{"event": "hub", "repo": "AdaptiveChunking/ali_mem-mixed_s2", "keep_steps": [0, 1, 2, 4, 6, 8, 16, 32, 64, 128, 256, 512, 1000, 1024, 2048, 3000, 4096, 6000], "t": 2.0}
{"event": "start", "run": "ali_mem-mixed_s2", "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": 2.0}
{"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": 2.0}
{"event": "probe", "step": 0, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 10933.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.628225, "cps": 85.414062, "mean_p": 0.513249, "boundary_entropy": 0.686295, "bpb": 8.12598}, "tr": {"chunks": 8187.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.452245, "cps": 63.960938, "mean_p": 0.498002, "boundary_entropy": 0.686826, "bpb": 8.097536}, "fi": {"chunks": 13540.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.719792, "cps": 105.78125, "mean_p": 0.517169, "boundary_entropy": 0.686661, "bpb": 8.065117}}, "bpb": {"en": 8.12598, "tr": 8.09754, "fi": 8.06512}, "gini_cpc": 0.099077, "gini_cps": 0.109267, "bpic": 1.737048, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.7198781817505053, "fi": 1.1457544508645294}, "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": 4.7}
{"event": "train", "step": 0, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 5.62735, "bpb": 8.11854, "chunk_loss": 0.062086, "F": 0.59833, "G": 0.50836, "chunks": 612.69, "bytes_per_chunk": 1.671, "boundary_entropy": 0.69105, "router_noise": 0.0, "grad_norm": 11.4935, "lr": 1e-06, "router_trainable": true, "sec_per_step": 0.015, "probe_sec": 0.0105, "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": 58765, "read_suppressed_frac": 0.0, "dropout_reads_logged": 0, "inventory": {"size": 11478, "admitted": 11478, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 2006, "staged_surface": 0, "admitted_split": 1198, "staged_split": 0, "admitted_merge": 7355, "staged_merge": 0, "admitted_skel": 919, "staged_skel": 0, "mean_count": 5.1197943687438965, "n_util_obs": 0, "util_count_spearman": NaN}, "probe_overhead_sec_total": 0.011, "probe_overhead_seq_frac": 0.0, "util_loss": -0.0, "alpha": 0.03, "t": 5.7}
{"event": "probe", "step": 1, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 10908.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.626788, "cps": 85.21875, "mean_p": 0.513143, "boundary_entropy": 0.686304, "bpb": 8.127096}, "tr": {"chunks": 8206.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.453295, "cps": 64.109375, "mean_p": 0.498082, "boundary_entropy": 0.686828, "bpb": 8.101378}, "fi": {"chunks": 13508.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.71809, "cps": 105.53125, "mean_p": 0.517038, "boundary_entropy": 0.686672, "bpb": 8.09858}}, "bpb": {"en": 8.1271, "tr": 8.10138, "fi": 8.09858}, "gini_cpc": 0.098172, "gini_cps": 0.108352, "bpic": 1.739072, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.7232025555756703, "fi": 1.1456663554824995}, "inventory": {"size": 11478, "admitted": 11478, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 2006, "staged_surface": 0, "admitted_split": 1198, "staged_split": 0, "admitted_merge": 7355, "staged_merge": 0, "admitted_skel": 919, "staged_skel": 0, "mean_count": 5.1197943687438965, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 7.6}
{"event": "probe", "step": 2, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 10880.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.62518, "cps": 85.0, "mean_p": 0.512985, "boundary_entropy": 0.686318, "bpb": 8.068778}, "tr": {"chunks": 8252.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.455836, "cps": 64.46875, "mean_p": 0.498262, "boundary_entropy": 0.686832, "bpb": 8.059751}, "fi": {"chunks": 13470.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.71607, "cps": 105.234375, "mean_p": 0.516807, "boundary_entropy": 0.686691, "bpb": 8.046688}}, "bpb": {"en": 8.06878, "tr": 8.05975, "fi": 8.04669}, "gini_cpc": 0.096539, "gini_cps": 0.106701, "bpic": 1.740139, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.7291281953592352, "fi": 1.1453835384959363}, "inventory": {"size": 20130, "admitted": 20130, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3, "admitted_surface": 3536, "staged_surface": 0, "admitted_split": 2221, "staged_split": 0, "admitted_merge": 12617, "staged_merge": 0, "admitted_skel": 1756, "staged_skel": 0, "mean_count": 5.563387870788574, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 10.5}
{"event": "probe", "step": 4, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 10714.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.615641, "cps": 83.703125, "mean_p": 0.512107, "boundary_entropy": 0.686379, "bpb": 7.864517}, "tr": {"chunks": 8402.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.464122, "cps": 65.640625, "mean_p": 0.49872, "boundary_entropy": 0.686845, "bpb": 7.909359}, "fi": {"chunks": 13210.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.702249, "cps": 103.203125, "mean_p": 0.515657, "boundary_entropy": 0.686776, "bpb": 7.85436}}, "bpb": {"en": 7.86452, "tr": 7.90936, "fi": 7.85436}, "gini_cpc": 0.089085, "gini_cps": 0.099157, "bpic": 1.754996, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.7538841349529471, "fi": 1.140678911007382}, "inventory": {"size": 34615, "admitted": 34615, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 9, "admitted_surface": 5995, "staged_surface": 0, "admitted_split": 3795, "staged_split": 0, "admitted_merge": 21901, "staged_merge": 0, "admitted_skel": 2924, "staged_skel": 0, "mean_count": 6.47979211807251, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 14.3}
{"event": "probe", "step": 6, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 10323.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.593174, "cps": 80.648438, "mean_p": 0.510373, "boundary_entropy": 0.686481, "bpb": 7.581708}, "tr": {"chunks": 8516.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.470419, "cps": 66.53125, "mean_p": 0.499244, "boundary_entropy": 0.686865, "bpb": 7.697746}, "fi": {"chunks": 12719.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.676147, "cps": 99.367188, "mean_p": 0.513522, "boundary_entropy": 0.686913, "bpb": 7.591148}}, "bpb": {"en": 7.58171, "tr": 7.69775, "fi": 7.59115}, "gini_cpc": 0.078835, "gini_cps": 0.088789, "bpic": 1.797706, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.7930549755625977, "fi": 1.13988037538431}, "inventory": {"size": 45894, "admitted": 45894, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 15, "admitted_surface": 7731, "staged_surface": 0, "admitted_split": 4804, "staged_split": 0, "admitted_merge": 29561, "staged_merge": 0, "admitted_skel": 3798, "staged_skel": 0, "mean_count": 7.385061264038086, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 18.0}
{"event": "probe", "step": 8, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 9663.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.555249, "cps": 75.492188, "mean_p": 0.507382, "boundary_entropy": 0.68661, "bpb": 7.314284}, "tr": {"chunks": 8615.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.475888, "cps": 67.304688, "mean_p": 0.499616, "boundary_entropy": 0.686893, "bpb": 7.486029}, "fi": {"chunks": 11794.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.626974, "cps": 92.140625, "mean_p": 0.510029, "boundary_entropy": 0.687087, "bpb": 7.34057}}, "bpb": {"en": 7.31428, "tr": 7.48603, "fi": 7.34057}, "gini_cpc": 0.060746, "gini_cps": 0.070475, "bpic": 1.886539, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.8570711391847022, "fi": 1.1291753286103166}, "inventory": {"size": 58322, "admitted": 58322, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 20, "admitted_surface": 9810, "staged_surface": 0, "admitted_split": 6083, "staged_split": 0, "admitted_merge": 37532, "staged_merge": 0, "admitted_skel": 4897, "staged_skel": 0, "mean_count": 7.673056602478027, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 21.6}
{"event": "probe", "step": 16, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 4072.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.233983, "cps": 31.8125, "mean_p": 0.486081, "boundary_entropy": 0.68629, "bpb": 6.72596}, "tr": {"chunks": 7643.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.422195, "cps": 59.710938, "mean_p": 0.495847, "boundary_entropy": 0.687021, "bpb": 7.014822}, "fi": {"chunks": 4654.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.247408, "cps": 36.359375, "mean_p": 0.486849, "boundary_entropy": 0.686917, "bpb": 6.789018}}, "bpb": {"en": 6.72596, "tr": 7.01482, "fi": 6.78902}, "gini_cpc": 0.138863, "gini_cps": 0.145438, "bpic": 3.46582, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.8043868734322819, "fi": 1.0573794029514354}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 35, "admitted_surface": 13416, "staged_surface": 0, "admitted_split": 10982, "staged_split": 0, "admitted_merge": 32223, "staged_merge": 0, "admitted_skel": 8915, "staged_skel": 0, "mean_count": 10.489578247070312, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 28.7}
{"event": "probe", "step": 32, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 2595.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.149112, "cps": 20.273438, "mean_p": 0.486794, "boundary_entropy": 0.686956, "bpb": 6.306159}, "tr": {"chunks": 4439.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.245208, "cps": 34.679688, "mean_p": 0.489324, "boundary_entropy": 0.687497, "bpb": 6.651777}, "fi": {"chunks": 1290.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.068577, "cps": 10.078125, "mean_p": 0.478158, "boundary_entropy": 0.686841, "bpb": 6.422764}}, "bpb": {"en": 6.30616, "tr": 6.65178, "fi": 6.42276}, "gini_cpc": 0.254385, "gini_cps": 0.252202, "bpic": 6.815473, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.644452568938371, "fi": 0.4599012445983057}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 65, "admitted_surface": 14016, "staged_surface": 0, "admitted_split": 12593, "staged_split": 0, "admitted_merge": 29604, "staged_merge": 0, "admitted_skel": 9323, "staged_skel": 0, "mean_count": 13.928192138671875, "n_util_obs": 3, "util_count_spearman": -0.5}, "t": 40.0}
{"event": "consolidate", "step": 50, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 50.4}
{"event": "train", "step": 50, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 4.21271, "bpb": 6.07765, "chunk_loss": 0.033746, "F": 0.21933, "G": 0.48783, "chunks": 224.59, "bytes_per_chunk": 4.559, "boundary_entropy": 0.69155, "router_noise": 0.0, "grad_norm": 2.4822, "lr": 5.1e-05, "router_trainable": true, "sec_per_step": 0.904, "probe_sec": 0.0678, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05077, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": 0.02139, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.02759, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.20365, "trunc_abs_frac": 0.26072, "mean_delta_full": 0.04016, "read_hit_rate": {"surface": 0.6601, "split": 0.76068, "merge": 0.35331, "skel": 0.46051}, "mean_gate": {"surface": 0.11913, "split": 0.11923, "merge": 0.11911, "skel": 0.11916}, "n_reads": 15665, "n_cand": 26729, "read_suppressed_frac": 0.1007, "dropout_reads_logged": 15665, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 106, "admitted_surface": 14425, "staged_surface": 0, "admitted_split": 12916, "staged_split": 0, "admitted_merge": 29062, "staged_merge": 0, "admitted_skel": 9133, "staged_skel": 0, "mean_count": 17.233795166015625, "n_util_obs": 4, "util_count_spearman": -0.8}, "probe_overhead_sec_total": 3.915, "probe_overhead_seq_frac": 0.4865, "util_loss": 0.075267, "alpha": 0.03, "t": 50.9}
{"event": "probe", "step": 64, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 5163.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.296673, "cps": 40.335938, "mean_p": 0.486818, "boundary_entropy": 0.685566, "bpb": 5.540883}, "tr": {"chunks": 5565.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.307408, "cps": 43.476562, "mean_p": 0.487038, "boundary_entropy": 0.686502, "bpb": 5.947531}, "fi": {"chunks": 4892.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.260061, "cps": 38.21875, "mean_p": 0.484226, "boundary_entropy": 0.686267, "bpb": 5.712375}}, "bpb": {"en": 5.54088, "tr": 5.94753, "fi": 5.71237}, "gini_cpc": 0.036527, "gini_cps": 0.028724, "bpic": 3.63201, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0361833568858012, "fi": 0.8765900970759286}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 134, "admitted_surface": 14811, "staged_surface": 0, "admitted_split": 13437, "staged_split": 0, "admitted_merge": 28052, "staged_merge": 0, "admitted_skel": 9236, "staged_skel": 0, "mean_count": 19.997390747070312, "n_util_obs": 6, "util_count_spearman": -0.2571428571428571}, "t": 61.2}
{"event": "consolidate", "step": 100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 84.7}
{"event": "train", "step": 100, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 3.25176, "bpb": 4.69129, "chunk_loss": 0.029952, "F": 0.16562, "G": 0.45447, "chunks": 169.59, "bytes_per_chunk": 6.038, "boundary_entropy": 0.68265, "router_noise": 0.0, "grad_norm": 2.1065, "lr": 0.000101, "router_trainable": true, "sec_per_step": 0.694, "probe_sec": 0.0557, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04159, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.00325, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.01636, "n_merge": 2, "mean_delta_merge": -0.0033, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -1.60417, "trunc_abs_frac": 4.59816, "mean_delta_full": 0.01367, "read_hit_rate": {"surface": 0.54802, "split": 0.74676, "merge": 0.19239, "skel": 0.49271}, "mean_gate": {"surface": 0.11944, "split": 0.11948, "merge": 0.11939, "skel": 0.11943}, "n_reads": 12553, "n_cand": 23437, "read_suppressed_frac": 0.1024, "dropout_reads_logged": 12553, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 210, "admitted_surface": 15588, "staged_surface": 0, "admitted_split": 15397, "staged_split": 0, "admitted_merge": 24403, "staged_merge": 0, "admitted_skel": 10148, "staged_skel": 0, "mean_count": 28.54510498046875, "n_util_obs": 11, "util_count_spearman": 0.11818181818181818}, "probe_overhead_sec_total": 7.511, "probe_overhead_seq_frac": 0.4932, "util_loss": 0.002997, "alpha": 0.03, "t": 85.6}
{"event": "probe", "step": 128, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3852.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.221341, "cps": 30.09375, "mean_p": 0.422454, "boundary_entropy": 0.648208, "bpb": 4.153655}, "tr": {"chunks": 5003.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.276363, "cps": 39.085938, "mean_p": 0.441076, "boundary_entropy": 0.659929, "bpb": 4.413441}, "fi": {"chunks": 4062.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.215937, "cps": 31.734375, "mean_p": 0.43086, "boundary_entropy": 0.660214, "bpb": 4.343686}}, "bpb": {"en": 4.15365, "tr": 4.41344, "fi": 4.34369}, "gini_cpc": 0.056448, "gini_cps": 0.059405, "bpic": 4.392041, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.2485840800555768, "fi": 0.9755867142758488}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 271, "admitted_surface": 15925, "staged_surface": 0, "admitted_split": 15386, "staged_split": 0, "admitted_merge": 24134, "staged_merge": 0, "admitted_skel": 10091, "staged_skel": 0, "mean_count": 35.40995788574219, "n_util_obs": 14, "util_count_spearman": 0.4153846153846154}, "t": 106.4}
{"event": "consolidate", "step": 150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 122.1}
{"event": "train", "step": 150, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.79779, "bpb": 4.03636, "chunk_loss": 0.032255, "F": 0.21243, "G": 0.3914, "chunks": 217.53, "bytes_per_chunk": 4.707, "boundary_entropy": 0.62422, "router_noise": 0.0, "grad_norm": 1.3967, "lr": 0.000151, "router_trainable": true, "sec_per_step": 0.751, "probe_sec": 0.0733, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05329, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.0172, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": -0.02591, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.47574, "trunc_abs_frac": 0.93282, "mean_delta_full": 0.11133, "read_hit_rate": {"surface": 0.63689, "split": 0.80977, "merge": 0.26229, "skel": 0.54227}, "mean_gate": {"surface": 0.11997, "split": 0.12007, "merge": 0.11993, "skel": 0.12011}, "n_reads": 17038, "n_cand": 28523, "read_suppressed_frac": 0.0998, "dropout_reads_logged": 17038, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 312, "admitted_surface": 16137, "staged_surface": 0, "admitted_split": 15552, "staged_split": 0, "admitted_merge": 23471, "staged_merge": 0, "admitted_skel": 10376, "staged_skel": 0, "mean_count": 41.19984436035156, "n_util_obs": 17, "util_count_spearman": -0.031862745098039214}, "probe_overhead_sec_total": 11.523, "probe_overhead_seq_frac": 0.4954, "util_loss": 0.001153, "alpha": 0.03, "t": 123.1}
{"event": "consolidate", "step": 200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 154.1}
{"event": "probe", "step": 200, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3991.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.229328, "cps": 31.179688, "mean_p": 0.373765, "boundary_entropy": 0.584726, "bpb": 3.626497}, "tr": {"chunks": 4588.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.253439, "cps": 35.84375, "mean_p": 0.378779, "boundary_entropy": 0.593468, "bpb": 3.776241}, "fi": {"chunks": 3586.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.190633, "cps": 28.015625, "mean_p": 0.39828, "boundary_entropy": 0.631516, "bpb": 3.838434}}, "bpb": {"en": 3.6265, "tr": 3.77624, "fi": 3.83843}, "gini_cpc": 0.062178, "gini_cps": 0.054912, "bpic": 4.663543, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.105134788372994, "fi": 0.8312674864998797}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 407, "admitted_surface": 16257, "staged_surface": 0, "admitted_split": 15949, "staged_split": 0, "admitted_merge": 22866, "staged_merge": 0, "admitted_skel": 10464, "staged_skel": 0, "mean_count": 51.39299011230469, "n_util_obs": 24, "util_count_spearman": -0.38782608695652177}, "t": 155.6}
{"event": "eval", "step": 200, "phase": 1, "bpb": {"en": 3.63291, "tr": 3.78792, "fi": 3.83617}, "bytes_per_chunk": {"en": 4.941, "tr": 4.772, "fi": 5.681}, "bpb_mean": 3.75233, "frozen_inventory": true, "t": 157.9}
{"event": "train", "step": 200, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.52274, "bpb": 3.63955, "chunk_loss": 0.031231, "F": 0.19485, "G": 0.36462, "chunks": 199.53, "bytes_per_chunk": 5.132, "boundary_entropy": 0.5842, "router_noise": 0.0, "grad_norm": 2.3874, "lr": 0.000201, "router_trainable": true, "sec_per_step": 0.724, "probe_sec": 0.0494, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03456, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.01068, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": -0.02051, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.13715, "trunc_abs_frac": 3.02315, "mean_delta_full": -0.08856, "read_hit_rate": {"surface": 0.59641, "split": 0.78708, "merge": 0.21009, "skel": 0.50969}, "mean_gate": {"surface": 0.12255, "split": 0.12269, "merge": 0.12247, "skel": 0.12313}, "n_reads": 14938, "n_cand": 26534, "read_suppressed_frac": 0.0991, "dropout_reads_logged": 14938, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 408, "admitted_surface": 16266, "staged_surface": 0, "admitted_split": 16008, "staged_split": 0, "admitted_merge": 22827, "staged_merge": 0, "admitted_skel": 10435, "staged_skel": 0, "mean_count": 51.61769104003906, "n_util_obs": 24, "util_count_spearman": -0.3791304347826087}, "probe_overhead_sec_total": 15.007, "probe_overhead_seq_frac": 0.4966, "util_loss": 0.000509, "alpha": 0.03, "t": 159.3}
{"event": "consolidate", "step": 250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 192.6}
{"event": "train", "step": 250, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.38377, "bpb": 3.43905, "chunk_loss": 0.030898, "F": 0.18768, "G": 0.3592, "chunks": 192.19, "bytes_per_chunk": 5.328, "boundary_entropy": 0.55748, "router_noise": 0.0, "grad_norm": 2.0063, "lr": 0.000251, "router_trainable": true, "sec_per_step": 0.685, "probe_sec": 0.0512, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0365, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.0166, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": -0.05344, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.3777, "trunc_abs_frac": 1.04581, "mean_delta_full": -0.06213, "read_hit_rate": {"surface": 0.57535, "split": 0.77068, "merge": 0.17318, "skel": 0.50271}, "mean_gate": {"surface": 0.12765, "split": 0.12687, "merge": 0.12749, "skel": 0.127}, "n_reads": 14200, "n_cand": 26106, "read_suppressed_frac": 0.0961, "dropout_reads_logged": 14200, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 493, "admitted_surface": 16405, "staged_surface": 0, "admitted_split": 16349, "staged_split": 0, "admitted_merge": 22113, "staged_merge": 0, "admitted_skel": 10669, "staged_skel": 0, "mean_count": 62.569915771484375, "n_util_obs": 32, "util_count_spearman": 0.01906158357771261}, "probe_overhead_sec_total": 18.621, "probe_overhead_seq_frac": 0.4973, "util_loss": 0.000493, "alpha": 0.03, "t": 193.5}
{"event": "probe", "step": 256, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3277.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.188301, "cps": 25.601562, "mean_p": 0.361951, "boundary_entropy": 0.555948, "bpb": 3.347365}, "tr": {"chunks": 4622.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.255317, "cps": 36.109375, "mean_p": 0.336179, "boundary_entropy": 0.532965, "bpb": 3.510349}, "fi": {"chunks": 4724.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.25113, "cps": 36.90625, "mean_p": 0.377887, "boundary_entropy": 0.589492, "bpb": 3.640985}}, "bpb": {"en": 3.34737, "tr": 3.51035, "fi": 3.64099}, "gini_cpc": 0.064307, "gini_cps": 0.076421, "bpic": 4.494336, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.3558981511063457, "fi": 1.3336617153751464}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 499, "admitted_surface": 16489, "staged_surface": 0, "admitted_split": 16371, "staged_split": 0, "admitted_merge": 22145, "staged_merge": 0, "admitted_skel": 10531, "staged_skel": 0, "mean_count": 63.71891784667969, "n_util_obs": 32, "util_count_spearman": -0.002565982404692082}, "t": 198.5}
{"event": "consolidate", "step": 300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 234.2}
{"event": "train", "step": 300, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.27952, "bpb": 3.28866, "chunk_loss": 0.030479, "F": 0.17996, "G": 0.30995, "chunks": 184.28, "bytes_per_chunk": 5.557, "boundary_entropy": 0.4885, "router_noise": 0.0, "grad_norm": 1.2463, "lr": 0.0003, "router_trainable": true, "sec_per_step": 0.825, "probe_sec": 0.1094, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0942, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.02048, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 3, "mean_delta_split": 0.03825, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.49378, "trunc_abs_frac": 0.21413, "mean_delta_full": 0.01961, "read_hit_rate": {"surface": 0.51373, "split": 0.76413, "merge": 0.13218, "skel": 0.466}, "mean_gate": {"surface": 0.13007, "split": 0.13024, "merge": 0.12908, "skel": 0.13094}, "n_reads": 13237, "n_cand": 25601, "read_suppressed_frac": 0.0984, "dropout_reads_logged": 13237, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 590, "admitted_surface": 16626, "staged_surface": 0, "admitted_split": 16486, "staged_split": 0, "admitted_merge": 21790, "staged_merge": 0, "admitted_skel": 10634, "staged_skel": 0, "mean_count": 73.26396179199219, "n_util_obs": 44, "util_count_spearman": -0.106553911205074}, "probe_overhead_sec_total": 22.703, "probe_overhead_seq_frac": 0.4977, "util_loss": 0.000633, "alpha": 0.03, "t": 234.8}
{"event": "consolidate", "step": 350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 269.7}
{"event": "train", "step": 350, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.1669, "bpb": 3.12618, "chunk_loss": 0.031118, "F": 0.20648, "G": 0.28761, "chunks": 211.44, "bytes_per_chunk": 4.843, "boundary_entropy": 0.41287, "router_noise": 0.0, "grad_norm": 1.0823, "lr": 0.0002999, "router_trainable": true, "sec_per_step": 0.711, "probe_sec": 0.1588, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.14027, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.0389, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.05359, "n_merge": 1, "mean_delta_merge": -0.01294, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.07848, "trunc_abs_frac": 0.22738, "mean_delta_full": -0.05314, "read_hit_rate": {"surface": 0.59222, "split": 0.81718, "merge": 0.18845, "skel": 0.54541}, "mean_gate": {"surface": 0.13942, "split": 0.14015, "merge": 0.13742, "skel": 0.14315}, "n_reads": 16953, "n_cand": 29195, "read_suppressed_frac": 0.1002, "dropout_reads_logged": 16953, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 685, "admitted_surface": 16690, "staged_surface": 0, "admitted_split": 16422, "staged_split": 0, "admitted_merge": 21800, "staged_merge": 0, "admitted_skel": 10624, "staged_skel": 0, "mean_count": 85.02862548828125, "n_util_obs": 48, "util_count_spearman": -0.0790273556231003}, "probe_overhead_sec_total": 26.588, "probe_overhead_seq_frac": 0.498, "util_loss": 0.000771, "alpha": 0.03, "t": 270.3}
{"event": "consolidate", "step": 400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 315.9}
{"event": "probe", "step": 400, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3065.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.176119, "cps": 23.945312, "mean_p": 0.259795, "boundary_entropy": 0.367902, "bpb": 2.915679}, "tr": {"chunks": 3419.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.188864, "cps": 26.710938, "mean_p": 0.266863, "boundary_entropy": 0.412263, "bpb": 3.170126}, "fi": {"chunks": 4108.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.218383, "cps": 32.09375, "mean_p": 0.275761, "boundary_entropy": 0.427746, "bpb": 3.37703}}, "bpb": {"en": 2.91568, "tr": 3.17013, "fi": 3.37703}, "gini_cpc": 0.048299, "gini_cps": 0.065647, "bpic": 5.356118, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0723639128968288, "fi": 1.2399728977430944}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 766, "admitted_surface": 16719, "staged_surface": 0, "admitted_split": 16304, "staged_split": 0, "admitted_merge": 21791, "staged_merge": 0, "admitted_skel": 10722, "staged_skel": 0, "mean_count": 97.36286926269531, "n_util_obs": 56, "util_count_spearman": 0.04948735475051264}, "t": 317.7}
{"event": "eval", "step": 400, "phase": 1, "bpb": {"en": 2.92758, "tr": 3.17845, "fi": 3.3694}, "bytes_per_chunk": {"en": 5.272, "tr": 5.551, "fi": 4.909}, "bpb_mean": 3.15848, "frozen_inventory": true, "t": 321.7}
{"event": "train", "step": 400, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.08657, "bpb": 3.01029, "chunk_loss": 0.030475, "F": 0.18625, "G": 0.26643, "chunks": 190.72, "bytes_per_chunk": 5.369, "boundary_entropy": 0.37312, "router_noise": 0.0, "grad_norm": 1.4932, "lr": 0.0002998, "router_trainable": true, "sec_per_step": 1.128, "probe_sec": 0.1057, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05958, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.04061, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": 0.14355, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.72683, "trunc_abs_frac": 1.81607, "mean_delta_full": -0.15674, "read_hit_rate": {"surface": 0.54554, "split": 0.79485, "merge": 0.1439, "skel": 0.52302}, "mean_gate": {"surface": 0.15227, "split": 0.15163, "merge": 0.15215, "skel": 0.15459}, "n_reads": 14859, "n_cand": 27074, "read_suppressed_frac": 0.1026, "dropout_reads_logged": 14859, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 768, "admitted_surface": 16755, "staged_surface": 0, "admitted_split": 16530, "staged_split": 0, "admitted_merge": 21402, "staged_merge": 0, "admitted_skel": 10849, "staged_skel": 0, "mean_count": 97.58729553222656, "n_util_obs": 56, "util_count_spearman": 0.04490772385509228}, "probe_overhead_sec_total": 31.177, "probe_overhead_seq_frac": 0.4983, "util_loss": 0.001388, "alpha": 0.03, "t": 326.8}
{"event": "consolidate", "step": 450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 376.3}
{"event": "train", "step": 450, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.99535, "bpb": 2.87868, "chunk_loss": 0.03075, "F": 0.1991, "G": 0.26639, "chunks": 203.88, "bytes_per_chunk": 5.023, "boundary_entropy": 0.33582, "router_noise": 0.0, "grad_norm": 1.1473, "lr": 0.0002995, "router_trainable": true, "sec_per_step": 1.019, "probe_sec": 0.1869, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.15303, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.03144, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": -0.02428, "n_merge": 1, "mean_delta_merge": 0.03406, "n_dropped_unpaired": 0, "n_full": 0, "read_hit_rate": {"surface": 0.58148, "split": 0.82989, "merge": 0.16146, "skel": 0.56592}, "mean_gate": {"surface": 0.15696, "split": 0.1559, "merge": 0.15482, "skel": 0.15871}, "n_reads": 17090, "n_cand": 29315, "read_suppressed_frac": 0.1004, "dropout_reads_logged": 17090, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 851, "admitted_surface": 16698, "staged_surface": 0, "admitted_split": 16335, "staged_split": 0, "admitted_merge": 21630, "staged_merge": 0, "admitted_skel": 10873, "staged_skel": 0, "mean_count": 110.3275146484375, "n_util_obs": 66, "util_count_spearman": 0.06535852207493999}, "probe_overhead_sec_total": 36.432, "probe_overhead_seq_frac": 0.4985, "util_loss": 0.001565, "alpha": 0.03, "t": 377.7}
{"event": "consolidate", "step": 500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 425.6}
{"event": "train", "step": 500, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.89839, "bpb": 2.73879, "chunk_loss": 0.031079, "F": 0.21161, "G": 0.26773, "chunks": 216.69, "bytes_per_chunk": 4.726, "boundary_entropy": 0.30626, "router_noise": 0.0, "grad_norm": 1.3513, "lr": 0.0002992, "router_trainable": true, "sec_per_step": 0.97, "probe_sec": 0.0758, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04504, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": 0.05191, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.14734, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 1.06388, "trunc_abs_frac": 1.50872, "mean_delta_full": 0.16553, "read_hit_rate": {"surface": 0.60533, "split": 0.85503, "merge": 0.16783, "skel": 0.6226}, "mean_gate": {"surface": 0.15946, "split": 0.15802, "merge": 0.1569, "skel": 0.15997}, "n_reads": 19144, "n_cand": 31387, "read_suppressed_frac": 0.0985, "dropout_reads_logged": 19144, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 916, "admitted_surface": 16705, "staged_surface": 0, "admitted_split": 16219, "staged_split": 0, "admitted_merge": 21790, "staged_merge": 0, "admitted_skel": 10822, "staged_skel": 0, "mean_count": 123.66964721679688, "n_util_obs": 73, "util_count_spearman": 0.02378748611625324}, "probe_overhead_sec_total": 41.416, "probe_overhead_seq_frac": 0.4986, "util_loss": 0.00184, "alpha": 0.03, "t": 426.2}
{"event": "probe", "step": 512, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3522.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.202379, "cps": 27.515625, "mean_p": 0.261073, "boundary_entropy": 0.309696, "bpb": 2.687688}, "tr": {"chunks": 3364.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.185826, "cps": 26.28125, "mean_p": 0.26169, "boundary_entropy": 0.360556, "bpb": 3.028831}, "fi": {"chunks": 4112.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.218596, "cps": 32.125, "mean_p": 0.272272, "boundary_entropy": 0.371475, "bpb": 3.212211}}, "bpb": {"en": 2.68769, "tr": 3.02883, "fi": 3.21221}, "gini_cpc": 0.036003, "gini_cps": 0.045342, "bpic": 5.158392, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9182061647805416, "fi": 1.0801299069548365}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 932, "admitted_surface": 16732, "staged_surface": 0, "admitted_split": 16257, "staged_split": 0, "admitted_merge": 21747, "staged_merge": 0, "admitted_skel": 10800, "staged_skel": 0, "mean_count": 126.67146301269531, "n_util_obs": 75, "util_count_spearman": 0.0413371266002845}, "t": 437.9}
{"event": "consolidate", "step": 550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 471.7}
{"event": "train", "step": 550, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.89697, "bpb": 2.73675, "chunk_loss": 0.031129, "F": 0.21432, "G": 0.26865, "chunks": 219.47, "bytes_per_chunk": 4.666, "boundary_entropy": 0.29686, "router_noise": 0.0, "grad_norm": 1.6149, "lr": 0.0002987, "router_trainable": true, "sec_per_step": 0.923, "probe_sec": 0.079, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05335, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.08317, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.10938, "n_merge": 3, "mean_delta_merge": -0.03867, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -0.20566, "trunc_abs_frac": 2.94055, "mean_delta_full": 0.14034, "read_hit_rate": {"surface": 0.61264, "split": 0.85735, "merge": 0.17617, "skel": 0.62265}, "mean_gate": {"surface": 0.1621, "split": 0.15844, "merge": 0.16001, "skel": 0.16043}, "n_reads": 19423, "n_cand": 31651, "read_suppressed_frac": 0.1011, "dropout_reads_logged": 19423, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 995, "admitted_surface": 16734, "staged_surface": 0, "admitted_split": 16312, "staged_split": 0, "admitted_merge": 21613, "staged_merge": 0, "admitted_skel": 10877, "staged_skel": 0, "mean_count": 137.2106475830078, "n_util_obs": 82, "util_count_spearman": 0.15059696781706772}, "probe_overhead_sec_total": 45.836, "probe_overhead_seq_frac": 0.4988, "util_loss": 0.002051, "alpha": 0.03, "t": 472.4}
{"event": "consolidate", "step": 600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 515.5}
{"event": "probe", "step": 600, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3482.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.20008, "cps": 27.203125, "mean_p": 0.256913, "boundary_entropy": 0.285297, "bpb": 2.576851}, "tr": {"chunks": 3427.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.189306, "cps": 26.773438, "mean_p": 0.257323, "boundary_entropy": 0.326803, "bpb": 2.923483}, "fi": {"chunks": 4209.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.223752, "cps": 32.882812, "mean_p": 0.274644, "boundary_entropy": 0.336059, "bpb": 3.135205}}, "bpb": {"en": 2.57685, "tr": 2.92348, "fi": 3.13521}, "gini_cpc": 0.037454, "gini_cps": 0.046891, "bpic": 5.102716, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9461476312566268, "fi": 1.1183104823576684}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1075, "admitted_surface": 16744, "staged_surface": 0, "admitted_split": 16446, "staged_split": 0, "admitted_merge": 21448, "staged_merge": 0, "admitted_skel": 10898, "staged_skel": 0, "mean_count": 150.74581909179688, "n_util_obs": 87, "util_count_spearman": 0.17977691915141794}, "t": 517.0}
{"event": "eval", "step": 600, "phase": 1, "bpb": {"en": 2.58902, "tr": 2.93184, "fi": 3.13586}, "bytes_per_chunk": {"en": 4.72, "tr": 5.905, "fi": 4.642}, "bpb_mean": 2.88558, "frozen_inventory": true, "t": 519.8}
{"event": "train", "step": 600, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.88251, "bpb": 2.71589, "chunk_loss": 0.030937, "F": 0.20966, "G": 0.26213, "chunks": 214.69, "bytes_per_chunk": 4.77, "boundary_entropy": 0.28407, "router_noise": 0.0, "grad_norm": 1.1754, "lr": 0.0002982, "router_trainable": true, "sec_per_step": 0.986, "probe_sec": 0.0775, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04164, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.05351, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 3, "mean_delta_merge": -0.01554, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.6767, "trunc_abs_frac": 0.6767, "mean_delta_full": 0.68115, "read_hit_rate": {"surface": 0.59579, "split": 0.8477, "merge": 0.15971, "skel": 0.61093}, "mean_gate": {"surface": 0.16727, "split": 0.16116, "merge": 0.16518, "skel": 0.16278}, "n_reads": 18571, "n_cand": 30955, "read_suppressed_frac": 0.1014, "dropout_reads_logged": 18571, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1078, "admitted_surface": 16773, "staged_surface": 0, "admitted_split": 16369, "staged_split": 0, "admitted_merge": 21492, "staged_merge": 0, "admitted_skel": 10902, "staged_skel": 0, "mean_count": 151.01597595214844, "n_util_obs": 87, "util_count_spearman": 0.17986804694904135}, "probe_overhead_sec_total": 51.107, "probe_overhead_seq_frac": 0.4989, "util_loss": 0.002335, "alpha": 0.03, "t": 521.7}
{"event": "consolidate", "step": 650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 563.1}
{"event": "train", "step": 650, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.83491, "bpb": 2.64722, "chunk_loss": 0.031039, "F": 0.21225, "G": 0.26334, "chunks": 217.34, "bytes_per_chunk": 4.711, "boundary_entropy": 0.26924, "router_noise": 0.0, "grad_norm": 1.2147, "lr": 0.0002975, "router_trainable": true, "sec_per_step": 0.853, "probe_sec": 0.1816, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.15228, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.02296, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.08618, "n_merge": 2, "mean_delta_merge": -0.05759, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.02669, "trunc_abs_frac": 0.02669, "mean_delta_full": -0.48535, "read_hit_rate": {"surface": 0.62097, "split": 0.85766, "merge": 0.18024, "skel": 0.62554}, "mean_gate": {"surface": 0.17038, "split": 0.16315, "merge": 0.16717, "skel": 0.16586}, "n_reads": 19296, "n_cand": 31293, "read_suppressed_frac": 0.0987, "dropout_reads_logged": 19296, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1149, "admitted_surface": 16725, "staged_surface": 0, "admitted_split": 16366, "staged_split": 0, "admitted_merge": 21549, "staged_merge": 0, "admitted_skel": 10896, "staged_skel": 0, "mean_count": 164.94728088378906, "n_util_obs": 96, "util_count_spearman": 0.13415672191043607}, "probe_overhead_sec_total": 55.811, "probe_overhead_seq_frac": 0.4989, "util_loss": 0.002812, "alpha": 0.03, "t": 564.3}
{"event": "consolidate", "step": 700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 603.1}
{"event": "train", "step": 700, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.9779, "bpb": 2.85351, "chunk_loss": 0.03126, "F": 0.21536, "G": 0.2657, "chunks": 220.53, "bytes_per_chunk": 4.643, "boundary_entropy": 0.27474, "router_noise": 0.0, "grad_norm": 1.0808, "lr": 0.0002967, "router_trainable": true, "sec_per_step": 0.8, "probe_sec": 0.1705, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.14154, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.02717, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": -0.26991, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -1.42325, "trunc_abs_frac": 1.05864, "mean_delta_full": 0.07052, "read_hit_rate": {"surface": 0.59843, "split": 0.84249, "merge": 0.15844, "skel": 0.60161}, "mean_gate": {"surface": 0.17102, "split": 0.1618, "merge": 0.16879, "skel": 0.16316}, "n_reads": 18552, "n_cand": 31229, "read_suppressed_frac": 0.0992, "dropout_reads_logged": 18552, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1211, "admitted_surface": 16731, "staged_surface": 0, "admitted_split": 16350, "staged_split": 0, "admitted_merge": 21554, "staged_merge": 0, "admitted_skel": 10901, "staged_skel": 0, "mean_count": 179.14930725097656, "n_util_obs": 101, "util_count_spearman": 0.12684915550378567}, "probe_overhead_sec_total": 60.474, "probe_overhead_seq_frac": 0.499, "util_loss": 0.002813, "alpha": 0.03, "t": 604.3}
{"event": "consolidate", "step": 750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 642.5}
{"event": "train", "step": 750, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.773, "bpb": 2.5579, "chunk_loss": 0.031295, "F": 0.22501, "G": 0.26544, "chunks": 230.41, "bytes_per_chunk": 4.444, "boundary_entropy": 0.2514, "router_noise": 0.0, "grad_norm": 1.1565, "lr": 0.0002959, "router_trainable": true, "sec_per_step": 0.793, "probe_sec": 0.1866, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.12641, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": 0.04508, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.02197, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -0.65271, "trunc_abs_frac": 10.29648, "mean_delta_full": -0.0062, "read_hit_rate": {"surface": 0.64092, "split": 0.87251, "merge": 0.19414, "skel": 0.64101}, "mean_gate": {"surface": 0.1762, "split": 0.16286, "merge": 0.17294, "skel": 0.16469}, "n_reads": 20700, "n_cand": 32814, "read_suppressed_frac": 0.1006, "dropout_reads_logged": 20700, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1284, "admitted_surface": 16697, "staged_surface": 0, "admitted_split": 16265, "staged_split": 0, "admitted_merge": 21679, "staged_merge": 0, "admitted_skel": 10895, "staged_skel": 0, "mean_count": 193.51856994628906, "n_util_obs": 109, "util_count_spearman": 0.12736539709016773}, "probe_overhead_sec_total": 65.147, "probe_overhead_seq_frac": 0.4991, "util_loss": 0.003282, "alpha": 0.03, "t": 644.0}
{"event": "consolidate", "step": 800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 682.2}
{"event": "probe", "step": 800, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3713.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.213354, "cps": 29.007812, "mean_p": 0.250012, "boundary_entropy": 0.235422, "bpb": 2.417014}, "tr": {"chunks": 3803.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.210076, "cps": 29.710938, "mean_p": 0.2432, "boundary_entropy": 0.27648, "bpb": 2.749237}, "fi": {"chunks": 4344.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.230929, "cps": 33.9375, "mean_p": 0.264958, "boundary_entropy": 0.278913, "bpb": 2.967034}}, "bpb": {"en": 2.41701, "tr": 2.74924, "fi": 2.96703}, "gini_cpc": 0.021245, "gini_cps": 0.035469, "bpic": 4.783474, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9846342648410756, "fi": 1.0823733836819782}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1357, "admitted_surface": 16774, "staged_surface": 0, "admitted_split": 16375, "staged_split": 0, "admitted_merge": 21462, "staged_merge": 0, "admitted_skel": 10925, "staged_skel": 0, "mean_count": 207.94000244140625, "n_util_obs": 124, "util_count_spearman": 0.08498505114083399}, "t": 683.7}
{"event": "eval", "step": 800, "phase": 1, "bpb": {"en": 2.42755, "tr": 2.75564, "fi": 2.97023}, "bytes_per_chunk": {"en": 4.463, "tr": 5.06, "fi": 4.506}, "bpb_mean": 2.71781, "frozen_inventory": true, "t": 686.5}
{"event": "train", "step": 800, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.84794, "bpb": 2.66601, "chunk_loss": 0.03129, "F": 0.22443, "G": 0.26225, "chunks": 229.81, "bytes_per_chunk": 4.456, "boundary_entropy": 0.24373, "router_noise": 0.0, "grad_norm": 1.1614, "lr": 0.0002949, "router_trainable": true, "sec_per_step": 0.891, "probe_sec": 0.0663, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04357, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.04216, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.13528, "n_merge": 1, "mean_delta_merge": 0.58923, "n_dropped_unpaired": 0, "n_full": 0, "read_hit_rate": {"surface": 0.61677, "split": 0.85741, "merge": 0.18477, "skel": 0.60717}, "mean_gate": {"surface": 0.1804, "split": 0.16197, "merge": 0.1775, "skel": 0.16418}, "n_reads": 19641, "n_cand": 32220, "read_suppressed_frac": 0.0988, "dropout_reads_logged": 19641, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1357, "admitted_surface": 16748, "staged_surface": 0, "admitted_split": 16334, "staged_split": 0, "admitted_merge": 21523, "staged_merge": 0, "admitted_skel": 10931, "staged_skel": 0, "mean_count": 208.2176513671875, "n_util_obs": 124, "util_count_spearman": 0.08378612712225499}, "probe_overhead_sec_total": 69.974, "probe_overhead_seq_frac": 0.4991, "util_loss": 0.003284, "alpha": 0.03, "t": 688.5}
{"event": "consolidate", "step": 850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 726.3}
{"event": "train", "step": 850, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.80048, "bpb": 2.59754, "chunk_loss": 0.030962, "F": 0.21436, "G": 0.25633, "chunks": 219.5, "bytes_per_chunk": 4.665, "boundary_entropy": 0.24298, "router_noise": 0.0, "grad_norm": 0.9207, "lr": 0.0002938, "router_trainable": true, "sec_per_step": 0.778, "probe_sec": 0.0737, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05081, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": -0.56972, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 5, "mean_delta_split": 0.03634, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -0.14447, "trunc_abs_frac": 0.95054, "mean_delta_full": 0.01489, "read_hit_rate": {"surface": 0.60083, "split": 0.85743, "merge": 0.15718, "skel": 0.62036}, "mean_gate": {"surface": 0.18025, "split": 0.16069, "merge": 0.17433, "skel": 0.1668}, "n_reads": 19239, "n_cand": 31725, "read_suppressed_frac": 0.0985, "dropout_reads_logged": 19239, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1421, "admitted_surface": 16781, "staged_surface": 0, "admitted_split": 16386, "staged_split": 0, "admitted_merge": 21371, "staged_merge": 0, "admitted_skel": 10998, "staged_skel": 0, "mean_count": 223.07154846191406, "n_util_obs": 129, "util_count_spearman": 0.03464333631484794}, "probe_overhead_sec_total": 74.483, "probe_overhead_seq_frac": 0.4992, "util_loss": 0.003388, "alpha": 0.03, "t": 727.4}
{"event": "consolidate", "step": 900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 772.2}
{"event": "train", "step": 900, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.60254, "bpb": 2.31198, "chunk_loss": 0.031034, "F": 0.21857, "G": 0.25267, "chunks": 223.81, "bytes_per_chunk": 4.575, "boundary_entropy": 0.22576, "router_noise": 0.0, "grad_norm": 1.1644, "lr": 0.0002927, "router_trainable": true, "sec_per_step": 0.91, "probe_sec": 0.074, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04735, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.03654, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": -0.05518, "n_merge": 1, "mean_delta_merge": 0.13043, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.11386, "trunc_abs_frac": 0.91252, "mean_delta_full": -0.04279, "read_hit_rate": {"surface": 0.66157, "split": 0.87725, "merge": 0.20766, "skel": 0.66033}, "mean_gate": {"surface": 0.18662, "split": 0.15897, "merge": 0.18117, "skel": 0.16726}, "n_reads": 20567, "n_cand": 31955, "read_suppressed_frac": 0.0981, "dropout_reads_logged": 20567, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1488, "admitted_surface": 16726, "staged_surface": 0, "admitted_split": 16371, "staged_split": 0, "admitted_merge": 21496, "staged_merge": 0, "admitted_skel": 10943, "staged_skel": 0, "mean_count": 238.26095581054688, "n_util_obs": 135, "util_count_spearman": -0.0005779930957875429}, "probe_overhead_sec_total": 79.66, "probe_overhead_seq_frac": 0.4992, "util_loss": 0.003636, "alpha": 0.03, "t": 772.9}
{"event": "consolidate", "step": 950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 810.2}
{"event": "train", "step": 950, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.67605, "bpb": 2.41802, "chunk_loss": 0.031008, "F": 0.21808, "G": 0.2493, "chunks": 223.31, "bytes_per_chunk": 4.586, "boundary_entropy": 0.21688, "router_noise": 0.0, "grad_norm": 1.0574, "lr": 0.0002914, "router_trainable": true, "sec_per_step": 0.764, "probe_sec": 0.0678, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04442, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.00852, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.01813, "n_merge": 2, "mean_delta_merge": -0.25304, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 1.37942, "trunc_abs_frac": 3.12463, "mean_delta_full": -0.15698, "read_hit_rate": {"surface": 0.63846, "split": 0.86963, "merge": 0.18611, "skel": 0.63247}, "mean_gate": {"surface": 0.18852, "split": 0.15714, "merge": 0.17905, "skel": 0.16631}, "n_reads": 19891, "n_cand": 31787, "read_suppressed_frac": 0.1015, "dropout_reads_logged": 19891, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1560, "admitted_surface": 16761, "staged_surface": 0, "admitted_split": 16315, "staged_split": 0, "admitted_merge": 21485, "staged_merge": 0, "admitted_skel": 10975, "staged_skel": 0, "mean_count": 253.44850158691406, "n_util_obs": 141, "util_count_spearman": 0.04473650413117}, "probe_overhead_sec_total": 83.915, "probe_overhead_seq_frac": 0.4993, "util_loss": 0.003832, "alpha": 0.03, "t": 811.1}
{"event": "consolidate", "step": 1000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 865.2}
{"event": "probe", "step": 1000, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3663.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.210481, "cps": 28.617188, "mean_p": 0.241489, "boundary_entropy": 0.212277, "bpb": 2.322756}, "tr": {"chunks": 3765.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.207977, "cps": 29.414062, "mean_p": 0.229493, "boundary_entropy": 0.2448, "bpb": 2.634935}, "fi": {"chunks": 4377.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.232683, "cps": 34.195312, "mean_p": 0.260283, "boundary_entropy": 0.25378, "bpb": 2.835527}}, "bpb": {"en": 2.32276, "tr": 2.63493, "fi": 2.83553}, "gini_cpc": 0.025296, "gini_cps": 0.040322, "bpic": 4.80576, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9881016639564947, "fi": 1.1054824814327233}, "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": 16757, "staged_surface": 0, "admitted_split": 16542, "staged_split": 0, "admitted_merge": 21235, "staged_merge": 0, "admitted_skel": 11002, "staged_skel": 0, "mean_count": 268.236328125, "n_util_obs": 152, "util_count_spearman": 0.09675435179300022}, "t": 867.3}
{"event": "eval", "step": 1000, "phase": 1, "bpb": {"en": 2.3335, "tr": 2.64237, "fi": 2.84117}, "bytes_per_chunk": {"en": 4.503, "tr": 5.267, "fi": 4.491}, "bpb_mean": 2.60568, "frozen_inventory": true, "t": 870.3}
{"event": "train", "step": 1000, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.73405, "bpb": 2.50171, "chunk_loss": 0.031018, "F": 0.21964, "G": 0.2502, "chunks": 224.91, "bytes_per_chunk": 4.553, "boundary_entropy": 0.2194, "router_noise": 0.0, "grad_norm": 0.9617, "lr": 0.0002901, "router_trainable": true, "sec_per_step": 1.235, "probe_sec": 0.219, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.17933, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.0686, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": 0.08356, "n_merge": 1, "mean_delta_merge": -0.06573, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 3.92091, "trunc_abs_frac": 0.71357, "mean_delta_full": 0.00839, "read_hit_rate": {"surface": 0.618, "split": 0.86223, "merge": 0.17552, "skel": 0.64862}, "mean_gate": {"surface": 0.18846, "split": 0.15472, "merge": 0.17919, "skel": 0.16471}, "n_reads": 20050, "n_cand": 32346, "read_suppressed_frac": 0.0981, "dropout_reads_logged": 20050, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1636, "admitted_surface": 16809, "staged_surface": 0, "admitted_split": 16342, "staged_split": 0, "admitted_merge": 21483, "staged_merge": 0, "admitted_skel": 10902, "staged_skel": 0, "mean_count": 268.53515625, "n_util_obs": 153, "util_count_spearman": 0.10064935064935066}, "probe_overhead_sec_total": 90.101, "probe_overhead_seq_frac": 0.4993, "util_loss": 0.004036, "alpha": 0.03, "t": 872.9}
{"event": "probe", "step": 1024, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3583.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.205884, "cps": 27.992188, "mean_p": 0.236072, "boundary_entropy": 0.212703, "bpb": 2.314107}, "tr": {"chunks": 3805.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.210186, "cps": 29.726562, "mean_p": 0.228022, "boundary_entropy": 0.242732, "bpb": 2.623257}, "fi": {"chunks": 4380.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.232842, "cps": 34.21875, "mean_p": 0.25886, "boundary_entropy": 0.248891, "bpb": 2.8122}}, "bpb": {"en": 2.31411, "tr": 2.62326, "fi": 2.8122}, "gini_cpc": 0.027696, "gini_cps": 0.045151, "bpic": 4.82087, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.02089581080285, "fi": 1.1309399330232563}, "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": 16775, "staged_surface": 0, "admitted_split": 16430, "staged_split": 0, "admitted_merge": 21334, "staged_merge": 0, "admitted_skel": 10997, "staged_skel": 0, "mean_count": 275.4134521484375, "n_util_obs": 156, "util_count_spearman": 0.09040167723104815}, "t": 903.8}
{"event": "consolidate", "step": 1050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 930.5}
{"event": "train", "step": 1050, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.6227, "bpb": 2.34106, "chunk_loss": 0.031173, "F": 0.22293, "G": 0.25057, "chunks": 228.28, "bytes_per_chunk": 4.486, "boundary_entropy": 0.2099, "router_noise": 0.0, "grad_norm": 0.9814, "lr": 0.0002886, "router_trainable": true, "sec_per_step": 1.165, "probe_sec": 0.0783, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05031, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": 0.08407, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.21179, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 1.38491, "trunc_abs_frac": 2.32485, "mean_delta_full": 0.31726, "read_hit_rate": {"surface": 0.65159, "split": 0.87828, "merge": 0.20177, "skel": 0.65374}, "mean_gate": {"surface": 0.19214, "split": 0.15334, "merge": 0.18036, "skel": 0.16664}, "n_reads": 20674, "n_cand": 32365, "read_suppressed_frac": 0.0994, "dropout_reads_logged": 20674, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1703, "admitted_surface": 16773, "staged_surface": 0, "admitted_split": 16299, "staged_split": 0, "admitted_merge": 21536, "staged_merge": 0, "admitted_skel": 10928, "staged_skel": 0, "mean_count": 283.65692138671875, "n_util_obs": 160, "util_count_spearman": 0.08701078680623657}, "probe_overhead_sec_total": 96.003, "probe_overhead_seq_frac": 0.4993, "util_loss": 0.004204, "alpha": 0.03, "t": 931.1}
{"event": "consolidate", "step": 1100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 969.2}
{"event": "train", "step": 1100, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.72804, "bpb": 2.49304, "chunk_loss": 0.031046, "F": 0.22128, "G": 0.24795, "chunks": 226.59, "bytes_per_chunk": 4.519, "boundary_entropy": 0.21138, "router_noise": 0.0, "grad_norm": 0.9507, "lr": 0.0002871, "router_trainable": true, "sec_per_step": 0.786, "probe_sec": 0.0689, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0434, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": 0.01017, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.0918, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.4227, "trunc_abs_frac": 0.43712, "mean_delta_full": 0.34753, "read_hit_rate": {"surface": 0.63596, "split": 0.86575, "merge": 0.19521, "skel": 0.62819}, "mean_gate": {"surface": 0.19585, "split": 0.15113, "merge": 0.18182, "skel": 0.16506}, "n_reads": 20090, "n_cand": 32152, "read_suppressed_frac": 0.0971, "dropout_reads_logged": 20090, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1762, "admitted_surface": 16745, "staged_surface": 0, "admitted_split": 16367, "staged_split": 0, "admitted_merge": 21457, "staged_merge": 0, "admitted_skel": 10967, "staged_skel": 0, "mean_count": 298.8760070800781, "n_util_obs": 165, "util_count_spearman": 0.08242470871258484}, "probe_overhead_sec_total": 100.629, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.004383, "alpha": 0.03, "t": 970.4}
{"event": "consolidate", "step": 1150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1008.2}
{"event": "train", "step": 1150, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.63546, "bpb": 2.35947, "chunk_loss": 0.030978, "F": 0.2218, "G": 0.24374, "chunks": 227.12, "bytes_per_chunk": 4.509, "boundary_entropy": 0.19645, "router_noise": 0.0, "grad_norm": 0.8369, "lr": 0.0002855, "router_trainable": true, "sec_per_step": 0.769, "probe_sec": 0.1044, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08313, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.04984, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.18652, "n_merge": 2, "mean_delta_merge": 0.39297, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.90539, "trunc_abs_frac": 0.90539, "mean_delta_full": 0.23999, "read_hit_rate": {"surface": 0.64677, "split": 0.8746, "merge": 0.20225, "skel": 0.66231}, "mean_gate": {"surface": 0.19755, "split": 0.14822, "merge": 0.18486, "skel": 0.16381}, "n_reads": 20723, "n_cand": 32455, "read_suppressed_frac": 0.0966, "dropout_reads_logged": 20723, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1832, "admitted_surface": 16791, "staged_surface": 0, "admitted_split": 16376, "staged_split": 0, "admitted_merge": 21440, "staged_merge": 0, "admitted_skel": 10929, "staged_skel": 0, "mean_count": 314.2191162109375, "n_util_obs": 174, "util_count_spearman": 0.05041487513881375}, "probe_overhead_sec_total": 105.562, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.004412, "alpha": 0.03, "t": 1008.8}
{"event": "consolidate", "step": 1200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1049.8}
{"event": "probe", "step": 1200, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3653.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.209906, "cps": 28.539062, "mean_p": 0.234422, "boundary_entropy": 0.18933, "bpb": 2.256098}, "tr": {"chunks": 4006.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.221289, "cps": 31.296875, "mean_p": 0.222902, "boundary_entropy": 0.210589, "bpb": 2.544469}, "fi": {"chunks": 4792.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.254745, "cps": 37.4375, "mean_p": 0.263038, "boundary_entropy": 0.219902, "bpb": 2.72197}}, "bpb": {"en": 2.2561, "tr": 2.54447, "fi": 2.72197}, "gini_cpc": 0.043578, "gini_cps": 0.060986, "bpic": 4.556421, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.054228715481063, "fi": 1.2136106360236218}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1902, "admitted_surface": 16750, "staged_surface": 0, "admitted_split": 16423, "staged_split": 0, "admitted_merge": 21382, "staged_merge": 0, "admitted_skel": 10981, "staged_skel": 0, "mean_count": 329.2275085449219, "n_util_obs": 178, "util_count_spearman": 0.06464637222598728}, "t": 1051.6}
{"event": "eval", "step": 1200, "phase": 1, "bpb": {"en": 2.266, "tr": 2.55477, "fi": 2.73143}, "bytes_per_chunk": {"en": 4.473, "tr": 4.954, "fi": 4.008}, "bpb_mean": 2.5174, "frozen_inventory": true, "t": 1054.9}
{"event": "train", "step": 1200, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.6087, "bpb": 2.32086, "chunk_loss": 0.031024, "F": 0.2244, "G": 0.24392, "chunks": 229.78, "bytes_per_chunk": 4.456, "boundary_entropy": 0.19067, "router_noise": 0.0, "grad_norm": 0.8755, "lr": 0.0002837, "router_trainable": true, "sec_per_step": 0.961, "probe_sec": 0.1336, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10617, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": -0.12543, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.04407, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 1.39916, "trunc_abs_frac": 0.59143, "mean_delta_full": -0.17002, "read_hit_rate": {"surface": 0.66016, "split": 0.87621, "merge": 0.22019, "skel": 0.66155}, "mean_gate": {"surface": 0.20755, "split": 0.15212, "merge": 0.19233, "skel": 0.17197}, "n_reads": 21113, "n_cand": 32714, "read_suppressed_frac": 0.1035, "dropout_reads_logged": 21113, "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": 16741, "staged_surface": 0, "admitted_split": 16393, "staged_split": 0, "admitted_merge": 21440, "staged_merge": 0, "admitted_skel": 10962, "staged_skel": 0, "mean_count": 329.5438232421875, "n_util_obs": 178, "util_count_spearman": 0.0656294252012652}, "probe_overhead_sec_total": 110.712, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.004551, "alpha": 0.03, "t": 1056.9}
{"event": "consolidate", "step": 1250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1096.8}
{"event": "train", "step": 1250, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.62291, "bpb": 2.34137, "chunk_loss": 0.030853, "F": 0.21588, "G": 0.23815, "chunks": 221.06, "bytes_per_chunk": 4.632, "boundary_entropy": 0.19006, "router_noise": 0.0, "grad_norm": 0.8821, "lr": 0.0002819, "router_trainable": true, "sec_per_step": 0.81, "probe_sec": 0.0717, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05158, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.02797, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": 0.09457, "n_merge": 1, "mean_delta_merge": -0.48096, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.94135, "trunc_abs_frac": 2.9332, "mean_delta_full": -0.06088, "read_hit_rate": {"surface": 0.64016, "split": 0.85855, "merge": 0.19358, "skel": 0.63586}, "mean_gate": {"surface": 0.19843, "split": 0.14354, "merge": 0.18044, "skel": 0.16191}, "n_reads": 19724, "n_cand": 31589, "read_suppressed_frac": 0.1007, "dropout_reads_logged": 19724, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1954, "admitted_surface": 16720, "staged_surface": 0, "admitted_split": 16538, "staged_split": 0, "admitted_merge": 21259, "staged_merge": 0, "admitted_skel": 11019, "staged_skel": 0, "mean_count": 344.81915283203125, "n_util_obs": 185, "util_count_spearman": 0.058746825366741216}, "probe_overhead_sec_total": 115.799, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.004796, "alpha": 0.03, "t": 1097.4}
{"event": "consolidate", "step": 1300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1137.7}
{"event": "train", "step": 1300, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.66297, "bpb": 2.39916, "chunk_loss": 0.031272, "F": 0.23096, "G": 0.25197, "chunks": 236.5, "bytes_per_chunk": 4.33, "boundary_entropy": 0.1853, "router_noise": 0.0, "grad_norm": 0.9122, "lr": 0.00028, "router_trainable": true, "sec_per_step": 0.822, "probe_sec": 0.1068, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08482, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.04119, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": 0.26065, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": -3.33026, "trunc_abs_frac": 3.33026, "mean_delta_full": -0.09277, "read_hit_rate": {"surface": 0.64756, "split": 0.88367, "merge": 0.20602, "skel": 0.6682}, "mean_gate": {"surface": 0.19728, "split": 0.13886, "merge": 0.17923, "skel": 0.1584}, "n_reads": 21597, "n_cand": 33589, "read_suppressed_frac": 0.102, "dropout_reads_logged": 21597, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2018, "admitted_surface": 16817, "staged_surface": 0, "admitted_split": 16348, "staged_split": 0, "admitted_merge": 21418, "staged_merge": 0, "admitted_skel": 10953, "staged_skel": 0, "mean_count": 360.1233825683594, "n_util_obs": 198, "util_count_spearman": 0.038306437587104884}, "probe_overhead_sec_total": 121.0, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.004749, "alpha": 0.03, "t": 1138.5}
{"event": "consolidate", "step": 1350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1178.3}
{"event": "train", "step": 1350, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.59322, "bpb": 2.29853, "chunk_loss": 0.030996, "F": 0.22256, "G": 0.24225, "chunks": 227.91, "bytes_per_chunk": 4.493, "boundary_entropy": 0.18346, "router_noise": 0.0, "grad_norm": 0.9965, "lr": 0.000278, "router_trainable": true, "sec_per_step": 0.815, "probe_sec": 0.0802, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05899, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.0266, "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": 0.75955, "trunc_abs_frac": 0.56464, "mean_delta_full": 0.1534, "read_hit_rate": {"surface": 0.64977, "split": 0.87548, "merge": 0.20916, "skel": 0.6495}, "mean_gate": {"surface": 0.20141, "split": 0.13783, "merge": 0.18204, "skel": 0.1614}, "n_reads": 20699, "n_cand": 32425, "read_suppressed_frac": 0.1021, "dropout_reads_logged": 20699, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2091, "admitted_surface": 16762, "staged_surface": 0, "admitted_split": 16450, "staged_split": 0, "admitted_merge": 21316, "staged_merge": 0, "admitted_skel": 11008, "staged_skel": 0, "mean_count": 375.4163818359375, "n_util_obs": 203, "util_count_spearman": 0.056026234324388684}, "probe_overhead_sec_total": 125.811, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.004895, "alpha": 0.03, "t": 1179.3}
{"event": "consolidate", "step": 1400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1218.5}
{"event": "probe", "step": 1400, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3565.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.20485, "cps": 27.851562, "mean_p": 0.228296, "boundary_entropy": 0.180778, "bpb": 2.204145}, "tr": {"chunks": 3852.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.212782, "cps": 30.09375, "mean_p": 0.215664, "boundary_entropy": 0.203631, "bpb": 2.484031}, "fi": {"chunks": 4534.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.241029, "cps": 35.421875, "mean_p": 0.256126, "boundary_entropy": 0.218863, "bpb": 2.629057}}, "bpb": {"en": 2.20414, "tr": 2.48403, "fi": 2.62906}, "gini_cpc": 0.036619, "gini_cps": 0.054054, "bpic": 4.74705, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0387243511280586, "fi": 1.1766145602939047}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2165, "admitted_surface": 16784, "staged_surface": 0, "admitted_split": 16469, "staged_split": 0, "admitted_merge": 21227, "staged_merge": 0, "admitted_skel": 11056, "staged_skel": 0, "mean_count": 390.4312438964844, "n_util_obs": 214, "util_count_spearman": 0.05562232337734296}, "t": 1220.3}
{"event": "eval", "step": 1400, "phase": 1, "bpb": {"en": 2.21291, "tr": 2.49221, "fi": 2.63961}, "bytes_per_chunk": {"en": 4.602, "tr": 5.086, "fi": 4.389}, "bpb_mean": 2.44824, "frozen_inventory": true, "t": 1223.4}
{"event": "train", "step": 1400, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.59327, "bpb": 2.2986, "chunk_loss": 0.030875, "F": 0.21689, "G": 0.23601, "chunks": 222.09, "bytes_per_chunk": 4.611, "boundary_entropy": 0.18364, "router_noise": 0.0, "grad_norm": 0.7963, "lr": 0.0002759, "router_trainable": true, "sec_per_step": 0.923, "probe_sec": 0.146, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08998, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": 0.0653, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.22119, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -0.37579, "trunc_abs_frac": 18.30669, "mean_delta_full": -0.14856, "read_hit_rate": {"surface": 0.64226, "split": 0.86049, "merge": 0.21142, "skel": 0.63464}, "mean_gate": {"surface": 0.20865, "split": 0.14047, "merge": 0.18453, "skel": 0.16665}, "n_reads": 19777, "n_cand": 31489, "read_suppressed_frac": 0.1015, "dropout_reads_logged": 19777, "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": 16765, "staged_surface": 0, "admitted_split": 16547, "staged_split": 0, "admitted_merge": 21165, "staged_merge": 0, "admitted_skel": 11059, "staged_skel": 0, "mean_count": 390.73681640625, "n_util_obs": 214, "util_count_spearman": 0.05644947254255235}, "probe_overhead_sec_total": 131.214, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.004934, "alpha": 0.03, "t": 1225.4}
{"event": "consolidate", "step": 1450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1268.9}
{"event": "train", "step": 1450, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.64617, "bpb": 2.37492, "chunk_loss": 0.030808, "F": 0.21786, "G": 0.23424, "chunks": 223.09, "bytes_per_chunk": 4.59, "boundary_entropy": 0.17612, "router_noise": 0.0, "grad_norm": 0.8249, "lr": 0.0002738, "router_trainable": true, "sec_per_step": 0.882, "probe_sec": 0.106, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07909, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.13485, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.11279, "n_merge": 1, "mean_delta_merge": -0.0874, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -0.32023, "trunc_abs_frac": 0.56598, "mean_delta_full": -0.15964, "read_hit_rate": {"surface": 0.63712, "split": 0.86039, "merge": 0.20283, "skel": 0.63347}, "mean_gate": {"surface": 0.20754, "split": 0.13494, "merge": 0.18805, "skel": 0.16298}, "n_reads": 19794, "n_cand": 31682, "read_suppressed_frac": 0.0985, "dropout_reads_logged": 19794, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2225, "admitted_surface": 16764, "staged_surface": 0, "admitted_split": 16555, "staged_split": 0, "admitted_merge": 21174, "staged_merge": 0, "admitted_skel": 11043, "staged_skel": 0, "mean_count": 406.0362243652344, "n_util_obs": 219, "util_count_spearman": 0.035855221817267816}, "probe_overhead_sec_total": 136.879, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.00511, "alpha": 0.03, "t": 1269.5}
{"event": "consolidate", "step": 1500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1310.0}
{"event": "train", "step": 1500, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.65781, "bpb": 2.39172, "chunk_loss": 0.031048, "F": 0.22324, "G": 0.24159, "chunks": 228.59, "bytes_per_chunk": 4.48, "boundary_entropy": 0.17748, "router_noise": 0.0, "grad_norm": 0.8404, "lr": 0.0002715, "router_trainable": true, "sec_per_step": 0.831, "probe_sec": 0.1432, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.11585, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.00869, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": -0.14368, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 1.88122, "trunc_abs_frac": 2.10822, "mean_delta_full": -0.0662, "read_hit_rate": {"surface": 0.63847, "split": 0.86404, "merge": 0.19928, "skel": 0.64309}, "mean_gate": {"surface": 0.2037, "split": 0.13093, "merge": 0.17792, "skel": 0.15895}, "n_reads": 20148, "n_cand": 32161, "read_suppressed_frac": 0.0994, "dropout_reads_logged": 20148, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2287, "admitted_surface": 16777, "staged_surface": 0, "admitted_split": 16497, "staged_split": 0, "admitted_merge": 21265, "staged_merge": 0, "admitted_skel": 10997, "staged_skel": 0, "mean_count": 421.15313720703125, "n_util_obs": 227, "util_count_spearman": 0.06712472149560449}, "probe_overhead_sec_total": 142.196, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.00515, "alpha": 0.03, "t": 1311.1}
{"event": "consolidate", "step": 1550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1361.3}
{"event": "train", "step": 1550, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.58222, "bpb": 2.28267, "chunk_loss": 0.030955, "F": 0.22073, "G": 0.23606, "chunks": 226.03, "bytes_per_chunk": 4.53, "boundary_entropy": 0.16751, "router_noise": 0.0, "grad_norm": 0.9102, "lr": 0.0002692, "router_trainable": true, "sec_per_step": 1.019, "probe_sec": 0.1277, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10138, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.01775, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 3, "mean_delta_split": 0.01166, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 0.2915, "trunc_abs_frac": 0.84943, "mean_delta_full": 0.15774, "read_hit_rate": {"surface": 0.64894, "split": 0.8688, "merge": 0.2197, "skel": 0.65768}, "mean_gate": {"surface": 0.20183, "split": 0.1287, "merge": 0.17349, "skel": 0.15828}, "n_reads": 20645, "n_cand": 32275, "read_suppressed_frac": 0.0979, "dropout_reads_logged": 20645, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2331, "admitted_surface": 16771, "staged_surface": 0, "admitted_split": 16548, "staged_split": 0, "admitted_merge": 21205, "staged_merge": 0, "admitted_skel": 11012, "staged_skel": 0, "mean_count": 436.4437255859375, "n_util_obs": 236, "util_count_spearman": 0.048913813790716854}, "probe_overhead_sec_total": 147.93, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.005147, "alpha": 0.03, "t": 1362.0}
{"event": "consolidate", "step": 1600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1415.0}
{"event": "probe", "step": 1600, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3552.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.204103, "cps": 27.75, "mean_p": 0.228394, "boundary_entropy": 0.172682, "bpb": 2.163054}, "tr": {"chunks": 3812.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.210573, "cps": 29.78125, "mean_p": 0.210537, "boundary_entropy": 0.184667, "bpb": 2.420522}, "fi": {"chunks": 4537.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.241189, "cps": 35.445312, "mean_p": 0.254115, "boundary_entropy": 0.196754, "bpb": 2.571734}}, "bpb": {"en": 2.16305, "tr": 2.42052, "fi": 2.57173}, "gini_cpc": 0.037697, "gini_cps": 0.055177, "bpic": 4.766994, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0317001736310691, "fi": 1.1817022404228694}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2377, "admitted_surface": 16810, "staged_surface": 0, "admitted_split": 16460, "staged_split": 0, "admitted_merge": 21234, "staged_merge": 0, "admitted_skel": 11032, "staged_skel": 0, "mean_count": 451.15130615234375, "n_util_obs": 247, "util_count_spearman": 0.08734953191245377}, "t": 1417.2}
{"event": "eval", "step": 1600, "phase": 1, "bpb": {"en": 2.17369, "tr": 2.43088, "fi": 2.58007}, "bytes_per_chunk": {"en": 4.608, "tr": 5.079, "fi": 4.297}, "bpb_mean": 2.39488, "frozen_inventory": true, "t": 1420.4}
{"event": "train", "step": 1600, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.50842, "bpb": 2.17619, "chunk_loss": 0.031069, "F": 0.224, "G": 0.24338, "chunks": 229.38, "bytes_per_chunk": 4.464, "boundary_entropy": 0.17242, "router_noise": 0.0, "grad_norm": 0.8392, "lr": 0.0002668, "router_trainable": true, "sec_per_step": 1.21, "probe_sec": 0.0957, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0696, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.06554, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 3, "mean_delta_merge": -0.07434, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -0.27962, "trunc_abs_frac": 0.25804, "mean_delta_full": -0.13354, "read_hit_rate": {"surface": 0.66393, "split": 0.88336, "merge": 0.21165, "skel": 0.67268}, "mean_gate": {"surface": 0.212, "split": 0.12844, "merge": 0.19005, "skel": 0.16135}, "n_reads": 21067, "n_cand": 32493, "read_suppressed_frac": 0.102, "dropout_reads_logged": 21067, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2378, "admitted_surface": 16730, "staged_surface": 0, "admitted_split": 16387, "staged_split": 0, "admitted_merge": 21438, "staged_merge": 0, "admitted_skel": 10981, "staged_skel": 0, "mean_count": 451.4832763671875, "n_util_obs": 247, "util_count_spearman": 0.08816259313174299}, "probe_overhead_sec_total": 154.481, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.00545, "alpha": 0.03, "t": 1422.5}
{"event": "consolidate", "step": 1650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1475.3}
{"event": "train", "step": 1650, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.5764, "bpb": 2.27426, "chunk_loss": 0.030844, "F": 0.21994, "G": 0.2348, "chunks": 225.22, "bytes_per_chunk": 4.547, "boundary_entropy": 0.16496, "router_noise": 0.0, "grad_norm": 0.8422, "lr": 0.0002643, "router_trainable": true, "sec_per_step": 1.071, "probe_sec": 0.1703, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.14042, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.03145, "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": 2, "trunc_ratio_sum": -3.94588, "trunc_abs_frac": 0.51437, "mean_delta_full": -0.00592, "read_hit_rate": {"surface": 0.65115, "split": 0.87269, "merge": 0.217, "skel": 0.65398}, "mean_gate": {"surface": 0.20761, "split": 0.12699, "merge": 0.17899, "skel": 0.16157}, "n_reads": 20728, "n_cand": 32330, "read_suppressed_frac": 0.0971, "dropout_reads_logged": 20728, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2424, "admitted_surface": 16805, "staged_surface": 0, "admitted_split": 16487, "staged_split": 0, "admitted_merge": 21201, "staged_merge": 0, "admitted_skel": 11043, "staged_skel": 0, "mean_count": 466.7941589355469, "n_util_obs": 253, "util_count_spearman": 0.08588540177884317}, "probe_overhead_sec_total": 161.267, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.005634, "alpha": 0.03, "t": 1476.1}
{"event": "consolidate", "step": 1700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1520.7}
{"event": "train", "step": 1700, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.53775, "bpb": 2.21851, "chunk_loss": 0.030643, "F": 0.20999, "G": 0.22536, "chunks": 215.03, "bytes_per_chunk": 4.762, "boundary_entropy": 0.16121, "router_noise": 0.0, "grad_norm": 0.7863, "lr": 0.0002618, "router_trainable": true, "sec_per_step": 0.915, "probe_sec": 0.1469, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.11592, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": 0.00154, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.03639, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.3504, "trunc_abs_frac": 10.50262, "mean_delta_full": -0.04248, "read_hit_rate": {"surface": 0.62885, "split": 0.85871, "merge": 0.19686, "skel": 0.62858}, "mean_gate": {"surface": 0.20328, "split": 0.12176, "merge": 0.17797, "skel": 0.15651}, "n_reads": 19147, "n_cand": 30797, "read_suppressed_frac": 0.0978, "dropout_reads_logged": 19147, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2470, "admitted_surface": 16841, "staged_surface": 0, "admitted_split": 16559, "staged_split": 0, "admitted_merge": 21044, "staged_merge": 0, "admitted_skel": 11092, "staged_skel": 0, "mean_count": 481.9320068359375, "n_util_obs": 259, "util_count_spearman": 0.08756813524705294}, "probe_overhead_sec_total": 166.465, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.005355, "alpha": 0.03, "t": 1521.9}
{"event": "consolidate", "step": 1750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1561.7}
{"event": "train", "step": 1750, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.56248, "bpb": 2.25418, "chunk_loss": 0.030752, "F": 0.21606, "G": 0.23114, "chunks": 221.25, "bytes_per_chunk": 4.628, "boundary_entropy": 0.1636, "router_noise": 0.0, "grad_norm": 0.7744, "lr": 0.0002591, "router_trainable": true, "sec_per_step": 0.81, "probe_sec": 0.0803, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05079, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": -0.05212, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.06433, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 1.08854, "trunc_abs_frac": 1.25115, "mean_delta_full": -0.24747, "read_hit_rate": {"surface": 0.63011, "split": 0.85983, "merge": 0.20453, "skel": 0.63433}, "mean_gate": {"surface": 0.20781, "split": 0.12071, "merge": 0.1803, "skel": 0.15639}, "n_reads": 19704, "n_cand": 31552, "read_suppressed_frac": 0.1009, "dropout_reads_logged": 19704, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2521, "admitted_surface": 16846, "staged_surface": 0, "admitted_split": 16611, "staged_split": 0, "admitted_merge": 21004, "staged_merge": 0, "admitted_skel": 11075, "staged_skel": 0, "mean_count": 497.0065612792969, "n_util_obs": 269, "util_count_spearman": 0.09527241353215658}, "probe_overhead_sec_total": 171.641, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.005564, "alpha": 0.03, "t": 1562.4}
{"event": "consolidate", "step": 1800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1602.0}
{"event": "probe", "step": 1800, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3621.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.208068, "cps": 28.289062, "mean_p": 0.226166, "boundary_entropy": 0.156585, "bpb": 2.121022}, "tr": {"chunks": 3978.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.219743, "cps": 31.078125, "mean_p": 0.209045, "boundary_entropy": 0.156185, "bpb": 2.359989}, "fi": {"chunks": 4762.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.25315, "cps": 37.203125, "mean_p": 0.25317, "boundary_entropy": 0.167608, "bpb": 2.510184}}, "bpb": {"en": 2.12102, "tr": 2.35999, "fi": 2.51018}, "gini_cpc": 0.044136, "gini_cps": 0.061538, "bpic": 4.589596, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0561116241724777, "fi": 1.216670850055695}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2587, "admitted_surface": 16785, "staged_surface": 0, "admitted_split": 16599, "staged_split": 0, "admitted_merge": 21041, "staged_merge": 0, "admitted_skel": 11111, "staged_skel": 0, "mean_count": 511.8961181640625, "n_util_obs": 276, "util_count_spearman": 0.0639317361339022}, "t": 1603.9}
{"event": "eval", "step": 1800, "phase": 1, "bpb": {"en": 2.13543, "tr": 2.37111, "fi": 2.51812}, "bytes_per_chunk": {"en": 4.526, "tr": 4.94, "fi": 4.089}, "bpb_mean": 2.34155, "frozen_inventory": true, "t": 1607.1}
{"event": "train", "step": 1800, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.53327, "bpb": 2.21203, "chunk_loss": 0.030955, "F": 0.22296, "G": 0.23616, "chunks": 228.31, "bytes_per_chunk": 4.485, "boundary_entropy": 0.156, "router_noise": 0.0, "grad_norm": 0.7456, "lr": 0.0002564, "router_trainable": true, "sec_per_step": 0.931, "probe_sec": 0.1024, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07336, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.04068, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 3, "mean_delta_merge": -0.02, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.27179, "trunc_abs_frac": 0.27717, "mean_delta_full": 0.33649, "read_hit_rate": {"surface": 0.65617, "split": 0.86854, "merge": 0.21734, "skel": 0.64775}, "mean_gate": {"surface": 0.20811, "split": 0.11979, "merge": 0.18098, "skel": 0.15546}, "n_reads": 20518, "n_cand": 32198, "read_suppressed_frac": 0.0992, "dropout_reads_logged": 20518, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2587, "admitted_surface": 16792, "staged_surface": 0, "admitted_split": 16561, "staged_split": 0, "admitted_merge": 21135, "staged_merge": 0, "admitted_skel": 11048, "staged_skel": 0, "mean_count": 512.205810546875, "n_util_obs": 276, "util_count_spearman": 0.06450565267414239}, "probe_overhead_sec_total": 176.486, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.005999, "alpha": 0.03, "t": 1608.9}
{"event": "consolidate", "step": 1850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1650.1}
{"event": "train", "step": 1850, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.54622, "bpb": 2.23072, "chunk_loss": 0.03086, "F": 0.21964, "G": 0.2321, "chunks": 224.91, "bytes_per_chunk": 4.553, "boundary_entropy": 0.14765, "router_noise": 0.0, "grad_norm": 0.7316, "lr": 0.0002537, "router_trainable": true, "sec_per_step": 0.838, "probe_sec": 0.0725, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0502, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.00574, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.04895, "n_merge": 1, "mean_delta_merge": 0.06787, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.89612, "trunc_abs_frac": 1.2625, "mean_delta_full": 0.01469, "read_hit_rate": {"surface": 0.64494, "split": 0.8694, "merge": 0.18982, "skel": 0.65553}, "mean_gate": {"surface": 0.2086, "split": 0.11749, "merge": 0.18272, "skel": 0.15559}, "n_reads": 20350, "n_cand": 32180, "read_suppressed_frac": 0.0986, "dropout_reads_logged": 20350, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2623, "admitted_surface": 16800, "staged_surface": 0, "admitted_split": 16518, "staged_split": 0, "admitted_merge": 21180, "staged_merge": 0, "admitted_skel": 11038, "staged_skel": 0, "mean_count": 527.4501342773438, "n_util_obs": 283, "util_count_spearman": 0.08226561413173683}, "probe_overhead_sec_total": 181.333, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.005725, "alpha": 0.03, "t": 1650.8}
{"event": "consolidate", "step": 1900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1690.8}
{"event": "train", "step": 1900, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.52908, "bpb": 2.206, "chunk_loss": 0.030796, "F": 0.21637, "G": 0.22819, "chunks": 221.56, "bytes_per_chunk": 4.622, "boundary_entropy": 0.1526, "router_noise": 0.0, "grad_norm": 0.8851, "lr": 0.0002508, "router_trainable": true, "sec_per_step": 0.814, "probe_sec": 0.1116, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08509, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": 0.02672, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.1199, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.86208, "trunc_abs_frac": 5.72596, "mean_delta_full": 0.15057, "read_hit_rate": {"surface": 0.64027, "split": 0.86284, "merge": 0.21065, "skel": 0.64471}, "mean_gate": {"surface": 0.20731, "split": 0.11463, "merge": 0.1774, "skel": 0.1536}, "n_reads": 19824, "n_cand": 31450, "read_suppressed_frac": 0.0992, "dropout_reads_logged": 19824, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2678, "admitted_surface": 16829, "staged_surface": 0, "admitted_split": 16629, "staged_split": 0, "admitted_merge": 21015, "staged_merge": 0, "admitted_skel": 11063, "staged_skel": 0, "mean_count": 542.7440185546875, "n_util_obs": 292, "util_count_spearman": 0.08195367789493876}, "probe_overhead_sec_total": 186.256, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.005818, "alpha": 0.03, "t": 1691.5}
{"event": "consolidate", "step": 1950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1730.6}
{"event": "train", "step": 1950, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.47756, "bpb": 2.13167, "chunk_loss": 0.030746, "F": 0.2142, "G": 0.22571, "chunks": 219.34, "bytes_per_chunk": 4.668, "boundary_entropy": 0.14325, "router_noise": 0.0, "grad_norm": 0.7348, "lr": 0.0002479, "router_trainable": true, "sec_per_step": 0.793, "probe_sec": 0.0969, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07866, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": -0.01204, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.03046, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.28674, "trunc_abs_frac": 1.56993, "mean_delta_full": -0.02969, "read_hit_rate": {"surface": 0.6525, "split": 0.8658, "merge": 0.21884, "skel": 0.64809}, "mean_gate": {"surface": 0.21017, "split": 0.11647, "merge": 0.17946, "skel": 0.15905}, "n_reads": 19780, "n_cand": 31082, "read_suppressed_frac": 0.1006, "dropout_reads_logged": 19780, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2719, "admitted_surface": 16803, "staged_surface": 0, "admitted_split": 16639, "staged_split": 0, "admitted_merge": 20988, "staged_merge": 0, "admitted_skel": 11106, "staged_skel": 0, "mean_count": 557.9273681640625, "n_util_obs": 298, "util_count_spearman": 0.07228285464754226}, "probe_overhead_sec_total": 191.09, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.006163, "alpha": 0.03, "t": 1731.2}
{"event": "consolidate", "step": 2000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1771.0}
{"event": "probe", "step": 2000, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3576.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.205482, "cps": 27.9375, "mean_p": 0.218964, "boundary_entropy": 0.14284, "bpb": 2.101312}, "tr": {"chunks": 3660.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.202176, "cps": 28.59375, "mean_p": 0.200395, "boundary_entropy": 0.163476, "bpb": 2.31054}, "fi": {"chunks": 4620.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.245601, "cps": 36.09375, "mean_p": 0.251488, "boundary_entropy": 0.164058, "bpb": 2.463361}}, "bpb": {"en": 2.10131, "tr": 2.31054, "fi": 2.46336}, "gini_cpc": 0.044316, "gini_cps": 0.058704, "bpic": 4.785088, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9839140088390556, "fi": 1.1952443575959946}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2771, "admitted_surface": 16842, "staged_surface": 0, "admitted_split": 16557, "staged_split": 0, "admitted_merge": 21013, "staged_merge": 0, "admitted_skel": 11124, "staged_skel": 0, "mean_count": 572.7587890625, "n_util_obs": 301, "util_count_spearman": 0.08342698730500979}, "t": 1772.8}
{"event": "eval", "step": 2000, "phase": 1, "bpb": {"en": 2.11146, "tr": 2.32356, "fi": 2.47397}, "bytes_per_chunk": {"en": 4.62, "tr": 5.356, "fi": 4.168}, "bpb_mean": 2.30299, "frozen_inventory": true, "t": 1775.8}
{"event": "train", "step": 2000, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.63987, "bpb": 2.36583, "chunk_loss": 0.030975, "F": 0.22272, "G": 0.23587, "chunks": 228.06, "bytes_per_chunk": 4.49, "boundary_entropy": 0.1495, "router_noise": 0.0, "grad_norm": 0.8007, "lr": 0.000245, "router_trainable": true, "sec_per_step": 0.929, "probe_sec": 0.1932, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.16805, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.02924, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 3, "mean_delta_split": 0.12912, "n_merge": 1, "mean_delta_merge": -0.24646, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.51769, "trunc_abs_frac": 0.47854, "mean_delta_full": -0.28113, "read_hit_rate": {"surface": 0.62497, "split": 0.86003, "merge": 0.18607, "skel": 0.62888}, "mean_gate": {"surface": 0.20773, "split": 0.11179, "merge": 0.18089, "skel": 0.1532}, "n_reads": 19839, "n_cand": 32169, "read_suppressed_frac": 0.1034, "dropout_reads_logged": 19839, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2773, "admitted_surface": 16846, "staged_surface": 0, "admitted_split": 16511, "staged_split": 0, "admitted_merge": 21096, "staged_merge": 0, "admitted_skel": 11083, "staged_skel": 0, "mean_count": 573.0409545898438, "n_util_obs": 301, "util_count_spearman": 0.08346042991353325}, "probe_overhead_sec_total": 196.22, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.006495, "alpha": 0.03, "t": 1777.6}
{"event": "probe", "step": 2048, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3591.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.206344, "cps": 28.054688, "mean_p": 0.218223, "boundary_entropy": 0.135224, "bpb": 2.084534}, "tr": {"chunks": 3803.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.210076, "cps": 29.710938, "mean_p": 0.205292, "boundary_entropy": 0.148119, "bpb": 2.304532}, "fi": {"chunks": 4632.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.246239, "cps": 36.1875, "mean_p": 0.249116, "boundary_entropy": 0.15543, "bpb": 2.455973}}, "bpb": {"en": 2.08453, "tr": 2.30453, "fi": 2.45597}, "gini_cpc": 0.040136, "gini_cps": 0.057708, "bpic": 4.717446, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0180860555151527, "fi": 1.193343253940497}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2828, "admitted_surface": 16847, "staged_surface": 0, "admitted_split": 16523, "staged_split": 0, "admitted_merge": 21092, "staged_merge": 0, "admitted_skel": 11074, "staged_skel": 0, "mean_count": 587.4480590820312, "n_util_obs": 310, "util_count_spearman": 0.09057757787636077}, "t": 1821.0}
{"event": "consolidate", "step": 2050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1824.5}
{"event": "train", "step": 2050, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.47693, "bpb": 2.13075, "chunk_loss": 0.030655, "F": 0.21124, "G": 0.21994, "chunks": 216.31, "bytes_per_chunk": 4.734, "boundary_entropy": 0.13732, "router_noise": 0.0, "grad_norm": 0.7892, "lr": 0.0002419, "router_trainable": true, "sec_per_step": 0.948, "probe_sec": 0.0792, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05912, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.00314, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": -0.10638, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.05006, "trunc_abs_frac": 0.99731, "mean_delta_full": -0.09058, "read_hit_rate": {"surface": 0.63803, "split": 0.86126, "merge": 0.20283, "skel": 0.63569}, "mean_gate": {"surface": 0.20425, "split": 0.1102, "merge": 0.1722, "skel": 0.1533}, "n_reads": 19393, "n_cand": 30924, "read_suppressed_frac": 0.0985, "dropout_reads_logged": 19393, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2828, "admitted_surface": 16831, "staged_surface": 0, "admitted_split": 16593, "staged_split": 0, "admitted_merge": 20972, "staged_merge": 0, "admitted_skel": 11140, "staged_skel": 0, "mean_count": 588.3624877929688, "n_util_obs": 310, "util_count_spearman": 0.0911090771421288}, "probe_overhead_sec_total": 201.788, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.005826, "alpha": 0.03, "t": 1825.0}
{"event": "consolidate", "step": 2100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1867.0}
{"event": "train", "step": 2100, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.52994, "bpb": 2.20724, "chunk_loss": 0.030684, "F": 0.21512, "G": 0.22511, "chunks": 220.28, "bytes_per_chunk": 4.649, "boundary_entropy": 0.13837, "router_noise": 0.0, "grad_norm": 0.7444, "lr": 0.0002388, "router_trainable": true, "sec_per_step": 0.861, "probe_sec": 0.0929, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0729, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.19224, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": -0.02643, "n_merge": 1, "mean_delta_merge": -0.51775, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.64957, "trunc_abs_frac": 0.32124, "mean_delta_full": -0.22278, "read_hit_rate": {"surface": 0.63389, "split": 0.85977, "merge": 0.19785, "skel": 0.64382}, "mean_gate": {"surface": 0.20744, "split": 0.10831, "merge": 0.17739, "skel": 0.15289}, "n_reads": 19630, "n_cand": 31394, "read_suppressed_frac": 0.1002, "dropout_reads_logged": 19630, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2879, "admitted_surface": 16838, "staged_surface": 0, "admitted_split": 16588, "staged_split": 0, "admitted_merge": 21023, "staged_merge": 0, "admitted_skel": 11087, "staged_skel": 0, "mean_count": 603.5552368164062, "n_util_obs": 318, "util_count_spearman": 0.09269920286667958}, "probe_overhead_sec_total": 206.782, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.006371, "alpha": 0.03, "t": 1868.1}
{"event": "consolidate", "step": 2150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1927.5}
{"event": "train", "step": 2150, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.48887, "bpb": 2.14799, "chunk_loss": 0.030773, "F": 0.21948, "G": 0.22777, "chunks": 224.75, "bytes_per_chunk": 4.556, "boundary_entropy": 0.13307, "router_noise": 0.0, "grad_norm": 0.7215, "lr": 0.0002357, "router_trainable": true, "sec_per_step": 1.23, "probe_sec": 0.1333, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08504, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.08668, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.03418, "n_merge": 3, "mean_delta_merge": 0.03805, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.59991, "trunc_abs_frac": 1.00442, "mean_delta_full": -0.06812, "read_hit_rate": {"surface": 0.65768, "split": 0.87389, "merge": 0.22713, "skel": 0.66246}, "mean_gate": {"surface": 0.20255, "split": 0.10609, "merge": 0.17485, "skel": 0.15074}, "n_reads": 20710, "n_cand": 32050, "read_suppressed_frac": 0.1026, "dropout_reads_logged": 20710, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2925, "admitted_surface": 16819, "staged_surface": 0, "admitted_split": 16562, "staged_split": 0, "admitted_merge": 21090, "staged_merge": 0, "admitted_skel": 11065, "staged_skel": 0, "mean_count": 618.9234619140625, "n_util_obs": 330, "util_count_spearman": 0.07711616017844725}, "probe_overhead_sec_total": 213.501, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.006168, "alpha": 0.03, "t": 1929.6}
{"event": "consolidate", "step": 2200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1986.8}
{"event": "probe", "step": 2200, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3533.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.203011, "cps": 27.601562, "mean_p": 0.213945, "boundary_entropy": 0.137803, "bpb": 2.071872}, "tr": {"chunks": 3781.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.20886, "cps": 29.539062, "mean_p": 0.207058, "boundary_entropy": 0.164161, "bpb": 2.282219}, "fi": {"chunks": 4559.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.242358, "cps": 35.617188, "mean_p": 0.244046, "boundary_entropy": 0.152758, "bpb": 2.427605}}, "bpb": {"en": 2.07187, "tr": 2.28222, "fi": 2.42761}, "gini_cpc": 0.040095, "gini_cps": 0.05761, "bpic": 4.778236, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.028813391759476, "fi": 1.1938181890461228}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2971, "admitted_surface": 16843, "staged_surface": 0, "admitted_split": 16574, "staged_split": 0, "admitted_merge": 20966, "staged_merge": 0, "admitted_skel": 11153, "staged_skel": 0, "mean_count": 633.73193359375, "n_util_obs": 339, "util_count_spearman": 0.0700741524036972}, "t": 1988.6}
{"event": "eval", "step": 2200, "phase": 1, "bpb": {"en": 2.0823, "tr": 2.29313, "fi": 2.4366}, "bytes_per_chunk": {"en": 4.644, "tr": 5.1, "fi": 4.269}, "bpb_mean": 2.27068, "frozen_inventory": true, "t": 1991.7}
{"event": "train", "step": 2200, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.46935, "bpb": 2.11982, "chunk_loss": 0.030625, "F": 0.21423, "G": 0.22264, "chunks": 219.38, "bytes_per_chunk": 4.668, "boundary_entropy": 0.13683, "router_noise": 0.0, "grad_norm": 0.6659, "lr": 0.0002325, "router_trainable": true, "sec_per_step": 1.289, "probe_sec": 0.1253, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10032, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.02249, "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": 0.35564, "trunc_abs_frac": 0.77229, "mean_delta_full": 0.05123, "read_hit_rate": {"surface": 0.65469, "split": 0.87189, "merge": 0.21406, "skel": 0.64677}, "mean_gate": {"surface": 0.20593, "split": 0.10664, "merge": 0.17615, "skel": 0.15281}, "n_reads": 20050, "n_cand": 31365, "read_suppressed_frac": 0.0998, "dropout_reads_logged": 20050, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2971, "admitted_surface": 16802, "staged_surface": 0, "admitted_split": 16548, "staged_split": 0, "admitted_merge": 21036, "staged_merge": 0, "admitted_skel": 11150, "staged_skel": 0, "mean_count": 634.0445556640625, "n_util_obs": 339, "util_count_spearman": 0.06995279054498796}, "probe_overhead_sec_total": 219.457, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.006931, "alpha": 0.03, "t": 1994.0}
{"event": "consolidate", "step": 2250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2035.8}
{"event": "train", "step": 2250, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.49538, "bpb": 2.15738, "chunk_loss": 0.030768, "F": 0.21725, "G": 0.22633, "chunks": 222.47, "bytes_per_chunk": 4.603, "boundary_entropy": 0.13001, "router_noise": 0.0, "grad_norm": 0.7368, "lr": 0.0002293, "router_trainable": true, "sec_per_step": 0.851, "probe_sec": 0.0702, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04996, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.04326, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": 0.47908, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.64816, "trunc_abs_frac": 0.68449, "mean_delta_full": 0.56954, "read_hit_rate": {"surface": 0.64654, "split": 0.86657, "merge": 0.20425, "skel": 0.6437}, "mean_gate": {"surface": 0.20201, "split": 0.10307, "merge": 0.17218, "skel": 0.14794}, "n_reads": 19962, "n_cand": 31616, "read_suppressed_frac": 0.0973, "dropout_reads_logged": 19962, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3023, "admitted_surface": 16796, "staged_surface": 0, "admitted_split": 16573, "staged_split": 0, "admitted_merge": 21052, "staged_merge": 0, "admitted_skel": 11115, "staged_skel": 0, "mean_count": 649.183837890625, "n_util_obs": 349, "util_count_spearman": 0.07750702217454515}, "probe_overhead_sec_total": 224.508, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.005917, "alpha": 0.03, "t": 2036.6}
{"event": "consolidate", "step": 2300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2074.6}
{"event": "train", "step": 2300, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.50384, "bpb": 2.16958, "chunk_loss": 0.030729, "F": 0.2179, "G": 0.22647, "chunks": 223.12, "bytes_per_chunk": 4.589, "boundary_entropy": 0.13158, "router_noise": 0.0, "grad_norm": 0.7086, "lr": 0.000226, "router_trainable": true, "sec_per_step": 0.781, "probe_sec": 0.0698, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05023, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.24213, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.08606, "n_merge": 2, "mean_delta_merge": 0.01178, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.40334, "trunc_abs_frac": 0.52363, "mean_delta_full": 0.38409, "read_hit_rate": {"surface": 0.64181, "split": 0.87047, "merge": 0.20167, "skel": 0.64367}, "mean_gate": {"surface": 0.20006, "split": 0.10127, "merge": 0.1699, "skel": 0.14689}, "n_reads": 20108, "n_cand": 31805, "read_suppressed_frac": 0.097, "dropout_reads_logged": 20108, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3072, "admitted_surface": 16821, "staged_surface": 0, "admitted_split": 16576, "staged_split": 0, "admitted_merge": 21037, "staged_merge": 0, "admitted_skel": 11102, "staged_skel": 0, "mean_count": 664.39111328125, "n_util_obs": 353, "util_count_spearman": 0.09586715015004244}, "probe_overhead_sec_total": 229.254, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.006582, "alpha": 0.03, "t": 2075.6}
{"event": "consolidate", "step": 2350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2113.8}
{"event": "train", "step": 2350, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.55421, "bpb": 2.24226, "chunk_loss": 0.03067, "F": 0.21518, "G": 0.22482, "chunks": 220.34, "bytes_per_chunk": 4.647, "boundary_entropy": 0.13204, "router_noise": 0.0, "grad_norm": 0.7272, "lr": 0.0002226, "router_trainable": true, "sec_per_step": 0.784, "probe_sec": 0.4502, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06141, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": 0.05859, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.18811, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 1.13555, "trunc_abs_frac": 0.45612, "mean_delta_full": 0.29785, "read_hit_rate": {"surface": 0.633, "split": 0.86432, "merge": 0.19064, "skel": 0.63747}, "mean_gate": {"surface": 0.19883, "split": 0.09812, "merge": 0.16869, "skel": 0.145}, "n_reads": 19782, "n_cand": 31638, "read_suppressed_frac": 0.1011, "dropout_reads_logged": 19782, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3119, "admitted_surface": 16822, "staged_surface": 0, "admitted_split": 16616, "staged_split": 0, "admitted_merge": 20974, "staged_merge": 0, "admitted_skel": 11124, "staged_skel": 0, "mean_count": 679.7029418945312, "n_util_obs": 362, "util_count_spearman": 0.11315500462228606}, "probe_overhead_sec_total": 234.367, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.006582, "alpha": 0.03, "t": 2114.8}
{"event": "consolidate", "step": 2400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2153.0}
{"event": "probe", "step": 2400, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3592.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.206401, "cps": 28.0625, "mean_p": 0.214722, "boundary_entropy": 0.122423, "bpb": 2.047891}, "tr": {"chunks": 3613.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.19958, "cps": 28.226562, "mean_p": 0.197155, "boundary_entropy": 0.146753, "bpb": 2.250585}, "fi": {"chunks": 4619.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.245548, "cps": 36.085938, "mean_p": 0.245247, "boundary_entropy": 0.134903, "bpb": 2.370309}}, "bpb": {"en": 2.04789, "tr": 2.25058, "fi": 2.37031}, "gini_cpc": 0.047036, "gini_cps": 0.057905, "bpic": 4.798038, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9669526375121228, "fi": 1.1896627707642509}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3160, "admitted_surface": 16821, "staged_surface": 0, "admitted_split": 16709, "staged_split": 0, "admitted_merge": 20851, "staged_merge": 0, "admitted_skel": 11155, "staged_skel": 0, "mean_count": 694.5975952148438, "n_util_obs": 375, "util_count_spearman": 0.12418444215408157}, "t": 2154.7}
{"event": "eval", "step": 2400, "phase": 1, "bpb": {"en": 2.05927, "tr": 2.25874, "fi": 2.38141}, "bytes_per_chunk": {"en": 4.59, "tr": 5.421, "fi": 4.281}, "bpb_mean": 2.23314, "frozen_inventory": true, "t": 2158.0}
{"event": "phase_transition", "step": 2400, "phase": 2, "objective": "global", "router_trainable": true, "encoder_trainable": true, "frozen_router_params": 0, "t": 2159.3}
{"event": "train", "step": 2400, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.49131, "bpb": 2.15151, "chunk_loss": 0.030797, "F": 0.21954, "G": 0.2278, "chunks": 224.81, "bytes_per_chunk": 4.555, "boundary_entropy": 0.12281, "router_noise": 0.0, "grad_norm": 0.733, "lr": 0.0002192, "router_trainable": true, "sec_per_step": 0.9, "probe_sec": 0.0928, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07218, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.06029, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.05109, "n_merge": 1, "mean_delta_merge": -0.00622, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": -0.04912, "trunc_abs_frac": 0.04912, "mean_delta_full": 0.12659, "read_hit_rate": {"surface": 0.64298, "split": 0.87428, "merge": 0.20281, "skel": 0.67295}, "mean_gate": {"surface": 0.20013, "split": 0.09843, "merge": 0.17264, "skel": 0.14487}, "n_reads": 20730, "n_cand": 32364, "read_suppressed_frac": 0.0966, "dropout_reads_logged": 20730, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3160, "admitted_surface": 16853, "staged_surface": 0, "admitted_split": 16620, "staged_split": 0, "admitted_merge": 20987, "staged_merge": 0, "admitted_skel": 11076, "staged_skel": 0, "mean_count": 694.920654296875, "n_util_obs": 375, "util_count_spearman": 0.12454125890786961}, "probe_overhead_sec_total": 239.67, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.006569, "alpha": 0.03, "t": 2159.8}
{"event": "consolidate", "step": 2450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2199.6}
{"event": "train", "step": 2450, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.46213, "bpb": 2.10941, "chunk_loss": 0.030614, "F": 0.21008, "G": 0.22101, "chunks": 215.12, "bytes_per_chunk": 4.76, "boundary_entropy": 0.13219, "router_noise": 0.0, "grad_norm": 0.7209, "lr": 0.0002158, "router_trainable": true, "sec_per_step": 0.809, "probe_sec": 0.0891, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06283, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.02331, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": -0.21376, "n_merge": 1, "mean_delta_merge": 0.13708, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.40583, "trunc_abs_frac": 0.38621, "mean_delta_full": 0.44051, "read_hit_rate": {"surface": 0.64113, "split": 0.86347, "merge": 0.20015, "skel": 0.65056}, "mean_gate": {"surface": 0.2016, "split": 0.09927, "merge": 0.17552, "skel": 0.1487}, "n_reads": 19649, "n_cand": 31086, "read_suppressed_frac": 0.1043, "dropout_reads_logged": 19649, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3205, "admitted_surface": 16818, "staged_surface": 0, "admitted_split": 16719, "staged_split": 0, "admitted_merge": 20895, "staged_merge": 0, "admitted_skel": 11104, "staged_skel": 0, "mean_count": 710.1165771484375, "n_util_obs": 378, "util_count_spearman": 0.11377119151273589}, "probe_overhead_sec_total": 244.884, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.005923, "alpha": 0.03, "t": 2200.3}
{"event": "consolidate", "step": 2500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2238.0}
{"event": "train", "step": 2500, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.43117, "bpb": 2.06475, "chunk_loss": 0.030739, "F": 0.21979, "G": 0.2263, "chunks": 225.06, "bytes_per_chunk": 4.55, "boundary_entropy": 0.11814, "router_noise": 0.0, "grad_norm": 0.7509, "lr": 0.0002123, "router_trainable": true, "sec_per_step": 0.767, "probe_sec": 0.0693, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04911, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.02715, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": 0.25577, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -1.55452, "trunc_abs_frac": 0.85217, "mean_delta_full": 0.02127, "read_hit_rate": {"surface": 0.65676, "split": 0.87373, "merge": 0.22163, "skel": 0.65956}, "mean_gate": {"surface": 0.20108, "split": 0.09647, "merge": 0.17168, "skel": 0.14601}, "n_reads": 20781, "n_cand": 32237, "read_suppressed_frac": 0.0975, "dropout_reads_logged": 20781, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3246, "admitted_surface": 16825, "staged_surface": 0, "admitted_split": 16630, "staged_split": 0, "admitted_merge": 20955, "staged_merge": 0, "admitted_skel": 11126, "staged_skel": 0, "mean_count": 725.27392578125, "n_util_obs": 385, "util_count_spearman": 0.11294475225342103}, "probe_overhead_sec_total": 249.847, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.006299, "alpha": 0.03, "t": 2238.6}
{"event": "consolidate", "step": 2550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2277.3}
{"event": "train", "step": 2550, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.45898, "bpb": 2.10486, "chunk_loss": 0.030699, "F": 0.21725, "G": 0.22365, "chunks": 222.47, "bytes_per_chunk": 4.603, "boundary_entropy": 0.1162, "router_noise": 0.0, "grad_norm": 0.6878, "lr": 0.0002088, "router_trainable": true, "sec_per_step": 0.785, "probe_sec": 0.0781, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05435, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": 0.07095, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 4, "mean_delta_split": 0.0409, "n_merge": 1, "mean_delta_merge": 0.07391, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": -0.19249, "trunc_abs_frac": 0.19249, "mean_delta_full": 0.31812, "read_hit_rate": {"surface": 0.64964, "split": 0.86668, "merge": 0.21488, "skel": 0.66942}, "mean_gate": {"surface": 0.20252, "split": 0.09837, "merge": 0.17158, "skel": 0.14975}, "n_reads": 20505, "n_cand": 32008, "read_suppressed_frac": 0.0998, "dropout_reads_logged": 20505, "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": 16820, "staged_surface": 0, "admitted_split": 16702, "staged_split": 0, "admitted_merge": 20928, "staged_merge": 0, "admitted_skel": 11086, "staged_skel": 0, "mean_count": 740.482177734375, "n_util_obs": 393, "util_count_spearman": 0.10627130772454091}, "probe_overhead_sec_total": 254.921, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.005573, "alpha": 0.03, "t": 2277.9}
{"event": "consolidate", "step": 2600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2318.3}
{"event": "probe", "step": 2600, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3600.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.206861, "cps": 28.125, "mean_p": 0.214239, "boundary_entropy": 0.115386, "bpb": 2.026202}, "tr": {"chunks": 3679.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.203226, "cps": 28.742188, "mean_p": 0.19851, "boundary_entropy": 0.133043, "bpb": 2.236044}, "fi": {"chunks": 4637.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.246505, "cps": 36.226562, "mean_p": 0.243058, "boundary_entropy": 0.126633, "bpb": 2.353686}}, "bpb": {"en": 2.0262, "tr": 2.23604, "fi": 2.35369}, "gini_cpc": 0.043943, "gini_cps": 0.058017, "bpic": 4.760994, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9824282807637775, "fi": 1.1916448266085446}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3341, "admitted_surface": 16833, "staged_surface": 0, "admitted_split": 16721, "staged_split": 0, "admitted_merge": 20890, "staged_merge": 0, "admitted_skel": 11092, "staged_skel": 0, "mean_count": 755.4317626953125, "n_util_obs": 398, "util_count_spearman": 0.10281241845031519}, "t": 2320.3}
{"event": "eval", "step": 2600, "phase": 2, "bpb": {"en": 2.03906, "tr": 2.24898, "fi": 2.36231}, "bytes_per_chunk": {"en": 4.573, "tr": 5.379, "fi": 4.243}, "bpb_mean": 2.21679, "frozen_inventory": true, "t": 2323.2}
{"event": "train", "step": 2600, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.41507, "bpb": 2.04152, "chunk_loss": 0.030648, "F": 0.21545, "G": 0.22272, "chunks": 220.62, "bytes_per_chunk": 4.641, "boundary_entropy": 0.11509, "router_noise": 0.0, "grad_norm": 0.7019, "lr": 0.0002053, "router_trainable": true, "sec_per_step": 0.943, "probe_sec": 0.1021, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07602, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.05561, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.01807, "n_merge": 1, "mean_delta_merge": 0.32666, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.21889, "trunc_abs_frac": 0.24873, "mean_delta_full": -0.46341, "read_hit_rate": {"surface": 0.65452, "split": 0.8749, "merge": 0.21741, "skel": 0.64903}, "mean_gate": {"surface": 0.19993, "split": 0.09709, "merge": 0.16991, "skel": 0.14516}, "n_reads": 20115, "n_cand": 31385, "read_suppressed_frac": 0.1041, "dropout_reads_logged": 20115, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3342, "admitted_surface": 16819, "staged_surface": 0, "admitted_split": 16634, "staged_split": 0, "admitted_merge": 20965, "staged_merge": 0, "admitted_skel": 11118, "staged_skel": 0, "mean_count": 755.739013671875, "n_util_obs": 399, "util_count_spearman": 0.09751237850494865}, "probe_overhead_sec_total": 259.934, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.005877, "alpha": 0.03, "t": 2325.1}
{"event": "consolidate", "step": 2650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2366.3}
{"event": "train", "step": 2650, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.45701, "bpb": 2.10202, "chunk_loss": 0.030507, "F": 0.20737, "G": 0.21371, "chunks": 212.34, "bytes_per_chunk": 4.822, "boundary_entropy": 0.11703, "router_noise": 0.0, "grad_norm": 0.681, "lr": 0.0002017, "router_trainable": true, "sec_per_step": 0.852, "probe_sec": 0.0918, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0571, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.02771, "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": 2, "trunc_ratio_sum": -2.00257, "trunc_abs_frac": 2.21141, "mean_delta_full": 0.02374, "read_hit_rate": {"surface": 0.62591, "split": 0.85086, "merge": 0.18898, "skel": 0.62257}, "mean_gate": {"surface": 0.20672, "split": 0.09849, "merge": 0.17686, "skel": 0.15213}, "n_reads": 18766, "n_cand": 30499, "read_suppressed_frac": 0.1023, "dropout_reads_logged": 18766, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3390, "admitted_surface": 16813, "staged_surface": 0, "admitted_split": 16764, "staged_split": 0, "admitted_merge": 20771, "staged_merge": 0, "admitted_skel": 11188, "staged_skel": 0, "mean_count": 770.7705078125, "n_util_obs": 410, "util_count_spearman": 0.08839704550385155}, "probe_overhead_sec_total": 265.204, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.0059, "alpha": 0.03, "t": 2367.7}
{"event": "consolidate", "step": 2700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2411.0}
{"event": "train", "step": 2700, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.44071, "bpb": 2.07851, "chunk_loss": 0.030724, "F": 0.21924, "G": 0.22457, "chunks": 224.5, "bytes_per_chunk": 4.561, "boundary_entropy": 0.11643, "router_noise": 0.0, "grad_norm": 0.6777, "lr": 0.0001981, "router_trainable": true, "sec_per_step": 0.88, "probe_sec": 0.1367, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.11557, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.00853, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.06592, "n_merge": 1, "mean_delta_merge": -0.04456, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 1.48927, "trunc_abs_frac": 4.11806, "mean_delta_full": -0.07397, "read_hit_rate": {"surface": 0.65227, "split": 0.86946, "merge": 0.21573, "skel": 0.66325}, "mean_gate": {"surface": 0.20064, "split": 0.09425, "merge": 0.17087, "skel": 0.14618}, "n_reads": 20504, "n_cand": 32019, "read_suppressed_frac": 0.0985, "dropout_reads_logged": 20504, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3434, "admitted_surface": 16809, "staged_surface": 0, "admitted_split": 16657, "staged_split": 0, "admitted_merge": 20979, "staged_merge": 0, "admitted_skel": 11091, "staged_skel": 0, "mean_count": 786.0224609375, "n_util_obs": 416, "util_count_spearman": 0.08418983772859906}, "probe_overhead_sec_total": 270.559, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.006278, "alpha": 0.03, "t": 2411.7}
{"event": "consolidate", "step": 2750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2469.2}
{"event": "train", "step": 2750, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.45161, "bpb": 2.09423, "chunk_loss": 0.030739, "F": 0.21741, "G": 0.22387, "chunks": 222.62, "bytes_per_chunk": 4.6, "boundary_entropy": 0.11362, "router_noise": 0.0, "grad_norm": 0.6787, "lr": 0.0001945, "router_trainable": true, "sec_per_step": 1.174, "probe_sec": 0.0806, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05564, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.04071, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": 0.19852, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.38991, "trunc_abs_frac": 0.45421, "mean_delta_full": -0.17987, "read_hit_rate": {"surface": 0.65877, "split": 0.87152, "merge": 0.2211, "skel": 0.65189}, "mean_gate": {"surface": 0.20078, "split": 0.09656, "merge": 0.17054, "skel": 0.14876}, "n_reads": 20317, "n_cand": 31648, "read_suppressed_frac": 0.0985, "dropout_reads_logged": 20317, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3472, "admitted_surface": 16802, "staged_surface": 0, "admitted_split": 16662, "staged_split": 0, "admitted_merge": 20951, "staged_merge": 0, "admitted_skel": 11121, "staged_skel": 0, "mean_count": 801.238037109375, "n_util_obs": 425, "util_count_spearman": 0.08063888324396992}, "probe_overhead_sec_total": 276.94, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.00573, "alpha": 0.03, "t": 2470.4}
{"event": "consolidate", "step": 2800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2526.9}
{"event": "probe", "step": 2800, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3545.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.203701, "cps": 27.695312, "mean_p": 0.212201, "boundary_entropy": 0.113382, "bpb": 2.008498}, "tr": {"chunks": 3598.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.198752, "cps": 28.109375, "mean_p": 0.194099, "boundary_entropy": 0.127736, "bpb": 2.193901}, "fi": {"chunks": 4590.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.244006, "cps": 35.859375, "mean_p": 0.241611, "boundary_entropy": 0.119323, "bpb": 2.312847}}, "bpb": {"en": 2.0085, "tr": 2.1939, "fi": 2.31285}, "gini_cpc": 0.046669, "gini_cps": 0.059377, "bpic": 4.835251, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9757049050227943, "fi": 1.197867226352795}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3513, "admitted_surface": 16858, "staged_surface": 0, "admitted_split": 16791, "staged_split": 0, "admitted_merge": 20799, "staged_merge": 0, "admitted_skel": 11088, "staged_skel": 0, "mean_count": 816.0219116210938, "n_util_obs": 430, "util_count_spearman": 0.08677826247566711}, "t": 2528.8}
{"event": "eval", "step": 2800, "phase": 2, "bpb": {"en": 2.02068, "tr": 2.20392, "fi": 2.32009}, "bytes_per_chunk": {"en": 4.596, "tr": 5.518, "fi": 4.299}, "bpb_mean": 2.18156, "frozen_inventory": true, "t": 2532.1}
{"event": "train", "step": 2800, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.40243, "bpb": 2.02328, "chunk_loss": 0.030639, "F": 0.21512, "G": 0.22118, "chunks": 220.28, "bytes_per_chunk": 4.649, "boundary_entropy": 0.11238, "router_noise": 0.0, "grad_norm": 0.6883, "lr": 0.0001909, "router_trainable": true, "sec_per_step": 1.295, "probe_sec": 0.1139, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07684, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.01724, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 3, "mean_delta_merge": -0.05182, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 2.43443, "trunc_abs_frac": 2.78578, "mean_delta_full": -0.01926, "read_hit_rate": {"surface": 0.65284, "split": 0.87415, "merge": 0.2146, "skel": 0.64444}, "mean_gate": {"surface": 0.19606, "split": 0.09332, "merge": 0.1662, "skel": 0.14466}, "n_reads": 20124, "n_cand": 31475, "read_suppressed_frac": 0.095, "dropout_reads_logged": 20124, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3514, "admitted_surface": 16806, "staged_surface": 0, "admitted_split": 16667, "staged_split": 0, "admitted_merge": 20920, "staged_merge": 0, "admitted_skel": 11143, "staged_skel": 0, "mean_count": 816.336669921875, "n_util_obs": 430, "util_count_spearman": 0.08732909146973097}, "probe_overhead_sec_total": 282.956, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.005879, "alpha": 0.03, "t": 2535.1}
{"event": "consolidate", "step": 2850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2577.2}
{"event": "train", "step": 2850, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.45514, "bpb": 2.09932, "chunk_loss": 0.030884, "F": 0.22455, "G": 0.23139, "chunks": 229.94, "bytes_per_chunk": 4.453, "boundary_entropy": 0.11537, "router_noise": 0.0, "grad_norm": 0.7198, "lr": 0.0001872, "router_trainable": true, "sec_per_step": 0.866, "probe_sec": 0.0906, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0559, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.0797, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": 0.13574, "n_merge": 1, "mean_delta_merge": 0.10522, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 1.67779, "trunc_abs_frac": 0.48166, "mean_delta_full": -0.07255, "read_hit_rate": {"surface": 0.65384, "split": 0.87607, "merge": 0.21483, "skel": 0.67428}, "mean_gate": {"surface": 0.19178, "split": 0.09118, "merge": 0.16283, "skel": 0.14142}, "n_reads": 21163, "n_cand": 32811, "read_suppressed_frac": 0.0986, "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": 3562, "admitted_surface": 16825, "staged_surface": 0, "admitted_split": 16615, "staged_split": 0, "admitted_merge": 21051, "staged_merge": 0, "admitted_skel": 11045, "staged_skel": 0, "mean_count": 831.5289916992188, "n_util_obs": 436, "util_count_spearman": 0.09511567593010632}, "probe_overhead_sec_total": 288.419, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.00629, "alpha": 0.03, "t": 2578.4}
{"event": "consolidate", "step": 2900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2616.9}
{"event": "train", "step": 2900, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.38227, "bpb": 1.9942, "chunk_loss": 0.030491, "F": 0.20807, "G": 0.21326, "chunks": 213.06, "bytes_per_chunk": 4.806, "boundary_entropy": 0.11085, "router_noise": 0.0, "grad_norm": 0.6492, "lr": 0.0001835, "router_trainable": true, "sec_per_step": 0.783, "probe_sec": 0.088, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05842, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.05538, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": 0.10602, "n_merge": 2, "mean_delta_merge": 0.11957, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -10.80966, "trunc_abs_frac": 2.35434, "mean_delta_full": 0.00201, "read_hit_rate": {"surface": 0.65119, "split": 0.86308, "merge": 0.21617, "skel": 0.64385}, "mean_gate": {"surface": 0.19558, "split": 0.09214, "merge": 0.16569, "skel": 0.14451}, "n_reads": 19335, "n_cand": 30448, "read_suppressed_frac": 0.0957, "dropout_reads_logged": 19335, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3605, "admitted_surface": 16788, "staged_surface": 0, "admitted_split": 16760, "staged_split": 0, "admitted_merge": 20820, "staged_merge": 0, "admitted_skel": 11168, "staged_skel": 0, "mean_count": 846.7112426757812, "n_util_obs": 445, "util_count_spearman": 0.08255547387211315}, "probe_overhead_sec_total": 293.231, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.006341, "alpha": 0.03, "t": 2617.5}
{"event": "consolidate", "step": 2950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2657.0}
{"event": "train", "step": 2950, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.53125, "bpb": 2.20913, "chunk_loss": 0.0308, "F": 0.21539, "G": 0.22227, "chunks": 220.56, "bytes_per_chunk": 4.643, "boundary_entropy": 0.11573, "router_noise": 0.0, "grad_norm": 0.6726, "lr": 0.0001799, "router_trainable": true, "sec_per_step": 0.803, "probe_sec": 0.1559, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.13203, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.02908, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": 0.04301, "n_merge": 2, "mean_delta_merge": 0.18695, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.56764, "trunc_abs_frac": 0.56764, "mean_delta_full": 0.51074, "read_hit_rate": {"surface": 0.62511, "split": 0.85455, "merge": 0.18587, "skel": 0.63953}, "mean_gate": {"surface": 0.19019, "split": 0.08948, "merge": 0.16193, "skel": 0.13983}, "n_reads": 19426, "n_cand": 31427, "read_suppressed_frac": 0.1001, "dropout_reads_logged": 19426, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3644, "admitted_surface": 16827, "staged_surface": 0, "admitted_split": 16738, "staged_split": 0, "admitted_merge": 20886, "staged_merge": 0, "admitted_skel": 11085, "staged_skel": 0, "mean_count": 861.7672729492188, "n_util_obs": 452, "util_count_spearman": 0.08712028107655066}, "probe_overhead_sec_total": 298.045, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.005778, "alpha": 0.03, "t": 2657.7}
{"event": "consolidate", "step": 3000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2693.7}
{"event": "probe", "step": 3000, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3609.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.207378, "cps": 28.195312, "mean_p": 0.213641, "boundary_entropy": 0.109253, "bpb": 1.990119}, "tr": {"chunks": 3658.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.202066, "cps": 28.578125, "mean_p": 0.195593, "boundary_entropy": 0.11491, "bpb": 2.171186}, "fi": {"chunks": 4684.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.249003, "cps": 36.59375, "mean_p": 0.2434, "boundary_entropy": 0.115527, "bpb": 2.290445}}, "bpb": {"en": 1.99012, "tr": 2.17119, "fi": 2.29044}, "gini_cpc": 0.047523, "gini_cps": 0.059967, "bpic": 4.74705, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.974384547203315, "fi": 1.2007213727239854}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3678, "admitted_surface": 16826, "staged_surface": 0, "admitted_split": 16662, "staged_split": 0, "admitted_merge": 20925, "staged_merge": 0, "admitted_skel": 11123, "staged_skel": 0, "mean_count": 876.6223754882812, "n_util_obs": 462, "util_count_spearman": 0.08239743809449901}, "t": 2695.3}
{"event": "eval", "step": 3000, "phase": 2, "bpb": {"en": 2.00325, "tr": 2.18165, "fi": 2.29571}, "bytes_per_chunk": {"en": 4.542, "tr": 5.381, "fi": 4.198}, "bpb_mean": 2.1602, "frozen_inventory": true, "t": 2698.1}
{"event": "train", "step": 3000, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.48511, "bpb": 2.14256, "chunk_loss": 0.0307, "F": 0.21588, "G": 0.22155, "chunks": 221.06, "bytes_per_chunk": 4.632, "boundary_entropy": 0.10792, "router_noise": 0.0, "grad_norm": 0.693, "lr": 0.0001761, "router_trainable": true, "sec_per_step": 0.851, "probe_sec": 0.5611, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.53476, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.03062, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.03333, "n_merge": 2, "mean_delta_merge": -0.17336, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.51483, "trunc_abs_frac": 0.66678, "mean_delta_full": -0.19543, "read_hit_rate": {"surface": 0.63959, "split": 0.8642, "merge": 0.20371, "skel": 0.64805}, "mean_gate": {"surface": 0.19137, "split": 0.08914, "merge": 0.16634, "skel": 0.13995}, "n_reads": 19833, "n_cand": 31472, "read_suppressed_frac": 0.0939, "dropout_reads_logged": 19833, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3679, "admitted_surface": 16817, "staged_surface": 0, "admitted_split": 16721, "staged_split": 0, "admitted_merge": 20872, "staged_merge": 0, "admitted_skel": 11126, "staged_skel": 0, "mean_count": 876.9139404296875, "n_util_obs": 462, "util_count_spearman": 0.08264909193494342}, "probe_overhead_sec_total": 303.105, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.006008, "alpha": 0.03, "t": 2700.2}
{"event": "consolidate", "step": 3050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2738.2}
{"event": "train", "step": 3050, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.43196, "bpb": 2.06588, "chunk_loss": 0.030726, "F": 0.21649, "G": 0.22103, "chunks": 221.69, "bytes_per_chunk": 4.619, "boundary_entropy": 0.1054, "router_noise": 0.0, "grad_norm": 0.6544, "lr": 0.0001724, "router_trainable": true, "sec_per_step": 0.773, "probe_sec": 0.0657, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04075, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.02656, "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": 0.26145, "trunc_abs_frac": 0.22302, "mean_delta_full": 0.10657, "read_hit_rate": {"surface": 0.64826, "split": 0.86932, "merge": 0.2091, "skel": 0.65152}, "mean_gate": {"surface": 0.19417, "split": 0.08966, "merge": 0.16516, "skel": 0.14356}, "n_reads": 20195, "n_cand": 31700, "read_suppressed_frac": 0.097, "dropout_reads_logged": 20195, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3720, "admitted_surface": 16824, "staged_surface": 0, "admitted_split": 16693, "staged_split": 0, "admitted_merge": 20902, "staged_merge": 0, "admitted_skel": 11117, "staged_skel": 0, "mean_count": 891.9867553710938, "n_util_obs": 470, "util_count_spearman": 0.08511141324849475}, "probe_overhead_sec_total": 308.432, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.006262, "alpha": 0.03, "t": 2738.8}
{"event": "consolidate", "step": 3100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2775.2}
{"event": "train", "step": 3100, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.48178, "bpb": 2.13776, "chunk_loss": 0.030491, "F": 0.20578, "G": 0.21138, "chunks": 210.72, "bytes_per_chunk": 4.86, "boundary_entropy": 0.10767, "router_noise": 0.0, "grad_norm": 0.7088, "lr": 0.0001687, "router_trainable": true, "sec_per_step": 0.746, "probe_sec": 0.1095, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09046, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.02828, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": 0.04565, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": -0.28524, "trunc_abs_frac": 0.28524, "mean_delta_full": -0.10999, "read_hit_rate": {"surface": 0.62807, "split": 0.8491, "merge": 0.20243, "skel": 0.62459}, "mean_gate": {"surface": 0.19288, "split": 0.08763, "merge": 0.1642, "skel": 0.1406}, "n_reads": 18622, "n_cand": 30135, "read_suppressed_frac": 0.0976, "dropout_reads_logged": 18622, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3769, "admitted_surface": 16836, "staged_surface": 0, "admitted_split": 16839, "staged_split": 0, "admitted_merge": 20665, "staged_merge": 0, "admitted_skel": 11196, "staged_skel": 0, "mean_count": 907.0419921875, "n_util_obs": 478, "util_count_spearman": 0.08125916049777229}, "probe_overhead_sec_total": 313.339, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.007354, "alpha": 0.03, "t": 2776.2}
{"event": "consolidate", "step": 3150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2813.0}
{"event": "train", "step": 3150, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.43826, "bpb": 2.07497, "chunk_loss": 0.030642, "F": 0.21411, "G": 0.21845, "chunks": 219.25, "bytes_per_chunk": 4.67, "boundary_entropy": 0.10613, "router_noise": 0.0, "grad_norm": 0.6629, "lr": 0.000165, "router_trainable": true, "sec_per_step": 0.756, "probe_sec": 0.0694, "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": 10, "mean_delta_loo": 0.08755, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.03186, "n_merge": 1, "mean_delta_merge": 0.05499, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.53863, "trunc_abs_frac": 1.11401, "mean_delta_full": 0.18317, "read_hit_rate": {"surface": 0.6366, "split": 0.86452, "merge": 0.20325, "skel": 0.6352}, "mean_gate": {"surface": 0.19073, "split": 0.08726, "merge": 0.16386, "skel": 0.14016}, "n_reads": 19696, "n_cand": 31352, "read_suppressed_frac": 0.1015, "dropout_reads_logged": 19696, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3805, "admitted_surface": 16865, "staged_surface": 0, "admitted_split": 16721, "staged_split": 0, "admitted_merge": 20771, "staged_merge": 0, "admitted_skel": 11179, "staged_skel": 0, "mean_count": 922.2579345703125, "n_util_obs": 483, "util_count_spearman": 0.06937433580147077}, "probe_overhead_sec_total": 318.667, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.006915, "alpha": 0.03, "t": 2814.0}
{"event": "consolidate", "step": 3200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2851.0}
{"event": "probe", "step": 3200, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3553.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.20416, "cps": 27.757812, "mean_p": 0.209855, "boundary_entropy": 0.106271, "bpb": 1.975495}, "tr": {"chunks": 3528.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.194885, "cps": 27.5625, "mean_p": 0.189323, "boundary_entropy": 0.113722, "bpb": 2.151198}, "fi": {"chunks": 4596.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.244325, "cps": 35.90625, "mean_p": 0.241361, "boundary_entropy": 0.114328, "bpb": 2.272461}}, "bpb": {"en": 1.9755, "tr": 2.1512, "fi": 2.27246}, "gini_cpc": 0.051231, "gini_cps": 0.060975, "bpic": 4.85844, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9545681457912621, "fi": 1.1967324003628432}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3850, "admitted_surface": 16843, "staged_surface": 0, "admitted_split": 16808, "staged_split": 0, "admitted_merge": 20759, "staged_merge": 0, "admitted_skel": 11126, "staged_skel": 0, "mean_count": 937.0728759765625, "n_util_obs": 494, "util_count_spearman": 0.053755151124709376}, "t": 2852.6}
{"event": "eval", "step": 3200, "phase": 2, "bpb": {"en": 1.98691, "tr": 2.15927, "fi": 2.28039}, "bytes_per_chunk": {"en": 4.605, "tr": 5.652, "fi": 4.307}, "bpb_mean": 2.14219, "frozen_inventory": true, "t": 2855.6}
{"event": "train", "step": 3200, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.42163, "bpb": 2.05097, "chunk_loss": 0.030631, "F": 0.21475, "G": 0.21988, "chunks": 219.91, "bytes_per_chunk": 4.657, "boundary_entropy": 0.10208, "router_noise": 0.0, "grad_norm": 0.6177, "lr": 0.0001613, "router_trainable": true, "sec_per_step": 0.871, "probe_sec": 0.1345, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09864, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": 0.11922, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.58386, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 1.09941, "trunc_abs_frac": 0.35623, "mean_delta_full": 0.09332, "read_hit_rate": {"surface": 0.65253, "split": 0.86747, "merge": 0.21512, "skel": 0.64193}, "mean_gate": {"surface": 0.19262, "split": 0.08798, "merge": 0.16679, "skel": 0.14079}, "n_reads": 19813, "n_cand": 31186, "read_suppressed_frac": 0.1038, "dropout_reads_logged": 19813, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3850, "admitted_surface": 16811, "staged_surface": 0, "admitted_split": 16743, "staged_split": 0, "admitted_merge": 20815, "staged_merge": 0, "admitted_skel": 11167, "staged_skel": 0, "mean_count": 937.374267578125, "n_util_obs": 494, "util_count_spearman": 0.05384051025140961}, "probe_overhead_sec_total": 323.917, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.006627, "alpha": 0.03, "t": 2857.5}
{"event": "consolidate", "step": 3250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2895.0}
{"event": "train", "step": 3250, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.42615, "bpb": 2.0575, "chunk_loss": 0.0306, "F": 0.21008, "G": 0.21459, "chunks": 215.12, "bytes_per_chunk": 4.76, "boundary_entropy": 0.10365, "router_noise": 0.0, "grad_norm": 0.6834, "lr": 0.0001576, "router_trainable": true, "sec_per_step": 0.766, "probe_sec": 0.0649, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04344, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.03922, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.35297, "n_merge": 1, "mean_delta_merge": 0.22052, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -38.52273, "trunc_abs_frac": 173.41592, "mean_delta_full": -0.00336, "read_hit_rate": {"surface": 0.64332, "split": 0.85926, "merge": 0.20601, "skel": 0.64744}, "mean_gate": {"surface": 0.18996, "split": 0.08646, "merge": 0.16241, "skel": 0.14033}, "n_reads": 19291, "n_cand": 30640, "read_suppressed_frac": 0.0991, "dropout_reads_logged": 19291, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3888, "admitted_surface": 16833, "staged_surface": 0, "admitted_split": 16820, "staged_split": 0, "admitted_merge": 20769, "staged_merge": 0, "admitted_skel": 11114, "staged_skel": 0, "mean_count": 952.4917602539062, "n_util_obs": 499, "util_count_spearman": 0.055830505730267925}, "probe_overhead_sec_total": 328.878, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.006301, "alpha": 0.03, "t": 2895.8}
{"event": "consolidate", "step": 3300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2932.1}
{"event": "train", "step": 3300, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.40434, "bpb": 2.02603, "chunk_loss": 0.030698, "F": 0.21692, "G": 0.2199, "chunks": 222.12, "bytes_per_chunk": 4.61, "boundary_entropy": 0.10268, "router_noise": 0.0, "grad_norm": 0.6426, "lr": 0.0001539, "router_trainable": true, "sec_per_step": 0.747, "probe_sec": 0.1019, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08049, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.01576, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.26996, "n_merge": 1, "mean_delta_merge": 0.25201, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.04345, "trunc_abs_frac": 2.0137, "mean_delta_full": 0.12959, "read_hit_rate": {"surface": 0.65503, "split": 0.86927, "merge": 0.21039, "skel": 0.6495}, "mean_gate": {"surface": 0.18701, "split": 0.08376, "merge": 0.16106, "skel": 0.13622}, "n_reads": 20183, "n_cand": 31656, "read_suppressed_frac": 0.1007, "dropout_reads_logged": 20183, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3924, "admitted_surface": 16811, "staged_surface": 0, "admitted_split": 16718, "staged_split": 0, "admitted_merge": 20854, "staged_merge": 0, "admitted_skel": 11153, "staged_skel": 0, "mean_count": 967.671875, "n_util_obs": 506, "util_count_spearman": 0.06312759395311798}, "probe_overhead_sec_total": 333.401, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.006868, "alpha": 0.03, "t": 2933.2}
{"event": "consolidate", "step": 3350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2988.5}
{"event": "train", "step": 3350, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.343, "bpb": 1.93754, "chunk_loss": 0.030673, "F": 0.2178, "G": 0.22028, "chunks": 223.03, "bytes_per_chunk": 4.591, "boundary_entropy": 0.09689, "router_noise": 0.0, "grad_norm": 0.6962, "lr": 0.0001501, "router_trainable": true, "sec_per_step": 1.119, "probe_sec": 0.0739, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0534, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.05379, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.0098, "n_merge": 2, "mean_delta_merge": 0.62234, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 1.90033, "trunc_abs_frac": 0.35522, "mean_delta_full": 0.10678, "read_hit_rate": {"surface": 0.67023, "split": 0.88146, "merge": 0.2306, "skel": 0.66907}, "mean_gate": {"surface": 0.18582, "split": 0.08357, "merge": 0.15628, "skel": 0.13708}, "n_reads": 20827, "n_cand": 31855, "read_suppressed_frac": 0.1003, "dropout_reads_logged": 20827, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3970, "admitted_surface": 16760, "staged_surface": 0, "admitted_split": 16701, "staged_split": 0, "admitted_merge": 20918, "staged_merge": 0, "admitted_skel": 11157, "staged_skel": 0, "mean_count": 982.9249267578125, "n_util_obs": 510, "util_count_spearman": 0.06534729118634901}, "probe_overhead_sec_total": 340.074, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.007014, "alpha": 0.03, "t": 2989.1}
{"event": "consolidate", "step": 3400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3036.9}
{"event": "probe", "step": 3400, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3574.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.205367, "cps": 27.921875, "mean_p": 0.208351, "boundary_entropy": 0.098085, "bpb": 1.962233}, "tr": {"chunks": 3574.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.197426, "cps": 27.921875, "mean_p": 0.190252, "boundary_entropy": 0.111134, "bpb": 2.149271}, "fi": {"chunks": 4633.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.246292, "cps": 36.195312, "mean_p": 0.242505, "boundary_entropy": 0.106424, "bpb": 2.240909}}, "bpb": {"en": 1.96223, "tr": 2.14927, "fi": 2.24091}, "gini_cpc": 0.05019, "gini_cps": 0.059927, "bpic": 4.81555, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9613323758493066, "fi": 1.1992783366195892}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4006, "admitted_surface": 16778, "staged_surface": 0, "admitted_split": 16750, "staged_split": 0, "admitted_merge": 20815, "staged_merge": 0, "admitted_skel": 11193, "staged_skel": 0, "mean_count": 997.8128051757812, "n_util_obs": 514, "util_count_spearman": 0.05960706505095769}, "t": 3038.6}
{"event": "eval", "step": 3400, "phase": 2, "bpb": {"en": 1.97404, "tr": 2.15835, "fi": 2.2505}, "bytes_per_chunk": {"en": 4.583, "tr": 5.63, "fi": 4.266}, "bpb_mean": 2.12763, "frozen_inventory": true, "t": 3041.6}
{"event": "train", "step": 3400, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.41127, "bpb": 2.03603, "chunk_loss": 0.03059, "F": 0.21152, "G": 0.21505, "chunks": 216.59, "bytes_per_chunk": 4.728, "boundary_entropy": 0.09847, "router_noise": 0.0, "grad_norm": 0.6856, "lr": 0.0001465, "router_trainable": true, "sec_per_step": 1.111, "probe_sec": 0.0802, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05146, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": 0.10547, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.03104, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.08145, "trunc_abs_frac": 0.08811, "mean_delta_full": 0.49292, "read_hit_rate": {"surface": 0.64502, "split": 0.86117, "merge": 0.21348, "skel": 0.64728}, "mean_gate": {"surface": 0.18778, "split": 0.0845, "merge": 0.16106, "skel": 0.13943}, "n_reads": 19711, "n_cand": 31113, "read_suppressed_frac": 0.099, "dropout_reads_logged": 19711, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4006, "admitted_surface": 16816, "staged_surface": 0, "admitted_split": 16821, "staged_split": 0, "admitted_merge": 20732, "staged_merge": 0, "admitted_skel": 11167, "staged_skel": 0, "mean_count": 998.1171875, "n_util_obs": 514, "util_count_spearman": 0.05989792337897577}, "probe_overhead_sec_total": 345.793, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.006632, "alpha": 0.03, "t": 3044.6}
{"event": "consolidate", "step": 3450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3089.9}
{"event": "train", "step": 3450, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.37389, "bpb": 1.9821, "chunk_loss": 0.030578, "F": 0.21176, "G": 0.21601, "chunks": 216.84, "bytes_per_chunk": 4.722, "boundary_entropy": 0.09581, "router_noise": 0.0, "grad_norm": 0.7061, "lr": 0.0001428, "router_trainable": true, "sec_per_step": 0.919, "probe_sec": 0.1262, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10567, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.01624, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": 0.04462, "n_merge": 1, "mean_delta_merge": 0.28394, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 1.34342, "trunc_abs_frac": 1.34342, "mean_delta_full": 0.13721, "read_hit_rate": {"surface": 0.64674, "split": 0.85985, "merge": 0.21789, "skel": 0.65297}, "mean_gate": {"surface": 0.18773, "split": 0.08354, "merge": 0.16053, "skel": 0.13909}, "n_reads": 19799, "n_cand": 31177, "read_suppressed_frac": 0.1026, "dropout_reads_logged": 19799, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4039, "admitted_surface": 16822, "staged_surface": 0, "admitted_split": 16833, "staged_split": 0, "admitted_merge": 20720, "staged_merge": 0, "admitted_skel": 11161, "staged_skel": 0, "mean_count": 1013.2218017578125, "n_util_obs": 521, "util_count_spearman": 0.034291372316113435}, "probe_overhead_sec_total": 351.195, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.00651, "alpha": 0.03, "t": 3090.6}
{"event": "consolidate", "step": 3500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3128.0}
{"event": "train", "step": 3500, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.35676, "bpb": 1.95739, "chunk_loss": 0.030494, "F": 0.2084, "G": 0.21155, "chunks": 213.41, "bytes_per_chunk": 4.798, "boundary_entropy": 0.09664, "router_noise": 0.0, "grad_norm": 0.6607, "lr": 0.0001391, "router_trainable": true, "sec_per_step": 0.759, "probe_sec": 0.0784, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0554, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.04292, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.17621, "n_merge": 1, "mean_delta_merge": -0.35397, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.68354, "trunc_abs_frac": 1.42741, "mean_delta_full": 0.18689, "read_hit_rate": {"surface": 0.66162, "split": 0.8708, "merge": 0.2303, "skel": 0.65718}, "mean_gate": {"surface": 0.19167, "split": 0.08612, "merge": 0.16315, "skel": 0.14217}, "n_reads": 19801, "n_cand": 30632, "read_suppressed_frac": 0.0965, "dropout_reads_logged": 19801, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4079, "admitted_surface": 16776, "staged_surface": 0, "admitted_split": 16801, "staged_split": 0, "admitted_merge": 20791, "staged_merge": 0, "admitted_skel": 11168, "staged_skel": 0, "mean_count": 1028.3961181640625, "n_util_obs": 531, "util_count_spearman": 0.020329887338061813}, "probe_overhead_sec_total": 356.134, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.005775, "alpha": 0.03, "t": 3128.5}
{"event": "consolidate", "step": 3550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3166.2}
{"event": "train", "step": 3550, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.39934, "bpb": 2.01883, "chunk_loss": 0.030607, "F": 0.21173, "G": 0.21567, "chunks": 216.81, "bytes_per_chunk": 4.723, "boundary_entropy": 0.09631, "router_noise": 0.0, "grad_norm": 0.6596, "lr": 0.0001355, "router_trainable": true, "sec_per_step": 0.764, "probe_sec": 0.0786, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05773, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.01946, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": -0.06805, "n_merge": 1, "mean_delta_merge": 0.00818, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": -4.11352, "trunc_abs_frac": 4.11352, "mean_delta_full": -0.02393, "read_hit_rate": {"surface": 0.64625, "split": 0.86349, "merge": 0.21734, "skel": 0.65098}, "mean_gate": {"surface": 0.18224, "split": 0.08032, "merge": 0.15459, "skel": 0.13369}, "n_reads": 19709, "n_cand": 31019, "read_suppressed_frac": 0.0967, "dropout_reads_logged": 19709, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4120, "admitted_surface": 16827, "staged_surface": 0, "admitted_split": 16794, "staged_split": 0, "admitted_merge": 20759, "staged_merge": 0, "admitted_skel": 11156, "staged_skel": 0, "mean_count": 1043.44189453125, "n_util_obs": 536, "util_count_spearman": 0.00773519785822095}, "probe_overhead_sec_total": 361.394, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.006544, "alpha": 0.03, "t": 3166.7}
{"event": "consolidate", "step": 3600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3203.7}
{"event": "probe", "step": 3600, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3565.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.20485, "cps": 27.851562, "mean_p": 0.208411, "boundary_entropy": 0.097253, "bpb": 1.946516}, "tr": {"chunks": 3548.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.19599, "cps": 27.71875, "mean_p": 0.189158, "boundary_entropy": 0.103603, "bpb": 2.107954}, "fi": {"chunks": 4587.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.243847, "cps": 35.835938, "mean_p": 0.240436, "boundary_entropy": 0.102318, "bpb": 2.215877}}, "bpb": {"en": 1.94652, "tr": 2.10795, "fi": 2.21588}, "gini_cpc": 0.049489, "gini_cps": 0.059202, "bpic": 4.848889, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9567481821916802, "fi": 1.190368546111191}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4156, "admitted_surface": 16807, "staged_surface": 0, "admitted_split": 16707, "staged_split": 0, "admitted_merge": 20896, "staged_merge": 0, "admitted_skel": 11126, "staged_skel": 0, "mean_count": 1058.318603515625, "n_util_obs": 546, "util_count_spearman": 0.0146135436452734}, "t": 3205.5}
{"event": "eval", "step": 3600, "phase": 2, "bpb": {"en": 1.95912, "tr": 2.11762, "fi": 2.22478}, "bytes_per_chunk": {"en": 4.572, "tr": 5.641, "fi": 4.318}, "bpb_mean": 2.10051, "frozen_inventory": true, "t": 3208.2}
{"event": "train", "step": 3600, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.39683, "bpb": 2.01519, "chunk_loss": 0.030689, "F": 0.21765, "G": 0.22104, "chunks": 222.88, "bytes_per_chunk": 4.595, "boundary_entropy": 0.09649, "router_noise": 0.0, "grad_norm": 0.6118, "lr": 0.0001319, "router_trainable": true, "sec_per_step": 0.867, "probe_sec": 0.0864, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06317, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.06024, "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": 1.01187, "trunc_abs_frac": 1.21168, "mean_delta_full": 0.15426, "read_hit_rate": {"surface": 0.64465, "split": 0.87025, "merge": 0.20628, "skel": 0.6578}, "mean_gate": {"surface": 0.18498, "split": 0.0804, "merge": 0.15784, "skel": 0.13351}, "n_reads": 20333, "n_cand": 31929, "read_suppressed_frac": 0.1006, "dropout_reads_logged": 20333, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4156, "admitted_surface": 16834, "staged_surface": 0, "admitted_split": 16708, "staged_split": 0, "admitted_merge": 20883, "staged_merge": 0, "admitted_skel": 11111, "staged_skel": 0, "mean_count": 1058.622802734375, "n_util_obs": 546, "util_count_spearman": 0.014230367828622859}, "probe_overhead_sec_total": 366.978, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.006143, "alpha": 0.03, "t": 3210.1}
{"event": "consolidate", "step": 3650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3247.1}
{"event": "train", "step": 3650, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.43027, "bpb": 2.06345, "chunk_loss": 0.030688, "F": 0.21609, "G": 0.2185, "chunks": 221.28, "bytes_per_chunk": 4.628, "boundary_entropy": 0.0937, "router_noise": 0.0, "grad_norm": 0.6969, "lr": 0.0001283, "router_trainable": true, "sec_per_step": 0.754, "probe_sec": 0.1496, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.11277, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.07058, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.06007, "n_merge": 3, "mean_delta_merge": -0.0599, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.64191, "trunc_abs_frac": 0.94037, "mean_delta_full": -0.27936, "read_hit_rate": {"surface": 0.64279, "split": 0.86616, "merge": 0.20336, "skel": 0.65421}, "mean_gate": {"surface": 0.17754, "split": 0.0773, "merge": 0.14987, "skel": 0.12889}, "n_reads": 20154, "n_cand": 31791, "read_suppressed_frac": 0.1012, "dropout_reads_logged": 20154, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4193, "admitted_surface": 16828, "staged_surface": 0, "admitted_split": 16767, "staged_split": 0, "admitted_merge": 20799, "staged_merge": 0, "admitted_skel": 11142, "staged_skel": 0, "mean_count": 1073.9208984375, "n_util_obs": 550, "util_count_spearman": 0.00936994866831832}, "probe_overhead_sec_total": 372.171, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.006773, "alpha": 0.03, "t": 3247.8}
{"event": "consolidate", "step": 3700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3285.5}
{"event": "train", "step": 3700, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.38251, "bpb": 1.99453, "chunk_loss": 0.0307, "F": 0.21878, "G": 0.22128, "chunks": 224.03, "bytes_per_chunk": 4.571, "boundary_entropy": 0.09834, "router_noise": 0.0, "grad_norm": 0.6583, "lr": 0.0001247, "router_trainable": true, "sec_per_step": 0.775, "probe_sec": 0.0775, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05458, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.054, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": 0.00131, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -0.33234, "trunc_abs_frac": 0.66954, "mean_delta_full": -0.1167, "read_hit_rate": {"surface": 0.66162, "split": 0.87241, "merge": 0.21619, "skel": 0.64673}, "mean_gate": {"surface": 0.18792, "split": 0.08089, "merge": 0.16108, "skel": 0.13744}, "n_reads": 20208, "n_cand": 31603, "read_suppressed_frac": 0.1041, "dropout_reads_logged": 20208, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4228, "admitted_surface": 16796, "staged_surface": 0, "admitted_split": 16706, "staged_split": 0, "admitted_merge": 20889, "staged_merge": 0, "admitted_skel": 11145, "staged_skel": 0, "mean_count": 1089.1087646484375, "n_util_obs": 557, "util_count_spearman": 0.0018097759636655996}, "probe_overhead_sec_total": 377.035, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.006829, "alpha": 0.03, "t": 3286.5}
{"event": "consolidate", "step": 3750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3324.3}
{"event": "train", "step": 3750, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.36811, "bpb": 1.97377, "chunk_loss": 0.03055, "F": 0.21185, "G": 0.21472, "chunks": 216.94, "bytes_per_chunk": 4.72, "boundary_entropy": 0.09435, "router_noise": 0.0, "grad_norm": 0.6598, "lr": 0.0001212, "router_trainable": true, "sec_per_step": 0.768, "probe_sec": 0.1429, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.12253, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.00568, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": 0.00452, "n_merge": 1, "mean_delta_merge": -0.4137, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 7.14481, "trunc_abs_frac": 0.53746, "mean_delta_full": -0.03235, "read_hit_rate": {"surface": 0.64906, "split": 0.86691, "merge": 0.21809, "skel": 0.65143}, "mean_gate": {"surface": 0.18721, "split": 0.07963, "merge": 0.15685, "skel": 0.13567}, "n_reads": 19891, "n_cand": 31161, "read_suppressed_frac": 0.1001, "dropout_reads_logged": 19891, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4270, "admitted_surface": 16825, "staged_surface": 0, "admitted_split": 16791, "staged_split": 0, "admitted_merge": 20746, "staged_merge": 0, "admitted_skel": 11174, "staged_skel": 0, "mean_count": 1104.1505126953125, "n_util_obs": 559, "util_count_spearman": 0.005968285063593183}, "probe_overhead_sec_total": 382.421, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.006403, "alpha": 0.03, "t": 3324.9}
{"event": "consolidate", "step": 3800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3362.0}
{"event": "probe", "step": 3800, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3570.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.205137, "cps": 27.890625, "mean_p": 0.207482, "boundary_entropy": 0.09514, "bpb": 1.937295}, "tr": {"chunks": 3620.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.199967, "cps": 28.28125, "mean_p": 0.19146, "boundary_entropy": 0.095406, "bpb": 2.100296}, "fi": {"chunks": 4700.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.249854, "cps": 36.71875, "mean_p": 0.242085, "boundary_entropy": 0.099286, "bpb": 2.196754}}, "bpb": {"en": 1.93729, "tr": 2.1003, "fi": 2.19675}, "gini_cpc": 0.050779, "gini_cps": 0.063359, "bpic": 4.771405, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.974796414726748, "fi": 1.2179848282941905}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4312, "admitted_surface": 16822, "staged_surface": 0, "admitted_split": 16762, "staged_split": 0, "admitted_merge": 20837, "staged_merge": 0, "admitted_skel": 11115, "staged_skel": 0, "mean_count": 1118.9527587890625, "n_util_obs": 564, "util_count_spearman": 0.018954026478575307}, "t": 3363.7}
{"event": "eval", "step": 3800, "phase": 2, "bpb": {"en": 1.95014, "tr": 2.10895, "fi": 2.20587}, "bytes_per_chunk": {"en": 4.587, "tr": 5.508, "fi": 4.218}, "bpb_mean": 2.08832, "frozen_inventory": true, "t": 3366.8}
{"event": "train", "step": 3800, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.44048, "bpb": 2.07818, "chunk_loss": 0.030753, "F": 0.22064, "G": 0.2221, "chunks": 225.94, "bytes_per_chunk": 4.532, "boundary_entropy": 0.09182, "router_noise": 0.0, "grad_norm": 0.689, "lr": 0.0001177, "router_trainable": true, "sec_per_step": 0.871, "probe_sec": 0.1008, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07732, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": -0.03364, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.05463, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -5.63858, "trunc_abs_frac": 3.11831, "mean_delta_full": 0.02651, "read_hit_rate": {"surface": 0.63893, "split": 0.87422, "merge": 0.19202, "skel": 0.66345}, "mean_gate": {"surface": 0.18021, "split": 0.07603, "merge": 0.15481, "skel": 0.13012}, "n_reads": 20637, "n_cand": 32483, "read_suppressed_frac": 0.1002, "dropout_reads_logged": 20637, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4313, "admitted_surface": 16844, "staged_surface": 0, "admitted_split": 16663, "staged_split": 0, "admitted_merge": 20933, "staged_merge": 0, "admitted_skel": 11096, "staged_skel": 0, "mean_count": 1119.23828125, "n_util_obs": 565, "util_count_spearman": 0.023683428281781802}, "probe_overhead_sec_total": 387.342, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.00689, "alpha": 0.03, "t": 3368.4}
{"event": "consolidate", "step": 3850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3406.1}
{"event": "train", "step": 3850, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.53551, "bpb": 2.21527, "chunk_loss": 0.03083, "F": 0.21371, "G": 0.21651, "chunks": 218.84, "bytes_per_chunk": 4.679, "boundary_entropy": 0.09305, "router_noise": 0.0, "grad_norm": 0.6909, "lr": 0.0001142, "router_trainable": true, "sec_per_step": 0.773, "probe_sec": 0.1027, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08225, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": -0.04688, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.08228, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -11.33279, "trunc_abs_frac": 0.32884, "mean_delta_full": -0.00235, "read_hit_rate": {"surface": 0.62631, "split": 0.85022, "merge": 0.18576, "skel": 0.63145}, "mean_gate": {"surface": 0.18186, "split": 0.07807, "merge": 0.15329, "skel": 0.13412}, "n_reads": 18892, "n_cand": 30811, "read_suppressed_frac": 0.1042, "dropout_reads_logged": 18892, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4349, "admitted_surface": 16829, "staged_surface": 0, "admitted_split": 16824, "staged_split": 0, "admitted_merge": 20774, "staged_merge": 0, "admitted_skel": 11109, "staged_skel": 0, "mean_count": 1134.435302734375, "n_util_obs": 576, "util_count_spearman": 0.01587396101577941}, "probe_overhead_sec_total": 392.304, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.00642, "alpha": 0.03, "t": 3407.1}
{"event": "consolidate", "step": 3900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3449.0}
{"event": "train", "step": 3900, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.42095, "bpb": 2.05, "chunk_loss": 0.030615, "F": 0.21326, "G": 0.21558, "chunks": 218.38, "bytes_per_chunk": 4.689, "boundary_entropy": 0.09508, "router_noise": 0.0, "grad_norm": 0.6898, "lr": 0.0001108, "router_trainable": true, "sec_per_step": 0.85, "probe_sec": 0.0713, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04703, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": 0.00992, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.13486, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.58055, "trunc_abs_frac": 0.82493, "mean_delta_full": 0.23224, "read_hit_rate": {"surface": 0.63614, "split": 0.87033, "merge": 0.19541, "skel": 0.6487}, "mean_gate": {"surface": 0.17537, "split": 0.0747, "merge": 0.14713, "skel": 0.12799}, "n_reads": 19846, "n_cand": 31414, "read_suppressed_frac": 0.0988, "dropout_reads_logged": 19846, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4392, "admitted_surface": 16857, "staged_surface": 0, "admitted_split": 16724, "staged_split": 0, "admitted_merge": 20816, "staged_merge": 0, "admitted_skel": 11139, "staged_skel": 0, "mean_count": 1149.4219970703125, "n_util_obs": 583, "util_count_spearman": 0.013673801258946667}, "probe_overhead_sec_total": 397.499, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.00637, "alpha": 0.03, "t": 3449.6}
{"event": "consolidate", "step": 3950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3490.3}
{"event": "train", "step": 3950, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.4792, "bpb": 2.13403, "chunk_loss": 0.030611, "F": 0.21133, "G": 0.21365, "chunks": 216.41, "bytes_per_chunk": 4.732, "boundary_entropy": 0.09207, "router_noise": 0.0, "grad_norm": 0.7704, "lr": 0.0001074, "router_trainable": true, "sec_per_step": 0.825, "probe_sec": 0.0701, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04751, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.13419, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": 0.17266, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 1.0, "trunc_abs_frac": 1.0, "mean_delta_full": 0.01709, "read_hit_rate": {"surface": 0.62513, "split": 0.85779, "merge": 0.19472, "skel": 0.63189}, "mean_gate": {"surface": 0.17587, "split": 0.07412, "merge": 0.14949, "skel": 0.12713}, "n_reads": 19200, "n_cand": 30957, "read_suppressed_frac": 0.101, "dropout_reads_logged": 19200, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4426, "admitted_surface": 16825, "staged_surface": 0, "admitted_split": 16803, "staged_split": 0, "admitted_merge": 20740, "staged_merge": 0, "admitted_skel": 11168, "staged_skel": 0, "mean_count": 1164.67822265625, "n_util_obs": 589, "util_count_spearman": 0.022371784477896988}, "probe_overhead_sec_total": 402.4, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.006521, "alpha": 0.03, "t": 3490.9}
{"event": "consolidate", "step": 4000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3543.6}
{"event": "probe", "step": 4000, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3531.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.202896, "cps": 27.585938, "mean_p": 0.205653, "boundary_entropy": 0.09436, "bpb": 1.923057}, "tr": {"chunks": 3565.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.196929, "cps": 27.851562, "mean_p": 0.189319, "boundary_entropy": 0.095868, "bpb": 2.083364}, "fi": {"chunks": 4589.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.243953, "cps": 35.851562, "mean_p": 0.239222, "boundary_entropy": 0.098406, "bpb": 2.179902}}, "bpb": {"en": 1.92306, "tr": 2.08336, "fi": 2.1799}, "gini_cpc": 0.048696, "gini_cps": 0.060362, "bpic": 4.855113, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9705890455686147, "fi": 1.2023546211392158}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4460, "admitted_surface": 16823, "staged_surface": 0, "admitted_split": 16864, "staged_split": 0, "admitted_merge": 20674, "staged_merge": 0, "admitted_skel": 11175, "staged_skel": 0, "mean_count": 1179.469970703125, "n_util_obs": 599, "util_count_spearman": 0.021797310810547314}, "t": 3545.4}
{"event": "eval", "step": 4000, "phase": 2, "bpb": {"en": 1.93587, "tr": 2.09258, "fi": 2.1863}, "bytes_per_chunk": {"en": 4.628, "tr": 5.623, "fi": 4.326}, "bpb_mean": 2.07159, "frozen_inventory": true, "t": 3548.3}
{"event": "train", "step": 4000, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.39999, "bpb": 2.01976, "chunk_loss": 0.030683, "F": 0.21475, "G": 0.21753, "chunks": 219.91, "bytes_per_chunk": 4.657, "boundary_entropy": 0.09138, "router_noise": 0.0, "grad_norm": 0.649, "lr": 0.000104, "router_trainable": true, "sec_per_step": 1.195, "probe_sec": 0.0676, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04593, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": 0.07817, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 3, "mean_delta_split": -0.15438, "n_merge": 2, "mean_delta_merge": -0.06821, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.06407, "trunc_abs_frac": 0.46621, "mean_delta_full": -0.24402, "read_hit_rate": {"surface": 0.65039, "split": 0.86726, "merge": 0.2144, "skel": 0.64552}, "mean_gate": {"surface": 0.18366, "split": 0.07781, "merge": 0.15268, "skel": 0.13411}, "n_reads": 19973, "n_cand": 31397, "read_suppressed_frac": 0.1006, "dropout_reads_logged": 19973, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4460, "admitted_surface": 16820, "staged_surface": 0, "admitted_split": 16807, "staged_split": 0, "admitted_merge": 20742, "staged_merge": 0, "admitted_skel": 11167, "staged_skel": 0, "mean_count": 1179.7706298828125, "n_util_obs": 599, "util_count_spearman": 0.021912106683497426}, "probe_overhead_sec_total": 408.033, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.006563, "alpha": 0.03, "t": 3550.6}
{"event": "consolidate", "step": 4050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3600.4}
{"event": "train", "step": 4050, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.41533, "bpb": 2.04189, "chunk_loss": 0.030653, "F": 0.21503, "G": 0.21784, "chunks": 220.19, "bytes_per_chunk": 4.651, "boundary_entropy": 0.09262, "router_noise": 0.0, "grad_norm": 0.6852, "lr": 0.0001007, "router_trainable": true, "sec_per_step": 1.01, "probe_sec": 0.2216, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.17553, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": 0.10817, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": 0.05585, "n_merge": 3, "mean_delta_merge": 0.05564, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.04237, "trunc_abs_frac": 0.04237, "mean_delta_full": 0.44373, "read_hit_rate": {"surface": 0.64813, "split": 0.85916, "merge": 0.22028, "skel": 0.64059}, "mean_gate": {"surface": 0.1807, "split": 0.07631, "merge": 0.15366, "skel": 0.13105}, "n_reads": 19806, "n_cand": 31330, "read_suppressed_frac": 0.1027, "dropout_reads_logged": 19806, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4501, "admitted_surface": 16820, "staged_surface": 0, "admitted_split": 16861, "staged_split": 0, "admitted_merge": 20661, "staged_merge": 0, "admitted_skel": 11194, "staged_skel": 0, "mean_count": 1194.7982177734375, "n_util_obs": 607, "util_count_spearman": 0.018939974346375243}, "probe_overhead_sec_total": 413.921, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.006257, "alpha": 0.03, "t": 3601.1}
{"event": "probe", "step": 4096, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3525.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.202551, "cps": 27.539062, "mean_p": 0.204543, "boundary_entropy": 0.093208, "bpb": 1.917697}, "tr": {"chunks": 3570.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.197205, "cps": 27.890625, "mean_p": 0.189822, "boundary_entropy": 0.097112, "bpb": 2.080351}, "fi": {"chunks": 4598.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.244431, "cps": 35.921875, "mean_p": 0.239011, "boundary_entropy": 0.095348, "bpb": 2.17238}}, "bpb": {"en": 1.9177, "tr": 2.08035, "fi": 2.17238}, "gini_cpc": 0.048875, "gini_cps": 0.061176, "bpic": 4.851792, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9736047040516382, "fi": 1.206763267757548}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4528, "admitted_surface": 16854, "staged_surface": 0, "admitted_split": 16800, "staged_split": 0, "admitted_merge": 20725, "staged_merge": 0, "admitted_skel": 11157, "staged_skel": 0, "mean_count": 1208.34716796875, "n_util_obs": 613, "util_count_spearman": 0.012327276083176141}, "t": 3644.0}
{"event": "consolidate", "step": 4100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3648.8}
{"event": "train", "step": 4100, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.35286, "bpb": 1.95177, "chunk_loss": 0.03055, "F": 0.21057, "G": 0.21263, "chunks": 215.62, "bytes_per_chunk": 4.749, "boundary_entropy": 0.08802, "router_noise": 0.0, "grad_norm": 0.6749, "lr": 9.75e-05, "router_trainable": true, "sec_per_step": 0.965, "probe_sec": 0.1346, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.11335, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.06502, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": 0.23855, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.91859, "trunc_abs_frac": 0.91859, "mean_delta_full": -0.07422, "read_hit_rate": {"surface": 0.65026, "split": 0.86711, "merge": 0.21109, "skel": 0.64987}, "mean_gate": {"surface": 0.1797, "split": 0.07542, "merge": 0.15155, "skel": 0.13094}, "n_reads": 19720, "n_cand": 30964, "read_suppressed_frac": 0.0992, "dropout_reads_logged": 19720, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4534, "admitted_surface": 16801, "staged_surface": 0, "admitted_split": 16834, "staged_split": 0, "admitted_merge": 20709, "staged_merge": 0, "admitted_skel": 11192, "staged_skel": 0, "mean_count": 1209.81201171875, "n_util_obs": 614, "util_count_spearman": 0.01682507130210974}, "probe_overhead_sec_total": 419.611, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.007178, "alpha": 0.03, "t": 3649.4}
{"event": "consolidate", "step": 4150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3686.2}
{"event": "train", "step": 4150, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.41369, "bpb": 2.03952, "chunk_loss": 0.030812, "F": 0.22006, "G": 0.22216, "chunks": 225.34, "bytes_per_chunk": 4.544, "boundary_entropy": 0.08886, "router_noise": 0.0, "grad_norm": 0.7457, "lr": 9.43e-05, "router_trainable": true, "sec_per_step": 0.747, "probe_sec": 0.0667, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04687, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.05756, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 3, "mean_delta_split": -0.00201, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -1.68768, "trunc_abs_frac": 1.11, "mean_delta_full": -0.05059, "read_hit_rate": {"surface": 0.6409, "split": 0.86465, "merge": 0.20344, "skel": 0.65118}, "mean_gate": {"surface": 0.1738, "split": 0.07169, "merge": 0.14582, "skel": 0.12599}, "n_reads": 20375, "n_cand": 32248, "read_suppressed_frac": 0.0954, "dropout_reads_logged": 20375, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4570, "admitted_surface": 16850, "staged_surface": 0, "admitted_split": 16820, "staged_split": 0, "admitted_merge": 20711, "staged_merge": 0, "admitted_skel": 11155, "staged_skel": 0, "mean_count": 1224.933837890625, "n_util_obs": 618, "util_count_spearman": 0.022983665593338247}, "probe_overhead_sec_total": 424.69, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.006252, "alpha": 0.03, "t": 3686.8}
{"event": "consolidate", "step": 4200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3724.0}
{"event": "probe", "step": 4200, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3574.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.205367, "cps": 27.921875, "mean_p": 0.206972, "boundary_entropy": 0.087621, "bpb": 1.915342}, "tr": {"chunks": 3605.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.199138, "cps": 28.164062, "mean_p": 0.19111, "boundary_entropy": 0.094122, "bpb": 2.068247}, "fi": {"chunks": 4686.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.24911, "cps": 36.609375, "mean_p": 0.242144, "boundary_entropy": 0.094002, "bpb": 2.167784}}, "bpb": {"en": 1.91534, "tr": 2.06825, "fi": 2.16778}, "gini_cpc": 0.050969, "gini_cps": 0.062481, "bpic": 4.781458, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9696707372514691, "fi": 1.212997687329893}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4613, "admitted_surface": 16829, "staged_surface": 0, "admitted_split": 16869, "staged_split": 0, "admitted_merge": 20639, "staged_merge": 0, "admitted_skel": 11199, "staged_skel": 0, "mean_count": 1239.778076171875, "n_util_obs": 628, "util_count_spearman": 0.02797739655473839}, "t": 3725.5}
{"event": "eval", "step": 4200, "phase": 2, "bpb": {"en": 1.92663, "tr": 2.07654, "fi": 2.17611}, "bytes_per_chunk": {"en": 4.598, "tr": 5.584, "fi": 4.233}, "bpb_mean": 2.05976, "frozen_inventory": true, "t": 3728.4}
{"event": "phase_transition", "step": 4200, "phase": 3, "objective": "global", "router_trainable": true, "encoder_trainable": true, "frozen_router_params": 0, "t": 3729.5}
{"event": "train", "step": 4200, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.34174, "bpb": 1.93572, "chunk_loss": 0.030475, "F": 0.20682, "G": 0.20982, "chunks": 211.78, "bytes_per_chunk": 4.835, "boundary_entropy": 0.08456, "router_noise": 0.0, "grad_norm": 0.6658, "lr": 9.12e-05, "router_trainable": true, "sec_per_step": 0.878, "probe_sec": 0.0863, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06194, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.11107, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.03351, "n_merge": 1, "mean_delta_merge": 0.54599, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.54805, "trunc_abs_frac": 0.46979, "mean_delta_full": -0.25745, "read_hit_rate": {"surface": 0.6424, "split": 0.86214, "merge": 0.20781, "skel": 0.64578}, "mean_gate": {"surface": 0.18158, "split": 0.0758, "merge": 0.15324, "skel": 0.13379}, "n_reads": 19333, "n_cand": 30564, "read_suppressed_frac": 0.1004, "dropout_reads_logged": 19333, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4614, "admitted_surface": 16835, "staged_surface": 0, "admitted_split": 16864, "staged_split": 0, "admitted_merge": 20628, "staged_merge": 0, "admitted_skel": 11209, "staged_skel": 0, "mean_count": 1240.0732421875, "n_util_obs": 628, "util_count_spearman": 0.02784258277979101}, "probe_overhead_sec_total": 429.495, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.007242, "alpha": 0.03, "t": 3730.7}
{"event": "consolidate", "step": 4250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3770.0}
{"event": "train", "step": 4250, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.37574, "bpb": 1.98478, "chunk_loss": 0.030667, "F": 0.21481, "G": 0.21692, "chunks": 219.97, "bytes_per_chunk": 4.655, "boundary_entropy": 0.08532, "router_noise": 0.0, "grad_norm": 0.647, "lr": 8.81e-05, "router_trainable": true, "sec_per_step": 0.799, "probe_sec": 0.0907, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0676, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.00923, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": 0.03158, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.45503, "trunc_abs_frac": 0.59694, "mean_delta_full": -0.22982, "read_hit_rate": {"surface": 0.65035, "split": 0.86618, "merge": 0.21333, "skel": 0.65597}, "mean_gate": {"surface": 0.17694, "split": 0.07378, "merge": 0.15186, "skel": 0.12993}, "n_reads": 20089, "n_cand": 31512, "read_suppressed_frac": 0.1029, "dropout_reads_logged": 20089, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4654, "admitted_surface": 16802, "staged_surface": 0, "admitted_split": 16831, "staged_split": 0, "admitted_merge": 20737, "staged_merge": 0, "admitted_skel": 11166, "staged_skel": 0, "mean_count": 1255.21044921875, "n_util_obs": 641, "util_count_spearman": 0.039868160146966626}, "probe_overhead_sec_total": 434.969, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.006753, "alpha": 0.03, "t": 3770.6}
{"event": "consolidate", "step": 4300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3807.1}
{"event": "train", "step": 4300, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.37712, "bpb": 1.98676, "chunk_loss": 0.030477, "F": 0.20712, "G": 0.20964, "chunks": 212.09, "bytes_per_chunk": 4.828, "boundary_entropy": 0.08605, "router_noise": 0.0, "grad_norm": 0.7326, "lr": 8.5e-05, "router_trainable": true, "sec_per_step": 0.753, "probe_sec": 0.0643, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04357, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.05181, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": -0.21722, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 1.01467, "trunc_abs_frac": 0.75537, "mean_delta_full": 0.03467, "read_hit_rate": {"surface": 0.63375, "split": 0.85549, "merge": 0.19173, "skel": 0.63723}, "mean_gate": {"surface": 0.17485, "split": 0.07317, "merge": 0.1473, "skel": 0.12768}, "n_reads": 18965, "n_cand": 30479, "read_suppressed_frac": 0.0978, "dropout_reads_logged": 18965, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4690, "admitted_surface": 16820, "staged_surface": 0, "admitted_split": 16890, "staged_split": 0, "admitted_merge": 20646, "staged_merge": 0, "admitted_skel": 11180, "staged_skel": 0, "mean_count": 1270.478271484375, "n_util_obs": 648, "util_count_spearman": 0.03571111470909095}, "probe_overhead_sec_total": 440.012, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.007039, "alpha": 0.03, "t": 3808.3}
{"event": "consolidate", "step": 4350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3845.1}
{"event": "train", "step": 4350, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.39844, "bpb": 2.01752, "chunk_loss": 0.030599, "F": 0.21393, "G": 0.21609, "chunks": 219.06, "bytes_per_chunk": 4.674, "boundary_entropy": 0.08398, "router_noise": 0.0, "grad_norm": 0.6648, "lr": 8.21e-05, "router_trainable": true, "sec_per_step": 0.751, "probe_sec": 0.1202, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09521, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.01515, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": -0.03412, "n_merge": 2, "mean_delta_merge": -0.21918, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.26683, "trunc_abs_frac": 3.38572, "mean_delta_full": -0.11334, "read_hit_rate": {"surface": 0.64001, "split": 0.86584, "merge": 0.20155, "skel": 0.66475}, "mean_gate": {"surface": 0.17376, "split": 0.07217, "merge": 0.14695, "skel": 0.12638}, "n_reads": 19944, "n_cand": 31438, "read_suppressed_frac": 0.0967, "dropout_reads_logged": 19944, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4721, "admitted_surface": 16834, "staged_surface": 0, "admitted_split": 16786, "staged_split": 0, "admitted_merge": 20805, "staged_merge": 0, "admitted_skel": 11111, "staged_skel": 0, "mean_count": 1285.42041015625, "n_util_obs": 660, "util_count_spearman": 0.03823627273199502}, "probe_overhead_sec_total": 445.027, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.006263, "alpha": 0.03, "t": 3845.8}
{"event": "consolidate", "step": 4400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3882.0}
{"event": "probe", "step": 4400, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3563.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.204735, "cps": 27.835938, "mean_p": 0.204875, "boundary_entropy": 0.084418, "bpb": 1.904808}, "tr": {"chunks": 3569.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.19715, "cps": 27.882812, "mean_p": 0.189508, "boundary_entropy": 0.091184, "bpb": 2.056392}, "fi": {"chunks": 4590.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.244006, "cps": 35.859375, "mean_p": 0.238051, "boundary_entropy": 0.089611, "bpb": 2.150087}}, "bpb": {"en": 1.90481, "tr": 2.05639, "fi": 2.15009}, "gini_cpc": 0.048364, "gini_cps": 0.058409, "bpic": 4.839788, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9629512347477338, "fi": 1.1918156939154247}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4749, "admitted_surface": 16755, "staged_surface": 0, "admitted_split": 16794, "staged_split": 0, "admitted_merge": 20841, "staged_merge": 0, "admitted_skel": 11146, "staged_skel": 0, "mean_count": 1300.206298828125, "n_util_obs": 667, "util_count_spearman": 0.04446358935678172}, "t": 3883.8}
{"event": "eval", "step": 4400, "phase": 3, "bpb": {"en": 1.91707, "tr": 2.06436, "fi": 2.15886}, "bytes_per_chunk": {"en": 4.584, "tr": 5.643, "fi": 4.318}, "bpb_mean": 2.04676, "frozen_inventory": true, "t": 3886.7}
{"event": "train", "step": 4400, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.34933, "bpb": 1.94668, "chunk_loss": 0.030543, "F": 0.21448, "G": 0.21438, "chunks": 219.62, "bytes_per_chunk": 4.662, "boundary_entropy": 0.08304, "router_noise": 0.0, "grad_norm": 0.6027, "lr": 7.92e-05, "router_trainable": true, "sec_per_step": 0.858, "probe_sec": 0.0917, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06741, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.09084, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.65173, "n_merge": 1, "mean_delta_merge": 0.34644, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -10.34091, "trunc_abs_frac": 10.59615, "mean_delta_full": 0.01074, "read_hit_rate": {"surface": 0.6578, "split": 0.87508, "merge": 0.21468, "skel": 0.65894}, "mean_gate": {"surface": 0.17546, "split": 0.07297, "merge": 0.15064, "skel": 0.1281}, "n_reads": 20292, "n_cand": 31503, "read_suppressed_frac": 0.0998, "dropout_reads_logged": 20292, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4748, "admitted_surface": 16781, "staged_surface": 0, "admitted_split": 16754, "staged_split": 0, "admitted_merge": 20858, "staged_merge": 0, "admitted_skel": 11143, "staged_skel": 0, "mean_count": 1300.51318359375, "n_util_obs": 668, "util_count_spearman": 0.04827223273539113}, "probe_overhead_sec_total": 450.405, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.007307, "alpha": 0.03, "t": 3888.7}
{"event": "consolidate", "step": 4450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3924.4}
{"event": "train", "step": 4450, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.3467, "bpb": 1.94288, "chunk_loss": 0.030434, "F": 0.2056, "G": 0.20676, "chunks": 210.53, "bytes_per_chunk": 4.864, "boundary_entropy": 0.08387, "router_noise": 0.0, "grad_norm": 0.6637, "lr": 7.63e-05, "router_trainable": true, "sec_per_step": 0.726, "probe_sec": 0.077, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05709, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.01577, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.02539, "n_merge": 1, "mean_delta_merge": 0.03123, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.06876, "trunc_abs_frac": 0.34582, "mean_delta_full": -0.19987, "read_hit_rate": {"surface": 0.64549, "split": 0.85817, "merge": 0.21523, "skel": 0.63046}, "mean_gate": {"surface": 0.17485, "split": 0.0734, "merge": 0.14822, "skel": 0.1284}, "n_reads": 18969, "n_cand": 30120, "read_suppressed_frac": 0.0985, "dropout_reads_logged": 18969, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4783, "admitted_surface": 16791, "staged_surface": 0, "admitted_split": 16875, "staged_split": 0, "admitted_merge": 20644, "staged_merge": 0, "admitted_skel": 11226, "staged_skel": 0, "mean_count": 1315.64892578125, "n_util_obs": 673, "util_count_spearman": 0.049070159575768}, "probe_overhead_sec_total": 455.546, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.00648, "alpha": 0.03, "t": 3925.1}
{"event": "consolidate", "step": 4500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3963.6}
{"event": "train", "step": 4500, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32469, "bpb": 1.91113, "chunk_loss": 0.030657, "F": 0.21771, "G": 0.2193, "chunks": 222.94, "bytes_per_chunk": 4.593, "boundary_entropy": 0.08334, "router_noise": 0.0, "grad_norm": 0.6265, "lr": 7.36e-05, "router_trainable": true, "sec_per_step": 0.793, "probe_sec": 0.0787, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04612, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.04486, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.02417, "n_merge": 1, "mean_delta_merge": 0.36859, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -2.12401, "trunc_abs_frac": 1.01221, "mean_delta_full": -0.03864, "read_hit_rate": {"surface": 0.65615, "split": 0.87682, "merge": 0.21273, "skel": 0.66252}, "mean_gate": {"surface": 0.17346, "split": 0.07151, "merge": 0.14813, "skel": 0.12544}, "n_reads": 20602, "n_cand": 31973, "read_suppressed_frac": 0.0984, "dropout_reads_logged": 20602, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4818, "admitted_surface": 16836, "staged_surface": 0, "admitted_split": 16734, "staged_split": 0, "admitted_merge": 20798, "staged_merge": 0, "admitted_skel": 11168, "staged_skel": 0, "mean_count": 1330.61767578125, "n_util_obs": 682, "util_count_spearman": 0.047543588390995914}, "probe_overhead_sec_total": 460.377, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.006357, "alpha": 0.03, "t": 3964.7}
{"event": "consolidate", "step": 4550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4023.1}
{"event": "train", "step": 4550, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35343, "bpb": 1.95259, "chunk_loss": 0.030499, "F": 0.20782, "G": 0.21048, "chunks": 212.81, "bytes_per_chunk": 4.812, "boundary_entropy": 0.08155, "router_noise": 0.0, "grad_norm": 0.6574, "lr": 7.09e-05, "router_trainable": true, "sec_per_step": 1.207, "probe_sec": 0.231, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.12222, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.04743, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.14941, "n_merge": 2, "mean_delta_merge": -0.27502, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.44716, "trunc_abs_frac": 0.60488, "mean_delta_full": -0.29533, "read_hit_rate": {"surface": 0.62364, "split": 0.85276, "merge": 0.1976, "skel": 0.6285}, "mean_gate": {"surface": 0.16844, "split": 0.0693, "merge": 0.14322, "skel": 0.12231}, "n_reads": 18906, "n_cand": 30564, "read_suppressed_frac": 0.0999, "dropout_reads_logged": 18906, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4857, "admitted_surface": 16816, "staged_surface": 0, "admitted_split": 16890, "staged_split": 0, "admitted_merge": 20600, "staged_merge": 0, "admitted_skel": 11230, "staged_skel": 0, "mean_count": 1345.7198486328125, "n_util_obs": 690, "util_count_spearman": 0.04834110029115072}, "probe_overhead_sec_total": 466.836, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.006546, "alpha": 0.03, "t": 4025.0}
{"event": "consolidate", "step": 4600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4073.5}
{"event": "probe", "step": 4600, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3565.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.20485, "cps": 27.851562, "mean_p": 0.206136, "boundary_entropy": 0.085366, "bpb": 1.894291}, "tr": {"chunks": 3579.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.197702, "cps": 27.960938, "mean_p": 0.189061, "boundary_entropy": 0.086813, "bpb": 2.047489}, "fi": {"chunks": 4609.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.245016, "cps": 36.007812, "mean_p": 0.2398, "boundary_entropy": 0.089878, "bpb": 2.13789}}, "bpb": {"en": 1.89429, "tr": 2.04749, "fi": 2.13789}, "gini_cpc": 0.04871, "gini_cps": 0.059219, "bpic": 4.827023, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9651075941555873, "fi": 1.1960777477711968}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4890, "admitted_surface": 16836, "staged_surface": 0, "admitted_split": 16785, "staged_split": 0, "admitted_merge": 20761, "staged_merge": 0, "admitted_skel": 11154, "staged_skel": 0, "mean_count": 1360.51416015625, "n_util_obs": 694, "util_count_spearman": 0.04050355207587757}, "t": 4075.3}
{"event": "eval", "step": 4600, "phase": 3, "bpb": {"en": 1.90635, "tr": 2.05433, "fi": 2.14596}, "bytes_per_chunk": {"en": 4.579, "tr": 5.63, "fi": 4.305}, "bpb_mean": 2.03554, "frozen_inventory": true, "t": 4078.1}
{"event": "train", "step": 4600, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35377, "bpb": 1.95307, "chunk_loss": 0.030735, "F": 0.22098, "G": 0.22246, "chunks": 226.28, "bytes_per_chunk": 4.525, "boundary_entropy": 0.08628, "router_noise": 0.0, "grad_norm": 0.6582, "lr": 6.82e-05, "router_trainable": true, "sec_per_step": 1.102, "probe_sec": 0.1694, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.14235, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": 0.01114, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": -0.00737, "n_merge": 3, "mean_delta_merge": 0.0351, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 1.57256, "trunc_abs_frac": 1.57256, "mean_delta_full": -0.0614, "read_hit_rate": {"surface": 0.65931, "split": 0.8752, "merge": 0.23534, "skel": 0.65878}, "mean_gate": {"surface": 0.16944, "split": 0.07013, "merge": 0.14285, "skel": 0.1247}, "n_reads": 20780, "n_cand": 32107, "read_suppressed_frac": 0.1031, "dropout_reads_logged": 20780, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4890, "admitted_surface": 16852, "staged_surface": 0, "admitted_split": 16749, "staged_split": 0, "admitted_merge": 20763, "staged_merge": 0, "admitted_skel": 11172, "staged_skel": 0, "mean_count": 1360.834228515625, "n_util_obs": 694, "util_count_spearman": 0.04034244737764094}, "probe_overhead_sec_total": 472.485, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006958, "alpha": 0.03, "t": 4080.2}
{"event": "consolidate", "step": 4650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4123.7}
{"event": "train", "step": 4650, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.3884, "bpb": 2.00304, "chunk_loss": 0.030558, "F": 0.21207, "G": 0.21424, "chunks": 217.16, "bytes_per_chunk": 4.715, "boundary_entropy": 0.08363, "router_noise": 0.0, "grad_norm": 0.6681, "lr": 6.57e-05, "router_trainable": true, "sec_per_step": 0.89, "probe_sec": 0.0631, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0412, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.04856, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.04883, "n_merge": 1, "mean_delta_merge": 0.3689, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.84772, "trunc_abs_frac": 0.63084, "mean_delta_full": 0.05902, "read_hit_rate": {"surface": 0.63582, "split": 0.85886, "merge": 0.21336, "skel": 0.64413}, "mean_gate": {"surface": 0.17357, "split": 0.07116, "merge": 0.14812, "skel": 0.12679}, "n_reads": 19500, "n_cand": 30992, "read_suppressed_frac": 0.0983, "dropout_reads_logged": 19500, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4922, "admitted_surface": 16869, "staged_surface": 0, "admitted_split": 16886, "staged_split": 0, "admitted_merge": 20618, "staged_merge": 0, "admitted_skel": 11163, "staged_skel": 0, "mean_count": 1375.876953125, "n_util_obs": 702, "util_count_spearman": 0.04336092496110123}, "probe_overhead_sec_total": 477.778, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006184, "alpha": 0.03, "t": 4124.7}
{"event": "consolidate", "step": 4700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4160.9}
{"event": "train", "step": 4700, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.4058, "bpb": 2.02813, "chunk_loss": 0.030716, "F": 0.22003, "G": 0.22059, "chunks": 225.31, "bytes_per_chunk": 4.545, "boundary_entropy": 0.08285, "router_noise": 0.0, "grad_norm": 0.6404, "lr": 6.32e-05, "router_trainable": true, "sec_per_step": 0.738, "probe_sec": 0.181, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.15839, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.08985, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": -0.03912, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -2.75673, "trunc_abs_frac": 0.74259, "mean_delta_full": 0.02417, "read_hit_rate": {"surface": 0.65269, "split": 0.87679, "merge": 0.21676, "skel": 0.65218}, "mean_gate": {"surface": 0.16704, "split": 0.06869, "merge": 0.1424, "skel": 0.1232}, "n_reads": 20634, "n_cand": 32127, "read_suppressed_frac": 0.1004, "dropout_reads_logged": 20634, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4953, "admitted_surface": 16841, "staged_surface": 0, "admitted_split": 16692, "staged_split": 0, "admitted_merge": 20819, "staged_merge": 0, "admitted_skel": 11184, "staged_skel": 0, "mean_count": 1390.90234375, "n_util_obs": 708, "util_count_spearman": 0.04007764657923737}, "probe_overhead_sec_total": 482.723, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006252, "alpha": 0.03, "t": 4161.6}
{"event": "consolidate", "step": 4750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4198.9}
{"event": "train", "step": 4750, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.4, "bpb": 2.01978, "chunk_loss": 0.030607, "F": 0.21515, "G": 0.21603, "chunks": 220.31, "bytes_per_chunk": 4.648, "boundary_entropy": 0.08295, "router_noise": 0.0, "grad_norm": 0.6566, "lr": 6.08e-05, "router_trainable": true, "sec_per_step": 0.759, "probe_sec": 0.0702, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04905, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.04624, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": -0.26569, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.45085, "trunc_abs_frac": 0.73767, "mean_delta_full": -0.30396, "read_hit_rate": {"surface": 0.64121, "split": 0.86215, "merge": 0.20784, "skel": 0.6472}, "mean_gate": {"surface": 0.16851, "split": 0.06921, "merge": 0.14431, "skel": 0.12304}, "n_reads": 19751, "n_cand": 31333, "read_suppressed_frac": 0.1003, "dropout_reads_logged": 19751, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4981, "admitted_surface": 16846, "staged_surface": 0, "admitted_split": 16829, "staged_split": 0, "admitted_merge": 20709, "staged_merge": 0, "admitted_skel": 11152, "staged_skel": 0, "mean_count": 1405.879638671875, "n_util_obs": 720, "util_count_spearman": 0.029933458831368782}, "probe_overhead_sec_total": 487.534, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006591, "alpha": 0.03, "t": 4199.5}
{"event": "consolidate", "step": 4800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4237.1}
{"event": "probe", "step": 4800, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3539.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.203356, "cps": 27.648438, "mean_p": 0.203623, "boundary_entropy": 0.084743, "bpb": 1.887177}, "tr": {"chunks": 3573.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.197371, "cps": 27.914062, "mean_p": 0.188428, "boundary_entropy": 0.088251, "bpb": 2.03232}, "fi": {"chunks": 4582.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.243581, "cps": 35.796875, "mean_p": 0.237631, "boundary_entropy": 0.090329, "bpb": 2.125582}}, "bpb": {"en": 1.88718, "tr": 2.03232, "fi": 2.12558}, "gini_cpc": 0.047814, "gini_cps": 0.059461, "bpic": 4.851377, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9705681206300008, "fi": 1.1978067578400564}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5014, "admitted_surface": 16826, "staged_surface": 0, "admitted_split": 16785, "staged_split": 0, "admitted_merge": 20718, "staged_merge": 0, "admitted_skel": 11207, "staged_skel": 0, "mean_count": 1420.5970458984375, "n_util_obs": 728, "util_count_spearman": 0.028348950605829502}, "t": 4238.8}
{"event": "eval", "step": 4800, "phase": 3, "bpb": {"en": 1.89994, "tr": 2.04139, "fi": 2.13193}, "bytes_per_chunk": {"en": 4.62, "tr": 5.635, "fi": 4.344}, "bpb_mean": 2.02442, "frozen_inventory": true, "t": 4241.7}
{"event": "train", "step": 4800, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.37373, "bpb": 1.98187, "chunk_loss": 0.030662, "F": 0.21942, "G": 0.2195, "chunks": 224.69, "bytes_per_chunk": 4.557, "boundary_entropy": 0.08462, "router_noise": 0.0, "grad_norm": 0.6492, "lr": 5.85e-05, "router_trainable": true, "sec_per_step": 0.875, "probe_sec": 0.104, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08269, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.02369, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.45587, "n_merge": 1, "mean_delta_merge": -0.25854, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.63776, "trunc_abs_frac": 0.57152, "mean_delta_full": 0.02332, "read_hit_rate": {"surface": 0.65465, "split": 0.87372, "merge": 0.20951, "skel": 0.65162}, "mean_gate": {"surface": 0.16935, "split": 0.06916, "merge": 0.14218, "skel": 0.12495}, "n_reads": 20396, "n_cand": 31924, "read_suppressed_frac": 0.102, "dropout_reads_logged": 20396, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5015, "admitted_surface": 16824, "staged_surface": 0, "admitted_split": 16709, "staged_split": 0, "admitted_merge": 20829, "staged_merge": 0, "admitted_skel": 11174, "staged_skel": 0, "mean_count": 1420.9013671875, "n_util_obs": 728, "util_count_spearman": 0.028764924165147632}, "probe_overhead_sec_total": 492.85, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006583, "alpha": 0.03, "t": 4243.3}
{"event": "consolidate", "step": 4850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4280.1}
{"event": "train", "step": 4850, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.34256, "bpb": 1.9369, "chunk_loss": 0.030576, "F": 0.21225, "G": 0.2135, "chunks": 217.34, "bytes_per_chunk": 4.711, "boundary_entropy": 0.08264, "router_noise": 0.0, "grad_norm": 0.6738, "lr": 5.62e-05, "router_trainable": true, "sec_per_step": 0.747, "probe_sec": 0.0796, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05907, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.03415, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.40009, "n_merge": 1, "mean_delta_merge": -0.19423, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 2.40887, "trunc_abs_frac": 0.53506, "mean_delta_full": 0.02545, "read_hit_rate": {"surface": 0.6503, "split": 0.85824, "merge": 0.20955, "skel": 0.6464}, "mean_gate": {"surface": 0.17159, "split": 0.06982, "merge": 0.14599, "skel": 0.12562}, "n_reads": 19670, "n_cand": 31120, "read_suppressed_frac": 0.1012, "dropout_reads_logged": 19670, "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": 16789, "staged_surface": 0, "admitted_split": 16856, "staged_split": 0, "admitted_merge": 20699, "staged_merge": 0, "admitted_skel": 11192, "staged_skel": 0, "mean_count": 1436.073974609375, "n_util_obs": 734, "util_count_spearman": 0.023494238327966956}, "probe_overhead_sec_total": 498.666, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006695, "alpha": 0.03, "t": 4280.6}
{"event": "consolidate", "step": 4900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4317.0}
{"event": "train", "step": 4900, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.37853, "bpb": 1.9888, "chunk_loss": 0.030612, "F": 0.20923, "G": 0.21042, "chunks": 214.25, "bytes_per_chunk": 4.779, "boundary_entropy": 0.08058, "router_noise": 0.0, "grad_norm": 0.6729, "lr": 5.41e-05, "router_trainable": true, "sec_per_step": 0.741, "probe_sec": 0.1507, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.1296, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.03684, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": 0.11433, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -4.12915, "trunc_abs_frac": 18.7272, "mean_delta_full": -0.04285, "read_hit_rate": {"surface": 0.63159, "split": 0.85245, "merge": 0.20716, "skel": 0.63744}, "mean_gate": {"surface": 0.16549, "split": 0.06654, "merge": 0.14014, "skel": 0.12017}, "n_reads": 19079, "n_cand": 30614, "read_suppressed_frac": 0.1007, "dropout_reads_logged": 19079, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5080, "admitted_surface": 16839, "staged_surface": 0, "admitted_split": 16907, "staged_split": 0, "admitted_merge": 20602, "staged_merge": 0, "admitted_skel": 11188, "staged_skel": 0, "mean_count": 1451.25, "n_util_obs": 747, "util_count_spearman": 0.006048912314219741}, "probe_overhead_sec_total": 503.851, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006372, "alpha": 0.03, "t": 4317.7}
{"event": "consolidate", "step": 4950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4353.5}
{"event": "train", "step": 4950, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.33066, "bpb": 1.91973, "chunk_loss": 0.030629, "F": 0.21426, "G": 0.21574, "chunks": 219.41, "bytes_per_chunk": 4.667, "boundary_entropy": 0.08184, "router_noise": 0.0, "grad_norm": 0.6725, "lr": 5.2e-05, "router_trainable": true, "sec_per_step": 0.74, "probe_sec": 0.0668, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04569, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.00335, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": -0.17897, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -2.68509, "trunc_abs_frac": 1.58322, "mean_delta_full": -0.04325, "read_hit_rate": {"surface": 0.65989, "split": 0.8664, "merge": 0.2238, "skel": 0.64167}, "mean_gate": {"surface": 0.17174, "split": 0.06949, "merge": 0.14555, "skel": 0.12564}, "n_reads": 19770, "n_cand": 30990, "read_suppressed_frac": 0.0975, "dropout_reads_logged": 19770, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5111, "admitted_surface": 16812, "staged_surface": 0, "admitted_split": 16846, "staged_split": 0, "admitted_merge": 20690, "staged_merge": 0, "admitted_skel": 11188, "staged_skel": 0, "mean_count": 1466.151123046875, "n_util_obs": 754, "util_count_spearman": -0.006386427608188324}, "probe_overhead_sec_total": 508.865, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006328, "alpha": 0.03, "t": 4354.7}
{"event": "consolidate", "step": 5000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4390.8}
{"event": "probe", "step": 5000, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3544.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.203643, "cps": 27.6875, "mean_p": 0.203651, "boundary_entropy": 0.082179, "bpb": 1.880126}, "tr": {"chunks": 3585.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.198033, "cps": 28.007812, "mean_p": 0.188475, "boundary_entropy": 0.084803, "bpb": 2.024465}, "fi": {"chunks": 4598.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.244431, "cps": 35.921875, "mean_p": 0.237967, "boundary_entropy": 0.083778, "bpb": 2.113776}}, "bpb": {"en": 1.88013, "tr": 2.02446, "fi": 2.11378}, "gini_cpc": 0.047874, "gini_cps": 0.059919, "bpic": 4.837725, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9724538847121232, "fi": 1.200293600125665}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5144, "admitted_surface": 16827, "staged_surface": 0, "admitted_split": 16848, "staged_split": 0, "admitted_merge": 20640, "staged_merge": 0, "admitted_skel": 11221, "staged_skel": 0, "mean_count": 1480.8243408203125, "n_util_obs": 760, "util_count_spearman": -0.009498973432953991}, "t": 4392.4}
{"event": "eval", "step": 5000, "phase": 3, "bpb": {"en": 1.89347, "tr": 2.03326, "fi": 2.12125}, "bytes_per_chunk": {"en": 4.624, "tr": 5.63, "fi": 4.319}, "bpb_mean": 2.01599, "frozen_inventory": true, "t": 4395.3}
{"event": "train", "step": 5000, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32206, "bpb": 1.90733, "chunk_loss": 0.030628, "F": 0.21713, "G": 0.21686, "chunks": 222.34, "bytes_per_chunk": 4.605, "boundary_entropy": 0.08017, "router_noise": 0.0, "grad_norm": 0.603, "lr": 5e-05, "router_trainable": true, "sec_per_step": 0.844, "probe_sec": 0.089, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05736, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": -0.02996, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.05603, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.41165, "trunc_abs_frac": 0.5814, "mean_delta_full": -0.1876, "read_hit_rate": {"surface": 0.65763, "split": 0.87492, "merge": 0.2221, "skel": 0.64951}, "mean_gate": {"surface": 0.16423, "split": 0.06611, "merge": 0.14078, "skel": 0.11947}, "n_reads": 20184, "n_cand": 31446, "read_suppressed_frac": 0.0995, "dropout_reads_logged": 20184, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5144, "admitted_surface": 16834, "staged_surface": 0, "admitted_split": 16740, "staged_split": 0, "admitted_merge": 20772, "staged_merge": 0, "admitted_skel": 11190, "staged_skel": 0, "mean_count": 1481.1279296875, "n_util_obs": 760, "util_count_spearman": -0.009122688272078065}, "probe_overhead_sec_total": 513.589, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006689, "alpha": 0.03, "t": 4396.9}
{"event": "consolidate", "step": 5050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4433.1}
{"event": "train", "step": 5050, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.33894, "bpb": 1.93168, "chunk_loss": 0.030611, "F": 0.21533, "G": 0.21578, "chunks": 220.5, "bytes_per_chunk": 4.644, "boundary_entropy": 0.08048, "router_noise": 0.0, "grad_norm": 0.6554, "lr": 4.81e-05, "router_trainable": true, "sec_per_step": 0.739, "probe_sec": 0.1408, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.11981, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.04579, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 3, "mean_delta_merge": 0.04463, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": -0.01321, "trunc_abs_frac": 0.01321, "mean_delta_full": 0.14673, "read_hit_rate": {"surface": 0.65376, "split": 0.87237, "merge": 0.21239, "skel": 0.65424}, "mean_gate": {"surface": 0.16807, "split": 0.06789, "merge": 0.14217, "skel": 0.12407}, "n_reads": 20270, "n_cand": 31627, "read_suppressed_frac": 0.0964, "dropout_reads_logged": 20270, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5182, "admitted_surface": 16814, "staged_surface": 0, "admitted_split": 16776, "staged_split": 0, "admitted_merge": 20771, "staged_merge": 0, "admitted_skel": 11175, "staged_skel": 0, "mean_count": 1496.0924072265625, "n_util_obs": 764, "util_count_spearman": -0.015593760859572519}, "probe_overhead_sec_total": 518.243, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006604, "alpha": 0.03, "t": 4433.8}
{"event": "consolidate", "step": 5100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4474.4}
{"event": "train", "step": 5100, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.37043, "bpb": 1.97711, "chunk_loss": 0.030611, "F": 0.2128, "G": 0.21483, "chunks": 217.91, "bytes_per_chunk": 4.699, "boundary_entropy": 0.0799, "router_noise": 0.0, "grad_norm": 0.6817, "lr": 4.63e-05, "router_trainable": true, "sec_per_step": 0.824, "probe_sec": 0.1049, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08024, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.04146, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 3, "mean_delta_split": 0.01837, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.10728, "trunc_abs_frac": 0.63369, "mean_delta_full": -0.17545, "read_hit_rate": {"surface": 0.6306, "split": 0.86147, "merge": 0.21059, "skel": 0.63563}, "mean_gate": {"surface": 0.1642, "split": 0.06647, "merge": 0.13596, "skel": 0.1207}, "n_reads": 19410, "n_cand": 30977, "read_suppressed_frac": 0.1021, "dropout_reads_logged": 19410, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5217, "admitted_surface": 16892, "staged_surface": 0, "admitted_split": 16844, "staged_split": 0, "admitted_merge": 20598, "staged_merge": 0, "admitted_skel": 11202, "staged_skel": 0, "mean_count": 1510.974853515625, "n_util_obs": 768, "util_count_spearman": -0.016595842429817976}, "probe_overhead_sec_total": 523.248, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006901, "alpha": 0.03, "t": 4475.0}
{"event": "consolidate", "step": 5150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4513.0}
{"event": "train", "step": 5150, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.3622, "bpb": 1.96524, "chunk_loss": 0.030696, "F": 0.21857, "G": 0.21887, "chunks": 223.81, "bytes_per_chunk": 4.575, "boundary_entropy": 0.08011, "router_noise": 0.0, "grad_norm": 0.6724, "lr": 4.45e-05, "router_trainable": true, "sec_per_step": 0.78, "probe_sec": 0.0814, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05775, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.02424, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": 0.14166, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -1.01088, "trunc_abs_frac": 2.17189, "mean_delta_full": 0.03813, "read_hit_rate": {"surface": 0.65708, "split": 0.86505, "merge": 0.21978, "skel": 0.64174}, "mean_gate": {"surface": 0.1668, "split": 0.06703, "merge": 0.14167, "skel": 0.12242}, "n_reads": 20073, "n_cand": 31570, "read_suppressed_frac": 0.0987, "dropout_reads_logged": 20073, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5244, "admitted_surface": 16806, "staged_surface": 0, "admitted_split": 16841, "staged_split": 0, "admitted_merge": 20713, "staged_merge": 0, "admitted_skel": 11176, "staged_skel": 0, "mean_count": 1525.87109375, "n_util_obs": 771, "util_count_spearman": -0.020513058777822436}, "probe_overhead_sec_total": 527.981, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006055, "alpha": 0.03, "t": 4514.0}
{"event": "consolidate", "step": 5200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4563.5}
{"event": "probe", "step": 5200, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3542.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.203528, "cps": 27.671875, "mean_p": 0.204031, "boundary_entropy": 0.082847, "bpb": 1.873081}, "tr": {"chunks": 3585.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.198033, "cps": 28.007812, "mean_p": 0.188448, "boundary_entropy": 0.085248, "bpb": 2.016622}, "fi": {"chunks": 4595.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.244272, "cps": 35.898438, "mean_p": 0.238243, "boundary_entropy": 0.083436, "bpb": 2.105184}}, "bpb": {"en": 1.87308, "tr": 2.01662, "fi": 2.10518}, "gini_cpc": 0.04773, "gini_cps": 0.059887, "bpic": 4.839788, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9730029834612548, "fi": 1.2001877663215963}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5286, "admitted_surface": 16833, "staged_surface": 0, "admitted_split": 16826, "staged_split": 0, "admitted_merge": 20693, "staged_merge": 0, "admitted_skel": 11184, "staged_skel": 0, "mean_count": 1540.42626953125, "n_util_obs": 778, "util_count_spearman": -0.01720988131914639}, "t": 4565.4}
{"event": "eval", "step": 5200, "phase": 3, "bpb": {"en": 1.88737, "tr": 2.02572, "fi": 2.11206}, "bytes_per_chunk": {"en": 4.627, "tr": 5.624, "fi": 4.323}, "bpb_mean": 2.00838, "frozen_inventory": true, "t": 4568.1}
{"event": "train", "step": 5200, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.31953, "bpb": 1.90367, "chunk_loss": 0.030537, "F": 0.21179, "G": 0.21321, "chunks": 216.88, "bytes_per_chunk": 4.722, "boundary_entropy": 0.08064, "router_noise": 0.0, "grad_norm": 0.6567, "lr": 4.29e-05, "router_trainable": true, "sec_per_step": 1.119, "probe_sec": 0.084, "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": -0.13729, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": -0.01404, "n_merge": 1, "mean_delta_merge": -0.01086, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.50607, "trunc_abs_frac": 0.50607, "mean_delta_full": -0.29163, "read_hit_rate": {"surface": 0.64693, "split": 0.8652, "merge": 0.21001, "skel": 0.66759}, "mean_gate": {"surface": 0.16221, "split": 0.06443, "merge": 0.13539, "skel": 0.11831}, "n_reads": 19983, "n_cand": 31287, "read_suppressed_frac": 0.0997, "dropout_reads_logged": 19983, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5288, "admitted_surface": 16876, "staged_surface": 0, "admitted_split": 16840, "staged_split": 0, "admitted_merge": 20658, "staged_merge": 0, "admitted_skel": 11162, "staged_skel": 0, "mean_count": 1540.7318115234375, "n_util_obs": 778, "util_count_spearman": -0.01716276405179806}, "probe_overhead_sec_total": 533.445, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006418, "alpha": 0.03, "t": 4570.0}
{"event": "consolidate", "step": 5250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4618.8}
{"event": "train", "step": 5250, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.31026, "bpb": 1.89031, "chunk_loss": 0.03056, "F": 0.21225, "G": 0.21302, "chunks": 217.34, "bytes_per_chunk": 4.711, "boundary_entropy": 0.07868, "router_noise": 0.0, "grad_norm": 0.6245, "lr": 4.14e-05, "router_trainable": true, "sec_per_step": 0.998, "probe_sec": 0.0937, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05314, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.03235, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": 0.06459, "n_merge": 1, "mean_delta_merge": 0.13702, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -0.34659, "trunc_abs_frac": 1.31992, "mean_delta_full": -0.08755, "read_hit_rate": {"surface": 0.65535, "split": 0.86987, "merge": 0.21318, "skel": 0.65146}, "mean_gate": {"surface": 0.16257, "split": 0.06525, "merge": 0.13691, "skel": 0.11925}, "n_reads": 19870, "n_cand": 31089, "read_suppressed_frac": 0.1014, "dropout_reads_logged": 19870, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5320, "admitted_surface": 16814, "staged_surface": 0, "admitted_split": 16805, "staged_split": 0, "admitted_merge": 20691, "staged_merge": 0, "admitted_skel": 11226, "staged_skel": 0, "mean_count": 1555.711181640625, "n_util_obs": 784, "util_count_spearman": -0.012862778574104873}, "probe_overhead_sec_total": 538.688, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006065, "alpha": 0.03, "t": 4619.9}
{"event": "consolidate", "step": 5300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4670.4}
{"event": "train", "step": 5300, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.36746, "bpb": 1.97283, "chunk_loss": 0.030609, "F": 0.21356, "G": 0.21481, "chunks": 218.69, "bytes_per_chunk": 4.682, "boundary_entropy": 0.07882, "router_noise": 0.0, "grad_norm": 0.6601, "lr": 3.99e-05, "router_trainable": true, "sec_per_step": 1.023, "probe_sec": 0.1078, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08311, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.0428, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": -0.04807, "n_merge": 1, "mean_delta_merge": -0.03967, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 0.89881, "trunc_abs_frac": 2.01467, "mean_delta_full": 0.06132, "read_hit_rate": {"surface": 0.63997, "split": 0.86474, "merge": 0.20782, "skel": 0.65733}, "mean_gate": {"surface": 0.16089, "split": 0.06302, "merge": 0.13395, "skel": 0.11624}, "n_reads": 19934, "n_cand": 31440, "read_suppressed_frac": 0.1006, "dropout_reads_logged": 19934, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5355, "admitted_surface": 16870, "staged_surface": 0, "admitted_split": 16841, "staged_split": 0, "admitted_merge": 20650, "staged_merge": 0, "admitted_skel": 11175, "staged_skel": 0, "mean_count": 1570.6502685546875, "n_util_obs": 791, "util_count_spearman": -0.01041182073784941}, "probe_overhead_sec_total": 544.028, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006249, "alpha": 0.03, "t": 4671.0}
{"event": "consolidate", "step": 5350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4708.9}
{"event": "train", "step": 5350, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.30345, "bpb": 1.88048, "chunk_loss": 0.030459, "F": 0.20877, "G": 0.2097, "chunks": 213.78, "bytes_per_chunk": 4.79, "boundary_entropy": 0.07957, "router_noise": 0.0, "grad_norm": 0.7069, "lr": 3.86e-05, "router_trainable": true, "sec_per_step": 0.767, "probe_sec": 0.071, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0439, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.06599, "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": 2.37623, "trunc_abs_frac": 5.11475, "mean_delta_full": -0.05289, "read_hit_rate": {"surface": 0.64004, "split": 0.86903, "merge": 0.21057, "skel": 0.65708}, "mean_gate": {"surface": 0.16107, "split": 0.06321, "merge": 0.13444, "skel": 0.11644}, "n_reads": 19634, "n_cand": 30828, "read_suppressed_frac": 0.0995, "dropout_reads_logged": 19634, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5381, "admitted_surface": 16872, "staged_surface": 0, "admitted_split": 16852, "staged_split": 0, "admitted_merge": 20626, "staged_merge": 0, "admitted_skel": 11186, "staged_skel": 0, "mean_count": 1585.585205078125, "n_util_obs": 798, "util_count_spearman": -0.005326643650415999}, "probe_overhead_sec_total": 549.201, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006315, "alpha": 0.03, "t": 4709.4}
{"event": "consolidate", "step": 5400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4746.6}
{"event": "probe", "step": 5400, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3534.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.203068, "cps": 27.609375, "mean_p": 0.203339, "boundary_entropy": 0.081518, "bpb": 1.868468}, "tr": {"chunks": 3575.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.197481, "cps": 27.929688, "mean_p": 0.18767, "boundary_entropy": 0.083989, "bpb": 2.010236}, "fi": {"chunks": 4585.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.24374, "cps": 35.820312, "mean_p": 0.237796, "boundary_entropy": 0.082201, "bpb": 2.099022}}, "bpb": {"en": 1.86847, "tr": 2.01024, "fi": 2.09902}, "gini_cpc": 0.047866, "gini_cps": 0.059917, "bpic": 4.851377, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9724853547428612, "fi": 1.2002868043379236}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5410, "admitted_surface": 16812, "staged_surface": 0, "admitted_split": 16836, "staged_split": 0, "admitted_merge": 20637, "staged_merge": 0, "admitted_skel": 11251, "staged_skel": 0, "mean_count": 1600.204833984375, "n_util_obs": 809, "util_count_spearman": 0.001225827177099816}, "t": 4748.3}
{"event": "eval", "step": 5400, "phase": 3, "bpb": {"en": 1.88233, "tr": 2.01978, "fi": 2.10667}, "bytes_per_chunk": {"en": 4.636, "tr": 5.651, "fi": 4.32}, "bpb_mean": 2.00293, "frozen_inventory": true, "t": 4751.1}
{"event": "train", "step": 5400, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.37519, "bpb": 1.98398, "chunk_loss": 0.030795, "F": 0.21896, "G": 0.22042, "chunks": 224.22, "bytes_per_chunk": 4.567, "boundary_entropy": 0.07966, "router_noise": 0.0, "grad_norm": 0.7189, "lr": 3.73e-05, "router_trainable": true, "sec_per_step": 0.867, "probe_sec": 0.0874, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06425, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.05575, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.26001, "n_merge": 2, "mean_delta_merge": -0.32147, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.11291, "trunc_abs_frac": 0.46952, "mean_delta_full": -0.3112, "read_hit_rate": {"surface": 0.64525, "split": 0.86554, "merge": 0.20335, "skel": 0.65248}, "mean_gate": {"surface": 0.1607, "split": 0.06305, "merge": 0.13471, "skel": 0.11583}, "n_reads": 20151, "n_cand": 31870, "read_suppressed_frac": 0.099, "dropout_reads_logged": 20151, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5411, "admitted_surface": 16825, "staged_surface": 0, "admitted_split": 16814, "staged_split": 0, "admitted_merge": 20742, "staged_merge": 0, "admitted_skel": 11155, "staged_skel": 0, "mean_count": 1600.488037109375, "n_util_obs": 809, "util_count_spearman": 0.0012078886297469493}, "probe_overhead_sec_total": 554.024, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.00703, "alpha": 0.03, "t": 4752.8}
{"event": "consolidate", "step": 5450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4788.6}
{"event": "train", "step": 5450, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.30632, "bpb": 1.88462, "chunk_loss": 0.030683, "F": 0.21667, "G": 0.21774, "chunks": 221.88, "bytes_per_chunk": 4.615, "boundary_entropy": 0.08074, "router_noise": 0.0, "grad_norm": 0.6802, "lr": 3.62e-05, "router_trainable": true, "sec_per_step": 0.729, "probe_sec": 0.0828, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05892, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": 0.37744, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 4, "mean_delta_split": 0.18276, "n_merge": 1, "mean_delta_merge": -0.04346, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.02524, "trunc_abs_frac": 0.35232, "mean_delta_full": 0.14447, "read_hit_rate": {"surface": 0.66242, "split": 0.86901, "merge": 0.22186, "skel": 0.66504}, "mean_gate": {"surface": 0.16459, "split": 0.06463, "merge": 0.13831, "skel": 0.11889}, "n_reads": 20310, "n_cand": 31545, "read_suppressed_frac": 0.0994, "dropout_reads_logged": 20310, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5442, "admitted_surface": 16818, "staged_surface": 0, "admitted_split": 16870, "staged_split": 0, "admitted_merge": 20664, "staged_merge": 0, "admitted_skel": 11184, "staged_skel": 0, "mean_count": 1615.4873046875, "n_util_obs": 815, "util_count_spearman": -0.008258634187992386}, "probe_overhead_sec_total": 558.911, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006363, "alpha": 0.03, "t": 4789.2}
{"event": "consolidate", "step": 5500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4826.4}
{"event": "train", "step": 5500, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.2976, "bpb": 1.87205, "chunk_loss": 0.030538, "F": 0.21313, "G": 0.21408, "chunks": 218.25, "bytes_per_chunk": 4.692, "boundary_entropy": 0.07999, "router_noise": 0.0, "grad_norm": 0.6681, "lr": 3.51e-05, "router_trainable": true, "sec_per_step": 0.754, "probe_sec": 0.0852, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06425, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.01845, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.06323, "n_merge": 1, "mean_delta_merge": 0.03687, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -0.5678, "trunc_abs_frac": 1.09096, "mean_delta_full": -0.13953, "read_hit_rate": {"surface": 0.64413, "split": 0.86862, "merge": 0.22211, "skel": 0.64468}, "mean_gate": {"surface": 0.15993, "split": 0.06298, "merge": 0.1356, "skel": 0.11568}, "n_reads": 19908, "n_cand": 31242, "read_suppressed_frac": 0.0981, "dropout_reads_logged": 19908, "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": 16867, "staged_surface": 0, "admitted_split": 16854, "staged_split": 0, "admitted_merge": 20566, "staged_merge": 0, "admitted_skel": 11249, "staged_skel": 0, "mean_count": 1630.4200439453125, "n_util_obs": 822, "util_count_spearman": -0.00041468719482250617}, "probe_overhead_sec_total": 564.512, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006404, "alpha": 0.03, "t": 4826.9}
{"event": "consolidate", "step": 5550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4864.1}
{"event": "train", "step": 5550, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35142, "bpb": 1.94969, "chunk_loss": 0.030489, "F": 0.20697, "G": 0.20804, "chunks": 211.94, "bytes_per_chunk": 4.832, "boundary_entropy": 0.0785, "router_noise": 0.0, "grad_norm": 0.6796, "lr": 3.41e-05, "router_trainable": true, "sec_per_step": 0.764, "probe_sec": 0.082, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05666, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.09246, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": -0.19286, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.31566, "trunc_abs_frac": 8.93864, "mean_delta_full": -0.12259, "read_hit_rate": {"surface": 0.63941, "split": 0.85619, "merge": 0.21346, "skel": 0.62845}, "mean_gate": {"surface": 0.16551, "split": 0.0656, "merge": 0.14069, "skel": 0.12079}, "n_reads": 18819, "n_cand": 30073, "read_suppressed_frac": 0.103, "dropout_reads_logged": 18819, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5499, "admitted_surface": 16827, "staged_surface": 0, "admitted_split": 16851, "staged_split": 0, "admitted_merge": 20607, "staged_merge": 0, "admitted_skel": 11251, "staged_skel": 0, "mean_count": 1645.271484375, "n_util_obs": 833, "util_count_spearman": 0.017951279329004553}, "probe_overhead_sec_total": 569.129, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006818, "alpha": 0.03, "t": 4865.2}
{"event": "consolidate", "step": 5600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4901.7}
{"event": "probe", "step": 5600, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3532.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.202954, "cps": 27.59375, "mean_p": 0.203147, "boundary_entropy": 0.081935, "bpb": 1.864559}, "tr": {"chunks": 3589.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.198254, "cps": 28.039062, "mean_p": 0.188399, "boundary_entropy": 0.081496, "bpb": 2.004745}, "fi": {"chunks": 4603.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.244697, "cps": 35.960938, "mean_p": 0.237831, "boundary_entropy": 0.082414, "bpb": 2.089986}}, "bpb": {"en": 1.86456, "tr": 2.00474, "fi": 2.08999}, "gini_cpc": 0.047936, "gini_cps": 0.060901, "bpic": 4.838963, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.976846516682662, "fi": 1.2056812768221383}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5530, "admitted_surface": 16833, "staged_surface": 0, "admitted_split": 16828, "staged_split": 0, "admitted_merge": 20696, "staged_merge": 0, "admitted_skel": 11179, "staged_skel": 0, "mean_count": 1659.8890380859375, "n_util_obs": 839, "util_count_spearman": 0.01620930678612751}, "t": 4903.4}
{"event": "eval", "step": 5600, "phase": 3, "bpb": {"en": 1.87866, "tr": 2.01365, "fi": 2.09788}, "bytes_per_chunk": {"en": 4.64, "tr": 5.622, "fi": 4.326}, "bpb_mean": 1.99673, "frozen_inventory": true, "t": 4906.2}
{"event": "train", "step": 5600, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.34142, "bpb": 1.93526, "chunk_loss": 0.030607, "F": 0.21423, "G": 0.21548, "chunks": 219.38, "bytes_per_chunk": 4.668, "boundary_entropy": 0.08017, "router_noise": 0.0, "grad_norm": 0.7054, "lr": 3.33e-05, "router_trainable": true, "sec_per_step": 0.855, "probe_sec": 0.1228, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07772, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.09994, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": -0.01616, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 7.63275, "trunc_abs_frac": 1.58026, "mean_delta_full": 0.00601, "read_hit_rate": {"surface": 0.63938, "split": 0.86172, "merge": 0.20687, "skel": 0.66349}, "mean_gate": {"surface": 0.16053, "split": 0.0628, "merge": 0.13646, "skel": 0.11531}, "n_reads": 19928, "n_cand": 31470, "read_suppressed_frac": 0.0982, "dropout_reads_logged": 19928, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5532, "admitted_surface": 16853, "staged_surface": 0, "admitted_split": 16852, "staged_split": 0, "admitted_merge": 20660, "staged_merge": 0, "admitted_skel": 11171, "staged_skel": 0, "mean_count": 1660.1895751953125, "n_util_obs": 839, "util_count_spearman": 0.016125654683235402}, "probe_overhead_sec_total": 573.907, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006304, "alpha": 0.03, "t": 4907.9}
{"event": "consolidate", "step": 5650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4944.6}
{"event": "train", "step": 5650, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32692, "bpb": 1.91435, "chunk_loss": 0.030645, "F": 0.21606, "G": 0.21681, "chunks": 221.25, "bytes_per_chunk": 4.628, "boundary_entropy": 0.07808, "router_noise": 0.0, "grad_norm": 0.6783, "lr": 3.25e-05, "router_trainable": true, "sec_per_step": 0.747, "probe_sec": 0.1146, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09206, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.03842, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.22722, "n_merge": 1, "mean_delta_merge": -0.00183, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -0.93116, "trunc_abs_frac": 1.81905, "mean_delta_full": -0.04089, "read_hit_rate": {"surface": 0.65011, "split": 0.86883, "merge": 0.2205, "skel": 0.65318}, "mean_gate": {"surface": 0.15745, "split": 0.06199, "merge": 0.13113, "skel": 0.11463}, "n_reads": 20175, "n_cand": 31556, "read_suppressed_frac": 0.0989, "dropout_reads_logged": 20175, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5555, "admitted_surface": 16859, "staged_surface": 0, "admitted_split": 16809, "staged_split": 0, "admitted_merge": 20670, "staged_merge": 0, "admitted_skel": 11198, "staged_skel": 0, "mean_count": 1675.093017578125, "n_util_obs": 850, "util_count_spearman": 0.014515228189907104}, "probe_overhead_sec_total": 579.108, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006794, "alpha": 0.03, "t": 4945.3}
{"event": "consolidate", "step": 5700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4981.6}
{"event": "train", "step": 5700, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.33546, "bpb": 1.92667, "chunk_loss": 0.03049, "F": 0.20987, "G": 0.21132, "chunks": 214.91, "bytes_per_chunk": 4.765, "boundary_entropy": 0.07672, "router_noise": 0.0, "grad_norm": 0.7147, "lr": 3.18e-05, "router_trainable": true, "sec_per_step": 0.746, "probe_sec": 0.1324, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10914, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.02609, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.49005, "n_merge": 1, "mean_delta_merge": -0.21576, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -0.02017, "trunc_abs_frac": 1.01985, "mean_delta_full": -0.19675, "read_hit_rate": {"surface": 0.6504, "split": 0.86377, "merge": 0.21327, "skel": 0.64404}, "mean_gate": {"surface": 0.15557, "split": 0.06177, "merge": 0.1303, "skel": 0.1139}, "n_reads": 19538, "n_cand": 30779, "read_suppressed_frac": 0.1009, "dropout_reads_logged": 19538, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5584, "admitted_surface": 16813, "staged_surface": 0, "admitted_split": 16851, "staged_split": 0, "admitted_merge": 20647, "staged_merge": 0, "admitted_skel": 11225, "staged_skel": 0, "mean_count": 1690.04345703125, "n_util_obs": 853, "util_count_spearman": 0.008544502240017512}, "probe_overhead_sec_total": 584.276, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006489, "alpha": 0.03, "t": 4982.6}
{"event": "consolidate", "step": 5750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 5023.4}
{"event": "train", "step": 5750, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38784, "bpb": 2.00223, "chunk_loss": 0.030835, "F": 0.22052, "G": 0.22109, "chunks": 225.81, "bytes_per_chunk": 4.535, "boundary_entropy": 0.07924, "router_noise": 0.0, "grad_norm": 0.7163, "lr": 3.13e-05, "router_trainable": true, "sec_per_step": 0.83, "probe_sec": 0.1315, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10195, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.03929, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.12546, "n_merge": 2, "mean_delta_merge": -0.15598, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -0.01385, "trunc_abs_frac": 0.47837, "mean_delta_full": -0.13608, "read_hit_rate": {"surface": 0.6422, "split": 0.86615, "merge": 0.20986, "skel": 0.64964}, "mean_gate": {"surface": 0.15822, "split": 0.0618, "merge": 0.13298, "skel": 0.11319}, "n_reads": 20213, "n_cand": 31969, "read_suppressed_frac": 0.1025, "dropout_reads_logged": 20213, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5616, "admitted_surface": 16874, "staged_surface": 0, "admitted_split": 16810, "staged_split": 0, "admitted_merge": 20678, "staged_merge": 0, "admitted_skel": 11174, "staged_skel": 0, "mean_count": 1704.926025390625, "n_util_obs": 860, "util_count_spearman": 0.008811811587389844}, "probe_overhead_sec_total": 589.703, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006354, "alpha": 0.03, "t": 5024.1}
{"event": "consolidate", "step": 5800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 5062.3}
{"event": "probe", "step": 5800, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3500.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.201115, "cps": 27.34375, "mean_p": 0.201265, "boundary_entropy": 0.078644, "bpb": 1.861231}, "tr": {"chunks": 3581.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.197813, "cps": 27.976562, "mean_p": 0.187642, "boundary_entropy": 0.080216, "bpb": 2.001417}, "fi": {"chunks": 4576.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.243262, "cps": 35.75, "mean_p": 0.236794, "boundary_entropy": 0.079411, "bpb": 2.089068}}, "bpb": {"en": 1.86123, "tr": 2.00142, "fi": 2.08907}, "gini_cpc": 0.047182, "gini_cps": 0.061537, "bpic": 4.866775, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9835803536903908, "fi": 1.2095677756935532}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5646, "admitted_surface": 16818, "staged_surface": 0, "admitted_split": 16767, "staged_split": 0, "admitted_merge": 20740, "staged_merge": 0, "admitted_skel": 11211, "staged_skel": 0, "mean_count": 1719.50732421875, "n_util_obs": 868, "util_count_spearman": 0.012453521974439788}, "t": 5064.2}
{"event": "eval", "step": 5800, "phase": 3, "bpb": {"en": 1.87491, "tr": 2.00972, "fi": 2.09532}, "bytes_per_chunk": {"en": 4.66, "tr": 5.62, "fi": 4.331}, "bpb_mean": 1.99332, "frozen_inventory": true, "t": 5067.2}
{"event": "train", "step": 5800, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.31675, "bpb": 1.89967, "chunk_loss": 0.030609, "F": 0.21277, "G": 0.21364, "chunks": 217.88, "bytes_per_chunk": 4.7, "boundary_entropy": 0.07936, "router_noise": 0.0, "grad_norm": 0.7137, "lr": 3.08e-05, "router_trainable": true, "sec_per_step": 0.896, "probe_sec": 0.0647, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04244, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.03604, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.1073, "n_merge": 2, "mean_delta_merge": -0.14243, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 1.00324, "trunc_abs_frac": 1.00324, "mean_delta_full": -0.09412, "read_hit_rate": {"surface": 0.65173, "split": 0.86875, "merge": 0.21048, "skel": 0.6596}, "mean_gate": {"surface": 0.1619, "split": 0.06395, "merge": 0.13693, "skel": 0.11776}, "n_reads": 19888, "n_cand": 31133, "read_suppressed_frac": 0.1012, "dropout_reads_logged": 19888, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5647, "admitted_surface": 16860, "staged_surface": 0, "admitted_split": 16802, "staged_split": 0, "admitted_merge": 20682, "staged_merge": 0, "admitted_skel": 11192, "staged_skel": 0, "mean_count": 1719.818359375, "n_util_obs": 868, "util_count_spearman": 0.012466109571448392}, "probe_overhead_sec_total": 595.104, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006395, "alpha": 0.03, "t": 5068.9}
{"event": "consolidate", "step": 5850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 5128.3}
{"event": "train", "step": 5850, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.33555, "bpb": 1.9268, "chunk_loss": 0.030604, "F": 0.21622, "G": 0.21685, "chunks": 221.41, "bytes_per_chunk": 4.625, "boundary_entropy": 0.07795, "router_noise": 0.0, "grad_norm": 0.6924, "lr": 3.05e-05, "router_trainable": true, "sec_per_step": 1.2, "probe_sec": 0.1219, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09666, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.02287, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.03906, "n_merge": 1, "mean_delta_merge": -0.00327, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.3618, "trunc_abs_frac": 0.53698, "mean_delta_full": -0.23511, "read_hit_rate": {"surface": 0.64483, "split": 0.86851, "merge": 0.20481, "skel": 0.65282}, "mean_gate": {"surface": 0.15334, "split": 0.0598, "merge": 0.12993, "skel": 0.11042}, "n_reads": 20268, "n_cand": 31881, "read_suppressed_frac": 0.099, "dropout_reads_logged": 20268, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5681, "admitted_surface": 16859, "staged_surface": 0, "admitted_split": 16786, "staged_split": 0, "admitted_merge": 20703, "staged_merge": 0, "admitted_skel": 11188, "staged_skel": 0, "mean_count": 1734.6326904296875, "n_util_obs": 874, "util_count_spearman": 0.010015374653646103}, "probe_overhead_sec_total": 600.985, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006939, "alpha": 0.03, "t": 5128.9}
{"event": "consolidate", "step": 5900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 5176.0}
{"event": "train", "step": 5900, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.26672, "bpb": 1.82749, "chunk_loss": 0.030456, "F": 0.2074, "G": 0.20883, "chunks": 212.38, "bytes_per_chunk": 4.822, "boundary_entropy": 0.07726, "router_noise": 0.0, "grad_norm": 0.6859, "lr": 3.02e-05, "router_trainable": true, "sec_per_step": 0.967, "probe_sec": 0.0981, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06192, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.05828, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.02063, "n_merge": 2, "mean_delta_merge": 0.3118, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.80923, "trunc_abs_frac": 0.22606, "mean_delta_full": 0.10046, "read_hit_rate": {"surface": 0.63616, "split": 0.86351, "merge": 0.20945, "skel": 0.65022}, "mean_gate": {"surface": 0.15872, "split": 0.06132, "merge": 0.13184, "skel": 0.11362}, "n_reads": 19357, "n_cand": 30603, "read_suppressed_frac": 0.0993, "dropout_reads_logged": 19357, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5706, "admitted_surface": 16881, "staged_surface": 0, "admitted_split": 16845, "staged_split": 0, "admitted_merge": 20588, "staged_merge": 0, "admitted_skel": 11222, "staged_skel": 0, "mean_count": 1749.583984375, "n_util_obs": 883, "util_count_spearman": 0.019476402979315282}, "probe_overhead_sec_total": 606.94, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006665, "alpha": 0.03, "t": 5177.3}
{"event": "consolidate", "step": 5950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 5224.5}
{"event": "train", "step": 5950, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.37569, "bpb": 1.9847, "chunk_loss": 0.030599, "F": 0.2135, "G": 0.21337, "chunks": 218.62, "bytes_per_chunk": 4.684, "boundary_entropy": 0.07842, "router_noise": 0.0, "grad_norm": 0.7139, "lr": 3.01e-05, "router_trainable": true, "sec_per_step": 0.958, "probe_sec": 0.1347, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10517, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.02006, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.02893, "n_merge": 1, "mean_delta_merge": -0.10071, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.11806, "trunc_abs_frac": 0.41166, "mean_delta_full": -0.3089, "read_hit_rate": {"surface": 0.63986, "split": 0.86937, "merge": 0.20384, "skel": 0.65806}, "mean_gate": {"surface": 0.15651, "split": 0.06131, "merge": 0.13021, "skel": 0.11309}, "n_reads": 20085, "n_cand": 31566, "read_suppressed_frac": 0.099, "dropout_reads_logged": 20085, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5735, "admitted_surface": 16870, "staged_surface": 0, "admitted_split": 16821, "staged_split": 0, "admitted_merge": 20648, "staged_merge": 0, "admitted_skel": 11197, "staged_skel": 0, "mean_count": 1764.390869140625, "n_util_obs": 887, "util_count_spearman": 0.012799674760325988}, "probe_overhead_sec_total": 612.397, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006764, "alpha": 0.03, "t": 5225.2}
{"event": "consolidate", "step": 6000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 5263.1}
{"event": "train", "step": 5999, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.37195, "bpb": 1.97931, "chunk_loss": 0.030669, "F": 0.21689, "G": 0.21743, "chunks": 222.09, "bytes_per_chunk": 4.611, "boundary_entropy": 0.07704, "router_noise": 0.0, "grad_norm": 0.6727, "lr": 3e-05, "router_trainable": true, "sec_per_step": 0.758, "probe_sec": 0.1183, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09094, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.20628, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": -0.09424, "n_merge": 2, "mean_delta_merge": 0.03046, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -0.04253, "trunc_abs_frac": 0.85763, "mean_delta_full": 0.18372, "read_hit_rate": {"surface": 0.64763, "split": 0.86409, "merge": 0.21227, "skel": 0.64157}, "mean_gate": {"surface": 0.15831, "split": 0.06301, "merge": 0.13247, "skel": 0.1147}, "n_reads": 19694, "n_cand": 31211, "read_suppressed_frac": 0.0989, "dropout_reads_logged": 19694, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5758, "admitted_surface": 16864, "staged_surface": 0, "admitted_split": 16789, "staged_split": 0, "admitted_merge": 20665, "staged_merge": 0, "admitted_skel": 11218, "staged_skel": 0, "mean_count": 1778.971923828125, "n_util_obs": 893, "util_count_spearman": 0.020786862038113977}, "probe_overhead_sec_total": 617.19, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.007279, "alpha": 0.03, "t": 5263.1}
{"event": "probe", "step": 6000, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3511.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.201747, "cps": 27.429688, "mean_p": 0.201933, "boundary_entropy": 0.07901, "bpb": 1.859427}, "tr": {"chunks": 3573.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.197371, "cps": 27.914062, "mean_p": 0.187768, "boundary_entropy": 0.081404, "bpb": 1.996246}, "fi": {"chunks": 4586.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.243794, "cps": 35.828125, "mean_p": 0.237572, "boundary_entropy": 0.080168, "bpb": 2.085729}}, "bpb": {"en": 1.85943, "tr": 1.99625, "fi": 2.08573}, "gini_cpc": 0.048138, "gini_cps": 0.061411, "bpic": 4.861354, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9783083391938402, "fi": 1.2084131918850212}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5758, "admitted_surface": 16864, "staged_surface": 0, "admitted_split": 16789, "staged_split": 0, "admitted_merge": 20665, "staged_merge": 0, "admitted_skel": 11218, "staged_skel": 0, "mean_count": 1778.971923828125, "n_util_obs": 893, "util_count_spearman": 0.020786862038113977}, "t": 5264.7}
{"event": "eval", "step": 6000, "phase": 3, "bpb": {"en": 1.87251, "tr": 2.00485, "fi": 2.09297}, "bytes_per_chunk": {"en": 4.658, "tr": 5.639, "fi": 4.331}, "bpb_mean": 1.99011, "frozen_inventory": true, "t": 5267.9}
{"event": "reps_dumped", "dirs": ["/local/scratch/sas245/beetle-hnet/runs/ali_mem-mixed_s2/probe/reps/th0.30", "/local/scratch/sas245/beetle-hnet/runs/ali_mem-mixed_s2/probe/reps/th0.40", "/local/scratch/sas245/beetle-hnet/runs/ali_mem-mixed_s2/probe/reps/th0.50", "/local/scratch/sas245/beetle-hnet/runs/ali_mem-mixed_s2/probe/reps/th0.60", "/local/scratch/sas245/beetle-hnet/runs/ali_mem-mixed_s2/probe/reps/th0.70"], "t": 5270.0}
{"event": "final", "run": "ali_mem-mixed_s2", "arch": "ali", "objective": "ratio", "condition": "A", "step": 6000, "params": 22785088, "seed": 2, "per_lang": {"en": {"cpc": 0.20174682526001264, "cps": 27.4296875, "bpb": 1.878356182090285, "bytes_per_chunk": 4.625145559123469, "reads_per_chunk": 2.8671971488055332}, "tr": {"cpc": 0.19737060155775285, "cps": 27.9140625, "bpb": 1.9955787507496994, "bytes_per_chunk": 5.658863020647821, "reads_per_chunk": 2.5556455007609364}, "fi": {"cpc": 0.24379352506512147, "cps": 35.828125, "bpb": 2.1120582146214826, "bytes_per_chunk": 4.314275369474342, "reads_per_chunk": 2.8598219281787958}}, "bpb_mean": 1.9953310491538223, "gini_cpc": 0.048138261731177856, "gini_cps": 0.061411025421308274, "bpic": 4.8613538988860325, "mean_boundary_entropy": 0.07703839987516403, "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": 617.1895942687988, "probe_overhead_seq_frac": 0.4998854166666667, "wall_sec": 5276.054979562759, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5758, "admitted_surface": 16864, "staged_surface": 0, "admitted_split": 16789, "staged_split": 0, "admitted_merge": 20665, "staged_merge": 0, "admitted_skel": 11218, "staged_skel": 0, "mean_count": 1778.971923828125, "n_util_obs": 893, "util_count_spearman": 0.020786862038113977}, "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": 0.020786862038113977, "paired_vs_dropout": {"n": 1977, "pearson": 0.009875145515678285, "spearman": 0.022241340372118754, "mean_abs_discrepancy": 0.42345359921455383, "mean_paired": 0.011157582513988018, "mean_dropout": 0.016990255564451218}, "predictor_calibration": {"n": 4855, "all": {"n": 4855, "pearson": 0.09485912965858367, "spearman": 0.08244094868611303, "mae": 0.07025863230228424, "reliability": [{"pred": -0.0036273847799748182, "obs": -0.003619680181145668, "n": 486}, {"pred": 0.003719085594639182, "obs": 0.0016223189886659384, "n": 485}, {"pred": 0.006902734749019146, "obs": 0.0011194533435627818, "n": 486}, {"pred": 0.009201674722135067, "obs": 0.011101895943284035, "n": 485}, {"pred": 0.011011066846549511, "obs": 0.007273219991475344, "n": 485}, {"pred": 0.012653808109462261, "obs": 0.01118354219943285, "n": 486}, {"pred": 0.014498446136713028, "obs": 0.005092968232929707, "n": 485}, {"pred": 0.016844023019075394, "obs": 0.014006364159286022, "n": 486}, {"pred": 0.020055899396538734, "obs": 0.012852534651756287, "n": 485}, {"pred": 0.02752947062253952, "obs": 0.03669833764433861, "n": 486}]}, "admitted": {"n": 4855, "pearson": 0.09485912965858367, "spearman": 0.08244094868611303, "mae": 0.07025863230228424, "reliability": [{"pred": -0.0036273847799748182, "obs": -0.003619680181145668, "n": 486}, {"pred": 0.003719085594639182, "obs": 0.0016223189886659384, "n": 485}, {"pred": 0.006902734749019146, "obs": 0.0011194533435627818, "n": 486}, {"pred": 0.009201674722135067, "obs": 0.011101895943284035, "n": 485}, {"pred": 0.011011066846549511, "obs": 0.007273219991475344, "n": 485}, {"pred": 0.012653808109462261, "obs": 0.01118354219943285, "n": 486}, {"pred": 0.014498446136713028, "obs": 0.005092968232929707, "n": 485}, {"pred": 0.016844023019075394, "obs": 0.014006364159286022, "n": 486}, {"pred": 0.020055899396538734, "obs": 0.012852534651756287, "n": 485}, {"pred": 0.02752947062253952, "obs": 0.03669833764433861, "n": 486}]}, "staged": {"n": 0, "pearson": NaN, "spearman": NaN, "mae": NaN}}, "t": 5276.8}
{"event": "hub_done", "repo": "AdaptiveChunking/ali_mem-mixed_s2", "pushed": [0, 1, 2, 4, 6, 8, 16, 32, 64, 128, 256, 512, 1000, 1024, 2048, 3000, 4096, 6000], "failed": [], "ok": true, "t": 5287.5}