{ "phase": "stream_align", "seed": 0, "rows": 159851, "rows_by_lang": { "en": 159851 }, "rows_with_voice_prompt": 135746, "rows_with_tool_prompt": 11292, "rows_with_voice_hint": 67943, "rows_prompt_withheld": 24091, "prompt_frames": { "median": 48, "p10": 48, "p90": 48, "below_full": 0.0692 }, "by_lang": { "en": { "rows": 159851, "rows_with_voice_prompt": 135746, "prompt_share": 0.8492, "rows_with_voice_hint": 67943, "rows_prompt_withheld": 24091, "prompt_frames": { "median": 48, "p10": 48, "p90": 48, "below_full": 0.0692 } } }, "rows_skipped": null, "voice_hint_share_of_prompted": 0.5005, "tool_prompt_share_realized": 0.0706, "zone_a_frames_mean": 50.0, "zone_a_share_of_timeline": 0.0495, "opens": { "self": 80015, "user": 79836 }, "row_frames": { "p10": 756, "median": 1003, "p90": 1289, "max": 1498 }, "rows_in_target_range": 0.9101, "rows_over_hard_cap": 0, "rows_supply_bound": 57487, "frames_by_lang": { "en": 161759346 }, "frame_share_by_lang": { "en": 1.0 }, "frame_share_by_subset": { "test": { "en": 1.0 }, "train": { "en": 1.0 }, "valid": { "en": 1.0 } }, "frame_share_declared": { "ko": 0.549, "en": 0.451 }, "turns": { "p10": 8, "median": 12, "p90": 15, "max": 17 }, "frames_by_self_supply": { "1": { "p10": 111, "median": 245, "p90": 358, "max": 546, "rows": 14 }, "2-3": { "p10": 245, "median": 402, "p90": 662, "max": 739, "rows": 45 }, "4-7": { "p10": 754, "median": 990, "p90": 1271, "max": 1497, "rows": 106608 }, "8+": { "p10": 763, "median": 1033, "p90": 1318, "max": 1498, "rows": 53184 } }, "ctc_target_ratio": { "assistant": { "frame_ratio": { "rows": 159848, "p10": 0.0809, "median": 0.1226, "p90": 0.1665, "max": 0.2956, "share_over_one": 0.0 }, "speech_ratio": { "rows": 159848, "p10": 0.3113, "median": 0.3688, "p90": 0.435, "max": 1.0, "share_over_one": 0.0 } }, "user": { "frame_ratio": { "rows": 159847, "p10": 0.085, "median": 0.1296, "p90": 0.1723, "max": 0.2708, "share_over_one": 0.0 }, "speech_ratio": { "rows": 159847, "p10": 0.3257, "median": 0.3631, "p90": 0.4055, "max": 1.0, "share_over_one": 0.0 } } }, "frames_total": 161759346, "hours_of_timeline": 3594.65, "sources": { "en": [ { "name": "libritts_r", "rows": 149291, "build_report": { "source": "libritts_r", "reader": "project_amnesty.datasets.libritts_r.LibriTTSRDataset", "repo_commit": "5d6df8edeaea5cd77a9d9413930ce781fd791e44", "cache_version": 2, "rows": 159851, "rows_by_subset": { "train": 149291, "test": 4830, "valid": 5730 }, "stats": { "total": 160267, "cached": 70272, "encoded": 89579, "failed": 416, "drops": { "no_alignable_text": 416 } }, "seconds": 3721.3, "rows_per_second": 42.96, "processor": "/home/work/voice_research/haan/trainer_output/init_scale_match", "align_model": "facebook/wav2vec2-base-960h", "mimi": "kyutai/mimi", "target_lufs": -20.0, "encode_batch_size": 8, "device": "cuda", "source_root": "data/corpora", "limit": null, "python": "3.12.3", "torch": "2.11.0+cu128" } } ] }, "settings": { "delay_offsets": [ 0, 1, 1, 1, 1, 1, 1, 1 ], "gap_frames": [ 4, 13 ], "gap_frames_note": "declared, not measured: turn gaps are Nunchi's subject and modelling them here would write that phase's answer into an earlier phase's data", "skip_other_speaker": true, "system_prompt_pool": { "en": { "plain": 39, "with_tools": 36 }, "ja": { "plain": 6, "with_tools": 4 }, "ko": { "plain": 39, "with_tools": 31 } }, "tool_prompt_share_declared": 0.125, "num_codebooks": 8, "eligibility": { "en": [ { "name": "libritts_r", "rows": 149291, "rows_label_covered": null, "rows_pairable_in_split": null } ] }, "subsets": [ "train", "valid", "test" ] }, "repo_commit": "9131b9d4454ecba591f25a533e6026357030335a", "seconds": 2882.6, "python": "3.12.3", "pipeline_validation": true }