{"event": "start", "run": "ali_ali_s0", "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": "utility", "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": 2.8}
{"event": "probe_schedule", "n": 40, "steps": [0, 1, 2, 4, 8, 16, 32, 64, 128, 200, 256, 400, 600, 800, 1000, 1200, 1400, 1600, 1800, 2000, 2200, 2400, 2600, 2800, 3000, 3200, 3400, 3600, 3800, 4000, 4200, 4400, 4600, 4800, 5000, 5200, 5400, 5600, 5800, 6000], "t": 2.8}
{"event": "probe", "step": 0, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 6384.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.366833, "cps": 49.875, "mean_p": 0.496725, "boundary_entropy": 0.686995, "bpb": 8.160238}, "tr": {"chunks": 5290.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.292217, "cps": 41.328125, "mean_p": 0.489567, "boundary_entropy": 0.687288, "bpb": 8.171434}, "fi": {"chunks": 8895.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.472862, "cps": 69.492188, "mean_p": 0.500688, "boundary_entropy": 0.687512, "bpb": 8.214758}}, "bpb": {"en": 8.16024, "tr": 8.17143, "fi": 8.21476}, "gini_cpc": 0.106395, "gini_cps": 0.116842, "bpic": 2.758131, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.7965927738475616, "fi": 1.2890367847879758}, "inventory": {"size": 0, "admitted": 0, "staged": 0, "budget_B": 65536, "staging_capacity": 262144, "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": 5.4}
{"event": "train", "step": 0, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 5.66332, "bpb": 8.17044, "chunk_loss": 0.038537, "F": 0.28732, "G": 0.4892, "chunks": 294.22, "bytes_per_chunk": 3.48, "boundary_entropy": 0.69137, "router_noise": 0.0, "grad_norm": 11.0318, "lr": 1e-06, "router_trainable": true, "sec_per_step": 0.016, "probe_sec": 0.0079, "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": 34933, "read_suppressed_frac": 0.0, "dropout_reads_logged": 0, "inventory": {"size": 9451, "admitted": 0, "staged": 9451, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 3696, "admitted_split": 0, "staged_split": 3051, "admitted_merge": 0, "staged_merge": 203, "admitted_skel": 0, "staged_skel": 2501, "mean_count": 0.0, "n_util_obs": 0, "util_count_spearman": NaN}, "probe_overhead_sec_total": 0.008, "probe_overhead_seq_frac": 0.0, "util_loss": 0.0, "alpha": 0.03, "t": 7.2}
{"event": "probe", "step": 1, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 6303.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.362179, "cps": 49.242188, "mean_p": 0.496454, "boundary_entropy": 0.686986, "bpb": 8.128298}, "tr": {"chunks": 5228.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.288792, "cps": 40.84375, "mean_p": 0.489372, "boundary_entropy": 0.687277, "bpb": 8.149243}, "fi": {"chunks": 8818.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.468768, "cps": 68.890625, "mean_p": 0.500418, "boundary_entropy": 0.687508, "bpb": 8.187196}}, "bpb": {"en": 8.1283, "tr": 8.14924, "fi": 8.1872}, "gini_cpc": 0.107154, "gini_cps": 0.117614, "bpic": 2.78795, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.7973735778105943, "fi": 1.2943002174161062}, "inventory": {"size": 9451, "admitted": 0, "staged": 9451, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 3696, "admitted_split": 0, "staged_split": 3051, "admitted_merge": 0, "staged_merge": 203, "admitted_skel": 0, "staged_skel": 2501, "mean_count": 0.0, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 8.8}
{"event": "probe", "step": 2, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 6171.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.354594, "cps": 48.210938, "mean_p": 0.495925, "boundary_entropy": 0.686967, "bpb": 8.064225}, "tr": {"chunks": 5127.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.283213, "cps": 40.054688, "mean_p": 0.488973, "boundary_entropy": 0.687254, "bpb": 8.105313}, "fi": {"chunks": 8677.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.461273, "cps": 67.789062, "mean_p": 0.499889, "boundary_entropy": 0.687499, "bpb": 8.130996}}, "bpb": {"en": 8.06423, "tr": 8.10531, "fi": 8.131}, "gini_cpc": 0.108005, "gini_cps": 0.118481, "bpic": 2.84015, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.7986956880537021, "fi": 1.3008472038994343}, "inventory": {"size": 16581, "admitted": 0, "staged": 16581, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2, "admitted_surface": 0, "staged_surface": 6592, "admitted_split": 0, "staged_split": 5118, "admitted_merge": 0, "staged_merge": 496, "admitted_skel": 0, "staged_skel": 4375, "mean_count": 0.0, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 12.8}
{"event": "probe", "step": 4, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 5666.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.325576, "cps": 44.265625, "mean_p": 0.494021, "boundary_entropy": 0.686888, "bpb": 7.848938}, "tr": {"chunks": 4747.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.262222, "cps": 37.085938, "mean_p": 0.487542, "boundary_entropy": 0.687166, "bpb": 7.951705}, "fi": {"chunks": 8073.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.429164, "cps": 63.070312, "mean_p": 0.497939, "boundary_entropy": 0.687458, "bpb": 7.941667}}, "bpb": {"en": 7.84894, "tr": 7.95171, "fi": 7.94167}, "gini_cpc": 0.109438, "gini_cps": 0.119947, "bpic": 3.068917, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.8054085400911859, "fi": 1.318167558718441}, "inventory": {"size": 30150, "admitted": 0, "staged": 30150, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 7, "admitted_surface": 0, "staged_surface": 11943, "admitted_split": 0, "staged_split": 9218, "admitted_merge": 0, "staged_merge": 918, "admitted_skel": 0, "staged_skel": 8071, "mean_count": 0.0, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 17.3}
{"event": "probe", "step": 8, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 4367.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.250934, "cps": 34.117188, "mean_p": 0.488765, "boundary_entropy": 0.686599, "bpb": 7.288629}, "tr": {"chunks": 3765.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.207977, "cps": 29.414062, "mean_p": 0.483652, "boundary_entropy": 0.686884, "bpb": 7.525083}, "fi": {"chunks": 6361.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.338153, "cps": 49.695312, "mean_p": 0.492186, "boundary_entropy": 0.687252, "bpb": 7.435063}}, "bpb": {"en": 7.28863, "tr": 7.52508, "fi": 7.43506}, "gini_cpc": 0.10888, "gini_cps": 0.119414, "bpic": 3.914441, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.8288107156108632, "fi": 1.3475796388397634}, "inventory": {"size": 55805, "admitted": 0, "staged": 55805, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 15, "admitted_surface": 0, "staged_surface": 21428, "admitted_split": 0, "staged_split": 17813, "admitted_merge": 0, "staged_merge": 1411, "admitted_skel": 0, "staged_skel": 15153, "mean_count": 0.0, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 22.8}
{"event": "probe", "step": 16, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 4438.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.255014, "cps": 34.671875, "mean_p": 0.490998, "boundary_entropy": 0.686932, "bpb": 6.660667}, "tr": {"chunks": 4545.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.251063, "cps": 35.507812, "mean_p": 0.488404, "boundary_entropy": 0.687353, "bpb": 6.978479}, "fi": {"chunks": 5618.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.298655, "cps": 43.890625, "mean_p": 0.491256, "boundary_entropy": 0.687412, "bpb": 6.839306}}, "bpb": {"en": 6.66067, "tr": 6.97848, "fi": 6.83931}, "gini_cpc": 0.039427, "gini_cps": 0.053878, "bpic": 3.885487, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9845100604405361, "fi": 1.1711342272407372}, "inventory": {"size": 102435, "admitted": 0, "staged": 102435, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 27, "admitted_surface": 0, "staged_surface": 37484, "admitted_split": 0, "staged_split": 34864, "admitted_merge": 0, "staged_merge": 1838, "admitted_skel": 0, "staged_skel": 28249, "mean_count": 0.0, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 30.5}
{"event": "probe", "step": 32, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3990.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.229271, "cps": 31.171875, "mean_p": 0.494538, "boundary_entropy": 0.687549, "bpb": 6.307242}, "tr": {"chunks": 4056.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.224051, "cps": 31.6875, "mean_p": 0.491867, "boundary_entropy": 0.687955, "bpb": 6.636336}, "fi": {"chunks": 4298.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.228483, "cps": 33.578125, "mean_p": 0.492836, "boundary_entropy": 0.687994, "bpb": 6.431583}}, "bpb": {"en": 6.30724, "tr": 6.63634, "fi": 6.43158}, "gini_cpc": 0.005104, "gini_cps": 0.016634, "bpic": 4.595917, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9772341143971901, "fi": 0.9965652795536766}, "inventory": {"size": 191344, "admitted": 0, "staged": 191344, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 48, "admitted_surface": 0, "staged_surface": 69360, "admitted_split": 0, "staged_split": 65854, "admitted_merge": 0, "staged_merge": 3289, "admitted_skel": 0, "staged_skel": 52841, "mean_count": 0.0, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 43.8}
{"event": "consolidate", "step": 50, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 4, "swapped": 0, "uncertain_staged": 262136, "t": 61.0}
{"event": "train", "step": 50, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 4.2256, "bpb": 6.09626, "chunk_loss": 0.030533, "F": 0.17389, "G": 0.49125, "chunks": 178.06, "bytes_per_chunk": 5.751, "boundary_entropy": 0.69209, "router_noise": 0.0, "grad_norm": 2.4397, "lr": 5.1e-05, "router_trainable": true, "sec_per_step": 1.101, "probe_sec": 0.2901, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.27417, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.00117, "n_shadow": 4, "mean_delta_shadow": 0.00298, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.02893, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.01302, "trunc_abs_frac": 0.27951, "mean_delta_full": -0.01758, "read_hit_rate": {"surface": 0.03691, "split": 0.1355, "merge": 0.0, "skel": 0.0}, "mean_gate": {"surface": 0.12078, "split": 0.12091, "merge": null, "skel": null}, "n_reads": 1362, "n_cand": 23073, "read_suppressed_frac": 0.1065, "dropout_reads_logged": 1362, "inventory": {"size": 262148, "admitted": 4, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 76, "admitted_surface": 1, "staged_surface": 93796, "admitted_split": 3, "staged_split": 90104, "admitted_merge": 0, "staged_merge": 5086, "admitted_skel": 0, "staged_skel": 73158, "mean_count": 16163.5, "n_util_obs": 4, "util_count_spearman": -0.8}, "probe_overhead_sec_total": 9.246, "probe_overhead_seq_frac": 0.4902, "util_loss": 0.079971, "alpha": 0.03, "t": 62.2}
{"event": "probe", "step": 64, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 4080.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.234442, "cps": 31.875, "mean_p": 0.488411, "boundary_entropy": 0.686628, "bpb": 5.623135}, "tr": {"chunks": 3925.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.216815, "cps": 30.664062, "mean_p": 0.488225, "boundary_entropy": 0.687519, "bpb": 6.009882}, "fi": {"chunks": 5369.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.285418, "cps": 41.945312, "mean_p": 0.493013, "boundary_entropy": 0.687702, "bpb": 5.816986}}, "bpb": {"en": 5.62313, "tr": 6.00988, "fi": 5.81699}, "gini_cpc": 0.062084, "gini_cps": 0.07198, "bpic": 4.241962, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9248111703942473, "fi": 1.2174341436643765}, "inventory": {"size": 262148, "admitted": 4, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 83, "admitted_surface": 1, "staged_surface": 94378, "admitted_split": 3, "staged_split": 88383, "admitted_merge": 0, "staged_merge": 6147, "admitted_skel": 0, "staged_skel": 73236, "mean_count": 20844.25, "n_util_obs": 4, "util_count_spearman": -0.8}, "t": 76.9}
{"event": "consolidate", "step": 100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 65532, "swapped": 4, "uncertain_staged": 262138, "t": 118.4}
{"event": "train", "step": 100, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 3.20511, "bpb": 4.624, "chunk_loss": 0.031787, "F": 0.19931, "G": 0.41995, "chunks": 204.09, "bytes_per_chunk": 5.017, "boundary_entropy": 0.66526, "router_noise": 0.0, "grad_norm": 2.3669, "lr": 0.000101, "router_trainable": true, "sec_per_step": 1.145, "probe_sec": 0.3241, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.30035, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": 0.00775, "n_shadow": 3, "mean_delta_shadow": 0.01958, "n_split": 2, "mean_delta_split": 0.01361, "n_merge": 1, "mean_delta_merge": 0.04462, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.3043, "trunc_abs_frac": 0.82032, "mean_delta_full": 0.08296, "read_hit_rate": {"surface": 0.6224, "split": 0.40584, "merge": 0.09324, "skel": 0.4107}, "mean_gate": {"surface": 0.12025, "split": 0.12031, "merge": 0.12042, "skel": 0.12038}, "n_reads": 11241, "n_cand": 29155, "read_suppressed_frac": 0.1014, "dropout_reads_logged": 11241, "inventory": {"size": 266108, "admitted": 65536, "staged": 200572, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 110, "admitted_surface": 62683, "staged_surface": 35013, "admitted_split": 2167, "staged_split": 80220, "admitted_merge": 205, "staged_merge": 11327, "admitted_skel": 481, "staged_skel": 74012, "mean_count": 13.640411376953125, "n_util_obs": 0, "util_count_spearman": NaN}, "probe_overhead_sec_total": 23.549, "probe_overhead_seq_frac": 0.495, "util_loss": 0.002149, "alpha": 0.03, "t": 119.5}
{"event": "probe", "step": 128, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3647.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.209562, "cps": 28.492188, "mean_p": 0.38032, "boundary_entropy": 0.611505, "bpb": 4.176299}, "tr": {"chunks": 3341.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.184555, "cps": 26.101562, "mean_p": 0.413167, "boundary_entropy": 0.650916, "bpb": 4.502704}, "fi": {"chunks": 3941.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.209505, "cps": 30.789062, "mean_p": 0.396803, "boundary_entropy": 0.636218, "bpb": 4.412504}}, "bpb": {"en": 4.1763, "tr": 4.5027, "fi": 4.4125}, "gini_cpc": 0.027618, "gini_cps": 0.0366, "bpic": 5.19096, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.8806721874725895, "fi": 0.9997304227699498}, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 152, "admitted_surface": 62683, "staged_surface": 65190, "admitted_split": 2167, "staged_split": 92631, "admitted_merge": 205, "staged_merge": 15738, "admitted_skel": 481, "staged_skel": 88585, "mean_count": 18.021087646484375, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 148.4}
{"event": "consolidate", "step": 150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 2, "uncertain_staged": 262134, "t": 185.4}
{"event": "train", "step": 150, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.82433, "bpb": 4.07464, "chunk_loss": 0.031309, "F": 0.19693, "G": 0.36325, "chunks": 201.66, "bytes_per_chunk": 5.078, "boundary_entropy": 0.59386, "router_noise": 0.0, "grad_norm": 3.8099, "lr": 0.000151, "router_trainable": true, "sec_per_step": 1.342, "probe_sec": 0.2991, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.27537, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": -0.01802, "n_shadow": 5, "mean_delta_shadow": -0.0024, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 1.61887, "trunc_abs_frac": 0.70344, "mean_delta_full": -0.01617, "read_hit_rate": {"surface": 0.50343, "split": 0.30796, "merge": 0.09114, "skel": 0.33975}, "mean_gate": {"surface": 0.12019, "split": 0.12026, "merge": 0.12068, "skel": 0.12015}, "n_reads": 8397, "n_cand": 27243, "read_suppressed_frac": 0.1021, "dropout_reads_logged": 8397, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 177, "admitted_surface": 62684, "staged_surface": 75700, "admitted_split": 2166, "staged_split": 85590, "admitted_merge": 205, "staged_merge": 19250, "admitted_skel": 481, "staged_skel": 81604, "mean_count": 21.008987426757812, "n_util_obs": 0, "util_count_spearman": NaN}, "probe_overhead_sec_total": 39.353, "probe_overhead_seq_frac": 0.4967, "util_loss": 0.000602, "alpha": 0.03, "t": 186.6}
{"event": "consolidate", "step": 200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 7, "uncertain_staged": 262132, "t": 252.6}
{"event": "probe", "step": 200, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3051.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.175315, "cps": 23.835938, "mean_p": 0.367612, "boundary_entropy": 0.571143, "bpb": 3.642661}, "tr": {"chunks": 3153.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.17417, "cps": 24.632812, "mean_p": 0.371075, "boundary_entropy": 0.591142, "bpb": 3.798778}, "fi": {"chunks": 3541.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.188241, "cps": 27.664062, "mean_p": 0.378119, "boundary_entropy": 0.607733, "bpb": 3.904962}}, "bpb": {"en": 3.64266, "tr": 3.79878, "fi": 3.90496}, "gini_cpc": 0.017445, "gini_cps": 0.033521, "bpic": 5.821652, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9934713146682611, "fi": 1.0737321470360839}, "inventory": {"size": 327673, "admitted": 65536, "staged": 262137, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 246, "admitted_surface": 62690, "staged_surface": 71025, "admitted_split": 2164, "staged_split": 86368, "admitted_merge": 205, "staged_merge": 24655, "admitted_skel": 477, "staged_skel": 80089, "mean_count": 24.567611694335938, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 255.0}
{"event": "eval", "step": 200, "phase": 1, "bpb": {"en": 3.64884, "tr": 3.80794, "fi": 3.90415}, "bytes_per_chunk": {"en": 5.552, "tr": 6.084, "fi": 5.632}, "bpb_mean": 3.78698, "frozen_inventory": true, "t": 259.1}
{"event": "train", "step": 200, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.55774, "bpb": 3.69004, "chunk_loss": 0.030534, "F": 0.17749, "G": 0.37309, "chunks": 181.75, "bytes_per_chunk": 5.634, "boundary_entropy": 0.57974, "router_noise": 0.0, "grad_norm": 1.6361, "lr": 0.000201, "router_trainable": true, "sec_per_step": 1.54, "probe_sec": 0.3275, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.30425, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.00053, "n_shadow": 5, "mean_delta_shadow": 0.01893, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.09277, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -0.12881, "trunc_abs_frac": 0.83003, "mean_delta_full": -0.05011, "read_hit_rate": {"surface": 0.43828, "split": 0.28518, "merge": 0.03025, "skel": 0.25207}, "mean_gate": {"surface": 0.12703, "split": 0.12788, "merge": 0.12912, "skel": 0.12726}, "n_reads": 6527, "n_cand": 25415, "read_suppressed_frac": 0.0936, "dropout_reads_logged": 6527, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 243, "admitted_surface": 62690, "staged_surface": 71698, "admitted_split": 2164, "staged_split": 86001, "admitted_merge": 205, "staged_merge": 24699, "admitted_skel": 477, "staged_skel": 79746, "mean_count": 24.667205810546875, "n_util_obs": 0, "util_count_spearman": NaN}, "probe_overhead_sec_total": 54.197, "probe_overhead_seq_frac": 0.4975, "util_loss": 0.000714, "alpha": 0.03, "t": 263.6}
{"event": "consolidate", "step": 250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262125, "t": 333.5}
{"event": "train", "step": 250, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.3835, "bpb": 3.43866, "chunk_loss": 0.031893, "F": 0.21582, "G": 0.34756, "chunks": 221.0, "bytes_per_chunk": 4.633, "boundary_entropy": 0.52815, "router_noise": 0.0, "grad_norm": 1.443, "lr": 0.000251, "router_trainable": true, "sec_per_step": 1.436, "probe_sec": 0.1338, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07901, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.02759, "n_shadow": 3, "mean_delta_shadow": 0.00202, "n_split": 2, "mean_delta_split": -0.06863, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 1.10931, "trunc_abs_frac": 0.39251, "mean_delta_full": 0.02334, "read_hit_rate": {"surface": 0.50327, "split": 0.44167, "merge": 0.01869, "skel": 0.35437}, "mean_gate": {"surface": 0.12812, "split": 0.12737, "merge": 0.1255, "skel": 0.12779}, "n_reads": 10676, "n_cand": 30910, "read_suppressed_frac": 0.0993, "dropout_reads_logged": 10676, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 255, "admitted_surface": 61748, "staged_surface": 78342, "admitted_split": 3088, "staged_split": 77706, "admitted_merge": 216, "staged_merge": 30708, "admitted_skel": 484, "staged_skel": 75388, "mean_count": 34.58135986328125, "n_util_obs": 14, "util_count_spearman": 0.05054945054945055}, "probe_overhead_sec_total": 67.766, "probe_overhead_seq_frac": 0.498, "util_loss": 0.000906, "alpha": 0.03, "t": 335.4}
{"event": "probe", "step": 256, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3345.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.192208, "cps": 26.132812, "mean_p": 0.323317, "boundary_entropy": 0.505138, "bpb": 3.337573}, "tr": {"chunks": 3447.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.19041, "cps": 26.929688, "mean_p": 0.336308, "boundary_entropy": 0.518083, "bpb": 3.524092}, "fi": {"chunks": 3900.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.207326, "cps": 30.46875, "mean_p": 0.356923, "boundary_entropy": 0.558207, "bpb": 3.646832}}, "bpb": {"en": 3.33757, "tr": 3.52409, "fi": 3.64683}, "gini_cpc": 0.019115, "gini_cps": 0.034605, "bpic": 5.306023, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9906465469514381, "fi": 1.0786504318506513}, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 260, "admitted_surface": 61748, "staged_surface": 79030, "admitted_split": 3088, "staged_split": 76400, "admitted_merge": 216, "staged_merge": 31881, "admitted_skel": 484, "staged_skel": 74833, "mean_count": 35.4063720703125, "n_util_obs": 17, "util_count_spearman": 0.14215686274509803}, "t": 346.2}
{"event": "consolidate", "step": 300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262130, "t": 405.4}
{"event": "train", "step": 300, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.28378, "bpb": 3.29479, "chunk_loss": 0.031282, "F": 0.21082, "G": 0.29613, "chunks": 215.88, "bytes_per_chunk": 4.743, "boundary_entropy": 0.4178, "router_noise": 0.0, "grad_norm": 1.1458, "lr": 0.0003, "router_trainable": true, "sec_per_step": 1.426, "probe_sec": 0.2486, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.22405, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": -0.01724, "n_shadow": 7, "mean_delta_shadow": 0.02042, "n_split": 1, "mean_delta_split": 0.033, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.33088, "trunc_abs_frac": 0.72063, "mean_delta_full": -0.14111, "read_hit_rate": {"surface": 0.50087, "split": 0.39771, "merge": 0.02528, "skel": 0.31364}, "mean_gate": {"surface": 0.13884, "split": 0.13786, "merge": 0.13678, "skel": 0.13925}, "n_reads": 9923, "n_cand": 30641, "read_suppressed_frac": 0.0951, "dropout_reads_logged": 9923, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 302, "admitted_surface": 61791, "staged_surface": 85886, "admitted_split": 3100, "staged_split": 68484, "admitted_merge": 330, "staged_merge": 39099, "admitted_skel": 315, "staged_skel": 68675, "mean_count": 39.9107666015625, "n_util_obs": 18, "util_count_spearman": -0.37048503611971106}, "probe_overhead_sec_total": 84.341, "probe_overhead_seq_frac": 0.4983, "util_loss": 0.000977, "alpha": 0.03, "t": 406.7}
{"event": "consolidate", "step": 350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262129, "t": 470.2}
{"event": "train", "step": 350, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.19428, "bpb": 3.16568, "chunk_loss": 0.030997, "F": 0.2041, "G": 0.28439, "chunks": 209.0, "bytes_per_chunk": 4.9, "boundary_entropy": 0.38806, "router_noise": 0.0, "grad_norm": 1.1132, "lr": 0.0002999, "router_trainable": true, "sec_per_step": 1.301, "probe_sec": 0.324, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.30207, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 3, "mean_delta_loo": -0.01925, "n_shadow": 9, "mean_delta_shadow": -0.02896, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.10413, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.68137, "trunc_abs_frac": 2.95892, "mean_delta_full": -0.15942, "read_hit_rate": {"surface": 0.46927, "split": 0.21964, "merge": 0.045, "skel": 0.27761}, "mean_gate": {"surface": 0.1622, "split": 0.16155, "merge": 0.15153, "skel": 0.17095}, "n_reads": 7291, "n_cand": 29604, "read_suppressed_frac": 0.1023, "dropout_reads_logged": 7291, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 333, "admitted_surface": 61754, "staged_surface": 81848, "admitted_split": 3054, "staged_split": 63792, "admitted_merge": 540, "staged_merge": 51231, "admitted_skel": 188, "staged_skel": 65273, "mean_count": 36.07566833496094, "n_util_obs": 13, "util_count_spearman": -0.19230769230769232}, "probe_overhead_sec_total": 102.066, "probe_overhead_seq_frac": 0.4986, "util_loss": 0.001354, "alpha": 0.03, "t": 471.7}
{"event": "consolidate", "step": 400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262129, "t": 533.5}
{"event": "probe", "step": 400, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3688.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.211917, "cps": 28.8125, "mean_p": 0.266444, "boundary_entropy": 0.318394, "bpb": 2.896806}, "tr": {"chunks": 3932.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.217202, "cps": 30.71875, "mean_p": 0.29684, "boundary_entropy": 0.42322, "bpb": 3.198738}, "fi": {"chunks": 4351.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.231301, "cps": 33.992188, "mean_p": 0.303073, "boundary_entropy": 0.410235, "bpb": 3.37658}}, "bpb": {"en": 2.89681, "tr": 3.19874, "fi": 3.37658}, "gini_cpc": 0.019567, "gini_cps": 0.036923, "bpic": 4.73912, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0249346263122219, "fi": 1.0914664926404}, "inventory": {"size": 326656, "admitted": 65536, "staged": 261120, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 360, "admitted_surface": 61230, "staged_surface": 77180, "admitted_split": 3310, "staged_split": 58573, "admitted_merge": 823, "staged_merge": 63689, "admitted_skel": 173, "staged_skel": 61678, "mean_count": 40.467926025390625, "n_util_obs": 19, "util_count_spearman": -0.3263157894736842}, "t": 536.3}
{"event": "eval", "step": 400, "phase": 1, "bpb": {"en": 2.91211, "tr": 3.20446, "fi": 3.36894}, "bytes_per_chunk": {"en": 4.494, "tr": 5.449, "fi": 4.624}, "bpb_mean": 3.16184, "frozen_inventory": true, "t": 540.8}
{"event": "train", "step": 400, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.04616, "bpb": 2.95198, "chunk_loss": 0.031292, "F": 0.21854, "G": 0.27857, "chunks": 223.78, "bytes_per_chunk": 4.576, "boundary_entropy": 0.33043, "router_noise": 0.0, "grad_norm": 1.217, "lr": 0.0002998, "router_trainable": true, "sec_per_step": 1.474, "probe_sec": 0.3342, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.31364, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": 0.03223, "n_shadow": 3, "mean_delta_shadow": -0.02736, "n_split": 0, "mean_delta_split": NaN, "n_merge": 4, "mean_delta_merge": -0.05458, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.36771, "trunc_abs_frac": 0.8715, "mean_delta_full": 0.10681, "read_hit_rate": {"surface": 0.48955, "split": 0.22404, "merge": 0.05242, "skel": 0.2953}, "mean_gate": {"surface": 0.19944, "split": 0.19706, "merge": 0.18176, "skel": 0.21719}, "n_reads": 8133, "n_cand": 31664, "read_suppressed_frac": 0.1008, "dropout_reads_logged": 8133, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 364, "admitted_surface": 61230, "staged_surface": 77472, "admitted_split": 3310, "staged_split": 58611, "admitted_merge": 823, "staged_merge": 64231, "admitted_skel": 173, "staged_skel": 61830, "mean_count": 40.59202575683594, "n_util_obs": 19, "util_count_spearman": -0.3263157894736842}, "probe_overhead_sec_total": 119.275, "probe_overhead_seq_frac": 0.4988, "util_loss": 0.001938, "alpha": 0.03, "t": 545.4}
{"event": "consolidate", "step": 450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262124, "t": 606.3}
{"event": "train", "step": 450, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.01652, "bpb": 2.90923, "chunk_loss": 0.031264, "F": 0.21823, "G": 0.27231, "chunks": 223.47, "bytes_per_chunk": 4.582, "boundary_entropy": 0.31184, "router_noise": 0.0, "grad_norm": 1.3484, "lr": 0.0002995, "router_trainable": true, "sec_per_step": 1.252, "probe_sec": 0.3683, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.34094, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 3, "mean_delta_loo": -0.21696, "n_shadow": 5, "mean_delta_shadow": 0.01299, "n_split": 2, "mean_delta_split": -0.21973, "n_merge": 2, "mean_delta_merge": -0.10663, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 17.49273, "trunc_abs_frac": 0.8866, "mean_delta_full": -0.00769, "read_hit_rate": {"surface": 0.37604, "split": 0.23463, "merge": 0.04191, "skel": 0.22307}, "mean_gate": {"surface": 0.22204, "split": 0.20881, "merge": 0.20457, "skel": 0.23656}, "n_reads": 7070, "n_cand": 31926, "read_suppressed_frac": 0.102, "dropout_reads_logged": 7070, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 301, "admitted_surface": 60420, "staged_surface": 74564, "admitted_split": 4233, "staged_split": 53122, "admitted_merge": 816, "staged_merge": 76533, "admitted_skel": 67, "staged_skel": 57925, "mean_count": 38.21919250488281, "n_util_obs": 21, "util_count_spearman": -0.17532467532467533}, "probe_overhead_sec_total": 136.111, "probe_overhead_seq_frac": 0.4989, "util_loss": 0.002705, "alpha": 0.03, "t": 608.0}
{"event": "consolidate", "step": 500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262122, "t": 670.5}
{"event": "train", "step": 500, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.99772, "bpb": 2.8821, "chunk_loss": 0.031449, "F": 0.22803, "G": 0.27009, "chunks": 233.5, "bytes_per_chunk": 4.385, "boundary_entropy": 0.29834, "router_noise": 0.0, "grad_norm": 1.1929, "lr": 0.0002992, "router_trainable": true, "sec_per_step": 1.285, "probe_sec": 0.3496, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.3156, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.1456, "n_shadow": 4, "mean_delta_shadow": 0.18318, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": 0.05261, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.47021, "trunc_abs_frac": 1.66385, "mean_delta_full": 0.09528, "read_hit_rate": {"surface": 0.29758, "split": 0.2086, "merge": 0.01741, "skel": 0.15403}, "mean_gate": {"surface": 0.23156, "split": 0.22017, "merge": 0.19867, "skel": 0.25102}, "n_reads": 5817, "n_cand": 32910, "read_suppressed_frac": 0.1037, "dropout_reads_logged": 5817, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 204, "admitted_surface": 59904, "staged_surface": 70971, "admitted_split": 5177, "staged_split": 48263, "admitted_merge": 399, "staged_merge": 87645, "admitted_skel": 56, "staged_skel": 55265, "mean_count": 35.4287109375, "n_util_obs": 24, "util_count_spearman": -0.13043478260869565}, "probe_overhead_sec_total": 152.959, "probe_overhead_seq_frac": 0.499, "util_loss": 0.003822, "alpha": 0.03, "t": 672.2}
{"event": "consolidate", "step": 550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262123, "t": 728.9}
{"event": "train", "step": 550, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.97556, "bpb": 2.85013, "chunk_loss": 0.031659, "F": 0.237, "G": 0.27118, "chunks": 242.69, "bytes_per_chunk": 4.219, "boundary_entropy": 0.28562, "router_noise": 0.0, "grad_norm": 1.4243, "lr": 0.0002987, "router_trainable": true, "sec_per_step": 1.161, "probe_sec": 0.1464, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09685, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.137, "n_shadow": 5, "mean_delta_shadow": -0.01177, "n_split": 1, "mean_delta_split": 0.11948, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 1.08623, "trunc_abs_frac": 2.4241, "mean_delta_full": 0.12168, "read_hit_rate": {"surface": 0.26817, "split": 0.11872, "merge": 0.00987, "skel": 0.12733}, "mean_gate": {"surface": 0.24558, "split": 0.21776, "merge": 0.20097, "skel": 0.25093}, "n_reads": 4415, "n_cand": 34110, "read_suppressed_frac": 0.1085, "dropout_reads_logged": 4415, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 212, "admitted_surface": 60286, "staged_surface": 66733, "admitted_split": 4964, "staged_split": 45864, "admitted_merge": 249, "staged_merge": 96835, "admitted_skel": 37, "staged_skel": 52712, "mean_count": 27.48150634765625, "n_util_obs": 26, "util_count_spearman": -0.01264957264957265}, "probe_overhead_sec_total": 167.441, "probe_overhead_seq_frac": 0.4991, "util_loss": 0.00562, "alpha": 0.03, "t": 730.3}
{"event": "consolidate", "step": 600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262122, "t": 770.7}
{"event": "probe", "step": 600, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3926.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.225593, "cps": 30.671875, "mean_p": 0.255632, "boundary_entropy": 0.256932, "bpb": 2.590021}, "tr": {"chunks": 3991.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.220461, "cps": 31.179688, "mean_p": 0.257863, "boundary_entropy": 0.323947, "bpb": 2.930419}, "fi": {"chunks": 4553.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.242039, "cps": 35.570312, "mean_p": 0.274259, "boundary_entropy": 0.316025, "bpb": 3.159733}}, "bpb": {"en": 2.59002, "tr": 2.93042, "fi": 3.15973}, "gini_cpc": 0.020907, "gini_cps": 0.03352, "bpic": 4.549479, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9772484747872092, "fi": 1.0729008560446838}, "inventory": {"size": 326656, "admitted": 65536, "staged": 261120, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 235, "admitted_surface": 60389, "staged_surface": 63849, "admitted_split": 5009, "staged_split": 43668, "admitted_merge": 88, "staged_merge": 103042, "admitted_skel": 50, "staged_skel": 50561, "mean_count": 33.58183288574219, "n_util_obs": 34, "util_count_spearman": 0.0016806722689075631}, "t": 772.8}
{"event": "eval", "step": 600, "phase": 1, "bpb": {"en": 2.60283, "tr": 2.94167, "fi": 3.16071}, "bytes_per_chunk": {"en": 4.172, "tr": 5.042, "fi": 4.359}, "bpb_mean": 2.90174, "frozen_inventory": true, "t": 775.8}
{"event": "train", "step": 600, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.8749, "bpb": 2.70491, "chunk_loss": 0.031441, "F": 0.23291, "G": 0.26414, "chunks": 238.5, "bytes_per_chunk": 4.294, "boundary_entropy": 0.26754, "router_noise": 0.0, "grad_norm": 1.4401, "lr": 0.0002982, "router_trainable": true, "sec_per_step": 1.001, "probe_sec": 0.1702, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.15026, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 11, "mean_delta_loo": -0.04317, "n_shadow": 3, "mean_delta_shadow": -0.19434, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.87151, "trunc_abs_frac": 0.91289, "mean_delta_full": -0.1937, "read_hit_rate": {"surface": 0.33145, "split": 0.14473, "merge": 0.00436, "skel": 0.23254}, "mean_gate": {"surface": 0.25184, "split": 0.22721, "merge": 0.21705, "skel": 0.26469}, "n_reads": 5715, "n_cand": 33670, "read_suppressed_frac": 0.0931, "dropout_reads_logged": 5715, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 235, "admitted_surface": 60389, "staged_surface": 64033, "admitted_split": 5009, "staged_split": 43736, "admitted_merge": 88, "staged_merge": 103616, "admitted_skel": 50, "staged_skel": 50759, "mean_count": 33.669036865234375, "n_util_obs": 34, "util_count_spearman": 0.006569900687547746}, "probe_overhead_sec_total": 173.573, "probe_overhead_seq_frac": 0.4992, "util_loss": 0.006534, "alpha": 0.03, "t": 780.3}
{"event": "consolidate", "step": 650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262122, "t": 823.8}
{"event": "train", "step": 650, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.89881, "bpb": 2.7394, "chunk_loss": 0.031693, "F": 0.24078, "G": 0.26703, "chunks": 246.56, "bytes_per_chunk": 4.153, "boundary_entropy": 0.25728, "router_noise": 0.0, "grad_norm": 1.0929, "lr": 0.0002975, "router_trainable": true, "sec_per_step": 0.886, "probe_sec": 0.1761, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.15273, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.06407, "n_shadow": 5, "mean_delta_shadow": 0.05293, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.07251, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -0.21965, "trunc_abs_frac": 0.11561, "mean_delta_full": -0.0352, "read_hit_rate": {"surface": 0.3324, "split": 0.16111, "merge": 0.00843, "skel": 0.21551}, "mean_gate": {"surface": 0.23139, "split": 0.20939, "merge": 0.19743, "skel": 0.23118}, "n_reads": 6046, "n_cand": 34673, "read_suppressed_frac": 0.092, "dropout_reads_logged": 6046, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 268, "admitted_surface": 60405, "staged_surface": 61283, "admitted_split": 4998, "staged_split": 42349, "admitted_merge": 87, "staged_merge": 109161, "admitted_skel": 46, "staged_skel": 49351, "mean_count": 37.09004211425781, "n_util_obs": 40, "util_count_spearman": -0.1606003752345216}, "probe_overhead_sec_total": 180.383, "probe_overhead_seq_frac": 0.4992, "util_loss": 0.007623, "alpha": 0.03, "t": 824.6}
{"event": "consolidate", "step": 700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262122, "t": 876.6}
{"event": "train", "step": 700, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.8501, "bpb": 2.66913, "chunk_loss": 0.031553, "F": 0.23993, "G": 0.25998, "chunks": 245.69, "bytes_per_chunk": 4.168, "boundary_entropy": 0.244, "router_noise": 0.0, "grad_norm": 1.1875, "lr": 0.0002967, "router_trainable": true, "sec_per_step": 1.052, "probe_sec": 0.0923, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.059, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 3, "mean_delta_loo": -0.08438, "n_shadow": 8, "mean_delta_shadow": -0.04027, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": 0.0434, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.38848, "trunc_abs_frac": 0.81541, "mean_delta_full": -0.33781, "read_hit_rate": {"surface": 0.46066, "split": 0.53801, "merge": 0.01667, "skel": 0.28902}, "mean_gate": {"surface": 0.22727, "split": 0.20059, "merge": 0.20388, "skel": 0.23028}, "n_reads": 12469, "n_cand": 34417, "read_suppressed_frac": 0.0995, "dropout_reads_logged": 12469, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 283, "admitted_surface": 59763, "staged_surface": 59775, "admitted_split": 5608, "staged_split": 40533, "admitted_merge": 102, "staged_merge": 113896, "admitted_skel": 63, "staged_skel": 47940, "mean_count": 70.23348999023438, "n_util_obs": 55, "util_count_spearman": 0.05772005772005772}, "probe_overhead_sec_total": 188.398, "probe_overhead_seq_frac": 0.4993, "util_loss": 0.007516, "alpha": 0.03, "t": 877.2}
{"event": "consolidate", "step": 750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262126, "t": 918.0}
{"event": "train", "step": 750, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.80131, "bpb": 2.59874, "chunk_loss": 0.031101, "F": 0.22089, "G": 0.25286, "chunks": 226.19, "bytes_per_chunk": 4.527, "boundary_entropy": 0.24444, "router_noise": 0.0, "grad_norm": 0.9663, "lr": 0.0002959, "router_trainable": true, "sec_per_step": 0.847, "probe_sec": 0.1618, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09426, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": 0.01515, "n_shadow": 6, "mean_delta_shadow": -0.02501, "n_split": 1, "mean_delta_split": -0.01404, "n_merge": 1, "mean_delta_merge": -0.18396, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 4.0234, "trunc_abs_frac": 3.68708, "mean_delta_full": -0.02283, "read_hit_rate": {"surface": 0.44754, "split": 0.57111, "merge": 0.01436, "skel": 0.29254}, "mean_gate": {"surface": 0.19086, "split": 0.16118, "merge": 0.18077, "skel": 0.17879}, "n_reads": 12073, "n_cand": 32355, "read_suppressed_frac": 0.1006, "dropout_reads_logged": 12073, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 297, "admitted_surface": 58825, "staged_surface": 59484, "admitted_split": 6531, "staged_split": 38994, "admitted_merge": 113, "staged_merge": 116329, "admitted_skel": 67, "staged_skel": 47337, "mean_count": 80.87019348144531, "n_util_obs": 70, "util_count_spearman": 0.015764150118099902}, "probe_overhead_sec_total": 195.058, "probe_overhead_seq_frac": 0.4993, "util_loss": 0.007365, "alpha": 0.03, "t": 919.6}
{"event": "consolidate", "step": 800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 210, "uncertain_staged": 262127, "t": 966.3}
{"event": "probe", "step": 800, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3903.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.224272, "cps": 30.492188, "mean_p": 0.243484, "boundary_entropy": 0.221169, "bpb": 2.41647}, "tr": {"chunks": 3701.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.204441, "cps": 28.914062, "mean_p": 0.241409, "boundary_entropy": 0.266843, "bpb": 2.768627}, "fi": {"chunks": 4786.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.254426, "cps": 37.390625, "mean_p": 0.277242, "boundary_entropy": 0.270317, "bpb": 2.95086}}, "bpb": {"en": 2.41647, "tr": 2.76863, "fi": 2.95086}, "gini_cpc": 0.048779, "gini_cps": 0.05838, "bpic": 4.578854, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9115785608553124, "fi": 1.1344526652128535}, "inventory": {"size": 327470, "admitted": 65536, "staged": 261934, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 339, "admitted_surface": 58800, "staged_surface": 57959, "admitted_split": 6536, "staged_split": 38111, "admitted_merge": 130, "staged_merge": 120104, "admitted_skel": 70, "staged_skel": 45760, "mean_count": 92.51084899902344, "n_util_obs": 83, "util_count_spearman": -0.08068510977708744}, "t": 968.0}
{"event": "eval", "step": 800, "phase": 1, "bpb": {"en": 2.42984, "tr": 2.77878, "fi": 2.95519}, "bytes_per_chunk": {"en": 4.207, "tr": 5.374, "fi": 4.099}, "bpb_mean": 2.72127, "frozen_inventory": true, "t": 971.1}
{"event": "train", "step": 800, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.79431, "bpb": 2.58864, "chunk_loss": 0.031351, "F": 0.23257, "G": 0.25565, "chunks": 238.16, "bytes_per_chunk": 4.3, "boundary_entropy": 0.22951, "router_noise": 0.0, "grad_norm": 0.9479, "lr": 0.0002949, "router_trainable": true, "sec_per_step": 1.134, "probe_sec": 0.0846, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05232, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.06585, "n_shadow": 3, "mean_delta_shadow": -0.10725, "n_split": 2, "mean_delta_split": 0.03859, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.32021, "trunc_abs_frac": 1.701, "mean_delta_full": -0.15804, "read_hit_rate": {"surface": 0.49374, "split": 0.58863, "merge": 0.02223, "skel": 0.30579}, "mean_gate": {"surface": 0.19529, "split": 0.16349, "merge": 0.17861, "skel": 0.17959}, "n_reads": 13252, "n_cand": 33637, "read_suppressed_frac": 0.1017, "dropout_reads_logged": 13252, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 339, "admitted_surface": 58800, "staged_surface": 58027, "admitted_split": 6536, "staged_split": 38128, "admitted_merge": 130, "staged_merge": 120187, "admitted_skel": 70, "staged_skel": 45802, "mean_count": 92.71308898925781, "n_util_obs": 83, "util_count_spearman": -0.08659421620861535}, "probe_overhead_sec_total": 202.049, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.007302, "alpha": 0.03, "t": 976.3}
{"event": "consolidate", "step": 850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262126, "t": 1025.3}
{"event": "train", "step": 850, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.76146, "bpb": 2.54124, "chunk_loss": 0.031492, "F": 0.24249, "G": 0.25371, "chunks": 248.31, "bytes_per_chunk": 4.124, "boundary_entropy": 0.22393, "router_noise": 0.0, "grad_norm": 1.0085, "lr": 0.0002938, "router_trainable": true, "sec_per_step": 0.999, "probe_sec": 0.1262, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10024, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.06725, "n_shadow": 3, "mean_delta_shadow": 0.09668, "n_split": 1, "mean_delta_split": 0.39545, "n_merge": 1, "mean_delta_merge": -0.11487, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.61337, "trunc_abs_frac": 0.77169, "mean_delta_full": 0.24744, "read_hit_rate": {"surface": 0.50619, "split": 0.59363, "merge": 0.02728, "skel": 0.30263}, "mean_gate": {"surface": 0.20726, "split": 0.16559, "merge": 0.18519, "skel": 0.18661}, "n_reads": 13810, "n_cand": 34685, "read_suppressed_frac": 0.0993, "dropout_reads_logged": 13810, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 369, "admitted_surface": 59063, "staged_surface": 56770, "admitted_split": 6271, "staged_split": 38031, "admitted_merge": 132, "staged_merge": 121931, "admitted_skel": 70, "staged_skel": 45412, "mean_count": 103.45114135742188, "n_util_obs": 90, "util_count_spearman": -0.05040128410914928}, "probe_overhead_sec_total": 208.809, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.007846, "alpha": 0.03, "t": 1026.2}
{"event": "consolidate", "step": 900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262127, "t": 1075.5}
{"event": "train", "step": 900, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.60811, "bpb": 2.32001, "chunk_loss": 0.031283, "F": 0.2341, "G": 0.24893, "chunks": 239.72, "bytes_per_chunk": 4.272, "boundary_entropy": 0.20814, "router_noise": 0.0, "grad_norm": 0.9795, "lr": 0.0002927, "router_trainable": true, "sec_per_step": 0.998, "probe_sec": 0.1158, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0927, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 2, "mean_delta_loo": -0.06693, "n_shadow": 5, "mean_delta_shadow": 0.26593, "n_split": 3, "mean_delta_split": -0.07799, "n_merge": 1, "mean_delta_merge": -0.9782, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.51589, "trunc_abs_frac": 0.57549, "mean_delta_full": -0.146, "read_hit_rate": {"surface": 0.46433, "split": 0.53222, "merge": 0.02301, "skel": 0.29064}, "mean_gate": {"surface": 0.21387, "split": 0.16335, "merge": 0.19249, "skel": 0.19053}, "n_reads": 12328, "n_cand": 33943, "read_suppressed_frac": 0.1017, "dropout_reads_logged": 12328, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 369, "admitted_surface": 60077, "staged_surface": 55217, "admitted_split": 5264, "staged_split": 37858, "admitted_merge": 129, "staged_merge": 124336, "admitted_skel": 66, "staged_skel": 44733, "mean_count": 105.69123840332031, "n_util_obs": 76, "util_count_spearman": -0.18933697881066303}, "probe_overhead_sec_total": 215.334, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.008603, "alpha": 0.03, "t": 1076.1}
{"event": "consolidate", "step": 950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262127, "t": 1122.1}
{"event": "train", "step": 950, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.77481, "bpb": 2.56051, "chunk_loss": 0.031321, "F": 0.23434, "G": 0.25137, "chunks": 239.97, "bytes_per_chunk": 4.267, "boundary_entropy": 0.20918, "router_noise": 0.0, "grad_norm": 1.1298, "lr": 0.0002914, "router_trainable": true, "sec_per_step": 0.941, "probe_sec": 0.1933, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.16987, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": 0.1008, "n_shadow": 5, "mean_delta_shadow": -0.11037, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.74877, "trunc_abs_frac": 1.10975, "mean_delta_full": -0.18473, "read_hit_rate": {"surface": 0.36851, "split": 0.17041, "merge": 0.00105, "skel": 0.14702}, "mean_gate": {"surface": 0.22858, "split": 0.17573, "merge": 0.22813, "skel": 0.20786}, "n_reads": 5848, "n_cand": 33750, "read_suppressed_frac": 0.1045, "dropout_reads_logged": 5848, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 281, "admitted_surface": 60698, "staged_surface": 54816, "admitted_split": 4743, "staged_split": 38437, "admitted_merge": 77, "staged_merge": 123860, "admitted_skel": 18, "staged_skel": 45031, "mean_count": 56.839813232421875, "n_util_obs": 57, "util_count_spearman": -0.1400051853772362}, "probe_overhead_sec_total": 222.946, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.009091, "alpha": 0.03, "t": 1123.1}
{"event": "consolidate", "step": 1000, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262126, "t": 1171.3}
{"event": "probe", "step": 1000, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3846.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.220996, "cps": 30.046875, "mean_p": 0.23561, "boundary_entropy": 0.201836, "bpb": 2.334495}, "tr": {"chunks": 3699.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.204331, "cps": 28.898438, "mean_p": 0.234149, "boundary_entropy": 0.240944, "bpb": 2.650243}, "fi": {"chunks": 4790.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.254638, "cps": 37.421875, "mean_p": 0.276696, "boundary_entropy": 0.248848, "bpb": 2.812634}}, "bpb": {"en": 2.33449, "tr": 2.65024, "fi": 2.81263}, "gini_cpc": 0.049324, "gini_cps": 0.058965, "bpic": 4.59927, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9245887827006203, "fi": 1.152228120914747}, "inventory": {"size": 326655, "admitted": 65536, "staged": 261119, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 312, "admitted_surface": 59774, "staged_surface": 55423, "admitted_split": 4759, "staged_split": 38757, "admitted_merge": 985, "staged_merge": 121950, "admitted_skel": 18, "staged_skel": 44989, "mean_count": 61.63551330566406, "n_util_obs": 65, "util_count_spearman": -0.02001748251748252}, "t": 1173.2}
{"event": "eval", "step": 1000, "phase": 1, "bpb": {"en": 2.3477, "tr": 2.66195, "fi": 2.82162}, "bytes_per_chunk": {"en": 4.285, "tr": 5.347, "fi": 4.133}, "bpb_mean": 2.61042, "frozen_inventory": true, "t": 1176.3}
{"event": "train", "step": 1000, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.71067, "bpb": 2.46798, "chunk_loss": 0.031618, "F": 0.24161, "G": 0.25687, "chunks": 247.41, "bytes_per_chunk": 4.139, "boundary_entropy": 0.20977, "router_noise": 0.0, "grad_norm": 1.0491, "lr": 0.0002901, "router_trainable": true, "sec_per_step": 1.195, "probe_sec": 0.1284, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09281, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.05307, "n_shadow": 3, "mean_delta_shadow": -0.0203, "n_split": 1, "mean_delta_split": 0.14449, "n_merge": 2, "mean_delta_merge": -0.14966, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 1.61373, "trunc_abs_frac": 3.6114, "mean_delta_full": 0.14551, "read_hit_rate": {"surface": 0.40317, "split": 0.17803, "merge": 0.05552, "skel": 0.14529}, "mean_gate": {"surface": 0.2419, "split": 0.17543, "merge": 0.21958, "skel": 0.21419}, "n_reads": 6735, "n_cand": 34278, "read_suppressed_frac": 0.1017, "dropout_reads_logged": 6735, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 314, "admitted_surface": 59774, "staged_surface": 55645, "admitted_split": 4759, "staged_split": 38906, "admitted_merge": 985, "staged_merge": 122519, "admitted_skel": 18, "staged_skel": 45074, "mean_count": 61.73828125, "n_util_obs": 65, "util_count_spearman": -0.02277097902097902}, "probe_overhead_sec_total": 229.65, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.009309, "alpha": 0.03, "t": 1182.9}
{"event": "consolidate", "step": 1050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262125, "t": 1234.0}
{"event": "train", "step": 1050, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.73676, "bpb": 2.50561, "chunk_loss": 0.031271, "F": 0.23383, "G": 0.25054, "chunks": 239.44, "bytes_per_chunk": 4.277, "boundary_entropy": 0.20371, "router_noise": 0.0, "grad_norm": 0.9642, "lr": 0.0002886, "router_trainable": true, "sec_per_step": 1.033, "probe_sec": 0.1006, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08056, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": -0.0551, "n_shadow": 4, "mean_delta_shadow": 0.09184, "n_split": 2, "mean_delta_split": -0.02649, "n_merge": 1, "mean_delta_merge": -0.03198, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.90659, "trunc_abs_frac": 3.2866, "mean_delta_full": 0.03104, "read_hit_rate": {"surface": 0.37654, "split": 0.19251, "merge": 0.0354, "skel": 0.15018}, "mean_gate": {"surface": 0.24936, "split": 0.16759, "merge": 0.22489, "skel": 0.22942}, "n_reads": 6476, "n_cand": 33756, "read_suppressed_frac": 0.1064, "dropout_reads_logged": 6476, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 326, "admitted_surface": 58866, "staged_surface": 55453, "admitted_split": 4758, "staged_split": 38722, "admitted_merge": 1895, "staged_merge": 123233, "admitted_skel": 17, "staged_skel": 44736, "mean_count": 66.15470886230469, "n_util_obs": 75, "util_count_spearman": 0.021963015647226174}, "probe_overhead_sec_total": 236.93, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.008673, "alpha": 0.03, "t": 1234.5}
{"event": "consolidate", "step": 1100, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262124, "t": 1285.4}
{"event": "train", "step": 1100, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.71953, "bpb": 2.48075, "chunk_loss": 0.031491, "F": 0.24051, "G": 0.25346, "chunks": 246.28, "bytes_per_chunk": 4.158, "boundary_entropy": 0.19576, "router_noise": 0.0, "grad_norm": 1.0026, "lr": 0.0002871, "router_trainable": true, "sec_per_step": 1.039, "probe_sec": 0.0727, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05246, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.04285, "n_shadow": 4, "mean_delta_shadow": -0.12094, "n_split": 2, "mean_delta_split": 0.04393, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.70018, "trunc_abs_frac": 0.70146, "mean_delta_full": -0.33936, "read_hit_rate": {"surface": 0.33418, "split": 0.14356, "merge": 0.04094, "skel": 0.13434}, "mean_gate": {"surface": 0.24991, "split": 0.16321, "merge": 0.19244, "skel": 0.2315}, "n_reads": 5620, "n_cand": 34676, "read_suppressed_frac": 0.1016, "dropout_reads_logged": 5620, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 333, "admitted_surface": 59834, "staged_surface": 54955, "admitted_split": 3747, "staged_split": 39258, "admitted_merge": 1939, "staged_merge": 122963, "admitted_skel": 16, "staged_skel": 44968, "mean_count": 61.5950927734375, "n_util_obs": 69, "util_count_spearman": 0.006247716477895506}, "probe_overhead_sec_total": 244.781, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.010309, "alpha": 0.03, "t": 1286.5}
{"event": "consolidate", "step": 1150, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262124, "t": 1332.6}
{"event": "train", "step": 1150, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.59072, "bpb": 2.29492, "chunk_loss": 0.031095, "F": 0.22775, "G": 0.24583, "chunks": 233.22, "bytes_per_chunk": 4.391, "boundary_entropy": 0.19203, "router_noise": 0.0, "grad_norm": 0.8688, "lr": 0.0002855, "router_trainable": true, "sec_per_step": 0.956, "probe_sec": 0.157, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.13146, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": 0.05558, "n_shadow": 5, "mean_delta_shadow": -0.10884, "n_split": 2, "mean_delta_split": 0.13617, "n_merge": 1, "mean_delta_merge": -0.32819, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 1.49795, "trunc_abs_frac": 1.00867, "mean_delta_full": -0.12927, "read_hit_rate": {"surface": 0.20441, "split": 0.05723, "merge": 0.02419, "skel": 0.08661}, "mean_gate": {"surface": 0.2648, "split": 0.17185, "merge": 0.19678, "skel": 0.24287}, "n_reads": 2892, "n_cand": 32896, "read_suppressed_frac": 0.1065, "dropout_reads_logged": 2892, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 250, "admitted_surface": 60466, "staged_surface": 54925, "admitted_split": 3126, "staged_split": 40947, "admitted_merge": 1934, "staged_merge": 120597, "admitted_skel": 10, "staged_skel": 45675, "mean_count": 33.64311218261719, "n_util_obs": 50, "util_count_spearman": -0.011572629051620648}, "probe_overhead_sec_total": 251.649, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.012163, "alpha": 0.03, "t": 1334.3}
{"event": "consolidate", "step": 1200, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 1024, "uncertain_staged": 262122, "t": 1386.2}
{"event": "probe", "step": 1200, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3655.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.210021, "cps": 28.554688, "mean_p": 0.229833, "boundary_entropy": 0.182986, "bpb": 2.272709}, "tr": {"chunks": 3788.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.209247, "cps": 29.59375, "mean_p": 0.231692, "boundary_entropy": 0.219551, "bpb": 2.559129}, "fi": {"chunks": 4840.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.257296, "cps": 37.8125, "mean_p": 0.276291, "boundary_entropy": 0.218051, "bpb": 2.739965}}, "bpb": {"en": 2.27271, "tr": 2.55913, "fi": 2.73996}, "gini_cpc": 0.047346, "gini_cps": 0.064317, "bpic": 4.618741, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9963138275559984, "fi": 1.2250962686573135}, "inventory": {"size": 326654, "admitted": 65536, "staged": 261118, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 266, "admitted_surface": 61173, "staged_surface": 54468, "admitted_split": 2669, "staged_split": 41331, "admitted_merge": 1685, "staged_merge": 119586, "admitted_skel": 9, "staged_skel": 45733, "mean_count": 33.14518737792969, "n_util_obs": 49, "util_count_spearman": -0.11071428571428571}, "t": 1388.1}
{"event": "eval", "step": 1200, "phase": 1, "bpb": {"en": 2.2852, "tr": 2.5687, "fi": 2.74799}, "bytes_per_chunk": {"en": 4.478, "tr": 5.205, "fi": 4.141}, "bpb_mean": 2.53396, "frozen_inventory": true, "t": 1391.4}
{"event": "train", "step": 1200, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.62781, "bpb": 2.34843, "chunk_loss": 0.031024, "F": 0.22336, "G": 0.24425, "chunks": 228.72, "bytes_per_chunk": 4.477, "boundary_entropy": 0.18769, "router_noise": 0.0, "grad_norm": 0.7706, "lr": 0.0002837, "router_trainable": true, "sec_per_step": 1.24, "probe_sec": 0.1165, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09099, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.0585, "n_shadow": 4, "mean_delta_shadow": -0.03517, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.33028, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": -0.45365, "trunc_abs_frac": 7.62658, "mean_delta_full": 0.04873, "read_hit_rate": {"surface": 0.21326, "split": 0.05071, "merge": 0.01199, "skel": 0.10297}, "mean_gate": {"surface": 0.28055, "split": 0.16938, "merge": 0.20785, "skel": 0.26341}, "n_reads": 2807, "n_cand": 32130, "read_suppressed_frac": 0.0962, "dropout_reads_logged": 2807, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 267, "admitted_surface": 61173, "staged_surface": 54650, "admitted_split": 2669, "staged_split": 41476, "admitted_merge": 1685, "staged_merge": 120126, "admitted_skel": 9, "staged_skel": 45892, "mean_count": 33.18803405761719, "n_util_obs": 49, "util_count_spearman": -0.10806122448979592}, "probe_overhead_sec_total": 258.776, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.014955, "alpha": 0.03, "t": 1396.3}
{"event": "consolidate", "step": 1250, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262120, "t": 1443.6}
{"event": "train", "step": 1250, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.68408, "bpb": 2.42961, "chunk_loss": 0.031047, "F": 0.22482, "G": 0.24354, "chunks": 230.22, "bytes_per_chunk": 4.448, "boundary_entropy": 0.19173, "router_noise": 0.0, "grad_norm": 0.8798, "lr": 0.0002819, "router_trainable": true, "sec_per_step": 0.967, "probe_sec": 0.1219, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08204, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": 0.12438, "n_shadow": 3, "mean_delta_shadow": -0.20022, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.48141, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.24013, "trunc_abs_frac": 2.27989, "mean_delta_full": 0.10199, "read_hit_rate": {"surface": 0.20518, "split": 0.01643, "merge": 0.00493, "skel": 0.09937}, "mean_gate": {"surface": 0.26884, "split": 0.23483, "merge": 0.21093, "skel": 0.2586}, "n_reads": 2284, "n_cand": 32582, "read_suppressed_frac": 0.0994, "dropout_reads_logged": 2284, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 293, "admitted_surface": 62096, "staged_surface": 54425, "admitted_split": 2408, "staged_split": 42024, "admitted_merge": 1023, "staged_merge": 120296, "admitted_skel": 9, "staged_skel": 45399, "mean_count": 30.284408569335938, "n_util_obs": 42, "util_count_spearman": -0.13750911595494691}, "probe_overhead_sec_total": 265.011, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.017035, "alpha": 0.03, "t": 1444.6}
{"event": "consolidate", "step": 1300, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262115, "t": 1498.2}
{"event": "train", "step": 1300, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.62351, "bpb": 2.34222, "chunk_loss": 0.030858, "F": 0.21661, "G": 0.23659, "chunks": 221.81, "bytes_per_chunk": 4.617, "boundary_entropy": 0.18788, "router_noise": 0.0, "grad_norm": 0.8513, "lr": 0.00028, "router_trainable": true, "sec_per_step": 1.094, "probe_sec": 0.517, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08615, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.00415, "n_shadow": 6, "mean_delta_shadow": -0.05631, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.19894, "n_dropped_unpaired": 0, "n_full": 7, "trunc_ratio_sum": 0.87456, "trunc_abs_frac": 2.13849, "mean_delta_full": -0.13766, "read_hit_rate": {"surface": 0.31758, "split": 0.03009, "merge": 0.00327, "skel": 0.13243}, "mean_gate": {"surface": 0.25687, "split": 0.20364, "merge": 0.21163, "skel": 0.23421}, "n_reads": 3339, "n_cand": 31697, "read_suppressed_frac": 0.0991, "dropout_reads_logged": 3339, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 334, "admitted_surface": 61725, "staged_surface": 54657, "admitted_split": 2445, "staged_split": 42372, "admitted_merge": 1354, "staged_merge": 119399, "admitted_skel": 12, "staged_skel": 45716, "mean_count": 36.53253173828125, "n_util_obs": 44, "util_count_spearman": -0.15687103594080337}, "probe_overhead_sec_total": 271.676, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.016389, "alpha": 0.03, "t": 1499.3}
{"event": "consolidate", "step": 1350, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262113, "t": 1550.4}
{"event": "train", "step": 1350, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.60892, "bpb": 2.32119, "chunk_loss": 0.031124, "F": 0.22687, "G": 0.24306, "chunks": 232.31, "bytes_per_chunk": 4.408, "boundary_entropy": 0.17975, "router_noise": 0.0, "grad_norm": 0.8874, "lr": 0.000278, "router_trainable": true, "sec_per_step": 1.041, "probe_sec": 0.0994, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07521, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": -0.0152, "n_shadow": 7, "mean_delta_shadow": 0.01511, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 6.95322, "trunc_abs_frac": 1.04232, "mean_delta_full": -0.00417, "read_hit_rate": {"surface": 0.48284, "split": 0.05808, "merge": 0.00651, "skel": 0.19963}, "mean_gate": {"surface": 0.22929, "split": 0.1978, "merge": 0.19631, "skel": 0.20669}, "n_reads": 5446, "n_cand": 32931, "read_suppressed_frac": 0.0935, "dropout_reads_logged": 5446, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 366, "admitted_surface": 61215, "staged_surface": 54558, "admitted_split": 2808, "staged_split": 42593, "admitted_merge": 1491, "staged_merge": 119268, "admitted_skel": 22, "staged_skel": 45725, "mean_count": 45.405517578125, "n_util_obs": 50, "util_count_spearman": -0.18626650660264105}, "probe_overhead_sec_total": 278.556, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.015899, "alpha": 0.03, "t": 1551.4}
{"event": "consolidate", "step": 1400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262113, "t": 1596.9}
{"event": "probe", "step": 1400, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3675.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.21117, "cps": 28.710938, "mean_p": 0.224801, "boundary_entropy": 0.160444, "bpb": 2.222268}, "tr": {"chunks": 3796.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.209689, "cps": 29.65625, "mean_p": 0.223684, "boundary_entropy": 0.200424, "bpb": 2.500534}, "fi": {"chunks": 4831.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.256818, "cps": 37.742188, "mean_p": 0.269174, "boundary_entropy": 0.194211, "bpb": 2.639119}}, "bpb": {"en": 2.22227, "tr": 2.50053, "fi": 2.63912}, "gini_cpc": 0.046363, "gini_cps": 0.062646, "bpic": 4.611608, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9929844078160459, "fi": 1.2161634041740665}, "inventory": {"size": 326656, "admitted": 65536, "staged": 261120, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 404, "admitted_surface": 60255, "staged_surface": 54592, "admitted_split": 3036, "staged_split": 42290, "admitted_merge": 2223, "staged_merge": 118645, "admitted_skel": 22, "staged_skel": 45593, "mean_count": 49.63652038574219, "n_util_obs": 54, "util_count_spearman": -0.14160472651038689}, "t": 1599.0}
{"event": "eval", "step": 1400, "phase": 1, "bpb": {"en": 2.23302, "tr": 2.50997, "fi": 2.64933}, "bytes_per_chunk": {"en": 4.455, "tr": 5.314, "fi": 4.156}, "bpb_mean": 2.4641, "frozen_inventory": true, "t": 1602.9}
{"event": "train", "step": 1400, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.69452, "bpb": 2.44468, "chunk_loss": 0.031012, "F": 0.22449, "G": 0.23981, "chunks": 229.88, "bytes_per_chunk": 4.455, "boundary_entropy": 0.16998, "router_noise": 0.0, "grad_norm": 0.8939, "lr": 0.0002759, "router_trainable": true, "sec_per_step": 1.162, "probe_sec": 0.1056, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06495, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.01795, "n_shadow": 6, "mean_delta_shadow": 0.05179, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.19, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.22131, "trunc_abs_frac": 0.22131, "mean_delta_full": 0.21484, "read_hit_rate": {"surface": 0.48525, "split": 0.07278, "merge": 0.01851, "skel": 0.19414}, "mean_gate": {"surface": 0.21914, "split": 0.17262, "merge": 0.19366, "skel": 0.18985}, "n_reads": 5609, "n_cand": 32287, "read_suppressed_frac": 0.0968, "dropout_reads_logged": 5609, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 403, "admitted_surface": 60255, "staged_surface": 54757, "admitted_split": 3036, "staged_split": 42495, "admitted_merge": 2223, "staged_merge": 119182, "admitted_skel": 22, "staged_skel": 45710, "mean_count": 49.72210693359375, "n_util_obs": 55, "util_count_spearman": -0.08095238095238096}, "probe_overhead_sec_total": 285.477, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.015149, "alpha": 0.03, "t": 1609.5}
{"event": "consolidate", "step": 1450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262112, "t": 1654.9}
{"event": "train", "step": 1450, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.64537, "bpb": 2.37377, "chunk_loss": 0.03103, "F": 0.22687, "G": 0.24034, "chunks": 232.31, "bytes_per_chunk": 4.408, "boundary_entropy": 0.17074, "router_noise": 0.0, "grad_norm": 0.7685, "lr": 0.0002738, "router_trainable": true, "sec_per_step": 0.919, "probe_sec": 0.1283, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.1056, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.05072, "n_shadow": 5, "mean_delta_shadow": 0.01456, "n_split": 2, "mean_delta_split": -0.00189, "n_merge": 1, "mean_delta_merge": -0.349, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.14935, "trunc_abs_frac": 0.25959, "mean_delta_full": 0.15417, "read_hit_rate": {"surface": 0.49709, "split": 0.07102, "merge": 0.11211, "skel": 0.21322}, "mean_gate": {"surface": 0.21489, "split": 0.16383, "merge": 0.1597, "skel": 0.18081}, "n_reads": 6526, "n_cand": 32596, "read_suppressed_frac": 0.1019, "dropout_reads_logged": 6526, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 462, "admitted_surface": 59251, "staged_surface": 54968, "admitted_split": 3085, "staged_split": 43036, "admitted_merge": 3178, "staged_merge": 118705, "admitted_skel": 22, "staged_skel": 45435, "mean_count": 55.36408996582031, "n_util_obs": 66, "util_count_spearman": 0.1156664231291097}, "probe_overhead_sec_total": 291.726, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.015285, "alpha": 0.03, "t": 1655.4}
{"event": "consolidate", "step": 1500, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262111, "t": 1711.9}
{"event": "train", "step": 1500, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.60076, "bpb": 2.30941, "chunk_loss": 0.030764, "F": 0.21521, "G": 0.23087, "chunks": 220.38, "bytes_per_chunk": 4.647, "boundary_entropy": 0.16739, "router_noise": 0.0, "grad_norm": 0.8111, "lr": 0.0002715, "router_trainable": true, "sec_per_step": 1.143, "probe_sec": 0.0841, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05995, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.04085, "n_shadow": 4, "mean_delta_shadow": 0.00412, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.74899, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 0.62213, "trunc_abs_frac": 0.61508, "mean_delta_full": -0.18343, "read_hit_rate": {"surface": 0.48675, "split": 0.09352, "merge": 0.1162, "skel": 0.18216}, "mean_gate": {"surface": 0.20473, "split": 0.14576, "merge": 0.14732, "skel": 0.18645}, "n_reads": 6329, "n_cand": 31437, "read_suppressed_frac": 0.1016, "dropout_reads_logged": 6329, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 497, "admitted_surface": 58381, "staged_surface": 55072, "admitted_split": 3691, "staged_split": 42435, "admitted_merge": 3442, "staged_merge": 119043, "admitted_skel": 22, "staged_skel": 45594, "mean_count": 61.44728088378906, "n_util_obs": 69, "util_count_spearman": 0.1092802338326635}, "probe_overhead_sec_total": 299.679, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.01471, "alpha": 0.03, "t": 1712.6}
{"event": "consolidate", "step": 1550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 986, "uncertain_staged": 262108, "t": 1771.5}
{"event": "train", "step": 1550, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.57256, "bpb": 2.26873, "chunk_loss": 0.030788, "F": 0.21768, "G": 0.22924, "chunks": 222.91, "bytes_per_chunk": 4.594, "boundary_entropy": 0.16196, "router_noise": 0.0, "grad_norm": 0.7036, "lr": 0.0002692, "router_trainable": true, "sec_per_step": 1.204, "probe_sec": 0.0896, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0595, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.01764, "n_shadow": 2, "mean_delta_shadow": -0.26038, "n_split": 1, "mean_delta_split": 0.06746, "n_merge": 2, "mean_delta_merge": -0.50087, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.86744, "trunc_abs_frac": 0.82294, "mean_delta_full": -0.65991, "read_hit_rate": {"surface": 0.54542, "split": 0.41606, "merge": 0.13892, "skel": 0.24035}, "mean_gate": {"surface": 0.19851, "split": 0.12611, "merge": 0.14352, "skel": 0.16917}, "n_reads": 11179, "n_cand": 31580, "read_suppressed_frac": 0.1015, "dropout_reads_logged": 11179, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 534, "admitted_surface": 57485, "staged_surface": 55205, "admitted_split": 4476, "staged_split": 42021, "admitted_merge": 3547, "staged_merge": 118791, "admitted_skel": 28, "staged_skel": 46127, "mean_count": 98.253662109375, "n_util_obs": 82, "util_count_spearman": -0.011525840630119345}, "probe_overhead_sec_total": 308.363, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.01375, "alpha": 0.03, "t": 1772.8}
{"event": "consolidate", "step": 1600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 57, "uncertain_staged": 262117, "t": 1830.0}
{"event": "probe", "step": 1600, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3634.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.208815, "cps": 28.390625, "mean_p": 0.220567, "boundary_entropy": 0.151267, "bpb": 2.18134}, "tr": {"chunks": 3838.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.212009, "cps": 29.984375, "mean_p": 0.217557, "boundary_entropy": 0.180202, "bpb": 2.435388}, "fi": {"chunks": 4843.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.257456, "cps": 37.835938, "mean_p": 0.265591, "boundary_entropy": 0.17631, "bpb": 2.578172}}, "bpb": {"en": 2.18134, "tr": 2.43539, "fi": 2.57817}, "gini_cpc": 0.047808, "gini_cps": 0.065449, "bpic": 4.60674, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.015298199920099, "fi": 1.232939546636418}, "inventory": {"size": 327623, "admitted": 65536, "staged": 262087, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 569, "admitted_surface": 57433, "staged_surface": 55226, "admitted_split": 4499, "staged_split": 42475, "admitted_merge": 3557, "staged_merge": 118453, "admitted_skel": 47, "staged_skel": 45933, "mean_count": 111.66145324707031, "n_util_obs": 95, "util_count_spearman": -0.018211086226203807}, "t": 1832.1}
{"event": "eval", "step": 1600, "phase": 1, "bpb": {"en": 2.19267, "tr": 2.44515, "fi": 2.58571}, "bytes_per_chunk": {"en": 4.512, "tr": 5.223, "fi": 4.139}, "bpb_mean": 2.40785, "frozen_inventory": true, "t": 1835.6}
{"event": "train", "step": 1600, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.57539, "bpb": 2.27281, "chunk_loss": 0.030892, "F": 0.22049, "G": 0.2339, "chunks": 225.78, "bytes_per_chunk": 4.535, "boundary_entropy": 0.15546, "router_noise": 0.0, "grad_norm": 0.8428, "lr": 0.0002668, "router_trainable": true, "sec_per_step": 1.382, "probe_sec": 0.1679, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.1183, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.01597, "n_shadow": 6, "mean_delta_shadow": -0.03713, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.36538, "trunc_abs_frac": 0.36538, "mean_delta_full": 0.05078, "read_hit_rate": {"surface": 0.53733, "split": 0.42829, "merge": 0.14439, "skel": 0.24195}, "mean_gate": {"surface": 0.18908, "split": 0.11258, "merge": 0.13084, "skel": 0.15454}, "n_reads": 11495, "n_cand": 32078, "read_suppressed_frac": 0.1012, "dropout_reads_logged": 11495, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 569, "admitted_surface": 57433, "staged_surface": 55263, "admitted_split": 4499, "staged_split": 42510, "admitted_merge": 3557, "staged_merge": 118380, "admitted_skel": 47, "staged_skel": 45991, "mean_count": 111.83689880371094, "n_util_obs": 95, "util_count_spearman": -0.019526875699888018}, "probe_overhead_sec_total": 315.138, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.012541, "alpha": 0.03, "t": 1841.9}
{"event": "consolidate", "step": 1650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 67, "uncertain_staged": 262120, "t": 1901.8}
{"event": "train", "step": 1650, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.59808, "bpb": 2.30554, "chunk_loss": 0.030791, "F": 0.21713, "G": 0.22949, "chunks": 222.34, "bytes_per_chunk": 4.605, "boundary_entropy": 0.15831, "router_noise": 0.0, "grad_norm": 0.8408, "lr": 0.0002643, "router_trainable": true, "sec_per_step": 1.217, "probe_sec": 0.0726, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04841, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.29314, "n_shadow": 2, "mean_delta_shadow": 0.09286, "n_split": 3, "mean_delta_split": -0.01813, "n_merge": 1, "mean_delta_merge": 0.09766, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -31.53788, "trunc_abs_frac": 0.2784, "mean_delta_full": -0.00201, "read_hit_rate": {"surface": 0.5245, "split": 0.45306, "merge": 0.13544, "skel": 0.25486}, "mean_gate": {"surface": 0.19283, "split": 0.10894, "merge": 0.12788, "skel": 0.1527}, "n_reads": 11547, "n_cand": 31608, "read_suppressed_frac": 0.0986, "dropout_reads_logged": 11547, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 610, "admitted_surface": 57370, "staged_surface": 55358, "admitted_split": 4507, "staged_split": 41975, "admitted_merge": 3583, "staged_merge": 119093, "admitted_skel": 76, "staged_skel": 45718, "mean_count": 121.82470703125, "n_util_obs": 106, "util_count_spearman": 0.016958503056967573}, "probe_overhead_sec_total": 321.141, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.011662, "alpha": 0.03, "t": 1902.7}
{"event": "consolidate", "step": 1700, "retired": 1, "split_retired": 0, "merge_admitted": 0, "admitted": 1, "swapped": 97, "uncertain_staged": 262121, "t": 1957.2}
{"event": "train", "step": 1700, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.60823, "bpb": 2.32018, "chunk_loss": 0.030782, "F": 0.21741, "G": 0.22992, "chunks": 222.62, "bytes_per_chunk": 4.6, "boundary_entropy": 0.15035, "router_noise": 0.0, "grad_norm": 0.7545, "lr": 0.0002618, "router_trainable": true, "sec_per_step": 1.1, "probe_sec": 0.1276, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10615, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": 0.11869, "n_shadow": 7, "mean_delta_shadow": 0.00325, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": 0.10739, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.45655, "trunc_abs_frac": 0.42457, "mean_delta_full": 0.29724, "read_hit_rate": {"surface": 0.5275, "split": 0.50278, "merge": 0.12465, "skel": 0.2617}, "mean_gate": {"surface": 0.19936, "split": 0.10794, "merge": 0.12457, "skel": 0.15062}, "n_reads": 12158, "n_cand": 31684, "read_suppressed_frac": 0.1011, "dropout_reads_logged": 12158, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 641, "admitted_surface": 57281, "staged_surface": 55287, "admitted_split": 4514, "staged_split": 41893, "admitted_merge": 3646, "staged_merge": 119018, "admitted_skel": 95, "staged_skel": 45946, "mean_count": 141.93284606933594, "n_util_obs": 113, "util_count_spearman": -0.004736685609260708}, "probe_overhead_sec_total": 327.544, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.011596, "alpha": 0.03, "t": 1957.7}
{"event": "consolidate", "step": 1750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 182, "uncertain_staged": 262122, "t": 2020.2}
{"event": "train", "step": 1750, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.55609, "bpb": 2.24496, "chunk_loss": 0.030742, "F": 0.21759, "G": 0.2285, "chunks": 222.81, "bytes_per_chunk": 4.596, "boundary_entropy": 0.14442, "router_noise": 0.0, "grad_norm": 0.7433, "lr": 0.0002591, "router_trainable": true, "sec_per_step": 1.267, "probe_sec": 0.1353, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.11066, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": -0.05329, "n_shadow": 5, "mean_delta_shadow": -0.08848, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.1095, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 1.46335, "trunc_abs_frac": 0.21468, "mean_delta_full": 0.04413, "read_hit_rate": {"surface": 0.53677, "split": 0.52797, "merge": 0.12496, "skel": 0.27708}, "mean_gate": {"surface": 0.19062, "split": 0.10073, "merge": 0.12361, "skel": 0.14447}, "n_reads": 12688, "n_cand": 31816, "read_suppressed_frac": 0.0987, "dropout_reads_logged": 12688, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 684, "admitted_surface": 57112, "staged_surface": 55445, "admitted_split": 4542, "staged_split": 42216, "admitted_merge": 3787, "staged_merge": 118193, "admitted_skel": 95, "staged_skel": 46290, "mean_count": 152.59217834472656, "n_util_obs": 116, "util_count_spearman": -0.031345865528774074}, "probe_overhead_sec_total": 334.057, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.011321, "alpha": 0.03, "t": 2021.1}
{"event": "consolidate", "step": 1800, "retired": 1, "split_retired": 0, "merge_admitted": 0, "admitted": 1, "swapped": 427, "uncertain_staged": 262122, "t": 2087.2}
{"event": "probe", "step": 1800, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3594.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.206516, "cps": 28.078125, "mean_p": 0.216535, "boundary_entropy": 0.141764, "bpb": 2.143269}, "tr": {"chunks": 3817.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.210849, "cps": 29.820312, "mean_p": 0.213636, "boundary_entropy": 0.165122, "bpb": 2.38558}, "fi": {"chunks": 4844.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.257509, "cps": 37.84375, "mean_p": 0.26348, "boundary_entropy": 0.160503, "bpb": 2.509956}}, "bpb": {"en": 2.14327, "tr": 2.38558, "fi": 2.50996}, "gini_cpc": 0.050373, "gini_cps": 0.067999, "bpic": 4.629294, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.020980990154926, "fi": 1.2469191604493128}, "inventory": {"size": 327252, "admitted": 65536, "staged": 261716, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 729, "admitted_surface": 56700, "staged_surface": 55347, "admitted_split": 4648, "staged_split": 41738, "admitted_merge": 4083, "staged_merge": 118716, "admitted_skel": 105, "staged_skel": 45915, "mean_count": 162.25674438476562, "n_util_obs": 119, "util_count_spearman": -0.02325527570021436}, "t": 2089.1}
{"event": "eval", "step": 1800, "phase": 1, "bpb": {"en": 2.15524, "tr": 2.39471, "fi": 2.51999}, "bytes_per_chunk": {"en": 4.572, "tr": 5.239, "fi": 4.157}, "bpb_mean": 2.35665, "frozen_inventory": true, "t": 2091.9}
{"event": "train", "step": 1800, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.5275, "bpb": 2.20371, "chunk_loss": 0.030658, "F": 0.21249, "G": 0.22333, "chunks": 217.59, "bytes_per_chunk": 4.706, "boundary_entropy": 0.14292, "router_noise": 0.0, "grad_norm": 0.7413, "lr": 0.0002564, "router_trainable": true, "sec_per_step": 1.541, "probe_sec": 0.2126, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10907, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": 0.02176, "n_shadow": 6, "mean_delta_shadow": -0.15495, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": -0.02368, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -0.60441, "trunc_abs_frac": 0.44719, "mean_delta_full": -0.12329, "read_hit_rate": {"surface": 0.53268, "split": 0.53422, "merge": 0.14843, "skel": 0.28076}, "mean_gate": {"surface": 0.20302, "split": 0.10426, "merge": 0.12731, "skel": 0.15308}, "n_reads": 12637, "n_cand": 31148, "read_suppressed_frac": 0.0973, "dropout_reads_logged": 12637, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 730, "admitted_surface": 56700, "staged_surface": 55448, "admitted_split": 4648, "staged_split": 41867, "admitted_merge": 4083, "staged_merge": 118824, "admitted_skel": 105, "staged_skel": 46005, "mean_count": 162.4495849609375, "n_util_obs": 119, "util_count_spearman": -0.022546645776954848}, "probe_overhead_sec_total": 342.046, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.011386, "alpha": 0.03, "t": 2098.1}
{"event": "consolidate", "step": 1850, "retired": 1, "split_retired": 0, "merge_admitted": 0, "admitted": 1, "swapped": 727, "uncertain_staged": 262122, "t": 2158.2}
{"event": "train", "step": 1850, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.55953, "bpb": 2.24993, "chunk_loss": 0.030848, "F": 0.22253, "G": 0.23126, "chunks": 227.88, "bytes_per_chunk": 4.494, "boundary_entropy": 0.13549, "router_noise": 0.0, "grad_norm": 0.8285, "lr": 0.0002537, "router_trainable": true, "sec_per_step": 1.219, "probe_sec": 0.1011, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0594, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 11, "mean_delta_loo": -0.08424, "n_shadow": 2, "mean_delta_shadow": 0.01562, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.20288, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.31711, "trunc_abs_frac": 0.98809, "mean_delta_full": -0.08321, "read_hit_rate": {"surface": 0.54353, "split": 0.55558, "merge": 0.13641, "skel": 0.31929}, "mean_gate": {"surface": 0.18811, "split": 0.0942, "merge": 0.1217, "skel": 0.13728}, "n_reads": 13644, "n_cand": 32480, "read_suppressed_frac": 0.0925, "dropout_reads_logged": 13644, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 765, "admitted_surface": 55978, "staged_surface": 55451, "admitted_split": 5087, "staged_split": 41714, "admitted_merge": 4231, "staged_merge": 119208, "admitted_skel": 240, "staged_skel": 45771, "mean_count": 173.4622802734375, "n_util_obs": 126, "util_count_spearman": 0.01009373828271466}, "probe_overhead_sec_total": 348.385, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.010729, "alpha": 0.03, "t": 2159.1}
{"event": "consolidate", "step": 1900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262122, "t": 2216.4}
{"event": "train", "step": 1900, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.49669, "bpb": 2.15927, "chunk_loss": 0.030769, "F": 0.21881, "G": 0.22661, "chunks": 224.06, "bytes_per_chunk": 4.57, "boundary_entropy": 0.13243, "router_noise": 0.0, "grad_norm": 0.713, "lr": 0.0002508, "router_trainable": true, "sec_per_step": 1.165, "probe_sec": 0.1157, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0631, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.07701, "n_shadow": 3, "mean_delta_shadow": 0.05019, "n_split": 2, "mean_delta_split": -0.06036, "n_merge": 1, "mean_delta_merge": -0.39441, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 1.45176, "trunc_abs_frac": 12.56058, "mean_delta_full": 0.01594, "read_hit_rate": {"surface": 0.54105, "split": 0.55805, "merge": 0.13158, "skel": 0.33944}, "mean_gate": {"surface": 0.20005, "split": 0.09864, "merge": 0.12921, "skel": 0.14814}, "n_reads": 13572, "n_cand": 32085, "read_suppressed_frac": 0.1024, "dropout_reads_logged": 13572, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 796, "admitted_surface": 55006, "staged_surface": 55669, "admitted_split": 5516, "staged_split": 41488, "admitted_merge": 4652, "staged_merge": 118994, "admitted_skel": 362, "staged_skel": 45993, "mean_count": 184.38284301757812, "n_util_obs": 137, "util_count_spearman": -0.004293688278231}, "probe_overhead_sec_total": 354.452, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.010474, "alpha": 0.03, "t": 2217.3}
{"event": "consolidate", "step": 1950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262121, "t": 2278.2}
{"event": "train", "step": 1950, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.57489, "bpb": 2.27209, "chunk_loss": 0.030896, "F": 0.22189, "G": 0.23224, "chunks": 227.22, "bytes_per_chunk": 4.507, "boundary_entropy": 0.13784, "router_noise": 0.0, "grad_norm": 0.7426, "lr": 0.0002479, "router_trainable": true, "sec_per_step": 1.235, "probe_sec": 0.1231, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09121, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.03611, "n_shadow": 2, "mean_delta_shadow": -0.06, "n_split": 2, "mean_delta_split": 0.10193, "n_merge": 1, "mean_delta_merge": -0.01971, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.69969, "trunc_abs_frac": 0.95413, "mean_delta_full": -0.13708, "read_hit_rate": {"surface": 0.5422, "split": 0.54779, "merge": 0.14, "skel": 0.35429}, "mean_gate": {"surface": 0.20398, "split": 0.09874, "merge": 0.13603, "skel": 0.14836}, "n_reads": 13750, "n_cand": 32484, "read_suppressed_frac": 0.099, "dropout_reads_logged": 13750, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 812, "admitted_surface": 54006, "staged_surface": 55807, "admitted_split": 5619, "staged_split": 41942, "admitted_merge": 5502, "staged_merge": 118619, "admitted_skel": 409, "staged_skel": 45776, "mean_count": 194.6803741455078, "n_util_obs": 142, "util_count_spearman": 0.007881548354382715}, "probe_overhead_sec_total": 361.232, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.01051, "alpha": 0.03, "t": 2279.1}
{"event": "consolidate", "step": 2000, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 1024, "uncertain_staged": 262121, "t": 2346.7}
{"event": "probe", "step": 2000, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3613.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.207608, "cps": 28.226562, "mean_p": 0.214134, "boundary_entropy": 0.122744, "bpb": 2.11487}, "tr": {"chunks": 3810.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.210462, "cps": 29.765625, "mean_p": 0.209467, "boundary_entropy": 0.153249, "bpb": 2.345592}, "fi": {"chunks": 4916.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.261336, "cps": 38.40625, "mean_p": 0.262921, "boundary_entropy": 0.143322, "bpb": 2.470014}}, "bpb": {"en": 2.11487, "tr": 2.34559, "fi": 2.47001}, "gini_cpc": 0.052721, "gini_cps": 0.0704, "bpic": 4.597779, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0137493362817211, "fi": 1.258798304864964}, "inventory": {"size": 326654, "admitted": 65536, "staged": 261118, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 842, "admitted_surface": 52999, "staged_surface": 55694, "admitted_split": 5686, "staged_split": 41834, "admitted_merge": 6441, "staged_merge": 117783, "admitted_skel": 410, "staged_skel": 45807, "mean_count": 203.6713409423828, "n_util_obs": 143, "util_count_spearman": 0.028214156735283494}, "t": 2348.6}
{"event": "eval", "step": 2000, "phase": 1, "bpb": {"en": 2.12715, "tr": 2.35664, "fi": 2.48052}, "bytes_per_chunk": {"en": 4.561, "tr": 5.263, "fi": 4.115}, "bpb_mean": 2.32144, "frozen_inventory": true, "t": 2351.7}
{"event": "train", "step": 2000, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.54028, "bpb": 2.22216, "chunk_loss": 0.030911, "F": 0.22647, "G": 0.23138, "chunks": 231.91, "bytes_per_chunk": 4.416, "boundary_entropy": 0.12203, "router_noise": 0.0, "grad_norm": 0.8444, "lr": 0.000245, "router_trainable": true, "sec_per_step": 1.562, "probe_sec": 0.1355, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08299, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.05207, "n_shadow": 6, "mean_delta_shadow": 0.07195, "n_split": 1, "mean_delta_split": -0.09888, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 2.37752, "trunc_abs_frac": 0.6343, "mean_delta_full": -0.04236, "read_hit_rate": {"surface": 0.55177, "split": 0.56202, "merge": 0.15713, "skel": 0.34573}, "mean_gate": {"surface": 0.19876, "split": 0.09503, "merge": 0.12687, "skel": 0.14055}, "n_reads": 14197, "n_cand": 32776, "read_suppressed_frac": 0.0991, "dropout_reads_logged": 14197, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 843, "admitted_surface": 52999, "staged_surface": 55824, "admitted_split": 5686, "staged_split": 41951, "admitted_merge": 6441, "staged_merge": 118402, "admitted_skel": 410, "staged_skel": 45967, "mean_count": 203.8880615234375, "n_util_obs": 143, "util_count_spearman": 0.02738930548987937}, "probe_overhead_sec_total": 368.333, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.010213, "alpha": 0.03, "t": 2357.2}
{"event": "consolidate", "step": 2050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262121, "t": 2413.3}
{"event": "train", "step": 2050, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.55013, "bpb": 2.23637, "chunk_loss": 0.030562, "F": 0.20709, "G": 0.21852, "chunks": 212.06, "bytes_per_chunk": 4.829, "boundary_entropy": 0.13187, "router_noise": 0.0, "grad_norm": 0.7859, "lr": 0.0002419, "router_trainable": true, "sec_per_step": 1.136, "probe_sec": 0.0979, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05634, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": 0.04298, "n_shadow": 4, "mean_delta_shadow": -0.04222, "n_split": 2, "mean_delta_split": -0.03641, "n_merge": 1, "mean_delta_merge": -0.13722, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.05833, "trunc_abs_frac": 0.14877, "mean_delta_full": -0.02197, "read_hit_rate": {"surface": 0.51466, "split": 0.56555, "merge": 0.1327, "skel": 0.30949}, "mean_gate": {"surface": 0.22277, "split": 0.10697, "merge": 0.14259, "skel": 0.15726}, "n_reads": 12638, "n_cand": 30441, "read_suppressed_frac": 0.1015, "dropout_reads_logged": 12638, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 881, "admitted_surface": 52018, "staged_surface": 56065, "admitted_split": 6129, "staged_split": 41739, "admitted_merge": 6975, "staged_merge": 118336, "admitted_skel": 414, "staged_skel": 46004, "mean_count": 214.23191833496094, "n_util_obs": 150, "util_count_spearman": 0.03928174585537135}, "probe_overhead_sec_total": 374.771, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.009675, "alpha": 0.03, "t": 2414.0}
{"event": "consolidate", "step": 2100, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 1024, "uncertain_staged": 262121, "t": 2470.8}
{"event": "train", "step": 2100, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.4733, "bpb": 2.12552, "chunk_loss": 0.03062, "F": 0.21292, "G": 0.21906, "chunks": 218.03, "bytes_per_chunk": 4.697, "boundary_entropy": 0.12064, "router_noise": 0.0, "grad_norm": 0.716, "lr": 0.0002388, "router_trainable": true, "sec_per_step": 1.159, "probe_sec": 0.1077, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06291, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.02803, "n_shadow": 5, "mean_delta_shadow": 0.02099, "n_split": 2, "mean_delta_split": 0.03616, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.07119, "trunc_abs_frac": 0.39429, "mean_delta_full": -0.21368, "read_hit_rate": {"surface": 0.52052, "split": 0.54617, "merge": 0.1354, "skel": 0.32434}, "mean_gate": {"surface": 0.19952, "split": 0.09639, "merge": 0.12559, "skel": 0.14132}, "n_reads": 12859, "n_cand": 31222, "read_suppressed_frac": 0.1043, "dropout_reads_logged": 12859, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 903, "admitted_surface": 51053, "staged_surface": 55956, "admitted_split": 6714, "staged_split": 41207, "admitted_merge": 7355, "staged_merge": 118807, "admitted_skel": 414, "staged_skel": 46174, "mean_count": 224.10426330566406, "n_util_obs": 154, "util_count_spearman": 0.046707284315617686}, "probe_overhead_sec_total": 381.429, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.009383, "alpha": 0.03, "t": 2471.9}
{"event": "consolidate", "step": 2150, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262119, "t": 2517.5}
{"event": "train", "step": 2150, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.47932, "bpb": 2.1342, "chunk_loss": 0.03088, "F": 0.22461, "G": 0.22993, "chunks": 230.0, "bytes_per_chunk": 4.452, "boundary_entropy": 0.12263, "router_noise": 0.0, "grad_norm": 0.7146, "lr": 0.0002357, "router_trainable": true, "sec_per_step": 0.926, "probe_sec": 0.1715, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.14375, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.00346, "n_shadow": 5, "mean_delta_shadow": 0.08963, "n_split": 1, "mean_delta_split": -0.04053, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 1.31401, "trunc_abs_frac": 0.55283, "mean_delta_full": 0.08582, "read_hit_rate": {"surface": 0.5408, "split": 0.56499, "merge": 0.15419, "skel": 0.35178}, "mean_gate": {"surface": 0.19626, "split": 0.09337, "merge": 0.13232, "skel": 0.13416}, "n_reads": 14131, "n_cand": 32702, "read_suppressed_frac": 0.1037, "dropout_reads_logged": 14131, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 928, "admitted_surface": 50072, "staged_surface": 55996, "admitted_split": 7495, "staged_split": 40936, "admitted_merge": 7553, "staged_merge": 118961, "admitted_skel": 416, "staged_skel": 46251, "mean_count": 234.38853454589844, "n_util_obs": 156, "util_count_spearman": 0.020827867427415405}, "probe_overhead_sec_total": 388.466, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.009156, "alpha": 0.03, "t": 2518.2}
{"event": "consolidate", "step": 2200, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262119, "t": 2561.4}
{"event": "probe", "step": 2200, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3580.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.205712, "cps": 27.96875, "mean_p": 0.21233, "boundary_entropy": 0.120176, "bpb": 2.090515}, "tr": {"chunks": 3854.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.212893, "cps": 30.109375, "mean_p": 0.206866, "boundary_entropy": 0.143084, "bpb": 2.298709}, "fi": {"chunks": 5015.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.266599, "cps": 39.179688, "mean_p": 0.264186, "boundary_entropy": 0.128596, "bpb": 2.436188}}, "bpb": {"en": 2.09051, "tr": 2.29871, "fi": 2.43619}, "gini_cpc": 0.05924, "gini_cps": 0.076847, "bpic": 4.557153, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0349092113193374, "fi": 1.2959855148345687}, "inventory": {"size": 326655, "admitted": 65536, "staged": 261119, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 960, "admitted_surface": 49085, "staged_surface": 55900, "admitted_split": 8260, "staged_split": 40576, "admitted_merge": 7775, "staged_merge": 118505, "admitted_skel": 416, "staged_skel": 46138, "mean_count": 244.77955627441406, "n_util_obs": 157, "util_count_spearman": 0.03826927728403178}, "t": 2563.1}
{"event": "eval", "step": 2200, "phase": 1, "bpb": {"en": 2.10171, "tr": 2.30899, "fi": 2.44673}, "bytes_per_chunk": {"en": 4.584, "tr": 5.199, "fi": 4.014}, "bpb_mean": 2.28581, "frozen_inventory": true, "t": 2566.2}
{"event": "train", "step": 2200, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.52761, "bpb": 2.20387, "chunk_loss": 0.030906, "F": 0.2215, "G": 0.22769, "chunks": 226.81, "bytes_per_chunk": 4.515, "boundary_entropy": 0.12025, "router_noise": 0.0, "grad_norm": 0.794, "lr": 0.0002325, "router_trainable": true, "sec_per_step": 1.059, "probe_sec": 0.1363, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.11251, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": 0.01689, "n_shadow": 7, "mean_delta_shadow": 0.16759, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.29419, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.51476, "trunc_abs_frac": 0.6811, "mean_delta_full": 0.43294, "read_hit_rate": {"surface": 0.54304, "split": 0.57306, "merge": 0.15847, "skel": 0.35016}, "mean_gate": {"surface": 0.2006, "split": 0.09325, "merge": 0.13286, "skel": 0.13998}, "n_reads": 14100, "n_cand": 32314, "read_suppressed_frac": 0.1043, "dropout_reads_logged": 14100, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 961, "admitted_surface": 49085, "staged_surface": 56127, "admitted_split": 8260, "staged_split": 40701, "admitted_merge": 7775, "staged_merge": 119043, "admitted_skel": 416, "staged_skel": 46273, "mean_count": 244.9947509765625, "n_util_obs": 157, "util_count_spearman": 0.03829408517790361}, "probe_overhead_sec_total": 395.951, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.009354, "alpha": 0.03, "t": 2571.1}
{"event": "consolidate", "step": 2250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262119, "t": 2610.4}
{"event": "train", "step": 2250, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.49281, "bpb": 2.15367, "chunk_loss": 0.030686, "F": 0.21725, "G": 0.22156, "chunks": 222.47, "bytes_per_chunk": 4.603, "boundary_entropy": 0.11154, "router_noise": 0.0, "grad_norm": 0.7231, "lr": 0.0002293, "router_trainable": true, "sec_per_step": 0.799, "probe_sec": 0.1446, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09678, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": 0.16748, "n_shadow": 7, "mean_delta_shadow": 0.0678, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.60864, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.07152, "trunc_abs_frac": 0.12066, "mean_delta_full": -0.18945, "read_hit_rate": {"surface": 0.53232, "split": 0.58033, "merge": 0.1666, "skel": 0.33264}, "mean_gate": {"surface": 0.19459, "split": 0.08969, "merge": 0.12575, "skel": 0.14461}, "n_reads": 13805, "n_cand": 31652, "read_suppressed_frac": 0.1016, "dropout_reads_logged": 13805, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 988, "admitted_surface": 48099, "staged_surface": 55975, "admitted_split": 8707, "staged_split": 40652, "admitted_merge": 8311, "staged_merge": 119126, "admitted_skel": 419, "staged_skel": 46391, "mean_count": 255.5474090576172, "n_util_obs": 166, "util_count_spearman": 0.0623292298207528}, "probe_overhead_sec_total": 402.51, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.010048, "alpha": 0.03, "t": 2611.1}
{"event": "consolidate", "step": 2300, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262120, "t": 2652.9}
{"event": "train", "step": 2300, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.53881, "bpb": 2.22004, "chunk_loss": 0.03072, "F": 0.21683, "G": 0.22273, "chunks": 222.03, "bytes_per_chunk": 4.612, "boundary_entropy": 0.11103, "router_noise": 0.0, "grad_norm": 0.701, "lr": 0.000226, "router_trainable": true, "sec_per_step": 0.847, "probe_sec": 0.0894, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06569, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 3, "mean_delta_loo": -0.00487, "n_shadow": 7, "mean_delta_shadow": 0.02981, "n_split": 1, "mean_delta_split": 0.09222, "n_merge": 1, "mean_delta_merge": 0.14972, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 10.0, "trunc_abs_frac": 0.90171, "mean_delta_full": -0.00214, "read_hit_rate": {"surface": 0.52396, "split": 0.58506, "merge": 0.16915, "skel": 0.32782}, "mean_gate": {"surface": 0.2095, "split": 0.09614, "merge": 0.13819, "skel": 0.14716}, "n_reads": 13680, "n_cand": 31413, "read_suppressed_frac": 0.0993, "dropout_reads_logged": 13680, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1011, "admitted_surface": 47124, "staged_surface": 56415, "admitted_split": 9353, "staged_split": 40281, "admitted_merge": 8634, "staged_merge": 119104, "admitted_skel": 425, "staged_skel": 46344, "mean_count": 265.716796875, "n_util_obs": 170, "util_count_spearman": 0.0926301699612229}, "probe_overhead_sec_total": 409.519, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.009883, "alpha": 0.03, "t": 2653.4}
{"event": "consolidate", "step": 2350, "retired": 1, "split_retired": 1, "merge_admitted": 0, "admitted": 2, "swapped": 1024, "uncertain_staged": 262117, "t": 2693.7}
{"event": "train", "step": 2350, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.45596, "bpb": 2.1005, "chunk_loss": 0.030692, "F": 0.21768, "G": 0.22216, "chunks": 222.91, "bytes_per_chunk": 4.594, "boundary_entropy": 0.109, "router_noise": 0.0, "grad_norm": 0.7105, "lr": 0.0002226, "router_trainable": true, "sec_per_step": 0.815, "probe_sec": 0.0843, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06256, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.01546, "n_shadow": 4, "mean_delta_shadow": 0.02878, "n_split": 1, "mean_delta_split": 0.14453, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.71852, "trunc_abs_frac": 0.79503, "mean_delta_full": 0.25244, "read_hit_rate": {"surface": 0.52302, "split": 0.575, "merge": 0.16424, "skel": 0.3542}, "mean_gate": {"surface": 0.18382, "split": 0.08647, "merge": 0.12142, "skel": 0.1315}, "n_reads": 13934, "n_cand": 32016, "read_suppressed_frac": 0.0958, "dropout_reads_logged": 13934, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1036, "admitted_surface": 46137, "staged_surface": 56557, "admitted_split": 9951, "staged_split": 40270, "admitted_merge": 9021, "staged_merge": 118609, "admitted_skel": 427, "staged_skel": 46708, "mean_count": 275.62896728515625, "n_util_obs": 176, "util_count_spearman": 0.08272727272727273}, "probe_overhead_sec_total": 416.024, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.009805, "alpha": 0.03, "t": 2694.2}
{"event": "consolidate", "step": 2400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262118, "t": 2746.8}
{"event": "probe", "step": 2400, "phase": 2, "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": 0.210942, "boundary_entropy": 0.102552, "bpb": 2.064113}, "tr": {"chunks": 3788.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.209247, "cps": 29.59375, "mean_p": 0.204394, "boundary_entropy": 0.133335, "bpb": 2.258815}, "fi": {"chunks": 4960.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.263676, "cps": 38.75, "mean_p": 0.261397, "boundary_entropy": 0.124396, "bpb": 2.390449}}, "bpb": {"en": 2.06411, "tr": 2.25881, "fi": 2.39045}, "gini_cpc": 0.056758, "gini_cps": 0.074447, "bpic": 4.600762, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.016334646864966, "fi": 1.2806991022363816}, "inventory": {"size": 326656, "admitted": 65536, "staged": 261120, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1061, "admitted_surface": 45141, "staged_surface": 56260, "admitted_split": 10053, "staged_split": 40708, "admitted_merge": 9706, "staged_merge": 117488, "admitted_skel": 636, "staged_skel": 46664, "mean_count": 286.4940185546875, "n_util_obs": 179, "util_count_spearman": 0.07755742344979585}, "t": 2749.0}
{"event": "eval", "step": 2400, "phase": 1, "bpb": {"en": 2.07609, "tr": 2.26753, "fi": 2.40198}, "bytes_per_chunk": {"en": 4.595, "tr": 5.306, "fi": 4.059}, "bpb_mean": 2.24853, "frozen_inventory": true, "t": 2752.1}
{"event": "phase_transition", "step": 2400, "phase": 2, "objective": "global", "router_trainable": true, "encoder_trainable": true, "frozen_router_params": 0, "t": 2757.7}
{"event": "train", "step": 2400, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.46253, "bpb": 2.10999, "chunk_loss": 0.030626, "F": 0.21414, "G": 0.21961, "chunks": 219.28, "bytes_per_chunk": 4.67, "boundary_entropy": 0.1078, "router_noise": 0.0, "grad_norm": 0.6812, "lr": 0.0002192, "router_trainable": true, "sec_per_step": 1.281, "probe_sec": 0.0894, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06002, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.03158, "n_shadow": 7, "mean_delta_shadow": -0.08685, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.11353, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.34423, "trunc_abs_frac": 2.28031, "mean_delta_full": -0.03015, "read_hit_rate": {"surface": 0.535, "split": 0.59107, "merge": 0.14958, "skel": 0.39672}, "mean_gate": {"surface": 0.19494, "split": 0.09288, "merge": 0.12783, "skel": 0.13648}, "n_reads": 14015, "n_cand": 31312, "read_suppressed_frac": 0.0999, "dropout_reads_logged": 14015, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1062, "admitted_surface": 45141, "staged_surface": 56435, "admitted_split": 10053, "staged_split": 40744, "admitted_merge": 9706, "staged_merge": 118079, "admitted_skel": 636, "staged_skel": 46886, "mean_count": 286.7079162597656, "n_util_obs": 179, "util_count_spearman": 0.07831063126399264}, "probe_overhead_sec_total": 423.023, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.009878, "alpha": 0.03, "t": 2758.2}
{"event": "consolidate", "step": 2450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262117, "t": 2803.0}
{"event": "train", "step": 2450, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.46381, "bpb": 2.11183, "chunk_loss": 0.030922, "F": 0.2261, "G": 0.2305, "chunks": 231.53, "bytes_per_chunk": 4.423, "boundary_entropy": 0.10869, "router_noise": 0.0, "grad_norm": 0.7625, "lr": 0.0002158, "router_trainable": true, "sec_per_step": 0.91, "probe_sec": 0.1609, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.12131, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.15057, "n_shadow": 3, "mean_delta_shadow": 0.01688, "n_split": 2, "mean_delta_split": -0.00197, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.26257, "trunc_abs_frac": 0.98695, "mean_delta_full": 0.18308, "read_hit_rate": {"surface": 0.55632, "split": 0.5884, "merge": 0.17549, "skel": 0.49054}, "mean_gate": {"surface": 0.19654, "split": 0.09293, "merge": 0.13022, "skel": 0.13321}, "n_reads": 15504, "n_cand": 32820, "read_suppressed_frac": 0.0966, "dropout_reads_logged": 15504, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1101, "admitted_surface": 44159, "staged_surface": 56536, "admitted_split": 10154, "staged_split": 40768, "admitted_merge": 10151, "staged_merge": 118686, "admitted_skel": 1072, "staged_skel": 46154, "mean_count": 298.64752197265625, "n_util_obs": 186, "util_count_spearman": 0.09081025134156064}, "probe_overhead_sec_total": 429.267, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.008911, "alpha": 0.03, "t": 2803.7}
{"event": "consolidate", "step": 2500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262118, "t": 2880.0}
{"event": "train", "step": 2500, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.5621, "bpb": 2.25363, "chunk_loss": 0.030819, "F": 0.22263, "G": 0.22623, "chunks": 227.97, "bytes_per_chunk": 4.492, "boundary_entropy": 0.1084, "router_noise": 0.0, "grad_norm": 0.7114, "lr": 0.0002123, "router_trainable": true, "sec_per_step": 1.539, "probe_sec": 0.0816, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05556, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.05455, "n_shadow": 5, "mean_delta_shadow": 0.07906, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.03149, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 1.6029, "trunc_abs_frac": 0.78565, "mean_delta_full": 0.05546, "read_hit_rate": {"surface": 0.53793, "split": 0.57987, "merge": 0.16651, "skel": 0.44163}, "mean_gate": {"surface": 0.19326, "split": 0.09064, "merge": 0.12929, "skel": 0.12926}, "n_reads": 14541, "n_cand": 32006, "read_suppressed_frac": 0.099, "dropout_reads_logged": 14541, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1110, "admitted_surface": 43172, "staged_surface": 56603, "admitted_split": 10311, "staged_split": 40840, "admitted_merge": 10738, "staged_merge": 118518, "admitted_skel": 1315, "staged_skel": 46183, "mean_count": 309.9722900390625, "n_util_obs": 191, "util_count_spearman": 0.1294812620556627}, "probe_overhead_sec_total": 436.918, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.010147, "alpha": 0.03, "t": 2880.6}
{"event": "consolidate", "step": 2550, "retired": 1, "split_retired": 1, "merge_admitted": 0, "admitted": 2, "swapped": 1024, "uncertain_staged": 262117, "t": 2928.2}
{"event": "train", "step": 2550, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.48999, "bpb": 2.1496, "chunk_loss": 0.030659, "F": 0.21698, "G": 0.2224, "chunks": 222.19, "bytes_per_chunk": 4.609, "boundary_entropy": 0.10732, "router_noise": 0.0, "grad_norm": 0.7053, "lr": 0.0002088, "router_trainable": true, "sec_per_step": 0.987, "probe_sec": 0.3038, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.27347, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.0355, "n_shadow": 4, "mean_delta_shadow": -0.02299, "n_split": 0, "mean_delta_split": NaN, "n_merge": 3, "mean_delta_merge": -0.0049, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 3.83903, "trunc_abs_frac": 1.0111, "mean_delta_full": -0.01273, "read_hit_rate": {"surface": 0.53179, "split": 0.58605, "merge": 0.15853, "skel": 0.48117}, "mean_gate": {"surface": 0.19184, "split": 0.09016, "merge": 0.1277, "skel": 0.12969}, "n_reads": 14721, "n_cand": 31884, "read_suppressed_frac": 0.1014, "dropout_reads_logged": 14721, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1138, "admitted_surface": 42178, "staged_surface": 56754, "admitted_split": 10359, "staged_split": 40627, "admitted_merge": 11008, "staged_merge": 119152, "admitted_skel": 1991, "staged_skel": 45611, "mean_count": 320.1825866699219, "n_util_obs": 193, "util_count_spearman": 0.09875457166192725}, "probe_overhead_sec_total": 446.289, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.009959, "alpha": 0.03, "t": 2930.0}
{"event": "consolidate", "step": 2600, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 1024, "uncertain_staged": 262116, "t": 2989.1}
{"event": "probe", "step": 2600, "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": 0.209949, "boundary_entropy": 0.098555, "bpb": 2.040633}, "tr": {"chunks": 3887.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.214716, "cps": 30.367188, "mean_p": 0.202682, "boundary_entropy": 0.125334, "bpb": 2.240711}, "fi": {"chunks": 4956.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.263463, "cps": 38.71875, "mean_p": 0.259801, "boundary_entropy": 0.114599, "bpb": 2.351421}}, "bpb": {"en": 2.04063, "tr": 2.24071, "fi": 2.35142}, "gini_cpc": 0.055569, "gini_cps": 0.073127, "bpic": 4.562284, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0402836706364853, "fi": 1.2764599895881419}, "inventory": {"size": 326654, "admitted": 65536, "staged": 261118, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1139, "admitted_surface": 41408, "staged_surface": 56632, "admitted_split": 10595, "staged_split": 40278, "admitted_merge": 11295, "staged_merge": 119055, "admitted_skel": 2238, "staged_skel": 45153, "mean_count": 324.2825927734375, "n_util_obs": 182, "util_count_spearman": 0.05428418153714775}, "t": 2991.4}
{"event": "eval", "step": 2600, "phase": 2, "bpb": {"en": 2.05184, "tr": 2.25027, "fi": 2.36498}, "bytes_per_chunk": {"en": 4.571, "tr": 5.125, "fi": 4.054}, "bpb_mean": 2.22236, "frozen_inventory": true, "t": 2995.0}
{"event": "train", "step": 2600, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.49964, "bpb": 2.16352, "chunk_loss": 0.030867, "F": 0.22293, "G": 0.22672, "chunks": 228.28, "bytes_per_chunk": 4.486, "boundary_entropy": 0.10182, "router_noise": 0.0, "grad_norm": 0.7105, "lr": 0.0002053, "router_trainable": true, "sec_per_step": 1.392, "probe_sec": 0.1591, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.13361, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": 0.08009, "n_shadow": 8, "mean_delta_shadow": 0.02321, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.13049, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.36567, "trunc_abs_frac": 0.2244, "mean_delta_full": 0.07002, "read_hit_rate": {"surface": 0.52069, "split": 0.57163, "merge": 0.17636, "skel": 0.48049}, "mean_gate": {"surface": 0.19118, "split": 0.08661, "merge": 0.12949, "skel": 0.12795}, "n_reads": 14866, "n_cand": 32544, "read_suppressed_frac": 0.1016, "dropout_reads_logged": 14866, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1140, "admitted_surface": 41408, "staged_surface": 56885, "admitted_split": 10595, "staged_split": 40393, "admitted_merge": 11295, "staged_merge": 119585, "admitted_skel": 2238, "staged_skel": 45281, "mean_count": 324.50946044921875, "n_util_obs": 182, "util_count_spearman": 0.05420256765243632}, "probe_overhead_sec_total": 460.707, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.0096, "alpha": 0.03, "t": 2999.6}
{"event": "consolidate", "step": 2650, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262117, "t": 3051.0}
{"event": "train", "step": 2650, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.43679, "bpb": 2.07285, "chunk_loss": 0.030597, "F": 0.21439, "G": 0.21677, "chunks": 219.53, "bytes_per_chunk": 4.664, "boundary_entropy": 0.10319, "router_noise": 0.0, "grad_norm": 0.6718, "lr": 0.0002017, "router_trainable": true, "sec_per_step": 1.05, "probe_sec": 0.2173, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.18452, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 3, "mean_delta_loo": 0.05444, "n_shadow": 3, "mean_delta_shadow": 0.00826, "n_split": 3, "mean_delta_split": 0.09041, "n_merge": 1, "mean_delta_merge": 0.0354, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 1.32148, "trunc_abs_frac": 2.26829, "mean_delta_full": 0.03949, "read_hit_rate": {"surface": 0.4975, "split": 0.56676, "merge": 0.1665, "skel": 0.45813}, "mean_gate": {"surface": 0.18092, "split": 0.08436, "merge": 0.11745, "skel": 0.12555}, "n_reads": 13907, "n_cand": 31320, "read_suppressed_frac": 0.0996, "dropout_reads_logged": 13907, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1148, "admitted_surface": 40493, "staged_surface": 56992, "admitted_split": 10728, "staged_split": 40982, "admitted_merge": 11568, "staged_merge": 119247, "admitted_skel": 2747, "staged_skel": 44923, "mean_count": 330.4892578125, "n_util_obs": 175, "util_count_spearman": -0.016022176456661968}, "probe_overhead_sec_total": 471.318, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.009571, "alpha": 0.03, "t": 3052.1}
{"event": "consolidate", "step": 2700, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262116, "t": 3111.9}
{"event": "train", "step": 2700, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.51296, "bpb": 2.18274, "chunk_loss": 0.030683, "F": 0.21414, "G": 0.21955, "chunks": 219.28, "bytes_per_chunk": 4.67, "boundary_entropy": 0.109, "router_noise": 0.0, "grad_norm": 0.7567, "lr": 0.0001981, "router_trainable": true, "sec_per_step": 1.214, "probe_sec": 0.2208, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.19716, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.004, "n_shadow": 6, "mean_delta_shadow": 0.03694, "n_split": 1, "mean_delta_split": 0.14148, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -0.23013, "trunc_abs_frac": 0.63212, "mean_delta_full": -0.19541, "read_hit_rate": {"surface": 0.47645, "split": 0.57737, "merge": 0.14958, "skel": 0.44869}, "mean_gate": {"surface": 0.20065, "split": 0.09174, "merge": 0.13506, "skel": 0.13822}, "n_reads": 13743, "n_cand": 31370, "read_suppressed_frac": 0.1014, "dropout_reads_logged": 13743, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1163, "admitted_surface": 39600, "staged_surface": 57054, "admitted_split": 11090, "staged_split": 40836, "admitted_merge": 11910, "staged_merge": 119319, "admitted_skel": 2936, "staged_skel": 44935, "mean_count": 339.5269775390625, "n_util_obs": 176, "util_count_spearman": -0.03941668500990535}, "probe_overhead_sec_total": 485.728, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.010055, "alpha": 0.03, "t": 3112.8}
{"event": "consolidate", "step": 2750, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262115, "t": 3163.9}
{"event": "train", "step": 2750, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.45522, "bpb": 2.09944, "chunk_loss": 0.030683, "F": 0.21817, "G": 0.22182, "chunks": 223.41, "bytes_per_chunk": 4.584, "boundary_entropy": 0.10053, "router_noise": 0.0, "grad_norm": 0.7194, "lr": 0.0001945, "router_trainable": true, "sec_per_step": 1.037, "probe_sec": 0.201, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.17747, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.00253, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.09631, "n_merge": 1, "mean_delta_merge": 0.18811, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 388.25, "trunc_abs_frac": 0.53947, "mean_delta_full": 0.00015, "read_hit_rate": {"surface": 0.46607, "split": 0.57082, "merge": 0.16316, "skel": 0.4505}, "mean_gate": {"surface": 0.18843, "split": 0.0862, "merge": 0.12985, "skel": 0.12961}, "n_reads": 13892, "n_cand": 31829, "read_suppressed_frac": 0.096, "dropout_reads_logged": 13892, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1170, "admitted_surface": 39012, "staged_surface": 56888, "admitted_split": 11161, "staged_split": 40681, "admitted_merge": 12211, "staged_merge": 119794, "admitted_skel": 3152, "staged_skel": 44781, "mean_count": 348.0810546875, "n_util_obs": 179, "util_count_spearman": -0.06998831941306138}, "probe_overhead_sec_total": 496.392, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.010136, "alpha": 0.03, "t": 3164.7}
{"event": "consolidate", "step": 2800, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262114, "t": 3215.4}
{"event": "probe", "step": 2800, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3572.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.205252, "cps": 27.90625, "mean_p": 0.207939, "boundary_entropy": 0.097259, "bpb": 2.02106}, "tr": {"chunks": 3771.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.208308, "cps": 29.460938, "mean_p": 0.198632, "boundary_entropy": 0.13114, "bpb": 2.222636}, "fi": {"chunks": 4941.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.262665, "cps": 38.601562, "mean_p": 0.258213, "boundary_entropy": 0.109235, "bpb": 2.327642}}, "bpb": {"en": 2.02106, "tr": 2.22264, "fi": 2.32764}, "gini_cpc": 0.056602, "gini_cps": 0.074297, "bpic": 4.618365, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0148892467322888, "fi": 1.2797220128581457}, "inventory": {"size": 326655, "admitted": 65536, "staged": 261119, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1148, "admitted_surface": 38674, "staged_surface": 57158, "admitted_split": 10759, "staged_split": 40283, "admitted_merge": 12699, "staged_merge": 119163, "admitted_skel": 3404, "staged_skel": 44515, "mean_count": 328.48858642578125, "n_util_obs": 154, "util_count_spearman": -0.1840935787804858}, "t": 3217.6}
{"event": "eval", "step": 2800, "phase": 2, "bpb": {"en": 2.03389, "tr": 2.23146, "fi": 2.33923}, "bytes_per_chunk": {"en": 4.611, "tr": 5.292, "fi": 4.063}, "bpb_mean": 2.20153, "frozen_inventory": true, "t": 3221.3}
{"event": "train", "step": 2800, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.44383, "bpb": 2.083, "chunk_loss": 0.030725, "F": 0.22043, "G": 0.22259, "chunks": 225.72, "bytes_per_chunk": 4.537, "boundary_entropy": 0.0973, "router_noise": 0.0, "grad_norm": 0.7244, "lr": 0.0001909, "router_trainable": true, "sec_per_step": 1.218, "probe_sec": 0.1431, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.11684, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 11, "mean_delta_loo": 0.15506, "n_shadow": 3, "mean_delta_shadow": -0.14851, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.20089, "trunc_abs_frac": 0.60858, "mean_delta_full": 0.16748, "read_hit_rate": {"surface": 0.46002, "split": 0.50558, "merge": 0.16245, "skel": 0.42577}, "mean_gate": {"surface": 0.18129, "split": 0.08551, "merge": 0.12499, "skel": 0.12844}, "n_reads": 12983, "n_cand": 32023, "read_suppressed_frac": 0.1053, "dropout_reads_logged": 12983, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1147, "admitted_surface": 38674, "staged_surface": 57297, "admitted_split": 10759, "staged_split": 40450, "admitted_merge": 12699, "staged_merge": 119746, "admitted_skel": 3404, "staged_skel": 44651, "mean_count": 328.6866760253906, "n_util_obs": 154, "util_count_spearman": -0.1841888661623007}, "probe_overhead_sec_total": 507.363, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.01035, "alpha": 0.03, "t": 3225.6}
{"event": "consolidate", "step": 2850, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262114, "t": 3275.8}
{"event": "train", "step": 2850, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.54431, "bpb": 2.22797, "chunk_loss": 0.030881, "F": 0.22223, "G": 0.22547, "chunks": 227.56, "bytes_per_chunk": 4.5, "boundary_entropy": 0.09891, "router_noise": 0.0, "grad_norm": 0.6596, "lr": 0.0001872, "router_trainable": true, "sec_per_step": 1.025, "probe_sec": 0.1666, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.13991, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": -0.15628, "n_shadow": 3, "mean_delta_shadow": -0.03759, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": 0.06409, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 2.46412, "trunc_abs_frac": 1.22848, "mean_delta_full": -0.03573, "read_hit_rate": {"surface": 0.42772, "split": 0.44772, "merge": 0.15337, "skel": 0.42531}, "mean_gate": {"surface": 0.1912, "split": 0.08981, "merge": 0.13337, "skel": 0.13269}, "n_reads": 12143, "n_cand": 32465, "read_suppressed_frac": 0.1016, "dropout_reads_logged": 12143, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1140, "admitted_surface": 38265, "staged_surface": 57673, "admitted_split": 10538, "staged_split": 40870, "admitted_merge": 13041, "staged_merge": 118808, "admitted_skel": 3692, "staged_skel": 44793, "mean_count": 315.06439208984375, "n_util_obs": 150, "util_count_spearman": -0.16037512778345706}, "probe_overhead_sec_total": 518.318, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.010123, "alpha": 0.03, "t": 3276.8}
{"event": "consolidate", "step": 2900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262113, "t": 3337.3}
{"event": "train", "step": 2900, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.48843, "bpb": 2.14735, "chunk_loss": 0.030979, "F": 0.22836, "G": 0.23034, "chunks": 233.84, "bytes_per_chunk": 4.379, "boundary_entropy": 0.10269, "router_noise": 0.0, "grad_norm": 0.6578, "lr": 0.0001835, "router_trainable": true, "sec_per_step": 1.229, "probe_sec": 0.2477, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.22396, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": -0.00854, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 3, "mean_delta_split": 0.00304, "n_merge": 2, "mean_delta_merge": 0.09113, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.49045, "trunc_abs_frac": 7.99845, "mean_delta_full": 0.15942, "read_hit_rate": {"surface": 0.439, "split": 0.41289, "merge": 0.15743, "skel": 0.41186}, "mean_gate": {"surface": 0.20094, "split": 0.09315, "merge": 0.12937, "skel": 0.14402}, "n_reads": 11898, "n_cand": 32949, "read_suppressed_frac": 0.0964, "dropout_reads_logged": 11898, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1152, "admitted_surface": 38438, "staged_surface": 57282, "admitted_split": 9682, "staged_split": 41504, "admitted_merge": 13311, "staged_merge": 118944, "admitted_skel": 4105, "staged_skel": 44414, "mean_count": 307.1258544921875, "n_util_obs": 146, "util_count_spearman": -0.10897400055912776}, "probe_overhead_sec_total": 528.94, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.010414, "alpha": 0.03, "t": 3338.3}
{"event": "consolidate", "step": 2950, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262113, "t": 3391.3}
{"event": "train", "step": 2950, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.42502, "bpb": 2.05588, "chunk_loss": 0.030669, "F": 0.21576, "G": 0.21813, "chunks": 220.94, "bytes_per_chunk": 4.635, "boundary_entropy": 0.09761, "router_noise": 0.0, "grad_norm": 0.6558, "lr": 0.0001799, "router_trainable": true, "sec_per_step": 1.086, "probe_sec": 0.1576, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.12724, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.06604, "n_shadow": 5, "mean_delta_shadow": 0.02124, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": 0.03748, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 1.75877, "trunc_abs_frac": 1.29646, "mean_delta_full": -0.06262, "read_hit_rate": {"surface": 0.39812, "split": 0.20894, "merge": 0.12718, "skel": 0.39597}, "mean_gate": {"surface": 0.20709, "split": 0.09524, "merge": 0.13861, "skel": 0.14733}, "n_reads": 8350, "n_cand": 31574, "read_suppressed_frac": 0.1035, "dropout_reads_logged": 8350, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1120, "admitted_surface": 38935, "staged_surface": 56722, "admitted_split": 8829, "staged_split": 42314, "admitted_merge": 13208, "staged_merge": 118791, "admitted_skel": 4564, "staged_skel": 44317, "mean_count": 236.68484497070312, "n_util_obs": 129, "util_count_spearman": -0.09799600010631232}, "probe_overhead_sec_total": 539.009, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.011298, "alpha": 0.03, "t": 3392.6}
{"event": "consolidate", "step": 3000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262112, "t": 3448.1}
{"event": "probe", "step": 3000, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3585.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.205999, "cps": 28.007812, "mean_p": 0.208633, "boundary_entropy": 0.095043, "bpb": 2.002661}, "tr": {"chunks": 3785.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.209081, "cps": 29.570312, "mean_p": 0.198562, "boundary_entropy": 0.125179, "bpb": 2.191187}, "fi": {"chunks": 4970.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.264207, "cps": 38.828125, "mean_p": 0.2593, "boundary_entropy": 0.10377, "bpb": 2.304062}}, "bpb": {"en": 2.00266, "tr": 2.19119, "fi": 2.30406}, "gini_cpc": 0.057127, "gini_cps": 0.074824, "bpic": 4.597407, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0149631917962696, "fi": 1.2825652399145964}, "inventory": {"size": 326656, "admitted": 65536, "staged": 261120, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1015, "admitted_surface": 39606, "staged_surface": 56193, "admitted_split": 8166, "staged_split": 42509, "admitted_merge": 13024, "staged_merge": 117810, "admitted_skel": 4740, "staged_skel": 44608, "mean_count": 220.44517517089844, "n_util_obs": 124, "util_count_spearman": -0.14865145554681353}, "t": 3450.0}
{"event": "eval", "step": 3000, "phase": 2, "bpb": {"en": 2.01433, "tr": 2.19946, "fi": 2.31427}, "bytes_per_chunk": {"en": 4.589, "tr": 5.26, "fi": 4.05}, "bpb_mean": 2.17602, "frozen_inventory": true, "t": 3453.7}
{"event": "train", "step": 3000, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.45467, "bpb": 2.09864, "chunk_loss": 0.030633, "F": 0.21701, "G": 0.21893, "chunks": 222.22, "bytes_per_chunk": 4.608, "boundary_entropy": 0.09573, "router_noise": 0.0, "grad_norm": 0.681, "lr": 0.0001761, "router_trainable": true, "sec_per_step": 1.348, "probe_sec": 0.2496, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.22315, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.00953, "n_shadow": 5, "mean_delta_shadow": -0.07486, "n_split": 2, "mean_delta_split": -0.07974, "n_merge": 1, "mean_delta_merge": 0.02509, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -0.83343, "trunc_abs_frac": 1.11875, "mean_delta_full": 0.08845, "read_hit_rate": {"surface": 0.3629, "split": 0.1783, "merge": 0.12232, "skel": 0.39014}, "mean_gate": {"surface": 0.20107, "split": 0.08992, "merge": 0.13805, "skel": 0.13889}, "n_reads": 7715, "n_cand": 31767, "read_suppressed_frac": 0.0988, "dropout_reads_logged": 7715, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1017, "admitted_surface": 39606, "staged_surface": 56534, "admitted_split": 8166, "staged_split": 42719, "admitted_merge": 13024, "staged_merge": 118107, "admitted_skel": 4740, "staged_skel": 44784, "mean_count": 220.5629425048828, "n_util_obs": 124, "util_count_spearman": -0.1490102281667978}, "probe_overhead_sec_total": 549.541, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.012713, "alpha": 0.03, "t": 3460.0}
{"event": "consolidate", "step": 3050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262109, "t": 3505.4}
{"event": "train", "step": 3050, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.44695, "bpb": 2.0875, "chunk_loss": 0.030659, "F": 0.21783, "G": 0.22041, "chunks": 223.06, "bytes_per_chunk": 4.591, "boundary_entropy": 0.09796, "router_noise": 0.0, "grad_norm": 0.7155, "lr": 0.0001724, "router_trainable": true, "sec_per_step": 0.923, "probe_sec": 0.1079, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08158, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": -0.0813, "n_shadow": 7, "mean_delta_shadow": -0.02431, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 12.58795, "trunc_abs_frac": 0.99449, "mean_delta_full": -0.02533, "read_hit_rate": {"surface": 0.33437, "split": 0.10542, "merge": 0.07789, "skel": 0.30791}, "mean_gate": {"surface": 0.21535, "split": 0.11338, "merge": 0.15317, "skel": 0.16014}, "n_reads": 5882, "n_cand": 31837, "read_suppressed_frac": 0.0993, "dropout_reads_logged": 5882, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 911, "admitted_surface": 40462, "staged_surface": 56326, "admitted_split": 7738, "staged_split": 43334, "admitted_merge": 12599, "staged_merge": 117822, "admitted_skel": 4737, "staged_skel": 44662, "mean_count": 180.67990112304688, "n_util_obs": 112, "util_count_spearman": -0.11096115078415963}, "probe_overhead_sec_total": 556.448, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.014075, "alpha": 0.03, "t": 3506.1}
{"event": "consolidate", "step": 3100, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262105, "t": 3559.1}
{"event": "train", "step": 3100, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.53716, "bpb": 2.21766, "chunk_loss": 0.030702, "F": 0.21515, "G": 0.2169, "chunks": 220.31, "bytes_per_chunk": 4.648, "boundary_entropy": 0.09974, "router_noise": 0.0, "grad_norm": 0.7488, "lr": 0.0001687, "router_trainable": true, "sec_per_step": 1.071, "probe_sec": 0.1036, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0782, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.05302, "n_shadow": 6, "mean_delta_shadow": -0.12082, "n_split": 1, "mean_delta_split": -0.03186, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.31501, "trunc_abs_frac": 0.57073, "mean_delta_full": 0.32367, "read_hit_rate": {"surface": 0.30977, "split": 0.06512, "merge": 0.06928, "skel": 0.21004}, "mean_gate": {"surface": 0.22232, "split": 0.11003, "merge": 0.17253, "skel": 0.17107}, "n_reads": 4541, "n_cand": 31265, "read_suppressed_frac": 0.1033, "dropout_reads_logged": 4541, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 842, "admitted_surface": 40903, "staged_surface": 56729, "admitted_split": 7343, "staged_split": 43537, "admitted_merge": 12598, "staged_merge": 117277, "admitted_skel": 4692, "staged_skel": 44601, "mean_count": 146.2000732421875, "n_util_obs": 101, "util_count_spearman": 0.04161910308677927}, "probe_overhead_sec_total": 562.68, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.017076, "alpha": 0.03, "t": 3559.7}
{"event": "consolidate", "step": 3150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262099, "t": 3611.8}
{"event": "train", "step": 3150, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.45232, "bpb": 2.09525, "chunk_loss": 0.030691, "F": 0.21774, "G": 0.21891, "chunks": 222.97, "bytes_per_chunk": 4.593, "boundary_entropy": 0.09618, "router_noise": 0.0, "grad_norm": 0.688, "lr": 0.000165, "router_trainable": true, "sec_per_step": 1.055, "probe_sec": 0.1296, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10955, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": -0.13423, "n_shadow": 4, "mean_delta_shadow": -0.03026, "n_split": 2, "mean_delta_split": 0.06996, "n_merge": 1, "mean_delta_merge": 0.08026, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 1.13957, "trunc_abs_frac": 1.81373, "mean_delta_full": -0.05969, "read_hit_rate": {"surface": 0.29579, "split": 0.05969, "merge": 0.0618, "skel": 0.21727}, "mean_gate": {"surface": 0.23708, "split": 0.11372, "merge": 0.16873, "skel": 0.18113}, "n_reads": 4435, "n_cand": 31800, "read_suppressed_frac": 0.0965, "dropout_reads_logged": 4435, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 865, "admitted_surface": 40129, "staged_surface": 56944, "admitted_split": 7575, "staged_split": 43650, "admitted_merge": 12740, "staged_merge": 117163, "admitted_skel": 5092, "staged_skel": 44387, "mean_count": 147.33920288085938, "n_util_obs": 101, "util_count_spearman": -0.009854397204426326}, "probe_overhead_sec_total": 569.206, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.016741, "alpha": 0.03, "t": 3612.4}
{"event": "consolidate", "step": 3200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262096, "t": 3662.6}
{"event": "probe", "step": 3200, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3575.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.205424, "cps": 27.929688, "mean_p": 0.206896, "boundary_entropy": 0.093588, "bpb": 1.983345}, "tr": {"chunks": 3760.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.2077, "cps": 29.375, "mean_p": 0.192731, "boundary_entropy": 0.132524, "bpb": 2.169466}, "fi": {"chunks": 4998.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.265696, "cps": 39.046875, "mean_p": 0.258443, "boundary_entropy": 0.105014, "bpb": 2.277007}}, "bpb": {"en": 1.98334, "tr": 2.16947, "fi": 2.27701}, "gini_cpc": 0.059192, "gini_cps": 0.076921, "bpic": 4.600016, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0110796456485016, "fi": 1.2933987664560203}, "inventory": {"size": 326656, "admitted": 65536, "staged": 261120, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 893, "admitted_surface": 39219, "staged_surface": 56908, "admitted_split": 7694, "staged_split": 43591, "admitted_merge": 13158, "staged_merge": 116251, "admitted_skel": 5465, "staged_skel": 44370, "mean_count": 150.7373809814453, "n_util_obs": 105, "util_count_spearman": 0.018059299191374664}, "t": 3664.4}
{"event": "eval", "step": 3200, "phase": 2, "bpb": {"en": 1.99629, "tr": 2.17914, "fi": 2.28667}, "bytes_per_chunk": {"en": 4.598, "tr": 5.303, "fi": 4.029}, "bpb_mean": 2.15403, "frozen_inventory": true, "t": 3667.5}
{"event": "train", "step": 3200, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.42095, "bpb": 2.05, "chunk_loss": 0.030867, "F": 0.22501, "G": 0.22467, "chunks": 230.41, "bytes_per_chunk": 4.444, "boundary_entropy": 0.09722, "router_noise": 0.0, "grad_norm": 0.6534, "lr": 0.0001613, "router_trainable": true, "sec_per_step": 1.22, "probe_sec": 0.1204, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09078, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 3, "mean_delta_loo": 0.06889, "n_shadow": 2, "mean_delta_shadow": 0.06433, "n_split": 2, "mean_delta_split": -0.0798, "n_merge": 3, "mean_delta_merge": -0.35144, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.04654, "trunc_abs_frac": 0.06762, "mean_delta_full": -0.13788, "read_hit_rate": {"surface": 0.31644, "split": 0.0712, "merge": 0.07388, "skel": 0.21341}, "mean_gate": {"surface": 0.22132, "split": 0.10126, "merge": 0.15601, "skel": 0.16738}, "n_reads": 4941, "n_cand": 32875, "read_suppressed_frac": 0.0996, "dropout_reads_logged": 4941, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 896, "admitted_surface": 39219, "staged_surface": 57080, "admitted_split": 7694, "staged_split": 43613, "admitted_merge": 13158, "staged_merge": 116984, "admitted_skel": 5465, "staged_skel": 44467, "mean_count": 150.8127899169922, "n_util_obs": 105, "util_count_spearman": 0.018059299191374664}, "probe_overhead_sec_total": 575.626, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.016559, "alpha": 0.03, "t": 3673.4}
{"event": "consolidate", "step": 3250, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262096, "t": 3720.6}
{"event": "train", "step": 3250, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.40136, "bpb": 2.02173, "chunk_loss": 0.030537, "F": 0.21332, "G": 0.21345, "chunks": 218.44, "bytes_per_chunk": 4.688, "boundary_entropy": 0.09543, "router_noise": 0.0, "grad_norm": 0.6915, "lr": 0.0001576, "router_trainable": true, "sec_per_step": 0.962, "probe_sec": 0.2292, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.1949, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 3, "mean_delta_loo": -0.04015, "n_shadow": 11, "mean_delta_shadow": -0.06419, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.13317, "trunc_abs_frac": 0.2137, "mean_delta_full": -0.2933, "read_hit_rate": {"surface": 0.42828, "split": 0.21843, "merge": 0.10959, "skel": 0.26812}, "mean_gate": {"surface": 0.21119, "split": 0.0968, "merge": 0.13705, "skel": 0.158}, "n_reads": 7788, "n_cand": 31228, "read_suppressed_frac": 0.1018, "dropout_reads_logged": 7788, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 918, "admitted_surface": 38385, "staged_surface": 57150, "admitted_split": 8058, "staged_split": 43122, "admitted_merge": 13541, "staged_merge": 117164, "admitted_skel": 5552, "staged_skel": 44708, "mean_count": 168.81483459472656, "n_util_obs": 114, "util_count_spearman": 0.15617747716733157}, "probe_overhead_sec_total": 583.008, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.016481, "alpha": 0.03, "t": 3721.5}
{"event": "consolidate", "step": 3300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 272, "uncertain_staged": 262099, "t": 3767.0}
{"event": "train", "step": 3300, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.44198, "bpb": 2.08033, "chunk_loss": 0.030748, "F": 0.22092, "G": 0.22149, "chunks": 226.22, "bytes_per_chunk": 4.527, "boundary_entropy": 0.09475, "router_noise": 0.0, "grad_norm": 0.6571, "lr": 0.0001539, "router_trainable": true, "sec_per_step": 0.928, "probe_sec": 0.1246, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09238, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": -0.01778, "n_shadow": 5, "mean_delta_shadow": 0.10542, "n_split": 1, "mean_delta_split": 0.0144, "n_merge": 1, "mean_delta_merge": -0.05493, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.76516, "trunc_abs_frac": 0.73122, "mean_delta_full": 0.13034, "read_hit_rate": {"surface": 0.49105, "split": 0.29348, "merge": 0.13171, "skel": 0.33169}, "mean_gate": {"surface": 0.19173, "split": 0.0834, "merge": 0.12862, "skel": 0.13397}, "n_reads": 9966, "n_cand": 32347, "read_suppressed_frac": 0.0983, "dropout_reads_logged": 9966, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 936, "admitted_surface": 38253, "staged_surface": 56934, "admitted_split": 8097, "staged_split": 43001, "admitted_merge": 13620, "staged_merge": 117561, "admitted_skel": 5566, "staged_skel": 44648, "mean_count": 199.81333923339844, "n_util_obs": 129, "util_count_spearman": 0.18572786225402504}, "probe_overhead_sec_total": 588.907, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.01566, "alpha": 0.03, "t": 3767.9}
{"event": "consolidate", "step": 3350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 667, "uncertain_staged": 262105, "t": 3819.0}
{"event": "train", "step": 3350, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.46693, "bpb": 2.11633, "chunk_loss": 0.030864, "F": 0.22372, "G": 0.22465, "chunks": 229.09, "bytes_per_chunk": 4.47, "boundary_entropy": 0.09894, "router_noise": 0.0, "grad_norm": 0.7195, "lr": 0.0001501, "router_trainable": true, "sec_per_step": 1.053, "probe_sec": 0.1201, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0606, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": -0.1157, "n_shadow": 4, "mean_delta_shadow": -0.17176, "n_split": 2, "mean_delta_split": 0.11078, "n_merge": 1, "mean_delta_merge": 0.11829, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.98444, "trunc_abs_frac": 4.33033, "mean_delta_full": -0.17419, "read_hit_rate": {"surface": 0.48842, "split": 0.36555, "merge": 0.14421, "skel": 0.33494}, "mean_gate": {"surface": 0.2018, "split": 0.08158, "merge": 0.13673, "skel": 0.13623}, "n_reads": 10935, "n_cand": 32312, "read_suppressed_frac": 0.101, "dropout_reads_logged": 10935, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 973, "admitted_surface": 37615, "staged_surface": 57361, "admitted_split": 8651, "staged_split": 43095, "admitted_merge": 13694, "staged_merge": 116661, "admitted_skel": 5576, "staged_skel": 45027, "mean_count": 209.65261840820312, "n_util_obs": 139, "util_count_spearman": 0.21313284775905983}, "probe_overhead_sec_total": 595.224, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.015072, "alpha": 0.03, "t": 3820.5}
{"event": "consolidate", "step": 3400, "retired": 1, "split_retired": 0, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262107, "t": 3862.3}
{"event": "probe", "step": 3400, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3578.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.205597, "cps": 27.953125, "mean_p": 0.207459, "boundary_entropy": 0.092356, "bpb": 1.963022}, "tr": {"chunks": 3767.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.208087, "cps": 29.429688, "mean_p": 0.192184, "boundary_entropy": 0.123727, "bpb": 2.133982}, "fi": {"chunks": 4961.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.263729, "cps": 38.757812, "mean_p": 0.256832, "boundary_entropy": 0.102468, "bpb": 2.248123}}, "bpb": {"en": 1.96302, "tr": 2.13398, "fi": 2.24812}, "gini_cpc": 0.05721, "gini_cps": 0.074923, "bpic": 4.610109, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0121126494757793, "fi": 1.28274735424568}, "inventory": {"size": 326655, "admitted": 65536, "staged": 261119, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1006, "admitted_surface": 36602, "staged_surface": 56943, "admitted_split": 9640, "staged_split": 41755, "admitted_merge": 13712, "staged_merge": 117701, "admitted_skel": 5582, "staged_skel": 44720, "mean_count": 222.93466186523438, "n_util_obs": 148, "util_count_spearman": 0.23320759466492927}, "t": 3864.0}
{"event": "eval", "step": 3400, "phase": 2, "bpb": {"en": 1.97606, "tr": 2.14548, "fi": 2.25597}, "bytes_per_chunk": {"en": 4.588, "tr": 5.277, "fi": 4.067}, "bpb_mean": 2.12584, "frozen_inventory": true, "t": 3866.6}
{"event": "train", "step": 3400, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.38402, "bpb": 1.99671, "chunk_loss": 0.030612, "F": 0.21463, "G": 0.21539, "chunks": 219.78, "bytes_per_chunk": 4.659, "boundary_entropy": 0.09313, "router_noise": 0.0, "grad_norm": 0.6625, "lr": 0.0001465, "router_trainable": true, "sec_per_step": 1.012, "probe_sec": 0.0946, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06773, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 3, "mean_delta_loo": -0.07292, "n_shadow": 3, "mean_delta_shadow": 0.03761, "n_split": 3, "mean_delta_split": 0.07291, "n_merge": 1, "mean_delta_merge": -0.43163, "n_dropped_unpaired": 0, "n_full": 0, "read_hit_rate": {"surface": 0.48707, "split": 0.47561, "merge": 0.14019, "skel": 0.32517}, "mean_gate": {"surface": 0.17723, "split": 0.07193, "merge": 0.11891, "skel": 0.1243}, "n_reads": 11884, "n_cand": 31412, "read_suppressed_frac": 0.0978, "dropout_reads_logged": 11884, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1009, "admitted_surface": 36602, "staged_surface": 57155, "admitted_split": 9640, "staged_split": 41986, "admitted_merge": 13712, "staged_merge": 118111, "admitted_skel": 5582, "staged_skel": 44892, "mean_count": 223.11599731445312, "n_util_obs": 148, "util_count_spearman": 0.23330014029918153}, "probe_overhead_sec_total": 601.958, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.013598, "alpha": 0.03, "t": 3871.2}
{"event": "consolidate", "step": 3450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 398, "uncertain_staged": 262109, "t": 3914.1}
{"event": "train", "step": 3450, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.38282, "bpb": 1.99499, "chunk_loss": 0.030506, "F": 0.21283, "G": 0.21254, "chunks": 217.94, "bytes_per_chunk": 4.699, "boundary_entropy": 0.09197, "router_noise": 0.0, "grad_norm": 0.6412, "lr": 0.0001428, "router_trainable": true, "sec_per_step": 0.878, "probe_sec": 0.0928, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06052, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.02646, "n_shadow": 3, "mean_delta_shadow": -0.06108, "n_split": 1, "mean_delta_split": 0.04797, "n_merge": 1, "mean_delta_merge": -0.30353, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.59517, "trunc_abs_frac": 0.68652, "mean_delta_full": 0.0752, "read_hit_rate": {"surface": 0.47825, "split": 0.4966, "merge": 0.13256, "skel": 0.40128}, "mean_gate": {"surface": 0.17372, "split": 0.06661, "merge": 0.11949, "skel": 0.11504}, "n_reads": 12346, "n_cand": 31202, "read_suppressed_frac": 0.1025, "dropout_reads_logged": 12346, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1060, "admitted_surface": 36229, "staged_surface": 57483, "admitted_split": 9832, "staged_split": 41681, "admitted_merge": 13729, "staged_merge": 118046, "admitted_skel": 5746, "staged_skel": 44934, "mean_count": 233.9353485107422, "n_util_obs": 154, "util_count_spearman": 0.2427938917502485}, "probe_overhead_sec_total": 607.291, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.013178, "alpha": 0.03, "t": 3915.0}
{"event": "consolidate", "step": 3500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 201, "uncertain_staged": 262110, "t": 3963.0}
{"event": "train", "step": 3500, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.46472, "bpb": 2.11314, "chunk_loss": 0.030758, "F": 0.21716, "G": 0.21801, "chunks": 222.38, "bytes_per_chunk": 4.605, "boundary_entropy": 0.09055, "router_noise": 0.0, "grad_norm": 0.6719, "lr": 0.0001391, "router_trainable": true, "sec_per_step": 0.971, "probe_sec": 0.1748, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.14786, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": 0.04813, "n_shadow": 5, "mean_delta_shadow": -0.11183, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": -0.08649, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": -1.67438, "trunc_abs_frac": 1.67438, "mean_delta_full": -0.07385, "read_hit_rate": {"surface": 0.48899, "split": 0.50574, "merge": 0.12961, "skel": 0.44494}, "mean_gate": {"surface": 0.18024, "split": 0.06851, "merge": 0.12119, "skel": 0.11559}, "n_reads": 12938, "n_cand": 31709, "read_suppressed_frac": 0.1016, "dropout_reads_logged": 12938, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1090, "admitted_surface": 36064, "staged_surface": 57481, "admitted_split": 9907, "staged_split": 41694, "admitted_merge": 13736, "staged_merge": 118065, "admitted_skel": 5829, "staged_skel": 44904, "mean_count": 244.32156372070312, "n_util_obs": 161, "util_count_spearman": 0.21267253278122844}, "probe_overhead_sec_total": 613.505, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.013269, "alpha": 0.03, "t": 3963.6}
{"event": "consolidate", "step": 3550, "retired": 1, "split_retired": 1, "merge_admitted": 0, "admitted": 2, "swapped": 18, "uncertain_staged": 262110, "t": 4006.7}
{"event": "train", "step": 3550, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.46548, "bpb": 2.11424, "chunk_loss": 0.030961, "F": 0.22791, "G": 0.22806, "chunks": 233.38, "bytes_per_chunk": 4.388, "boundary_entropy": 0.09253, "router_noise": 0.0, "grad_norm": 0.7053, "lr": 0.0001355, "router_trainable": true, "sec_per_step": 0.874, "probe_sec": 0.1627, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.13663, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.01505, "n_shadow": 5, "mean_delta_shadow": 0.02272, "n_split": 2, "mean_delta_split": 0.07437, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 2.29265, "trunc_abs_frac": 0.81931, "mean_delta_full": -0.02175, "read_hit_rate": {"surface": 0.51547, "split": 0.49922, "merge": 0.1429, "skel": 0.45554}, "mean_gate": {"surface": 0.17847, "split": 0.06687, "merge": 0.12084, "skel": 0.11334}, "n_reads": 13843, "n_cand": 33266, "read_suppressed_frac": 0.1011, "dropout_reads_logged": 13843, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1105, "admitted_surface": 36047, "staged_surface": 57498, "admitted_split": 9915, "staged_split": 41677, "admitted_merge": 13742, "staged_merge": 117921, "admitted_skel": 5832, "staged_skel": 45048, "mean_count": 253.79586791992188, "n_util_obs": 162, "util_count_spearman": 0.20082815734989648}, "probe_overhead_sec_total": 620.371, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.01314, "alpha": 0.03, "t": 4007.3}
{"event": "consolidate", "step": 3600, "retired": 1, "split_retired": 0, "merge_admitted": 0, "admitted": 1, "swapped": 17, "uncertain_staged": 262111, "t": 4049.8}
{"event": "probe", "step": 3600, "phase": 2, "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": 0.204627, "boundary_entropy": 0.090634, "bpb": 1.945954}, "tr": {"chunks": 3808.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.210352, "cps": 29.75, "mean_p": 0.192604, "boundary_entropy": 0.125657, "bpb": 2.113745}, "fi": {"chunks": 4945.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.262878, "cps": 38.632812, "mean_p": 0.2548, "boundary_entropy": 0.098725, "bpb": 2.231045}}, "bpb": {"en": 1.94595, "tr": 2.11374, "fi": 2.23105}, "gini_cpc": 0.058465, "gini_cps": 0.076074, "bpic": 4.614233, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0335272973557765, "fi": 1.291605768108878}, "inventory": {"size": 327662, "admitted": 65536, "staged": 262126, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1145, "admitted_surface": 36034, "staged_surface": 57378, "admitted_split": 9916, "staged_split": 41951, "admitted_merge": 13748, "staged_merge": 117725, "admitted_skel": 5838, "staged_skel": 45072, "mean_count": 264.94744873046875, "n_util_obs": 168, "util_count_spearman": 0.15446884409681882}, "t": 4051.7}
{"event": "eval", "step": 3600, "phase": 2, "bpb": {"en": 1.9598, "tr": 2.12255, "fi": 2.23975}, "bytes_per_chunk": {"en": 4.633, "tr": 5.246, "fi": 4.063}, "bpb_mean": 2.10737, "frozen_inventory": true, "t": 4054.6}
{"event": "train", "step": 3600, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.44309, "bpb": 2.08193, "chunk_loss": 0.030602, "F": 0.21619, "G": 0.21588, "chunks": 221.38, "bytes_per_chunk": 4.626, "boundary_entropy": 0.0913, "router_noise": 0.0, "grad_norm": 0.6702, "lr": 0.0001319, "router_trainable": true, "sec_per_step": 1.054, "probe_sec": 0.1492, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.11386, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.00266, "n_shadow": 3, "mean_delta_shadow": -0.03322, "n_split": 2, "mean_delta_split": 0.17999, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -1.83001, "trunc_abs_frac": 0.22523, "mean_delta_full": -0.02759, "read_hit_rate": {"surface": 0.48383, "split": 0.52074, "merge": 0.12251, "skel": 0.40316}, "mean_gate": {"surface": 0.17685, "split": 0.06585, "merge": 0.11651, "skel": 0.11288}, "n_reads": 12616, "n_cand": 31253, "read_suppressed_frac": 0.1021, "dropout_reads_logged": 12616, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1145, "admitted_surface": 36034, "staged_surface": 57441, "admitted_split": 9916, "staged_split": 41949, "admitted_merge": 13748, "staged_merge": 117741, "admitted_skel": 5838, "staged_skel": 45013, "mean_count": 265.13995361328125, "n_util_obs": 168, "util_count_spearman": 0.15396510444875253}, "probe_overhead_sec_total": 626.349, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.012445, "alpha": 0.03, "t": 4060.0}
{"event": "consolidate", "step": 3650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 55, "uncertain_staged": 262112, "t": 4118.1}
{"event": "train", "step": 3650, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.46392, "bpb": 2.11199, "chunk_loss": 0.030763, "F": 0.22076, "G": 0.2212, "chunks": 226.06, "bytes_per_chunk": 4.53, "boundary_entropy": 0.09234, "router_noise": 0.0, "grad_norm": 0.6972, "lr": 0.0001283, "router_trainable": true, "sec_per_step": 1.184, "probe_sec": 0.1173, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05855, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 3, "mean_delta_loo": -0.0174, "n_shadow": 5, "mean_delta_shadow": -0.00109, "n_split": 2, "mean_delta_split": 0.23026, "n_merge": 1, "mean_delta_merge": -0.25195, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 5.04887, "trunc_abs_frac": 4.31333, "mean_delta_full": -0.01717, "read_hit_rate": {"surface": 0.49806, "split": 0.52466, "merge": 0.13445, "skel": 0.47558}, "mean_gate": {"surface": 0.19468, "split": 0.0752, "merge": 0.12916, "skel": 0.12748}, "n_reads": 13623, "n_cand": 32167, "read_suppressed_frac": 0.1036, "dropout_reads_logged": 13623, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1168, "admitted_surface": 35989, "staged_surface": 57365, "admitted_split": 9926, "staged_split": 42031, "admitted_merge": 13755, "staged_merge": 117647, "admitted_skel": 5866, "staged_skel": 45101, "mean_count": 276.24456787109375, "n_util_obs": 171, "util_count_spearman": 0.12977063969728958}, "probe_overhead_sec_total": 632.34, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.011792, "alpha": 0.03, "t": 4119.2}
{"event": "consolidate", "step": 3700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 103, "uncertain_staged": 262112, "t": 4161.3}
{"event": "train", "step": 3700, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.36266, "bpb": 1.9659, "chunk_loss": 0.03055, "F": 0.21121, "G": 0.21157, "chunks": 216.28, "bytes_per_chunk": 4.735, "boundary_entropy": 0.09086, "router_noise": 0.0, "grad_norm": 0.6737, "lr": 0.0001247, "router_trainable": true, "sec_per_step": 0.851, "probe_sec": 0.088, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06628, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": 0.14331, "n_shadow": 5, "mean_delta_shadow": 0.00234, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -0.39428, "trunc_abs_frac": 0.47789, "mean_delta_full": -0.17334, "read_hit_rate": {"surface": 0.47423, "split": 0.5105, "merge": 0.13257, "skel": 0.46745}, "mean_gate": {"surface": 0.18239, "split": 0.06934, "merge": 0.12349, "skel": 0.11643}, "n_reads": 12688, "n_cand": 30859, "read_suppressed_frac": 0.0962, "dropout_reads_logged": 12688, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1185, "admitted_surface": 35892, "staged_surface": 57445, "admitted_split": 9926, "staged_split": 41728, "admitted_merge": 13761, "staged_merge": 117820, "admitted_skel": 5957, "staged_skel": 45151, "mean_count": 286.59307861328125, "n_util_obs": 177, "util_count_spearman": 0.14345535915236923}, "probe_overhead_sec_total": 637.928, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.011782, "alpha": 0.03, "t": 4161.8}
{"event": "consolidate", "step": 3750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 55, "uncertain_staged": 262112, "t": 4210.9}
{"event": "train", "step": 3750, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.3882, "bpb": 2.00275, "chunk_loss": 0.030852, "F": 0.22363, "G": 0.22491, "chunks": 229.0, "bytes_per_chunk": 4.472, "boundary_entropy": 0.08381, "router_noise": 0.0, "grad_norm": 0.6531, "lr": 0.0001212, "router_trainable": true, "sec_per_step": 0.998, "probe_sec": 0.1319, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07425, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": -0.00856, "n_shadow": 6, "mean_delta_shadow": -0.11169, "n_split": 1, "mean_delta_split": 0.0423, "n_merge": 1, "mean_delta_merge": -0.18188, "n_dropped_unpaired": 0, "n_full": 0, "read_hit_rate": {"surface": 0.51192, "split": 0.50703, "merge": 0.15419, "skel": 0.51514}, "mean_gate": {"surface": 0.1651, "split": 0.06138, "merge": 0.11389, "skel": 0.10499}, "n_reads": 14009, "n_cand": 32527, "read_suppressed_frac": 0.1, "dropout_reads_logged": 14009, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1205, "admitted_surface": 35843, "staged_surface": 57714, "admitted_split": 9932, "staged_split": 41998, "admitted_merge": 13764, "staged_merge": 117162, "admitted_skel": 5997, "staged_skel": 45270, "mean_count": 296.82861328125, "n_util_obs": 186, "util_count_spearman": 0.17559781620502682}, "probe_overhead_sec_total": 643.178, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.011318, "alpha": 0.03, "t": 4211.7}
{"event": "consolidate", "step": 3800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 321, "uncertain_staged": 262112, "t": 4270.6}
{"event": "probe", "step": 3800, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3568.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.205022, "cps": 27.875, "mean_p": 0.20591, "boundary_entropy": 0.084119, "bpb": 1.932491}, "tr": {"chunks": 3677.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.203116, "cps": 28.726562, "mean_p": 0.189762, "boundary_entropy": 0.121448, "bpb": 2.09135}, "fi": {"chunks": 4976.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.264526, "cps": 38.875, "mean_p": 0.257692, "boundary_entropy": 0.091647, "bpb": 2.206536}}, "bpb": {"en": 1.93249, "tr": 2.09135, "fi": 2.20654}, "gini_cpc": 0.060863, "gini_cps": 0.076808, "bpic": 4.642173, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9907004332953757, "fi": 1.2902318627136635}, "inventory": {"size": 327359, "admitted": 65536, "staged": 261823, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1241, "admitted_surface": 35559, "staged_surface": 57443, "admitted_split": 9958, "staged_split": 41545, "admitted_merge": 13770, "staged_merge": 118410, "admitted_skel": 6249, "staged_skel": 44425, "mean_count": 307.1689453125, "n_util_obs": 188, "util_count_spearman": 0.16988407226204472}, "t": 4272.4}
{"event": "eval", "step": 3800, "phase": 2, "bpb": {"en": 1.9458, "tr": 2.10112, "fi": 2.21208}, "bytes_per_chunk": {"en": 4.603, "tr": 5.387, "fi": 4.051}, "bpb_mean": 2.08634, "frozen_inventory": true, "t": 4276.2}
{"event": "train", "step": 3800, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.36514, "bpb": 1.96948, "chunk_loss": 0.030607, "F": 0.2124, "G": 0.21239, "chunks": 217.5, "bytes_per_chunk": 4.708, "boundary_entropy": 0.08715, "router_noise": 0.0, "grad_norm": 0.6598, "lr": 0.0001177, "router_trainable": true, "sec_per_step": 1.393, "probe_sec": 0.1147, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05717, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.08353, "n_shadow": 8, "mean_delta_shadow": -0.00631, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.79841, "trunc_abs_frac": 0.79841, "mean_delta_full": -0.36877, "read_hit_rate": {"surface": 0.50462, "split": 0.49599, "merge": 0.14327, "skel": 0.523}, "mean_gate": {"surface": 0.17515, "split": 0.06723, "merge": 0.11552, "skel": 0.11323}, "n_reads": 13163, "n_cand": 31037, "read_suppressed_frac": 0.1003, "dropout_reads_logged": 13163, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1241, "admitted_surface": 35559, "staged_surface": 57501, "admitted_split": 9958, "staged_split": 41632, "admitted_merge": 13770, "staged_merge": 118517, "admitted_skel": 6249, "staged_skel": 44494, "mean_count": 307.36981201171875, "n_util_obs": 188, "util_count_spearman": 0.16987143019683695}, "probe_overhead_sec_total": 649.631, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.011177, "alpha": 0.03, "t": 4281.4}
{"event": "consolidate", "step": 3850, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 4, "uncertain_staged": 262112, "t": 4338.3}
{"event": "train", "step": 3850, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.35715, "bpb": 1.95796, "chunk_loss": 0.030421, "F": 0.20441, "G": 0.20572, "chunks": 209.31, "bytes_per_chunk": 4.892, "boundary_entropy": 0.09168, "router_noise": 0.0, "grad_norm": 0.7269, "lr": 0.0001142, "router_trainable": true, "sec_per_step": 1.156, "probe_sec": 0.096, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05922, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": 0.01975, "n_shadow": 2, "mean_delta_shadow": 0.00189, "n_split": 2, "mean_delta_split": -0.05505, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.34689, "trunc_abs_frac": 0.34689, "mean_delta_full": 0.27905, "read_hit_rate": {"surface": 0.473, "split": 0.51019, "merge": 0.12089, "skel": 0.50331}, "mean_gate": {"surface": 0.17447, "split": 0.06667, "merge": 0.11103, "skel": 0.11245}, "n_reads": 12447, "n_cand": 30014, "read_suppressed_frac": 0.0979, "dropout_reads_logged": 12447, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1278, "admitted_surface": 35558, "staged_surface": 57682, "admitted_split": 9958, "staged_split": 42166, "admitted_merge": 13770, "staged_merge": 117630, "admitted_skel": 6250, "staged_skel": 44666, "mean_count": 317.62646484375, "n_util_obs": 192, "util_count_spearman": 0.18707782871714185}, "probe_overhead_sec_total": 655.837, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.010709, "alpha": 0.03, "t": 4339.2}
{"event": "consolidate", "step": 3900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 81, "uncertain_staged": 262112, "t": 4393.8}
{"event": "train", "step": 3900, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.39121, "bpb": 2.00709, "chunk_loss": 0.03059, "F": 0.21271, "G": 0.21254, "chunks": 217.81, "bytes_per_chunk": 4.701, "boundary_entropy": 0.09027, "router_noise": 0.0, "grad_norm": 0.7028, "lr": 0.0001108, "router_trainable": true, "sec_per_step": 1.117, "probe_sec": 0.0979, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06288, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 11, "mean_delta_loo": 0.04561, "n_shadow": 3, "mean_delta_shadow": 0.06619, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.18449, "trunc_abs_frac": 0.20748, "mean_delta_full": 0.5032, "read_hit_rate": {"surface": 0.49207, "split": 0.51364, "merge": 0.14162, "skel": 0.52875}, "mean_gate": {"surface": 0.18114, "split": 0.06961, "merge": 0.12039, "skel": 0.11474}, "n_reads": 13296, "n_cand": 31025, "read_suppressed_frac": 0.0955, "dropout_reads_logged": 13296, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1302, "admitted_surface": 35595, "staged_surface": 57414, "admitted_split": 9987, "staged_split": 42527, "admitted_merge": 13773, "staged_merge": 117433, "admitted_skel": 6181, "staged_skel": 44770, "mean_count": 328.52276611328125, "n_util_obs": 196, "util_count_spearman": 0.17336180860473827}, "probe_overhead_sec_total": 661.594, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009957, "alpha": 0.03, "t": 4395.0}
{"event": "consolidate", "step": 3950, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 64, "uncertain_staged": 262113, "t": 4453.8}
{"event": "train", "step": 3950, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.41618, "bpb": 2.04312, "chunk_loss": 0.030719, "F": 0.21964, "G": 0.21991, "chunks": 224.91, "bytes_per_chunk": 4.553, "boundary_entropy": 0.08397, "router_noise": 0.0, "grad_norm": 0.6762, "lr": 0.0001074, "router_trainable": true, "sec_per_step": 1.214, "probe_sec": 0.101, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05242, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 3, "mean_delta_loo": 0.04184, "n_shadow": 7, "mean_delta_shadow": -0.04242, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": -0.0918, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.05082, "trunc_abs_frac": 0.05082, "mean_delta_full": 0.17114, "read_hit_rate": {"surface": 0.49351, "split": 0.51886, "merge": 0.14538, "skel": 0.53208}, "mean_gate": {"surface": 0.1635, "split": 0.06153, "merge": 0.10802, "skel": 0.10344}, "n_reads": 13646, "n_cand": 31653, "read_suppressed_frac": 0.0985, "dropout_reads_logged": 13646, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1327, "admitted_surface": 35630, "staged_surface": 57465, "admitted_split": 9983, "staged_split": 42145, "admitted_merge": 13777, "staged_merge": 117742, "admitted_skel": 6146, "staged_skel": 44792, "mean_count": 338.57940673828125, "n_util_obs": 198, "util_count_spearman": 0.16562198780396367}, "probe_overhead_sec_total": 667.856, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.010574, "alpha": 0.03, "t": 4455.8}
{"event": "consolidate", "step": 4000, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262113, "t": 4510.1}
{"event": "probe", "step": 4000, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3572.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.205252, "cps": 27.90625, "mean_p": 0.205227, "boundary_entropy": 0.083347, "bpb": 1.916353}, "tr": {"chunks": 3538.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.195437, "cps": 27.640625, "mean_p": 0.189675, "boundary_entropy": 0.121828, "bpb": 2.076347}, "fi": {"chunks": 4974.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.26442, "cps": 38.859375, "mean_p": 0.257515, "boundary_entropy": 0.09094, "bpb": 2.187329}}, "bpb": {"en": 1.91635, "tr": 2.07635, "fi": 2.18733}, "gini_cpc": 0.069144, "gini_cps": 0.079223, "bpic": 4.694803, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9521819556984454, "fi": 1.2882690329804527}, "inventory": {"size": 326655, "admitted": 65536, "staged": 261119, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1353, "admitted_surface": 35660, "staged_surface": 57243, "admitted_split": 9447, "staged_split": 42032, "admitted_merge": 14369, "staged_merge": 117431, "admitted_skel": 6060, "staged_skel": 44413, "mean_count": 348.11322021484375, "n_util_obs": 204, "util_count_spearman": 0.18868196563739037}, "t": 4511.8}
{"event": "eval", "step": 4000, "phase": 2, "bpb": {"en": 1.93224, "tr": 2.08545, "fi": 2.19408}, "bytes_per_chunk": {"en": 4.615, "tr": 5.528, "fi": 4.049}, "bpb_mean": 2.07059, "frozen_inventory": true, "t": 4514.9}
{"event": "train", "step": 4000, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.43892, "bpb": 2.07593, "chunk_loss": 0.030697, "F": 0.22018, "G": 0.21969, "chunks": 225.47, "bytes_per_chunk": 4.542, "boundary_entropy": 0.08588, "router_noise": 0.0, "grad_norm": 0.7251, "lr": 0.000104, "router_trainable": true, "sec_per_step": 1.294, "probe_sec": 0.1261, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06683, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.05569, "n_shadow": 1, "mean_delta_shadow": -0.06653, "n_split": 2, "mean_delta_split": 0.02036, "n_merge": 1, "mean_delta_merge": 0.01611, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.11392, "trunc_abs_frac": 0.11392, "mean_delta_full": -0.23413, "read_hit_rate": {"surface": 0.49436, "split": 0.51818, "merge": 0.15424, "skel": 0.52899}, "mean_gate": {"surface": 0.17256, "split": 0.064, "merge": 0.11428, "skel": 0.1086}, "n_reads": 13671, "n_cand": 31627, "read_suppressed_frac": 0.1002, "dropout_reads_logged": 13671, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1354, "admitted_surface": 35660, "staged_surface": 57460, "admitted_split": 9447, "staged_split": 42117, "admitted_merge": 14369, "staged_merge": 117989, "admitted_skel": 6060, "staged_skel": 44578, "mean_count": 348.32183837890625, "n_util_obs": 204, "util_count_spearman": 0.18849962188407743}, "probe_overhead_sec_total": 674.166, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009813, "alpha": 0.03, "t": 4520.5}
{"event": "consolidate", "step": 4050, "retired": 1, "split_retired": 2, "merge_admitted": 0, "admitted": 3, "swapped": 1024, "uncertain_staged": 262113, "t": 4569.4}
{"event": "train", "step": 4050, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.35027, "bpb": 1.94803, "chunk_loss": 0.03064, "F": 0.21527, "G": 0.21571, "chunks": 220.44, "bytes_per_chunk": 4.645, "boundary_entropy": 0.07968, "router_noise": 0.0, "grad_norm": 0.6261, "lr": 0.0001007, "router_trainable": true, "sec_per_step": 1.004, "probe_sec": 0.1482, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09716, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.01261, "n_shadow": 3, "mean_delta_shadow": -0.03439, "n_split": 1, "mean_delta_split": 0.05359, "n_merge": 1, "mean_delta_merge": -0.09293, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.98554, "trunc_abs_frac": 0.99561, "mean_delta_full": -0.08441, "read_hit_rate": {"surface": 0.50385, "split": 0.5122, "merge": 0.14406, "skel": 0.54462}, "mean_gate": {"surface": 0.16166, "split": 0.06164, "merge": 0.10885, "skel": 0.1034}, "n_reads": 13598, "n_cand": 31357, "read_suppressed_frac": 0.0982, "dropout_reads_logged": 13598, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1362, "admitted_surface": 35490, "staged_surface": 57432, "admitted_split": 9297, "staged_split": 42064, "admitted_merge": 14631, "staged_merge": 117936, "admitted_skel": 6118, "staged_skel": 44712, "mean_count": 357.25634765625, "n_util_obs": 210, "util_count_spearman": 0.1970994473795905}, "probe_overhead_sec_total": 680.597, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009974, "alpha": 0.03, "t": 4570.7}
{"event": "consolidate", "step": 4100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 346, "uncertain_staged": 262113, "t": 4631.8}
{"event": "train", "step": 4100, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.43415, "bpb": 2.06904, "chunk_loss": 0.030812, "F": 0.22299, "G": 0.22296, "chunks": 228.34, "bytes_per_chunk": 4.484, "boundary_entropy": 0.08481, "router_noise": 0.0, "grad_norm": 0.7153, "lr": 9.75e-05, "router_trainable": true, "sec_per_step": 1.242, "probe_sec": 0.498, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05248, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.00521, "n_shadow": 6, "mean_delta_shadow": 0.03191, "n_split": 1, "mean_delta_split": 0.1994, "n_merge": 1, "mean_delta_merge": -0.04718, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.05525, "trunc_abs_frac": 1.40747, "mean_delta_full": -0.09246, "read_hit_rate": {"surface": 0.51519, "split": 0.58667, "merge": 0.14372, "skel": 0.56367}, "mean_gate": {"surface": 0.16912, "split": 0.06407, "merge": 0.12063, "skel": 0.1009}, "n_reads": 15221, "n_cand": 32514, "read_suppressed_frac": 0.1024, "dropout_reads_logged": 15221, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1375, "admitted_surface": 35203, "staged_surface": 57432, "admitted_split": 9304, "staged_split": 41933, "admitted_merge": 14840, "staged_merge": 118391, "admitted_skel": 6189, "staged_skel": 44388, "mean_count": 383.2366943359375, "n_util_obs": 212, "util_count_spearman": 0.17840081713566744}, "probe_overhead_sec_total": 686.941, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009235, "alpha": 0.03, "t": 4632.8}
{"event": "consolidate", "step": 4150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 86, "uncertain_staged": 262114, "t": 4679.1}
{"event": "train", "step": 4150, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.37836, "bpb": 1.98856, "chunk_loss": 0.030898, "F": 0.22617, "G": 0.2265, "chunks": 231.59, "bytes_per_chunk": 4.422, "boundary_entropy": 0.07835, "router_noise": 0.0, "grad_norm": 0.6532, "lr": 9.43e-05, "router_trainable": true, "sec_per_step": 0.938, "probe_sec": 0.0726, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04269, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": -0.01441, "n_shadow": 5, "mean_delta_shadow": -0.01185, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.61007, "trunc_abs_frac": 1.00114, "mean_delta_full": -0.05991, "read_hit_rate": {"surface": 0.52012, "split": 0.58741, "merge": 0.15938, "skel": 0.56852}, "mean_gate": {"surface": 0.16091, "split": 0.05989, "merge": 0.11267, "skel": 0.0993}, "n_reads": 15586, "n_cand": 32912, "read_suppressed_frac": 0.0969, "dropout_reads_logged": 15586, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1409, "admitted_surface": 35137, "staged_surface": 57634, "admitted_split": 9304, "staged_split": 41755, "admitted_merge": 14897, "staged_merge": 118193, "admitted_skel": 6198, "staged_skel": 44562, "mean_count": 394.41424560546875, "n_util_obs": 216, "util_count_spearman": 0.1709487404436722}, "probe_overhead_sec_total": 692.192, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009327, "alpha": 0.03, "t": 4679.7}
{"event": "consolidate", "step": 4200, "retired": 1, "split_retired": 0, "merge_admitted": 0, "admitted": 1, "swapped": 503, "uncertain_staged": 262114, "t": 4732.8}
{"event": "probe", "step": 4200, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3566.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.204907, "cps": 27.859375, "mean_p": 0.205226, "boundary_entropy": 0.077451, "bpb": 1.906393}, "tr": {"chunks": 3514.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.194111, "cps": 27.453125, "mean_p": 0.187281, "boundary_entropy": 0.116165, "bpb": 2.052654}, "fi": {"chunks": 4983.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.264898, "cps": 38.929688, "mean_p": 0.257214, "boundary_entropy": 0.088681, "bpb": 2.162487}}, "bpb": {"en": 1.90639, "tr": 2.05265, "fi": 2.16249}, "gini_cpc": 0.07108, "gini_cps": 0.081185, "bpic": 4.702976, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9473140686299675, "fi": 1.2927715472262937}, "inventory": {"size": 327176, "admitted": 65536, "staged": 261640, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1425, "admitted_surface": 34798, "staged_surface": 57540, "admitted_split": 9306, "staged_split": 41847, "admitted_merge": 15175, "staged_merge": 118184, "admitted_skel": 6257, "staged_skel": 44069, "mean_count": 408.796630859375, "n_util_obs": 222, "util_count_spearman": 0.1493289446336136}, "t": 4734.6}
{"event": "eval", "step": 4200, "phase": 2, "bpb": {"en": 1.9182, "tr": 2.0639, "fi": 2.16989}, "bytes_per_chunk": {"en": 4.611, "tr": 5.562, "fi": 4.048}, "bpb_mean": 2.05066, "frozen_inventory": true, "t": 4737.6}
{"event": "phase_transition", "step": 4200, "phase": 3, "objective": "global", "router_trainable": true, "encoder_trainable": true, "frozen_router_params": 0, "t": 4741.9}
{"event": "train", "step": 4200, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.36369, "bpb": 1.9674, "chunk_loss": 0.030527, "F": 0.21255, "G": 0.21286, "chunks": 217.66, "bytes_per_chunk": 4.705, "boundary_entropy": 0.07989, "router_noise": 0.0, "grad_norm": 0.6761, "lr": 9.12e-05, "router_trainable": true, "sec_per_step": 1.283, "probe_sec": 0.1578, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.094, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.07779, "n_shadow": 5, "mean_delta_shadow": -0.03381, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.14087, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.11009, "trunc_abs_frac": 0.17403, "mean_delta_full": -0.33337, "read_hit_rate": {"surface": 0.49993, "split": 0.59086, "merge": 0.13882, "skel": 0.5382}, "mean_gate": {"surface": 0.16011, "split": 0.06166, "merge": 0.1087, "skel": 0.10185}, "n_reads": 14324, "n_cand": 31057, "read_suppressed_frac": 0.0984, "dropout_reads_logged": 14324, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1426, "admitted_surface": 34798, "staged_surface": 57603, "admitted_split": 9306, "staged_split": 41920, "admitted_merge": 15175, "staged_merge": 118398, "admitted_skel": 6257, "staged_skel": 44223, "mean_count": 409.01519775390625, "n_util_obs": 222, "util_count_spearman": 0.14859846962194628}, "probe_overhead_sec_total": 697.972, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009669, "alpha": 0.03, "t": 4743.8}
{"event": "consolidate", "step": 4250, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 304, "uncertain_staged": 262115, "t": 4806.2}
{"event": "train", "step": 4250, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.36556, "bpb": 1.97009, "chunk_loss": 0.030512, "F": 0.20779, "G": 0.2073, "chunks": 212.78, "bytes_per_chunk": 4.812, "boundary_entropy": 0.08819, "router_noise": 0.0, "grad_norm": 0.6757, "lr": 8.81e-05, "router_trainable": true, "sec_per_step": 1.282, "probe_sec": 0.1189, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05636, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": 0.17509, "n_shadow": 1, "mean_delta_shadow": -0.11182, "n_split": 1, "mean_delta_split": 0.2428, "n_merge": 1, "mean_delta_merge": -0.12175, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.82802, "trunc_abs_frac": 1.19712, "mean_delta_full": 0.06317, "read_hit_rate": {"surface": 0.4955, "split": 0.58525, "merge": 0.1321, "skel": 0.53424}, "mean_gate": {"surface": 0.17882, "split": 0.06704, "merge": 0.1197, "skel": 0.11368}, "n_reads": 13948, "n_cand": 30550, "read_suppressed_frac": 0.0992, "dropout_reads_logged": 13948, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1452, "admitted_surface": 34598, "staged_surface": 57700, "admitted_split": 9307, "staged_split": 41982, "admitted_merge": 15355, "staged_merge": 117984, "admitted_skel": 6276, "staged_skel": 44478, "mean_count": 420.8258972167969, "n_util_obs": 226, "util_count_spearman": 0.13874182942835236}, "probe_overhead_sec_total": 704.451, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008644, "alpha": 0.03, "t": 4807.9}
{"event": "consolidate", "step": 4300, "retired": 2, "split_retired": 2, "merge_admitted": 0, "admitted": 4, "swapped": 605, "uncertain_staged": 262116, "t": 4860.0}
{"event": "train", "step": 4300, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.45382, "bpb": 2.09742, "chunk_loss": 0.030655, "F": 0.21439, "G": 0.21562, "chunks": 219.53, "bytes_per_chunk": 4.664, "boundary_entropy": 0.08939, "router_noise": 0.0, "grad_norm": 0.7581, "lr": 8.5e-05, "router_trainable": true, "sec_per_step": 1.056, "probe_sec": 0.0919, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06062, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 1, "mean_delta_loo": 0.07617, "n_shadow": 7, "mean_delta_shadow": 0.01843, "n_split": 1, "mean_delta_split": 0.03308, "n_merge": 2, "mean_delta_merge": -0.08466, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.11408, "trunc_abs_frac": 0.64697, "mean_delta_full": -0.09818, "read_hit_rate": {"surface": 0.49421, "split": 0.59395, "merge": 0.13805, "skel": 0.53117}, "mean_gate": {"surface": 0.18345, "split": 0.06848, "merge": 0.12326, "skel": 0.1131}, "n_reads": 14326, "n_cand": 31195, "read_suppressed_frac": 0.1019, "dropout_reads_logged": 14326, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1471, "admitted_surface": 34180, "staged_surface": 57494, "admitted_split": 9274, "staged_split": 42220, "admitted_merge": 15706, "staged_merge": 117809, "admitted_skel": 6376, "staged_skel": 44621, "mean_count": 430.43756103515625, "n_util_obs": 226, "util_count_spearman": 0.13940408805481427}, "probe_overhead_sec_total": 710.489, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008199, "alpha": 0.03, "t": 4860.7}
{"event": "consolidate", "step": 4350, "retired": 1, "split_retired": 0, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262116, "t": 4919.2}
{"event": "train", "step": 4350, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.31329, "bpb": 1.89468, "chunk_loss": 0.030633, "F": 0.21503, "G": 0.21436, "chunks": 220.19, "bytes_per_chunk": 4.651, "boundary_entropy": 0.08132, "router_noise": 0.0, "grad_norm": 0.7167, "lr": 8.21e-05, "router_trainable": true, "sec_per_step": 1.18, "probe_sec": 0.1003, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06796, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.01024, "n_shadow": 6, "mean_delta_shadow": -0.04319, "n_split": 1, "mean_delta_split": 0.05562, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.42309, "trunc_abs_frac": 1.09112, "mean_delta_full": -0.11646, "read_hit_rate": {"surface": 0.50656, "split": 0.57151, "merge": 0.14623, "skel": 0.55329}, "mean_gate": {"surface": 0.16114, "split": 0.0599, "merge": 0.1062, "skel": 0.10409}, "n_reads": 14462, "n_cand": 31468, "read_suppressed_frac": 0.0991, "dropout_reads_logged": 14462, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1472, "admitted_surface": 34575, "staged_surface": 57306, "admitted_split": 8951, "staged_split": 42341, "admitted_merge": 15663, "staged_merge": 117853, "admitted_skel": 6347, "staged_skel": 44644, "mean_count": 440.20367431640625, "n_util_obs": 231, "util_count_spearman": 0.13129539126540626}, "probe_overhead_sec_total": 717.645, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009503, "alpha": 0.03, "t": 4919.7}
{"event": "consolidate", "step": 4400, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 1024, "uncertain_staged": 262116, "t": 4980.3}
{"event": "probe", "step": 4400, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3564.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.204792, "cps": 27.84375, "mean_p": 0.204656, "boundary_entropy": 0.077415, "bpb": 1.89181}, "tr": {"chunks": 3582.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.197868, "cps": 27.984375, "mean_p": 0.188532, "boundary_entropy": 0.117043, "bpb": 2.038858}, "fi": {"chunks": 4985.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.265005, "cps": 38.945312, "mean_p": 0.256482, "boundary_entropy": 0.085663, "bpb": 2.146717}}, "bpb": {"en": 1.89181, "tr": 2.03886, "fi": 2.14672}, "gini_cpc": 0.067036, "gini_cps": 0.078092, "bpic": 4.676614, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9661875898687478, "fi": 1.2940161722124404}, "inventory": {"size": 326654, "admitted": 65536, "staged": 261118, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1503, "admitted_surface": 34159, "staged_surface": 57294, "admitted_split": 8797, "staged_split": 41949, "admitted_merge": 16201, "staged_merge": 117311, "admitted_skel": 6379, "staged_skel": 44564, "mean_count": 450.41180419921875, "n_util_obs": 238, "util_count_spearman": 0.15200914529460324}, "t": 4982.0}
{"event": "eval", "step": 4400, "phase": 3, "bpb": {"en": 1.90551, "tr": 2.04835, "fi": 2.15272}, "bytes_per_chunk": {"en": 4.62, "tr": 5.475, "fi": 4.044}, "bpb_mean": 2.03553, "frozen_inventory": true, "t": 4984.9}
{"event": "train", "step": 4400, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.42173, "bpb": 2.05112, "chunk_loss": 0.030879, "F": 0.22156, "G": 0.22169, "chunks": 226.88, "bytes_per_chunk": 4.513, "boundary_entropy": 0.08083, "router_noise": 0.0, "grad_norm": 0.7047, "lr": 7.92e-05, "router_trainable": true, "sec_per_step": 1.42, "probe_sec": 0.1074, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06264, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.004, "n_shadow": 6, "mean_delta_shadow": -0.02548, "n_split": 1, "mean_delta_split": -0.01553, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": -2.95597, "trunc_abs_frac": 2.95597, "mean_delta_full": 0.01941, "read_hit_rate": {"surface": 0.51079, "split": 0.58118, "merge": 0.15106, "skel": 0.55232}, "mean_gate": {"surface": 0.16874, "split": 0.06357, "merge": 0.12185, "skel": 0.10161}, "n_reads": 14892, "n_cand": 32091, "read_suppressed_frac": 0.1003, "dropout_reads_logged": 14892, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1503, "admitted_surface": 34159, "staged_surface": 57480, "admitted_split": 8797, "staged_split": 41961, "admitted_merge": 16201, "staged_merge": 118026, "admitted_skel": 6379, "staged_skel": 44677, "mean_count": 450.6390380859375, "n_util_obs": 238, "util_count_spearman": 0.15181817655826696}, "probe_overhead_sec_total": 723.684, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009129, "alpha": 0.03, "t": 4990.7}
{"event": "consolidate", "step": 4450, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262116, "t": 5052.0}
{"event": "train", "step": 4450, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35671, "bpb": 1.95732, "chunk_loss": 0.030739, "F": 0.21973, "G": 0.21846, "chunks": 225.0, "bytes_per_chunk": 4.551, "boundary_entropy": 0.07887, "router_noise": 0.0, "grad_norm": 0.6617, "lr": 7.63e-05, "router_trainable": true, "sec_per_step": 1.239, "probe_sec": 0.1242, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07171, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": -0.03579, "n_shadow": 2, "mean_delta_shadow": 0.08511, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": -0.96344, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.86623, "trunc_abs_frac": 0.62027, "mean_delta_full": -0.99402, "read_hit_rate": {"surface": 0.52107, "split": 0.58626, "merge": 0.15359, "skel": 0.55612}, "mean_gate": {"surface": 0.16661, "split": 0.06356, "merge": 0.11382, "skel": 0.10819}, "n_reads": 14917, "n_cand": 31780, "read_suppressed_frac": 0.0986, "dropout_reads_logged": 14917, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1531, "admitted_surface": 33548, "staged_surface": 57502, "admitted_split": 8798, "staged_split": 41656, "admitted_merge": 16746, "staged_merge": 118067, "admitted_skel": 6444, "staged_skel": 44919, "mean_count": 461.45562744140625, "n_util_obs": 241, "util_count_spearman": 0.13312214944617812}, "probe_overhead_sec_total": 729.709, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009488, "alpha": 0.03, "t": 5052.7}
{"event": "consolidate", "step": 4500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262116, "t": 5117.2}
{"event": "train", "step": 4500, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.37127, "bpb": 1.97832, "chunk_loss": 0.030764, "F": 0.22018, "G": 0.21975, "chunks": 225.47, "bytes_per_chunk": 4.542, "boundary_entropy": 0.08254, "router_noise": 0.0, "grad_norm": 0.7514, "lr": 7.36e-05, "router_trainable": true, "sec_per_step": 1.312, "probe_sec": 0.1104, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08742, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.09611, "n_shadow": 4, "mean_delta_shadow": 0.00407, "n_split": 1, "mean_delta_split": -0.24695, "n_merge": 1, "mean_delta_merge": 0.55058, "n_dropped_unpaired": 0, "n_full": 0, "read_hit_rate": {"surface": 0.51636, "split": 0.57826, "merge": 0.1393, "skel": 0.55912}, "mean_gate": {"surface": 0.16928, "split": 0.06316, "merge": 0.11709, "skel": 0.10664}, "n_reads": 15011, "n_cand": 32339, "read_suppressed_frac": 0.0966, "dropout_reads_logged": 15011, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1543, "admitted_surface": 34091, "staged_surface": 57186, "admitted_split": 8454, "staged_split": 41732, "admitted_merge": 16499, "staged_merge": 118445, "admitted_skel": 6492, "staged_skel": 44781, "mean_count": 472.1550598144531, "n_util_obs": 249, "util_count_spearman": 0.14420134943672044}, "probe_overhead_sec_total": 735.599, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008618, "alpha": 0.03, "t": 5118.2}
{"event": "consolidate", "step": 4550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262115, "t": 5176.6}
{"event": "train", "step": 4550, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35488, "bpb": 1.95469, "chunk_loss": 0.030692, "F": 0.216, "G": 0.21722, "chunks": 221.19, "bytes_per_chunk": 4.63, "boundary_entropy": 0.08206, "router_noise": 0.0, "grad_norm": 0.6592, "lr": 7.09e-05, "router_trainable": true, "sec_per_step": 1.192, "probe_sec": 0.1778, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07694, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.0326, "n_shadow": 4, "mean_delta_shadow": 0.01669, "n_split": 1, "mean_delta_split": -0.11957, "n_merge": 1, "mean_delta_merge": 0.71387, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.46164, "trunc_abs_frac": 0.68836, "mean_delta_full": 0.29831, "read_hit_rate": {"surface": 0.50341, "split": 0.58072, "merge": 0.15141, "skel": 0.5539}, "mean_gate": {"surface": 0.16276, "split": 0.05982, "merge": 0.11321, "skel": 0.09999}, "n_reads": 14603, "n_cand": 31517, "read_suppressed_frac": 0.0983, "dropout_reads_logged": 14603, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1553, "admitted_surface": 34687, "staged_surface": 57178, "admitted_split": 8154, "staged_split": 42460, "admitted_merge": 16148, "staged_merge": 117704, "admitted_skel": 6547, "staged_skel": 44802, "mean_count": 481.6873779296875, "n_util_obs": 254, "util_count_spearman": 0.1676676853520058}, "probe_overhead_sec_total": 741.09, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.007865, "alpha": 0.03, "t": 5177.8}
{"event": "consolidate", "step": 4600, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262115, "t": 5232.6}
{"event": "probe", "step": 4600, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3569.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.20508, "cps": 27.882812, "mean_p": 0.204935, "boundary_entropy": 0.074269, "bpb": 1.881232}, "tr": {"chunks": 3725.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.205767, "cps": 29.101562, "mean_p": 0.190363, "boundary_entropy": 0.113849, "bpb": 2.020991}, "fi": {"chunks": 4996.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.265589, "cps": 39.03125, "mean_p": 0.257116, "boundary_entropy": 0.084017, "bpb": 2.133805}}, "bpb": {"en": 1.88123, "tr": 2.02099, "fi": 2.13381}, "gini_cpc": 0.059636, "gini_cps": 0.077407, "bpic": 4.616111, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0033519473350148, "fi": 1.2950547183312124}, "inventory": {"size": 326655, "admitted": 65536, "staged": 261119, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1565, "admitted_surface": 35354, "staged_surface": 56713, "admitted_split": 7914, "staged_split": 41948, "admitted_merge": 15651, "staged_merge": 117949, "admitted_skel": 6617, "staged_skel": 44509, "mean_count": 490.0595703125, "n_util_obs": 255, "util_count_spearman": 0.17374583389044582}, "t": 5234.4}
{"event": "eval", "step": 4600, "phase": 3, "bpb": {"en": 1.89531, "tr": 2.03149, "fi": 2.14086}, "bytes_per_chunk": {"en": 4.619, "tr": 5.32, "fi": 4.027}, "bpb_mean": 2.02255, "frozen_inventory": true, "t": 5237.7}
{"event": "train", "step": 4600, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.40529, "bpb": 2.0274, "chunk_loss": 0.030607, "F": 0.21271, "G": 0.21188, "chunks": 217.81, "bytes_per_chunk": 4.701, "boundary_entropy": 0.08134, "router_noise": 0.0, "grad_norm": 0.7043, "lr": 6.82e-05, "router_trainable": true, "sec_per_step": 1.296, "probe_sec": 0.1129, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06955, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.09927, "n_shadow": 2, "mean_delta_shadow": -0.02356, "n_split": 1, "mean_delta_split": -0.49988, "n_merge": 1, "mean_delta_merge": -0.10428, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.08508, "trunc_abs_frac": 0.34242, "mean_delta_full": 0.46667, "read_hit_rate": {"surface": 0.48487, "split": 0.5822, "merge": 0.12786, "skel": 0.53563}, "mean_gate": {"surface": 0.17306, "split": 0.06443, "merge": 0.11118, "skel": 0.11323}, "n_reads": 14101, "n_cand": 31195, "read_suppressed_frac": 0.1006, "dropout_reads_logged": 14101, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1567, "admitted_surface": 35354, "staged_surface": 56967, "admitted_split": 7914, "staged_split": 42133, "admitted_merge": 15651, "staged_merge": 118374, "admitted_skel": 6617, "staged_skel": 44670, "mean_count": 490.2747497558594, "n_util_obs": 255, "util_count_spearman": 0.17202698008337194}, "probe_overhead_sec_total": 746.452, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009531, "alpha": 0.03, "t": 5242.7}
{"event": "consolidate", "step": 4650, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262115, "t": 5307.5}
{"event": "train", "step": 4650, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.33228, "bpb": 1.92208, "chunk_loss": 0.030544, "F": 0.20874, "G": 0.21147, "chunks": 213.75, "bytes_per_chunk": 4.791, "boundary_entropy": 0.08098, "router_noise": 0.0, "grad_norm": 0.6499, "lr": 6.57e-05, "router_trainable": true, "sec_per_step": 1.316, "probe_sec": 0.1359, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.11134, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.09849, "n_shadow": 5, "mean_delta_shadow": -0.01202, "n_split": 1, "mean_delta_split": -0.12357, "n_merge": 1, "mean_delta_merge": -0.06314, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 0.08121, "trunc_abs_frac": 0.43858, "mean_delta_full": -0.17151, "read_hit_rate": {"surface": 0.48619, "split": 0.56851, "merge": 0.13253, "skel": 0.53998}, "mean_gate": {"surface": 0.16195, "split": 0.06116, "merge": 0.11376, "skel": 0.10113}, "n_reads": 13732, "n_cand": 30600, "read_suppressed_frac": 0.0956, "dropout_reads_logged": 13732, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1574, "admitted_surface": 36041, "staged_surface": 57218, "admitted_split": 7552, "staged_split": 42637, "admitted_merge": 15336, "staged_merge": 117425, "admitted_skel": 6607, "staged_skel": 44864, "mean_count": 498.887451171875, "n_util_obs": 258, "util_count_spearman": 0.14974552871575766}, "probe_overhead_sec_total": 751.871, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008994, "alpha": 0.03, "t": 5308.4}
{"event": "consolidate", "step": 4700, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 1024, "uncertain_staged": 262115, "t": 5359.8}
{"event": "train", "step": 4700, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.34317, "bpb": 1.93779, "chunk_loss": 0.030601, "F": 0.21341, "G": 0.21207, "chunks": 218.53, "bytes_per_chunk": 4.686, "boundary_entropy": 0.07963, "router_noise": 0.0, "grad_norm": 0.6755, "lr": 6.32e-05, "router_trainable": true, "sec_per_step": 1.044, "probe_sec": 0.1324, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07606, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.01724, "n_shadow": 4, "mean_delta_shadow": 0.02094, "n_split": 1, "mean_delta_split": -0.04565, "n_merge": 1, "mean_delta_merge": 0.57202, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.03906, "trunc_abs_frac": 0.73128, "mean_delta_full": -0.32837, "read_hit_rate": {"surface": 0.49361, "split": 0.57949, "merge": 0.13523, "skel": 0.53479}, "mean_gate": {"surface": 0.17083, "split": 0.06292, "merge": 0.11281, "skel": 0.1084}, "n_reads": 14178, "n_cand": 31219, "read_suppressed_frac": 0.0983, "dropout_reads_logged": 14178, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1573, "admitted_surface": 36734, "staged_surface": 57142, "admitted_split": 7191, "staged_split": 42264, "admitted_merge": 15208, "staged_merge": 117901, "admitted_skel": 6403, "staged_skel": 44837, "mean_count": 505.0762939453125, "n_util_obs": 252, "util_count_spearman": 0.11430470744627703}, "probe_overhead_sec_total": 757.586, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008171, "alpha": 0.03, "t": 5360.6}
{"event": "consolidate", "step": 4750, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 1024, "uncertain_staged": 262115, "t": 5411.0}
{"event": "train", "step": 4750, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.30903, "bpb": 1.88853, "chunk_loss": 0.030669, "F": 0.21777, "G": 0.21597, "chunks": 223.0, "bytes_per_chunk": 4.592, "boundary_entropy": 0.07947, "router_noise": 0.0, "grad_norm": 0.6818, "lr": 6.08e-05, "router_trainable": true, "sec_per_step": 1.025, "probe_sec": 0.0728, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04699, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": -0.03169, "n_shadow": 6, "mean_delta_shadow": -0.0047, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": 0.17319, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -0.16641, "trunc_abs_frac": 1.51384, "mean_delta_full": 0.08868, "read_hit_rate": {"surface": 0.48029, "split": 0.56588, "merge": 0.14989, "skel": 0.54354}, "mean_gate": {"surface": 0.17248, "split": 0.06409, "merge": 0.10936, "skel": 0.10867}, "n_reads": 14261, "n_cand": 31686, "read_suppressed_frac": 0.104, "dropout_reads_logged": 14261, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1597, "admitted_surface": 36912, "staged_surface": 56902, "admitted_split": 7195, "staged_split": 42092, "admitted_merge": 15148, "staged_merge": 118378, "admitted_skel": 6281, "staged_skel": 44772, "mean_count": 514.22998046875, "n_util_obs": 255, "util_count_spearman": 0.0952480893932376}, "probe_overhead_sec_total": 763.411, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008663, "alpha": 0.03, "t": 5411.9}
{"event": "consolidate", "step": 4800, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 1024, "uncertain_staged": 262114, "t": 5444.6}
{"event": "probe", "step": 4800, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3568.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.205022, "cps": 27.875, "mean_p": 0.204814, "boundary_entropy": 0.072666, "bpb": 1.87143}, "tr": {"chunks": 3569.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.19715, "cps": 27.882812, "mean_p": 0.186925, "boundary_entropy": 0.112012, "bpb": 2.011485}, "fi": {"chunks": 4983.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.264898, "cps": 38.929688, "mean_p": 0.25696, "boundary_entropy": 0.081927, "bpb": 2.121942}}, "bpb": {"en": 1.87143, "tr": 2.01148, "fi": 2.12194}, "gini_cpc": 0.067708, "gini_cps": 0.077833, "bpic": 4.680858, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9616018075689954, "fi": 1.2920468994980276}, "inventory": {"size": 326654, "admitted": 65536, "staged": 261118, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1612, "admitted_surface": 37392, "staged_surface": 56744, "admitted_split": 7198, "staged_split": 42254, "admitted_merge": 14792, "staged_merge": 117425, "admitted_skel": 6154, "staged_skel": 44695, "mean_count": 522.8035888671875, "n_util_obs": 257, "util_count_spearman": 0.08315423943534522}, "t": 5446.1}
{"event": "eval", "step": 4800, "phase": 3, "bpb": {"en": 1.88512, "tr": 2.02121, "fi": 2.12821}, "bytes_per_chunk": {"en": 4.619, "tr": 5.536, "fi": 4.033}, "bpb_mean": 2.01151, "frozen_inventory": true, "t": 5448.7}
{"event": "train", "step": 4800, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.37135, "bpb": 1.97844, "chunk_loss": 0.030957, "F": 0.22742, "G": 0.22663, "chunks": 232.88, "bytes_per_chunk": 4.397, "boundary_entropy": 0.07804, "router_noise": 0.0, "grad_norm": 0.7107, "lr": 5.85e-05, "router_trainable": true, "sec_per_step": 0.838, "probe_sec": 0.0673, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04495, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 3, "mean_delta_loo": 0.03174, "n_shadow": 5, "mean_delta_shadow": -0.03577, "n_split": 1, "mean_delta_split": 0.07758, "n_merge": 2, "mean_delta_merge": 0.08878, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.59239, "trunc_abs_frac": 0.59239, "mean_delta_full": 0.24048, "read_hit_rate": {"surface": 0.49865, "split": 0.56753, "merge": 0.13711, "skel": 0.54871}, "mean_gate": {"surface": 0.16962, "split": 0.06323, "merge": 0.11238, "skel": 0.10621}, "n_reads": 15002, "n_cand": 33120, "read_suppressed_frac": 0.1009, "dropout_reads_logged": 15002, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1615, "admitted_surface": 37392, "staged_surface": 56960, "admitted_split": 7198, "staged_split": 42308, "admitted_merge": 14792, "staged_merge": 118115, "admitted_skel": 6154, "staged_skel": 44761, "mean_count": 523.032470703125, "n_util_obs": 257, "util_count_spearman": 0.08337339381051488}, "probe_overhead_sec_total": 767.336, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008199, "alpha": 0.03, "t": 5453.8}
{"event": "consolidate", "step": 4850, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262114, "t": 5492.2}
{"event": "train", "step": 4850, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32631, "bpb": 1.91346, "chunk_loss": 0.030827, "F": 0.22348, "G": 0.22203, "chunks": 228.84, "bytes_per_chunk": 4.475, "boundary_entropy": 0.07975, "router_noise": 0.0, "grad_norm": 0.6833, "lr": 5.62e-05, "router_trainable": true, "sec_per_step": 0.785, "probe_sec": 0.0644, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04233, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.03417, "n_shadow": 5, "mean_delta_shadow": 0.0334, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": -0.06581, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 1.39218, "trunc_abs_frac": 1.39218, "mean_delta_full": -0.11548, "read_hit_rate": {"surface": 0.48718, "split": 0.58336, "merge": 0.1536, "skel": 0.54635}, "mean_gate": {"surface": 0.16961, "split": 0.06193, "merge": 0.1125, "skel": 0.10564}, "n_reads": 14756, "n_cand": 32180, "read_suppressed_frac": 0.0962, "dropout_reads_logged": 14756, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1638, "admitted_surface": 37711, "staged_surface": 56965, "admitted_split": 7203, "staged_split": 42407, "admitted_merge": 14425, "staged_merge": 117870, "admitted_skel": 6197, "staged_skel": 44902, "mean_count": 533.1815185546875, "n_util_obs": 260, "util_count_spearman": 0.09030106271485581}, "probe_overhead_sec_total": 773.103, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008745, "alpha": 0.03, "t": 5493.0}
{"event": "consolidate", "step": 4900, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262114, "t": 5524.6}
{"event": "train", "step": 4900, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.33568, "bpb": 1.92698, "chunk_loss": 0.030551, "F": 0.21329, "G": 0.21171, "chunks": 218.41, "bytes_per_chunk": 4.689, "boundary_entropy": 0.07615, "router_noise": 0.0, "grad_norm": 0.6934, "lr": 5.41e-05, "router_trainable": true, "sec_per_step": 0.652, "probe_sec": 0.0683, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04576, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 3, "mean_delta_loo": 0.00903, "n_shadow": 6, "mean_delta_shadow": 0.05712, "n_split": 2, "mean_delta_split": -0.08321, "n_merge": 1, "mean_delta_merge": -0.00339, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 1.05558, "trunc_abs_frac": 5.44425, "mean_delta_full": 0.07056, "read_hit_rate": {"surface": 0.46112, "split": 0.57213, "merge": 0.12693, "skel": 0.51779}, "mean_gate": {"surface": 0.1604, "split": 0.06091, "merge": 0.10378, "skel": 0.1019}, "n_reads": 13701, "n_cand": 31185, "read_suppressed_frac": 0.1014, "dropout_reads_logged": 13701, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1637, "admitted_surface": 38172, "staged_surface": 56964, "admitted_split": 7064, "staged_split": 42446, "admitted_merge": 14062, "staged_merge": 117899, "admitted_skel": 6238, "staged_skel": 44835, "mean_count": 542.7496337890625, "n_util_obs": 267, "util_count_spearman": 0.07954442110187437}, "probe_overhead_sec_total": 777.16, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.007609, "alpha": 0.03, "t": 5525.6}
{"event": "consolidate", "step": 4950, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262113, "t": 5558.6}
{"event": "train", "step": 4950, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32203, "bpb": 1.90729, "chunk_loss": 0.030771, "F": 0.22263, "G": 0.22123, "chunks": 227.97, "bytes_per_chunk": 4.492, "boundary_entropy": 0.07389, "router_noise": 0.0, "grad_norm": 0.6825, "lr": 5.2e-05, "router_trainable": true, "sec_per_step": 0.667, "probe_sec": 0.0774, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04749, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.05481, "n_shadow": 6, "mean_delta_shadow": -0.14456, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.08252, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.47983, "trunc_abs_frac": 0.35313, "mean_delta_full": -0.64299, "read_hit_rate": {"surface": 0.4746, "split": 0.57189, "merge": 0.14645, "skel": 0.53082}, "mean_gate": {"surface": 0.1528, "split": 0.05653, "merge": 0.10583, "skel": 0.09584}, "n_reads": 14387, "n_cand": 32154, "read_suppressed_frac": 0.1001, "dropout_reads_logged": 14387, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1645, "admitted_surface": 38547, "staged_surface": 56851, "admitted_split": 6946, "staged_split": 42406, "admitted_merge": 13894, "staged_merge": 117912, "admitted_skel": 6149, "staged_skel": 44975, "mean_count": 551.8387451171875, "n_util_obs": 270, "util_count_spearman": 0.08057815150642213}, "probe_overhead_sec_total": 781.461, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008599, "alpha": 0.03, "t": 5559.0}
{"event": "consolidate", "step": 5000, "retired": 0, "split_retired": 3, "merge_admitted": 0, "admitted": 3, "swapped": 1024, "uncertain_staged": 262113, "t": 5601.0}
{"event": "probe", "step": 5000, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3559.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.204505, "cps": 27.804688, "mean_p": 0.203839, "boundary_entropy": 0.072794, "bpb": 1.861563}, "tr": {"chunks": 3660.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.202176, "cps": 28.59375, "mean_p": 0.187639, "boundary_entropy": 0.110369, "bpb": 1.999922}, "fi": {"chunks": 4953.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.263303, "cps": 38.695312, "mean_p": 0.2559, "boundary_entropy": 0.082117, "bpb": 2.108062}}, "bpb": {"en": 1.86156, "tr": 1.99992, "fi": 2.10806}, "gini_cpc": 0.060824, "gini_cps": 0.07635, "bpic": 4.660861, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9886137947761907, "fi": 1.2875158280602261}, "inventory": {"size": 326653, "admitted": 65536, "staged": 261117, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1656, "admitted_surface": 39299, "staged_surface": 56397, "admitted_split": 6725, "staged_split": 42053, "admitted_merge": 13350, "staged_merge": 117705, "admitted_skel": 6162, "staged_skel": 44962, "mean_count": 559.3359985351562, "n_util_obs": 271, "util_count_spearman": 0.07849046137520199}, "t": 5602.8}
{"event": "eval", "step": 5000, "phase": 3, "bpb": {"en": 1.87623, "tr": 2.00904, "fi": 2.11469}, "bytes_per_chunk": {"en": 4.626, "tr": 5.381, "fi": 4.053}, "bpb_mean": 1.99999, "frozen_inventory": true, "t": 5605.3}
{"event": "train", "step": 5000, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32081, "bpb": 1.90552, "chunk_loss": 0.030674, "F": 0.21823, "G": 0.21612, "chunks": 223.47, "bytes_per_chunk": 4.582, "boundary_entropy": 0.07741, "router_noise": 0.0, "grad_norm": 0.6643, "lr": 5e-05, "router_trainable": true, "sec_per_step": 1.044, "probe_sec": 0.1371, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10815, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": -0.00161, "n_shadow": 5, "mean_delta_shadow": 0.04508, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.19678, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.75014, "trunc_abs_frac": 1.28576, "mean_delta_full": -0.14901, "read_hit_rate": {"surface": 0.47282, "split": 0.57457, "merge": 0.13772, "skel": 0.51195}, "mean_gate": {"surface": 0.17255, "split": 0.063, "merge": 0.1194, "skel": 0.10907}, "n_reads": 14058, "n_cand": 31695, "read_suppressed_frac": 0.0997, "dropout_reads_logged": 14058, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1655, "admitted_surface": 39299, "staged_surface": 56622, "admitted_split": 6725, "staged_split": 42180, "admitted_merge": 13350, "staged_merge": 118288, "admitted_skel": 6162, "staged_skel": 45054, "mean_count": 559.550537109375, "n_util_obs": 271, "util_count_spearman": 0.07829481680878475}, "probe_overhead_sec_total": 786.072, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008872, "alpha": 0.03, "t": 5611.2}
{"event": "consolidate", "step": 5050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262113, "t": 5647.1}
{"event": "train", "step": 5050, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.33939, "bpb": 1.93233, "chunk_loss": 0.030629, "F": 0.21591, "G": 0.21573, "chunks": 221.09, "bytes_per_chunk": 4.632, "boundary_entropy": 0.07551, "router_noise": 0.0, "grad_norm": 0.6456, "lr": 4.81e-05, "router_trainable": true, "sec_per_step": 0.728, "probe_sec": 0.0675, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0468, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.03878, "n_shadow": 4, "mean_delta_shadow": -0.01396, "n_split": 1, "mean_delta_split": 0.05084, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.08464, "trunc_abs_frac": 0.24798, "mean_delta_full": -0.15588, "read_hit_rate": {"surface": 0.4589, "split": 0.56544, "merge": 0.13821, "skel": 0.51115}, "mean_gate": {"surface": 0.1621, "split": 0.06159, "merge": 0.10532, "skel": 0.09961}, "n_reads": 13692, "n_cand": 31343, "read_suppressed_frac": 0.0982, "dropout_reads_logged": 13692, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1662, "admitted_surface": 40077, "staged_surface": 56561, "admitted_split": 6334, "staged_split": 42190, "admitted_merge": 13026, "staged_merge": 118536, "admitted_skel": 6099, "staged_skel": 44857, "mean_count": 565.3984375, "n_util_obs": 268, "util_count_spearman": 0.0503389239841476}, "probe_overhead_sec_total": 791.227, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.007835, "alpha": 0.03, "t": 5647.6}
{"event": "consolidate", "step": 5100, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 1024, "uncertain_staged": 262113, "t": 5683.9}
{"event": "train", "step": 5100, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.30497, "bpb": 1.88267, "chunk_loss": 0.030469, "F": 0.21011, "G": 0.20854, "chunks": 215.16, "bytes_per_chunk": 4.759, "boundary_entropy": 0.0744, "router_noise": 0.0, "grad_norm": 0.6433, "lr": 4.63e-05, "router_trainable": true, "sec_per_step": 0.743, "probe_sec": 0.1031, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08282, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 3, "mean_delta_loo": -0.03507, "n_shadow": 3, "mean_delta_shadow": -0.10543, "n_split": 2, "mean_delta_split": 0.01904, "n_merge": 2, "mean_delta_merge": 0.0881, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 1.02793, "trunc_abs_frac": 1.02793, "mean_delta_full": -0.19885, "read_hit_rate": {"surface": 0.44535, "split": 0.55705, "merge": 0.12403, "skel": 0.50429}, "mean_gate": {"surface": 0.15791, "split": 0.05924, "merge": 0.10357, "skel": 0.09947}, "n_reads": 13017, "n_cand": 30530, "read_suppressed_frac": 0.1002, "dropout_reads_logged": 13017, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1677, "admitted_surface": 40866, "staged_surface": 56543, "admitted_split": 5789, "staged_split": 42796, "admitted_merge": 12840, "staged_merge": 117845, "admitted_skel": 6041, "staged_skel": 44960, "mean_count": 565.895263671875, "n_util_obs": 257, "util_count_spearman": 0.00013290652429644376}, "probe_overhead_sec_total": 795.326, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.007807, "alpha": 0.03, "t": 5684.7}
{"event": "consolidate", "step": 5150, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262113, "t": 5717.5}
{"event": "train", "step": 5150, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.29995, "bpb": 1.87544, "chunk_loss": 0.030671, "F": 0.21695, "G": 0.2161, "chunks": 222.16, "bytes_per_chunk": 4.609, "boundary_entropy": 0.07554, "router_noise": 0.0, "grad_norm": 0.6692, "lr": 4.45e-05, "router_trainable": true, "sec_per_step": 0.693, "probe_sec": 0.1307, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07234, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.0698, "n_shadow": 8, "mean_delta_shadow": 0.04501, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 5.53439, "trunc_abs_frac": 0.40226, "mean_delta_full": 0.02307, "read_hit_rate": {"surface": 0.44963, "split": 0.54063, "merge": 0.12392, "skel": 0.49676}, "mean_gate": {"surface": 0.15581, "split": 0.0569, "merge": 0.10413, "skel": 0.09738}, "n_reads": 13363, "n_cand": 31777, "read_suppressed_frac": 0.0997, "dropout_reads_logged": 13363, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1679, "admitted_surface": 41707, "staged_surface": 56428, "admitted_split": 5285, "staged_split": 42485, "admitted_merge": 12709, "staged_merge": 118031, "admitted_skel": 5835, "staged_skel": 45200, "mean_count": 565.675537109375, "n_util_obs": 257, "util_count_spearman": 0.00760465950644344}, "probe_overhead_sec_total": 799.414, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009284, "alpha": 0.03, "t": 5719.4}
{"event": "consolidate", "step": 5200, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262112, "t": 5762.7}
{"event": "probe", "step": 5200, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3557.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.20439, "cps": 27.789062, "mean_p": 0.203565, "boundary_entropy": 0.07325, "bpb": 1.855764}, "tr": {"chunks": 3385.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.186986, "cps": 26.445312, "mean_p": 0.184352, "boundary_entropy": 0.111111, "bpb": 2.002747}, "fi": {"chunks": 4972.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.264313, "cps": 38.84375, "mean_p": 0.25592, "boundary_entropy": 0.081351, "bpb": 2.101225}}, "bpb": {"en": 1.85576, "tr": 2.00275, "fi": 2.10123}, "gini_cpc": 0.078622, "gini_cps": 0.088803, "bpic": 4.761793, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.91484680692997, "fi": 1.2931815252881649}, "inventory": {"size": 326655, "admitted": 65536, "staged": 261119, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1670, "admitted_surface": 42507, "staged_surface": 56191, "admitted_split": 4793, "staged_split": 43235, "admitted_merge": 12584, "staged_merge": 116484, "admitted_skel": 5652, "staged_skel": 45209, "mean_count": 568.90771484375, "n_util_obs": 257, "util_count_spearman": 0.0005454117107282003}, "t": 5764.3}
{"event": "eval", "step": 5200, "phase": 3, "bpb": {"en": 1.87026, "tr": 2.01222, "fi": 2.10765}, "bytes_per_chunk": {"en": 4.624, "tr": 5.873, "fi": 4.05}, "bpb_mean": 1.99671, "frozen_inventory": true, "t": 5766.8}
{"event": "train", "step": 5200, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.33745, "bpb": 1.92953, "chunk_loss": 0.030873, "F": 0.22278, "G": 0.22248, "chunks": 228.12, "bytes_per_chunk": 4.489, "boundary_entropy": 0.07752, "router_noise": 0.0, "grad_norm": 0.7095, "lr": 4.29e-05, "router_trainable": true, "sec_per_step": 1.028, "probe_sec": 0.0704, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04644, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.09294, "n_shadow": 5, "mean_delta_shadow": 0.05493, "n_split": 1, "mean_delta_split": -0.04822, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.69026, "trunc_abs_frac": 1.3319, "mean_delta_full": 0.42057, "read_hit_rate": {"surface": 0.45666, "split": 0.53582, "merge": 0.13599, "skel": 0.49688}, "mean_gate": {"surface": 0.16637, "split": 0.06061, "merge": 0.11076, "skel": 0.10207}, "n_reads": 13697, "n_cand": 32432, "read_suppressed_frac": 0.0962, "dropout_reads_logged": 13697, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1669, "admitted_surface": 42507, "staged_surface": 56462, "admitted_split": 4793, "staged_split": 43334, "admitted_merge": 12584, "staged_merge": 117072, "admitted_skel": 5652, "staged_skel": 45276, "mean_count": 569.1167602539062, "n_util_obs": 257, "util_count_spearman": 0.0007762307950480414}, "probe_overhead_sec_total": 803.813, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009312, "alpha": 0.03, "t": 5770.7}
{"event": "consolidate", "step": 5250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262111, "t": 5805.3}
{"event": "train", "step": 5250, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.37125, "bpb": 1.9783, "chunk_loss": 0.030739, "F": 0.22137, "G": 0.22096, "chunks": 226.69, "bytes_per_chunk": 4.517, "boundary_entropy": 0.07456, "router_noise": 0.0, "grad_norm": 0.69, "lr": 4.14e-05, "router_trainable": true, "sec_per_step": 0.701, "probe_sec": 0.0793, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04739, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": 0.04713, "n_shadow": 4, "mean_delta_shadow": -0.00665, "n_split": 2, "mean_delta_split": 0.02213, "n_merge": 1, "mean_delta_merge": -0.08655, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.10284, "trunc_abs_frac": 4.05199, "mean_delta_full": 0.1739, "read_hit_rate": {"surface": 0.45929, "split": 0.52151, "merge": 0.13115, "skel": 0.46687}, "mean_gate": {"surface": 0.16558, "split": 0.06094, "merge": 0.10543, "skel": 0.10286}, "n_reads": 13197, "n_cand": 32107, "read_suppressed_frac": 0.0985, "dropout_reads_logged": 13197, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1678, "admitted_surface": 43328, "staged_surface": 56209, "admitted_split": 4007, "staged_split": 43559, "admitted_merge": 12575, "staged_merge": 117066, "admitted_skel": 5626, "staged_skel": 45310, "mean_count": 568.1736450195312, "n_util_obs": 259, "util_count_spearman": -0.011387543945683481}, "probe_overhead_sec_total": 808.291, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008427, "alpha": 0.03, "t": 5805.8}
{"event": "consolidate", "step": 5300, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262109, "t": 5840.6}
{"event": "train", "step": 5300, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.36091, "bpb": 1.96337, "chunk_loss": 0.030588, "F": 0.21143, "G": 0.2128, "chunks": 216.5, "bytes_per_chunk": 4.73, "boundary_entropy": 0.07532, "router_noise": 0.0, "grad_norm": 0.6587, "lr": 3.99e-05, "router_trainable": true, "sec_per_step": 0.705, "probe_sec": 0.0761, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04799, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": 0.16618, "n_shadow": 5, "mean_delta_shadow": -0.10981, "n_split": 2, "mean_delta_split": 0.04736, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.93104, "trunc_abs_frac": 0.74226, "mean_delta_full": 0.17511, "read_hit_rate": {"surface": 0.42662, "split": 0.47275, "merge": 0.10999, "skel": 0.4251}, "mean_gate": {"surface": 0.15806, "split": 0.05891, "merge": 0.10231, "skel": 0.09914}, "n_reads": 11555, "n_cand": 30908, "read_suppressed_frac": 0.1026, "dropout_reads_logged": 11555, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1672, "admitted_surface": 44110, "staged_surface": 56019, "admitted_split": 3364, "staged_split": 44125, "admitted_merge": 12438, "staged_merge": 116884, "admitted_skel": 5624, "staged_skel": 45116, "mean_count": 555.0571899414062, "n_util_obs": 248, "util_count_spearman": -0.04815658087489045}, "probe_overhead_sec_total": 812.237, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008902, "alpha": 0.03, "t": 5841.1}
{"event": "consolidate", "step": 5350, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262107, "t": 5886.4}
{"event": "train", "step": 5350, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32297, "bpb": 1.90864, "chunk_loss": 0.030826, "F": 0.21484, "G": 0.21597, "chunks": 220.0, "bytes_per_chunk": 4.655, "boundary_entropy": 0.07698, "router_noise": 0.0, "grad_norm": 0.7439, "lr": 3.86e-05, "router_trainable": true, "sec_per_step": 0.917, "probe_sec": 0.0715, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04297, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 11, "mean_delta_loo": 0.05615, "n_shadow": 3, "mean_delta_shadow": -0.01835, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.46723, "trunc_abs_frac": 0.37623, "mean_delta_full": 0.1427, "read_hit_rate": {"surface": 0.44035, "split": 0.42875, "merge": 0.11024, "skel": 0.38256}, "mean_gate": {"surface": 0.17782, "split": 0.06069, "merge": 0.11664, "skel": 0.10759}, "n_reads": 10910, "n_cand": 31002, "read_suppressed_frac": 0.0985, "dropout_reads_logged": 10910, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1632, "admitted_surface": 44993, "staged_surface": 55763, "admitted_split": 2931, "staged_split": 44373, "admitted_merge": 12061, "staged_merge": 116957, "admitted_skel": 5551, "staged_skel": 45051, "mean_count": 539.8118896484375, "n_util_obs": 232, "util_count_spearman": -0.10224990341906133}, "probe_overhead_sec_total": 816.93, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009288, "alpha": 0.03, "t": 5886.9}
{"event": "consolidate", "step": 5400, "retired": 0, "split_retired": 3, "merge_admitted": 0, "admitted": 3, "swapped": 1024, "uncertain_staged": 262105, "t": 5918.8}
{"event": "probe", "step": 5400, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3559.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.204505, "cps": 27.804688, "mean_p": 0.20376, "boundary_entropy": 0.071609, "bpb": 1.85193}, "tr": {"chunks": 3495.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.193062, "cps": 27.304688, "mean_p": 0.186004, "boundary_entropy": 0.108557, "bpb": 1.988926}, "fi": {"chunks": 4977.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.264579, "cps": 38.882812, "mean_p": 0.256259, "boundary_entropy": 0.08032, "bpb": 2.09294}}, "bpb": {"en": 1.85193, "tr": 1.98893, "fi": 2.09294}, "gini_cpc": 0.072006, "gini_cps": 0.082121, "bpic": 4.715485, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9440451400936576, "fi": 1.2937545480023713}, "inventory": {"size": 326653, "admitted": 65536, "staged": 261117, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1631, "admitted_surface": 45878, "staged_surface": 55552, "admitted_split": 2158, "staged_split": 44400, "admitted_merge": 12022, "staged_merge": 116085, "admitted_skel": 5478, "staged_skel": 45080, "mean_count": 527.2996826171875, "n_util_obs": 218, "util_count_spearman": -0.11706412751951879}, "t": 5920.3}
{"event": "eval", "step": 5400, "phase": 3, "bpb": {"en": 1.86646, "tr": 1.99918, "fi": 2.09861}, "bytes_per_chunk": {"en": 4.622, "tr": 5.613, "fi": 4.046}, "bpb_mean": 1.98808, "frozen_inventory": true, "t": 5922.7}
{"event": "train", "step": 5400, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.24773, "bpb": 1.80009, "chunk_loss": 0.03056, "F": 0.20813, "G": 0.20826, "chunks": 213.12, "bytes_per_chunk": 4.805, "boundary_entropy": 0.0756, "router_noise": 0.0, "grad_norm": 0.6406, "lr": 3.73e-05, "router_trainable": true, "sec_per_step": 0.795, "probe_sec": 0.0908, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05953, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": 0.01939, "n_shadow": 9, "mean_delta_shadow": 0.02671, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.57458, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.70627, "trunc_abs_frac": 0.58726, "mean_delta_full": 0.18445, "read_hit_rate": {"surface": 0.40822, "split": 0.41773, "merge": 0.12152, "skel": 0.37415}, "mean_gate": {"surface": 0.16613, "split": 0.0605, "merge": 0.10991, "skel": 0.10556}, "n_reads": 10378, "n_cand": 30325, "read_suppressed_frac": 0.1001, "dropout_reads_logged": 10378, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1631, "admitted_surface": 45878, "staged_surface": 55746, "admitted_split": 2158, "staged_split": 44568, "admitted_merge": 12022, "staged_merge": 116535, "admitted_skel": 5478, "staged_skel": 45295, "mean_count": 527.4580078125, "n_util_obs": 218, "util_count_spearman": -0.11708613521178639}, "probe_overhead_sec_total": 821.307, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009661, "alpha": 0.03, "t": 5926.6}
{"event": "consolidate", "step": 5450, "retired": 0, "split_retired": 4, "merge_admitted": 0, "admitted": 4, "swapped": 1024, "uncertain_staged": 262105, "t": 5960.4}
{"event": "train", "step": 5450, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.29314, "bpb": 1.8656, "chunk_loss": 0.030622, "F": 0.21683, "G": 0.21669, "chunks": 222.03, "bytes_per_chunk": 4.612, "boundary_entropy": 0.07095, "router_noise": 0.0, "grad_norm": 0.6493, "lr": 3.62e-05, "router_trainable": true, "sec_per_step": 0.684, "probe_sec": 0.0783, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04562, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.07187, "n_shadow": 6, "mean_delta_shadow": 0.09233, "n_split": 1, "mean_delta_split": -0.63159, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.62152, "trunc_abs_frac": 1.26276, "mean_delta_full": 0.20586, "read_hit_rate": {"surface": 0.39785, "split": 0.36485, "merge": 0.09771, "skel": 0.33931}, "mean_gate": {"surface": 0.16733, "split": 0.06063, "merge": 0.11357, "skel": 0.10466}, "n_reads": 9839, "n_cand": 31862, "read_suppressed_frac": 0.0908, "dropout_reads_logged": 9839, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1592, "admitted_surface": 46717, "staged_surface": 55559, "admitted_split": 1835, "staged_split": 44677, "admitted_merge": 11769, "staged_merge": 116850, "admitted_skel": 5215, "staged_skel": 45058, "mean_count": 494.702392578125, "n_util_obs": 213, "util_count_spearman": -0.09036961287152416}, "probe_overhead_sec_total": 825.387, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009878, "alpha": 0.03, "t": 5960.9}
{"event": "consolidate", "step": 5500, "retired": 0, "split_retired": 3, "merge_admitted": 0, "admitted": 3, "swapped": 1024, "uncertain_staged": 262104, "t": 6002.9}
{"event": "train", "step": 5500, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.31192, "bpb": 1.89269, "chunk_loss": 0.030614, "F": 0.21545, "G": 0.21399, "chunks": 220.62, "bytes_per_chunk": 4.641, "boundary_entropy": 0.07572, "router_noise": 0.0, "grad_norm": 0.6684, "lr": 3.51e-05, "router_trainable": true, "sec_per_step": 0.85, "probe_sec": 0.0858, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06146, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.01592, "n_shadow": 6, "mean_delta_shadow": -0.00015, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.07349, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -0.31503, "trunc_abs_frac": 1.18011, "mean_delta_full": -0.07027, "read_hit_rate": {"surface": 0.39229, "split": 0.29226, "merge": 0.09949, "skel": 0.31362}, "mean_gate": {"surface": 0.17619, "split": 0.06154, "merge": 0.11313, "skel": 0.11529}, "n_reads": 8607, "n_cand": 31276, "read_suppressed_frac": 0.0991, "dropout_reads_logged": 8607, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1550, "admitted_surface": 47539, "staged_surface": 55355, "admitted_split": 1552, "staged_split": 45159, "admitted_merge": 11370, "staged_merge": 116561, "admitted_skel": 5075, "staged_skel": 45069, "mean_count": 447.36456298828125, "n_util_obs": 206, "util_count_spearman": -0.05827372472819128}, "probe_overhead_sec_total": 829.647, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009874, "alpha": 0.03, "t": 6003.4}
{"event": "consolidate", "step": 5550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262102, "t": 6045.4}
{"event": "train", "step": 5550, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.28584, "bpb": 1.85508, "chunk_loss": 0.030767, "F": 0.22504, "G": 0.22243, "chunks": 230.44, "bytes_per_chunk": 4.444, "boundary_entropy": 0.07108, "router_noise": 0.0, "grad_norm": 0.6735, "lr": 3.41e-05, "router_trainable": true, "sec_per_step": 0.85, "probe_sec": 0.1019, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08162, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.11812, "n_shadow": 7, "mean_delta_shadow": 0.00716, "n_split": 1, "mean_delta_split": 0.03604, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.39813, "trunc_abs_frac": 0.60021, "mean_delta_full": 0.29766, "read_hit_rate": {"surface": 0.39812, "split": 0.19405, "merge": 0.08331, "skel": 0.31726}, "mean_gate": {"surface": 0.16089, "split": 0.05821, "merge": 0.10891, "skel": 0.10718}, "n_reads": 7716, "n_cand": 32723, "read_suppressed_frac": 0.0982, "dropout_reads_logged": 7716, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1489, "admitted_surface": 48290, "staged_surface": 55194, "admitted_split": 1524, "staged_split": 44895, "admitted_merge": 10707, "staged_merge": 116877, "admitted_skel": 5015, "staged_skel": 45178, "mean_count": 387.87139892578125, "n_util_obs": 199, "util_count_spearman": -0.03825644824747051}, "probe_overhead_sec_total": 834.007, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009648, "alpha": 0.03, "t": 6045.9}
{"event": "consolidate", "step": 5600, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262099, "t": 6080.8}
{"event": "probe", "step": 5600, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3558.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.204448, "cps": 27.796875, "mean_p": 0.203502, "boundary_entropy": 0.072248, "bpb": 1.84816}, "tr": {"chunks": 3521.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.194498, "cps": 27.507812, "mean_p": 0.1859, "boundary_entropy": 0.108095, "bpb": 1.983802}, "fi": {"chunks": 4969.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.264154, "cps": 38.820312, "mean_p": 0.255648, "boundary_entropy": 0.079746, "bpb": 2.086816}}, "bpb": {"en": 1.84816, "tr": 1.9838, "fi": 2.08682}, "gini_cpc": 0.070031, "gini_cps": 0.080124, "bpic": 4.708831, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9513353837451966, "fi": 1.2920380086890784}, "inventory": {"size": 326655, "admitted": 65536, "staged": 261119, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1486, "admitted_surface": 48278, "staged_surface": 54865, "admitted_split": 1514, "staged_split": 44780, "admitted_merge": 10802, "staged_merge": 116595, "admitted_skel": 4942, "staged_skel": 44879, "mean_count": 384.8026123046875, "n_util_obs": 202, "util_count_spearman": -0.027378612506369658}, "t": 6082.3}
{"event": "eval", "step": 5600, "phase": 3, "bpb": {"en": 1.86322, "tr": 1.99347, "fi": 2.09322}, "bytes_per_chunk": {"en": 4.626, "tr": 5.571, "fi": 4.048}, "bpb_mean": 1.9833, "frozen_inventory": true, "t": 6084.8}
{"event": "train", "step": 5600, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.319, "bpb": 1.90292, "chunk_loss": 0.030737, "F": 0.21967, "G": 0.2188, "chunks": 224.94, "bytes_per_chunk": 4.552, "boundary_entropy": 0.07808, "router_noise": 0.0, "grad_norm": 0.692, "lr": 3.33e-05, "router_trainable": true, "sec_per_step": 0.879, "probe_sec": 0.0733, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05261, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": 0.01529, "n_shadow": 4, "mean_delta_shadow": 0.00124, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.09726, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.14344, "trunc_abs_frac": 0.14344, "mean_delta_full": 0.28467, "read_hit_rate": {"surface": 0.3892, "split": 0.20396, "merge": 0.07205, "skel": 0.28558}, "mean_gate": {"surface": 0.18748, "split": 0.06786, "merge": 0.11535, "skel": 0.11844}, "n_reads": 7295, "n_cand": 31761, "read_suppressed_frac": 0.1072, "dropout_reads_logged": 7295, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1486, "admitted_surface": 48278, "staged_surface": 55121, "admitted_split": 1514, "staged_split": 44824, "admitted_merge": 10802, "staged_merge": 117203, "admitted_skel": 4942, "staged_skel": 44996, "mean_count": 384.9139404296875, "n_util_obs": 203, "util_count_spearman": -0.019207382566463017}, "probe_overhead_sec_total": 838.165, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009878, "alpha": 0.03, "t": 6089.8}
{"event": "consolidate", "step": 5650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262097, "t": 6128.3}
{"event": "train", "step": 5650, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.3042, "bpb": 1.88157, "chunk_loss": 0.030633, "F": 0.21625, "G": 0.21532, "chunks": 221.44, "bytes_per_chunk": 4.624, "boundary_entropy": 0.07417, "router_noise": 0.0, "grad_norm": 0.6578, "lr": 3.25e-05, "router_trainable": true, "sec_per_step": 0.788, "probe_sec": 0.0885, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04864, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": 0.12757, "n_shadow": 4, "mean_delta_shadow": -0.27751, "n_split": 1, "mean_delta_split": -0.24005, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 1.33105, "trunc_abs_frac": 0.84213, "mean_delta_full": -0.1734, "read_hit_rate": {"surface": 0.39269, "split": 0.18395, "merge": 0.07135, "skel": 0.28336}, "mean_gate": {"surface": 0.17619, "split": 0.065, "merge": 0.11809, "skel": 0.11466}, "n_reads": 6988, "n_cand": 31435, "read_suppressed_frac": 0.0957, "dropout_reads_logged": 6988, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1506, "admitted_surface": 48152, "staged_surface": 55234, "admitted_split": 1511, "staged_split": 45070, "admitted_merge": 10888, "staged_merge": 116566, "admitted_skel": 4985, "staged_skel": 45274, "mean_count": 388.33489990234375, "n_util_obs": 208, "util_count_spearman": -0.027590648916995218}, "probe_overhead_sec_total": 842.461, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009318, "alpha": 0.03, "t": 6129.2}
{"event": "consolidate", "step": 5700, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262096, "t": 6170.0}
{"event": "train", "step": 5700, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38255, "bpb": 1.99459, "chunk_loss": 0.030703, "F": 0.21976, "G": 0.21826, "chunks": 225.03, "bytes_per_chunk": 4.55, "boundary_entropy": 0.07605, "router_noise": 0.0, "grad_norm": 0.6954, "lr": 3.18e-05, "router_trainable": true, "sec_per_step": 0.835, "probe_sec": 0.0815, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04658, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.1473, "n_shadow": 3, "mean_delta_shadow": -0.04026, "n_split": 2, "mean_delta_split": -0.07895, "n_merge": 1, "mean_delta_merge": -0.01931, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -0.14001, "trunc_abs_frac": 0.39525, "mean_delta_full": 0.23148, "read_hit_rate": {"surface": 0.39336, "split": 0.18091, "merge": 0.07552, "skel": 0.27411}, "mean_gate": {"surface": 0.18459, "split": 0.06717, "merge": 0.11975, "skel": 0.12013}, "n_reads": 7058, "n_cand": 31969, "read_suppressed_frac": 0.1006, "dropout_reads_logged": 7058, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1533, "admitted_surface": 48007, "staged_surface": 55198, "admitted_split": 1514, "staged_split": 44858, "admitted_merge": 11004, "staged_merge": 117056, "admitted_skel": 5011, "staged_skel": 45032, "mean_count": 391.7771911621094, "n_util_obs": 209, "util_count_spearman": -0.00037725432462274567}, "probe_overhead_sec_total": 846.569, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009361, "alpha": 0.03, "t": 6171.0}
{"event": "consolidate", "step": 5750, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 1024, "uncertain_staged": 262091, "t": 6202.0}
{"event": "train", "step": 5750, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.36381, "bpb": 1.96756, "chunk_loss": 0.030664, "F": 0.21588, "G": 0.21578, "chunks": 221.06, "bytes_per_chunk": 4.632, "boundary_entropy": 0.07562, "router_noise": 0.0, "grad_norm": 0.701, "lr": 3.13e-05, "router_trainable": true, "sec_per_step": 0.639, "probe_sec": 0.0655, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04417, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": -0.16176, "n_shadow": 4, "mean_delta_shadow": 0.00652, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": 0.03391, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 1.1949, "trunc_abs_frac": 1.02328, "mean_delta_full": -0.05981, "read_hit_rate": {"surface": 0.37489, "split": 0.18432, "merge": 0.05991, "skel": 0.26667}, "mean_gate": {"surface": 0.17946, "split": 0.06643, "merge": 0.11412, "skel": 0.12166}, "n_reads": 6674, "n_cand": 31314, "read_suppressed_frac": 0.1013, "dropout_reads_logged": 6674, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1531, "admitted_surface": 47896, "staged_surface": 55209, "admitted_split": 1508, "staged_split": 45043, "admitted_merge": 11094, "staged_merge": 116962, "admitted_skel": 5038, "staged_skel": 44930, "mean_count": 391.86651611328125, "n_util_obs": 205, "util_count_spearman": -0.01867643576492875}, "probe_overhead_sec_total": 850.533, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009646, "alpha": 0.03, "t": 6202.9}
{"event": "consolidate", "step": 5800, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262091, "t": 6236.6}
{"event": "probe", "step": 5800, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3567.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.204965, "cps": 27.867188, "mean_p": 0.204087, "boundary_entropy": 0.071326, "bpb": 1.843317}, "tr": {"chunks": 3448.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.190466, "cps": 26.9375, "mean_p": 0.185395, "boundary_entropy": 0.107697, "bpb": 1.982896}, "fi": {"chunks": 4968.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.264101, "cps": 38.8125, "mean_p": 0.255625, "boundary_entropy": 0.079884, "bpb": 2.080212}}, "bpb": {"en": 1.84332, "tr": 1.9829, "fi": 2.08021}, "gini_cpc": 0.074432, "gini_cps": 0.084564, "bpic": 4.734374, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9292610125955733, "fi": 1.2885186668734663}, "inventory": {"size": 326655, "admitted": 65536, "staged": 261119, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1368, "admitted_surface": 48521, "staged_surface": 55079, "admitted_split": 1288, "staged_split": 45106, "admitted_merge": 10719, "staged_merge": 115952, "admitted_skel": 5008, "staged_skel": 44982, "mean_count": 365.67022705078125, "n_util_obs": 204, "util_count_spearman": -0.0768345689831862}, "t": 6238.2}
{"event": "eval", "step": 5800, "phase": 3, "bpb": {"en": 1.85829, "tr": 1.99253, "fi": 2.0864}, "bytes_per_chunk": {"en": 4.618, "tr": 5.723, "fi": 4.049}, "bpb_mean": 1.97907, "frozen_inventory": true, "t": 6240.7}
{"event": "train", "step": 5800, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.2836, "bpb": 1.85184, "chunk_loss": 0.030589, "F": 0.21445, "G": 0.21375, "chunks": 219.59, "bytes_per_chunk": 4.663, "boundary_entropy": 0.07464, "router_noise": 0.0, "grad_norm": 0.7264, "lr": 3.08e-05, "router_trainable": true, "sec_per_step": 0.843, "probe_sec": 0.0674, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04335, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.09944, "n_shadow": 3, "mean_delta_shadow": 0.08002, "n_split": 1, "mean_delta_split": -0.3678, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.10142, "trunc_abs_frac": 0.10142, "mean_delta_full": 0.30994, "read_hit_rate": {"surface": 0.35911, "split": 0.14635, "merge": 0.04826, "skel": 0.28186}, "mean_gate": {"surface": 0.1717, "split": 0.06525, "merge": 0.10502, "skel": 0.12001}, "n_reads": 6129, "n_cand": 31436, "read_suppressed_frac": 0.0945, "dropout_reads_logged": 6129, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1367, "admitted_surface": 48521, "staged_surface": 55254, "admitted_split": 1288, "staged_split": 45307, "admitted_merge": 10719, "staged_merge": 116547, "admitted_skel": 5008, "staged_skel": 45036, "mean_count": 365.76373291015625, "n_util_obs": 204, "util_count_spearman": -0.07943120057106105}, "probe_overhead_sec_total": 855.206, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009747, "alpha": 0.03, "t": 6245.1}
{"event": "consolidate", "step": 5850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262086, "t": 6275.7}
{"event": "train", "step": 5850, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.3372, "bpb": 1.92917, "chunk_loss": 0.030491, "F": 0.20743, "G": 0.20762, "chunks": 212.41, "bytes_per_chunk": 4.821, "boundary_entropy": 0.07273, "router_noise": 0.0, "grad_norm": 0.7069, "lr": 3.05e-05, "router_trainable": true, "sec_per_step": 0.621, "probe_sec": 0.0703, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0441, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": 0.10382, "n_shadow": 2, "mean_delta_shadow": -0.03711, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": 0.13454, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.80817, "trunc_abs_frac": 1.38913, "mean_delta_full": -0.32422, "read_hit_rate": {"surface": 0.35358, "split": 0.13553, "merge": 0.03832, "skel": 0.25063}, "mean_gate": {"surface": 0.17426, "split": 0.06336, "merge": 0.11014, "skel": 0.11718}, "n_reads": 5523, "n_cand": 30315, "read_suppressed_frac": 0.1018, "dropout_reads_logged": 5523, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1370, "admitted_surface": 48442, "staged_surface": 55066, "admitted_split": 1280, "staged_split": 45190, "admitted_merge": 10826, "staged_merge": 116859, "admitted_skel": 4988, "staged_skel": 45029, "mean_count": 362.10479736328125, "n_util_obs": 199, "util_count_spearman": -0.09231460331962844}, "probe_overhead_sec_total": 858.889, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.010674, "alpha": 0.03, "t": 6276.1}
{"event": "consolidate", "step": 5900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262084, "t": 6315.5}
{"event": "train", "step": 5900, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.36714, "bpb": 1.97237, "chunk_loss": 0.030725, "F": 0.22134, "G": 0.2196, "chunks": 226.66, "bytes_per_chunk": 4.518, "boundary_entropy": 0.0757, "router_noise": 0.0, "grad_norm": 0.7231, "lr": 3.02e-05, "router_trainable": true, "sec_per_step": 0.805, "probe_sec": 0.0681, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04802, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.24899, "n_shadow": 2, "mean_delta_shadow": 0.16113, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.24158, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 2.5182, "trunc_abs_frac": 5.52955, "mean_delta_full": 0.10847, "read_hit_rate": {"surface": 0.32129, "split": 0.11864, "merge": 0.04855, "skel": 0.22178}, "mean_gate": {"surface": 0.18535, "split": 0.06998, "merge": 0.11811, "skel": 0.12562}, "n_reads": 5361, "n_cand": 32342, "read_suppressed_frac": 0.0903, "dropout_reads_logged": 5361, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1264, "admitted_surface": 48555, "staged_surface": 55245, "admitted_split": 1205, "staged_split": 44965, "admitted_merge": 10926, "staged_merge": 116869, "admitted_skel": 4850, "staged_skel": 45065, "mean_count": 337.0045166015625, "n_util_obs": 197, "util_count_spearman": -0.04013905244525187}, "probe_overhead_sec_total": 863.029, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.011372, "alpha": 0.03, "t": 6316.3}
{"event": "consolidate", "step": 5950, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262081, "t": 6348.4}
{"event": "train", "step": 5950, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.29685, "bpb": 1.87096, "chunk_loss": 0.030594, "F": 0.2132, "G": 0.21259, "chunks": 218.31, "bytes_per_chunk": 4.691, "boundary_entropy": 0.07685, "router_noise": 0.0, "grad_norm": 0.76, "lr": 3.01e-05, "router_trainable": true, "sec_per_step": 0.649, "probe_sec": 0.0694, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04562, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 3, "mean_delta_loo": 0.03343, "n_shadow": 8, "mean_delta_shadow": -0.09785, "n_split": 1, "mean_delta_split": -0.03979, "n_merge": 1, "mean_delta_merge": -0.14355, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.31956, "trunc_abs_frac": 0.71922, "mean_delta_full": 0.20718, "read_hit_rate": {"surface": 0.27538, "split": 0.06225, "merge": 0.02832, "skel": 0.20458}, "mean_gate": {"surface": 0.19117, "split": 0.08511, "merge": 0.11369, "skel": 0.13048}, "n_reads": 3944, "n_cand": 31240, "read_suppressed_frac": 0.1062, "dropout_reads_logged": 3944, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1083, "admitted_surface": 48824, "staged_surface": 55634, "admitted_split": 1069, "staged_split": 44982, "admitted_merge": 10774, "staged_merge": 116352, "admitted_skel": 4869, "staged_skel": 45176, "mean_count": 271.68994140625, "n_util_obs": 175, "util_count_spearman": 0.08103224361845052}, "probe_overhead_sec_total": 866.662, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.012744, "alpha": 0.03, "t": 6348.8}
{"event": "consolidate", "step": 6000, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 1024, "uncertain_staged": 262080, "t": 6381.2}
{"event": "train", "step": 5999, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32511, "bpb": 1.91173, "chunk_loss": 0.030771, "F": 0.22058, "G": 0.21976, "chunks": 225.88, "bytes_per_chunk": 4.533, "boundary_entropy": 0.0767, "router_noise": 0.0, "grad_norm": 0.7134, "lr": 3e-05, "router_trainable": true, "sec_per_step": 0.649, "probe_sec": 0.0802, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04218, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": -0.0235, "n_shadow": 5, "mean_delta_shadow": -0.0826, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -0.05878, "trunc_abs_frac": 0.26491, "mean_delta_full": -0.104, "read_hit_rate": {"surface": 0.27682, "split": 0.05692, "merge": 0.03057, "skel": 0.19276}, "mean_gate": {"surface": 0.19257, "split": 0.08236, "merge": 0.11463, "skel": 0.13607}, "n_reads": 3982, "n_cand": 32350, "read_suppressed_frac": 0.0881, "dropout_reads_logged": 3982, "inventory": {"size": 326654, "admitted": 65536, "staged": 261118, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1014, "admitted_surface": 49700, "staged_surface": 55590, "admitted_split": 1056, "staged_split": 44818, "admitted_merge": 9928, "staged_merge": 115773, "admitted_skel": 4852, "staged_skel": 44937, "mean_count": 256.76251220703125, "n_util_obs": 158, "util_count_spearman": 0.014961383353692579}, "probe_overhead_sec_total": 870.34, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.012397, "alpha": 0.03, "t": 6381.2}
{"event": "probe", "step": 6000, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3562.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.204677, "cps": 27.828125, "mean_p": 0.203437, "boundary_entropy": 0.071286, "bpb": 1.842787}, "tr": {"chunks": 3469.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.191626, "cps": 27.101562, "mean_p": 0.184538, "boundary_entropy": 0.107026, "bpb": 1.977688}, "fi": {"chunks": 4988.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.265164, "cps": 38.96875, "mean_p": 0.255982, "boundary_entropy": 0.078953, "bpb": 2.078593}}, "bpb": {"en": 1.84279, "tr": 1.97769, "fi": 2.07859}, "gini_cpc": 0.074116, "gini_cps": 0.084255, "bpic": 4.720193, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9362330184787326, "fi": 1.2955219225994945}, "inventory": {"size": 326654, "admitted": 65536, "staged": 261118, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1014, "admitted_surface": 49700, "staged_surface": 55590, "admitted_split": 1056, "staged_split": 44818, "admitted_merge": 9928, "staged_merge": 115773, "admitted_skel": 4852, "staged_skel": 44937, "mean_count": 256.76251220703125, "n_util_obs": 158, "util_count_spearman": 0.014961383353692579}, "t": 6382.7}
{"event": "eval", "step": 6000, "phase": 3, "bpb": {"en": 1.85726, "tr": 1.98786, "fi": 2.08449}, "bytes_per_chunk": {"en": 4.626, "tr": 5.667, "fi": 4.038}, "bpb_mean": 1.97654, "frozen_inventory": true, "t": 6385.1}
{"event": "reps_dumped", "dirs": ["/local/scratch/sas245/beetle-hnet/runs/ali_ali_s0/probe/reps/th0.30", "/local/scratch/sas245/beetle-hnet/runs/ali_ali_s0/probe/reps/th0.40", "/local/scratch/sas245/beetle-hnet/runs/ali_ali_s0/probe/reps/th0.50", "/local/scratch/sas245/beetle-hnet/runs/ali_ali_s0/probe/reps/th0.60", "/local/scratch/sas245/beetle-hnet/runs/ali_ali_s0/probe/reps/th0.70"], "t": 6389.2}
{"event": "final", "run": "ali_ali_s0", "arch": "ali", "objective": "ratio", "condition": "A", "step": 6000, "params": 22785088, "seed": 0, "per_lang": {"en": {"cpc": 0.20467735447911278, "cps": 27.828125, "bpb": 1.8633615092116778, "bytes_per_chunk": 4.578294736106745, "reads_per_chunk": 0.5181634007474938}, "tr": {"cpc": 0.1916256973982213, "cps": 27.1015625, "bpb": 1.9796491264364238, "bytes_per_chunk": 5.653005552859993, "reads_per_chunk": 0.4338239258181034}, "fi": {"cpc": 0.26516399978735844, "cps": 38.96875, "bpb": 2.102437171915831, "bytes_per_chunk": 4.022063166939004, "reads_per_chunk": 0.5245065322623952}}, "bpb_mean": 1.9818159358546443, "gini_cpc": 0.07411636725171777, "gini_cps": 0.08425548437196673, "bpic": 4.720193027706132, "mean_boundary_entropy": 0.07669578492641449, "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": 870.3404922485352, "probe_overhead_seq_frac": 0.4999166666666667, "wall_sec": 6394.039115190506, "inventory": {"size": 326654, "admitted": 65536, "staged": 261118, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1014, "admitted_surface": 49700, "staged_surface": 55590, "admitted_split": 1056, "staged_split": 44818, "admitted_merge": 9928, "staged_merge": 115773, "admitted_skel": 4852, "staged_skel": 44937, "mean_count": 256.76251220703125, "n_util_obs": 158, "util_count_spearman": 0.014961383353692579}, "ali_params": 268096, "ali_config": {"enabled": true, "budget_B": 65536, "staging_capacity": 262144, "policy": "utility", "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": 0.014961383353692579, "paired_vs_dropout": {"n": 345, "pearson": 0.10045638979176129, "spearman": 0.007687744940570643, "mean_abs_discrepancy": 0.4151502549648285, "mean_paired": 0.07601567357778549, "mean_dropout": 0.06797131896018982}, "predictor_calibration": {"n": 876, "all": {"n": 876, "pearson": 0.13681471618952112, "spearman": 0.165901986756622, "mae": 0.1180039495229721, "reliability": [{"pred": 0.008998926728963852, "obs": 0.013264304026961327, "n": 88}, {"pred": 0.024100078269839287, "obs": 0.03168987110257149, "n": 87}, {"pred": 0.029995130375027657, "obs": 0.0712234303355217, "n": 88}, {"pred": 0.033803485333919525, "obs": 0.035896942019462585, "n": 87}, {"pred": 0.03712466359138489, "obs": 0.05713080242276192, "n": 88}, {"pred": 0.03993796929717064, "obs": 0.06282036006450653, "n": 88}, {"pred": 0.043663058429956436, "obs": 0.058250121772289276, "n": 87}, {"pred": 0.05019814893603325, "obs": 0.07246245443820953, "n": 87}, {"pred": 0.06242417171597481, "obs": 0.10975954681634903, "n": 88}, {"pred": 0.09033317118883133, "obs": 0.1332487165927887, "n": 88}]}, "admitted": {"n": 729, "pearson": 0.08544018632175758, "spearman": 0.09002875609737895, "mae": 0.12969015538692474, "reliability": [{"pred": 0.02281586453318596, "obs": 0.05334170162677765, "n": 73}, {"pred": 0.0300038643181324, "obs": 0.0836091861128807, "n": 73}, {"pred": 0.03323758766055107, "obs": 0.045768171548843384, "n": 73}, {"pred": 0.03632484748959541, "obs": 0.04184611141681671, "n": 73}, {"pred": 0.038959965109825134, "obs": 0.10057094693183899, "n": 72}, {"pred": 0.04187716171145439, "obs": 0.04433668404817581, "n": 73}, {"pred": 0.046852365136146545, "obs": 0.08143503963947296, "n": 73}, {"pred": 0.05364459753036499, "obs": 0.11871179938316345, "n": 73}, {"pred": 0.06735952943563461, "obs": 0.10214866697788239, "n": 73}, {"pred": 0.09312524646520615, "obs": 0.12422539293766022, "n": 73}]}, "staged": {"n": 147, "pearson": -0.08867445450936956, "spearman": -0.134311555055195, "mae": 0.060049932450056076, "reliability": [{"pred": -0.015993375331163406, "obs": 0.007137191481888294, "n": 15}, {"pred": 0.0036264939699321985, "obs": -0.025403428822755814, "n": 15}, {"pred": 0.013626416213810444, "obs": -0.009133473969995975, "n": 14}, {"pred": 0.018318302929401398, "obs": 0.0204047542065382, "n": 15}, {"pred": 0.020198455080389977, "obs": 0.02128237672150135, "n": 14}, {"pred": 0.021048178896307945, "obs": -0.015680933371186256, "n": 15}, {"pred": 0.02490708976984024, "obs": -7.826586806913838e-05, "n": 15}, {"pred": 0.031698454171419144, "obs": -0.03762463852763176, "n": 14}, {"pred": 0.03959312662482262, "obs": -0.02788812853395939, "n": 15}, {"pred": 0.04810865968465805, "obs": -0.026685316115617752, "n": 15}]}}, "t": 6395.0}