vs_previous": false, "generation_passed_prompts": 0, "generation_scored_prompts": 0, "generation_total_prompts": 40, "language_consistency_en": 1.0, "language_consistency_it": 0.85, "language_switch_rate_en": 0.0, "language_switch_rate_it": 0.0, "loop_rate": 0.375, "model_type": "pretrained", "ppl_en": 80.87098332928396, "ppl_it": 35.982162207895115, "ppl_mixed": 85.07813449956652, "repeated_4gram_rate": 0.75, "run_dir": null, "run_name": null, "selected_by": "manual_path", "selection_metric_name": null, "selection_metric_value": null, "source_loss_books_en": 4.411039806547619, "source_loss_books_it": 4.39037105015346, "source_loss_code": 7.720833333333333, "source_loss_web_en": 5.489309210526316, "source_loss_web_it": 5.4087171052631575, "source_loss_wiki_en": 2.9984019886363638, "source_loss_wiki_it": 2.92021484375, "val_loss_en": 4.392855086416568, "val_loss_it": 3.5830233214331453, "val_loss_mixed": 4.443570063664363, "validation_loss_regression_vs_previous": false } ] }