| {"event": "hub", "repo": "AdaptiveChunking/ali_mem-mixed_s3", "keep_steps": [0, 1, 2, 4, 6, 8, 16, 32, 64, 128, 256, 512, 1000, 1024, 2048, 3000, 4096, 6000], "t": 1.7} |
| {"event": "start", "run": "ali_mem-mixed_s3", "arch": "ali", "objective": "ratio", "condition": "A", "params": 22785088, "langs": ["en", "tr", "fi"], "premium": {"en": 1.0, "tr": 1.12, "fi": 1.104}, "schedule": {"tau1": 0.4, "tau2": 0.7, "p1_objective": "global", "p2_objective": "global", "p3_objective": "global", "freeze_in_p3": false, "freeze_scope": "router"}, "corpus": "/local/scratch/sas245/beetle-hnet/scratch/data/tier1_corpus", "premium_kind": "flores_per_sent", "doc_filtered": {}, "weights": {"en": 0.74, "tr": 0.1536, "fi": 0.1064}, "device": "cuda", "ali_params": 268096, "ali_config": {"enabled": true, "budget_B": 65536, "staging_capacity": 262144, "policy": "frequency", "value_mode": "learned", "candidate_types": ["surface", "split", "merge", "skel"], "gate": "vector", "retrieval": true, "memory_scope": "persistent", "read_dropout": 0.1, "write_rule": "mean", "eta_min": 0.0, "consolidate_every": 50, "probe_seqs": 2, "probe_reps": 8, "horizon": 32, "predictor": "mlp", "lambda_util": 0.1, "max_span": 64, "byte_classes_path": "/local/scratch/sas245/beetle-hnet/scratch/data/byte_classes.json", "gate_hidden": 0, "type_dim": 16, "gate_bias_init": -2.0}, "t": 1.7} |
| {"event": "probe_schedule", "n": 45, "steps": [0, 1, 2, 4, 6, 8, 16, 32, 64, 128, 200, 256, 400, 512, 600, 800, 1000, 1024, 1200, 1400, 1600, 1800, 2000, 2048, 2200, 2400, 2600, 2800, 3000, 3200, 3400, 3600, 3800, 4000, 4096, 4200, 4400, 4600, 4800, 5000, 5200, 5400, 5600, 5800, 6000], "t": 1.7} |
| {"event": "probe", "step": 0, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 15894.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.913291, "cps": 124.171875, "mean_p": 0.532199, "boundary_entropy": 0.685519, "bpb": 8.014977}, "tr": {"chunks": 15976.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.882506, "cps": 124.8125, "mean_p": 0.523031, "boundary_entropy": 0.686614, "bpb": 8.046717}, "fi": {"chunks": 12092.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.642815, "cps": 94.46875, "mean_p": 0.5099, "boundary_entropy": 0.687389, "bpb": 8.01587}}, "bpb": {"en": 8.01498, "tr": 8.04672, "fi": 8.01587}, "gini_cpc": 0.073942, "gini_cps": 0.058899, "bpic": 1.290478, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9662920621975919, "fi": 0.7038452216831197}, "inventory": {"size": 0, "admitted": 0, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 0.0, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 4.6} |
| {"event": "train", "step": 0, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 5.5638, "bpb": 8.02686, "chunk_loss": 0.092026, "F": 0.95703, "G": 0.52969, "chunks": 980.0, "bytes_per_chunk": 1.045, "boundary_entropy": 0.6902, "router_noise": 0.0, "grad_norm": 13.6073, "lr": 1e-06, "router_trainable": true, "sec_per_step": 0.018, "probe_sec": 0.0136, "probe_extra_seqs": 0, "n_rows": 0, "extra_fwd_seqs": 0, "wall_s": 0.0, "overhead_frac_est": 0.0, "chunk_mismatch": 0, "n_loo": 0, "mean_delta_loo": NaN, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "read_hit_rate": {"surface": 0.0, "split": 0.0, "merge": 0.0, "skel": 0.0}, "mean_gate": {"surface": null, "split": null, "merge": null, "skel": null}, "n_reads": 0, "n_cand": 65310, "read_suppressed_frac": 0.0, "dropout_reads_logged": 0, "inventory": {"size": 2916, "admitted": 2916, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 366, "staged_surface": 0, "admitted_split": 162, "staged_split": 0, "admitted_merge": 2302, "staged_merge": 0, "admitted_skel": 86, "staged_skel": 0, "mean_count": 22.397119522094727, "n_util_obs": 0, "util_count_spearman": NaN}, "probe_overhead_sec_total": 0.014, "probe_overhead_seq_frac": 0.0, "util_loss": -0.0, "alpha": 0.03, "t": 5.8} |
| {"event": "probe", "step": 1, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 15874.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.912142, "cps": 124.015625, "mean_p": 0.532017, "boundary_entropy": 0.685541, "bpb": 7.963626}, "tr": {"chunks": 15941.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.880572, "cps": 124.539062, "mean_p": 0.522881, "boundary_entropy": 0.686629, "bpb": 8.014483}, "fi": {"chunks": 12057.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.640955, "cps": 94.195312, "mean_p": 0.509762, "boundary_entropy": 0.687395, "bpb": 7.977594}}, "bpb": {"en": 7.96363, "tr": 8.01448, "fi": 7.97759}, "gini_cpc": 0.074288, "gini_cps": 0.05902, "bpic": 1.293125, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9653899082407584, "fi": 0.7026921820082177}, "inventory": {"size": 2916, "admitted": 2916, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 366, "staged_surface": 0, "admitted_split": 162, "staged_split": 0, "admitted_merge": 2302, "staged_merge": 0, "admitted_skel": 86, "staged_skel": 0, "mean_count": 22.397119522094727, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 8.5} |
| {"event": "probe", "step": 2, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 15834.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.909843, "cps": 123.703125, "mean_p": 0.531622, "boundary_entropy": 0.685589, "bpb": 7.891623}, "tr": {"chunks": 15884.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.877424, "cps": 124.09375, "mean_p": 0.522583, "boundary_entropy": 0.686659, "bpb": 7.957093}, "fi": {"chunks": 11978.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.636755, "cps": 93.578125, "mean_p": 0.509493, "boundary_entropy": 0.687405, "bpb": 7.915756}}, "bpb": {"en": 7.89162, "tr": 7.95709, "fi": 7.91576}, "gini_cpc": 0.075106, "gini_cps": 0.059594, "bpic": 1.298334, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9643680344821516, "fi": 0.6998515115024347}, "inventory": {"size": 5514, "admitted": 5514, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2, "admitted_surface": 791, "staged_surface": 0, "admitted_split": 359, "staged_split": 0, "admitted_merge": 4155, "staged_merge": 0, "admitted_skel": 209, "staged_skel": 0, "mean_count": 23.671926498413086, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 12.7} |
| {"event": "probe", "step": 4, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 15717.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.90312, "cps": 122.789062, "mean_p": 0.530114, "boundary_entropy": 0.685768, "bpb": 7.650801}, "tr": {"chunks": 15675.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.865879, "cps": 122.460938, "mean_p": 0.521366, "boundary_entropy": 0.686775, "bpb": 7.760076}, "fi": {"chunks": 11647.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.619159, "cps": 90.992188, "mean_p": 0.508427, "boundary_entropy": 0.687444, "bpb": 7.706061}}, "bpb": {"en": 7.6508, "tr": 7.76008, "fi": 7.70606}, "gini_cpc": 0.079269, "gini_cps": 0.063044, "bpic": 1.318153, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9587634403154471, "fi": 0.6855776626698898}, "inventory": {"size": 9210, "admitted": 9210, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5, "admitted_surface": 1371, "staged_surface": 0, "admitted_split": 684, "staged_split": 0, "admitted_merge": 6728, "staged_merge": 0, "admitted_skel": 427, "staged_skel": 0, "mean_count": 28.255048751831055, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 18.3} |
| {"event": "probe", "step": 6, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 15435.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.886916, "cps": 120.585938, "mean_p": 0.527556, "boundary_entropy": 0.686051, "bpb": 7.36185}, "tr": {"chunks": 15260.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.842954, "cps": 119.21875, "mean_p": 0.519299, "boundary_entropy": 0.686957, "bpb": 7.522493}, "fi": {"chunks": 10995.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.584498, "cps": 85.898438, "mean_p": 0.506537, "boundary_entropy": 0.687504, "bpb": 7.449735}}, "bpb": {"en": 7.36185, "tr": 7.52249, "fi": 7.44973}, "gini_cpc": 0.087113, "gini_cps": 0.071, "bpic": 1.360806, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9504329158056639, "fi": 0.6590234018833655}, "inventory": {"size": 12969, "admitted": 12969, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 10, "admitted_surface": 1987, "staged_surface": 0, "admitted_split": 1009, "staged_split": 0, "admitted_merge": 9319, "staged_merge": 0, "admitted_skel": 654, "staged_skel": 0, "mean_count": 30.05983543395996, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 25.5} |
| {"event": "probe", "step": 8, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 14977.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.860599, "cps": 117.007812, "mean_p": 0.523898, "boundary_entropy": 0.686407, "bpb": 7.113883}, "tr": {"chunks": 14610.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.807049, "cps": 114.140625, "mean_p": 0.516345, "boundary_entropy": 0.687186, "bpb": 7.318998}, "fi": {"chunks": 10056.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.534581, "cps": 78.5625, "mean_p": 0.503777, "boundary_entropy": 0.687567, "bpb": 7.230409}}, "bpb": {"en": 7.11388, "tr": 7.319, "fi": 7.23041}, "gini_cpc": 0.098693, "gini_cps": 0.082755, "bpic": 1.431072, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9377756567509095, "fi": 0.6211731448785325}, "inventory": {"size": 16521, "admitted": 16521, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 13, "admitted_surface": 2562, "staged_surface": 0, "admitted_split": 1326, "staged_split": 0, "admitted_merge": 11739, "staged_merge": 0, "admitted_skel": 894, "staged_skel": 0, "mean_count": 31.428665161132812, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 30.9} |
| {"event": "probe", "step": 16, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 5004.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.287537, "cps": 39.09375, "mean_p": 0.494076, "boundary_entropy": 0.687133, "bpb": 6.587279}, "tr": {"chunks": 5929.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.327515, "cps": 46.320312, "mean_p": 0.492622, "boundary_entropy": 0.687505, "bpb": 6.902532}, "fi": {"chunks": 2817.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.149753, "cps": 22.007812, "mean_p": 0.48157, "boundary_entropy": 0.686775, "bpb": 6.783761}}, "bpb": {"en": 6.58728, "tr": 6.90253, "fi": 6.78376}, "gini_cpc": 0.154952, "gini_cps": 0.150885, "bpic": 4.125964, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.1390367019205714, "fi": 0.5208129599664058}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 28, "admitted_surface": 12565, "staged_surface": 0, "admitted_split": 8742, "staged_split": 0, "admitted_merge": 36934, "staged_merge": 0, "admitted_skel": 7295, "staged_skel": 0, "mean_count": 14.2711181640625, "n_util_obs": 1, "util_count_spearman": NaN}, "t": 41.0} |
| {"event": "probe", "step": 32, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 7863.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.451819, "cps": 61.429688, "mean_p": 0.502605, "boundary_entropy": 0.687629, "bpb": 6.270993}, "tr": {"chunks": 10305.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.569243, "cps": 80.507812, "mean_p": 0.503798, "boundary_entropy": 0.688074, "bpb": 6.619802}, "fi": {"chunks": 7393.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.393015, "cps": 57.757812, "mean_p": 0.499399, "boundary_entropy": 0.688087, "bpb": 6.459566}}, "bpb": {"en": 6.27099, "tr": 6.6198, "fi": 6.45957}, "gini_cpc": 0.083083, "gini_cps": 0.075949, "bpic": 2.219475, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.2598919156972028, "fi": 0.8698505998473013}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 54, "admitted_surface": 13603, "staged_surface": 0, "admitted_split": 10633, "staged_split": 0, "admitted_merge": 32692, "staged_merge": 0, "admitted_skel": 8608, "staged_skel": 0, "mean_count": 16.522720336914062, "n_util_obs": 7, "util_count_spearman": 0.0}, "t": 56.3} |
| {"event": "consolidate", "step": 50, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 74.7} |
| {"event": "train", "step": 50, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 4.22726, "bpb": 6.09865, "chunk_loss": 0.033409, "F": 0.21451, "G": 0.49186, "chunks": 219.66, "bytes_per_chunk": 4.662, "boundary_entropy": 0.69206, "router_noise": 0.0, "grad_norm": 2.5068, "lr": 5.1e-05, "router_trainable": true, "sec_per_step": 1.4, "probe_sec": 0.1422, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.12158, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.00253, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": 0.00449, "n_merge": 1, "mean_delta_merge": -0.00562, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.37618, "trunc_abs_frac": 0.15746, "mean_delta_full": 0.03223, "read_hit_rate": {"surface": 0.6243, "split": 0.78674, "merge": 0.27542, "skel": 0.51363}, "mean_gate": {"surface": 0.12015, "split": 0.12016, "merge": 0.12006, "skel": 0.12006}, "n_reads": 16427, "n_cand": 28244, "read_suppressed_frac": 0.1018, "dropout_reads_logged": 16427, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 93, "admitted_surface": 14346, "staged_surface": 0, "admitted_split": 13776, "staged_split": 0, "admitted_merge": 27347, "staged_merge": 0, "admitted_skel": 10067, "staged_skel": 0, "mean_count": 19.972991943359375, "n_util_obs": 8, "util_count_spearman": 0.023809523809523808}, "probe_overhead_sec_total": 9.528, "probe_overhead_seq_frac": 0.4847, "util_loss": 0.001465, "alpha": 0.03, "t": 75.8} |
| {"event": "probe", "step": 64, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3383.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.194392, "cps": 26.429688, "mean_p": 0.48589, "boundary_entropy": 0.686575, "bpb": 5.607825}, "tr": {"chunks": 4539.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.250732, "cps": 35.460938, "mean_p": 0.488511, "boundary_entropy": 0.687355, "bpb": 6.001038}, "fi": {"chunks": 4908.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.260911, "cps": 38.34375, "mean_p": 0.488518, "boundary_entropy": 0.687209, "bpb": 5.801251}}, "bpb": {"en": 5.60783, "tr": 6.00104, "fi": 5.80125}, "gini_cpc": 0.06281, "gini_cps": 0.079241, "bpic": 4.421824, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.2898278610641454, "fi": 1.3421924546415658}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 123, "admitted_surface": 14520, "staged_surface": 0, "admitted_split": 13877, "staged_split": 0, "admitted_merge": 27116, "staged_merge": 0, "admitted_skel": 10023, "staged_skel": 0, "mean_count": 22.975570678710938, "n_util_obs": 9, "util_count_spearman": 0.06666666666666667}, "t": 89.9} |
| {"event": "consolidate", "step": 100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 123.3} |
| {"event": "train", "step": 100, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 3.31581, "bpb": 4.7837, "chunk_loss": 0.030414, "F": 0.17245, "G": 0.46796, "chunks": 176.59, "bytes_per_chunk": 5.799, "boundary_entropy": 0.68797, "router_noise": 0.0, "grad_norm": 1.7311, "lr": 0.000101, "router_trainable": true, "sec_per_step": 0.976, "probe_sec": 0.1741, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.13613, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": 0.01552, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.02808, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -0.34733, "trunc_abs_frac": 0.60507, "mean_delta_full": -0.06024, "read_hit_rate": {"surface": 0.54351, "split": 0.75893, "merge": 0.15419, "skel": 0.516}, "mean_gate": {"surface": 0.12009, "split": 0.12016, "merge": 0.12, "skel": 0.11996}, "n_reads": 13420, "n_cand": 25038, "read_suppressed_frac": 0.0988, "dropout_reads_logged": 13420, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 190, "admitted_surface": 15538, "staged_surface": 0, "admitted_split": 15493, "staged_split": 0, "admitted_merge": 23615, "staged_merge": 0, "admitted_skel": 10890, "staged_skel": 0, "mean_count": 31.40557861328125, "n_util_obs": 14, "util_count_spearman": 0.01098901098901099}, "probe_overhead_sec_total": 17.686, "probe_overhead_seq_frac": 0.4923, "util_loss": 0.000375, "alpha": 0.03, "t": 124.6} |
| {"event": "probe", "step": 128, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3334.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.191576, "cps": 26.046875, "mean_p": 0.41811, "boundary_entropy": 0.659618, "bpb": 4.13663}, "tr": {"chunks": 4596.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.253881, "cps": 35.90625, "mean_p": 0.422708, "boundary_entropy": 0.661866, "bpb": 4.452644}, "fi": {"chunks": 4409.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.234384, "cps": 34.445312, "mean_p": 0.420432, "boundary_entropy": 0.663113, "bpb": 4.379469}}, "bpb": {"en": 4.13663, "tr": 4.45264, "fi": 4.37947}, "gini_cpc": 0.061097, "gini_cps": 0.068185, "bpic": 4.597779, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.3252200358138615, "fi": 1.2234514502660045}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 239, "admitted_surface": 15663, "staged_surface": 0, "admitted_split": 15706, "staged_split": 0, "admitted_merge": 23160, "staged_merge": 0, "admitted_skel": 11007, "staged_skel": 0, "mean_count": 38.22346496582031, "n_util_obs": 14, "util_count_spearman": -0.11208791208791209}, "t": 151.9} |
| {"event": "consolidate", "step": 150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 172.9} |
| {"event": "train", "step": 150, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.74275, "bpb": 3.95696, "chunk_loss": 0.031843, "F": 0.20911, "G": 0.3641, "chunks": 214.12, "bytes_per_chunk": 4.782, "boundary_entropy": 0.61248, "router_noise": 0.0, "grad_norm": 1.3886, "lr": 0.000151, "router_trainable": true, "sec_per_step": 0.98, "probe_sec": 0.1459, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.12373, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.00258, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 3, "mean_delta_split": -0.00905, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.01847, "trunc_abs_frac": 0.01847, "mean_delta_full": -0.13965, "read_hit_rate": {"surface": 0.68446, "split": 0.83666, "merge": 0.2939, "skel": 0.61588}, "mean_gate": {"surface": 0.12024, "split": 0.12019, "merge": 0.12011, "skel": 0.12006}, "n_reads": 18097, "n_cand": 28471, "read_suppressed_frac": 0.1021, "dropout_reads_logged": 18097, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 284, "admitted_surface": 15805, "staged_surface": 0, "admitted_split": 16056, "staged_split": 0, "admitted_merge": 22407, "staged_merge": 0, "admitted_skel": 11268, "staged_skel": 0, "mean_count": 44.59889221191406, "n_util_obs": 21, "util_count_spearman": -0.08441558441558442}, "probe_overhead_sec_total": 26.525, "probe_overhead_seq_frac": 0.4948, "util_loss": 0.000244, "alpha": 0.03, "t": 173.6} |
| {"event": "consolidate", "step": 200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 223.1} |
| {"event": "probe", "step": 200, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3303.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.189795, "cps": 25.804688, "mean_p": 0.341529, "boundary_entropy": 0.530353, "bpb": 3.6423}, "tr": {"chunks": 4405.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.24333, "cps": 34.414062, "mean_p": 0.332623, "boundary_entropy": 0.528902, "bpb": 3.784781}, "fi": {"chunks": 4422.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.235075, "cps": 34.546875, "mean_p": 0.338186, "boundary_entropy": 0.542383, "bpb": 3.851711}}, "bpb": {"en": 3.6423, "tr": 3.78478, "fi": 3.85171}, "gini_cpc": 0.053412, "gini_cps": 0.0615, "bpic": 4.676999, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.2820675493842555, "fi": 1.2385752611268523}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 380, "admitted_surface": 15967, "staged_surface": 0, "admitted_split": 16411, "staged_split": 0, "admitted_merge": 21868, "staged_merge": 0, "admitted_skel": 11290, "staged_skel": 0, "mean_count": 56.91438293457031, "n_util_obs": 24, "util_count_spearman": -0.3373913043478261}, "t": 225.0} |
| {"event": "eval", "step": 200, "phase": 1, "bpb": {"en": 3.64734, "tr": 3.79426, "fi": 3.85018}, "bytes_per_chunk": {"en": 5.088, "tr": 4.472, "fi": 4.797}, "bpb_mean": 3.76393, "frozen_inventory": true, "t": 228.5} |
| {"event": "train", "step": 200, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.58724, "bpb": 3.73261, "chunk_loss": 0.031121, "F": 0.1976, "G": 0.33365, "chunks": 202.34, "bytes_per_chunk": 5.061, "boundary_entropy": 0.52735, "router_noise": 0.0, "grad_norm": 1.9576, "lr": 0.000201, "router_trainable": true, "sec_per_step": 1.142, "probe_sec": 0.1208, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08188, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": -0.01117, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.05151, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.2227, "trunc_abs_frac": 0.64201, "mean_delta_full": 0.11346, "read_hit_rate": {"surface": 0.61198, "split": 0.81367, "merge": 0.22321, "skel": 0.54342}, "mean_gate": {"surface": 0.1215, "split": 0.12156, "merge": 0.12134, "skel": 0.12179}, "n_reads": 16360, "n_cand": 27794, "read_suppressed_frac": 0.1012, "dropout_reads_logged": 16360, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 381, "admitted_surface": 15998, "staged_surface": 0, "admitted_split": 16409, "staged_split": 0, "admitted_merge": 21829, "staged_merge": 0, "admitted_skel": 11300, "staged_skel": 0, "mean_count": 57.16796875, "n_util_obs": 24, "util_count_spearman": -0.34956521739130436}, "probe_overhead_sec_total": 35.236, "probe_overhead_seq_frac": 0.4961, "util_loss": 0.000252, "alpha": 0.03, "t": 230.7} |
| {"event": "consolidate", "step": 250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 281.8} |
| {"event": "train", "step": 250, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.38125, "bpb": 3.43541, "chunk_loss": 0.031224, "F": 0.21082, "G": 0.28992, "chunks": 215.88, "bytes_per_chunk": 4.743, "boundary_entropy": 0.41777, "router_noise": 0.0, "grad_norm": 1.8564, "lr": 0.000251, "router_trainable": true, "sec_per_step": 1.038, "probe_sec": 0.1803, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.15669, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.01121, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.07443, "n_merge": 1, "mean_delta_merge": -0.08179, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -0.06024, "trunc_abs_frac": 1.08904, "mean_delta_full": 0.05137, "read_hit_rate": {"surface": 0.63554, "split": 0.83783, "merge": 0.22092, "skel": 0.61061}, "mean_gate": {"surface": 0.12358, "split": 0.12362, "merge": 0.12328, "skel": 0.12389}, "n_reads": 18549, "n_cand": 30158, "read_suppressed_frac": 0.0965, "dropout_reads_logged": 18549, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 464, "admitted_surface": 16141, "staged_surface": 0, "admitted_split": 16371, "staged_split": 0, "admitted_merge": 21728, "staged_merge": 0, "admitted_skel": 11296, "staged_skel": 0, "mean_count": 71.76887512207031, "n_util_obs": 25, "util_count_spearman": -0.13307692307692306}, "probe_overhead_sec_total": 44.481, "probe_overhead_seq_frac": 0.4969, "util_loss": 0.000358, "alpha": 0.03, "t": 282.6} |
| {"event": "probe", "step": 256, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3987.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.229098, "cps": 31.148438, "mean_p": 0.291555, "boundary_entropy": 0.384851, "bpb": 3.318141}, "tr": {"chunks": 4023.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.222228, "cps": 31.429688, "mean_p": 0.303017, "boundary_entropy": 0.442328, "bpb": 3.503191}, "fi": {"chunks": 4149.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.220562, "cps": 32.414062, "mean_p": 0.307122, "boundary_entropy": 0.472793, "bpb": 3.656966}}, "bpb": {"en": 3.31814, "tr": 3.50319, "fi": 3.65697}, "gini_cpc": 0.00847, "gini_cps": 0.008882, "bpic": 4.665844, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9700125778885781, "fi": 0.962740930244335}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 472, "admitted_surface": 16111, "staged_surface": 0, "admitted_split": 16312, "staged_split": 0, "admitted_merge": 21803, "staged_merge": 0, "admitted_skel": 11310, "staged_skel": 0, "mean_count": 73.27088928222656, "n_util_obs": 26, "util_count_spearman": -0.007863247863247864}, "t": 289.5} |
| {"event": "consolidate", "step": 300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 344.3} |
| {"event": "train", "step": 300, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.29716, "bpb": 3.31411, "chunk_loss": 0.03145, "F": 0.22522, "G": 0.27525, "chunks": 230.62, "bytes_per_chunk": 4.44, "boundary_entropy": 0.34906, "router_noise": 0.0, "grad_norm": 1.2201, "lr": 0.0003, "router_trainable": true, "sec_per_step": 1.257, "probe_sec": 0.1691, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.14383, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": 0.00655, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.01581, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -0.12592, "trunc_abs_frac": 0.75556, "mean_delta_full": 0.04411, "read_hit_rate": {"surface": 0.64222, "split": 0.84815, "merge": 0.22089, "skel": 0.61883}, "mean_gate": {"surface": 0.13479, "split": 0.13502, "merge": 0.13366, "skel": 0.13652}, "n_reads": 20003, "n_cand": 32175, "read_suppressed_frac": 0.0972, "dropout_reads_logged": 20003, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 561, "admitted_surface": 16107, "staged_surface": 0, "admitted_split": 16121, "staged_split": 0, "admitted_merge": 22055, "staged_merge": 0, "admitted_skel": 11253, "staged_skel": 0, "mean_count": 87.11204528808594, "n_util_obs": 36, "util_count_spearman": -0.09987129987129988}, "probe_overhead_sec_total": 53.28, "probe_overhead_seq_frac": 0.4974, "util_loss": 0.000533, "alpha": 0.03, "t": 345.5} |
| {"event": "consolidate", "step": 350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 398.2} |
| {"event": "train", "step": 350, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.22979, "bpb": 3.21691, "chunk_loss": 0.031376, "F": 0.23172, "G": 0.26308, "chunks": 237.28, "bytes_per_chunk": 4.316, "boundary_entropy": 0.29783, "router_noise": 0.0, "grad_norm": 1.2758, "lr": 0.0002999, "router_trainable": true, "sec_per_step": 1.073, "probe_sec": 0.161, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.13846, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.03951, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": 0.03632, "n_merge": 1, "mean_delta_merge": -0.03702, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -3.78065, "trunc_abs_frac": 0.51263, "mean_delta_full": -0.00757, "read_hit_rate": {"surface": 0.65732, "split": 0.8568, "merge": 0.23483, "skel": 0.641}, "mean_gate": {"surface": 0.1528, "split": 0.15274, "merge": 0.14899, "skel": 0.15667}, "n_reads": 20964, "n_cand": 33028, "read_suppressed_frac": 0.0979, "dropout_reads_logged": 20964, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 647, "admitted_surface": 16166, "staged_surface": 0, "admitted_split": 16126, "staged_split": 0, "admitted_merge": 21983, "staged_merge": 0, "admitted_skel": 11261, "staged_skel": 0, "mean_count": 102.59231567382812, "n_util_obs": 42, "util_count_spearman": 0.04837533425168139}, "probe_overhead_sec_total": 62.826, "probe_overhead_seq_frac": 0.4978, "util_loss": 0.00084, "alpha": 0.03, "t": 399.1} |
| {"event": "consolidate", "step": 400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 449.6} |
| {"event": "probe", "step": 400, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 4229.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.243004, "cps": 33.039062, "mean_p": 0.261096, "boundary_entropy": 0.263276, "bpb": 2.880941}, "tr": {"chunks": 4349.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.240236, "cps": 33.976562, "mean_p": 0.257206, "boundary_entropy": 0.317181, "bpb": 3.190844}, "fi": {"chunks": 4514.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.239966, "cps": 35.265625, "mean_p": 0.266273, "boundary_entropy": 0.350067, "bpb": 3.352804}}, "bpb": {"en": 2.88094, "tr": 3.19084, "fi": 3.3528}, "gini_cpc": 0.002801, "gini_cps": 0.014513, "bpic": 4.333333, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9886106650670691, "fi": 0.9874977308787004}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 733, "admitted_surface": 16092, "staged_surface": 0, "admitted_split": 15978, "staged_split": 0, "admitted_merge": 22220, "staged_merge": 0, "admitted_skel": 11246, "staged_skel": 0, "mean_count": 118.38560485839844, "n_util_obs": 48, "util_count_spearman": 0.14687364307425096}, "t": 451.8} |
| {"event": "eval", "step": 400, "phase": 1, "bpb": {"en": 2.89448, "tr": 3.19977, "fi": 3.34435}, "bytes_per_chunk": {"en": 4.065, "tr": 4.701, "fi": 4.588}, "bpb_mean": 3.1462, "frozen_inventory": true, "t": 455.6} |
| {"event": "train", "step": 400, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.15138, "bpb": 3.10379, "chunk_loss": 0.031596, "F": 0.23935, "G": 0.26084, "chunks": 245.09, "bytes_per_chunk": 4.178, "boundary_entropy": 0.28494, "router_noise": 0.0, "grad_norm": 1.0892, "lr": 0.0002998, "router_trainable": true, "sec_per_step": 1.175, "probe_sec": 0.2029, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.17999, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.05176, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": 0.06158, "n_merge": 1, "mean_delta_merge": 0.08643, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 2.88727, "trunc_abs_frac": 0.57347, "mean_delta_full": -0.04061, "read_hit_rate": {"surface": 0.67904, "split": 0.86252, "merge": 0.24682, "skel": 0.66528}, "mean_gate": {"surface": 0.1661, "split": 0.16489, "merge": 0.16099, "skel": 0.16939}, "n_reads": 21964, "n_cand": 33944, "read_suppressed_frac": 0.0997, "dropout_reads_logged": 21964, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 735, "admitted_surface": 16096, "staged_surface": 0, "admitted_split": 16092, "staged_split": 0, "admitted_merge": 22091, "staged_merge": 0, "admitted_skel": 11257, "staged_skel": 0, "mean_count": 118.70260620117188, "n_util_obs": 48, "util_count_spearman": 0.17792010421189752}, "probe_overhead_sec_total": 72.528, "probe_overhead_seq_frac": 0.4981, "util_loss": 0.001308, "alpha": 0.03, "t": 457.9} |
| {"event": "consolidate", "step": 450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 511.2} |
| {"event": "train", "step": 450, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.03224, "bpb": 2.93191, "chunk_loss": 0.031596, "F": 0.24133, "G": 0.26207, "chunks": 247.12, "bytes_per_chunk": 4.144, "boundary_entropy": 0.25742, "router_noise": 0.0, "grad_norm": 1.2514, "lr": 0.0002995, "router_trainable": true, "sec_per_step": 1.092, "probe_sec": 0.2337, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.19963, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.01499, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.1629, "n_merge": 1, "mean_delta_merge": 0.16609, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.04039, "trunc_abs_frac": 0.85113, "mean_delta_full": -0.14355, "read_hit_rate": {"surface": 0.68728, "split": 0.87283, "merge": 0.27104, "skel": 0.66265}, "mean_gate": {"surface": 0.17374, "split": 0.17057, "merge": 0.16749, "skel": 0.17395}, "n_reads": 22526, "n_cand": 34209, "read_suppressed_frac": 0.1011, "dropout_reads_logged": 22526, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 813, "admitted_surface": 16110, "staged_surface": 0, "admitted_split": 15914, "staged_split": 0, "admitted_merge": 22219, "staged_merge": 0, "admitted_skel": 11293, "staged_skel": 0, "mean_count": 135.51470947265625, "n_util_obs": 57, "util_count_spearman": 0.08756805807622504}, "probe_overhead_sec_total": 82.36, "probe_overhead_seq_frac": 0.4983, "util_loss": 0.001627, "alpha": 0.03, "t": 512.5} |
| {"event": "consolidate", "step": 500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 566.4} |
| {"event": "train", "step": 500, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.962, "bpb": 2.83057, "chunk_loss": 0.031792, "F": 0.2496, "G": 0.26396, "chunks": 255.59, "bytes_per_chunk": 4.006, "boundary_entropy": 0.24122, "router_noise": 0.0, "grad_norm": 1.0124, "lr": 0.0002992, "router_trainable": true, "sec_per_step": 1.1, "probe_sec": 0.2263, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.20208, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.11751, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.02148, "n_merge": 2, "mean_delta_merge": -0.006, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 1.44225, "trunc_abs_frac": 0.37664, "mean_delta_full": -0.05778, "read_hit_rate": {"surface": 0.69842, "split": 0.88013, "merge": 0.27702, "skel": 0.6768}, "mean_gate": {"surface": 0.17895, "split": 0.17427, "merge": 0.17223, "skel": 0.17645}, "n_reads": 23491, "n_cand": 35217, "read_suppressed_frac": 0.1016, "dropout_reads_logged": 23491, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 889, "admitted_surface": 16122, "staged_surface": 0, "admitted_split": 15804, "staged_split": 0, "admitted_merge": 22428, "staged_merge": 0, "admitted_skel": 11182, "staged_skel": 0, "mean_count": 152.87879943847656, "n_util_obs": 65, "util_count_spearman": -0.10026223776223776}, "probe_overhead_sec_total": 92.411, "probe_overhead_seq_frac": 0.4984, "util_loss": 0.00206, "alpha": 0.03, "t": 567.5} |
| {"event": "probe", "step": 512, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 4287.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.246337, "cps": 33.492188, "mean_p": 0.259233, "boundary_entropy": 0.230321, "bpb": 2.679221}, "tr": {"chunks": 4648.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.256753, "cps": 36.3125, "mean_p": 0.259521, "boundary_entropy": 0.275948, "bpb": 3.012475}, "fi": {"chunks": 5254.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.279305, "cps": 41.046875, "mean_p": 0.276284, "boundary_entropy": 0.291148, "bpb": 3.200679}}, "bpb": {"en": 2.67922, "tr": 3.01248, "fi": 3.20068}, "gini_cpc": 0.028091, "gini_cps": 0.045434, "bpic": 3.998309, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.042284320724884, "fi": 1.1338322919688868}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 903, "admitted_surface": 16082, "staged_surface": 0, "admitted_split": 15773, "staged_split": 0, "admitted_merge": 22532, "staged_merge": 0, "admitted_skel": 11149, "staged_skel": 0, "mean_count": 156.74375915527344, "n_util_obs": 65, "util_count_spearman": -0.07591783216783217}, "t": 582.1} |
| {"event": "consolidate", "step": 550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 620.1} |
| {"event": "train", "step": 550, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.90359, "bpb": 2.7463, "chunk_loss": 0.031716, "F": 0.24872, "G": 0.25952, "chunks": 254.69, "bytes_per_chunk": 4.021, "boundary_entropy": 0.22992, "router_noise": 0.0, "grad_norm": 1.1199, "lr": 0.0002987, "router_trainable": true, "sec_per_step": 1.065, "probe_sec": 0.1491, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.12361, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.02716, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 3, "mean_delta_split": -0.06057, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.0273, "trunc_abs_frac": 0.82006, "mean_delta_full": 0.23364, "read_hit_rate": {"surface": 0.7152, "split": 0.88645, "merge": 0.27381, "skel": 0.69956}, "mean_gate": {"surface": 0.1854, "split": 0.17818, "merge": 0.17848, "skel": 0.1809}, "n_reads": 23798, "n_cand": 35189, "read_suppressed_frac": 0.1027, "dropout_reads_logged": 23798, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 963, "admitted_surface": 16108, "staged_surface": 0, "admitted_split": 15793, "staged_split": 0, "admitted_merge": 22451, "staged_merge": 0, "admitted_skel": 11184, "staged_skel": 0, "mean_count": 170.6178741455078, "n_util_obs": 68, "util_count_spearman": 0.0373706913005306}, "probe_overhead_sec_total": 99.778, "probe_overhead_seq_frac": 0.4986, "util_loss": 0.002254, "alpha": 0.03, "t": 620.7} |
| {"event": "consolidate", "step": 600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 670.4} |
| {"event": "probe", "step": 600, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 4301.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.247141, "cps": 33.601562, "mean_p": 0.256718, "boundary_entropy": 0.212116, "bpb": 2.568815}, "tr": {"chunks": 4741.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.26189, "cps": 37.039062, "mean_p": 0.256408, "boundary_entropy": 0.250858, "bpb": 2.903585}, "fi": {"chunks": 5229.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.277976, "cps": 40.851562, "mean_p": 0.276281, "boundary_entropy": 0.266273, "bpb": 3.107492}}, "bpb": {"en": 2.56881, "tr": 2.90358, "fi": 3.10749}, "gini_cpc": 0.02612, "gini_cps": 0.043351, "bpic": 3.975335, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.059678398954095, "fi": 1.124764073754577}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1044, "admitted_surface": 16148, "staged_surface": 0, "admitted_split": 15787, "staged_split": 0, "admitted_merge": 22351, "staged_merge": 0, "admitted_skel": 11250, "staged_skel": 0, "mean_count": 188.2033233642578, "n_util_obs": 81, "util_count_spearman": 0.08168473351400181}, "t": 672.3} |
| {"event": "eval", "step": 600, "phase": 1, "bpb": {"en": 2.58012, "tr": 2.91431, "fi": 3.10918}, "bytes_per_chunk": {"en": 3.99, "tr": 4.354, "fi": 3.847}, "bpb_mean": 2.86787, "frozen_inventory": true, "t": 676.3} |
| {"event": "train", "step": 600, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.83253, "bpb": 2.64379, "chunk_loss": 0.031703, "F": 0.24988, "G": 0.25855, "chunks": 255.88, "bytes_per_chunk": 4.002, "boundary_entropy": 0.21352, "router_noise": 0.0, "grad_norm": 1.2502, "lr": 0.0002982, "router_trainable": true, "sec_per_step": 1.152, "probe_sec": 0.1796, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.15494, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.11676, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.01587, "n_merge": 1, "mean_delta_merge": -0.05566, "n_dropped_unpaired": 0, "n_full": 7, "trunc_ratio_sum": 0.90671, "trunc_abs_frac": 1.02882, "mean_delta_full": 0.12558, "read_hit_rate": {"surface": 0.71812, "split": 0.90056, "merge": 0.28188, "skel": 0.7005}, "mean_gate": {"surface": 0.1891, "split": 0.17921, "merge": 0.18123, "skel": 0.1793}, "n_reads": 24251, "n_cand": 35412, "read_suppressed_frac": 0.0983, "dropout_reads_logged": 24251, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1045, "admitted_surface": 16115, "staged_surface": 0, "admitted_split": 15667, "staged_split": 0, "admitted_merge": 22515, "staged_merge": 0, "admitted_skel": 11239, "staged_skel": 0, "mean_count": 188.57595825195312, "n_util_obs": 81, "util_count_spearman": 0.08274616079494128}, "probe_overhead_sec_total": 109.88, "probe_overhead_seq_frac": 0.4987, "util_loss": 0.002675, "alpha": 0.03, "t": 678.3} |
| {"event": "consolidate", "step": 650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 729.5} |
| {"event": "train", "step": 650, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.81963, "bpb": 2.62516, "chunk_loss": 0.031625, "F": 0.24878, "G": 0.25452, "chunks": 254.75, "bytes_per_chunk": 4.02, "boundary_entropy": 0.19858, "router_noise": 0.0, "grad_norm": 1.1194, "lr": 0.0002975, "router_trainable": true, "sec_per_step": 1.041, "probe_sec": 0.1455, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.11648, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.09708, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": 0.11337, "n_merge": 2, "mean_delta_merge": -0.21332, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 2.06447, "trunc_abs_frac": 1.33278, "mean_delta_full": 0.0116, "read_hit_rate": {"surface": 0.72069, "split": 0.88926, "merge": 0.30786, "skel": 0.6991}, "mean_gate": {"surface": 0.19544, "split": 0.18248, "merge": 0.18632, "skel": 0.18287}, "n_reads": 24186, "n_cand": 35214, "read_suppressed_frac": 0.1044, "dropout_reads_logged": 24186, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1121, "admitted_surface": 16084, "staged_surface": 0, "admitted_split": 15761, "staged_split": 0, "admitted_merge": 22424, "staged_merge": 0, "admitted_skel": 11267, "staged_skel": 0, "mean_count": 206.77999877929688, "n_util_obs": 95, "util_count_spearman": 0.13919372900335947}, "probe_overhead_sec_total": 119.7, "probe_overhead_seq_frac": 0.4988, "util_loss": 0.002973, "alpha": 0.03, "t": 730.4} |
| {"event": "consolidate", "step": 700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 779.8} |
| {"event": "train", "step": 700, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.70432, "bpb": 2.45882, "chunk_loss": 0.031744, "F": 0.25244, "G": 0.25788, "chunks": 258.5, "bytes_per_chunk": 3.961, "boundary_entropy": 0.19481, "router_noise": 0.0, "grad_norm": 1.0181, "lr": 0.0002967, "router_trainable": true, "sec_per_step": 1.013, "probe_sec": 0.2177, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.19083, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.00164, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.09149, "n_merge": 1, "mean_delta_merge": 0.06403, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -0.20434, "trunc_abs_frac": 2.18609, "mean_delta_full": -0.10116, "read_hit_rate": {"surface": 0.74478, "split": 0.90176, "merge": 0.31116, "skel": 0.72724}, "mean_gate": {"surface": 0.19949, "split": 0.18386, "merge": 0.18904, "skel": 0.18506}, "n_reads": 25286, "n_cand": 35967, "read_suppressed_frac": 0.0991, "dropout_reads_logged": 25286, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1182, "admitted_surface": 16047, "staged_surface": 0, "admitted_split": 15654, "staged_split": 0, "admitted_merge": 22606, "staged_merge": 0, "admitted_skel": 11229, "staged_skel": 0, "mean_count": 225.2954864501953, "n_util_obs": 104, "util_count_spearman": 0.09397204737010562}, "probe_overhead_sec_total": 129.785, "probe_overhead_seq_frac": 0.4989, "util_loss": 0.003325, "alpha": 0.03, "t": 781.1} |
| {"event": "consolidate", "step": 750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 830.0} |
| {"event": "train", "step": 750, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.75304, "bpb": 2.52909, "chunk_loss": 0.031952, "F": 0.25882, "G": 0.2614, "chunks": 265.03, "bytes_per_chunk": 3.864, "boundary_entropy": 0.18458, "router_noise": 0.0, "grad_norm": 1.0075, "lr": 0.0002959, "router_trainable": true, "sec_per_step": 0.994, "probe_sec": 0.1087, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07599, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.07113, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.01123, "n_merge": 1, "mean_delta_merge": -0.87363, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 1.17699, "trunc_abs_frac": 0.75844, "mean_delta_full": -0.40755, "read_hit_rate": {"surface": 0.73666, "split": 0.90164, "merge": 0.32078, "skel": 0.72461}, "mean_gate": {"surface": 0.20595, "split": 0.1863, "merge": 0.19269, "skel": 0.18627}, "n_reads": 25622, "n_cand": 36505, "read_suppressed_frac": 0.1002, "dropout_reads_logged": 25622, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1256, "admitted_surface": 16112, "staged_surface": 0, "admitted_split": 15560, "staged_split": 0, "admitted_merge": 22692, "staged_merge": 0, "admitted_skel": 11172, "staged_skel": 0, "mean_count": 244.0406494140625, "n_util_obs": 114, "util_count_spearman": 0.1066442558878921}, "probe_overhead_sec_total": 140.072, "probe_overhead_seq_frac": 0.499, "util_loss": 0.003721, "alpha": 0.03, "t": 830.7} |
| {"event": "consolidate", "step": 800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 881.2} |
| {"event": "probe", "step": 800, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 4288.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.246394, "cps": 33.5, "mean_p": 0.252634, "boundary_entropy": 0.179696, "bpb": 2.39566}, "tr": {"chunks": 4933.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.272496, "cps": 38.539062, "mean_p": 0.254849, "boundary_entropy": 0.208501, "bpb": 2.722013}, "fi": {"chunks": 5341.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.28393, "cps": 41.726562, "mean_p": 0.278327, "boundary_entropy": 0.228332, "bpb": 2.938837}}, "bpb": {"en": 2.39566, "tr": 2.72201, "fi": 2.93884}, "gini_cpc": 0.03117, "gini_cps": 0.048208, "bpic": 3.895893, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.1059357766008933, "fi": 1.1523384097864302}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1329, "admitted_surface": 16071, "staged_surface": 0, "admitted_split": 15494, "staged_split": 0, "admitted_merge": 22864, "staged_merge": 0, "admitted_skel": 11107, "staged_skel": 0, "mean_count": 262.3577575683594, "n_util_obs": 123, "util_count_spearman": 0.11385123370006836}, "t": 883.3} |
| {"event": "eval", "step": 800, "phase": 1, "bpb": {"en": 2.4088, "tr": 2.73123, "fi": 2.94305}, "bytes_per_chunk": {"en": 4.017, "tr": 4.099, "fi": 3.763}, "bpb_mean": 2.69436, "frozen_inventory": true, "t": 887.2} |
| {"event": "train", "step": 800, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.73536, "bpb": 2.5036, "chunk_loss": 0.031532, "F": 0.24533, "G": 0.25015, "chunks": 251.22, "bytes_per_chunk": 4.076, "boundary_entropy": 0.18273, "router_noise": 0.0, "grad_norm": 0.993, "lr": 0.0002949, "router_trainable": true, "sec_per_step": 1.182, "probe_sec": 0.1801, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.14481, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.04868, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": -0.06964, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.1258, "trunc_abs_frac": 0.54141, "mean_delta_full": -0.30688, "read_hit_rate": {"surface": 0.7185, "split": 0.89572, "merge": 0.29404, "skel": 0.6998}, "mean_gate": {"surface": 0.20836, "split": 0.18614, "merge": 0.19528, "skel": 0.19013}, "n_reads": 24099, "n_cand": 35067, "read_suppressed_frac": 0.1002, "dropout_reads_logged": 24099, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1330, "admitted_surface": 16118, "staged_surface": 0, "admitted_split": 15569, "staged_split": 0, "admitted_merge": 22661, "staged_merge": 0, "admitted_skel": 11188, "staged_skel": 0, "mean_count": 262.7143859863281, "n_util_obs": 123, "util_count_spearman": 0.11384478466677846}, "probe_overhead_sec_total": 150.404, "probe_overhead_seq_frac": 0.499, "util_loss": 0.00397, "alpha": 0.03, "t": 889.8} |
| {"event": "consolidate", "step": 850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 939.6} |
| {"event": "train", "step": 850, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.75324, "bpb": 2.5294, "chunk_loss": 0.032324, "F": 0.26282, "G": 0.2628, "chunks": 269.12, "bytes_per_chunk": 3.805, "boundary_entropy": 0.18073, "router_noise": 0.0, "grad_norm": 0.9419, "lr": 0.0002938, "router_trainable": true, "sec_per_step": 1.021, "probe_sec": 0.1915, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.15194, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": 0.05564, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.27507, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.12925, "trunc_abs_frac": 0.21443, "mean_delta_full": 0.31857, "read_hit_rate": {"surface": 0.73147, "split": 0.90199, "merge": 0.30604, "skel": 0.71831}, "mean_gate": {"surface": 0.20905, "split": 0.18399, "merge": 0.19328, "skel": 0.18609}, "n_reads": 25626, "n_cand": 36850, "read_suppressed_frac": 0.0983, "dropout_reads_logged": 25626, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1398, "admitted_surface": 16085, "staged_surface": 0, "admitted_split": 15519, "staged_split": 0, "admitted_merge": 22778, "staged_merge": 0, "admitted_skel": 11154, "staged_skel": 0, "mean_count": 281.43218994140625, "n_util_obs": 133, "util_count_spearman": 0.1387153501742466}, "probe_overhead_sec_total": 160.373, "probe_overhead_seq_frac": 0.4991, "util_loss": 0.004199, "alpha": 0.03, "t": 940.9} |
| {"event": "consolidate", "step": 900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 990.8} |
| {"event": "train", "step": 900, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.67825, "bpb": 2.4212, "chunk_loss": 0.031581, "F": 0.24893, "G": 0.25182, "chunks": 254.91, "bytes_per_chunk": 4.017, "boundary_entropy": 0.17259, "router_noise": 0.0, "grad_norm": 0.895, "lr": 0.0002927, "router_trainable": true, "sec_per_step": 1.017, "probe_sec": 0.2426, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.21758, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.07745, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.02246, "n_merge": 1, "mean_delta_merge": 0.16678, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -0.62573, "trunc_abs_frac": 0.78746, "mean_delta_full": 0.02087, "read_hit_rate": {"surface": 0.73637, "split": 0.90051, "merge": 0.30112, "skel": 0.71375}, "mean_gate": {"surface": 0.21615, "split": 0.18581, "merge": 0.19925, "skel": 0.19031}, "n_reads": 24542, "n_cand": 35287, "read_suppressed_frac": 0.1007, "dropout_reads_logged": 24542, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1461, "admitted_surface": 16094, "staged_surface": 0, "admitted_split": 15530, "staged_split": 0, "admitted_merge": 22707, "staged_merge": 0, "admitted_skel": 11205, "staged_skel": 0, "mean_count": 300.16326904296875, "n_util_obs": 140, "util_count_spearman": 0.17159038726465636}, "probe_overhead_sec_total": 170.464, "probe_overhead_seq_frac": 0.4991, "util_loss": 0.004503, "alpha": 0.03, "t": 991.8} |
| {"event": "consolidate", "step": 950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1040.2} |
| {"event": "train", "step": 950, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.69311, "bpb": 2.44264, "chunk_loss": 0.031496, "F": 0.24472, "G": 0.25164, "chunks": 250.59, "bytes_per_chunk": 4.086, "boundary_entropy": 0.1756, "router_noise": 0.0, "grad_norm": 1.2279, "lr": 0.0002914, "router_trainable": true, "sec_per_step": 0.997, "probe_sec": 0.2659, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.24111, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.04149, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": -0.01553, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -0.20498, "trunc_abs_frac": 0.464, "mean_delta_full": 0.14507, "read_hit_rate": {"surface": 0.71817, "split": 0.89117, "merge": 0.29478, "skel": 0.7004}, "mean_gate": {"surface": 0.21659, "split": 0.18241, "merge": 0.19797, "skel": 0.18847}, "n_reads": 23884, "n_cand": 34891, "read_suppressed_frac": 0.1009, "dropout_reads_logged": 23884, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1535, "admitted_surface": 16107, "staged_surface": 0, "admitted_split": 15568, "staged_split": 0, "admitted_merge": 22574, "staged_merge": 0, "admitted_skel": 11287, "staged_skel": 0, "mean_count": 318.76666259765625, "n_util_obs": 146, "util_count_spearman": 0.17896522803736517}, "probe_overhead_sec_total": 180.024, "probe_overhead_seq_frac": 0.4992, "util_loss": 0.004513, "alpha": 0.03, "t": 1041.6} |
| {"event": "consolidate", "step": 1000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1091.4} |
| {"event": "probe", "step": 1000, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 4270.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.24536, "cps": 33.359375, "mean_p": 0.249948, "boundary_entropy": 0.164704, "bpb": 2.313055}, "tr": {"chunks": 4910.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.271226, "cps": 38.359375, "mean_p": 0.253512, "boundary_entropy": 0.191316, "bpb": 2.62079}, "fi": {"chunks": 5502.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.292488, "cps": 42.984375, "mean_p": 0.280618, "boundary_entropy": 0.196833, "bpb": 2.797087}}, "bpb": {"en": 2.31305, "tr": 2.62079, "fi": 2.79709}, "gini_cpc": 0.038833, "gini_cps": 0.055942, "bpic": 3.864051, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.1054196640328096, "fi": 1.1920787540600155}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1603, "admitted_surface": 16077, "staged_surface": 0, "admitted_split": 15510, "staged_split": 0, "admitted_merge": 22751, "staged_merge": 0, "admitted_skel": 11198, "staged_skel": 0, "mean_count": 337.0010986328125, "n_util_obs": 159, "util_count_spearman": 0.2621865297348937}, "t": 1093.6} |
| {"event": "eval", "step": 1000, "phase": 1, "bpb": {"en": 2.32545, "tr": 2.62795, "fi": 2.80343}, "bytes_per_chunk": {"en": 4.015, "tr": 4.11, "fi": 3.632}, "bpb_mean": 2.58561, "frozen_inventory": true, "t": 1097.5} |
| {"event": "train", "step": 1000, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.70689, "bpb": 2.46252, "chunk_loss": 0.031717, "F": 0.25293, "G": 0.25501, "chunks": 259.0, "bytes_per_chunk": 3.954, "boundary_entropy": 0.16777, "router_noise": 0.0, "grad_norm": 1.0136, "lr": 0.0002901, "router_trainable": true, "sec_per_step": 1.162, "probe_sec": 0.6638, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.62755, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.08143, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": -0.11884, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.88681, "trunc_abs_frac": 0.41753, "mean_delta_full": 0.15057, "read_hit_rate": {"surface": 0.73159, "split": 0.9, "merge": 0.3108, "skel": 0.71329}, "mean_gate": {"surface": 0.22068, "split": 0.18541, "merge": 0.19819, "skel": 0.19198}, "n_reads": 25030, "n_cand": 35931, "read_suppressed_frac": 0.1, "dropout_reads_logged": 25030, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1604, "admitted_surface": 16105, "staged_surface": 0, "admitted_split": 15534, "staged_split": 0, "admitted_merge": 22661, "staged_merge": 0, "admitted_skel": 11236, "staged_skel": 0, "mean_count": 337.385498046875, "n_util_obs": 159, "util_count_spearman": 0.2625895629328875}, "probe_overhead_sec_total": 190.249, "probe_overhead_seq_frac": 0.4992, "util_loss": 0.004846, "alpha": 0.03, "t": 1099.7} |
| {"event": "probe", "step": 1024, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 4267.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.245188, "cps": 33.335938, "mean_p": 0.248971, "boundary_entropy": 0.160116, "bpb": 2.299288}, "tr": {"chunks": 4926.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.27211, "cps": 38.484375, "mean_p": 0.252489, "boundary_entropy": 0.182616, "bpb": 2.589304}, "fi": {"chunks": 5436.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.28898, "cps": 42.46875, "mean_p": 0.279749, "boundary_entropy": 0.19789, "bpb": 2.769259}}, "bpb": {"en": 2.29929, "tr": 2.5893, "fi": 2.76926}, "gini_cpc": 0.036209, "gini_cps": 0.053273, "bpic": 3.87805, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.1098015663074023, "fi": 1.1786070700919862}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1635, "admitted_surface": 16145, "staged_surface": 0, "admitted_split": 15485, "staged_split": 0, "admitted_merge": 22661, "staged_merge": 0, "admitted_skel": 11245, "staged_skel": 0, "mean_count": 345.94146728515625, "n_util_obs": 164, "util_count_spearman": 0.25354427108130534}, "t": 1117.7} |
| {"event": "consolidate", "step": 1050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1141.9} |
| {"event": "train", "step": 1050, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.6482, "bpb": 2.37785, "chunk_loss": 0.031814, "F": 0.25638, "G": 0.25702, "chunks": 262.53, "bytes_per_chunk": 3.9, "boundary_entropy": 0.16412, "router_noise": 0.0, "grad_norm": 0.858, "lr": 0.0002886, "router_trainable": true, "sec_per_step": 0.86, "probe_sec": 0.1746, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.14226, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.00539, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.08667, "n_merge": 1, "mean_delta_merge": -0.12323, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.25378, "trunc_abs_frac": 0.46884, "mean_delta_full": -0.12921, "read_hit_rate": {"surface": 0.74513, "split": 0.90236, "merge": 0.31738, "skel": 0.73006}, "mean_gate": {"surface": 0.21826, "split": 0.17621, "merge": 0.20045, "skel": 0.18442}, "n_reads": 25486, "n_cand": 36208, "read_suppressed_frac": 0.1005, "dropout_reads_logged": 25486, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1670, "admitted_surface": 16064, "staged_surface": 0, "admitted_split": 15504, "staged_split": 0, "admitted_merge": 22803, "staged_merge": 0, "admitted_skel": 11165, "staged_skel": 0, "mean_count": 356.0977478027344, "n_util_obs": 169, "util_count_spearman": 0.2187235841081995}, "probe_overhead_sec_total": 197.441, "probe_overhead_seq_frac": 0.4993, "util_loss": 0.004966, "alpha": 0.03, "t": 1142.7} |
| {"event": "consolidate", "step": 1100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1190.3} |
| {"event": "train", "step": 1100, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.64655, "bpb": 2.37548, "chunk_loss": 0.031666, "F": 0.25089, "G": 0.25314, "chunks": 256.91, "bytes_per_chunk": 3.986, "boundary_entropy": 0.16045, "router_noise": 0.0, "grad_norm": 0.8275, "lr": 0.0002871, "router_trainable": true, "sec_per_step": 0.972, "probe_sec": 0.2395, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.19761, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": 0.02886, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.02191, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.18624, "trunc_abs_frac": 0.57352, "mean_delta_full": 0.21847, "read_hit_rate": {"surface": 0.73672, "split": 0.89403, "merge": 0.3212, "skel": 0.71553}, "mean_gate": {"surface": 0.22199, "split": 0.1778, "merge": 0.19645, "skel": 0.18733}, "n_reads": 24798, "n_cand": 35527, "read_suppressed_frac": 0.1006, "dropout_reads_logged": 24798, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1733, "admitted_surface": 16059, "staged_surface": 0, "admitted_split": 15556, "staged_split": 0, "admitted_merge": 22736, "staged_merge": 0, "admitted_skel": 11185, "staged_skel": 0, "mean_count": 374.81195068359375, "n_util_obs": 175, "util_count_spearman": 0.16919390953873711}, "probe_overhead_sec_total": 206.489, "probe_overhead_seq_frac": 0.4993, "util_loss": 0.005017, "alpha": 0.03, "t": 1191.3} |
| {"event": "consolidate", "step": 1150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1237.2} |
| {"event": "train", "step": 1150, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.66834, "bpb": 2.40691, "chunk_loss": 0.031608, "F": 0.25009, "G": 0.25145, "chunks": 256.09, "bytes_per_chunk": 3.999, "boundary_entropy": 0.16082, "router_noise": 0.0, "grad_norm": 0.8834, "lr": 0.0002855, "router_trainable": true, "sec_per_step": 0.934, "probe_sec": 0.1274, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09982, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.00346, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": -0.05652, "n_merge": 1, "mean_delta_merge": -0.23297, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.3533, "trunc_abs_frac": 9.01262, "mean_delta_full": 0.22467, "read_hit_rate": {"surface": 0.72081, "split": 0.89513, "merge": 0.28643, "skel": 0.69894}, "mean_gate": {"surface": 0.22911, "split": 0.17984, "merge": 0.20545, "skel": 0.18762}, "n_reads": 24275, "n_cand": 35474, "read_suppressed_frac": 0.0987, "dropout_reads_logged": 24275, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1791, "admitted_surface": 16105, "staged_surface": 0, "admitted_split": 15416, "staged_split": 0, "admitted_merge": 22819, "staged_merge": 0, "admitted_skel": 11196, "staged_skel": 0, "mean_count": 393.67816162109375, "n_util_obs": 187, "util_count_spearman": 0.12985072983458062}, "probe_overhead_sec_total": 216.102, "probe_overhead_seq_frac": 0.4993, "util_loss": 0.005172, "alpha": 0.03, "t": 1238.0} |
| {"event": "consolidate", "step": 1200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1284.7} |
| {"event": "probe", "step": 1200, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 4311.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.247716, "cps": 33.679688, "mean_p": 0.24773, "boundary_entropy": 0.154316, "bpb": 2.246186}, "tr": {"chunks": 4841.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.267414, "cps": 37.820312, "mean_p": 0.241447, "boundary_entropy": 0.197284, "bpb": 2.530405}, "fi": {"chunks": 5329.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.283292, "cps": 41.632812, "mean_p": 0.278143, "boundary_entropy": 0.193828, "bpb": 2.692133}}, "bpb": {"en": 2.24619, "tr": 2.5304, "fi": 2.69213}, "gini_cpc": 0.029705, "gini_cps": 0.046866, "bpic": 3.917685, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0795198402891426, "fi": 1.143615239857844}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1842, "admitted_surface": 16119, "staged_surface": 0, "admitted_split": 15425, "staged_split": 0, "admitted_merge": 22848, "staged_merge": 0, "admitted_skel": 11144, "staged_skel": 0, "mean_count": 412.11517333984375, "n_util_obs": 197, "util_count_spearman": 0.055988246706158235}, "t": 1286.9} |
| {"event": "eval", "step": 1200, "phase": 1, "bpb": {"en": 2.25809, "tr": 2.53658, "fi": 2.69936}, "bytes_per_chunk": {"en": 4.003, "tr": 4.143, "fi": 3.767}, "bpb_mean": 2.49801, "frozen_inventory": true, "t": 1290.7} |
| {"event": "train", "step": 1200, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.62816, "bpb": 2.34894, "chunk_loss": 0.031874, "F": 0.25641, "G": 0.25737, "chunks": 262.56, "bytes_per_chunk": 3.9, "boundary_entropy": 0.15988, "router_noise": 0.0, "grad_norm": 0.7914, "lr": 0.0002837, "router_trainable": true, "sec_per_step": 1.09, "probe_sec": 0.1675, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.14274, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.04192, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 3, "mean_delta_merge": 0.04898, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 1.43138, "trunc_abs_frac": 0.85148, "mean_delta_full": -0.07693, "read_hit_rate": {"surface": 0.73226, "split": 0.89893, "merge": 0.30403, "skel": 0.71166}, "mean_gate": {"surface": 0.22441, "split": 0.17236, "merge": 0.19786, "skel": 0.18573}, "n_reads": 25086, "n_cand": 36180, "read_suppressed_frac": 0.0985, "dropout_reads_logged": 25086, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1842, "admitted_surface": 16129, "staged_surface": 0, "admitted_split": 15396, "staged_split": 0, "admitted_merge": 22830, "staged_merge": 0, "admitted_skel": 11181, "staged_skel": 0, "mean_count": 412.4990234375, "n_util_obs": 197, "util_count_spearman": 0.055680601728462495}, "probe_overhead_sec_total": 224.818, "probe_overhead_seq_frac": 0.4993, "util_loss": 0.005324, "alpha": 0.03, "t": 1292.5} |
| {"event": "consolidate", "step": 1250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1343.2} |
| {"event": "train", "step": 1250, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.63216, "bpb": 2.35471, "chunk_loss": 0.031619, "F": 0.25281, "G": 0.25056, "chunks": 258.88, "bytes_per_chunk": 3.956, "boundary_entropy": 0.15466, "router_noise": 0.0, "grad_norm": 0.8404, "lr": 0.0002819, "router_trainable": true, "sec_per_step": 1.044, "probe_sec": 0.1425, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10372, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": 0.04113, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.30487, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.22016, "trunc_abs_frac": 1.20865, "mean_delta_full": -0.19295, "read_hit_rate": {"surface": 0.72685, "split": 0.90136, "merge": 0.29246, "skel": 0.72371}, "mean_gate": {"surface": 0.22617, "split": 0.17165, "merge": 0.20152, "skel": 0.18682}, "n_reads": 24893, "n_cand": 35909, "read_suppressed_frac": 0.0997, "dropout_reads_logged": 24893, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1903, "admitted_surface": 16143, "staged_surface": 0, "admitted_split": 15400, "staged_split": 0, "admitted_merge": 22833, "staged_merge": 0, "admitted_skel": 11160, "staged_skel": 0, "mean_count": 431.4033508300781, "n_util_obs": 206, "util_count_spearman": 0.07945241208427281}, "probe_overhead_sec_total": 234.566, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.00549, "alpha": 0.03, "t": 1344.7} |
| {"event": "consolidate", "step": 1300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1390.4} |
| {"event": "train", "step": 1300, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.59151, "bpb": 2.29606, "chunk_loss": 0.031431, "F": 0.24734, "G": 0.24545, "chunks": 253.28, "bytes_per_chunk": 4.043, "boundary_entropy": 0.15482, "router_noise": 0.0, "grad_norm": 0.7944, "lr": 0.00028, "router_trainable": true, "sec_per_step": 0.932, "probe_sec": 0.137, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10655, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": -0.00578, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 3, "mean_delta_split": 0.00891, "n_merge": 2, "mean_delta_merge": -0.04367, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": -0.94553, "trunc_abs_frac": 0.94553, "mean_delta_full": 0.03137, "read_hit_rate": {"surface": 0.72724, "split": 0.89794, "merge": 0.30618, "skel": 0.71312}, "mean_gate": {"surface": 0.22966, "split": 0.17322, "merge": 0.20258, "skel": 0.19002}, "n_reads": 24457, "n_cand": 35233, "read_suppressed_frac": 0.1013, "dropout_reads_logged": 24457, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1971, "admitted_surface": 16124, "staged_surface": 0, "admitted_split": 15483, "staged_split": 0, "admitted_merge": 22734, "staged_merge": 0, "admitted_skel": 11195, "staged_skel": 0, "mean_count": 450.31903076171875, "n_util_obs": 213, "util_count_spearman": 0.08168691216244744}, "probe_overhead_sec_total": 243.346, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.005792, "alpha": 0.03, "t": 1391.3} |
| {"event": "consolidate", "step": 1350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1441.5} |
| {"event": "train", "step": 1350, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.61894, "bpb": 2.33564, "chunk_loss": 0.031912, "F": 0.26132, "G": 0.25552, "chunks": 267.59, "bytes_per_chunk": 3.827, "boundary_entropy": 0.15295, "router_noise": 0.0, "grad_norm": 0.8043, "lr": 0.000278, "router_trainable": true, "sec_per_step": 1.023, "probe_sec": 0.2587, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.22469, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.05798, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": -0.10971, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.11843, "trunc_abs_frac": 0.36114, "mean_delta_full": -0.1478, "read_hit_rate": {"surface": 0.75021, "split": 0.90517, "merge": 0.32263, "skel": 0.72887}, "mean_gate": {"surface": 0.23209, "split": 0.17283, "merge": 0.20393, "skel": 0.18867}, "n_reads": 25979, "n_cand": 36753, "read_suppressed_frac": 0.1005, "dropout_reads_logged": 25979, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2040, "admitted_surface": 16050, "staged_surface": 0, "admitted_split": 15408, "staged_split": 0, "admitted_merge": 22913, "staged_merge": 0, "admitted_skel": 11165, "staged_skel": 0, "mean_count": 468.9852294921875, "n_util_obs": 220, "util_count_spearman": 0.05677465161752027}, "probe_overhead_sec_total": 252.8, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.005721, "alpha": 0.03, "t": 1442.4} |
| {"event": "consolidate", "step": 1400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1491.3} |
| {"event": "probe", "step": 1400, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 4215.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.2422, "cps": 32.929688, "mean_p": 0.245547, "boundary_entropy": 0.144867, "bpb": 2.191099}, "tr": {"chunks": 4810.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.265702, "cps": 37.578125, "mean_p": 0.240512, "boundary_entropy": 0.188069, "bpb": 2.470704}, "fi": {"chunks": 5410.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.287598, "cps": 42.265625, "mean_p": 0.281941, "boundary_entropy": 0.185547, "bpb": 2.636355}}, "bpb": {"en": 2.1911, "tr": 2.4707, "fi": 2.63635}, "gini_cpc": 0.038046, "gini_cps": 0.05519, "bpic": 3.93017, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0970364716097665, "fi": 1.1874406793480987}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2105, "admitted_surface": 16073, "staged_surface": 0, "admitted_split": 15428, "staged_split": 0, "admitted_merge": 22819, "staged_merge": 0, "admitted_skel": 11216, "staged_skel": 0, "mean_count": 487.2125549316406, "n_util_obs": 227, "util_count_spearman": 0.04586207931312958}, "t": 1493.3} |
| {"event": "eval", "step": 1400, "phase": 1, "bpb": {"en": 2.20588, "tr": 2.47821, "fi": 2.6433}, "bytes_per_chunk": {"en": 4.032, "tr": 4.179, "fi": 3.725}, "bpb_mean": 2.44246, "frozen_inventory": true, "t": 1497.0} |
| {"event": "train", "step": 1400, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.60308, "bpb": 2.31276, "chunk_loss": 0.031483, "F": 0.24869, "G": 0.248, "chunks": 254.66, "bytes_per_chunk": 4.021, "boundary_entropy": 0.14705, "router_noise": 0.0, "grad_norm": 0.7985, "lr": 0.0002759, "router_trainable": true, "sec_per_step": 1.132, "probe_sec": 0.1829, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.15619, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.11546, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.29944, "n_merge": 3, "mean_delta_merge": 0.21268, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": -3.92147, "trunc_abs_frac": 3.92147, "mean_delta_full": -0.01904, "read_hit_rate": {"surface": 0.7302, "split": 0.896, "merge": 0.30167, "skel": 0.70757}, "mean_gate": {"surface": 0.23639, "split": 0.17431, "merge": 0.2057, "skel": 0.19098}, "n_reads": 24449, "n_cand": 35324, "read_suppressed_frac": 0.1021, "dropout_reads_logged": 24449, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2106, "admitted_surface": 16058, "staged_surface": 0, "admitted_split": 15474, "staged_split": 0, "admitted_merge": 22773, "staged_merge": 0, "admitted_skel": 11231, "staged_skel": 0, "mean_count": 487.5919494628906, "n_util_obs": 227, "util_count_spearman": 0.046221651356569665}, "probe_overhead_sec_total": 262.059, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.00665, "alpha": 0.03, "t": 1499.0} |
| {"event": "consolidate", "step": 1450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1546.2} |
| {"event": "train", "step": 1450, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.60609, "bpb": 2.3171, "chunk_loss": 0.031702, "F": 0.25601, "G": 0.253, "chunks": 262.16, "bytes_per_chunk": 3.906, "boundary_entropy": 0.15335, "router_noise": 0.0, "grad_norm": 0.8017, "lr": 0.0002738, "router_trainable": true, "sec_per_step": 0.961, "probe_sec": 0.1626, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.13768, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.00464, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.10498, "n_merge": 1, "mean_delta_merge": -0.19775, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": -0.16762, "trunc_abs_frac": 0.69901, "mean_delta_full": 0.04606, "read_hit_rate": {"surface": 0.74106, "split": 0.89855, "merge": 0.3194, "skel": 0.72315}, "mean_gate": {"surface": 0.22792, "split": 0.16505, "merge": 0.19736, "skel": 0.18283}, "n_reads": 25246, "n_cand": 36041, "read_suppressed_frac": 0.1003, "dropout_reads_logged": 25246, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2166, "admitted_surface": 16065, "staged_surface": 0, "admitted_split": 15404, "staged_split": 0, "admitted_merge": 22888, "staged_merge": 0, "admitted_skel": 11179, "staged_skel": 0, "mean_count": 506.3005065917969, "n_util_obs": 236, "util_count_spearman": 0.0844873942675072}, "probe_overhead_sec_total": 271.177, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.006523, "alpha": 0.03, "t": 1547.1} |
| {"event": "consolidate", "step": 1500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1595.4} |
| {"event": "train", "step": 1500, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.59582, "bpb": 2.30227, "chunk_loss": 0.031756, "F": 0.25479, "G": 0.25218, "chunks": 260.91, "bytes_per_chunk": 3.925, "boundary_entropy": 0.14633, "router_noise": 0.0, "grad_norm": 0.7529, "lr": 0.0002715, "router_trainable": true, "sec_per_step": 0.983, "probe_sec": 0.1948, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.16422, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.00153, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.14014, "n_merge": 2, "mean_delta_merge": -0.41556, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": -0.12877, "trunc_abs_frac": 0.12877, "mean_delta_full": -0.04456, "read_hit_rate": {"surface": 0.73512, "split": 0.9, "merge": 0.30103, "skel": 0.72338}, "mean_gate": {"surface": 0.22681, "split": 0.16147, "merge": 0.19455, "skel": 0.18108}, "n_reads": 25021, "n_cand": 35961, "read_suppressed_frac": 0.0974, "dropout_reads_logged": 25021, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2221, "admitted_surface": 16073, "staged_surface": 0, "admitted_split": 15390, "staged_split": 0, "admitted_merge": 22919, "staged_merge": 0, "admitted_skel": 11154, "staged_skel": 0, "mean_count": 524.8585815429688, "n_util_obs": 244, "util_count_spearman": 0.08780428972252491}, "probe_overhead_sec_total": 280.786, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.006652, "alpha": 0.03, "t": 1596.3} |
| {"event": "consolidate", "step": 1550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1632.0} |
| {"event": "train", "step": 1550, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.46413, "bpb": 2.11229, "chunk_loss": 0.031455, "F": 0.245, "G": 0.24925, "chunks": 250.88, "bytes_per_chunk": 4.082, "boundary_entropy": 0.14646, "router_noise": 0.0, "grad_norm": 0.7632, "lr": 0.0002692, "router_trainable": true, "sec_per_step": 0.725, "probe_sec": 0.0832, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05382, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.02321, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": -0.4252, "trunc_abs_frac": 0.4252, "mean_delta_full": 0.18359, "read_hit_rate": {"surface": 0.73187, "split": 0.89338, "merge": 0.31655, "skel": 0.71563}, "mean_gate": {"surface": 0.22322, "split": 0.1553, "merge": 0.19433, "skel": 0.17817}, "n_reads": 24274, "n_cand": 34883, "read_suppressed_frac": 0.099, "dropout_reads_logged": 24274, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2291, "admitted_surface": 16096, "staged_surface": 0, "admitted_split": 15455, "staged_split": 0, "admitted_merge": 22789, "staged_merge": 0, "admitted_skel": 11196, "staged_skel": 0, "mean_count": 543.0960083007812, "n_util_obs": 253, "util_count_spearman": 0.11593461884458119}, "probe_overhead_sec_total": 286.348, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.006092, "alpha": 0.03, "t": 1632.5} |
| {"event": "consolidate", "step": 1600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1672.1} |
| {"event": "probe", "step": 1600, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 4222.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.242602, "cps": 32.984375, "mean_p": 0.240844, "boundary_entropy": 0.139481, "bpb": 2.159708}, "tr": {"chunks": 4478.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.247362, "cps": 34.984375, "mean_p": 0.235751, "boundary_entropy": 0.176428, "bpb": 2.409363}, "fi": {"chunks": 5333.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.283504, "cps": 41.664062, "mean_p": 0.277513, "boundary_entropy": 0.183248, "bpb": 2.567787}}, "bpb": {"en": 2.15971, "tr": 2.40936, "fi": 2.56779}, "gini_cpc": 0.035255, "gini_cps": 0.05278, "bpic": 4.042756, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0196225435938406, "fi": 1.1685992183182174}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2336, "admitted_surface": 16073, "staged_surface": 0, "admitted_split": 15444, "staged_split": 0, "admitted_merge": 22913, "staged_merge": 0, "admitted_skel": 11106, "staged_skel": 0, "mean_count": 561.0714111328125, "n_util_obs": 262, "util_count_spearman": 0.1270679930113021}, "t": 1674.0} |
| {"event": "eval", "step": 1600, "phase": 1, "bpb": {"en": 2.17151, "tr": 2.41958, "fi": 2.57173}, "bytes_per_chunk": {"en": 4.029, "tr": 4.615, "fi": 3.747}, "bpb_mean": 2.38761, "frozen_inventory": true, "t": 1676.9} |
| {"event": "train", "step": 1600, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.56451, "bpb": 2.25711, "chunk_loss": 0.031595, "F": 0.25198, "G": 0.25038, "chunks": 258.03, "bytes_per_chunk": 3.969, "boundary_entropy": 0.139, "router_noise": 0.0, "grad_norm": 0.7709, "lr": 0.0002668, "router_trainable": true, "sec_per_step": 0.925, "probe_sec": 0.1525, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10302, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.01003, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.00201, "n_merge": 1, "mean_delta_merge": 0.06065, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 11.72121, "trunc_abs_frac": 0.61058, "mean_delta_full": -0.00671, "read_hit_rate": {"surface": 0.7335, "split": 0.90143, "merge": 0.31161, "skel": 0.71519}, "mean_gate": {"surface": 0.22764, "split": 0.15587, "merge": 0.1955, "skel": 0.17987}, "n_reads": 24883, "n_cand": 35678, "read_suppressed_frac": 0.0995, "dropout_reads_logged": 24883, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2338, "admitted_surface": 16113, "staged_surface": 0, "admitted_split": 15349, "staged_split": 0, "admitted_merge": 22897, "staged_merge": 0, "admitted_skel": 11177, "staged_skel": 0, "mean_count": 561.45263671875, "n_util_obs": 262, "util_count_spearman": 0.1326988745785756}, "probe_overhead_sec_total": 292.506, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.006704, "alpha": 0.03, "t": 1678.8} |
| {"event": "consolidate", "step": 1650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1719.6} |
| {"event": "train", "step": 1650, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.49972, "bpb": 2.16363, "chunk_loss": 0.031231, "F": 0.24106, "G": 0.23885, "chunks": 246.84, "bytes_per_chunk": 4.148, "boundary_entropy": 0.13371, "router_noise": 0.0, "grad_norm": 0.7386, "lr": 0.0002643, "router_trainable": true, "sec_per_step": 0.83, "probe_sec": 0.1322, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10459, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.03476, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": 0.05473, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -3.62727, "trunc_abs_frac": 0.50974, "mean_delta_full": 0.0179, "read_hit_rate": {"surface": 0.72544, "split": 0.89655, "merge": 0.30453, "skel": 0.70941}, "mean_gate": {"surface": 0.22817, "split": 0.15618, "merge": 0.19422, "skel": 0.17937}, "n_reads": 23868, "n_cand": 34453, "read_suppressed_frac": 0.1015, "dropout_reads_logged": 23868, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2385, "admitted_surface": 16097, "staged_surface": 0, "admitted_split": 15437, "staged_split": 0, "admitted_merge": 22809, "staged_merge": 0, "admitted_skel": 11193, "staged_skel": 0, "mean_count": 579.7200317382812, "n_util_obs": 267, "util_count_spearman": 0.1376439386309481}, "probe_overhead_sec_total": 299.237, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.006427, "alpha": 0.03, "t": 1720.3} |
| {"event": "consolidate", "step": 1700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1759.2} |
| {"event": "train", "step": 1700, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.55175, "bpb": 2.2387, "chunk_loss": 0.031281, "F": 0.24081, "G": 0.24237, "chunks": 246.59, "bytes_per_chunk": 4.153, "boundary_entropy": 0.13523, "router_noise": 0.0, "grad_norm": 0.738, "lr": 0.0002618, "router_trainable": true, "sec_per_step": 0.79, "probe_sec": 0.0922, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06871, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.04496, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 3, "mean_delta_merge": -0.46605, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.77086, "trunc_abs_frac": 0.46419, "mean_delta_full": -0.0907, "read_hit_rate": {"surface": 0.71994, "split": 0.88717, "merge": 0.29309, "skel": 0.69468}, "mean_gate": {"surface": 0.22691, "split": 0.15381, "merge": 0.19405, "skel": 0.17944}, "n_reads": 23481, "n_cand": 34398, "read_suppressed_frac": 0.0994, "dropout_reads_logged": 23481, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2431, "admitted_surface": 16087, "staged_surface": 0, "admitted_split": 15480, "staged_split": 0, "admitted_merge": 22747, "staged_merge": 0, "admitted_skel": 11222, "staged_skel": 0, "mean_count": 597.9363403320312, "n_util_obs": 272, "util_count_spearman": 0.15195462232579662}, "probe_overhead_sec_total": 305.447, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.00697, "alpha": 0.03, "t": 1759.8} |
| {"event": "consolidate", "step": 1750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1801.1} |
| {"event": "train", "step": 1750, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.54811, "bpb": 2.23345, "chunk_loss": 0.03131, "F": 0.24207, "G": 0.24264, "chunks": 247.88, "bytes_per_chunk": 4.131, "boundary_entropy": 0.14246, "router_noise": 0.0, "grad_norm": 0.7698, "lr": 0.0002591, "router_trainable": true, "sec_per_step": 0.838, "probe_sec": 0.0972, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07106, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": 0.08976, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.10352, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.01204, "trunc_abs_frac": 0.82671, "mean_delta_full": -0.18147, "read_hit_rate": {"surface": 0.71962, "split": 0.89853, "merge": 0.28826, "skel": 0.70917}, "mean_gate": {"surface": 0.22519, "split": 0.15003, "merge": 0.19337, "skel": 0.17948}, "n_reads": 23792, "n_cand": 34579, "read_suppressed_frac": 0.0968, "dropout_reads_logged": 23792, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2501, "admitted_surface": 16127, "staged_surface": 0, "admitted_split": 15343, "staged_split": 0, "admitted_merge": 22863, "staged_merge": 0, "admitted_skel": 11203, "staged_skel": 0, "mean_count": 615.9793090820312, "n_util_obs": 276, "util_count_spearman": 0.1160437373958342}, "probe_overhead_sec_total": 312.478, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.006701, "alpha": 0.03, "t": 1801.7} |
| {"event": "consolidate", "step": 1800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1842.5} |
| {"event": "probe", "step": 1800, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 4285.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.246222, "cps": 33.476562, "mean_p": 0.242933, "boundary_entropy": 0.127277, "bpb": 2.115343}, "tr": {"chunks": 4311.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.238137, "cps": 33.679688, "mean_p": 0.231575, "boundary_entropy": 0.172533, "bpb": 2.363428}, "fi": {"chunks": 5166.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.274627, "cps": 40.359375, "mean_p": 0.27548, "boundary_entropy": 0.170119, "bpb": 2.50598}}, "bpb": {"en": 2.11534, "tr": 2.36343, "fi": 2.50598}, "gini_cpc": 0.032051, "gini_cps": 0.042678, "bpic": 4.122366, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9671654311053353, "fi": 1.1153619183194718}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2545, "admitted_surface": 16134, "staged_surface": 0, "admitted_split": 15425, "staged_split": 0, "admitted_merge": 22780, "staged_merge": 0, "admitted_skel": 11197, "staged_skel": 0, "mean_count": 633.84423828125, "n_util_obs": 287, "util_count_spearman": 0.09894682259312389}, "t": 1844.3} |
| {"event": "eval", "step": 1800, "phase": 1, "bpb": {"en": 2.12917, "tr": 2.37181, "fi": 2.51576}, "bytes_per_chunk": {"en": 3.99, "tr": 4.642, "fi": 4.024}, "bpb_mean": 2.33891, "frozen_inventory": true, "t": 1847.2} |
| {"event": "train", "step": 1800, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.51627, "bpb": 2.18751, "chunk_loss": 0.031334, "F": 0.24216, "G": 0.24229, "chunks": 247.97, "bytes_per_chunk": 4.13, "boundary_entropy": 0.13493, "router_noise": 0.0, "grad_norm": 0.7051, "lr": 0.0002564, "router_trainable": true, "sec_per_step": 0.943, "probe_sec": 0.1195, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09435, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.01369, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": -0.03873, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.29885, "trunc_abs_frac": 0.27306, "mean_delta_full": 0.09027, "read_hit_rate": {"surface": 0.72479, "split": 0.89675, "merge": 0.29818, "skel": 0.70576}, "mean_gate": {"surface": 0.22558, "split": 0.14945, "merge": 0.19131, "skel": 0.17711}, "n_reads": 23756, "n_cand": 34428, "read_suppressed_frac": 0.1046, "dropout_reads_logged": 23756, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2546, "admitted_surface": 16117, "staged_surface": 0, "admitted_split": 15399, "staged_split": 0, "admitted_merge": 22826, "staged_merge": 0, "admitted_skel": 11194, "staged_skel": 0, "mean_count": 634.2129516601562, "n_util_obs": 288, "util_count_spearman": 0.10233383166753071}, "probe_overhead_sec_total": 318.831, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.006905, "alpha": 0.03, "t": 1848.9} |
| {"event": "consolidate", "step": 1850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1888.3} |
| {"event": "train", "step": 1850, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.53694, "bpb": 2.21734, "chunk_loss": 0.031268, "F": 0.24249, "G": 0.24077, "chunks": 248.31, "bytes_per_chunk": 4.124, "boundary_entropy": 0.14038, "router_noise": 0.0, "grad_norm": 0.704, "lr": 0.0002537, "router_trainable": true, "sec_per_step": 0.812, "probe_sec": 0.141, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10965, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": -0.03142, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.12769, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.10314, "trunc_abs_frac": 0.96696, "mean_delta_full": -0.20634, "read_hit_rate": {"surface": 0.71582, "split": 0.89122, "merge": 0.29701, "skel": 0.69782}, "mean_gate": {"surface": 0.22266, "split": 0.14558, "merge": 0.18805, "skel": 0.17444}, "n_reads": 23567, "n_cand": 34452, "read_suppressed_frac": 0.0988, "dropout_reads_logged": 23567, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2596, "admitted_surface": 16108, "staged_surface": 0, "admitted_split": 15377, "staged_split": 0, "admitted_merge": 22816, "staged_merge": 0, "admitted_skel": 11235, "staged_skel": 0, "mean_count": 652.3311767578125, "n_util_obs": 291, "util_count_spearman": 0.07284964534154656}, "probe_overhead_sec_total": 325.28, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.007325, "alpha": 0.03, "t": 1889.5} |
| {"event": "consolidate", "step": 1900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1929.2} |
| {"event": "train", "step": 1900, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.45808, "bpb": 2.10356, "chunk_loss": 0.031586, "F": 0.25119, "G": 0.2517, "chunks": 257.22, "bytes_per_chunk": 3.981, "boundary_entropy": 0.13641, "router_noise": 0.0, "grad_norm": 0.7355, "lr": 0.0002508, "router_trainable": true, "sec_per_step": 0.807, "probe_sec": 0.1274, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09659, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.07984, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.01022, "n_merge": 2, "mean_delta_merge": -0.16574, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 1.3108, "trunc_abs_frac": 0.74225, "mean_delta_full": -0.13501, "read_hit_rate": {"surface": 0.73765, "split": 0.90894, "merge": 0.31272, "skel": 0.7333}, "mean_gate": {"surface": 0.21584, "split": 0.13741, "merge": 0.18511, "skel": 0.16843}, "n_reads": 25368, "n_cand": 35934, "read_suppressed_frac": 0.0978, "dropout_reads_logged": 25368, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2647, "admitted_surface": 16102, "staged_surface": 0, "admitted_split": 15260, "staged_split": 0, "admitted_merge": 23025, "staged_merge": 0, "admitted_skel": 11149, "staged_skel": 0, "mean_count": 670.5355834960938, "n_util_obs": 301, "util_count_spearman": 0.06255351917449561}, "probe_overhead_sec_total": 332.026, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.006871, "alpha": 0.03, "t": 1929.8} |
| {"event": "consolidate", "step": 1950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1970.0} |
| {"event": "train", "step": 1950, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.45519, "bpb": 2.09939, "chunk_loss": 0.031195, "F": 0.24002, "G": 0.23784, "chunks": 245.78, "bytes_per_chunk": 4.166, "boundary_entropy": 0.13068, "router_noise": 0.0, "grad_norm": 0.6978, "lr": 0.0002479, "router_trainable": true, "sec_per_step": 0.824, "probe_sec": 0.0988, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07574, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.02686, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.04939, "n_merge": 1, "mean_delta_merge": -0.49487, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -0.33148, "trunc_abs_frac": 1.2069, "mean_delta_full": 0.15665, "read_hit_rate": {"surface": 0.73586, "split": 0.89624, "merge": 0.31829, "skel": 0.71282}, "mean_gate": {"surface": 0.22176, "split": 0.13977, "merge": 0.18816, "skel": 0.17224}, "n_reads": 23846, "n_cand": 34155, "read_suppressed_frac": 0.1006, "dropout_reads_logged": 23846, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2701, "admitted_surface": 16106, "staged_surface": 0, "admitted_split": 15420, "staged_split": 0, "admitted_merge": 22802, "staged_merge": 0, "admitted_skel": 11208, "staged_skel": 0, "mean_count": 688.6743774414062, "n_util_obs": 311, "util_count_spearman": 0.05476889566195655}, "probe_overhead_sec_total": 338.331, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.006694, "alpha": 0.03, "t": 1971.0} |
| {"event": "consolidate", "step": 2000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2009.7} |
| {"event": "probe", "step": 2000, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 4212.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.242027, "cps": 32.90625, "mean_p": 0.23891, "boundary_entropy": 0.119445, "bpb": 2.091306}, "tr": {"chunks": 4311.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.238137, "cps": 33.679688, "mean_p": 0.230737, "boundary_entropy": 0.164371, "bpb": 2.323591}, "fi": {"chunks": 5102.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.271224, "cps": 39.859375, "mean_p": 0.270777, "boundary_entropy": 0.163572, "bpb": 2.461655}}, "bpb": {"en": 2.09131, "tr": 2.32359, "fi": 2.46166}, "gini_cpc": 0.029356, "gini_cps": 0.043547, "bpic": 4.163817, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.983927794939782, "fi": 1.1206353771608144}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2757, "admitted_surface": 16097, "staged_surface": 0, "admitted_split": 15367, "staged_split": 0, "admitted_merge": 22913, "staged_merge": 0, "admitted_skel": 11159, "staged_skel": 0, "mean_count": 706.456298828125, "n_util_obs": 317, "util_count_spearman": 0.06387693617486266}, "t": 2011.7} |
| {"event": "eval", "step": 2000, "phase": 1, "bpb": {"en": 2.10356, "tr": 2.33597, "fi": 2.46917}, "bytes_per_chunk": {"en": 4.063, "tr": 4.63, "fi": 3.999}, "bpb_mean": 2.3029, "frozen_inventory": true, "t": 2015.8} |
| {"event": "train", "step": 2000, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.57426, "bpb": 2.27118, "chunk_loss": 0.031485, "F": 0.24658, "G": 0.24711, "chunks": 252.5, "bytes_per_chunk": 4.055, "boundary_entropy": 0.12744, "router_noise": 0.0, "grad_norm": 0.7579, "lr": 0.000245, "router_trainable": true, "sec_per_step": 1.019, "probe_sec": 0.1775, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.1215, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.01114, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.06274, "n_merge": 1, "mean_delta_merge": 0.12881, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 1.0, "trunc_abs_frac": 1.0, "mean_delta_full": -0.0979, "read_hit_rate": {"surface": 0.70924, "split": 0.89101, "merge": 0.28705, "skel": 0.6895}, "mean_gate": {"surface": 0.22445, "split": 0.14053, "merge": 0.19052, "skel": 0.17293}, "n_reads": 23734, "n_cand": 34977, "read_suppressed_frac": 0.1001, "dropout_reads_logged": 23734, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2757, "admitted_surface": 16133, "staged_surface": 0, "admitted_split": 15345, "staged_split": 0, "admitted_merge": 22831, "staged_merge": 0, "admitted_skel": 11227, "staged_skel": 0, "mean_count": 706.8186645507812, "n_util_obs": 317, "util_count_spearman": 0.06377635443948869}, "probe_overhead_sec_total": 344.732, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.006317, "alpha": 0.03, "t": 2022.0} |
| {"event": "probe", "step": 2048, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 4190.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.240763, "cps": 32.734375, "mean_p": 0.239208, "boundary_entropy": 0.127786, "bpb": 2.087655}, "tr": {"chunks": 4377.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.241783, "cps": 34.195312, "mean_p": 0.235493, "boundary_entropy": 0.175096, "bpb": 2.317344}, "fi": {"chunks": 5085.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.270321, "cps": 39.726562, "mean_p": 0.273965, "boundary_entropy": 0.171223, "bpb": 2.441873}}, "bpb": {"en": 2.08766, "tr": 2.31734, "fi": 2.44187}, "gini_cpc": 0.026173, "gini_cps": 0.043705, "bpic": 4.155582, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0042367086139417, "fi": 1.1227657889197773}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2798, "admitted_surface": 16132, "staged_surface": 0, "admitted_split": 15290, "staged_split": 0, "admitted_merge": 22997, "staged_merge": 0, "admitted_skel": 11117, "staged_skel": 0, "mean_count": 723.765869140625, "n_util_obs": 323, "util_count_spearman": 0.063107653966158}, "t": 2079.3} |
| {"event": "consolidate", "step": 2050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2083.3} |
| {"event": "train", "step": 2050, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.52263, "bpb": 2.19669, "chunk_loss": 0.031444, "F": 0.24463, "G": 0.24575, "chunks": 250.5, "bytes_per_chunk": 4.088, "boundary_entropy": 0.13187, "router_noise": 0.0, "grad_norm": 0.6977, "lr": 0.0002419, "router_trainable": true, "sec_per_step": 1.268, "probe_sec": 0.1512, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08129, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": 0.21389, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": -0.08315, "n_merge": 3, "mean_delta_merge": -0.13343, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 20.33333, "trunc_abs_frac": 20.33333, "mean_delta_full": -0.00073, "read_hit_rate": {"surface": 0.71455, "split": 0.89751, "merge": 0.28999, "skel": 0.70766}, "mean_gate": {"surface": 0.22132, "split": 0.13554, "merge": 0.18676, "skel": 0.17069}, "n_reads": 24016, "n_cand": 34959, "read_suppressed_frac": 0.1003, "dropout_reads_logged": 24016, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2802, "admitted_surface": 16143, "staged_surface": 0, "admitted_split": 15285, "staged_split": 0, "admitted_merge": 22923, "staged_merge": 0, "admitted_skel": 11185, "staged_skel": 0, "mean_count": 724.8599853515625, "n_util_obs": 323, "util_count_spearman": 0.06039876442302537}, "probe_overhead_sec_total": 351.825, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.006506, "alpha": 0.03, "t": 2085.4} |
| {"event": "consolidate", "step": 2100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2134.9} |
| {"event": "train", "step": 2100, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.54659, "bpb": 2.23126, "chunk_loss": 0.031339, "F": 0.2438, "G": 0.24292, "chunks": 249.66, "bytes_per_chunk": 4.102, "boundary_entropy": 0.12498, "router_noise": 0.0, "grad_norm": 0.7501, "lr": 0.0002388, "router_trainable": true, "sec_per_step": 1.003, "probe_sec": 0.1024, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07262, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.00669, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.0188, "n_merge": 2, "mean_delta_merge": -0.0481, "n_dropped_unpaired": 0, "n_full": 0, "read_hit_rate": {"surface": 0.70139, "split": 0.88565, "merge": 0.28139, "skel": 0.68329}, "mean_gate": {"surface": 0.22149, "split": 0.13546, "merge": 0.1877, "skel": 0.17202}, "n_reads": 23458, "n_cand": 34832, "read_suppressed_frac": 0.0998, "dropout_reads_logged": 23458, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2859, "admitted_surface": 16138, "staged_surface": 0, "admitted_split": 15349, "staged_split": 0, "admitted_merge": 22842, "staged_merge": 0, "admitted_skel": 11207, "staged_skel": 0, "mean_count": 742.9285888671875, "n_util_obs": 331, "util_count_spearman": 0.042259183770377406}, "probe_overhead_sec_total": 358.505, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.007568, "alpha": 0.03, "t": 2135.5} |
| {"event": "consolidate", "step": 2150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2185.4} |
| {"event": "train", "step": 2150, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.51993, "bpb": 2.1928, "chunk_loss": 0.031544, "F": 0.25015, "G": 0.24848, "chunks": 256.16, "bytes_per_chunk": 3.998, "boundary_entropy": 0.12379, "router_noise": 0.0, "grad_norm": 0.7179, "lr": 0.0002357, "router_trainable": true, "sec_per_step": 1.024, "probe_sec": 0.1939, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10008, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.09695, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 3.61616, "trunc_abs_frac": 1.53237, "mean_delta_full": -0.0274, "read_hit_rate": {"surface": 0.72492, "split": 0.89671, "merge": 0.30173, "skel": 0.71441}, "mean_gate": {"surface": 0.21975, "split": 0.13291, "merge": 0.18483, "skel": 0.16962}, "n_reads": 24643, "n_cand": 35596, "read_suppressed_frac": 0.0963, "dropout_reads_logged": 24643, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2915, "admitted_surface": 16145, "staged_surface": 0, "admitted_split": 15311, "staged_split": 0, "admitted_merge": 22954, "staged_merge": 0, "admitted_skel": 11126, "staged_skel": 0, "mean_count": 760.9835205078125, "n_util_obs": 335, "util_count_spearman": 0.05682846757062286}, "probe_overhead_sec_total": 365.245, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.006914, "alpha": 0.03, "t": 2186.7} |
| {"event": "consolidate", "step": 2200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2238.9} |
| {"event": "probe", "step": 2200, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 4218.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.242372, "cps": 32.953125, "mean_p": 0.236852, "boundary_entropy": 0.114411, "bpb": 2.068833}, "tr": {"chunks": 4325.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.238911, "cps": 33.789062, "mean_p": 0.229575, "boundary_entropy": 0.160598, "bpb": 2.277413}, "fi": {"chunks": 5013.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.266493, "cps": 39.164062, "mean_p": 0.267696, "boundary_entropy": 0.157168, "bpb": 2.424177}}, "bpb": {"en": 2.06883, "tr": 2.27741, "fi": 2.42418}, "gini_cpc": 0.024591, "gini_cps": 0.039097, "bpic": 4.18501, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9857189486837961, "fi": 1.0995205886087203}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2962, "admitted_surface": 16126, "staged_surface": 0, "admitted_split": 15334, "staged_split": 0, "admitted_merge": 22942, "staged_merge": 0, "admitted_skel": 11134, "staged_skel": 0, "mean_count": 778.7451171875, "n_util_obs": 340, "util_count_spearman": 0.047947595720093245}, "t": 2240.7} |
| {"event": "eval", "step": 2200, "phase": 1, "bpb": {"en": 2.08111, "tr": 2.28756, "fi": 2.43239}, "bytes_per_chunk": {"en": 4.052, "tr": 4.602, "fi": 4.097}, "bpb_mean": 2.26702, "frozen_inventory": true, "t": 2243.8} |
| {"event": "train", "step": 2200, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.59399, "bpb": 2.29964, "chunk_loss": 0.031603, "F": 0.24673, "G": 0.24556, "chunks": 252.66, "bytes_per_chunk": 4.053, "boundary_entropy": 0.12813, "router_noise": 0.0, "grad_norm": 0.7415, "lr": 0.0002325, "router_trainable": true, "sec_per_step": 1.179, "probe_sec": 0.178, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.14337, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.04353, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 3, "mean_delta_split": -0.0674, "n_merge": 1, "mean_delta_merge": 0.11389, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 1.14764, "trunc_abs_frac": 2.2022, "mean_delta_full": 0.03876, "read_hit_rate": {"surface": 0.70446, "split": 0.87956, "merge": 0.2764, "skel": 0.69241}, "mean_gate": {"surface": 0.22332, "split": 0.13372, "merge": 0.1913, "skel": 0.17271}, "n_reads": 23256, "n_cand": 34707, "read_suppressed_frac": 0.1014, "dropout_reads_logged": 23256, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2964, "admitted_surface": 16158, "staged_surface": 0, "admitted_split": 15368, "staged_split": 0, "admitted_merge": 22853, "staged_merge": 0, "admitted_skel": 11157, "staged_skel": 0, "mean_count": 779.0687866210938, "n_util_obs": 341, "util_count_spearman": 0.04920028205756759}, "probe_overhead_sec_total": 372.219, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.007878, "alpha": 0.03, "t": 2245.6} |
| {"event": "consolidate", "step": 2250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2289.3} |
| {"event": "train", "step": 2250, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.58095, "bpb": 2.28082, "chunk_loss": 0.031257, "F": 0.24121, "G": 0.2409, "chunks": 247.0, "bytes_per_chunk": 4.146, "boundary_entropy": 0.12897, "router_noise": 0.0, "grad_norm": 0.7443, "lr": 0.0002293, "router_trainable": true, "sec_per_step": 0.897, "probe_sec": 0.1307, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07212, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": -0.00047, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.0722, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.11024, "trunc_abs_frac": 0.11024, "mean_delta_full": -0.60156, "read_hit_rate": {"surface": 0.69334, "split": 0.88347, "merge": 0.27334, "skel": 0.68702}, "mean_gate": {"surface": 0.21829, "split": 0.12812, "merge": 0.18388, "skel": 0.16539}, "n_reads": 23001, "n_cand": 34356, "read_suppressed_frac": 0.0986, "dropout_reads_logged": 23001, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3027, "admitted_surface": 16216, "staged_surface": 0, "admitted_split": 15371, "staged_split": 0, "admitted_merge": 22772, "staged_merge": 0, "admitted_skel": 11177, "staged_skel": 0, "mean_count": 797.0719604492188, "n_util_obs": 344, "util_count_spearman": 0.030670612893431242}, "probe_overhead_sec_total": 378.495, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.008828, "alpha": 0.03, "t": 2290.5} |
| {"event": "consolidate", "step": 2300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2344.4} |
| {"event": "train", "step": 2300, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.50292, "bpb": 2.16826, "chunk_loss": 0.031594, "F": 0.25168, "G": 0.24919, "chunks": 257.72, "bytes_per_chunk": 3.973, "boundary_entropy": 0.11841, "router_noise": 0.0, "grad_norm": 0.6466, "lr": 0.000226, "router_trainable": true, "sec_per_step": 1.092, "probe_sec": 0.1089, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07857, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.09352, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": 0.01755, "n_merge": 1, "mean_delta_merge": -0.59595, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": -0.2267, "trunc_abs_frac": 0.2267, "mean_delta_full": -0.37073, "read_hit_rate": {"surface": 0.73548, "split": 0.89723, "merge": 0.30429, "skel": 0.72195}, "mean_gate": {"surface": 0.2196, "split": 0.12809, "merge": 0.18751, "skel": 0.16911}, "n_reads": 24864, "n_cand": 35708, "read_suppressed_frac": 0.0991, "dropout_reads_logged": 24864, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3074, "admitted_surface": 16126, "staged_surface": 0, "admitted_split": 15362, "staged_split": 0, "admitted_merge": 22937, "staged_merge": 0, "admitted_skel": 11111, "staged_skel": 0, "mean_count": 815.1078491210938, "n_util_obs": 352, "util_count_spearman": 0.03293696841714303}, "probe_overhead_sec_total": 386.243, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.007846, "alpha": 0.03, "t": 2345.0} |
| {"event": "consolidate", "step": 2350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2390.3} |
| {"event": "train", "step": 2350, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.3927, "bpb": 2.00924, "chunk_loss": 0.031409, "F": 0.24695, "G": 0.24231, "chunks": 252.88, "bytes_per_chunk": 4.049, "boundary_entropy": 0.12084, "router_noise": 0.0, "grad_norm": 0.6468, "lr": 0.0002226, "router_trainable": true, "sec_per_step": 0.931, "probe_sec": 0.1709, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.1425, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.05534, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 3, "mean_delta_split": 0.06431, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.61179, "trunc_abs_frac": 0.60481, "mean_delta_full": 0.27604, "read_hit_rate": {"surface": 0.74206, "split": 0.90798, "merge": 0.32561, "skel": 0.73395}, "mean_gate": {"surface": 0.224, "split": 0.13025, "merge": 0.1919, "skel": 0.17331}, "n_reads": 24947, "n_cand": 35172, "read_suppressed_frac": 0.1025, "dropout_reads_logged": 24947, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3132, "admitted_surface": 16163, "staged_surface": 0, "admitted_split": 15306, "staged_split": 0, "admitted_merge": 22919, "staged_merge": 0, "admitted_skel": 11148, "staged_skel": 0, "mean_count": 833.193603515625, "n_util_obs": 356, "util_count_spearman": 0.04470418310950865}, "probe_overhead_sec_total": 392.455, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.00725, "alpha": 0.03, "t": 2391.6} |
| {"event": "consolidate", "step": 2400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2442.4} |
| {"event": "probe", "step": 2400, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 4144.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.23812, "cps": 32.375, "mean_p": 0.233921, "boundary_entropy": 0.117334, "bpb": 2.050321}, "tr": {"chunks": 4336.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.239518, "cps": 33.875, "mean_p": 0.224023, "boundary_entropy": 0.159122, "bpb": 2.243133}, "fi": {"chunks": 5044.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.268141, "cps": 39.40625, "mean_p": 0.269766, "boundary_entropy": 0.158255, "bpb": 2.401828}}, "bpb": {"en": 2.05032, "tr": 2.24313, "fi": 2.40183}, "gini_cpc": 0.026836, "gini_cps": 0.044366, "bpic": 4.194913, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.005872872027653, "fi": 1.1260756511274825}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3182, "admitted_surface": 16119, "staged_surface": 0, "admitted_split": 15421, "staged_split": 0, "admitted_merge": 22767, "staged_merge": 0, "admitted_skel": 11229, "staged_skel": 0, "mean_count": 850.8404541015625, "n_util_obs": 357, "util_count_spearman": 0.0574781792699504}, "t": 2444.6} |
| {"event": "eval", "step": 2400, "phase": 1, "bpb": {"en": 2.06228, "tr": 2.25429, "fi": 2.41002}, "bytes_per_chunk": {"en": 4.118, "tr": 4.57, "fi": 4.093}, "bpb_mean": 2.2422, "frozen_inventory": true, "t": 2448.0} |
| {"event": "phase_transition", "step": 2400, "phase": 2, "objective": "global", "router_trainable": true, "encoder_trainable": true, "frozen_router_params": 0, "t": 2449.6} |
| {"event": "train", "step": 2400, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.46823, "bpb": 2.1182, "chunk_loss": 0.031188, "F": 0.23911, "G": 0.23588, "chunks": 244.84, "bytes_per_chunk": 4.182, "boundary_entropy": 0.12016, "router_noise": 0.0, "grad_norm": 0.6997, "lr": 0.0002192, "router_trainable": true, "sec_per_step": 1.181, "probe_sec": 0.1125, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07178, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": 0.04104, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.03586, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -0.44824, "trunc_abs_frac": 0.89835, "mean_delta_full": -0.18709, "read_hit_rate": {"surface": 0.71075, "split": 0.89155, "merge": 0.2916, "skel": 0.696}, "mean_gate": {"surface": 0.2184, "split": 0.12661, "merge": 0.18566, "skel": 0.16836}, "n_reads": 23248, "n_cand": 34094, "read_suppressed_frac": 0.1006, "dropout_reads_logged": 23248, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3183, "admitted_surface": 16152, "staged_surface": 0, "admitted_split": 15381, "staged_split": 0, "admitted_merge": 22750, "staged_merge": 0, "admitted_skel": 11253, "staged_skel": 0, "mean_count": 851.1800537109375, "n_util_obs": 357, "util_count_spearman": 0.05763853466815136}, "probe_overhead_sec_total": 399.968, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.007687, "alpha": 0.03, "t": 2450.7} |
| {"event": "consolidate", "step": 2450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2502.4} |
| {"event": "train", "step": 2450, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.49337, "bpb": 2.15448, "chunk_loss": 0.031378, "F": 0.24478, "G": 0.24428, "chunks": 250.66, "bytes_per_chunk": 4.085, "boundary_entropy": 0.11901, "router_noise": 0.0, "grad_norm": 0.6812, "lr": 0.0002158, "router_trainable": true, "sec_per_step": 1.051, "probe_sec": 0.1112, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07279, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.01549, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.11499, "n_merge": 1, "mean_delta_merge": -0.02982, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.36183, "trunc_abs_frac": 0.95904, "mean_delta_full": 0.25215, "read_hit_rate": {"surface": 0.71436, "split": 0.89025, "merge": 0.27611, "skel": 0.71158}, "mean_gate": {"surface": 0.21934, "split": 0.12666, "merge": 0.18692, "skel": 0.16692}, "n_reads": 23810, "n_cand": 34936, "read_suppressed_frac": 0.1, "dropout_reads_logged": 23810, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3229, "admitted_surface": 16126, "staged_surface": 0, "admitted_split": 15371, "staged_split": 0, "admitted_merge": 22873, "staged_merge": 0, "admitted_skel": 11166, "staged_skel": 0, "mean_count": 868.9486083984375, "n_util_obs": 367, "util_count_spearman": 0.02441051775001408}, "probe_overhead_sec_total": 406.908, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.007806, "alpha": 0.03, "t": 2503.2} |
| {"event": "consolidate", "step": 2500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2550.3} |
| {"event": "train", "step": 2500, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.39991, "bpb": 2.01964, "chunk_loss": 0.031396, "F": 0.2471, "G": 0.24249, "chunks": 253.03, "bytes_per_chunk": 4.047, "boundary_entropy": 0.11451, "router_noise": 0.0, "grad_norm": 0.6979, "lr": 0.0002123, "router_trainable": true, "sec_per_step": 0.955, "probe_sec": 0.1122, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07429, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.09095, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.27374, "n_merge": 2, "mean_delta_merge": -0.16314, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 0.60186, "trunc_abs_frac": 0.49446, "mean_delta_full": 0.07967, "read_hit_rate": {"surface": 0.73825, "split": 0.90355, "merge": 0.31271, "skel": 0.70746}, "mean_gate": {"surface": 0.21137, "split": 0.12171, "merge": 0.17977, "skel": 0.16303}, "n_reads": 24364, "n_cand": 34883, "read_suppressed_frac": 0.0997, "dropout_reads_logged": 24364, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3288, "admitted_surface": 16067, "staged_surface": 0, "admitted_split": 15258, "staged_split": 0, "admitted_merge": 23016, "staged_merge": 0, "admitted_skel": 11195, "staged_skel": 0, "mean_count": 886.8837280273438, "n_util_obs": 372, "util_count_spearman": 0.0259734961850852}, "probe_overhead_sec_total": 413.126, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.008152, "alpha": 0.03, "t": 2550.9} |
| {"event": "consolidate", "step": 2550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2599.6} |
| {"event": "train", "step": 2550, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.43514, "bpb": 2.07048, "chunk_loss": 0.031263, "F": 0.24063, "G": 0.23725, "chunks": 246.41, "bytes_per_chunk": 4.156, "boundary_entropy": 0.10943, "router_noise": 0.0, "grad_norm": 0.6816, "lr": 0.0002088, "router_trainable": true, "sec_per_step": 0.987, "probe_sec": 0.1748, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.13425, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.05943, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": 0.08301, "n_merge": 1, "mean_delta_merge": -0.44044, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 3.45676, "trunc_abs_frac": 1.38592, "mean_delta_full": -0.01133, "read_hit_rate": {"surface": 0.72991, "split": 0.89552, "merge": 0.31351, "skel": 0.69634}, "mean_gate": {"surface": 0.21359, "split": 0.12162, "merge": 0.18344, "skel": 0.16551}, "n_reads": 23653, "n_cand": 34157, "read_suppressed_frac": 0.0996, "dropout_reads_logged": 23653, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3345, "admitted_surface": 16121, "staged_surface": 0, "admitted_split": 15350, "staged_split": 0, "admitted_merge": 22814, "staged_merge": 0, "admitted_skel": 11251, "staged_skel": 0, "mean_count": 904.8501586914062, "n_util_obs": 382, "util_count_spearman": 0.05445355549166664}, "probe_overhead_sec_total": 419.264, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.008416, "alpha": 0.03, "t": 2600.3} |
| {"event": "consolidate", "step": 2600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2645.0} |
| {"event": "probe", "step": 2600, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 4123.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.236913, "cps": 32.210938, "mean_p": 0.233425, "boundary_entropy": 0.112359, "bpb": 2.023409}, "tr": {"chunks": 4337.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.239574, "cps": 33.882812, "mean_p": 0.22339, "boundary_entropy": 0.145168, "bpb": 2.217359}, "fi": {"chunks": 4964.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.263888, "cps": 38.78125, "mean_p": 0.26261, "boundary_entropy": 0.143849, "bpb": 2.351171}}, "bpb": {"en": 2.02341, "tr": 2.21736, "fi": 2.35117}, "gini_cpc": 0.024289, "gini_cps": 0.041766, "bpic": 4.226162, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0112293267180315, "fi": 1.1138601707555915}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3396, "admitted_surface": 16109, "staged_surface": 0, "admitted_split": 15373, "staged_split": 0, "admitted_merge": 22878, "staged_merge": 0, "admitted_skel": 11176, "staged_skel": 0, "mean_count": 922.2965087890625, "n_util_obs": 387, "util_count_spearman": 0.04248224333739543}, "t": 2646.7} |
| {"event": "eval", "step": 2600, "phase": 2, "bpb": {"en": 2.03402, "tr": 2.22786, "fi": 2.35946}, "bytes_per_chunk": {"en": 4.115, "tr": 4.559, "fi": 4.103}, "bpb_mean": 2.20711, "frozen_inventory": true, "t": 2649.6} |
| {"event": "train", "step": 2600, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.41062, "bpb": 2.03509, "chunk_loss": 0.031171, "F": 0.24023, "G": 0.23555, "chunks": 246.0, "bytes_per_chunk": 4.163, "boundary_entropy": 0.11248, "router_noise": 0.0, "grad_norm": 0.6432, "lr": 0.0002053, "router_trainable": true, "sec_per_step": 1.019, "probe_sec": 0.1113, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08526, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.0499, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.62833, "trunc_abs_frac": 0.65758, "mean_delta_full": -0.12207, "read_hit_rate": {"surface": 0.72436, "split": 0.89711, "merge": 0.2938, "skel": 0.70511}, "mean_gate": {"surface": 0.21172, "split": 0.12086, "merge": 0.18367, "skel": 0.16484}, "n_reads": 23613, "n_cand": 34251, "read_suppressed_frac": 0.1007, "dropout_reads_logged": 23613, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3396, "admitted_surface": 16109, "staged_surface": 0, "admitted_split": 15324, "staged_split": 0, "admitted_merge": 22914, "staged_merge": 0, "admitted_skel": 11189, "staged_skel": 0, "mean_count": 922.660400390625, "n_util_obs": 387, "util_count_spearman": 0.04301495802033867}, "probe_overhead_sec_total": 425.643, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.007437, "alpha": 0.03, "t": 2651.2} |
| {"event": "consolidate", "step": 2650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2701.1} |
| {"event": "train", "step": 2650, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.39366, "bpb": 2.01062, "chunk_loss": 0.031285, "F": 0.24335, "G": 0.23812, "chunks": 249.19, "bytes_per_chunk": 4.109, "boundary_entropy": 0.11223, "router_noise": 0.0, "grad_norm": 0.6558, "lr": 0.0002017, "router_trainable": true, "sec_per_step": 1.024, "probe_sec": 0.1615, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07748, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": 0.01929, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.15125, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 1.32604, "trunc_abs_frac": 0.85748, "mean_delta_full": 0.07675, "read_hit_rate": {"surface": 0.73898, "split": 0.90161, "merge": 0.30405, "skel": 0.72275}, "mean_gate": {"surface": 0.21133, "split": 0.11933, "merge": 0.18048, "skel": 0.16322}, "n_reads": 24312, "n_cand": 34745, "read_suppressed_frac": 0.0971, "dropout_reads_logged": 24312, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3435, "admitted_surface": 16082, "staged_surface": 0, "admitted_split": 15343, "staged_split": 0, "admitted_merge": 22927, "staged_merge": 0, "admitted_skel": 11184, "staged_skel": 0, "mean_count": 940.4069213867188, "n_util_obs": 398, "util_count_spearman": 0.027920136463845006}, "probe_overhead_sec_total": 432.057, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.008087, "alpha": 0.03, "t": 2702.5} |
| {"event": "consolidate", "step": 2700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2747.6} |
| {"event": "train", "step": 2700, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.5084, "bpb": 2.17616, "chunk_loss": 0.031469, "F": 0.24734, "G": 0.24555, "chunks": 253.28, "bytes_per_chunk": 4.043, "boundary_entropy": 0.1099, "router_noise": 0.0, "grad_norm": 0.6463, "lr": 0.0001981, "router_trainable": true, "sec_per_step": 0.916, "probe_sec": 0.0835, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05814, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.05169, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": 0.06146, "n_merge": 1, "mean_delta_merge": -0.17294, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -0.50498, "trunc_abs_frac": 2.75253, "mean_delta_full": 0.12472, "read_hit_rate": {"surface": 0.71349, "split": 0.89279, "merge": 0.28255, "skel": 0.71054}, "mean_gate": {"surface": 0.21408, "split": 0.11912, "merge": 0.18567, "skel": 0.16359}, "n_reads": 24244, "n_cand": 35442, "read_suppressed_frac": 0.1021, "dropout_reads_logged": 24244, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3481, "admitted_surface": 16114, "staged_surface": 0, "admitted_split": 15329, "staged_split": 0, "admitted_merge": 22936, "staged_merge": 0, "admitted_skel": 11157, "staged_skel": 0, "mean_count": 958.0941772460938, "n_util_obs": 402, "util_count_spearman": 0.03396041524281006}, "probe_overhead_sec_total": 437.737, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.007325, "alpha": 0.03, "t": 2748.3} |
| {"event": "consolidate", "step": 2750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2789.5} |
| {"event": "train", "step": 2750, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.43743, "bpb": 2.07377, "chunk_loss": 0.031215, "F": 0.23904, "G": 0.23749, "chunks": 244.78, "bytes_per_chunk": 4.183, "boundary_entropy": 0.11262, "router_noise": 0.0, "grad_norm": 0.6244, "lr": 0.0001945, "router_trainable": true, "sec_per_step": 0.838, "probe_sec": 0.0833, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05311, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.27066, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 3, "mean_delta_split": 0.03384, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 1.36983, "trunc_abs_frac": 1.46356, "mean_delta_full": 0.17878, "read_hit_rate": {"surface": 0.71311, "split": 0.89078, "merge": 0.27919, "skel": 0.69946}, "mean_gate": {"surface": 0.20684, "split": 0.11537, "merge": 0.17392, "skel": 0.15679}, "n_reads": 23278, "n_cand": 34212, "read_suppressed_frac": 0.0977, "dropout_reads_logged": 23278, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3535, "admitted_surface": 16113, "staged_surface": 0, "admitted_split": 15378, "staged_split": 0, "admitted_merge": 22814, "staged_merge": 0, "admitted_skel": 11231, "staged_skel": 0, "mean_count": 975.5823974609375, "n_util_obs": 410, "util_count_spearman": 0.043202158588789834}, "probe_overhead_sec_total": 443.5, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.007316, "alpha": 0.03, "t": 2790.1} |
| {"event": "consolidate", "step": 2800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2834.9} |
| {"event": "probe", "step": 2800, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3968.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.228007, "cps": 31.0, "mean_p": 0.228929, "boundary_entropy": 0.102296, "bpb": 2.017491}, "tr": {"chunks": 4364.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.241065, "cps": 34.09375, "mean_p": 0.225946, "boundary_entropy": 0.137656, "bpb": 2.187977}, "fi": {"chunks": 4981.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.264792, "cps": 38.914062, "mean_p": 0.262965, "boundary_entropy": 0.138763, "bpb": 2.330964}}, "bpb": {"en": 2.01749, "tr": 2.18798, "fi": 2.33096}, "gini_cpc": 0.033417, "gini_cps": 0.050727, "bpic": 4.261399, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0572717964229774, "fi": 1.161333930708868}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3572, "admitted_surface": 16152, "staged_surface": 0, "admitted_split": 15449, "staged_split": 0, "admitted_merge": 22677, "staged_merge": 0, "admitted_skel": 11258, "staged_skel": 0, "mean_count": 992.4317016601562, "n_util_obs": 415, "util_count_spearman": 0.026716902170425126}, "t": 2836.7} |
| {"event": "eval", "step": 2800, "phase": 2, "bpb": {"en": 2.03083, "tr": 2.19889, "fi": 2.33693}, "bytes_per_chunk": {"en": 4.302, "tr": 4.576, "fi": 4.126}, "bpb_mean": 2.18888, "frozen_inventory": true, "t": 2839.4} |
| {"event": "train", "step": 2800, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.46265, "bpb": 2.11016, "chunk_loss": 0.031223, "F": 0.23581, "G": 0.23812, "chunks": 241.47, "bytes_per_chunk": 4.241, "boundary_entropy": 0.10906, "router_noise": 0.0, "grad_norm": 0.6744, "lr": 0.0001909, "router_trainable": true, "sec_per_step": 1.028, "probe_sec": 0.148, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09588, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.01016, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": -0.08367, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.66415, "trunc_abs_frac": 1.17999, "mean_delta_full": -0.11027, "read_hit_rate": {"surface": 0.70227, "split": 0.88077, "merge": 0.27274, "skel": 0.68498}, "mean_gate": {"surface": 0.20462, "split": 0.11487, "merge": 0.17482, "skel": 0.15706}, "n_reads": 22424, "n_cand": 33515, "read_suppressed_frac": 0.1017, "dropout_reads_logged": 22424, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3573, "admitted_surface": 16128, "staged_surface": 0, "admitted_split": 15476, "staged_split": 0, "admitted_merge": 22688, "staged_merge": 0, "admitted_skel": 11244, "staged_skel": 0, "mean_count": 992.7769775390625, "n_util_obs": 415, "util_count_spearman": 0.026585357028898048}, "probe_overhead_sec_total": 449.362, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.006681, "alpha": 0.03, "t": 2841.5} |
| {"event": "consolidate", "step": 2850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2888.6} |
| {"event": "train", "step": 2850, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.42635, "bpb": 2.05779, "chunk_loss": 0.030913, "F": 0.22815, "G": 0.23062, "chunks": 233.62, "bytes_per_chunk": 4.383, "boundary_entropy": 0.10898, "router_noise": 0.0, "grad_norm": 0.6717, "lr": 0.0001872, "router_trainable": true, "sec_per_step": 0.959, "probe_sec": 0.1106, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0892, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.04836, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.15381, "n_merge": 1, "mean_delta_merge": 0.34851, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.17702, "trunc_abs_frac": 0.74051, "mean_delta_full": 0.32841, "read_hit_rate": {"surface": 0.68632, "split": 0.87921, "merge": 0.26862, "skel": 0.67411}, "mean_gate": {"surface": 0.20878, "split": 0.11465, "merge": 0.18392, "skel": 0.15957}, "n_reads": 21626, "n_cand": 32631, "read_suppressed_frac": 0.0972, "dropout_reads_logged": 21626, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3607, "admitted_surface": 16160, "staged_surface": 0, "admitted_split": 15470, "staged_split": 0, "admitted_merge": 22583, "staged_merge": 0, "admitted_skel": 11323, "staged_skel": 0, "mean_count": 1009.4939575195312, "n_util_obs": 425, "util_count_spearman": 0.037177158399899955}, "probe_overhead_sec_total": 454.862, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.007097, "alpha": 0.03, "t": 2889.5} |
| {"event": "consolidate", "step": 2900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2928.1} |
| {"event": "train", "step": 2900, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.44185, "bpb": 2.08015, "chunk_loss": 0.031028, "F": 0.23209, "G": 0.23236, "chunks": 237.66, "bytes_per_chunk": 4.309, "boundary_entropy": 0.10368, "router_noise": 0.0, "grad_norm": 0.663, "lr": 0.0001835, "router_trainable": true, "sec_per_step": 0.797, "probe_sec": 0.0928, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05913, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": 0.02325, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.14087, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -17.85, "trunc_abs_frac": 0.68187, "mean_delta_full": -0.00549, "read_hit_rate": {"surface": 0.69325, "split": 0.88249, "merge": 0.26416, "skel": 0.68031}, "mean_gate": {"surface": 0.2031, "split": 0.11404, "merge": 0.17672, "skel": 0.15905}, "n_reads": 22208, "n_cand": 33332, "read_suppressed_frac": 0.0992, "dropout_reads_logged": 22208, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3664, "admitted_surface": 16168, "staged_surface": 0, "admitted_split": 15426, "staged_split": 0, "admitted_merge": 22643, "staged_merge": 0, "admitted_skel": 11299, "staged_skel": 0, "mean_count": 1026.1422119140625, "n_util_obs": 433, "util_count_spearman": 0.02910043270763297}, "probe_overhead_sec_total": 460.663, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.007843, "alpha": 0.03, "t": 2929.3} |
| {"event": "consolidate", "step": 2950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2975.1} |
| {"event": "train", "step": 2950, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.51193, "bpb": 2.18125, "chunk_loss": 0.031279, "F": 0.23831, "G": 0.23985, "chunks": 244.03, "bytes_per_chunk": 4.196, "boundary_entropy": 0.10411, "router_noise": 0.0, "grad_norm": 0.719, "lr": 0.0001799, "router_trainable": true, "sec_per_step": 0.936, "probe_sec": 0.0805, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05732, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": 0.02518, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.19482, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 1.4742, "trunc_abs_frac": 0.66089, "mean_delta_full": -0.08988, "read_hit_rate": {"surface": 0.68767, "split": 0.87602, "merge": 0.25888, "skel": 0.67145}, "mean_gate": {"surface": 0.20234, "split": 0.11222, "merge": 0.17885, "skel": 0.15549}, "n_reads": 22214, "n_cand": 33755, "read_suppressed_frac": 0.0998, "dropout_reads_logged": 22214, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3704, "admitted_surface": 16149, "staged_surface": 0, "admitted_split": 15389, "staged_split": 0, "admitted_merge": 22737, "staged_merge": 0, "admitted_skel": 11261, "staged_skel": 0, "mean_count": 1042.8363037109375, "n_util_obs": 438, "util_count_spearman": 0.03980195562324165}, "probe_overhead_sec_total": 466.429, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.007195, "alpha": 0.03, "t": 2976.1} |
| {"event": "consolidate", "step": 3000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3021.4} |
| {"event": "probe", "step": 3000, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3963.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.227719, "cps": 30.960938, "mean_p": 0.227618, "boundary_entropy": 0.09572, "bpb": 1.998689}, "tr": {"chunks": 4376.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.241728, "cps": 34.1875, "mean_p": 0.226386, "boundary_entropy": 0.144197, "bpb": 2.1723}, "fi": {"chunks": 4947.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.262984, "cps": 38.648438, "mean_p": 0.26549, "boundary_entropy": 0.141005, "bpb": 2.29656}}, "bpb": {"en": 1.99869, "tr": 2.1723, "fi": 2.29656}, "gini_cpc": 0.032099, "gini_cps": 0.049375, "bpic": 4.270059, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0615166481747582, "fi": 1.1548619558409843}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3738, "admitted_surface": 16145, "staged_surface": 0, "admitted_split": 15401, "staged_split": 0, "admitted_merge": 22700, "staged_merge": 0, "admitted_skel": 11290, "staged_skel": 0, "mean_count": 1059.162841796875, "n_util_obs": 446, "util_count_spearman": 0.02439413271236818}, "t": 3023.0} |
| {"event": "eval", "step": 3000, "phase": 2, "bpb": {"en": 2.01192, "tr": 2.18203, "fi": 2.30419}, "bytes_per_chunk": {"en": 4.274, "tr": 4.572, "fi": 4.183}, "bpb_mean": 2.16605, "frozen_inventory": true, "t": 3026.0} |
| {"event": "train", "step": 3000, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.40496, "bpb": 2.02693, "chunk_loss": 0.031174, "F": 0.23734, "G": 0.23813, "chunks": 243.03, "bytes_per_chunk": 4.213, "boundary_entropy": 0.09831, "router_noise": 0.0, "grad_norm": 0.6837, "lr": 0.0001761, "router_trainable": true, "sec_per_step": 1.043, "probe_sec": 0.5806, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.5534, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": -0.02495, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.19421, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 0.55657, "trunc_abs_frac": 0.967, "mean_delta_full": -0.14807, "read_hit_rate": {"surface": 0.70962, "split": 0.88181, "merge": 0.28899, "skel": 0.68677}, "mean_gate": {"surface": 0.19769, "split": 0.10927, "merge": 0.17504, "skel": 0.15301}, "n_reads": 22613, "n_cand": 33537, "read_suppressed_frac": 0.098, "dropout_reads_logged": 22613, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3738, "admitted_surface": 16166, "staged_surface": 0, "admitted_split": 15443, "staged_split": 0, "admitted_merge": 22658, "staged_merge": 0, "admitted_skel": 11269, "staged_skel": 0, "mean_count": 1059.5067138671875, "n_util_obs": 446, "util_count_spearman": 0.024124756539363613}, "probe_overhead_sec_total": 472.663, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.007803, "alpha": 0.03, "t": 3028.2} |
| {"event": "consolidate", "step": 3050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3069.5} |
| {"event": "train", "step": 3050, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.43313, "bpb": 2.06757, "chunk_loss": 0.031122, "F": 0.23566, "G": 0.23516, "chunks": 241.31, "bytes_per_chunk": 4.243, "boundary_entropy": 0.09919, "router_noise": 0.0, "grad_norm": 0.6499, "lr": 0.0001724, "router_trainable": true, "sec_per_step": 0.839, "probe_sec": 0.1497, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.12015, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.05288, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": 0.12249, "n_merge": 1, "mean_delta_merge": -0.26276, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.18454, "trunc_abs_frac": 0.26366, "mean_delta_full": -0.49695, "read_hit_rate": {"surface": 0.69701, "split": 0.88587, "merge": 0.25868, "skel": 0.6802}, "mean_gate": {"surface": 0.19879, "split": 0.10864, "merge": 0.17581, "skel": 0.15325}, "n_reads": 22536, "n_cand": 33787, "read_suppressed_frac": 0.1004, "dropout_reads_logged": 22536, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3774, "admitted_surface": 16133, "staged_surface": 0, "admitted_split": 15359, "staged_split": 0, "admitted_merge": 22762, "staged_merge": 0, "admitted_skel": 11282, "staged_skel": 0, "mean_count": 1076.1112060546875, "n_util_obs": 458, "util_count_spearman": 0.025547235671500385}, "probe_overhead_sec_total": 478.078, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.008046, "alpha": 0.03, "t": 3070.2} |
| {"event": "consolidate", "step": 3100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3122.1} |
| {"event": "train", "step": 3100, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.47606, "bpb": 2.12951, "chunk_loss": 0.031356, "F": 0.23779, "G": 0.23886, "chunks": 243.5, "bytes_per_chunk": 4.205, "boundary_entropy": 0.10169, "router_noise": 0.0, "grad_norm": 0.6523, "lr": 0.0001687, "router_trainable": true, "sec_per_step": 1.051, "probe_sec": 0.0805, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05708, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.02432, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": -0.13281, "n_merge": 2, "mean_delta_merge": 0.04413, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.52344, "trunc_abs_frac": 0.67007, "mean_delta_full": -0.06378, "read_hit_rate": {"surface": 0.68879, "split": 0.8742, "merge": 0.26359, "skel": 0.67231}, "mean_gate": {"surface": 0.19751, "split": 0.10636, "merge": 0.17456, "skel": 0.1516}, "n_reads": 21987, "n_cand": 33435, "read_suppressed_frac": 0.0976, "dropout_reads_logged": 21987, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3820, "admitted_surface": 16194, "staged_surface": 0, "admitted_split": 15421, "staged_split": 0, "admitted_merge": 22655, "staged_merge": 0, "admitted_skel": 11266, "staged_skel": 0, "mean_count": 1092.80810546875, "n_util_obs": 465, "util_count_spearman": 0.0249807263781841}, "probe_overhead_sec_total": 484.504, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.007666, "alpha": 0.03, "t": 3122.8} |
| {"event": "consolidate", "step": 3150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3164.9} |
| {"event": "train", "step": 3150, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.44155, "bpb": 2.07972, "chunk_loss": 0.030989, "F": 0.23163, "G": 0.23135, "chunks": 237.19, "bytes_per_chunk": 4.317, "boundary_entropy": 0.09652, "router_noise": 0.0, "grad_norm": 0.6826, "lr": 0.000165, "router_trainable": true, "sec_per_step": 0.853, "probe_sec": 0.0917, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07044, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.18867, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": -0.17663, "n_merge": 1, "mean_delta_merge": 0.02747, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.95113, "trunc_abs_frac": 3.9426, "mean_delta_full": 0.34409, "read_hit_rate": {"surface": 0.69079, "split": 0.87094, "merge": 0.28089, "skel": 0.66315}, "mean_gate": {"surface": 0.19542, "split": 0.10497, "merge": 0.17165, "skel": 0.15189}, "n_reads": 21590, "n_cand": 32717, "read_suppressed_frac": 0.1011, "dropout_reads_logged": 21590, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3874, "admitted_surface": 16184, "staged_surface": 0, "admitted_split": 15513, "staged_split": 0, "admitted_merge": 22526, "staged_merge": 0, "admitted_skel": 11313, "staged_skel": 0, "mean_count": 1109.552490234375, "n_util_obs": 473, "util_count_spearman": 0.024721804993852927}, "probe_overhead_sec_total": 490.327, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.008698, "alpha": 0.03, "t": 3165.4} |
| {"event": "consolidate", "step": 3200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3208.9} |
| {"event": "probe", "step": 3200, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3904.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.224329, "cps": 30.5, "mean_p": 0.223767, "boundary_entropy": 0.091344, "bpb": 1.977785}, "tr": {"chunks": 4389.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.242446, "cps": 34.289062, "mean_p": 0.226776, "boundary_entropy": 0.132514, "bpb": 2.141545}, "fi": {"chunks": 4973.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.264367, "cps": 38.851562, "mean_p": 0.2612, "boundary_entropy": 0.135022, "bpb": 2.272047}}, "bpb": {"en": 1.97778, "tr": 2.14154, "fi": 2.27205}, "gini_cpc": 0.036507, "gini_cps": 0.053721, "bpic": 4.276496, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.080760194058045, "fi": 1.1784763923227688}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3913, "admitted_surface": 16152, "staged_surface": 0, "admitted_split": 15528, "staged_split": 0, "admitted_merge": 22565, "staged_merge": 0, "admitted_skel": 11291, "staged_skel": 0, "mean_count": 1125.8648681640625, "n_util_obs": 480, "util_count_spearman": 0.05088428532276784}, "t": 3210.8} |
| {"event": "eval", "step": 3200, "phase": 2, "bpb": {"en": 1.99066, "tr": 2.15244, "fi": 2.27856}, "bytes_per_chunk": {"en": 4.34, "tr": 4.562, "fi": 4.181}, "bpb_mean": 2.14056, "frozen_inventory": true, "t": 3214.4} |
| {"event": "train", "step": 3200, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.44539, "bpb": 2.08525, "chunk_loss": 0.031123, "F": 0.23514, "G": 0.23549, "chunks": 240.78, "bytes_per_chunk": 4.253, "boundary_entropy": 0.09554, "router_noise": 0.0, "grad_norm": 0.6266, "lr": 0.0001613, "router_trainable": true, "sec_per_step": 1.031, "probe_sec": 0.1322, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07316, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": -0.00343, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.43237, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.02493, "trunc_abs_frac": 0.4438, "mean_delta_full": -0.18159, "read_hit_rate": {"surface": 0.6996, "split": 0.88081, "merge": 0.26803, "skel": 0.66853}, "mean_gate": {"surface": 0.19406, "split": 0.10245, "merge": 0.17025, "skel": 0.15044}, "n_reads": 22075, "n_cand": 33253, "read_suppressed_frac": 0.1021, "dropout_reads_logged": 22075, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3913, "admitted_surface": 16124, "staged_surface": 0, "admitted_split": 15430, "staged_split": 0, "admitted_merge": 22652, "staged_merge": 0, "admitted_skel": 11330, "staged_skel": 0, "mean_count": 1126.19873046875, "n_util_obs": 480, "util_count_spearman": 0.0509804730055252}, "probe_overhead_sec_total": 496.251, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.006712, "alpha": 0.03, "t": 3217.0} |
| {"event": "consolidate", "step": 3250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3265.0} |
| {"event": "train", "step": 3250, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.40714, "bpb": 2.03008, "chunk_loss": 0.030879, "F": 0.22772, "G": 0.22862, "chunks": 233.19, "bytes_per_chunk": 4.391, "boundary_entropy": 0.08791, "router_noise": 0.0, "grad_norm": 0.6033, "lr": 0.0001576, "router_trainable": true, "sec_per_step": 0.981, "probe_sec": 0.0941, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04638, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.06361, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 3, "mean_delta_split": 0.03587, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.39961, "trunc_abs_frac": 0.93337, "mean_delta_full": -0.20854, "read_hit_rate": {"surface": 0.69731, "split": 0.87931, "merge": 0.27088, "skel": 0.67292}, "mean_gate": {"surface": 0.19714, "split": 0.10438, "merge": 0.17234, "skel": 0.15355}, "n_reads": 21662, "n_cand": 32555, "read_suppressed_frac": 0.1001, "dropout_reads_logged": 21662, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3953, "admitted_surface": 16150, "staged_surface": 0, "admitted_split": 15477, "staged_split": 0, "admitted_merge": 22575, "staged_merge": 0, "admitted_skel": 11334, "staged_skel": 0, "mean_count": 1142.72216796875, "n_util_obs": 487, "util_count_spearman": 0.042955436782368704}, "probe_overhead_sec_total": 502.496, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.007667, "alpha": 0.03, "t": 3266.0} |
| {"event": "consolidate", "step": 3300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3306.7} |
| {"event": "train", "step": 3300, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.37027, "bpb": 1.97688, "chunk_loss": 0.031062, "F": 0.23376, "G": 0.23361, "chunks": 239.38, "bytes_per_chunk": 4.278, "boundary_entropy": 0.08435, "router_noise": 0.0, "grad_norm": 0.6569, "lr": 0.0001539, "router_trainable": true, "sec_per_step": 0.825, "probe_sec": 0.1348, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.11147, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.04796, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.10718, "n_merge": 1, "mean_delta_merge": -0.66141, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 0.2182, "trunc_abs_frac": 0.17125, "mean_delta_full": -0.20691, "read_hit_rate": {"surface": 0.7053, "split": 0.88124, "merge": 0.28146, "skel": 0.68657}, "mean_gate": {"surface": 0.1874, "split": 0.09778, "merge": 0.16224, "skel": 0.14529}, "n_reads": 22406, "n_cand": 33332, "read_suppressed_frac": 0.101, "dropout_reads_logged": 22406, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3993, "admitted_surface": 16118, "staged_surface": 0, "admitted_split": 15524, "staged_split": 0, "admitted_merge": 22595, "staged_merge": 0, "admitted_skel": 11299, "staged_skel": 0, "mean_count": 1159.329833984375, "n_util_obs": 497, "util_count_spearman": 0.05641828037795369}, "probe_overhead_sec_total": 508.171, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.008682, "alpha": 0.03, "t": 3307.3} |
| {"event": "consolidate", "step": 3350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3348.5} |
| {"event": "train", "step": 3350, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.35387, "bpb": 1.95323, "chunk_loss": 0.030824, "F": 0.2244, "G": 0.22704, "chunks": 229.78, "bytes_per_chunk": 4.456, "boundary_entropy": 0.08817, "router_noise": 0.0, "grad_norm": 0.6055, "lr": 0.0001501, "router_trainable": true, "sec_per_step": 0.846, "probe_sec": 0.111, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08926, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.12904, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": -0.0076, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -0.49732, "trunc_abs_frac": 2.12885, "mean_delta_full": -0.04549, "read_hit_rate": {"surface": 0.69717, "split": 0.87994, "merge": 0.27164, "skel": 0.67804}, "mean_gate": {"surface": 0.19629, "split": 0.10202, "merge": 0.17158, "skel": 0.15207}, "n_reads": 21527, "n_cand": 32248, "read_suppressed_frac": 0.0998, "dropout_reads_logged": 21527, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4038, "admitted_surface": 16178, "staged_surface": 0, "admitted_split": 15524, "staged_split": 0, "admitted_merge": 22491, "staged_merge": 0, "admitted_skel": 11343, "staged_skel": 0, "mean_count": 1175.97265625, "n_util_obs": 501, "util_count_spearman": 0.05975284633329788}, "probe_overhead_sec_total": 513.768, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.007896, "alpha": 0.03, "t": 3349.6} |
| {"event": "consolidate", "step": 3400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3396.4} |
| {"event": "probe", "step": 3400, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3932.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.225938, "cps": 30.71875, "mean_p": 0.226317, "boundary_entropy": 0.082416, "bpb": 1.967542}, "tr": {"chunks": 4405.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.24333, "cps": 34.414062, "mean_p": 0.227748, "boundary_entropy": 0.117401, "bpb": 2.125659}, "fi": {"chunks": 4997.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.265642, "cps": 39.039062, "mean_p": 0.261812, "boundary_entropy": 0.121146, "bpb": 2.248283}}, "bpb": {"en": 1.96754, "tr": 2.12566, "fi": 2.24828}, "gini_cpc": 0.036017, "gini_cps": 0.053247, "bpic": 4.254687, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0769758686714639, "fi": 1.1757312919385008}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4081, "admitted_surface": 16159, "staged_surface": 0, "admitted_split": 15515, "staged_split": 0, "admitted_merge": 22532, "staged_merge": 0, "admitted_skel": 11330, "staged_skel": 0, "mean_count": 1192.305908203125, "n_util_obs": 510, "util_count_spearman": 0.07028491346586616}, "t": 3398.3} |
| {"event": "eval", "step": 3400, "phase": 2, "bpb": {"en": 1.97966, "tr": 2.13483, "fi": 2.25429}, "bytes_per_chunk": {"en": 4.321, "tr": 4.542, "fi": 4.142}, "bpb_mean": 2.12293, "frozen_inventory": true, "t": 3401.1} |
| {"event": "train", "step": 3400, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.44215, "bpb": 2.08058, "chunk_loss": 0.031126, "F": 0.23462, "G": 0.23579, "chunks": 240.25, "bytes_per_chunk": 4.262, "boundary_entropy": 0.08684, "router_noise": 0.0, "grad_norm": 0.6625, "lr": 0.0001465, "router_trainable": true, "sec_per_step": 1.076, "probe_sec": 0.0684, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04526, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.0364, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 1.18159, "trunc_abs_frac": 1.16097, "mean_delta_full": 0.0542, "read_hit_rate": {"surface": 0.69005, "split": 0.87774, "merge": 0.25223, "skel": 0.67603}, "mean_gate": {"surface": 0.19646, "split": 0.1008, "merge": 0.17166, "skel": 0.14879}, "n_reads": 21876, "n_cand": 33229, "read_suppressed_frac": 0.0988, "dropout_reads_logged": 21876, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4081, "admitted_surface": 16163, "staged_surface": 0, "admitted_split": 15500, "staged_split": 0, "admitted_merge": 22599, "staged_merge": 0, "admitted_skel": 11274, "staged_skel": 0, "mean_count": 1192.63232421875, "n_util_obs": 510, "util_count_spearman": 0.06986805860727827}, "probe_overhead_sec_total": 518.843, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.008233, "alpha": 0.03, "t": 3403.3} |
| {"event": "consolidate", "step": 3450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3439.5} |
| {"event": "train", "step": 3450, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.40605, "bpb": 2.02851, "chunk_loss": 0.030837, "F": 0.22604, "G": 0.22655, "chunks": 231.47, "bytes_per_chunk": 4.424, "boundary_entropy": 0.08748, "router_noise": 0.0, "grad_norm": 0.6225, "lr": 0.0001428, "router_trainable": true, "sec_per_step": 0.741, "probe_sec": 0.0723, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.049, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.11354, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.14355, "n_merge": 1, "mean_delta_merge": 0.00787, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 1.46789, "trunc_abs_frac": 1.33838, "mean_delta_full": -0.20245, "read_hit_rate": {"surface": 0.69586, "split": 0.88025, "merge": 0.26106, "skel": 0.67902}, "mean_gate": {"surface": 0.19348, "split": 0.1, "merge": 0.17073, "skel": 0.1498}, "n_reads": 21585, "n_cand": 32453, "read_suppressed_frac": 0.0987, "dropout_reads_logged": 21585, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4131, "admitted_surface": 16130, "staged_surface": 0, "admitted_split": 15528, "staged_split": 0, "admitted_merge": 22563, "staged_merge": 0, "admitted_skel": 11315, "staged_skel": 0, "mean_count": 1209.22216796875, "n_util_obs": 517, "util_count_spearman": 0.06046046437209518}, "probe_overhead_sec_total": 524.012, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.008693, "alpha": 0.03, "t": 3440.4} |
| {"event": "consolidate", "step": 3500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3476.4} |
| {"event": "train", "step": 3500, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.39793, "bpb": 2.01678, "chunk_loss": 0.030886, "F": 0.22781, "G": 0.22848, "chunks": 233.28, "bytes_per_chunk": 4.39, "boundary_entropy": 0.08612, "router_noise": 0.0, "grad_norm": 0.6503, "lr": 0.0001391, "router_trainable": true, "sec_per_step": 0.73, "probe_sec": 0.0701, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04631, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.03584, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.08118, "n_merge": 2, "mean_delta_merge": 0.16194, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.11642, "trunc_abs_frac": 6.52762, "mean_delta_full": -0.05155, "read_hit_rate": {"surface": 0.69366, "split": 0.87809, "merge": 0.26564, "skel": 0.6673}, "mean_gate": {"surface": 0.18594, "split": 0.09505, "merge": 0.16081, "skel": 0.14355}, "n_reads": 21371, "n_cand": 32324, "read_suppressed_frac": 0.1004, "dropout_reads_logged": 21371, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4169, "admitted_surface": 16154, "staged_surface": 0, "admitted_split": 15519, "staged_split": 0, "admitted_merge": 22542, "staged_merge": 0, "admitted_skel": 11321, "staged_skel": 0, "mean_count": 1225.829833984375, "n_util_obs": 521, "util_count_spearman": 0.068861106490721}, "probe_overhead_sec_total": 529.13, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.007964, "alpha": 0.03, "t": 3476.9} |
| {"event": "consolidate", "step": 3550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3523.6} |
| {"event": "train", "step": 3550, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.40909, "bpb": 2.03289, "chunk_loss": 0.03103, "F": 0.23343, "G": 0.23316, "chunks": 239.03, "bytes_per_chunk": 4.284, "boundary_entropy": 0.0942, "router_noise": 0.0, "grad_norm": 0.6512, "lr": 0.0001355, "router_trainable": true, "sec_per_step": 0.952, "probe_sec": 0.0892, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04972, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": -0.05094, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.31482, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.64335, "trunc_abs_frac": 0.74002, "mean_delta_full": -0.08044, "read_hit_rate": {"surface": 0.69516, "split": 0.88309, "merge": 0.26315, "skel": 0.69387}, "mean_gate": {"surface": 0.19281, "split": 0.09761, "merge": 0.17126, "skel": 0.14651}, "n_reads": 22422, "n_cand": 33510, "read_suppressed_frac": 0.0977, "dropout_reads_logged": 22422, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4210, "admitted_surface": 16184, "staged_surface": 0, "admitted_split": 15492, "staged_split": 0, "admitted_merge": 22611, "staged_merge": 0, "admitted_skel": 11249, "staged_skel": 0, "mean_count": 1242.470703125, "n_util_obs": 528, "util_count_spearman": 0.08136998375732445}, "probe_overhead_sec_total": 534.044, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.007346, "alpha": 0.03, "t": 3524.5} |
| {"event": "consolidate", "step": 3600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3560.0} |
| {"event": "probe", "step": 3600, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3894.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.223755, "cps": 30.421875, "mean_p": 0.22272, "boundary_entropy": 0.086132, "bpb": 1.954035}, "tr": {"chunks": 4385.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.242225, "cps": 34.257812, "mean_p": 0.222772, "boundary_entropy": 0.122818, "bpb": 2.106092}, "fi": {"chunks": 4973.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.264367, "cps": 38.851562, "mean_p": 0.258399, "boundary_entropy": 0.12333, "bpb": 2.224166}}, "bpb": {"en": 1.95403, "tr": 2.10609, "fi": 2.22417}, "gini_cpc": 0.037071, "gini_cps": 0.054281, "bpic": 4.281014, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0825481428092476, "fi": 1.1815027826471725}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4253, "admitted_surface": 16143, "staged_surface": 0, "admitted_split": 15592, "staged_split": 0, "admitted_merge": 22519, "staged_merge": 0, "admitted_skel": 11282, "staged_skel": 0, "mean_count": 1258.7037353515625, "n_util_obs": 537, "util_count_spearman": 0.07578708691389248}, "t": 3561.5} |
| {"event": "eval", "step": 3600, "phase": 2, "bpb": {"en": 1.96701, "tr": 2.11529, "fi": 2.23285}, "bytes_per_chunk": {"en": 4.365, "tr": 4.564, "fi": 4.157}, "bpb_mean": 2.10505, "frozen_inventory": true, "t": 3564.3} |
| {"event": "train", "step": 3600, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.42105, "bpb": 2.05015, "chunk_loss": 0.030947, "F": 0.22897, "G": 0.22873, "chunks": 234.47, "bytes_per_chunk": 4.367, "boundary_entropy": 0.10033, "router_noise": 0.0, "grad_norm": 0.6716, "lr": 0.0001319, "router_trainable": true, "sec_per_step": 0.83, "probe_sec": 0.1031, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0768, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.01732, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.09573, "n_merge": 1, "mean_delta_merge": -0.03894, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -0.36563, "trunc_abs_frac": 0.39682, "mean_delta_full": 0.11765, "read_hit_rate": {"surface": 0.68893, "split": 0.87304, "merge": 0.26133, "skel": 0.69167}, "mean_gate": {"surface": 0.19791, "split": 0.10074, "merge": 0.17322, "skel": 0.1493}, "n_reads": 21833, "n_cand": 32915, "read_suppressed_frac": 0.1012, "dropout_reads_logged": 21833, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4254, "admitted_surface": 16171, "staged_surface": 0, "admitted_split": 15620, "staged_split": 0, "admitted_merge": 22504, "staged_merge": 0, "admitted_skel": 11241, "staged_skel": 0, "mean_count": 1259.0439453125, "n_util_obs": 537, "util_count_spearman": 0.0762867990540254}, "probe_overhead_sec_total": 538.299, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.007748, "alpha": 0.03, "t": 3566.0} |
| {"event": "consolidate", "step": 3650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3603.9} |
| {"event": "train", "step": 3650, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.42612, "bpb": 2.05746, "chunk_loss": 0.031153, "F": 0.2359, "G": 0.23617, "chunks": 241.56, "bytes_per_chunk": 4.239, "boundary_entropy": 0.08873, "router_noise": 0.0, "grad_norm": 0.678, "lr": 0.0001283, "router_trainable": true, "sec_per_step": 0.786, "probe_sec": 0.0984, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06731, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.01989, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": 0.09941, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.80689, "trunc_abs_frac": 0.85265, "mean_delta_full": 0.21371, "read_hit_rate": {"surface": 0.69823, "split": 0.87953, "merge": 0.26415, "skel": 0.67702}, "mean_gate": {"surface": 0.19026, "split": 0.09582, "merge": 0.17225, "skel": 0.14565}, "n_reads": 22276, "n_cand": 33523, "read_suppressed_frac": 0.1019, "dropout_reads_logged": 22276, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4286, "admitted_surface": 16137, "staged_surface": 0, "admitted_split": 15488, "staged_split": 0, "admitted_merge": 22612, "staged_merge": 0, "admitted_skel": 11299, "staged_skel": 0, "mean_count": 1275.69482421875, "n_util_obs": 543, "util_count_spearman": 0.06976457538209688}, "probe_overhead_sec_total": 542.764, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.007969, "alpha": 0.03, "t": 3605.3} |
| {"event": "consolidate", "step": 3700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3652.1} |
| {"event": "train", "step": 3700, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.5036, "bpb": 2.16923, "chunk_loss": 0.031326, "F": 0.23993, "G": 0.23992, "chunks": 245.69, "bytes_per_chunk": 4.168, "boundary_entropy": 0.09367, "router_noise": 0.0, "grad_norm": 0.6969, "lr": 0.0001247, "router_trainable": true, "sec_per_step": 0.95, "probe_sec": 0.1378, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09476, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.10583, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.0372, "n_merge": 2, "mean_delta_merge": -0.11273, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 1.36686, "trunc_abs_frac": 1.36686, "mean_delta_full": 0.53955, "read_hit_rate": {"surface": 0.67599, "split": 0.87586, "merge": 0.24596, "skel": 0.67659}, "mean_gate": {"surface": 0.18716, "split": 0.0933, "merge": 0.16543, "skel": 0.1414}, "n_reads": 22178, "n_cand": 33977, "read_suppressed_frac": 0.0996, "dropout_reads_logged": 22178, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4324, "admitted_surface": 16201, "staged_surface": 0, "admitted_split": 15448, "staged_split": 0, "admitted_merge": 22648, "staged_merge": 0, "admitted_skel": 11239, "staged_skel": 0, "mean_count": 1292.282958984375, "n_util_obs": 550, "util_count_spearman": 0.0674850159078518}, "probe_overhead_sec_total": 547.906, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.006692, "alpha": 0.03, "t": 3652.8} |
| {"event": "consolidate", "step": 3750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3687.9} |
| {"event": "train", "step": 3750, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.34651, "bpb": 1.94261, "chunk_loss": 0.031162, "F": 0.23709, "G": 0.23762, "chunks": 242.78, "bytes_per_chunk": 4.218, "boundary_entropy": 0.08007, "router_noise": 0.0, "grad_norm": 0.6243, "lr": 0.0001212, "router_trainable": true, "sec_per_step": 0.717, "probe_sec": 0.0969, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0731, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.05356, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 3, "mean_delta_merge": 0.07719, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.61804, "trunc_abs_frac": 0.74172, "mean_delta_full": 0.11066, "read_hit_rate": {"surface": 0.71126, "split": 0.88552, "merge": 0.29124, "skel": 0.68282}, "mean_gate": {"surface": 0.18876, "split": 0.09369, "merge": 0.16679, "skel": 0.14535}, "n_reads": 22601, "n_cand": 33449, "read_suppressed_frac": 0.1042, "dropout_reads_logged": 22601, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4367, "admitted_surface": 16129, "staged_surface": 0, "admitted_split": 15487, "staged_split": 0, "admitted_merge": 22645, "staged_merge": 0, "admitted_skel": 11275, "staged_skel": 0, "mean_count": 1308.992919921875, "n_util_obs": 552, "util_count_spearman": 0.06617877291729693}, "probe_overhead_sec_total": 552.233, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.006783, "alpha": 0.03, "t": 3688.6} |
| {"event": "consolidate", "step": 3800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3721.8} |
| {"event": "probe", "step": 3800, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3914.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.224904, "cps": 30.578125, "mean_p": 0.223903, "boundary_entropy": 0.08237, "bpb": 1.937015}, "tr": {"chunks": 4401.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.243109, "cps": 34.382812, "mean_p": 0.222166, "boundary_entropy": 0.116624, "bpb": 2.081471}, "fi": {"chunks": 4974.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.26442, "cps": 38.859375, "mean_p": 0.25913, "boundary_entropy": 0.117613, "bpb": 2.206981}}, "bpb": {"en": 1.93701, "tr": 2.08147, "fi": 2.20698}, "gini_cpc": 0.035968, "gini_cps": 0.053177, "bpic": 4.269095, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0809462918019415, "fi": 1.1757018359239082}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4416, "admitted_surface": 16150, "staged_surface": 0, "admitted_split": 15492, "staged_split": 0, "admitted_merge": 22618, "staged_merge": 0, "admitted_skel": 11276, "staged_skel": 0, "mean_count": 1325.29931640625, "n_util_obs": 561, "util_count_spearman": 0.08751230339900307}, "t": 3723.5} |
| {"event": "eval", "step": 3800, "phase": 2, "bpb": {"en": 1.95046, "tr": 2.0921, "fi": 2.21552}, "bytes_per_chunk": {"en": 4.34, "tr": 4.543, "fi": 4.187}, "bpb_mean": 2.08603, "frozen_inventory": true, "t": 3726.1} |
| {"event": "train", "step": 3800, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.38003, "bpb": 1.99097, "chunk_loss": 0.030975, "F": 0.23087, "G": 0.22965, "chunks": 236.41, "bytes_per_chunk": 4.332, "boundary_entropy": 0.08543, "router_noise": 0.0, "grad_norm": 0.6453, "lr": 0.0001177, "router_trainable": true, "sec_per_step": 0.784, "probe_sec": 0.0783, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05402, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.03096, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.16785, "n_merge": 1, "mean_delta_merge": -0.65723, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.53064, "trunc_abs_frac": 4.84074, "mean_delta_full": -0.3978, "read_hit_rate": {"surface": 0.70742, "split": 0.88072, "merge": 0.27836, "skel": 0.68851}, "mean_gate": {"surface": 0.19041, "split": 0.09636, "merge": 0.16676, "skel": 0.14652}, "n_reads": 22045, "n_cand": 32802, "read_suppressed_frac": 0.0987, "dropout_reads_logged": 22045, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4417, "admitted_surface": 16133, "staged_surface": 0, "admitted_split": 15524, "staged_split": 0, "admitted_merge": 22626, "staged_merge": 0, "admitted_skel": 11253, "staged_skel": 0, "mean_count": 1325.6405029296875, "n_util_obs": 561, "util_count_spearman": 0.08721189354332766}, "probe_overhead_sec_total": 556.342, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.006975, "alpha": 0.03, "t": 3727.8} |
| {"event": "consolidate", "step": 3850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3765.3} |
| {"event": "train", "step": 3850, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.43838, "bpb": 2.07514, "chunk_loss": 0.03085, "F": 0.22748, "G": 0.2273, "chunks": 232.94, "bytes_per_chunk": 4.396, "boundary_entropy": 0.09096, "router_noise": 0.0, "grad_norm": 0.7298, "lr": 0.0001142, "router_trainable": true, "sec_per_step": 0.762, "probe_sec": 0.0808, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04418, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.01937, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.38037, "n_merge": 1, "mean_delta_merge": -0.01651, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.30804, "trunc_abs_frac": 0.64804, "mean_delta_full": 0.11869, "read_hit_rate": {"surface": 0.68108, "split": 0.871, "merge": 0.25237, "skel": 0.66857}, "mean_gate": {"surface": 0.19207, "split": 0.09532, "merge": 0.17041, "skel": 0.14757}, "n_reads": 21324, "n_cand": 32608, "read_suppressed_frac": 0.1061, "dropout_reads_logged": 21324, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4458, "admitted_surface": 16156, "staged_surface": 0, "admitted_split": 15556, "staged_split": 0, "admitted_merge": 22515, "staged_merge": 0, "admitted_skel": 11309, "staged_skel": 0, "mean_count": 1342.31640625, "n_util_obs": 566, "util_count_spearman": 0.09329009221312884}, "probe_overhead_sec_total": 561.011, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.008849, "alpha": 0.03, "t": 3765.9} |
| {"event": "consolidate", "step": 3900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3807.0} |
| {"event": "train", "step": 3900, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.40202, "bpb": 2.02269, "chunk_loss": 0.030972, "F": 0.23178, "G": 0.23124, "chunks": 237.34, "bytes_per_chunk": 4.314, "boundary_entropy": 0.08257, "router_noise": 0.0, "grad_norm": 0.6931, "lr": 0.0001108, "router_trainable": true, "sec_per_step": 0.844, "probe_sec": 0.1984, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.11702, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.0071, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.16028, "n_merge": 1, "mean_delta_merge": 0.20123, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 4.20898, "trunc_abs_frac": 1.87435, "mean_delta_full": 0.01562, "read_hit_rate": {"surface": 0.70356, "split": 0.88096, "merge": 0.27619, "skel": 0.67491}, "mean_gate": {"surface": 0.18297, "split": 0.09129, "merge": 0.15998, "skel": 0.14171}, "n_reads": 21971, "n_cand": 32881, "read_suppressed_frac": 0.0978, "dropout_reads_logged": 21971, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4504, "admitted_surface": 16106, "staged_surface": 0, "admitted_split": 15540, "staged_split": 0, "admitted_merge": 22582, "staged_merge": 0, "admitted_skel": 11308, "staged_skel": 0, "mean_count": 1358.847412109375, "n_util_obs": 569, "util_count_spearman": 0.08877803051261257}, "probe_overhead_sec_total": 565.84, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.008839, "alpha": 0.03, "t": 3808.1} |
| {"event": "consolidate", "step": 3950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3843.8} |
| {"event": "train", "step": 3950, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.29524, "bpb": 1.86864, "chunk_loss": 0.031009, "F": 0.23215, "G": 0.23165, "chunks": 237.72, "bytes_per_chunk": 4.308, "boundary_entropy": 0.0789, "router_noise": 0.0, "grad_norm": 0.5903, "lr": 0.0001074, "router_trainable": true, "sec_per_step": 0.735, "probe_sec": 0.0951, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05896, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.08686, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": -0.01225, "n_merge": 2, "mean_delta_merge": -0.32372, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -0.36303, "trunc_abs_frac": 0.54289, "mean_delta_full": 0.20853, "read_hit_rate": {"surface": 0.72449, "split": 0.89281, "merge": 0.29259, "skel": 0.70355}, "mean_gate": {"surface": 0.18103, "split": 0.08982, "merge": 0.16306, "skel": 0.13903}, "n_reads": 22775, "n_cand": 33154, "read_suppressed_frac": 0.1037, "dropout_reads_logged": 22775, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4548, "admitted_surface": 16083, "staged_surface": 0, "admitted_split": 15509, "staged_split": 0, "admitted_merge": 22665, "staged_merge": 0, "admitted_skel": 11279, "staged_skel": 0, "mean_count": 1375.438232421875, "n_util_obs": 578, "util_count_spearman": 0.10304714451916147}, "probe_overhead_sec_total": 570.237, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.00866, "alpha": 0.03, "t": 3844.8} |
| {"event": "consolidate", "step": 4000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3881.4} |
| {"event": "probe", "step": 4000, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3896.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.223869, "cps": 30.4375, "mean_p": 0.223331, "boundary_entropy": 0.079236, "bpb": 1.927932}, "tr": {"chunks": 4407.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.24344, "cps": 34.429688, "mean_p": 0.223833, "boundary_entropy": 0.11567, "bpb": 2.0766}, "fi": {"chunks": 4969.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.264154, "cps": 38.820312, "mean_p": 0.258761, "boundary_entropy": 0.113099, "bpb": 2.19395}}, "bpb": {"en": 1.92793, "tr": 2.0766, "fi": 2.19395}, "gini_cpc": 0.036716, "gini_cps": 0.053898, "bpic": 4.274563, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0874208881847778, "fi": 1.1799464155327877}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4595, "admitted_surface": 16117, "staged_surface": 0, "admitted_split": 15520, "staged_split": 0, "admitted_merge": 22616, "staged_merge": 0, "admitted_skel": 11283, "staged_skel": 0, "mean_count": 1391.649658203125, "n_util_obs": 587, "util_count_spearman": 0.09949630876948339}, "t": 3883.0} |
| {"event": "eval", "step": 4000, "phase": 2, "bpb": {"en": 1.93949, "tr": 2.08403, "fi": 2.20168}, "bytes_per_chunk": {"en": 4.355, "tr": 4.547, "fi": 4.165}, "bpb_mean": 2.07507, "frozen_inventory": true, "t": 3885.6} |
| {"event": "train", "step": 4000, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.35649, "bpb": 1.957, "chunk_loss": 0.030956, "F": 0.22983, "G": 0.23074, "chunks": 235.34, "bytes_per_chunk": 4.351, "boundary_entropy": 0.08618, "router_noise": 0.0, "grad_norm": 0.8506, "lr": 0.000104, "router_trainable": true, "sec_per_step": 0.85, "probe_sec": 0.0758, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05121, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.0948, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 3, "mean_delta_merge": -0.09098, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.05391, "trunc_abs_frac": 0.05391, "mean_delta_full": -0.21851, "read_hit_rate": {"surface": 0.69529, "split": 0.87562, "merge": 0.26798, "skel": 0.6887}, "mean_gate": {"surface": 0.18072, "split": 0.09002, "merge": 0.16162, "skel": 0.13709}, "n_reads": 21781, "n_cand": 32738, "read_suppressed_frac": 0.1025, "dropout_reads_logged": 21781, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4596, "admitted_surface": 16145, "staged_surface": 0, "admitted_split": 15575, "staged_split": 0, "admitted_merge": 22589, "staged_merge": 0, "admitted_skel": 11227, "staged_skel": 0, "mean_count": 1391.9898681640625, "n_util_obs": 587, "util_count_spearman": 0.09943881733726219}, "probe_overhead_sec_total": 574.809, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.007215, "alpha": 0.03, "t": 3887.3} |
| {"event": "consolidate", "step": 4050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3928.4} |
| {"event": "train", "step": 4050, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.4285, "bpb": 2.0609, "chunk_loss": 0.031052, "F": 0.23276, "G": 0.23251, "chunks": 238.34, "bytes_per_chunk": 4.296, "boundary_entropy": 0.08561, "router_noise": 0.0, "grad_norm": 0.6957, "lr": 0.0001007, "router_trainable": true, "sec_per_step": 0.834, "probe_sec": 0.0956, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06742, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.00794, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.02612, "n_merge": 3, "mean_delta_merge": -0.03981, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.3347, "trunc_abs_frac": 1.00926, "mean_delta_full": -0.22893, "read_hit_rate": {"surface": 0.69111, "split": 0.87645, "merge": 0.26471, "skel": 0.67601}, "mean_gate": {"surface": 0.18179, "split": 0.08967, "merge": 0.16192, "skel": 0.13896}, "n_reads": 21925, "n_cand": 33133, "read_suppressed_frac": 0.1013, "dropout_reads_logged": 21925, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4644, "admitted_surface": 16161, "staged_surface": 0, "admitted_split": 15529, "staged_split": 0, "admitted_merge": 22566, "staged_merge": 0, "admitted_skel": 11280, "staged_skel": 0, "mean_count": 1408.572509765625, "n_util_obs": 599, "util_count_spearman": 0.08605848217009511}, "probe_overhead_sec_total": 579.332, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.007629, "alpha": 0.03, "t": 3929.0} |
| {"event": "probe", "step": 4096, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3913.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.224846, "cps": 30.570312, "mean_p": 0.223754, "boundary_entropy": 0.078922, "bpb": 1.915963}, "tr": {"chunks": 4385.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.242225, "cps": 34.257812, "mean_p": 0.22024, "boundary_entropy": 0.122197, "bpb": 2.064768}, "fi": {"chunks": 4934.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.262293, "cps": 38.546875, "mean_p": 0.256423, "boundary_entropy": 0.116541, "bpb": 2.183871}}, "bpb": {"en": 1.91596, "tr": 2.06477, "fi": 2.18387}, "gini_cpc": 0.034228, "gini_cps": 0.051441, "bpic": 4.287485, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0772917117554843, "fi": 1.1665451006089154}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4681, "admitted_surface": 16154, "staged_surface": 0, "admitted_split": 15550, "staged_split": 0, "admitted_merge": 22558, "staged_merge": 0, "admitted_skel": 11274, "staged_skel": 0, "mean_count": 1423.52880859375, "n_util_obs": 602, "util_count_spearman": 0.07569306209108088}, "t": 3968.6} |
| {"event": "consolidate", "step": 4100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3972.6} |
| {"event": "train", "step": 4100, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.36401, "bpb": 1.96785, "chunk_loss": 0.030773, "F": 0.22394, "G": 0.22427, "chunks": 229.31, "bytes_per_chunk": 4.466, "boundary_entropy": 0.08208, "router_noise": 0.0, "grad_norm": 0.6635, "lr": 9.75e-05, "router_trainable": true, "sec_per_step": 0.893, "probe_sec": 0.0808, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04417, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.02403, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": -0.12662, "n_merge": 2, "mean_delta_merge": 0.00824, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -0.80207, "trunc_abs_frac": 0.61329, "mean_delta_full": 0.08981, "read_hit_rate": {"surface": 0.67999, "split": 0.8733, "merge": 0.25419, "skel": 0.66935}, "mean_gate": {"surface": 0.17456, "split": 0.08466, "merge": 0.15514, "skel": 0.13548}, "n_reads": 21260, "n_cand": 32394, "read_suppressed_frac": 0.0991, "dropout_reads_logged": 21260, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4684, "admitted_surface": 16147, "staged_surface": 0, "admitted_split": 15539, "staged_split": 0, "admitted_merge": 22499, "staged_merge": 0, "admitted_skel": 11351, "staged_skel": 0, "mean_count": 1425.068115234375, "n_util_obs": 602, "util_count_spearman": 0.07543179594214146}, "probe_overhead_sec_total": 584.648, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.007335, "alpha": 0.03, "t": 3973.6} |
| {"event": "consolidate", "step": 4150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4010.7} |
| {"event": "train", "step": 4150, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.35396, "bpb": 1.95335, "chunk_loss": 0.030901, "F": 0.2291, "G": 0.22864, "chunks": 234.59, "bytes_per_chunk": 4.365, "boundary_entropy": 0.08388, "router_noise": 0.0, "grad_norm": 0.6319, "lr": 9.43e-05, "router_trainable": true, "sec_per_step": 0.753, "probe_sec": 0.072, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04299, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": -0.03991, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.01721, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.40711, "trunc_abs_frac": 0.35434, "mean_delta_full": -0.14044, "read_hit_rate": {"surface": 0.70609, "split": 0.88031, "merge": 0.28738, "skel": 0.69292}, "mean_gate": {"surface": 0.18242, "split": 0.08915, "merge": 0.15932, "skel": 0.14011}, "n_reads": 22199, "n_cand": 32864, "read_suppressed_frac": 0.102, "dropout_reads_logged": 22199, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4733, "admitted_surface": 16127, "staged_surface": 0, "admitted_split": 15586, "staged_split": 0, "admitted_merge": 22532, "staged_merge": 0, "admitted_skel": 11291, "staged_skel": 0, "mean_count": 1441.7598876953125, "n_util_obs": 608, "util_count_spearman": 0.07828932302899738}, "probe_overhead_sec_total": 589.086, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.007159, "alpha": 0.03, "t": 4011.2} |
| {"event": "consolidate", "step": 4200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4052.7} |
| {"event": "probe", "step": 4200, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3907.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.224502, "cps": 30.523438, "mean_p": 0.223969, "boundary_entropy": 0.076257, "bpb": 1.915599}, "tr": {"chunks": 4399.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.242998, "cps": 34.367188, "mean_p": 0.223734, "boundary_entropy": 0.110277, "bpb": 2.051007}, "fi": {"chunks": 4970.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.264207, "cps": 38.828125, "mean_p": 0.256857, "boundary_entropy": 0.110337, "bpb": 2.170559}}, "bpb": {"en": 1.9156, "tr": 2.05101, "fi": 2.17056}, "gini_cpc": 0.036176, "gini_cps": 0.05338, "bpic": 4.273275, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0823908680217815, "fi": 1.176861117249508}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4774, "admitted_surface": 16130, "staged_surface": 0, "admitted_split": 15528, "staged_split": 0, "admitted_merge": 22581, "staged_merge": 0, "admitted_skel": 11297, "staged_skel": 0, "mean_count": 1458.141845703125, "n_util_obs": 618, "util_count_spearman": 0.07571365731535167}, "t": 4054.2} |
| {"event": "eval", "step": 4200, "phase": 2, "bpb": {"en": 1.92931, "tr": 2.06016, "fi": 2.17755}, "bytes_per_chunk": {"en": 4.349, "tr": 4.554, "fi": 4.176}, "bpb_mean": 2.05567, "frozen_inventory": true, "t": 4057.1} |
| {"event": "phase_transition", "step": 4200, "phase": 3, "objective": "global", "router_trainable": true, "encoder_trainable": true, "frozen_router_params": 0, "t": 4057.8} |
| {"event": "train", "step": 4200, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.36889, "bpb": 1.97489, "chunk_loss": 0.030817, "F": 0.22513, "G": 0.22522, "chunks": 230.53, "bytes_per_chunk": 4.442, "boundary_entropy": 0.08553, "router_noise": 0.0, "grad_norm": 0.6479, "lr": 9.12e-05, "router_trainable": true, "sec_per_step": 0.959, "probe_sec": 0.0865, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05613, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.10565, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": -0.12036, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 1.73103, "trunc_abs_frac": 1.35104, "mean_delta_full": 0.01416, "read_hit_rate": {"surface": 0.68155, "split": 0.87257, "merge": 0.25106, "skel": 0.67951}, "mean_gate": {"surface": 0.18386, "split": 0.09086, "merge": 0.16248, "skel": 0.14046}, "n_reads": 21320, "n_cand": 32433, "read_suppressed_frac": 0.104, "dropout_reads_logged": 21320, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4775, "admitted_surface": 16167, "staged_surface": 0, "admitted_split": 15599, "staged_split": 0, "admitted_merge": 22505, "staged_merge": 0, "admitted_skel": 11265, "staged_skel": 0, "mean_count": 1458.4642333984375, "n_util_obs": 619, "util_count_spearman": 0.0737490684134873}, "probe_overhead_sec_total": 593.692, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.008462, "alpha": 0.03, "t": 4059.2} |
| {"event": "consolidate", "step": 4250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4106.8} |
| {"event": "train", "step": 4250, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.40548, "bpb": 2.02767, "chunk_loss": 0.031127, "F": 0.23538, "G": 0.23448, "chunks": 241.03, "bytes_per_chunk": 4.248, "boundary_entropy": 0.07923, "router_noise": 0.0, "grad_norm": 0.6499, "lr": 8.81e-05, "router_trainable": true, "sec_per_step": 0.962, "probe_sec": 0.0877, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06353, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.16065, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": 0.07471, "n_merge": 1, "mean_delta_merge": 0.06897, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 1.58394, "trunc_abs_frac": 1.15588, "mean_delta_full": 0.07296, "read_hit_rate": {"surface": 0.70264, "split": 0.87437, "merge": 0.28069, "skel": 0.66312}, "mean_gate": {"surface": 0.18052, "split": 0.08807, "merge": 0.15887, "skel": 0.13776}, "n_reads": 21864, "n_cand": 32979, "read_suppressed_frac": 0.0992, "dropout_reads_logged": 21864, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4807, "admitted_surface": 16122, "staged_surface": 0, "admitted_split": 15605, "staged_split": 0, "admitted_merge": 22485, "staged_merge": 0, "admitted_skel": 11324, "staged_skel": 0, "mean_count": 1475.138671875, "n_util_obs": 626, "util_count_spearman": 0.06854137693310762}, "probe_overhead_sec_total": 599.146, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.007422, "alpha": 0.03, "t": 4107.3} |
| {"event": "consolidate", "step": 4300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4145.6} |
| {"event": "train", "step": 4300, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35844, "bpb": 1.95982, "chunk_loss": 0.030876, "F": 0.22806, "G": 0.22749, "chunks": 233.53, "bytes_per_chunk": 4.385, "boundary_entropy": 0.08103, "router_noise": 0.0, "grad_norm": 0.6311, "lr": 8.5e-05, "router_trainable": true, "sec_per_step": 0.778, "probe_sec": 0.0999, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05692, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": 0.03773, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.02472, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.08974, "trunc_abs_frac": 0.8897, "mean_delta_full": 0.1429, "read_hit_rate": {"surface": 0.69278, "split": 0.87699, "merge": 0.27345, "skel": 0.67762}, "mean_gate": {"surface": 0.17441, "split": 0.08447, "merge": 0.1517, "skel": 0.13378}, "n_reads": 21752, "n_cand": 32696, "read_suppressed_frac": 0.0982, "dropout_reads_logged": 21752, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4838, "admitted_surface": 16149, "staged_surface": 0, "admitted_split": 15572, "staged_split": 0, "admitted_merge": 22486, "staged_merge": 0, "admitted_skel": 11329, "staged_skel": 0, "mean_count": 1491.812255859375, "n_util_obs": 632, "util_count_spearman": 0.06499013507725095}, "probe_overhead_sec_total": 603.948, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.007558, "alpha": 0.03, "t": 4146.2} |
| {"event": "consolidate", "step": 4350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4202.7} |
| {"event": "train", "step": 4350, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.40987, "bpb": 2.03401, "chunk_loss": 0.030834, "F": 0.22681, "G": 0.22613, "chunks": 232.25, "bytes_per_chunk": 4.409, "boundary_entropy": 0.0808, "router_noise": 0.0, "grad_norm": 0.6339, "lr": 8.21e-05, "router_trainable": true, "sec_per_step": 1.148, "probe_sec": 0.0692, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04698, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.03082, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": 0.04007, "n_merge": 2, "mean_delta_merge": -0.47064, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": -8.45985, "trunc_abs_frac": 8.45985, "mean_delta_full": 0.01672, "read_hit_rate": {"surface": 0.67824, "split": 0.87171, "merge": 0.25136, "skel": 0.66906}, "mean_gate": {"surface": 0.17597, "split": 0.08591, "merge": 0.15549, "skel": 0.13424}, "n_reads": 21162, "n_cand": 32395, "read_suppressed_frac": 0.1015, "dropout_reads_logged": 21162, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4873, "admitted_surface": 16154, "staged_surface": 0, "admitted_split": 15589, "staged_split": 0, "admitted_merge": 22493, "staged_merge": 0, "admitted_skel": 11300, "staged_skel": 0, "mean_count": 1508.440673828125, "n_util_obs": 641, "util_count_spearman": 0.06917519226864215}, "probe_overhead_sec_total": 609.228, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.007373, "alpha": 0.03, "t": 4203.6} |
| {"event": "consolidate", "step": 4400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4241.9} |
| {"event": "probe", "step": 4400, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3897.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.223927, "cps": 30.445312, "mean_p": 0.223833, "boundary_entropy": 0.075763, "bpb": 1.904287}, "tr": {"chunks": 4409.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.243551, "cps": 34.445312, "mean_p": 0.223309, "boundary_entropy": 0.108001, "bpb": 2.039226}, "fi": {"chunks": 4982.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.264845, "cps": 38.921875, "mean_p": 0.258536, "boundary_entropy": 0.109756, "bpb": 2.158494}}, "bpb": {"en": 1.90429, "tr": 2.03923, "fi": 2.15849}, "gini_cpc": 0.03725, "gini_cps": 0.054435, "bpic": 4.269416, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0876352181471884, "fi": 1.1827298401926805}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4906, "admitted_surface": 16152, "staged_surface": 0, "admitted_split": 15589, "staged_split": 0, "admitted_merge": 22510, "staged_merge": 0, "admitted_skel": 11285, "staged_skel": 0, "mean_count": 1524.673828125, "n_util_obs": 652, "util_count_spearman": 0.055283131496643816}, "t": 4243.5} |
| {"event": "eval", "step": 4400, "phase": 3, "bpb": {"en": 1.91762, "tr": 2.04931, "fi": 2.16651}, "bytes_per_chunk": {"en": 4.359, "tr": 4.544, "fi": 4.186}, "bpb_mean": 2.04448, "frozen_inventory": true, "t": 4246.1} |
| {"event": "train", "step": 4400, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.33836, "bpb": 1.93085, "chunk_loss": 0.031032, "F": 0.23285, "G": 0.23351, "chunks": 238.44, "bytes_per_chunk": 4.295, "boundary_entropy": 0.07823, "router_noise": 0.0, "grad_norm": 0.6223, "lr": 7.92e-05, "router_trainable": true, "sec_per_step": 0.889, "probe_sec": 0.0654, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04296, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.04764, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.23082, "n_merge": 2, "mean_delta_merge": -0.0233, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 1.72493, "trunc_abs_frac": 1.72493, "mean_delta_full": 0.08521, "read_hit_rate": {"surface": 0.7065, "split": 0.88627, "merge": 0.27795, "skel": 0.69393}, "mean_gate": {"surface": 0.17547, "split": 0.08608, "merge": 0.1534, "skel": 0.13602}, "n_reads": 22363, "n_cand": 33132, "read_suppressed_frac": 0.1004, "dropout_reads_logged": 22363, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4906, "admitted_surface": 16142, "staged_surface": 0, "admitted_split": 15522, "staged_split": 0, "admitted_merge": 22620, "staged_merge": 0, "admitted_skel": 11252, "staged_skel": 0, "mean_count": 1525.022705078125, "n_util_obs": 652, "util_count_spearman": 0.05518740945084366}, "probe_overhead_sec_total": 614.215, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.006652, "alpha": 0.03, "t": 4248.0} |
| {"event": "consolidate", "step": 4450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4287.0} |
| {"event": "train", "step": 4450, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.31685, "bpb": 1.89981, "chunk_loss": 0.030703, "F": 0.2215, "G": 0.22075, "chunks": 226.81, "bytes_per_chunk": 4.515, "boundary_entropy": 0.07753, "router_noise": 0.0, "grad_norm": 0.6315, "lr": 7.63e-05, "router_trainable": true, "sec_per_step": 0.8, "probe_sec": 0.0971, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06469, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.01666, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.07648, "n_merge": 1, "mean_delta_merge": -0.27496, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.43971, "trunc_abs_frac": 0.55151, "mean_delta_full": 0.15717, "read_hit_rate": {"surface": 0.68821, "split": 0.87629, "merge": 0.26995, "skel": 0.66509}, "mean_gate": {"surface": 0.17755, "split": 0.08726, "merge": 0.15328, "skel": 0.13925}, "n_reads": 21118, "n_cand": 31904, "read_suppressed_frac": 0.1005, "dropout_reads_logged": 21118, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4950, "admitted_surface": 16135, "staged_surface": 0, "admitted_split": 15612, "staged_split": 0, "admitted_merge": 22425, "staged_merge": 0, "admitted_skel": 11364, "staged_skel": 0, "mean_count": 1541.740478515625, "n_util_obs": 658, "util_count_spearman": 0.06034014328600817}, "probe_overhead_sec_total": 619.094, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.007765, "alpha": 0.03, "t": 4288.0} |
| {"event": "consolidate", "step": 4500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4324.5} |
| {"event": "train", "step": 4500, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.41641, "bpb": 2.04345, "chunk_loss": 0.031037, "F": 0.23187, "G": 0.23194, "chunks": 237.44, "bytes_per_chunk": 4.313, "boundary_entropy": 0.08448, "router_noise": 0.0, "grad_norm": 0.6959, "lr": 7.36e-05, "router_trainable": true, "sec_per_step": 0.75, "probe_sec": 0.0876, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05968, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": 0.03152, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.06641, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -6.96787, "trunc_abs_frac": 4.82421, "mean_delta_full": -0.0076, "read_hit_rate": {"surface": 0.67975, "split": 0.87069, "merge": 0.24861, "skel": 0.66642}, "mean_gate": {"surface": 0.17536, "split": 0.08265, "merge": 0.15567, "skel": 0.13174}, "n_reads": 21594, "n_cand": 33121, "read_suppressed_frac": 0.0975, "dropout_reads_logged": 21594, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4988, "admitted_surface": 16130, "staged_surface": 0, "admitted_split": 15584, "staged_split": 0, "admitted_merge": 22528, "staged_merge": 0, "admitted_skel": 11294, "staged_skel": 0, "mean_count": 1558.3740234375, "n_util_obs": 665, "util_count_spearman": 0.0654511781213694}, "probe_overhead_sec_total": 623.278, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.007382, "alpha": 0.03, "t": 4325.5} |
| {"event": "consolidate", "step": 4550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4361.3} |
| {"event": "train", "step": 4550, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35956, "bpb": 1.96143, "chunk_loss": 0.030999, "F": 0.23141, "G": 0.23139, "chunks": 236.97, "bytes_per_chunk": 4.321, "boundary_entropy": 0.08055, "router_noise": 0.0, "grad_norm": 0.6899, "lr": 7.09e-05, "router_trainable": true, "sec_per_step": 0.735, "probe_sec": 0.1588, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.12324, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.00036, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 3, "mean_delta_split": 0.01947, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.29779, "trunc_abs_frac": 0.50321, "mean_delta_full": -0.17432, "read_hit_rate": {"surface": 0.70189, "split": 0.88644, "merge": 0.27703, "skel": 0.69321}, "mean_gate": {"surface": 0.16888, "split": 0.08038, "merge": 0.14847, "skel": 0.13149}, "n_reads": 22446, "n_cand": 33267, "read_suppressed_frac": 0.0997, "dropout_reads_logged": 22446, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5024, "admitted_surface": 16125, "staged_surface": 0, "admitted_split": 15536, "staged_split": 0, "admitted_merge": 22582, "staged_merge": 0, "admitted_skel": 11293, "staged_skel": 0, "mean_count": 1575.0318603515625, "n_util_obs": 671, "util_count_spearman": 0.05942934874732182}, "probe_overhead_sec_total": 627.589, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.008107, "alpha": 0.03, "t": 4362.3} |
| {"event": "consolidate", "step": 4600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4406.6} |
| {"event": "probe", "step": 4600, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3921.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.225306, "cps": 30.632812, "mean_p": 0.223839, "boundary_entropy": 0.074128, "bpb": 1.893914}, "tr": {"chunks": 4411.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.243661, "cps": 34.460938, "mean_p": 0.22216, "boundary_entropy": 0.106613, "bpb": 2.027571}, "fi": {"chunks": 4983.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.264898, "cps": 38.929688, "mean_p": 0.258299, "boundary_entropy": 0.107905, "bpb": 2.146791}}, "bpb": {"en": 1.89391, "tr": 2.02757, "fi": 2.14679}, "gini_cpc": 0.035967, "gini_cps": 0.053173, "bpic": 4.260759, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0814682759171874, "fi": 1.1757264313718345}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5051, "admitted_surface": 16148, "staged_surface": 0, "admitted_split": 15539, "staged_split": 0, "admitted_merge": 22594, "staged_merge": 0, "admitted_skel": 11255, "staged_skel": 0, "mean_count": 1591.3681640625, "n_util_obs": 685, "util_count_spearman": 0.06937390594131797}, "t": 4408.5} |
| {"event": "eval", "step": 4600, "phase": 3, "bpb": {"en": 1.90753, "tr": 2.0372, "fi": 2.15402}, "bytes_per_chunk": {"en": 4.336, "tr": 4.541, "fi": 4.185}, "bpb_mean": 2.03291, "frozen_inventory": true, "t": 4411.9} |
| {"event": "train", "step": 4600, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.39153, "bpb": 2.00755, "chunk_loss": 0.03108, "F": 0.233, "G": 0.23245, "chunks": 238.59, "bytes_per_chunk": 4.292, "boundary_entropy": 0.0766, "router_noise": 0.0, "grad_norm": 0.6668, "lr": 6.82e-05, "router_trainable": true, "sec_per_step": 1.079, "probe_sec": 0.2149, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.12397, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.00903, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": -0.03814, "n_merge": 1, "mean_delta_merge": -0.38931, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.56708, "trunc_abs_frac": 1.43101, "mean_delta_full": -0.11578, "read_hit_rate": {"surface": 0.69551, "split": 0.87425, "merge": 0.2709, "skel": 0.67698}, "mean_gate": {"surface": 0.17121, "split": 0.08041, "merge": 0.15092, "skel": 0.12999}, "n_reads": 22043, "n_cand": 33226, "read_suppressed_frac": 0.1045, "dropout_reads_logged": 22043, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5051, "admitted_surface": 16109, "staged_surface": 0, "admitted_split": 15593, "staged_split": 0, "admitted_merge": 22525, "staged_merge": 0, "admitted_skel": 11309, "staged_skel": 0, "mean_count": 1591.70654296875, "n_util_obs": 685, "util_count_spearman": 0.06945434444213959}, "probe_overhead_sec_total": 632.451, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.009063, "alpha": 0.03, "t": 4416.2} |
| {"event": "consolidate", "step": 4650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4457.7} |
| {"event": "train", "step": 4650, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.39893, "bpb": 2.01822, "chunk_loss": 0.030869, "F": 0.22696, "G": 0.22716, "chunks": 232.41, "bytes_per_chunk": 4.406, "boundary_entropy": 0.07966, "router_noise": 0.0, "grad_norm": 0.66, "lr": 6.57e-05, "router_trainable": true, "sec_per_step": 0.839, "probe_sec": 0.0729, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05112, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.00325, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.61049, "n_merge": 1, "mean_delta_merge": -0.12076, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.40252, "trunc_abs_frac": 1.7204, "mean_delta_full": 0.20638, "read_hit_rate": {"surface": 0.68386, "split": 0.87083, "merge": 0.26407, "skel": 0.66994}, "mean_gate": {"surface": 0.17155, "split": 0.08121, "merge": 0.14979, "skel": 0.13178}, "n_reads": 21341, "n_cand": 32479, "read_suppressed_frac": 0.0994, "dropout_reads_logged": 21341, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5082, "admitted_surface": 16149, "staged_surface": 0, "admitted_split": 15599, "staged_split": 0, "admitted_merge": 22471, "staged_merge": 0, "admitted_skel": 11317, "staged_skel": 0, "mean_count": 1608.38818359375, "n_util_obs": 693, "util_count_spearman": 0.07539122599411707}, "probe_overhead_sec_total": 638.173, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.007809, "alpha": 0.03, "t": 4458.2} |
| {"event": "consolidate", "step": 4700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4493.8} |
| {"event": "train", "step": 4700, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.4156, "bpb": 2.04228, "chunk_loss": 0.031121, "F": 0.23468, "G": 0.23583, "chunks": 240.31, "bytes_per_chunk": 4.261, "boundary_entropy": 0.0789, "router_noise": 0.0, "grad_norm": 0.6773, "lr": 6.32e-05, "router_trainable": true, "sec_per_step": 0.721, "probe_sec": 0.0733, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0494, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.08572, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.20959, "n_merge": 1, "mean_delta_merge": -0.27584, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.76003, "trunc_abs_frac": 0.8109, "mean_delta_full": -0.38534, "read_hit_rate": {"surface": 0.68373, "split": 0.87259, "merge": 0.25256, "skel": 0.66956}, "mean_gate": {"surface": 0.1694, "split": 0.08043, "merge": 0.14967, "skel": 0.12997}, "n_reads": 21737, "n_cand": 33249, "read_suppressed_frac": 0.1025, "dropout_reads_logged": 21737, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5118, "admitted_surface": 16142, "staged_surface": 0, "admitted_split": 15561, "staged_split": 0, "admitted_merge": 22558, "staged_merge": 0, "admitted_skel": 11275, "staged_skel": 0, "mean_count": 1625.03564453125, "n_util_obs": 701, "util_count_spearman": 0.07212386533537005}, "probe_overhead_sec_total": 642.306, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.008495, "alpha": 0.03, "t": 4494.2} |
| {"event": "consolidate", "step": 4750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4537.4} |
| {"event": "train", "step": 4750, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.36734, "bpb": 1.97266, "chunk_loss": 0.030919, "F": 0.22882, "G": 0.22756, "chunks": 234.31, "bytes_per_chunk": 4.37, "boundary_entropy": 0.07742, "router_noise": 0.0, "grad_norm": 0.6664, "lr": 6.08e-05, "router_trainable": true, "sec_per_step": 0.891, "probe_sec": 0.1093, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06528, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.00146, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 0.38055, "trunc_abs_frac": 0.55776, "mean_delta_full": -0.15312, "read_hit_rate": {"surface": 0.69609, "split": 0.87677, "merge": 0.27724, "skel": 0.66907}, "mean_gate": {"surface": 0.1755, "split": 0.08295, "merge": 0.15379, "skel": 0.13349}, "n_reads": 21573, "n_cand": 32474, "read_suppressed_frac": 0.1011, "dropout_reads_logged": 21573, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5153, "admitted_surface": 16149, "staged_surface": 0, "admitted_split": 15608, "staged_split": 0, "admitted_merge": 22455, "staged_merge": 0, "admitted_skel": 11324, "staged_skel": 0, "mean_count": 1641.6448974609375, "n_util_obs": 704, "util_count_spearman": 0.06939900725093479}, "probe_overhead_sec_total": 647.109, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.007309, "alpha": 0.03, "t": 4538.8} |
| {"event": "consolidate", "step": 4800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4582.1} |
| {"event": "probe", "step": 4800, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3873.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.222548, "cps": 30.257812, "mean_p": 0.221766, "boundary_entropy": 0.074993, "bpb": 1.88731}, "tr": {"chunks": 4390.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.242501, "cps": 34.296875, "mean_p": 0.219763, "boundary_entropy": 0.107791, "bpb": 2.015746}, "fi": {"chunks": 4940.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.262612, "cps": 38.59375, "mean_p": 0.254301, "boundary_entropy": 0.10957, "bpb": 2.135961}}, "bpb": {"en": 1.88731, "tr": 2.01575, "fi": 2.13596}, "gini_cpc": 0.036706, "gini_cps": 0.053877, "bpic": 4.296902, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0896589542934305, "fi": 1.1800263051255382}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5190, "admitted_surface": 16142, "staged_surface": 0, "admitted_split": 15540, "staged_split": 0, "admitted_merge": 22556, "staged_merge": 0, "admitted_skel": 11298, "staged_skel": 0, "mean_count": 1657.9912109375, "n_util_obs": 708, "util_count_spearman": 0.06885223238685717}, "t": 4583.5} |
| {"event": "eval", "step": 4800, "phase": 3, "bpb": {"en": 1.90115, "tr": 2.02547, "fi": 2.14297}, "bytes_per_chunk": {"en": 4.382, "tr": 4.546, "fi": 4.221}, "bpb_mean": 2.0232, "frozen_inventory": true, "t": 4586.0} |
| {"event": "train", "step": 4800, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.3577, "bpb": 1.95875, "chunk_loss": 0.030993, "F": 0.23221, "G": 0.23113, "chunks": 237.78, "bytes_per_chunk": 4.306, "boundary_entropy": 0.08, "router_noise": 0.0, "grad_norm": 0.6441, "lr": 5.85e-05, "router_trainable": true, "sec_per_step": 0.977, "probe_sec": 0.1346, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.11217, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.08591, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 3, "mean_delta_merge": 0.17402, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -0.04648, "trunc_abs_frac": 4.48702, "mean_delta_full": 0.12671, "read_hit_rate": {"surface": 0.69605, "split": 0.87864, "merge": 0.26441, "skel": 0.68047}, "mean_gate": {"surface": 0.17205, "split": 0.0806, "merge": 0.15003, "skel": 0.13244}, "n_reads": 21934, "n_cand": 33033, "read_suppressed_frac": 0.0996, "dropout_reads_logged": 21934, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5191, "admitted_surface": 16118, "staged_surface": 0, "admitted_split": 15554, "staged_split": 0, "admitted_merge": 22550, "staged_merge": 0, "admitted_skel": 11314, "staged_skel": 0, "mean_count": 1658.309326171875, "n_util_obs": 708, "util_count_spearman": 0.06884971507226019}, "probe_overhead_sec_total": 652.06, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.008108, "alpha": 0.03, "t": 4587.6} |
| {"event": "consolidate", "step": 4850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4621.5} |
| {"event": "train", "step": 4850, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.40635, "bpb": 2.02893, "chunk_loss": 0.031058, "F": 0.23508, "G": 0.23507, "chunks": 240.72, "bytes_per_chunk": 4.254, "boundary_entropy": 0.07852, "router_noise": 0.0, "grad_norm": 0.6573, "lr": 5.62e-05, "router_trainable": true, "sec_per_step": 0.704, "probe_sec": 0.0968, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04784, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.04739, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.10236, "n_merge": 2, "mean_delta_merge": -0.08279, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.03275, "trunc_abs_frac": 0.53876, "mean_delta_full": 0.33887, "read_hit_rate": {"surface": 0.69209, "split": 0.88309, "merge": 0.26417, "skel": 0.68537}, "mean_gate": {"surface": 0.16163, "split": 0.07575, "merge": 0.14109, "skel": 0.12393}, "n_reads": 22427, "n_cand": 33658, "read_suppressed_frac": 0.0989, "dropout_reads_logged": 22427, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5221, "admitted_surface": 16146, "staged_surface": 0, "admitted_split": 15501, "staged_split": 0, "admitted_merge": 22589, "staged_merge": 0, "admitted_skel": 11300, "staged_skel": 0, "mean_count": 1674.9139404296875, "n_util_obs": 716, "util_count_spearman": 0.07063666270310172}, "probe_overhead_sec_total": 656.306, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.00713, "alpha": 0.03, "t": 4622.8} |
| {"event": "consolidate", "step": 4900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4662.1} |
| {"event": "train", "step": 4900, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.36116, "bpb": 1.96374, "chunk_loss": 0.030935, "F": 0.2309, "G": 0.23038, "chunks": 236.44, "bytes_per_chunk": 4.331, "boundary_entropy": 0.07974, "router_noise": 0.0, "grad_norm": 0.6474, "lr": 5.41e-05, "router_trainable": true, "sec_per_step": 0.809, "probe_sec": 0.1173, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07477, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.04826, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.01917, "n_merge": 2, "mean_delta_merge": -0.00906, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 1.03128, "trunc_abs_frac": 0.49377, "mean_delta_full": -0.12024, "read_hit_rate": {"surface": 0.70069, "split": 0.88019, "merge": 0.26046, "skel": 0.69073}, "mean_gate": {"surface": 0.17586, "split": 0.08258, "merge": 0.15584, "skel": 0.13397}, "n_reads": 22089, "n_cand": 33080, "read_suppressed_frac": 0.0998, "dropout_reads_logged": 22089, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5254, "admitted_surface": 16107, "staged_surface": 0, "admitted_split": 15558, "staged_split": 0, "admitted_merge": 22621, "staged_merge": 0, "admitted_skel": 11250, "staged_skel": 0, "mean_count": 1691.6611328125, "n_util_obs": 726, "util_count_spearman": 0.06399287180935198}, "probe_overhead_sec_total": 661.43, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.008343, "alpha": 0.03, "t": 4663.2} |
| {"event": "consolidate", "step": 4950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4712.3} |
| {"event": "train", "step": 4950, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.39614, "bpb": 2.01421, "chunk_loss": 0.030962, "F": 0.22971, "G": 0.23025, "chunks": 235.22, "bytes_per_chunk": 4.353, "boundary_entropy": 0.07619, "router_noise": 0.0, "grad_norm": 0.6781, "lr": 5.2e-05, "router_trainable": true, "sec_per_step": 0.994, "probe_sec": 0.0749, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04743, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.07207, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 3, "mean_delta_merge": -0.08838, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 5.19444, "trunc_abs_frac": 5.20539, "mean_delta_full": -0.01538, "read_hit_rate": {"surface": 0.67872, "split": 0.87434, "merge": 0.24642, "skel": 0.66015}, "mean_gate": {"surface": 0.16762, "split": 0.07878, "merge": 0.14714, "skel": 0.12719}, "n_reads": 21330, "n_cand": 32747, "read_suppressed_frac": 0.0995, "dropout_reads_logged": 21330, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5295, "admitted_surface": 16128, "staged_surface": 0, "admitted_split": 15567, "staged_split": 0, "admitted_merge": 22520, "staged_merge": 0, "admitted_skel": 11321, "staged_skel": 0, "mean_count": 1708.166259765625, "n_util_obs": 732, "util_count_spearman": 0.05335135983936462}, "probe_overhead_sec_total": 666.448, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.007863, "alpha": 0.03, "t": 4712.9} |
| {"event": "consolidate", "step": 5000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4746.7} |
| {"event": "probe", "step": 5000, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3880.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.22295, "cps": 30.3125, "mean_p": 0.222396, "boundary_entropy": 0.073032, "bpb": 1.879704}, "tr": {"chunks": 4391.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.242556, "cps": 34.304688, "mean_p": 0.22096, "boundary_entropy": 0.10475, "bpb": 2.0057}, "fi": {"chunks": 4953.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.263303, "cps": 38.695312, "mean_p": 0.255266, "boundary_entropy": 0.107556, "bpb": 2.120832}}, "bpb": {"en": 1.8797, "tr": 2.0057, "fi": 2.12083}, "gini_cpc": 0.036913, "gini_cps": 0.054094, "bpic": 4.290079, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0879408408129656, "fi": 1.1809971216665838}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5321, "admitted_surface": 16115, "staged_surface": 0, "admitted_split": 15451, "staged_split": 0, "admitted_merge": 22702, "staged_merge": 0, "admitted_skel": 11268, "staged_skel": 0, "mean_count": 1724.400390625, "n_util_obs": 734, "util_count_spearman": 0.04549905915762416}, "t": 4748.5} |
| {"event": "eval", "step": 5000, "phase": 3, "bpb": {"en": 1.89285, "tr": 2.01408, "fi": 2.12858}, "bytes_per_chunk": {"en": 4.379, "tr": 4.552, "fi": 4.22}, "bpb_mean": 2.01184, "frozen_inventory": true, "t": 4751.4} |
| {"event": "train", "step": 5000, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35182, "bpb": 1.95026, "chunk_loss": 0.031032, "F": 0.23361, "G": 0.23313, "chunks": 239.22, "bytes_per_chunk": 4.281, "boundary_entropy": 0.07848, "router_noise": 0.0, "grad_norm": 0.6455, "lr": 5e-05, "router_trainable": true, "sec_per_step": 0.804, "probe_sec": 0.0927, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05532, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": 0.05459, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.11066, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 1.99305, "trunc_abs_frac": 1.10478, "mean_delta_full": 0.14044, "read_hit_rate": {"surface": 0.69212, "split": 0.88443, "merge": 0.27216, "skel": 0.68864}, "mean_gate": {"surface": 0.16718, "split": 0.07803, "merge": 0.14452, "skel": 0.12793}, "n_reads": 22443, "n_cand": 33524, "read_suppressed_frac": 0.1001, "dropout_reads_logged": 22443, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5322, "admitted_surface": 16161, "staged_surface": 0, "admitted_split": 15499, "staged_split": 0, "admitted_merge": 22607, "staged_merge": 0, "admitted_skel": 11269, "staged_skel": 0, "mean_count": 1724.7421875, "n_util_obs": 734, "util_count_spearman": 0.04526737045298536}, "probe_overhead_sec_total": 670.572, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.007876, "alpha": 0.03, "t": 4753.1} |
| {"event": "consolidate", "step": 5050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4801.5} |
| {"event": "train", "step": 5050, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.3231, "bpb": 1.90883, "chunk_loss": 0.030972, "F": 0.23224, "G": 0.232, "chunks": 237.81, "bytes_per_chunk": 4.306, "boundary_entropy": 0.07485, "router_noise": 0.0, "grad_norm": 0.6315, "lr": 4.81e-05, "router_trainable": true, "sec_per_step": 0.985, "probe_sec": 0.0673, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04358, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.01232, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": 0.22067, "n_merge": 1, "mean_delta_merge": -0.15292, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 6.02642, "trunc_abs_frac": 4.38971, "mean_delta_full": 0.00947, "read_hit_rate": {"surface": 0.69715, "split": 0.88747, "merge": 0.27153, "skel": 0.69395}, "mean_gate": {"surface": 0.15809, "split": 0.07331, "merge": 0.13896, "skel": 0.12132}, "n_reads": 22535, "n_cand": 33483, "read_suppressed_frac": 0.099, "dropout_reads_logged": 22535, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5358, "admitted_surface": 16123, "staged_surface": 0, "admitted_split": 15484, "staged_split": 0, "admitted_merge": 22645, "staged_merge": 0, "admitted_skel": 11284, "staged_skel": 0, "mean_count": 1741.342529296875, "n_util_obs": 743, "util_count_spearman": 0.03910971314032346}, "probe_overhead_sec_total": 675.934, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.008259, "alpha": 0.03, "t": 4802.4} |
| {"event": "consolidate", "step": 5100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4844.4} |
| {"event": "train", "step": 5100, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.39103, "bpb": 2.00683, "chunk_loss": 0.030905, "F": 0.22891, "G": 0.2284, "chunks": 234.41, "bytes_per_chunk": 4.368, "boundary_entropy": 0.07823, "router_noise": 0.0, "grad_norm": 0.6437, "lr": 4.63e-05, "router_trainable": true, "sec_per_step": 0.854, "probe_sec": 0.0767, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04385, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.10845, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 3, "mean_delta_merge": -0.14926, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.01315, "trunc_abs_frac": 0.09293, "mean_delta_full": 0.30176, "read_hit_rate": {"surface": 0.69393, "split": 0.87999, "merge": 0.25817, "skel": 0.68502}, "mean_gate": {"surface": 0.17094, "split": 0.07958, "merge": 0.14996, "skel": 0.13051}, "n_reads": 21856, "n_cand": 32878, "read_suppressed_frac": 0.1017, "dropout_reads_logged": 21856, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5395, "admitted_surface": 16128, "staged_surface": 0, "admitted_split": 15532, "staged_split": 0, "admitted_merge": 22591, "staged_merge": 0, "admitted_skel": 11285, "staged_skel": 0, "mean_count": 1757.8970947265625, "n_util_obs": 752, "util_count_spearman": 0.041917959559288696}, "probe_overhead_sec_total": 680.842, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.007604, "alpha": 0.03, "t": 4845.1} |
| {"event": "consolidate", "step": 5150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4885.1} |
| {"event": "train", "step": 5150, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35127, "bpb": 1.94947, "chunk_loss": 0.031217, "F": 0.23285, "G": 0.23195, "chunks": 238.44, "bytes_per_chunk": 4.295, "boundary_entropy": 0.07393, "router_noise": 0.0, "grad_norm": 0.6869, "lr": 4.45e-05, "router_trainable": true, "sec_per_step": 0.822, "probe_sec": 0.0808, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04807, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.07106, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": 0.05635, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.80599, "trunc_abs_frac": 1.00309, "mean_delta_full": 0.37335, "read_hit_rate": {"surface": 0.70163, "split": 0.87467, "merge": 0.27835, "skel": 0.67671}, "mean_gate": {"surface": 0.16828, "split": 0.07861, "merge": 0.14799, "skel": 0.12829}, "n_reads": 21733, "n_cand": 32695, "read_suppressed_frac": 0.1, "dropout_reads_logged": 21733, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5438, "admitted_surface": 16109, "staged_surface": 0, "admitted_split": 15597, "staged_split": 0, "admitted_merge": 22542, "staged_merge": 0, "admitted_skel": 11288, "staged_skel": 0, "mean_count": 1774.405517578125, "n_util_obs": 756, "util_count_spearman": 0.03967630555912376}, "probe_overhead_sec_total": 685.313, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.008914, "alpha": 0.03, "t": 4886.2} |
| {"event": "consolidate", "step": 5200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4943.1} |
| {"event": "probe", "step": 5200, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3882.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.223065, "cps": 30.328125, "mean_p": 0.222867, "boundary_entropy": 0.070653, "bpb": 1.871908}, "tr": {"chunks": 4396.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.242833, "cps": 34.34375, "mean_p": 0.221826, "boundary_entropy": 0.099764, "bpb": 1.995525}, "fi": {"chunks": 4988.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.265164, "cps": 38.96875, "mean_p": 0.256405, "boundary_entropy": 0.104015, "bpb": 2.108457}}, "bpb": {"en": 1.87191, "tr": 1.99553, "fi": 2.10846}, "gini_cpc": 0.038391, "gini_cps": 0.055581, "bpic": 4.276496, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0886185276232747, "fi": 1.1887298012105614}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5465, "admitted_surface": 16126, "staged_surface": 0, "admitted_split": 15530, "staged_split": 0, "admitted_merge": 22639, "staged_merge": 0, "admitted_skel": 11241, "staged_skel": 0, "mean_count": 1790.609130859375, "n_util_obs": 766, "util_count_spearman": 0.028885467095108877}, "t": 4944.7} |
| {"event": "eval", "step": 5200, "phase": 3, "bpb": {"en": 1.88625, "tr": 2.00616, "fi": 2.1162}, "bytes_per_chunk": {"en": 4.37, "tr": 4.549, "fi": 4.162}, "bpb_mean": 2.00287, "frozen_inventory": true, "t": 4947.3} |
| {"event": "train", "step": 5200, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.33713, "bpb": 1.92907, "chunk_loss": 0.030936, "F": 0.22858, "G": 0.22822, "chunks": 234.06, "bytes_per_chunk": 4.375, "boundary_entropy": 0.07011, "router_noise": 0.0, "grad_norm": 0.6446, "lr": 4.29e-05, "router_trainable": true, "sec_per_step": 1.274, "probe_sec": 0.1299, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06934, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.12481, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 3, "mean_delta_merge": -0.18913, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 1.0, "trunc_abs_frac": 1.0, "mean_delta_full": -0.3623, "read_hit_rate": {"surface": 0.68597, "split": 0.87545, "merge": 0.2634, "skel": 0.67302}, "mean_gate": {"surface": 0.16511, "split": 0.07807, "merge": 0.14466, "skel": 0.12696}, "n_reads": 21458, "n_cand": 32543, "read_suppressed_frac": 0.0994, "dropout_reads_logged": 21458, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5465, "admitted_surface": 16124, "staged_surface": 0, "admitted_split": 15576, "staged_split": 0, "admitted_merge": 22522, "staged_merge": 0, "admitted_skel": 11314, "staged_skel": 0, "mean_count": 1790.940673828125, "n_util_obs": 766, "util_count_spearman": 0.028911925653206088}, "probe_overhead_sec_total": 690.82, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.007274, "alpha": 0.03, "t": 4949.9} |
| {"event": "consolidate", "step": 5250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4996.6} |
| {"event": "train", "step": 5250, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35034, "bpb": 1.94812, "chunk_loss": 0.03097, "F": 0.23093, "G": 0.23036, "chunks": 236.47, "bytes_per_chunk": 4.33, "boundary_entropy": 0.07738, "router_noise": 0.0, "grad_norm": 0.6613, "lr": 4.14e-05, "router_trainable": true, "sec_per_step": 0.956, "probe_sec": 0.0951, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06331, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.12317, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": 0.01193, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.15142, "trunc_abs_frac": 0.17931, "mean_delta_full": -0.40735, "read_hit_rate": {"surface": 0.69874, "split": 0.88052, "merge": 0.26843, "skel": 0.68785}, "mean_gate": {"surface": 0.16709, "split": 0.0776, "merge": 0.14503, "skel": 0.12808}, "n_reads": 22088, "n_cand": 33051, "read_suppressed_frac": 0.1013, "dropout_reads_logged": 22088, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5491, "admitted_surface": 16096, "staged_surface": 0, "admitted_split": 15562, "staged_split": 0, "admitted_merge": 22610, "staged_merge": 0, "admitted_skel": 11268, "staged_skel": 0, "mean_count": 1807.589599609375, "n_util_obs": 770, "util_count_spearman": 0.035348506799340616}, "probe_overhead_sec_total": 695.763, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009234, "alpha": 0.03, "t": 4997.7} |
| {"event": "consolidate", "step": 5300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 5036.5} |
| {"event": "train", "step": 5300, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.3313, "bpb": 1.92066, "chunk_loss": 0.030922, "F": 0.22961, "G": 0.22972, "chunks": 235.12, "bytes_per_chunk": 4.355, "boundary_entropy": 0.07681, "router_noise": 0.0, "grad_norm": 0.6677, "lr": 3.99e-05, "router_trainable": true, "sec_per_step": 0.787, "probe_sec": 0.0816, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05145, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.00345, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 1.25341, "trunc_abs_frac": 0.73212, "mean_delta_full": 0.07153, "read_hit_rate": {"surface": 0.70141, "split": 0.88073, "merge": 0.26381, "skel": 0.68688}, "mean_gate": {"surface": 0.16359, "split": 0.07586, "merge": 0.14311, "skel": 0.12563}, "n_reads": 21945, "n_cand": 32862, "read_suppressed_frac": 0.1008, "dropout_reads_logged": 21945, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5531, "admitted_surface": 16101, "staged_surface": 0, "admitted_split": 15538, "staged_split": 0, "admitted_merge": 22638, "staged_merge": 0, "admitted_skel": 11259, "staged_skel": 0, "mean_count": 1824.1295166015625, "n_util_obs": 777, "util_count_spearman": 0.035331683165910704}, "probe_overhead_sec_total": 700.569, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.007557, "alpha": 0.03, "t": 5037.1} |
| {"event": "consolidate", "step": 5350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 5077.5} |
| {"event": "train", "step": 5350, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.3386, "bpb": 1.93119, "chunk_loss": 0.031059, "F": 0.23465, "G": 0.23415, "chunks": 240.28, "bytes_per_chunk": 4.262, "boundary_entropy": 0.07382, "router_noise": 0.0, "grad_norm": 0.6784, "lr": 3.86e-05, "router_trainable": true, "sec_per_step": 0.843, "probe_sec": 0.1707, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06991, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": 0.00324, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.12817, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.10061, "trunc_abs_frac": 1.30359, "mean_delta_full": -0.12604, "read_hit_rate": {"surface": 0.70641, "split": 0.88798, "merge": 0.27554, "skel": 0.69157}, "mean_gate": {"surface": 0.16502, "split": 0.07667, "merge": 0.14417, "skel": 0.12667}, "n_reads": 22492, "n_cand": 33353, "read_suppressed_frac": 0.0959, "dropout_reads_logged": 22492, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5557, "admitted_surface": 16129, "staged_surface": 0, "admitted_split": 15474, "staged_split": 0, "admitted_merge": 22664, "staged_merge": 0, "admitted_skel": 11269, "staged_skel": 0, "mean_count": 1840.744873046875, "n_util_obs": 781, "util_count_spearman": 0.03504597122553017}, "probe_overhead_sec_total": 705.212, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008169, "alpha": 0.03, "t": 5079.2} |
| {"event": "consolidate", "step": 5400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 5127.1} |
| {"event": "probe", "step": 5400, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3875.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.222663, "cps": 30.273438, "mean_p": 0.221476, "boundary_entropy": 0.073294, "bpb": 1.86792}, "tr": {"chunks": 4397.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.242888, "cps": 34.351562, "mean_p": 0.22169, "boundary_entropy": 0.10169, "bpb": 1.987958}, "fi": {"chunks": 4958.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.263569, "cps": 38.734375, "mean_p": 0.254988, "boundary_entropy": 0.107308, "bpb": 2.104366}}, "bpb": {"en": 1.86792, "tr": 1.98796, "fi": 2.10437}, "gini_cpc": 0.037403, "gini_cps": 0.054573, "bpic": 4.288133, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0908331500927488, "fi": 1.1837147310856206}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5591, "admitted_surface": 16113, "staged_surface": 0, "admitted_split": 15552, "staged_split": 0, "admitted_merge": 22657, "staged_merge": 0, "admitted_skel": 11214, "staged_skel": 0, "mean_count": 1857.0020751953125, "n_util_obs": 788, "util_count_spearman": 0.039215841837903845}, "t": 5129.0} |
| {"event": "eval", "step": 5400, "phase": 3, "bpb": {"en": 1.88205, "tr": 1.9965, "fi": 2.11121}, "bytes_per_chunk": {"en": 4.391, "tr": 4.551, "fi": 4.202}, "bpb_mean": 1.99659, "frozen_inventory": true, "t": 5132.8} |
| {"event": "train", "step": 5400, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32658, "bpb": 1.91386, "chunk_loss": 0.030971, "F": 0.23102, "G": 0.23106, "chunks": 236.56, "bytes_per_chunk": 4.329, "boundary_entropy": 0.07652, "router_noise": 0.0, "grad_norm": 0.6429, "lr": 3.73e-05, "router_trainable": true, "sec_per_step": 1.145, "probe_sec": 0.1725, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.11265, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.02053, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -0.67134, "trunc_abs_frac": 0.58379, "mean_delta_full": 0.11748, "read_hit_rate": {"surface": 0.69594, "split": 0.88031, "merge": 0.27005, "skel": 0.68274}, "mean_gate": {"surface": 0.16234, "split": 0.07528, "merge": 0.14142, "skel": 0.12237}, "n_reads": 21965, "n_cand": 32939, "read_suppressed_frac": 0.0978, "dropout_reads_logged": 21965, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5591, "admitted_surface": 16105, "staged_surface": 0, "admitted_split": 15554, "staged_split": 0, "admitted_merge": 22611, "staged_merge": 0, "admitted_skel": 11266, "staged_skel": 0, "mean_count": 1857.33837890625, "n_util_obs": 789, "util_count_spearman": 0.03678580813556898}, "probe_overhead_sec_total": 710.62, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008358, "alpha": 0.03, "t": 5136.4} |
| {"event": "consolidate", "step": 5450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 5175.9} |
| {"event": "train", "step": 5450, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.34533, "bpb": 1.94091, "chunk_loss": 0.030914, "F": 0.22818, "G": 0.22771, "chunks": 233.66, "bytes_per_chunk": 4.383, "boundary_entropy": 0.07471, "router_noise": 0.0, "grad_norm": 0.676, "lr": 3.62e-05, "router_trainable": true, "sec_per_step": 0.801, "probe_sec": 0.1048, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07556, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": 0.12141, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.03204, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.86312, "trunc_abs_frac": 0.42944, "mean_delta_full": 0.25456, "read_hit_rate": {"surface": 0.69174, "split": 0.87655, "merge": 0.27007, "skel": 0.68044}, "mean_gate": {"surface": 0.16483, "split": 0.0764, "merge": 0.14423, "skel": 0.12589}, "n_reads": 21690, "n_cand": 32640, "read_suppressed_frac": 0.0986, "dropout_reads_logged": 21690, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5614, "admitted_surface": 16108, "staged_surface": 0, "admitted_split": 15594, "staged_split": 0, "admitted_merge": 22552, "staged_merge": 0, "admitted_skel": 11282, "staged_skel": 0, "mean_count": 1873.8834228515625, "n_util_obs": 798, "util_count_spearman": 0.03420413200945587}, "probe_overhead_sec_total": 715.284, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.007535, "alpha": 0.03, "t": 5176.5} |
| {"event": "consolidate", "step": 5500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 5213.6} |
| {"event": "train", "step": 5500, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.34898, "bpb": 1.94617, "chunk_loss": 0.031047, "F": 0.23126, "G": 0.23101, "chunks": 236.81, "bytes_per_chunk": 4.324, "boundary_entropy": 0.07841, "router_noise": 0.0, "grad_norm": 0.72, "lr": 3.51e-05, "router_trainable": true, "sec_per_step": 0.772, "probe_sec": 0.0772, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04391, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.0627, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.48022, "n_merge": 1, "mean_delta_merge": 0.1452, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -0.93669, "trunc_abs_frac": 0.7859, "mean_delta_full": 0.03543, "read_hit_rate": {"surface": 0.69109, "split": 0.87484, "merge": 0.26471, "skel": 0.67171}, "mean_gate": {"surface": 0.16418, "split": 0.07707, "merge": 0.14319, "skel": 0.12436}, "n_reads": 21610, "n_cand": 32740, "read_suppressed_frac": 0.0972, "dropout_reads_logged": 21610, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5650, "admitted_surface": 16073, "staged_surface": 0, "admitted_split": 15593, "staged_split": 0, "admitted_merge": 22596, "staged_merge": 0, "admitted_skel": 11274, "staged_skel": 0, "mean_count": 1890.394287109375, "n_util_obs": 803, "util_count_spearman": 0.04371575644416258}, "probe_overhead_sec_total": 719.739, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008124, "alpha": 0.03, "t": 5215.1} |
| {"event": "consolidate", "step": 5550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 5285.4} |
| {"event": "train", "step": 5550, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.29276, "bpb": 1.86506, "chunk_loss": 0.030975, "F": 0.23157, "G": 0.23024, "chunks": 237.12, "bytes_per_chunk": 4.318, "boundary_entropy": 0.07574, "router_noise": 0.0, "grad_norm": 0.6392, "lr": 3.41e-05, "router_trainable": true, "sec_per_step": 1.44, "probe_sec": 0.1155, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06471, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.04334, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.23596, "n_merge": 1, "mean_delta_merge": 0.82513, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 2.02242, "trunc_abs_frac": 0.73905, "mean_delta_full": 0.09619, "read_hit_rate": {"surface": 0.70659, "split": 0.88375, "merge": 0.2823, "skel": 0.68437}, "mean_gate": {"surface": 0.16396, "split": 0.07599, "merge": 0.14206, "skel": 0.12571}, "n_reads": 22283, "n_cand": 33081, "read_suppressed_frac": 0.1006, "dropout_reads_logged": 22283, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5688, "admitted_surface": 16103, "staged_surface": 0, "admitted_split": 15565, "staged_split": 0, "admitted_merge": 22563, "staged_merge": 0, "admitted_skel": 11305, "staged_skel": 0, "mean_count": 1907.0919189453125, "n_util_obs": 813, "util_count_spearman": 0.03972100116187083}, "probe_overhead_sec_total": 726.354, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008338, "alpha": 0.03, "t": 5287.1} |
| {"event": "consolidate", "step": 5600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 5334.2} |
| {"event": "probe", "step": 5600, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3878.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.222835, "cps": 30.296875, "mean_p": 0.221765, "boundary_entropy": 0.07225, "bpb": 1.861404}, "tr": {"chunks": 4400.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.243054, "cps": 34.375, "mean_p": 0.220712, "boundary_entropy": 0.103417, "bpb": 1.982165}, "fi": {"chunks": 4956.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.263463, "cps": 38.71875, "mean_p": 0.253966, "boundary_entropy": 0.106894, "bpb": 2.097397}}, "bpb": {"en": 1.8614, "tr": 1.98217, "fi": 2.0974}, "gini_cpc": 0.037136, "gini_cps": 0.054305, "bpic": 4.286837, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0907329689883831, "fi": 1.1823218882415178}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5726, "admitted_surface": 16114, "staged_surface": 0, "admitted_split": 15532, "staged_split": 0, "admitted_merge": 22585, "staged_merge": 0, "admitted_skel": 11305, "staged_skel": 0, "mean_count": 1923.29248046875, "n_util_obs": 821, "util_count_spearman": 0.04725141741176392}, "t": 5336.0} |
| {"event": "eval", "step": 5600, "phase": 3, "bpb": {"en": 1.87614, "tr": 1.99098, "fi": 2.10523}, "bytes_per_chunk": {"en": 4.379, "tr": 4.551, "fi": 4.219}, "bpb_mean": 1.99078, "frozen_inventory": true, "t": 5338.5} |
| {"event": "train", "step": 5600, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.36848, "bpb": 1.9743, "chunk_loss": 0.031222, "F": 0.23688, "G": 0.2361, "chunks": 242.56, "bytes_per_chunk": 4.222, "boundary_entropy": 0.07751, "router_noise": 0.0, "grad_norm": 0.7043, "lr": 3.33e-05, "router_trainable": true, "sec_per_step": 1.055, "probe_sec": 0.0674, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04512, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.03508, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 3, "mean_delta_split": 0.29602, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.58567, "trunc_abs_frac": 0.5985, "mean_delta_full": 0.15674, "read_hit_rate": {"surface": 0.7066, "split": 0.88176, "merge": 0.28527, "skel": 0.68965}, "mean_gate": {"surface": 0.16176, "split": 0.07459, "merge": 0.13984, "skel": 0.12345}, "n_reads": 22488, "n_cand": 33419, "read_suppressed_frac": 0.0985, "dropout_reads_logged": 22488, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5726, "admitted_surface": 16107, "staged_surface": 0, "admitted_split": 15567, "staged_split": 0, "admitted_merge": 22593, "staged_merge": 0, "admitted_skel": 11269, "staged_skel": 0, "mean_count": 1923.641845703125, "n_util_obs": 821, "util_count_spearman": 0.047215713634586196}, "probe_overhead_sec_total": 731.602, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.00718, "alpha": 0.03, "t": 5339.9} |
| {"event": "consolidate", "step": 5650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 5377.5} |
| {"event": "train", "step": 5650, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.34188, "bpb": 1.93593, "chunk_loss": 0.030888, "F": 0.22736, "G": 0.22666, "chunks": 232.81, "bytes_per_chunk": 4.398, "boundary_entropy": 0.07514, "router_noise": 0.0, "grad_norm": 0.6667, "lr": 3.25e-05, "router_trainable": true, "sec_per_step": 0.766, "probe_sec": 0.1101, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06818, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.03328, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": 0.03436, "n_merge": 1, "mean_delta_merge": -1.25061, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 0.17946, "trunc_abs_frac": 0.25473, "mean_delta_full": 0.09898, "read_hit_rate": {"surface": 0.69507, "split": 0.87576, "merge": 0.27958, "skel": 0.6775}, "mean_gate": {"surface": 0.16441, "split": 0.07657, "merge": 0.14221, "skel": 0.12482}, "n_reads": 21430, "n_cand": 32213, "read_suppressed_frac": 0.0977, "dropout_reads_logged": 21430, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5759, "admitted_surface": 16116, "staged_surface": 0, "admitted_split": 15623, "staged_split": 0, "admitted_merge": 22508, "staged_merge": 0, "admitted_skel": 11289, "staged_skel": 0, "mean_count": 1940.2022705078125, "n_util_obs": 832, "util_count_spearman": 0.05689214760236996}, "probe_overhead_sec_total": 736.265, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.007769, "alpha": 0.03, "t": 5378.2} |
| {"event": "consolidate", "step": 5700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 5417.0} |
| {"event": "train", "step": 5700, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.34921, "bpb": 1.94649, "chunk_loss": 0.030958, "F": 0.23083, "G": 0.22965, "chunks": 236.38, "bytes_per_chunk": 4.332, "boundary_entropy": 0.07416, "router_noise": 0.0, "grad_norm": 0.7124, "lr": 3.18e-05, "router_trainable": true, "sec_per_step": 0.789, "probe_sec": 0.1439, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10356, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.11255, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.05634, "n_merge": 3, "mean_delta_merge": -0.19807, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -3.66384, "trunc_abs_frac": 0.43361, "mean_delta_full": -0.0072, "read_hit_rate": {"surface": 0.69968, "split": 0.88393, "merge": 0.27507, "skel": 0.67759}, "mean_gate": {"surface": 0.16084, "split": 0.07461, "merge": 0.14201, "skel": 0.12199}, "n_reads": 21962, "n_cand": 32836, "read_suppressed_frac": 0.0999, "dropout_reads_logged": 21962, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5792, "admitted_surface": 16104, "staged_surface": 0, "admitted_split": 15552, "staged_split": 0, "admitted_merge": 22570, "staged_merge": 0, "admitted_skel": 11310, "staged_skel": 0, "mean_count": 1956.8419189453125, "n_util_obs": 838, "util_count_spearman": 0.06167182763549376}, "probe_overhead_sec_total": 741.052, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006762, "alpha": 0.03, "t": 5417.6} |
| {"event": "consolidate", "step": 5750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 5464.1} |
| {"event": "train", "step": 5750, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35523, "bpb": 1.95519, "chunk_loss": 0.031074, "F": 0.23419, "G": 0.23342, "chunks": 239.81, "bytes_per_chunk": 4.27, "boundary_entropy": 0.07411, "router_noise": 0.0, "grad_norm": 0.6774, "lr": 3.13e-05, "router_trainable": true, "sec_per_step": 0.942, "probe_sec": 0.1308, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07328, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.04673, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": 0.08191, "n_merge": 2, "mean_delta_merge": -0.55989, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.31477, "trunc_abs_frac": 0.55242, "mean_delta_full": -0.26717, "read_hit_rate": {"surface": 0.7023, "split": 0.87766, "merge": 0.27608, "skel": 0.67595}, "mean_gate": {"surface": 0.1646, "split": 0.07593, "merge": 0.14363, "skel": 0.12534}, "n_reads": 21939, "n_cand": 32953, "read_suppressed_frac": 0.1028, "dropout_reads_logged": 21939, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5832, "admitted_surface": 16098, "staged_surface": 0, "admitted_split": 15597, "staged_split": 0, "admitted_merge": 22561, "staged_merge": 0, "admitted_skel": 11280, "staged_skel": 0, "mean_count": 1973.3792724609375, "n_util_obs": 842, "util_count_spearman": 0.06583917018901138}, "probe_overhead_sec_total": 746.015, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008351, "alpha": 0.03, "t": 5464.7} |
| {"event": "consolidate", "step": 5800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 5504.2} |
| {"event": "probe", "step": 5800, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3883.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.223122, "cps": 30.335938, "mean_p": 0.222199, "boundary_entropy": 0.069077, "bpb": 1.857636}, "tr": {"chunks": 4397.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.242888, "cps": 34.351562, "mean_p": 0.221478, "boundary_entropy": 0.097477, "bpb": 1.976013}, "fi": {"chunks": 4966.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.263994, "cps": 38.796875, "mean_p": 0.255349, "boundary_entropy": 0.103134, "bpb": 2.09355}}, "bpb": {"en": 1.85764, "tr": 1.97601, "fi": 2.09355}, "gini_cpc": 0.037326, "gini_cps": 0.054507, "bpic": 4.282953, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0885857472596967, "fi": 1.1831820201858045}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5872, "admitted_surface": 16092, "staged_surface": 0, "admitted_split": 15614, "staged_split": 0, "admitted_merge": 22529, "staged_merge": 0, "admitted_skel": 11301, "staged_skel": 0, "mean_count": 1989.555419921875, "n_util_obs": 842, "util_count_spearman": 0.06847304157700353}, "t": 5505.7} |
| {"event": "eval", "step": 5800, "phase": 3, "bpb": {"en": 1.87207, "tr": 1.98416, "fi": 2.10082}, "bytes_per_chunk": {"en": 4.377, "tr": 4.547, "fi": 4.192}, "bpb_mean": 1.98568, "frozen_inventory": true, "t": 5508.2} |
| {"event": "train", "step": 5800, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35106, "bpb": 1.94916, "chunk_loss": 0.030909, "F": 0.22968, "G": 0.22859, "chunks": 235.19, "bytes_per_chunk": 4.354, "boundary_entropy": 0.07482, "router_noise": 0.0, "grad_norm": 0.6524, "lr": 3.08e-05, "router_trainable": true, "sec_per_step": 0.9, "probe_sec": 0.0778, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04555, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.01077, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": 0.06247, "n_merge": 2, "mean_delta_merge": 0.25372, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.39802, "trunc_abs_frac": 0.39382, "mean_delta_full": 0.49554, "read_hit_rate": {"surface": 0.69335, "split": 0.88033, "merge": 0.26923, "skel": 0.67661}, "mean_gate": {"surface": 0.16719, "split": 0.0781, "merge": 0.14644, "skel": 0.12687}, "n_reads": 21816, "n_cand": 32779, "read_suppressed_frac": 0.1006, "dropout_reads_logged": 21816, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5872, "admitted_surface": 16103, "staged_surface": 0, "admitted_split": 15588, "staged_split": 0, "admitted_merge": 22589, "staged_merge": 0, "admitted_skel": 11256, "staged_skel": 0, "mean_count": 1989.88232421875, "n_util_obs": 842, "util_count_spearman": 0.06878088946584929}, "probe_overhead_sec_total": 750.465, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.007034, "alpha": 0.03, "t": 5509.8} |
| {"event": "consolidate", "step": 5850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 5545.2} |
| {"event": "train", "step": 5850, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.27487, "bpb": 1.83925, "chunk_loss": 0.030823, "F": 0.22696, "G": 0.22609, "chunks": 232.41, "bytes_per_chunk": 4.406, "boundary_entropy": 0.0716, "router_noise": 0.0, "grad_norm": 0.6391, "lr": 3.05e-05, "router_trainable": true, "sec_per_step": 0.719, "probe_sec": 0.0797, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.056, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.04441, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.23572, "n_merge": 2, "mean_delta_merge": 0.30341, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.3209, "trunc_abs_frac": 0.77465, "mean_delta_full": -0.21371, "read_hit_rate": {"surface": 0.70412, "split": 0.88225, "merge": 0.28333, "skel": 0.68765}, "mean_gate": {"surface": 0.16315, "split": 0.07521, "merge": 0.14262, "skel": 0.12532}, "n_reads": 21849, "n_cand": 32439, "read_suppressed_frac": 0.1013, "dropout_reads_logged": 21849, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5908, "admitted_surface": 16118, "staged_surface": 0, "admitted_split": 15610, "staged_split": 0, "admitted_merge": 22524, "staged_merge": 0, "admitted_skel": 11284, "staged_skel": 0, "mean_count": 2006.43701171875, "n_util_obs": 855, "util_count_spearman": 0.05239451813599762}, "probe_overhead_sec_total": 755.086, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.007528, "alpha": 0.03, "t": 5545.7} |
| {"event": "consolidate", "step": 5900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 5584.7} |
| {"event": "train", "step": 5900, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.33808, "bpb": 1.93044, "chunk_loss": 0.03091, "F": 0.22842, "G": 0.22794, "chunks": 233.91, "bytes_per_chunk": 4.378, "boundary_entropy": 0.06693, "router_noise": 0.0, "grad_norm": 0.6334, "lr": 3.02e-05, "router_trainable": true, "sec_per_step": 0.809, "probe_sec": 0.0861, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0538, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": -0.01304, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.30212, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.24918, "trunc_abs_frac": 0.24918, "mean_delta_full": 0.22241, "read_hit_rate": {"surface": 0.6973, "split": 0.87884, "merge": 0.28958, "skel": 0.67244}, "mean_gate": {"surface": 0.15831, "split": 0.07369, "merge": 0.13638, "skel": 0.12191}, "n_reads": 21699, "n_cand": 32439, "read_suppressed_frac": 0.0999, "dropout_reads_logged": 21699, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5946, "admitted_surface": 16136, "staged_surface": 0, "admitted_split": 15630, "staged_split": 0, "admitted_merge": 22438, "staged_merge": 0, "admitted_skel": 11332, "staged_skel": 0, "mean_count": 2022.9739990234375, "n_util_obs": 860, "util_count_spearman": 0.05646519944605934}, "probe_overhead_sec_total": 759.675, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.007379, "alpha": 0.03, "t": 5586.1} |
| {"event": "consolidate", "step": 5950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 5630.6} |
| {"event": "train", "step": 5950, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35381, "bpb": 1.95313, "chunk_loss": 0.030871, "F": 0.22815, "G": 0.22797, "chunks": 233.62, "bytes_per_chunk": 4.383, "boundary_entropy": 0.07029, "router_noise": 0.0, "grad_norm": 0.7178, "lr": 3.01e-05, "router_trainable": true, "sec_per_step": 0.905, "probe_sec": 0.1068, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08001, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": 0.0169, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.21973, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.2109, "trunc_abs_frac": 0.57654, "mean_delta_full": 0.14856, "read_hit_rate": {"surface": 0.69586, "split": 0.87547, "merge": 0.26592, "skel": 0.67545}, "mean_gate": {"surface": 0.15803, "split": 0.07316, "merge": 0.1391, "skel": 0.1205}, "n_reads": 21507, "n_cand": 32460, "read_suppressed_frac": 0.1001, "dropout_reads_logged": 21507, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5969, "admitted_surface": 16075, "staged_surface": 0, "admitted_split": 15601, "staged_split": 0, "admitted_merge": 22590, "staged_merge": 0, "admitted_skel": 11270, "staged_skel": 0, "mean_count": 2039.441650390625, "n_util_obs": 869, "util_count_spearman": 0.048165907615060205}, "probe_overhead_sec_total": 764.696, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.007768, "alpha": 0.03, "t": 5631.4} |
| {"event": "consolidate", "step": 6000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 5665.4} |
| {"event": "train", "step": 5999, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32049, "bpb": 1.90506, "chunk_loss": 0.030727, "F": 0.22278, "G": 0.22183, "chunks": 228.12, "bytes_per_chunk": 4.489, "boundary_entropy": 0.07048, "router_noise": 0.0, "grad_norm": 0.703, "lr": 3e-05, "router_trainable": true, "sec_per_step": 0.68, "probe_sec": 0.1431, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.11911, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.05153, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.21204, "n_merge": 3, "mean_delta_merge": -0.02728, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.45352, "trunc_abs_frac": 0.89796, "mean_delta_full": -0.14313, "read_hit_rate": {"surface": 0.69263, "split": 0.87537, "merge": 0.26838, "skel": 0.67698}, "mean_gate": {"surface": 0.15955, "split": 0.07358, "merge": 0.1411, "skel": 0.12249}, "n_reads": 21163, "n_cand": 31898, "read_suppressed_frac": 0.1003, "dropout_reads_logged": 21163, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 6000, "admitted_surface": 16127, "staged_surface": 0, "admitted_split": 15618, "staged_split": 0, "admitted_merge": 22479, "staged_merge": 0, "admitted_skel": 11312, "staged_skel": 0, "mean_count": 2055.7568359375, "n_util_obs": 878, "util_count_spearman": 0.04335744847761789}, "probe_overhead_sec_total": 769.202, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008074, "alpha": 0.03, "t": 5665.4} |
| {"event": "probe", "step": 6000, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3891.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.223582, "cps": 30.398438, "mean_p": 0.22245, "boundary_entropy": 0.070225, "bpb": 1.854305}, "tr": {"chunks": 4405.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.24333, "cps": 34.414062, "mean_p": 0.220503, "boundary_entropy": 0.097072, "bpb": 1.973619}, "fi": {"chunks": 4963.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.263835, "cps": 38.773438, "mean_p": 0.255145, "boundary_entropy": 0.10489, "bpb": 2.088535}}, "bpb": {"en": 1.8543, "tr": 1.97362, "fi": 2.08854}, "gini_cpc": 0.036723, "gini_cps": 0.0539, "bpic": 4.278754, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0883241109268045, "fi": 1.1800360662675964}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 6000, "admitted_surface": 16127, "staged_surface": 0, "admitted_split": 15618, "staged_split": 0, "admitted_merge": 22479, "staged_merge": 0, "admitted_skel": 11312, "staged_skel": 0, "mean_count": 2055.7568359375, "n_util_obs": 878, "util_count_spearman": 0.04335744847761789}, "t": 5666.8} |
| {"event": "eval", "step": 6000, "phase": 3, "bpb": {"en": 1.86849, "tr": 1.98298, "fi": 2.09716}, "bytes_per_chunk": {"en": 4.375, "tr": 4.545, "fi": 4.209}, "bpb_mean": 1.98288, "frozen_inventory": true, "t": 5669.3} |
| {"event": "reps_dumped", "dirs": ["/local/scratch/sas245/beetle-hnet/runs/ali_mem-mixed_s3/probe/reps/th0.30", "/local/scratch/sas245/beetle-hnet/runs/ali_mem-mixed_s3/probe/reps/th0.40", "/local/scratch/sas245/beetle-hnet/runs/ali_mem-mixed_s3/probe/reps/th0.50", "/local/scratch/sas245/beetle-hnet/runs/ali_mem-mixed_s3/probe/reps/th0.60", "/local/scratch/sas245/beetle-hnet/runs/ali_mem-mixed_s3/probe/reps/th0.70"], "t": 5671.6} |
| {"event": "final", "run": "ali_mem-mixed_s3", "arch": "ali", "objective": "ratio", "condition": "A", "step": 6000, "params": 22785088, "seed": 3, "per_lang": {"en": {"cpc": 0.22358214101017065, "cps": 30.3984375, "bpb": 1.8732869719582723, "bytes_per_chunk": 4.330739621020634, "reads_per_chunk": 2.9152500371710364}, "tr": {"cpc": 0.24332983483400542, "cps": 34.4140625, "bpb": 1.973897069938076, "bytes_per_chunk": 4.535324088891965, "reads_per_chunk": 2.8439200352938125}, "fi": {"cpc": 0.2638349901653288, "cps": 38.7734375, "bpb": 2.1163776448892273, "bytes_per_chunk": 4.19121926262271, "reads_per_chunk": 2.8692162568349695}}, "bpb_mean": 1.9878538955951919, "gini_cpc": 0.03672301633580277, "gini_cps": 0.05390049526108065, "bpic": 4.278754053850215, "mean_boundary_entropy": 0.07047656178474426, "dechunk": "ema", "router_noise": 0.0, "transitions": [{"step": 0, "phase": 1, "objective": "global", "router_trainable": true, "encoder_trainable": true, "frozen_router_params": 0}, {"step": 2400, "phase": 2, "objective": "global", "router_trainable": true, "encoder_trainable": true, "frozen_router_params": 0}, {"step": 4200, "phase": 3, "objective": "global", "router_trainable": true, "encoder_trainable": true, "frozen_router_params": 0}], "probe_overhead_sec_total": 769.2024283409119, "probe_overhead_seq_frac": 0.4998697916666667, "wall_sec": 5678.775774717331, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 6000, "admitted_surface": 16127, "staged_surface": 0, "admitted_split": 15618, "staged_split": 0, "admitted_merge": 22479, "staged_merge": 0, "admitted_skel": 11312, "staged_skel": 0, "mean_count": 2055.7568359375, "n_util_obs": 878, "util_count_spearman": 0.04335744847761789}, "ali_params": 268096, "ali_config": {"enabled": true, "budget_B": 65536, "staging_capacity": 262144, "policy": "frequency", "value_mode": "learned", "candidate_types": ["surface", "split", "merge", "skel"], "gate": "vector", "retrieval": true, "memory_scope": "persistent", "read_dropout": 0.1, "write_rule": "mean", "eta_min": 0.0, "consolidate_every": 50, "probe_seqs": 2, "probe_reps": 8, "horizon": 32, "predictor": "mlp", "lambda_util": 0.1, "max_span": 64, "byte_classes_path": "/local/scratch/sas245/beetle-hnet/scratch/data/byte_classes.json", "gate_hidden": 0, "type_dim": 16, "gate_bias_init": -2.0}, "util_vs_frequency_spearman": 0.04335744847761789, "paired_vs_dropout": {"n": 1977, "pearson": -0.009101918160019415, "spearman": -0.0197561423631308, "mean_abs_discrepancy": 0.3950422704219818, "mean_paired": 0.014537793584167957, "mean_dropout": 0.020082391798496246}, "predictor_calibration": {"n": 4833, "all": {"n": 4833, "pearson": 0.09692244709165086, "spearman": 0.07902655696036502, "mae": 0.07802729308605194, "reliability": [{"pred": -0.0058213802985847, "obs": 0.0046739233657717705, "n": 484}, {"pred": 0.0004906650865450501, "obs": 0.0051831030286848545, "n": 483}, {"pred": 0.003542698686942458, "obs": 0.0099435243755579, "n": 483}, {"pred": 0.005755372811108828, "obs": 0.0013852394185960293, "n": 483}, {"pred": 0.0076438006944954395, "obs": 0.008460395969450474, "n": 483}, {"pred": 0.009310845285654068, "obs": 0.013862266205251217, "n": 484}, {"pred": 0.0110578378662467, "obs": 0.017848312854766846, "n": 483}, {"pred": 0.01364006195217371, "obs": 0.016601432114839554, "n": 483}, {"pred": 0.01776207983493805, "obs": 0.028665343299508095, "n": 483}, {"pred": 0.02886735089123249, "obs": 0.03570443019270897, "n": 484}]}, "admitted": {"n": 4833, "pearson": 0.09692244709165086, "spearman": 0.07902655696036502, "mae": 0.07802729308605194, "reliability": [{"pred": -0.0058213802985847, "obs": 0.0046739233657717705, "n": 484}, {"pred": 0.0004906650865450501, "obs": 0.0051831030286848545, "n": 483}, {"pred": 0.003542698686942458, "obs": 0.0099435243755579, "n": 483}, {"pred": 0.005755372811108828, "obs": 0.0013852394185960293, "n": 483}, {"pred": 0.0076438006944954395, "obs": 0.008460395969450474, "n": 483}, {"pred": 0.009310845285654068, "obs": 0.013862266205251217, "n": 484}, {"pred": 0.0110578378662467, "obs": 0.017848312854766846, "n": 483}, {"pred": 0.01364006195217371, "obs": 0.016601432114839554, "n": 483}, {"pred": 0.01776207983493805, "obs": 0.028665343299508095, "n": 483}, {"pred": 0.02886735089123249, "obs": 0.03570443019270897, "n": 484}]}, "staged": {"n": 0, "pearson": NaN, "spearman": NaN, "mae": NaN}}, "t": 5679.7} |
| {"event": "hub_done", "repo": "AdaptiveChunking/ali_mem-mixed_s3", "pushed": [0, 1, 2, 4, 6, 8, 16, 32, 64, 128, 256, 512, 1000, 1024, 2048, 3000, 4096, 6000], "failed": [], "ok": true, "t": 5690.4} |
|
|
