{"event": "hub", "repo": "AdaptiveChunking/ali_ali-ridge_s2", "keep_steps": [0, 1, 2, 4, 6, 8, 16, 32, 64, 128, 256, 512, 1000, 1024, 2048, 3000, 4096, 6000], "t": 3.9}
{"event": "start", "run": "ali_ali-ridge_s2", "arch": "ali", "objective": "ratio", "condition": "A", "params": 22785088, "langs": ["en", "tr", "fi"], "premium": {"en": 1.0, "tr": 1.12, "fi": 1.104}, "schedule": {"tau1": 0.4, "tau2": 0.7, "p1_objective": "global", "p2_objective": "global", "p3_objective": "global", "freeze_in_p3": false, "freeze_scope": "router"}, "corpus": "/local/scratch/sas245/beetle-hnet/scratch/data/tier1_corpus", "premium_kind": "flores_per_sent", "doc_filtered": {}, "weights": {"en": 0.74, "tr": 0.1536, "fi": 0.1064}, "device": "cuda", "ali_params": 268096, "ali_config": {"enabled": true, "budget_B": 65536, "staging_capacity": 262144, "policy": "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": "ridge", "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": 3.9}
{"event": "probe_schedule", "n": 45, "steps": [0, 1, 2, 4, 6, 8, 16, 32, 64, 128, 200, 256, 400, 512, 600, 800, 1000, 1024, 1200, 1400, 1600, 1800, 2000, 2048, 2200, 2400, 2600, 2800, 3000, 3200, 3400, 3600, 3800, 4000, 4096, 4200, 4400, 4600, 4800, 5000, 5200, 5400, 5600, 5800, 6000], "t": 3.9}
{"event": "probe", "step": 0, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 10933.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.628225, "cps": 85.414062, "mean_p": 0.513249, "boundary_entropy": 0.686295, "bpb": 8.12598}, "tr": {"chunks": 8187.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.452245, "cps": 63.960938, "mean_p": 0.498002, "boundary_entropy": 0.686826, "bpb": 8.097536}, "fi": {"chunks": 13540.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.719792, "cps": 105.78125, "mean_p": 0.517169, "boundary_entropy": 0.686661, "bpb": 8.065117}}, "bpb": {"en": 8.12598, "tr": 8.09754, "fi": 8.06512}, "gini_cpc": 0.099077, "gini_cps": 0.109267, "bpic": 1.737048, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.7198781817505053, "fi": 1.1457544508645294}, "inventory": {"size": 0, "admitted": 0, "staged": 0, "budget_B": 65536, "staging_capacity": 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": 7.5}
{"event": "train", "step": 0, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 5.62735, "bpb": 8.11854, "chunk_loss": 0.062086, "F": 0.59833, "G": 0.50836, "chunks": 612.69, "bytes_per_chunk": 1.671, "boundary_entropy": 0.69105, "router_noise": 0.0, "grad_norm": 11.4935, "lr": 1e-06, "router_trainable": true, "sec_per_step": 0.046, "probe_sec": 0.1289, "probe_extra_seqs": 0, "n_rows": 0, "extra_fwd_seqs": 0, "wall_s": 0.0, "overhead_frac_est": 0.0, "chunk_mismatch": 0, "n_loo": 0, "mean_delta_loo": NaN, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "read_hit_rate": {"surface": 0.0, "split": 0.0, "merge": 0.0, "skel": 0.0}, "mean_gate": {"surface": null, "split": null, "merge": null, "skel": null}, "n_reads": 0, "n_cand": 58765, "read_suppressed_frac": 0.0, "dropout_reads_logged": 0, "inventory": {"size": 4975, "admitted": 0, "staged": 4975, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 2006, "admitted_split": 0, "staged_split": 1198, "admitted_merge": 0, "staged_merge": 852, "admitted_skel": 0, "staged_skel": 919, "mean_count": 0.0, "n_util_obs": 0, "util_count_spearman": NaN}, "probe_overhead_sec_total": 0.129, "probe_overhead_seq_frac": 0.0, "alpha": 0.03, "t": 11.9}
{"event": "probe", "step": 1, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 10906.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.626674, "cps": 85.203125, "mean_p": 0.513143, "boundary_entropy": 0.686304, "bpb": 8.094369}, "tr": {"chunks": 8206.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.453295, "cps": 64.109375, "mean_p": 0.498082, "boundary_entropy": 0.686828, "bpb": 8.076161}, "fi": {"chunks": 13505.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.717931, "cps": 105.507812, "mean_p": 0.517038, "boundary_entropy": 0.686672, "bpb": 8.036118}}, "bpb": {"en": 8.09437, "tr": 8.07616, "fi": 8.03612}, "gini_cpc": 0.098128, "gini_cps": 0.108308, "bpic": 1.739338, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.7233351802878608, "fi": 1.1456219654381847}, "inventory": {"size": 4975, "admitted": 0, "staged": 4975, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 2006, "admitted_split": 0, "staged_split": 1198, "admitted_merge": 0, "staged_merge": 852, "admitted_skel": 0, "staged_skel": 919, "mean_count": 0.0, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 14.0}
{"event": "probe", "step": 2, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 10880.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.62518, "cps": 85.0, "mean_p": 0.512978, "boundary_entropy": 0.686318, "bpb": 8.031168}, "tr": {"chunks": 8246.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.455505, "cps": 64.421875, "mean_p": 0.498244, "boundary_entropy": 0.686832, "bpb": 8.031929}, "fi": {"chunks": 13467.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.715911, "cps": 105.210938, "mean_p": 0.516798, "boundary_entropy": 0.686692, "bpb": 7.976287}}, "bpb": {"en": 8.03117, "tr": 8.03193, "fi": 7.97629}, "gini_cpc": 0.09663, "gini_cps": 0.106792, "bpic": 1.740619, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.7285980488284359, "fi": 1.1451284419394785}, "inventory": {"size": 8933, "admitted": 0, "staged": 8933, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 3533, "admitted_split": 0, "staged_split": 2211, "admitted_merge": 0, "staged_merge": 1438, "admitted_skel": 0, "staged_skel": 1751, "mean_count": 0.0, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 19.2}
{"event": "probe", "step": 4, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 10718.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.615871, "cps": 83.734375, "mean_p": 0.512102, "boundary_entropy": 0.68638, "bpb": 7.819679}, "tr": {"chunks": 8385.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.463183, "cps": 65.507812, "mean_p": 0.498671, "boundary_entropy": 0.686846, "bpb": 7.878763}, "fi": {"chunks": 13216.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.702568, "cps": 103.25, "mean_p": 0.515685, "boundary_entropy": 0.686774, "bpb": 7.773584}}, "bpb": {"en": 7.81968, "tr": 7.87876, "fi": 7.77358}, "gini_cpc": 0.089576, "gini_cps": 0.099652, "bpic": 1.755376, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.7520779969673853, "fi": 1.1407711096847934}, "inventory": {"size": 15209, "admitted": 0, "staged": 15209, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4, "admitted_surface": 0, "staged_surface": 5985, "admitted_split": 0, "staged_split": 3775, "admitted_merge": 0, "staged_merge": 2529, "admitted_skel": 0, "staged_skel": 2920, "mean_count": 0.0, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 25.2}
{"event": "probe", "step": 6, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 10325.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.593289, "cps": 80.664062, "mean_p": 0.510353, "boundary_entropy": 0.686483, "bpb": 7.54287}, "tr": {"chunks": 8507.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.469922, "cps": 66.460938, "mean_p": 0.499189, "boundary_entropy": 0.686869, "bpb": 7.674385}, "fi": {"chunks": 12733.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.676891, "cps": 99.476562, "mean_p": 0.513644, "boundary_entropy": 0.686908, "bpb": 7.518197}}, "bpb": {"en": 7.54287, "tr": 7.67438, "fi": 7.5182}, "gini_cpc": 0.079294, "gini_cps": 0.089255, "bpic": 1.797307, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.7920633918983102, "fi": 1.140914016231037}, "inventory": {"size": 19899, "admitted": 0, "staged": 19899, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 7, "admitted_surface": 0, "staged_surface": 7725, "admitted_split": 0, "staged_split": 4777, "admitted_merge": 0, "staged_merge": 3607, "admitted_skel": 0, "staged_skel": 3790, "mean_count": 0.0, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 31.7}
{"event": "probe", "step": 8, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 9631.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.55341, "cps": 75.242188, "mean_p": 0.507286, "boundary_entropy": 0.686614, "bpb": 7.293105}, "tr": {"chunks": 8636.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.477048, "cps": 67.46875, "mean_p": 0.499597, "boundary_entropy": 0.686902, "bpb": 7.478109}, "fi": {"chunks": 11871.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.631067, "cps": 92.742188, "mean_p": 0.510289, "boundary_entropy": 0.687081, "bpb": 7.295817}}, "bpb": {"en": 7.29311, "tr": 7.47811, "fi": 7.29582}, "gini_cpc": 0.061798, "gini_cps": 0.07156, "bpic": 1.882408, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.8620149930261254, "fi": 1.140323721739613}, "inventory": {"size": 25371, "admitted": 0, "staged": 25371, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 9, "admitted_surface": 0, "staged_surface": 9827, "admitted_split": 0, "staged_split": 6082, "admitted_merge": 0, "staged_merge": 4562, "admitted_skel": 0, "staged_skel": 4900, "mean_count": 0.0, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 39.6}
{"event": "probe", "step": 16, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3751.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.215538, "cps": 29.304688, "mean_p": 0.484382, "boundary_entropy": 0.68616, "bpb": 6.724684}, "tr": {"chunks": 7275.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.401867, "cps": 56.835938, "mean_p": 0.494376, "boundary_entropy": 0.686996, "bpb": 7.013467}, "fi": {"chunks": 4586.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.243794, "cps": 35.828125, "mean_p": 0.486535, "boundary_entropy": 0.686903, "bpb": 6.787277}}, "bpb": {"en": 6.72468, "tr": 7.01347, "fi": 6.78728}, "gini_cpc": 0.144241, "gini_cps": 0.150483, "bpic": 3.633871, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.8644876124510015, "fi": 1.1310953656913647}, "inventory": {"size": 63026, "admitted": 0, "staged": 63026, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 23, "admitted_surface": 0, "staged_surface": 23789, "admitted_split": 0, "staged_split": 17667, "admitted_merge": 0, "staged_merge": 6762, "admitted_skel": 0, "staged_skel": 14808, "mean_count": 0.0, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 49.5}
{"event": "probe", "step": 32, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 2600.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.1494, "cps": 20.3125, "mean_p": 0.486736, "boundary_entropy": 0.686947, "bpb": 6.304668}, "tr": {"chunks": 4647.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.256698, "cps": 36.304688, "mean_p": 0.490191, "boundary_entropy": 0.687561, "bpb": 6.651782}, "fi": {"chunks": 1398.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.074318, "cps": 10.921875, "mean_p": 0.479423, "boundary_entropy": 0.686975, "bpb": 6.42079}}, "bpb": {"en": 6.30467, "tr": 6.65178, "fi": 6.42079}, "gini_cpc": 0.253086, "gini_cps": 0.250549, "bpic": 6.562406, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.7181967502198956, "fi": 0.497446134217704}, "inventory": {"size": 138957, "admitted": 0, "staged": 138957, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 47, "admitted_surface": 0, "staged_surface": 50336, "admitted_split": 0, "staged_split": 43611, "admitted_merge": 0, "staged_merge": 8727, "admitted_skel": 0, "staged_skel": 36283, "mean_count": 0.0, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 62.3}
{"event": "consolidate", "step": 50, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 65536, "swapped": 0, "uncertain_staged": 230661, "t": 81.9}
{"event": "train", "step": 50, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 4.21235, "bpb": 6.07714, "chunk_loss": 0.034219, "F": 0.22595, "G": 0.48851, "chunks": 231.38, "bytes_per_chunk": 4.426, "boundary_entropy": 0.69161, "router_noise": 0.0, "grad_norm": 2.4788, "lr": 5.1e-05, "router_trainable": true, "sec_per_step": 1.428, "probe_sec": 0.2083, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.1533, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": 0.00706, "n_shadow": 4, "mean_delta_shadow": -0.0096, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.00623, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.79924, "trunc_abs_frac": 0.49714, "mean_delta_full": 0.00403, "read_hit_rate": {"surface": 0.0, "split": 0.08213, "merge": 0.05665, "skel": 0.0}, "mean_gate": {"surface": null, "split": 0.11899, "merge": 0.11903, "skel": null}, "n_reads": 1174, "n_cand": 27533, "read_suppressed_frac": 0.1031, "dropout_reads_logged": 1174, "inventory": {"size": 236243, "admitted": 65536, "staged": 170707, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 83, "admitted_surface": 525, "staged_surface": 83912, "admitted_split": 61267, "staged_split": 15604, "admitted_merge": 3744, "staged_merge": 6690, "admitted_skel": 0, "staged_skel": 64501, "mean_count": 1.6729888916015625, "n_util_obs": 0, "util_count_spearman": NaN}, "probe_overhead_sec_total": 6.543, "probe_overhead_seq_frac": 0.4902, "alpha": 0.03, "t": 83.3}
{"event": "probe", "step": 64, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 4792.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.275355, "cps": 37.4375, "mean_p": 0.485065, "boundary_entropy": 0.6855, "bpb": 5.552331}, "tr": {"chunks": 5377.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.297023, "cps": 42.007812, "mean_p": 0.486514, "boundary_entropy": 0.686478, "bpb": 5.95655}, "fi": {"chunks": 4700.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.249854, "cps": 36.71875, "mean_p": 0.483987, "boundary_entropy": 0.68633, "bpb": 5.717869}}, "bpb": {"en": 5.55233, "tr": 5.95655, "fi": 5.71787}, "gini_cpc": 0.038245, "gini_cps": 0.030354, "bpic": 3.815455, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.078690355789174, "fi": 0.9073885302609057}, "inventory": {"size": 308254, "admitted": 65536, "staged": 242718, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 89, "admitted_surface": 525, "staged_surface": 110385, "admitted_split": 61267, "staged_split": 38815, "admitted_merge": 3744, "staged_merge": 8168, "admitted_skel": 0, "staged_skel": 85350, "mean_count": 1.85919189453125, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 93.9}
{"event": "consolidate", "step": 100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 262134, "t": 133.1}
{"event": "train", "step": 100, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 3.24873, "bpb": 4.68693, "chunk_loss": 0.029599, "F": 0.16, "G": 0.45223, "chunks": 163.84, "bytes_per_chunk": 6.25, "boundary_entropy": 0.68184, "router_noise": 0.0, "grad_norm": 1.5584, "lr": 0.000101, "router_trainable": true, "sec_per_step": 1.015, "probe_sec": 0.1617, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10451, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.0092, "n_shadow": 4, "mean_delta_shadow": -0.00168, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.00745, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -4.09756, "trunc_abs_frac": 7.89362, "mean_delta_full": -0.005, "read_hit_rate": {"surface": 0.0, "split": 0.05948, "merge": 0.02416, "skel": 0.0}, "mean_gate": {"surface": null, "split": 0.11916, "merge": 0.11907, "skel": null}, "n_reads": 649, "n_cand": 22574, "read_suppressed_frac": 0.1032, "dropout_reads_logged": 649, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 118, "admitted_surface": 525, "staged_surface": 99819, "admitted_split": 61267, "staged_split": 76167, "admitted_merge": 3744, "staged_merge": 11601, "admitted_skel": 0, "staged_skel": 74557, "mean_count": 2.284637451171875, "n_util_obs": 0, "util_count_spearman": NaN}, "probe_overhead_sec_total": 13.179, "probe_overhead_seq_frac": 0.495, "alpha": 0.03, "t": 134.0}
{"event": "probe", "step": 128, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3858.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.221686, "cps": 30.140625, "mean_p": 0.417012, "boundary_entropy": 0.645049, "bpb": 4.167089}, "tr": {"chunks": 4820.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.266254, "cps": 37.65625, "mean_p": 0.436035, "boundary_entropy": 0.658845, "bpb": 4.398175}, "fi": {"chunks": 3844.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.204349, "cps": 30.03125, "mean_p": 0.428644, "boundary_entropy": 0.659666, "bpb": 4.352004}}, "bpb": {"en": 4.16709, "tr": 4.39817, "fi": 4.352}, "gini_cpc": 0.059615, "gini_cps": 0.051962, "bpic": 4.530586, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.2010425224452201, "fi": 0.9217929716349249}, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 137, "admitted_surface": 525, "staged_surface": 101248, "admitted_split": 61267, "staged_split": 74536, "admitted_merge": 3744, "staged_merge": 14661, "admitted_skel": 0, "staged_skel": 71699, "mean_count": 2.597412109375, "n_util_obs": 1, "util_count_spearman": NaN}, "t": 158.9}
{"event": "consolidate", "step": 150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1, "uncertain_staged": 262133, "t": 182.5}
{"event": "train", "step": 150, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.79952, "bpb": 4.03885, "chunk_loss": 0.032508, "F": 0.21716, "G": 0.39462, "chunks": 222.38, "bytes_per_chunk": 4.605, "boundary_entropy": 0.62637, "router_noise": 0.0, "grad_norm": 2.273, "lr": 0.000151, "router_trainable": true, "sec_per_step": 0.98, "probe_sec": 0.1024, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05824, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.01533, "n_shadow": 6, "mean_delta_shadow": 0.01346, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.03076, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.71765, "trunc_abs_frac": 0.71765, "mean_delta_full": 0.02075, "read_hit_rate": {"surface": 0.0, "split": 0.05905, "merge": 0.02836, "skel": 0.0}, "mean_gate": {"surface": null, "split": 0.11979, "merge": 0.11977, "skel": null}, "n_reads": 826, "n_cand": 28892, "read_suppressed_frac": 0.0847, "dropout_reads_logged": 826, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 157, "admitted_surface": 525, "staged_surface": 100553, "admitted_split": 61267, "staged_split": 72779, "admitted_merge": 3744, "staged_merge": 17778, "admitted_skel": 0, "staged_skel": 71034, "mean_count": 3.16583251953125, "n_util_obs": 2, "util_count_spearman": NaN}, "probe_overhead_sec_total": 21.832, "probe_overhead_seq_frac": 0.4967, "alpha": 0.03, "t": 183.1}
{"event": "consolidate", "step": 200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1, "uncertain_staged": 262133, "t": 225.3}
{"event": "probe", "step": 200, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3241.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.186232, "cps": 25.320312, "mean_p": 0.364215, "boundary_entropy": 0.581275, "bpb": 3.645697}, "tr": {"chunks": 4002.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.221068, "cps": 31.265625, "mean_p": 0.369271, "boundary_entropy": 0.582976, "bpb": 3.804344}, "fi": {"chunks": 3856.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.204986, "cps": 30.125, "mean_p": 0.398499, "boundary_entropy": 0.626574, "bpb": 3.87609}}, "bpb": {"en": 3.6457, "tr": 3.80434, "fi": 3.87609}, "gini_cpc": 0.03793, "gini_cps": 0.04571, "bpic": 5.111451, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.1870571330295976, "fi": 1.10070320478389}, "inventory": {"size": 327679, "admitted": 65536, "staged": 262143, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 196, "admitted_surface": 525, "staged_surface": 97610, "admitted_split": 61267, "staged_split": 73262, "admitted_merge": 3744, "staged_merge": 22807, "admitted_skel": 0, "staged_skel": 68464, "mean_count": 3.290557861328125, "n_util_obs": 2, "util_count_spearman": NaN}, "t": 227.5}
{"event": "eval", "step": 200, "phase": 1, "bpb": {"en": 3.65319, "tr": 3.81528, "fi": 3.87416}, "bytes_per_chunk": {"en": 6.11, "tr": 5.712, "fi": 5.451}, "bpb_mean": 3.78088, "frozen_inventory": true, "t": 230.9}
{"event": "train", "step": 200, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.53703, "bpb": 3.66016, "chunk_loss": 0.029634, "F": 0.15717, "G": 0.35433, "chunks": 160.94, "bytes_per_chunk": 6.363, "boundary_entropy": 0.58196, "router_noise": 0.0, "grad_norm": 1.7454, "lr": 0.000201, "router_trainable": true, "sec_per_step": 1.098, "probe_sec": 0.3618, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.22246, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.00085, "n_shadow": 2, "mean_delta_shadow": 0.02169, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.05786, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 11.63086, "trunc_abs_frac": 0.5975, "mean_delta_full": -0.00195, "read_hit_rate": {"surface": 0.0, "split": 0.06405, "merge": 0.00787, "skel": 0.0}, "mean_gate": {"surface": null, "split": 0.12109, "merge": 0.1211, "skel": null}, "n_reads": 588, "n_cand": 22047, "read_suppressed_frac": 0.1054, "dropout_reads_logged": 588, "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": 525, "staged_surface": 97451, "admitted_split": 61267, "staged_split": 73853, "admitted_merge": 3744, "staged_merge": 22150, "admitted_skel": 0, "staged_skel": 68690, "mean_count": 3.299530029296875, "n_util_obs": 3, "util_count_spearman": 0.5}, "probe_overhead_sec_total": 28.676, "probe_overhead_seq_frac": 0.4975, "alpha": 0.03, "t": 237.9}
{"event": "consolidate", "step": 250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262130, "t": 277.6}
{"event": "train", "step": 250, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.38918, "bpb": 3.44685, "chunk_loss": 0.030915, "F": 0.18933, "G": 0.34137, "chunks": 193.88, "bytes_per_chunk": 5.282, "boundary_entropy": 0.54488, "router_noise": 0.0, "grad_norm": 2.0996, "lr": 0.000251, "router_trainable": true, "sec_per_step": 0.801, "probe_sec": 0.0794, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04747, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.0033, "n_shadow": 6, "mean_delta_shadow": 0.00018, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.05652, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -0.05223, "trunc_abs_frac": 1.65592, "mean_delta_full": -0.05142, "read_hit_rate": {"surface": 0.00016, "split": 0.06997, "merge": 0.01042, "skel": 0.01834}, "mean_gate": {"surface": 0.12096, "split": 0.12221, "merge": 0.12226, "skel": 0.12115}, "n_reads": 850, "n_cand": 26628, "read_suppressed_frac": 0.1059, "dropout_reads_logged": 850, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 118, "admitted_surface": 526, "staged_surface": 93434, "admitted_split": 60452, "staged_split": 77696, "admitted_merge": 3744, "staged_merge": 26493, "admitted_skel": 814, "staged_skel": 64521, "mean_count": 4.7662506103515625, "n_util_obs": 8, "util_count_spearman": 0.2857142857142857}, "probe_overhead_sec_total": 34.907, "probe_overhead_seq_frac": 0.498, "alpha": 0.03, "t": 278.0}
{"event": "probe", "step": 256, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3247.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.186577, "cps": 25.367188, "mean_p": 0.342724, "boundary_entropy": 0.537399, "bpb": 3.352768}, "tr": {"chunks": 4193.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.231619, "cps": 32.757812, "mean_p": 0.323201, "boundary_entropy": 0.521026, "bpb": 3.511094}, "fi": {"chunks": 4228.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.224762, "cps": 33.03125, "mean_p": 0.368437, "boundary_entropy": 0.582335, "bpb": 3.660153}}, "bpb": {"en": 3.35277, "tr": 3.51109, "fi": 3.66015}, "gini_cpc": 0.046703, "gini_cps": 0.056051, "bpic": 4.862187, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.2414125814401427, "fi": 1.2046612112574866}, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 120, "admitted_surface": 526, "staged_surface": 93324, "admitted_split": 60452, "staged_split": 77307, "admitted_merge": 3744, "staged_merge": 27199, "admitted_skel": 814, "staged_skel": 64314, "mean_count": 4.831756591796875, "n_util_obs": 8, "util_count_spearman": 0.23809523809523808}, "t": 284.7}
{"event": "consolidate", "step": 300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262129, "t": 328.9}
{"event": "train", "step": 300, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.28376, "bpb": 3.29477, "chunk_loss": 0.030724, "F": 0.1893, "G": 0.30646, "chunks": 193.84, "bytes_per_chunk": 5.283, "boundary_entropy": 0.47423, "router_noise": 0.0, "grad_norm": 1.9289, "lr": 0.0003, "router_trainable": true, "sec_per_step": 1.043, "probe_sec": 0.1816, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.1296, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": 0.00012, "n_shadow": 4, "mean_delta_shadow": 0.04973, "n_split": 0, "mean_delta_split": NaN, "n_merge": 3, "mean_delta_merge": 0.00024, "n_dropped_unpaired": 0, "n_full": 0, "read_hit_rate": {"surface": 0.00681, "split": 0.13981, "merge": 0.01141, "skel": 0.01763}, "mean_gate": {"surface": 0.12013, "split": 0.12225, "merge": 0.12215, "skel": 0.12205}, "n_reads": 1667, "n_cand": 27046, "read_suppressed_frac": 0.1002, "dropout_reads_logged": 1667, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 142, "admitted_surface": 505, "staged_surface": 91827, "admitted_split": 60473, "staged_split": 72958, "admitted_merge": 3744, "staged_merge": 34474, "admitted_skel": 814, "staged_skel": 62885, "mean_count": 6.99078369140625, "n_util_obs": 17, "util_count_spearman": 0.27941176470588236}, "probe_overhead_sec_total": 41.667, "probe_overhead_seq_frac": 0.4983, "alpha": 0.03, "t": 330.2}
{"event": "consolidate", "step": 350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262128, "t": 373.7}
{"event": "train", "step": 350, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.17402, "bpb": 3.13645, "chunk_loss": 0.030858, "F": 0.19907, "G": 0.28573, "chunks": 203.84, "bytes_per_chunk": 5.023, "boundary_entropy": 0.42737, "router_noise": 0.0, "grad_norm": 1.1066, "lr": 0.0002999, "router_trainable": true, "sec_per_step": 0.89, "probe_sec": 0.1642, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09689, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": -0.00029, "n_shadow": 5, "mean_delta_shadow": 0.00936, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.40496, "trunc_abs_frac": 0.40496, "mean_delta_full": 0.15991, "read_hit_rate": {"surface": 0.00863, "split": 0.1164, "merge": 0.01208, "skel": 0.01638}, "mean_gate": {"surface": 0.12034, "split": 0.12802, "merge": 0.12761, "skel": 0.12838}, "n_reads": 1525, "n_cand": 28467, "read_suppressed_frac": 0.1082, "dropout_reads_logged": 1525, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 147, "admitted_surface": 500, "staged_surface": 88420, "admitted_split": 59596, "staged_split": 67366, "admitted_merge": 3963, "staged_merge": 46577, "admitted_skel": 1477, "staged_skel": 59781, "mean_count": 7.2101898193359375, "n_util_obs": 15, "util_count_spearman": 0.22142857142857142}, "probe_overhead_sec_total": 48.498, "probe_overhead_seq_frac": 0.4986, "alpha": 0.03, "t": 374.7}
{"event": "consolidate", "step": 400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262125, "t": 419.6}
{"event": "probe", "step": 400, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3355.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.192783, "cps": 26.210938, "mean_p": 0.269698, "boundary_entropy": 0.382677, "bpb": 2.910234}, "tr": {"chunks": 3273.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.180799, "cps": 25.570312, "mean_p": 0.271021, "boundary_entropy": 0.426751, "bpb": 3.196457}, "fi": {"chunks": 3992.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.212216, "cps": 31.1875, "mean_p": 0.281387, "boundary_entropy": 0.441397, "bpb": 3.39001}}, "bpb": {"en": 2.91023, "tr": 3.19646, "fi": 3.39001}, "gini_cpc": 0.035755, "gini_cps": 0.045135, "bpic": 5.341996, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9378363237421106, "fi": 1.1008046232263662}, "inventory": {"size": 326656, "admitted": 65536, "staged": 261120, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 154, "admitted_surface": 506, "staged_surface": 84016, "admitted_split": 58644, "staged_split": 63142, "admitted_merge": 3892, "staged_merge": 57687, "admitted_skel": 2494, "staged_skel": 56275, "mean_count": 6.6326904296875, "n_util_obs": 15, "util_count_spearman": 0.21785714285714286}, "t": 422.3}
{"event": "eval", "step": 400, "phase": 1, "bpb": {"en": 2.92252, "tr": 3.20329, "fi": 3.3811}, "bytes_per_chunk": {"en": 5.233, "tr": 6.429, "fi": 5.306}, "bpb_mean": 3.16897, "frozen_inventory": true, "t": 426.3}
{"event": "train", "step": 400, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.09343, "bpb": 3.02018, "chunk_loss": 0.03031, "F": 0.17947, "G": 0.26565, "chunks": 183.78, "bytes_per_chunk": 5.572, "boundary_entropy": 0.39074, "router_noise": 0.0, "grad_norm": 1.3156, "lr": 0.0002998, "router_trainable": true, "sec_per_step": 1.17, "probe_sec": 0.1167, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0469, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": -0.00804, "n_shadow": 5, "mean_delta_shadow": 0.02741, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -8.04167, "trunc_abs_frac": 0.22017, "mean_delta_full": -0.00269, "read_hit_rate": {"surface": 0.00684, "split": 0.10152, "merge": 0.00739, "skel": 0.01738}, "mean_gate": {"surface": 0.13056, "split": 0.13162, "merge": 0.13498, "skel": 0.13142}, "n_reads": 1202, "n_cand": 26217, "read_suppressed_frac": 0.0824, "dropout_reads_logged": 1202, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 153, "admitted_surface": 506, "staged_surface": 84254, "admitted_split": 58644, "staged_split": 63351, "admitted_merge": 3892, "staged_merge": 57958, "admitted_skel": 2494, "staged_skel": 56581, "mean_count": 6.651031494140625, "n_util_obs": 15, "util_count_spearman": 0.18928571428571428}, "probe_overhead_sec_total": 55.575, "probe_overhead_seq_frac": 0.4988, "alpha": 0.03, "t": 433.2}
{"event": "consolidate", "step": 450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262118, "t": 473.7}
{"event": "train", "step": 450, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.01085, "bpb": 2.90104, "chunk_loss": 0.030982, "F": 0.20999, "G": 0.26607, "chunks": 215.03, "bytes_per_chunk": 4.762, "boundary_entropy": 0.35159, "router_noise": 0.0, "grad_norm": 1.8337, "lr": 0.0002995, "router_trainable": true, "sec_per_step": 0.82, "probe_sec": 0.0913, "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": 9, "mean_delta_loo": -0.05553, "n_shadow": 5, "mean_delta_shadow": 0.02521, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 1.06024, "trunc_abs_frac": 1.96186, "mean_delta_full": 0.0966, "read_hit_rate": {"surface": 0.00307, "split": 0.10399, "merge": 0.01423, "skel": 0.01257}, "mean_gate": {"surface": 0.14492, "split": 0.14374, "merge": 0.1427, "skel": 0.13937}, "n_reads": 1410, "n_cand": 30427, "read_suppressed_frac": 0.1071, "dropout_reads_logged": 1410, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 125, "admitted_surface": 1528, "staged_surface": 78227, "admitted_split": 58208, "staged_split": 58641, "admitted_merge": 3862, "staged_merge": 69670, "admitted_skel": 1938, "staged_skel": 55606, "mean_count": 8.111129760742188, "n_util_obs": 20, "util_count_spearman": 0.16691729323308271}, "probe_overhead_sec_total": 61.787, "probe_overhead_seq_frac": 0.4989, "alpha": 0.03, "t": 474.2}
{"event": "consolidate", "step": 500, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262117, "t": 518.0}
{"event": "train", "step": 500, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.91021, "bpb": 2.75585, "chunk_loss": 0.031014, "F": 0.21344, "G": 0.2609, "chunks": 218.56, "bytes_per_chunk": 4.685, "boundary_entropy": 0.32439, "router_noise": 0.0, "grad_norm": 1.3557, "lr": 0.0002992, "router_trainable": true, "sec_per_step": 0.906, "probe_sec": 0.1587, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06114, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.00921, "n_shadow": 3, "mean_delta_shadow": -0.13609, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.06274, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.15597, "trunc_abs_frac": 0.2347, "mean_delta_full": -0.38428, "read_hit_rate": {"surface": 0.26329, "split": 0.10977, "merge": 0.00952, "skel": 0.01786}, "mean_gate": {"surface": 0.14723, "split": 0.14517, "merge": 0.1451, "skel": 0.14218}, "n_reads": 3333, "n_cand": 31315, "read_suppressed_frac": 0.0957, "dropout_reads_logged": 3333, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 127, "admitted_surface": 2550, "staged_surface": 74354, "admitted_split": 57200, "staged_split": 54396, "admitted_merge": 3862, "staged_merge": 79592, "admitted_skel": 1924, "staged_skel": 53802, "mean_count": 13.663482666015625, "n_util_obs": 23, "util_count_spearman": 0.11956521739130435}, "probe_overhead_sec_total": 69.019, "probe_overhead_seq_frac": 0.499, "alpha": 0.03, "t": 519.5}
{"event": "probe", "step": 512, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3440.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.197667, "cps": 26.875, "mean_p": 0.258092, "boundary_entropy": 0.33448, "bpb": 2.717594}, "tr": {"chunks": 3575.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.197481, "cps": 27.929688, "mean_p": 0.267365, "boundary_entropy": 0.379831, "bpb": 3.07943}, "fi": {"chunks": 4481.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.238212, "cps": 35.007812, "mean_p": 0.266421, "boundary_entropy": 0.376645, "bpb": 3.230345}}, "bpb": {"en": 2.71759, "tr": 3.07943, "fi": 3.23034}, "gini_cpc": 0.042873, "gini_cps": 0.060369, "bpic": 4.934934, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.999059082459672, "fi": 1.2051156825607974}, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 145, "admitted_surface": 2550, "staged_surface": 73602, "admitted_split": 57200, "staged_split": 53951, "admitted_merge": 3862, "staged_merge": 81294, "admitted_skel": 1924, "staged_skel": 53297, "mean_count": 14.142013549804688, "n_util_obs": 26, "util_count_spearman": 0.174017094017094}, "t": 530.9}
{"event": "consolidate", "step": 550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262111, "t": 573.8}
{"event": "train", "step": 550, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.90138, "bpb": 2.74312, "chunk_loss": 0.030685, "F": 0.19934, "G": 0.25602, "chunks": 204.12, "bytes_per_chunk": 5.017, "boundary_entropy": 0.31504, "router_noise": 0.0, "grad_norm": 1.3481, "lr": 0.0002987, "router_trainable": true, "sec_per_step": 1.102, "probe_sec": 0.1617, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08124, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 3, "mean_delta_loo": 0.24038, "n_shadow": 10, "mean_delta_shadow": -0.07114, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.06006, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.29571, "trunc_abs_frac": 0.29467, "mean_delta_full": -0.16058, "read_hit_rate": {"surface": 0.21385, "split": 0.1314, "merge": 0.01113, "skel": 0.02018}, "mean_gate": {"surface": 0.14331, "split": 0.13972, "merge": 0.14124, "skel": 0.13832}, "n_reads": 3072, "n_cand": 29416, "read_suppressed_frac": 0.1019, "dropout_reads_logged": 3072, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 170, "admitted_surface": 3571, "staged_surface": 70542, "admitted_split": 56197, "staged_split": 52620, "admitted_merge": 3862, "staged_merge": 86236, "admitted_skel": 1906, "staged_skel": 52746, "mean_count": 16.675094604492188, "n_util_obs": 31, "util_count_spearman": 0.14717741935483872}, "probe_overhead_sec_total": 75.988, "probe_overhead_seq_frac": 0.4991, "alpha": 0.03, "t": 574.6}
{"event": "consolidate", "step": 600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262107, "t": 622.5}
{"event": "probe", "step": 600, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3531.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.202896, "cps": 27.585938, "mean_p": 0.254698, "boundary_entropy": 0.300854, "bpb": 2.593318}, "tr": {"chunks": 3790.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.209358, "cps": 29.609375, "mean_p": 0.253498, "boundary_entropy": 0.317144, "bpb": 2.927066}, "fi": {"chunks": 4249.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.225878, "cps": 33.195312, "mean_p": 0.267365, "boundary_entropy": 0.3387, "bpb": 3.141858}}, "bpb": {"en": 2.59332, "tr": 2.92707, "fi": 3.14186}, "gini_cpc": 0.02401, "gini_cps": 0.041371, "bpic": 4.903371, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0318464187110938, "fi": 1.1132719078711109}, "inventory": {"size": 326656, "admitted": 65536, "staged": 261120, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 227, "admitted_surface": 4574, "staged_surface": 67825, "admitted_split": 56054, "staged_split": 50488, "admitted_merge": 3869, "staged_merge": 91884, "admitted_skel": 1039, "staged_skel": 50923, "mean_count": 18.907638549804688, "n_util_obs": 38, "util_count_spearman": 0.19312835102308787}, "t": 624.7}
{"event": "eval", "step": 600, "phase": 1, "bpb": {"en": 2.60467, "tr": 2.93362, "fi": 3.14372}, "bytes_per_chunk": {"en": 5.055, "tr": 5.392, "fi": 4.882}, "bpb_mean": 2.89401, "frozen_inventory": true, "t": 629.3}
{"event": "train", "step": 600, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.88964, "bpb": 2.72617, "chunk_loss": 0.030548, "F": 0.19394, "G": 0.2491, "chunks": 198.59, "bytes_per_chunk": 5.156, "boundary_entropy": 0.30103, "router_noise": 0.0, "grad_norm": 1.1154, "lr": 0.0002982, "router_trainable": true, "sec_per_step": 1.245, "probe_sec": 0.3296, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.1796, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.02848, "n_shadow": 8, "mean_delta_shadow": -0.00702, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 23.25864, "trunc_abs_frac": 0.86804, "mean_delta_full": -0.005, "read_hit_rate": {"surface": 0.19484, "split": 0.12952, "merge": 0.00954, "skel": 0.01504}, "mean_gate": {"surface": 0.15125, "split": 0.14677, "merge": 0.14108, "skel": 0.14785}, "n_reads": 2823, "n_cand": 28901, "read_suppressed_frac": 0.1091, "dropout_reads_logged": 2823, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 227, "admitted_surface": 4574, "staged_surface": 68185, "admitted_split": 56054, "staged_split": 50627, "admitted_merge": 3869, "staged_merge": 92191, "admitted_skel": 1039, "staged_skel": 51141, "mean_count": 18.950729370117188, "n_util_obs": 38, "util_count_spearman": 0.17387022650180545}, "probe_overhead_sec_total": 83.513, "probe_overhead_seq_frac": 0.4992, "alpha": 0.03, "t": 636.8}
{"event": "consolidate", "step": 650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262106, "t": 691.9}
{"event": "train", "step": 650, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.84, "bpb": 2.65457, "chunk_loss": 0.030622, "F": 0.19745, "G": 0.2512, "chunks": 202.19, "bytes_per_chunk": 5.065, "boundary_entropy": 0.28943, "router_noise": 0.0, "grad_norm": 1.2254, "lr": 0.0002975, "router_trainable": true, "sec_per_step": 1.114, "probe_sec": 0.095, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05703, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": 0.14906, "n_shadow": 7, "mean_delta_shadow": 0.02148, "n_split": 2, "mean_delta_split": -0.058, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.01005, "trunc_abs_frac": 0.06509, "mean_delta_full": -0.20251, "read_hit_rate": {"surface": 0.22025, "split": 0.1324, "merge": 0.00968, "skel": 0.01956}, "mean_gate": {"surface": 0.16146, "split": 0.15214, "merge": 0.15525, "skel": 0.1534}, "n_reads": 3064, "n_cand": 28946, "read_suppressed_frac": 0.0924, "dropout_reads_logged": 3064, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 209, "admitted_surface": 5287, "staged_surface": 66119, "admitted_split": 55394, "staged_split": 49132, "admitted_merge": 3869, "staged_merge": 96518, "admitted_skel": 986, "staged_skel": 50375, "mean_count": 21.430938720703125, "n_util_obs": 45, "util_count_spearman": -0.05059288537549407}, "probe_overhead_sec_total": 91.53, "probe_overhead_seq_frac": 0.4992, "alpha": 0.03, "t": 692.5}
{"event": "consolidate", "step": 700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262102, "t": 743.0}
{"event": "train", "step": 700, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.9905, "bpb": 2.87169, "chunk_loss": 0.030799, "F": 0.20209, "G": 0.25559, "chunks": 206.94, "bytes_per_chunk": 4.948, "boundary_entropy": 0.2894, "router_noise": 0.0, "grad_norm": 1.1134, "lr": 0.0002967, "router_trainable": true, "sec_per_step": 1.045, "probe_sec": 0.1358, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10187, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.10721, "n_shadow": 8, "mean_delta_shadow": -0.00093, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.38913, "trunc_abs_frac": 2.68221, "mean_delta_full": -0.22791, "read_hit_rate": {"surface": 0.17602, "split": 0.10081, "merge": 0.00747, "skel": 0.02222}, "mean_gate": {"surface": 0.16533, "split": 0.15105, "merge": 0.1537, "skel": 0.14203}, "n_reads": 2460, "n_cand": 29347, "read_suppressed_frac": 0.1012, "dropout_reads_logged": 2460, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 244, "admitted_surface": 6306, "staged_surface": 64586, "admitted_split": 54509, "staged_split": 47977, "admitted_merge": 3859, "staged_merge": 99911, "admitted_skel": 862, "staged_skel": 49670, "mean_count": 21.872024536132812, "n_util_obs": 42, "util_count_spearman": 0.14804310833806011}, "probe_overhead_sec_total": 99.557, "probe_overhead_seq_frac": 0.4993, "alpha": 0.03, "t": 744.7}
{"event": "consolidate", "step": 750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262104, "t": 799.1}
{"event": "train", "step": 750, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.78306, "bpb": 2.57241, "chunk_loss": 0.030749, "F": 0.20468, "G": 0.25293, "chunks": 209.59, "bytes_per_chunk": 4.886, "boundary_entropy": 0.26886, "router_noise": 0.0, "grad_norm": 1.1775, "lr": 0.0002959, "router_trainable": true, "sec_per_step": 1.109, "probe_sec": 0.236, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.12014, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.00504, "n_shadow": 4, "mean_delta_shadow": 0.02946, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": 0.01897, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": -0.02859, "trunc_abs_frac": 0.02859, "mean_delta_full": 0.32983, "read_hit_rate": {"surface": 0.18127, "split": 0.10772, "merge": 0.00963, "skel": 0.02737}, "mean_gate": {"surface": 0.174, "split": 0.15664, "merge": 0.15589, "skel": 0.1484}, "n_reads": 2668, "n_cand": 30020, "read_suppressed_frac": 0.0945, "dropout_reads_logged": 2668, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 265, "admitted_surface": 7320, "staged_surface": 63361, "admitted_split": 54469, "staged_split": 46664, "admitted_merge": 3744, "staged_merge": 103097, "admitted_skel": 3, "staged_skel": 49022, "mean_count": 23.218460083007812, "n_util_obs": 49, "util_count_spearman": 0.11051020408163266}, "probe_overhead_sec_total": 106.689, "probe_overhead_seq_frac": 0.4993, "alpha": 0.03, "t": 800.2}
{"event": "consolidate", "step": 800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262103, "t": 856.6}
{"event": "probe", "step": 800, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3384.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.194449, "cps": 26.4375, "mean_p": 0.245751, "boundary_entropy": 0.270162, "bpb": 2.439959}, "tr": {"chunks": 3864.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.213445, "cps": 30.1875, "mean_p": 0.244097, "boundary_entropy": 0.290556, "bpb": 2.769892}, "fi": {"chunks": 4161.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.2212, "cps": 32.507812, "mean_p": 0.262522, "boundary_entropy": 0.305015, "bpb": 2.97814}}, "bpb": {"en": 2.43996, "tr": 2.76989, "fi": 2.97814}, "gini_cpc": 0.028349, "gini_cps": 0.045403, "bpic": 4.972566, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0976915780974352, "fi": 1.137573844864424}, "inventory": {"size": 326656, "admitted": 65536, "staged": 261120, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 305, "admitted_surface": 8317, "staged_surface": 61965, "admitted_split": 54093, "staged_split": 45894, "admitted_merge": 3122, "staged_merge": 104722, "admitted_skel": 4, "staged_skel": 48539, "mean_count": 26.177001953125, "n_util_obs": 54, "util_count_spearman": 0.07840670859538784}, "t": 859.0}
{"event": "eval", "step": 800, "phase": 1, "bpb": {"en": 2.44849, "tr": 2.77831, "fi": 2.97985}, "bytes_per_chunk": {"en": 5.155, "tr": 5.24, "fi": 4.97}, "bpb_mean": 2.73555, "frozen_inventory": true, "t": 862.6}
{"event": "train", "step": 800, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.86284, "bpb": 2.6875, "chunk_loss": 0.030668, "F": 0.19861, "G": 0.24867, "chunks": 203.38, "bytes_per_chunk": 5.035, "boundary_entropy": 0.27074, "router_noise": 0.0, "grad_norm": 1.2333, "lr": 0.0002949, "router_trainable": true, "sec_per_step": 1.423, "probe_sec": 0.4127, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.23778, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": 0.00525, "n_shadow": 4, "mean_delta_shadow": 0.10986, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.2368, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 1.8057, "trunc_abs_frac": 2.14373, "mean_delta_full": 0.06301, "read_hit_rate": {"surface": 0.17773, "split": 0.10347, "merge": 0.00698, "skel": 0.04142}, "mean_gate": {"surface": 0.178, "split": 0.15447, "merge": 0.14782, "skel": 0.15084}, "n_reads": 2552, "n_cand": 28906, "read_suppressed_frac": 0.096, "dropout_reads_logged": 2552, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 305, "admitted_surface": 8317, "staged_surface": 62226, "admitted_split": 54093, "staged_split": 46105, "admitted_merge": 3122, "staged_merge": 105133, "admitted_skel": 4, "staged_skel": 48680, "mean_count": 26.2159423828125, "n_util_obs": 54, "util_count_spearman": 0.06186392224128073}, "probe_overhead_sec_total": 114.682, "probe_overhead_seq_frac": 0.4994, "alpha": 0.03, "t": 871.3}
{"event": "consolidate", "step": 850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262101, "t": 923.5}
{"event": "train", "step": 850, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.81064, "bpb": 2.6122, "chunk_loss": 0.030676, "F": 0.20114, "G": 0.24681, "chunks": 205.97, "bytes_per_chunk": 4.972, "boundary_entropy": 0.26299, "router_noise": 0.0, "grad_norm": 1.06, "lr": 0.0002938, "router_trainable": true, "sec_per_step": 1.064, "probe_sec": 0.1046, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06222, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.08431, "n_shadow": 2, "mean_delta_shadow": -0.11005, "n_split": 1, "mean_delta_split": 0.03857, "n_merge": 2, "mean_delta_merge": 0.33789, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 2.17879, "trunc_abs_frac": 3.89653, "mean_delta_full": -0.09786, "read_hit_rate": {"surface": 0.18417, "split": 0.11058, "merge": 0.00935, "skel": 0.0421}, "mean_gate": {"surface": 0.17748, "split": 0.15188, "merge": 0.16351, "skel": 0.14413}, "n_reads": 2779, "n_cand": 29919, "read_suppressed_frac": 0.1026, "dropout_reads_logged": 2779, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 306, "admitted_surface": 8902, "staged_surface": 62153, "admitted_split": 53109, "staged_split": 45072, "admitted_merge": 3521, "staged_merge": 106651, "admitted_skel": 4, "staged_skel": 48268, "mean_count": 28.334564208984375, "n_util_obs": 56, "util_count_spearman": -0.025085440874914558}, "probe_overhead_sec_total": 122.341, "probe_overhead_seq_frac": 0.4994, "alpha": 0.03, "t": 924.5}
{"event": "consolidate", "step": 900, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 1024, "uncertain_staged": 262097, "t": 976.5}
{"event": "train", "step": 900, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.60903, "bpb": 2.32134, "chunk_loss": 0.030555, "F": 0.19742, "G": 0.24163, "chunks": 202.16, "bytes_per_chunk": 5.065, "boundary_entropy": 0.25168, "router_noise": 0.0, "grad_norm": 1.1226, "lr": 0.0002927, "router_trainable": true, "sec_per_step": 1.062, "probe_sec": 0.1879, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08606, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.01147, "n_shadow": 5, "mean_delta_shadow": -0.07491, "n_split": 1, "mean_delta_split": 0.02071, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -0.01495, "trunc_abs_frac": 4.75786, "mean_delta_full": 0.15986, "read_hit_rate": {"surface": 0.2408, "split": 0.14204, "merge": 0.01296, "skel": 0.04931}, "mean_gate": {"surface": 0.18686, "split": 0.15315, "merge": 0.1692, "skel": 0.15043}, "n_reads": 3522, "n_cand": 29464, "read_suppressed_frac": 0.0999, "dropout_reads_logged": 3522, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 315, "admitted_surface": 9622, "staged_surface": 61363, "admitted_split": 52135, "staged_split": 44345, "admitted_merge": 3774, "staged_merge": 108605, "admitted_skel": 5, "staged_skel": 47831, "mean_count": 32.1488037109375, "n_util_obs": 63, "util_count_spearman": 0.061155913978494625}, "probe_overhead_sec_total": 130.741, "probe_overhead_seq_frac": 0.4994, "alpha": 0.03, "t": 977.7}
{"event": "consolidate", "step": 950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262098, "t": 1032.1}
{"event": "train", "step": 950, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.68631, "bpb": 2.43283, "chunk_loss": 0.030662, "F": 0.20267, "G": 0.24388, "chunks": 207.53, "bytes_per_chunk": 4.934, "boundary_entropy": 0.24558, "router_noise": 0.0, "grad_norm": 1.1403, "lr": 0.0002914, "router_trainable": true, "sec_per_step": 1.105, "probe_sec": 0.1283, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06342, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": -0.00861, "n_shadow": 5, "mean_delta_shadow": 0.15808, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.10779, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.52458, "trunc_abs_frac": 0.62619, "mean_delta_full": 0.35824, "read_hit_rate": {"surface": 0.22999, "split": 0.15273, "merge": 0.01384, "skel": 0.03659}, "mean_gate": {"surface": 0.18244, "split": 0.15222, "merge": 0.16679, "skel": 0.14384}, "n_reads": 3551, "n_cand": 29691, "read_suppressed_frac": 0.098, "dropout_reads_logged": 3551, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 322, "admitted_surface": 10150, "staged_surface": 61075, "admitted_split": 51166, "staged_split": 44306, "admitted_merge": 4216, "staged_merge": 109278, "admitted_skel": 4, "staged_skel": 47485, "mean_count": 34.43260192871094, "n_util_obs": 73, "util_count_spearman": 0.10934221893126003}, "probe_overhead_sec_total": 139.415, "probe_overhead_seq_frac": 0.4995, "alpha": 0.03, "t": 1032.9}
{"event": "consolidate", "step": 1000, "retired": 0, "split_retired": 6, "merge_admitted": 0, "admitted": 6, "swapped": 1024, "uncertain_staged": 262095, "t": 1087.9}
{"event": "probe", "step": 1000, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3409.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.195886, "cps": 26.632812, "mean_p": 0.238346, "boundary_entropy": 0.245442, "bpb": 2.336198}, "tr": {"chunks": 4044.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.223388, "cps": 31.59375, "mean_p": 0.235499, "boundary_entropy": 0.244067, "bpb": 2.657897}, "fi": {"chunks": 4164.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.22136, "cps": 32.53125, "mean_p": 0.253933, "boundary_entropy": 0.267563, "bpb": 2.855972}}, "bpb": {"en": 2.3362, "tr": 2.6579, "fi": 2.85597}, "gini_cpc": 0.02862, "gini_cps": 0.043327, "bpic": 4.883533, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.1404013282295677, "fi": 1.1300455680714208}, "inventory": {"size": 326650, "admitted": 65536, "staged": 261114, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 352, "admitted_surface": 10868, "staged_surface": 60091, "admitted_split": 50384, "staged_split": 44013, "admitted_merge": 4279, "staged_merge": 109828, "admitted_skel": 5, "staged_skel": 47182, "mean_count": 37.972198486328125, "n_util_obs": 74, "util_count_spearman": 0.12743522721731973}, "t": 1089.8}
{"event": "eval", "step": 1000, "phase": 1, "bpb": {"en": 2.34615, "tr": 2.66524, "fi": 2.86089}, "bytes_per_chunk": {"en": 5.121, "tr": 4.932, "fi": 4.919}, "bpb_mean": 2.62409, "frozen_inventory": true, "t": 1093.4}
{"event": "train", "step": 1000, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.74135, "bpb": 2.51223, "chunk_loss": 0.030565, "F": 0.19861, "G": 0.23965, "chunks": 203.38, "bytes_per_chunk": 5.035, "boundary_entropy": 0.24178, "router_noise": 0.0, "grad_norm": 1.0224, "lr": 0.0002901, "router_trainable": true, "sec_per_step": 1.353, "probe_sec": 0.2731, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.18155, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.059, "n_shadow": 5, "mean_delta_shadow": -0.02581, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.84387, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.61526, "trunc_abs_frac": 0.61526, "mean_delta_full": 0.95581, "read_hit_rate": {"surface": 0.2117, "split": 0.14221, "merge": 0.018, "skel": 0.04481}, "mean_gate": {"surface": 0.17192, "split": 0.14603, "merge": 0.16416, "skel": 0.14751}, "n_reads": 3345, "n_cand": 29531, "read_suppressed_frac": 0.0942, "dropout_reads_logged": 3345, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 353, "admitted_surface": 10868, "staged_surface": 60370, "admitted_split": 50384, "staged_split": 44128, "admitted_merge": 4279, "staged_merge": 110281, "admitted_skel": 5, "staged_skel": 47365, "mean_count": 38.02323913574219, "n_util_obs": 74, "util_count_spearman": 0.12571640133283968}, "probe_overhead_sec_total": 148.124, "probe_overhead_seq_frac": 0.4995, "alpha": 0.03, "t": 1100.6}
{"event": "probe", "step": 1024, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3457.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.198644, "cps": 27.007812, "mean_p": 0.239293, "boundary_entropy": 0.249329, "bpb": 2.329042}, "tr": {"chunks": 4124.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.227808, "cps": 32.21875, "mean_p": 0.231749, "boundary_entropy": 0.248274, "bpb": 2.622664}, "fi": {"chunks": 4208.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.223699, "cps": 32.875, "mean_p": 0.254351, "boundary_entropy": 0.268047, "bpb": 2.846681}}, "bpb": {"en": 2.32904, "tr": 2.62266, "fi": 2.84668}, "gini_cpc": 0.029905, "gini_cps": 0.042469, "bpic": 4.812283, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.146813629737501, "fi": 1.1261301560287282}, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 362, "admitted_surface": 10868, "staged_surface": 60519, "admitted_split": 50384, "staged_split": 43836, "admitted_merge": 4279, "staged_merge": 110711, "admitted_skel": 5, "staged_skel": 47078, "mean_count": 39.25666809082031, "n_util_obs": 77, "util_count_spearman": -0.0034965034965034965}, "t": 1130.5}
{"event": "consolidate", "step": 1050, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262095, "t": 1170.4}
{"event": "train", "step": 1050, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.64249, "bpb": 2.36962, "chunk_loss": 0.030696, "F": 0.19656, "G": 0.23923, "chunks": 201.28, "bytes_per_chunk": 5.087, "boundary_entropy": 0.24142, "router_noise": 0.0, "grad_norm": 1.1359, "lr": 0.0002886, "router_trainable": true, "sec_per_step": 1.409, "probe_sec": 0.0817, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05116, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.0123, "n_shadow": 2, "mean_delta_shadow": -0.10309, "n_split": 1, "mean_delta_split": 0.07971, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -5.04375, "trunc_abs_frac": 1.62148, "mean_delta_full": 0.00684, "read_hit_rate": {"surface": 0.23342, "split": 0.17781, "merge": 0.02133, "skel": 0.04983}, "mean_gate": {"surface": 0.16477, "split": 0.13547, "merge": 0.15281, "skel": 0.13167}, "n_reads": 3850, "n_cand": 28797, "read_suppressed_frac": 0.1005, "dropout_reads_logged": 3850, "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": 11487, "staged_surface": 59569, "admitted_split": 50147, "staged_split": 44292, "admitted_merge": 3897, "staged_merge": 111199, "admitted_skel": 5, "staged_skel": 47084, "mean_count": 42.88844299316406, "n_util_obs": 81, "util_count_spearman": 0.05413279132791328}, "probe_overhead_sec_total": 156.925, "probe_overhead_seq_frac": 0.4995, "alpha": 0.03, "t": 1171.0}
{"event": "consolidate", "step": 1100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262094, "t": 1233.8}
{"event": "train", "step": 1100, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.74526, "bpb": 2.51788, "chunk_loss": 0.0307, "F": 0.20474, "G": 0.2406, "chunks": 209.66, "bytes_per_chunk": 4.884, "boundary_entropy": 0.23855, "router_noise": 0.0, "grad_norm": 1.0386, "lr": 0.0002871, "router_trainable": true, "sec_per_step": 1.272, "probe_sec": 0.1332, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07189, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": 0.12185, "n_shadow": 2, "mean_delta_shadow": 0.29507, "n_split": 2, "mean_delta_split": -0.02682, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -1.24814, "trunc_abs_frac": 1.32226, "mean_delta_full": -0.0943, "read_hit_rate": {"surface": 0.26868, "split": 0.1678, "merge": 0.02182, "skel": 0.0516}, "mean_gate": {"surface": 0.17062, "split": 0.13537, "merge": 0.16256, "skel": 0.14069}, "n_reads": 4118, "n_cand": 29703, "read_suppressed_frac": 0.1047, "dropout_reads_logged": 4118, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 387, "admitted_surface": 12321, "staged_surface": 58639, "admitted_split": 49390, "staged_split": 43951, "admitted_merge": 3819, "staged_merge": 112868, "admitted_skel": 6, "staged_skel": 46686, "mean_count": 46.9259033203125, "n_util_obs": 87, "util_count_spearman": 0.014161259750674346}, "probe_overhead_sec_total": 165.605, "probe_overhead_seq_frac": 0.4995, "alpha": 0.03, "t": 1234.6}
{"event": "consolidate", "step": 1150, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 1024, "uncertain_staged": 262093, "t": 1302.4}
{"event": "train", "step": 1150, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.6422, "bpb": 2.36919, "chunk_loss": 0.03061, "F": 0.20309, "G": 0.23881, "chunks": 207.97, "bytes_per_chunk": 4.924, "boundary_entropy": 0.22732, "router_noise": 0.0, "grad_norm": 0.8587, "lr": 0.0002855, "router_trainable": true, "sec_per_step": 1.374, "probe_sec": 0.1318, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07012, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": -0.04566, "n_shadow": 4, "mean_delta_shadow": -0.03951, "n_split": 4, "mean_delta_split": 0.24224, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.03256, "trunc_abs_frac": 0.41518, "mean_delta_full": -0.19635, "read_hit_rate": {"surface": 0.26499, "split": 0.16415, "merge": 0.0179, "skel": 0.05212}, "mean_gate": {"surface": 0.16672, "split": 0.13042, "merge": 0.15324, "skel": 0.1502}, "n_reads": 4074, "n_cand": 30132, "read_suppressed_frac": 0.1063, "dropout_reads_logged": 4074, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 421, "admitted_surface": 13287, "staged_surface": 58260, "admitted_split": 48423, "staged_split": 44194, "admitted_merge": 3821, "staged_merge": 112942, "admitted_skel": 5, "staged_skel": 46748, "mean_count": 49.03009033203125, "n_util_obs": 89, "util_count_spearman": 0.044552264215185565}, "probe_overhead_sec_total": 173.963, "probe_overhead_seq_frac": 0.4996, "alpha": 0.03, "t": 1303.3}
{"event": "consolidate", "step": 1200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262088, "t": 1373.2}
{"event": "probe", "step": 1200, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3369.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.193587, "cps": 26.320312, "mean_p": 0.231573, "boundary_entropy": 0.228782, "bpb": 2.269407}, "tr": {"chunks": 4273.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.236038, "cps": 33.382812, "mean_p": 0.228113, "boundary_entropy": 0.22044, "bpb": 2.543588}, "fi": {"chunks": 4226.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.224656, "cps": 33.015625, "mean_p": 0.253118, "boundary_entropy": 0.249211, "bpb": 2.740239}}, "bpb": {"en": 2.26941, "tr": 2.54359, "fi": 2.74024}, "gini_cpc": 0.043254, "gini_cps": 0.050781, "bpic": 4.780249, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.2192856165046269, "fi": 1.1604881723350802}, "inventory": {"size": 326656, "admitted": 65536, "staged": 261120, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 434, "admitted_surface": 14301, "staged_surface": 57833, "admitted_split": 47419, "staged_split": 43480, "admitted_merge": 3810, "staged_merge": 113134, "admitted_skel": 6, "staged_skel": 46673, "mean_count": 54.13285827636719, "n_util_obs": 95, "util_count_spearman": -0.05031512568313872}, "t": 1376.1}
{"event": "eval", "step": 1200, "phase": 1, "bpb": {"en": 2.28163, "tr": 2.55233, "fi": 2.7474}, "bytes_per_chunk": {"en": 5.168, "tr": 4.638, "fi": 4.844}, "bpb_mean": 2.52712, "frozen_inventory": true, "t": 1382.0}
{"event": "train", "step": 1200, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.62071, "bpb": 2.33819, "chunk_loss": 0.03053, "F": 0.19986, "G": 0.23357, "chunks": 204.66, "bytes_per_chunk": 5.004, "boundary_entropy": 0.22459, "router_noise": 0.0, "grad_norm": 0.9458, "lr": 0.0002837, "router_trainable": true, "sec_per_step": 1.782, "probe_sec": 0.3257, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.20113, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.01391, "n_shadow": 5, "mean_delta_shadow": -0.19622, "n_split": 1, "mean_delta_split": 0.13208, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.94652, "trunc_abs_frac": 1.15114, "mean_delta_full": -0.09543, "read_hit_rate": {"surface": 0.25948, "split": 0.17628, "merge": 0.02174, "skel": 0.0679}, "mean_gate": {"surface": 0.17348, "split": 0.13143, "merge": 0.1581, "skel": 0.14296}, "n_reads": 4207, "n_cand": 29553, "read_suppressed_frac": 0.1017, "dropout_reads_logged": 4207, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 433, "admitted_surface": 14301, "staged_surface": 58092, "admitted_split": 47419, "staged_split": 43654, "admitted_merge": 3810, "staged_merge": 113632, "admitted_skel": 6, "staged_skel": 46766, "mean_count": 54.197052001953125, "n_util_obs": 96, "util_count_spearman": -0.056877373846988606}, "probe_overhead_sec_total": 183.858, "probe_overhead_seq_frac": 0.4996, "alpha": 0.03, "t": 1392.4}
{"event": "consolidate", "step": 1250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262088, "t": 1459.8}
{"event": "train", "step": 1250, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.63268, "bpb": 2.35546, "chunk_loss": 0.030506, "F": 0.20135, "G": 0.23081, "chunks": 206.19, "bytes_per_chunk": 4.966, "boundary_entropy": 0.21549, "router_noise": 0.0, "grad_norm": 0.9158, "lr": 0.0002819, "router_trainable": true, "sec_per_step": 1.365, "probe_sec": 0.1553, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10687, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 11, "mean_delta_loo": 0.03343, "n_shadow": 2, "mean_delta_shadow": -0.10543, "n_split": 1, "mean_delta_split": -0.06055, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 1.0, "trunc_abs_frac": 1.0, "mean_delta_full": -0.05908, "read_hit_rate": {"surface": 0.25891, "split": 0.18366, "merge": 0.01638, "skel": 0.06981}, "mean_gate": {"surface": 0.16658, "split": 0.12592, "merge": 0.15037, "skel": 0.13522}, "n_reads": 4266, "n_cand": 29544, "read_suppressed_frac": 0.0975, "dropout_reads_logged": 4266, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 451, "admitted_surface": 15305, "staged_surface": 57564, "admitted_split": 46400, "staged_split": 43700, "admitted_merge": 3824, "staged_merge": 114278, "admitted_skel": 7, "staged_skel": 46602, "mean_count": 58.88130187988281, "n_util_obs": 102, "util_count_spearman": -0.08826073926638808}, "probe_overhead_sec_total": 192.969, "probe_overhead_seq_frac": 0.4996, "alpha": 0.03, "t": 1460.7}
{"event": "consolidate", "step": 1300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262089, "t": 1526.3}
{"event": "train", "step": 1300, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.66921, "bpb": 2.40817, "chunk_loss": 0.030676, "F": 0.20584, "G": 0.23495, "chunks": 210.78, "bytes_per_chunk": 4.858, "boundary_entropy": 0.2103, "router_noise": 0.0, "grad_norm": 0.9619, "lr": 0.00028, "router_trainable": true, "sec_per_step": 1.331, "probe_sec": 0.1201, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0707, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": 0.06246, "n_shadow": 7, "mean_delta_shadow": 0.06941, "n_split": 1, "mean_delta_split": -0.19958, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.04672, "trunc_abs_frac": 1.16345, "mean_delta_full": 0.09318, "read_hit_rate": {"surface": 0.26158, "split": 0.18424, "merge": 0.02081, "skel": 0.07765}, "mean_gate": {"surface": 0.16398, "split": 0.12353, "merge": 0.15324, "skel": 0.14157}, "n_reads": 4448, "n_cand": 30231, "read_suppressed_frac": 0.0976, "dropout_reads_logged": 4448, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 453, "admitted_surface": 16216, "staged_surface": 57484, "admitted_split": 45485, "staged_split": 43369, "admitted_merge": 3826, "staged_merge": 114882, "admitted_skel": 9, "staged_skel": 46409, "mean_count": 63.101226806640625, "n_util_obs": 110, "util_count_spearman": -0.04079525730901878}, "probe_overhead_sec_total": 201.208, "probe_overhead_seq_frac": 0.4996, "alpha": 0.03, "t": 1527.2}
{"event": "consolidate", "step": 1350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262089, "t": 1581.2}
{"event": "train", "step": 1350, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.59932, "bpb": 2.30734, "chunk_loss": 0.030604, "F": 0.20239, "G": 0.23352, "chunks": 207.25, "bytes_per_chunk": 4.941, "boundary_entropy": 0.21063, "router_noise": 0.0, "grad_norm": 0.9415, "lr": 0.000278, "router_trainable": true, "sec_per_step": 1.09, "probe_sec": 0.0869, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06496, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.10085, "n_shadow": 6, "mean_delta_shadow": -0.06902, "n_split": 1, "mean_delta_split": 0.52063, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.8972, "trunc_abs_frac": 2.25055, "mean_delta_full": 0.21196, "read_hit_rate": {"surface": 0.30182, "split": 0.18843, "merge": 0.02543, "skel": 0.08328}, "mean_gate": {"surface": 0.16638, "split": 0.1216, "merge": 0.15665, "skel": 0.14313}, "n_reads": 4757, "n_cand": 29763, "read_suppressed_frac": 0.1081, "dropout_reads_logged": 4757, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 469, "admitted_surface": 16936, "staged_surface": 57302, "admitted_split": 44812, "staged_split": 43779, "admitted_merge": 3779, "staged_merge": 114566, "admitted_skel": 9, "staged_skel": 46497, "mean_count": 67.26609802246094, "n_util_obs": 115, "util_count_spearman": -0.06640101002130513}, "probe_overhead_sec_total": 209.073, "probe_overhead_seq_frac": 0.4996, "alpha": 0.03, "t": 1581.7}
{"event": "consolidate", "step": 1400, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262088, "t": 1643.6}
{"event": "probe", "step": 1400, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3442.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.197782, "cps": 26.890625, "mean_p": 0.230555, "boundary_entropy": 0.212502, "bpb": 2.218088}, "tr": {"chunks": 4125.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.227863, "cps": 32.226562, "mean_p": 0.223107, "boundary_entropy": 0.20585, "bpb": 2.471121}, "fi": {"chunks": 4242.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.225506, "cps": 33.140625, "mean_p": 0.247067, "boundary_entropy": 0.230101, "bpb": 2.647612}}, "bpb": {"en": 2.21809, "tr": 2.47112, "fi": 2.64761}, "gini_cpc": 0.030798, "gini_cps": 0.045163, "bpic": 4.804132, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.1520906596102238, "fi": 1.1401763670674845}, "inventory": {"size": 326655, "admitted": 65536, "staged": 261119, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 508, "admitted_surface": 17857, "staged_surface": 56548, "admitted_split": 43910, "staged_split": 43797, "admitted_merge": 3760, "staged_merge": 114442, "admitted_skel": 9, "staged_skel": 46332, "mean_count": 71.72763061523438, "n_util_obs": 122, "util_count_spearman": -0.07590682735170394}, "t": 1646.6}
{"event": "eval", "step": 1400, "phase": 1, "bpb": {"en": 2.22744, "tr": 2.47841, "fi": 2.65306}, "bytes_per_chunk": {"en": 5.079, "tr": 4.816, "fi": 4.877}, "bpb_mean": 2.45297, "frozen_inventory": true, "t": 1651.4}
{"event": "train", "step": 1400, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.60031, "bpb": 2.30876, "chunk_loss": 0.030513, "F": 0.19989, "G": 0.22704, "chunks": 204.69, "bytes_per_chunk": 5.003, "boundary_entropy": 0.20432, "router_noise": 0.0, "grad_norm": 0.9219, "lr": 0.0002759, "router_trainable": true, "sec_per_step": 1.542, "probe_sec": 0.2778, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.18582, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.05883, "n_shadow": 3, "mean_delta_shadow": -0.01425, "n_split": 1, "mean_delta_split": 0.29382, "n_merge": 1, "mean_delta_merge": -0.29846, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": -0.02222, "trunc_abs_frac": 0.02222, "mean_delta_full": -0.31445, "read_hit_rate": {"surface": 0.33032, "split": 0.18525, "merge": 0.01989, "skel": 0.08893}, "mean_gate": {"surface": 0.16543, "split": 0.11741, "merge": 0.15631, "skel": 0.14005}, "n_reads": 4821, "n_cand": 29236, "read_suppressed_frac": 0.1008, "dropout_reads_logged": 4821, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 509, "admitted_surface": 17857, "staged_surface": 56718, "admitted_split": 43910, "staged_split": 43974, "admitted_merge": 3760, "staged_merge": 114954, "admitted_skel": 9, "staged_skel": 46498, "mean_count": 71.80119323730469, "n_util_obs": 122, "util_count_spearman": -0.0745586063095423}, "probe_overhead_sec_total": 217.908, "probe_overhead_seq_frac": 0.4996, "alpha": 0.03, "t": 1658.8}
{"event": "consolidate", "step": 1450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262088, "t": 1720.2}
{"event": "train", "step": 1450, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.64789, "bpb": 2.37741, "chunk_loss": 0.030592, "F": 0.20621, "G": 0.23064, "chunks": 211.16, "bytes_per_chunk": 4.849, "boundary_entropy": 0.1978, "router_noise": 0.0, "grad_norm": 0.8662, "lr": 0.0002738, "router_trainable": true, "sec_per_step": 1.238, "probe_sec": 0.1053, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07011, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.00925, "n_shadow": 5, "mean_delta_shadow": 0.02462, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.1178, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.23284, "trunc_abs_frac": 1.13131, "mean_delta_full": 0.33875, "read_hit_rate": {"surface": 0.32848, "split": 0.18135, "merge": 0.02032, "skel": 0.06949}, "mean_gate": {"surface": 0.16358, "split": 0.11467, "merge": 0.15082, "skel": 0.13136}, "n_reads": 4793, "n_cand": 29970, "read_suppressed_frac": 0.1014, "dropout_reads_logged": 4793, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 519, "admitted_surface": 18877, "staged_surface": 56331, "admitted_split": 42896, "staged_split": 43775, "admitted_merge": 3755, "staged_merge": 115203, "admitted_skel": 8, "staged_skel": 46835, "mean_count": 73.98281860351562, "n_util_obs": 125, "util_count_spearman": -0.005874039938556067}, "probe_overhead_sec_total": 227.052, "probe_overhead_seq_frac": 0.4997, "alpha": 0.03, "t": 1720.7}
{"event": "consolidate", "step": 1500, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 1024, "uncertain_staged": 262084, "t": 1781.3}
{"event": "train", "step": 1500, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.66003, "bpb": 2.39491, "chunk_loss": 0.030966, "F": 0.21985, "G": 0.24177, "chunks": 225.12, "bytes_per_chunk": 4.549, "boundary_entropy": 0.19708, "router_noise": 0.0, "grad_norm": 0.893, "lr": 0.0002715, "router_trainable": true, "sec_per_step": 1.228, "probe_sec": 0.0854, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05553, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.00643, "n_shadow": 2, "mean_delta_shadow": 0.10553, "n_split": 1, "mean_delta_split": -0.14923, "n_merge": 2, "mean_delta_merge": 0.33704, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.63705, "trunc_abs_frac": 0.63705, "mean_delta_full": -0.23425, "read_hit_rate": {"surface": 0.34202, "split": 0.18266, "merge": 0.02619, "skel": 0.08584}, "mean_gate": {"surface": 0.16074, "split": 0.11276, "merge": 0.15248, "skel": 0.13006}, "n_reads": 5314, "n_cand": 31662, "read_suppressed_frac": 0.1084, "dropout_reads_logged": 5314, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 536, "admitted_surface": 19705, "staged_surface": 56227, "admitted_split": 42093, "staged_split": 44002, "admitted_merge": 3729, "staged_merge": 115081, "admitted_skel": 9, "staged_skel": 46834, "mean_count": 78.23432922363281, "n_util_obs": 131, "util_count_spearman": -0.0036833395612021567}, "probe_overhead_sec_total": 234.935, "probe_overhead_seq_frac": 0.4997, "alpha": 0.03, "t": 1782.1}
{"event": "consolidate", "step": 1550, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262082, "t": 1839.0}
{"event": "train", "step": 1550, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.58591, "bpb": 2.28799, "chunk_loss": 0.030619, "F": 0.20612, "G": 0.22746, "chunks": 211.06, "bytes_per_chunk": 4.852, "boundary_entropy": 0.18678, "router_noise": 0.0, "grad_norm": 0.9225, "lr": 0.0002692, "router_trainable": true, "sec_per_step": 1.153, "probe_sec": 0.1255, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07189, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.05997, "n_shadow": 6, "mean_delta_shadow": 0.11568, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.18814, "trunc_abs_frac": 0.18717, "mean_delta_full": 0.06512, "read_hit_rate": {"surface": 0.33829, "split": 0.1843, "merge": 0.02242, "skel": 0.09545}, "mean_gate": {"surface": 0.15745, "split": 0.10753, "merge": 0.14813, "skel": 0.12487}, "n_reads": 5064, "n_cand": 30176, "read_suppressed_frac": 0.1005, "dropout_reads_logged": 5064, "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": 20721, "staged_surface": 55532, "admitted_split": 41073, "staged_split": 44060, "admitted_merge": 3733, "staged_merge": 115586, "admitted_skel": 9, "staged_skel": 46966, "mean_count": 82.14901733398438, "n_util_obs": 133, "util_count_spearman": -0.019228047086924417}, "probe_overhead_sec_total": 242.975, "probe_overhead_seq_frac": 0.4997, "alpha": 0.03, "t": 1839.8}
{"event": "consolidate", "step": 1600, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 1024, "uncertain_staged": 262080, "t": 1901.4}
{"event": "probe", "step": 1600, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3395.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.195081, "cps": 26.523438, "mean_p": 0.223491, "boundary_entropy": 0.191341, "bpb": 2.172871}, "tr": {"chunks": 3979.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.219798, "cps": 31.085938, "mean_p": 0.216554, "boundary_entropy": 0.191852, "bpb": 2.421045}, "fi": {"chunks": 4272.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.227101, "cps": 33.375, "mean_p": 0.247927, "boundary_entropy": 0.212711, "bpb": 2.589058}}, "bpb": {"en": 2.17287, "tr": 2.42104, "fi": 2.58906}, "gini_cpc": 0.033251, "gini_cps": 0.050203, "bpic": 4.871372, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.1266985341691875, "fi": 1.1641359531042417}, "inventory": {"size": 326654, "admitted": 65536, "staged": 261118, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 605, "admitted_surface": 21739, "staged_surface": 55196, "admitted_split": 40056, "staged_split": 44369, "admitted_merge": 3732, "staged_merge": 114650, "admitted_skel": 9, "staged_skel": 46903, "mean_count": 85.78350830078125, "n_util_obs": 135, "util_count_spearman": -0.004362994296894153}, "t": 1904.0}
{"event": "eval", "step": 1600, "phase": 1, "bpb": {"en": 2.18634, "tr": 2.42894, "fi": 2.59554}, "bytes_per_chunk": {"en": 5.13, "tr": 5.01, "fi": 4.868}, "bpb_mean": 2.40361, "frozen_inventory": true, "t": 1908.6}
{"event": "train", "step": 1600, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.52068, "bpb": 2.19388, "chunk_loss": 0.030414, "F": 0.19644, "G": 0.22653, "chunks": 201.16, "bytes_per_chunk": 5.091, "boundary_entropy": 0.19542, "router_noise": 0.0, "grad_norm": 0.9523, "lr": 0.0002668, "router_trainable": true, "sec_per_step": 1.535, "probe_sec": 0.2961, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.11172, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": -0.00358, "n_shadow": 5, "mean_delta_shadow": -0.08983, "n_split": 1, "mean_delta_split": -0.18921, "n_merge": 2, "mean_delta_merge": -0.06495, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 8.60486, "trunc_abs_frac": 0.64178, "mean_delta_full": -0.01339, "read_hit_rate": {"surface": 0.37658, "split": 0.1948, "merge": 0.01899, "skel": 0.08494}, "mean_gate": {"surface": 0.16221, "split": 0.10983, "merge": 0.15092, "skel": 0.13785}, "n_reads": 5220, "n_cand": 29440, "read_suppressed_frac": 0.099, "dropout_reads_logged": 5220, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 607, "admitted_surface": 21739, "staged_surface": 55405, "admitted_split": 40056, "staged_split": 44394, "admitted_merge": 3732, "staged_merge": 115358, "admitted_skel": 9, "staged_skel": 46987, "mean_count": 85.8631591796875, "n_util_obs": 136, "util_count_spearman": 0.0004245985694847845}, "probe_overhead_sec_total": 251.409, "probe_overhead_seq_frac": 0.4997, "alpha": 0.03, "t": 1916.5}
{"event": "consolidate", "step": 1650, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262077, "t": 1971.2}
{"event": "train", "step": 1650, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.57816, "bpb": 2.27681, "chunk_loss": 0.030562, "F": 0.20605, "G": 0.22781, "chunks": 211.0, "bytes_per_chunk": 4.853, "boundary_entropy": 0.18561, "router_noise": 0.0, "grad_norm": 0.8493, "lr": 0.0002643, "router_trainable": true, "sec_per_step": 1.113, "probe_sec": 0.176, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.12361, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": 0.03473, "n_shadow": 5, "mean_delta_shadow": -0.01249, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": -0.1648, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 1.26411, "trunc_abs_frac": 2.39012, "mean_delta_full": -0.07733, "read_hit_rate": {"surface": 0.34003, "split": 0.19209, "merge": 0.02138, "skel": 0.08639}, "mean_gate": {"surface": 0.16082, "split": 0.10832, "merge": 0.14476, "skel": 0.1339}, "n_reads": 5135, "n_cand": 30367, "read_suppressed_frac": 0.1038, "dropout_reads_logged": 5135, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 626, "admitted_surface": 22759, "staged_surface": 54972, "admitted_split": 39035, "staged_split": 44087, "admitted_merge": 3733, "staged_merge": 116393, "admitted_skel": 9, "staged_skel": 46692, "mean_count": 89.93350219726562, "n_util_obs": 140, "util_count_spearman": 0.01712188353840163}, "probe_overhead_sec_total": 259.306, "probe_overhead_seq_frac": 0.4997, "alpha": 0.03, "t": 1972.1}
{"event": "consolidate", "step": 1700, "retired": 0, "split_retired": 4, "merge_admitted": 0, "admitted": 4, "swapped": 1024, "uncertain_staged": 262075, "t": 2025.6}
{"event": "train", "step": 1700, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.54161, "bpb": 2.22407, "chunk_loss": 0.03044, "F": 0.20081, "G": 0.2206, "chunks": 205.62, "bytes_per_chunk": 4.98, "boundary_entropy": 0.18056, "router_noise": 0.0, "grad_norm": 0.7625, "lr": 0.0002618, "router_trainable": true, "sec_per_step": 1.098, "probe_sec": 0.1366, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07203, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.01317, "n_shadow": 5, "mean_delta_shadow": -0.18472, "n_split": 1, "mean_delta_split": -0.04315, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 1.34839, "trunc_abs_frac": 5.82671, "mean_delta_full": -0.08677, "read_hit_rate": {"surface": 0.33109, "split": 0.1858, "merge": 0.0221, "skel": 0.08404}, "mean_gate": {"surface": 0.16101, "split": 0.10975, "merge": 0.1494, "skel": 0.12993}, "n_reads": 4831, "n_cand": 29287, "read_suppressed_frac": 0.1085, "dropout_reads_logged": 4831, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 645, "admitted_surface": 23780, "staged_surface": 54802, "admitted_split": 38012, "staged_split": 44653, "admitted_merge": 3734, "staged_merge": 115707, "admitted_skel": 10, "staged_skel": 46982, "mean_count": 92.93241882324219, "n_util_obs": 144, "util_count_spearman": 0.0014347721244272968}, "probe_overhead_sec_total": 266.796, "probe_overhead_seq_frac": 0.4997, "alpha": 0.03, "t": 2027.0}
{"event": "consolidate", "step": 1750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262074, "t": 2083.1}
{"event": "train", "step": 1750, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.56561, "bpb": 2.2587, "chunk_loss": 0.030576, "F": 0.20541, "G": 0.2275, "chunks": 210.34, "bytes_per_chunk": 4.868, "boundary_entropy": 0.17962, "router_noise": 0.0, "grad_norm": 0.7876, "lr": 0.0002591, "router_trainable": true, "sec_per_step": 1.159, "probe_sec": 0.1102, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06258, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": -0.01274, "n_shadow": 5, "mean_delta_shadow": -0.10213, "n_split": 2, "mean_delta_split": -0.01767, "n_merge": 1, "mean_delta_merge": -0.3537, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -0.04201, "trunc_abs_frac": 0.24062, "mean_delta_full": -0.21488, "read_hit_rate": {"surface": 0.32796, "split": 0.18021, "merge": 0.01875, "skel": 0.09024}, "mean_gate": {"surface": 0.16114, "split": 0.10598, "merge": 0.15103, "skel": 0.13036}, "n_reads": 4903, "n_cand": 30261, "read_suppressed_frac": 0.1042, "dropout_reads_logged": 4903, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 666, "admitted_surface": 24781, "staged_surface": 54703, "admitted_split": 36999, "staged_split": 44947, "admitted_merge": 3746, "staged_merge": 115499, "admitted_skel": 10, "staged_skel": 46995, "mean_count": 95.22300720214844, "n_util_obs": 142, "util_count_spearman": -0.010035813751726248}, "probe_overhead_sec_total": 275.687, "probe_overhead_seq_frac": 0.4997, "alpha": 0.03, "t": 2085.0}
{"event": "consolidate", "step": 1800, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262075, "t": 2145.4}
{"event": "probe", "step": 1800, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3461.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.198874, "cps": 27.039062, "mean_p": 0.224045, "boundary_entropy": 0.180615, "bpb": 2.133735}, "tr": {"chunks": 4131.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.228194, "cps": 32.273438, "mean_p": 0.218047, "boundary_entropy": 0.173991, "bpb": 2.36056}, "fi": {"chunks": 4316.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.22944, "cps": 33.71875, "mean_p": 0.244273, "boundary_entropy": 0.190089, "bpb": 2.532246}}, "bpb": {"en": 2.13373, "tr": 2.36056, "fi": 2.53225}, "gini_cpc": 0.031039, "gini_cps": 0.047867, "bpic": 4.764192, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.1474325468458497, "fi": 1.15369782393087}, "inventory": {"size": 326655, "admitted": 65536, "staged": 261119, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 691, "admitted_surface": 25547, "staged_surface": 54358, "admitted_split": 36248, "staged_split": 44319, "admitted_merge": 3732, "staged_merge": 115775, "admitted_skel": 9, "staged_skel": 46667, "mean_count": 97.66160583496094, "n_util_obs": 145, "util_count_spearman": -0.01754057931939179}, "t": 2148.0}
{"event": "eval", "step": 1800, "phase": 1, "bpb": {"en": 2.14858, "tr": 2.36841, "fi": 2.53701}, "bytes_per_chunk": {"en": 5.034, "tr": 4.853, "fi": 4.758}, "bpb_mean": 2.35133, "frozen_inventory": true, "t": 2152.2}
{"event": "train", "step": 1800, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.53775, "bpb": 2.2185, "chunk_loss": 0.030398, "F": 0.19647, "G": 0.21863, "chunks": 201.19, "bytes_per_chunk": 5.09, "boundary_entropy": 0.17613, "router_noise": 0.0, "grad_norm": 0.7901, "lr": 0.0002564, "router_trainable": true, "sec_per_step": 1.493, "probe_sec": 0.1392, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05555, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": -0.00361, "n_shadow": 4, "mean_delta_shadow": 0.03288, "n_split": 3, "mean_delta_split": -0.06917, "n_merge": 1, "mean_delta_merge": 0.23236, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -0.0238, "trunc_abs_frac": 0.61779, "mean_delta_full": -0.17993, "read_hit_rate": {"surface": 0.3491, "split": 0.17993, "merge": 0.01868, "skel": 0.07646}, "mean_gate": {"surface": 0.15916, "split": 0.10558, "merge": 0.14548, "skel": 0.13518}, "n_reads": 4764, "n_cand": 28948, "read_suppressed_frac": 0.094, "dropout_reads_logged": 4764, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 694, "admitted_surface": 25547, "staged_surface": 54576, "admitted_split": 36248, "staged_split": 44492, "admitted_merge": 3732, "staged_merge": 116248, "admitted_skel": 9, "staged_skel": 46828, "mean_count": 97.73429870605469, "n_util_obs": 145, "util_count_spearman": -0.017548452110827784}, "probe_overhead_sec_total": 284.866, "probe_overhead_seq_frac": 0.4997, "alpha": 0.03, "t": 2159.7}
{"event": "consolidate", "step": 1850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262075, "t": 2216.7}
{"event": "train", "step": 1850, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.55356, "bpb": 2.24131, "chunk_loss": 0.030483, "F": 0.19894, "G": 0.22225, "chunks": 203.72, "bytes_per_chunk": 5.027, "boundary_entropy": 0.17474, "router_noise": 0.0, "grad_norm": 0.7846, "lr": 0.0002537, "router_trainable": true, "sec_per_step": 1.153, "probe_sec": 0.1469, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07422, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.01932, "n_shadow": 2, "mean_delta_shadow": -0.07623, "n_split": 2, "mean_delta_split": 0.11722, "n_merge": 1, "mean_delta_merge": 0.1275, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": -0.43723, "trunc_abs_frac": 0.43723, "mean_delta_full": 0.11255, "read_hit_rate": {"surface": 0.3396, "split": 0.17963, "merge": 0.01859, "skel": 0.06378}, "mean_gate": {"surface": 0.15502, "split": 0.10063, "merge": 0.15136, "skel": 0.12041}, "n_reads": 4698, "n_cand": 29399, "read_suppressed_frac": 0.0973, "dropout_reads_logged": 4698, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 703, "admitted_surface": 26387, "staged_surface": 54482, "admitted_split": 35468, "staged_split": 44552, "admitted_merge": 3673, "staged_merge": 116202, "admitted_skel": 8, "staged_skel": 46908, "mean_count": 100.21632385253906, "n_util_obs": 148, "util_count_spearman": -0.021962929920743918}, "probe_overhead_sec_total": 293.275, "probe_overhead_seq_frac": 0.4997, "alpha": 0.03, "t": 2217.3}
{"event": "consolidate", "step": 1900, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262073, "t": 2278.5}
{"event": "train", "step": 1900, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.53084, "bpb": 2.20854, "chunk_loss": 0.030752, "F": 0.20789, "G": 0.22662, "chunks": 212.88, "bytes_per_chunk": 4.81, "boundary_entropy": 0.17171, "router_noise": 0.0, "grad_norm": 0.8177, "lr": 0.0002508, "router_trainable": true, "sec_per_step": 1.246, "probe_sec": 0.1136, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07119, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.01696, "n_shadow": 4, "mean_delta_shadow": -0.0235, "n_split": 2, "mean_delta_split": -0.00319, "n_merge": 1, "mean_delta_merge": -0.04144, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.82623, "trunc_abs_frac": 0.54163, "mean_delta_full": 0.02863, "read_hit_rate": {"surface": 0.37522, "split": 0.187, "merge": 0.0283, "skel": 0.07449}, "mean_gate": {"surface": 0.16474, "split": 0.10501, "merge": 0.14905, "skel": 0.12631}, "n_reads": 5307, "n_cand": 30327, "read_suppressed_frac": 0.1023, "dropout_reads_logged": 5307, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 713, "admitted_surface": 27107, "staged_surface": 54643, "admitted_split": 34797, "staged_split": 44023, "admitted_merge": 3624, "staged_merge": 116486, "admitted_skel": 8, "staged_skel": 46992, "mean_count": 104.84146118164062, "n_util_obs": 153, "util_count_spearman": -0.05339217026523528}, "probe_overhead_sec_total": 301.974, "probe_overhead_seq_frac": 0.4997, "alpha": 0.03, "t": 2279.6}
{"event": "consolidate", "step": 1950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262073, "t": 2331.2}
{"event": "train", "step": 1950, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.47932, "bpb": 2.13421, "chunk_loss": 0.030455, "F": 0.19739, "G": 0.21791, "chunks": 202.12, "bytes_per_chunk": 5.066, "boundary_entropy": 0.1731, "router_noise": 0.0, "grad_norm": 0.708, "lr": 0.0002479, "router_trainable": true, "sec_per_step": 1.059, "probe_sec": 0.2463, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.12283, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.06771, "n_shadow": 2, "mean_delta_shadow": -0.09509, "n_split": 2, "mean_delta_split": 0.22858, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.70748, "trunc_abs_frac": 0.84062, "mean_delta_full": -0.20886, "read_hit_rate": {"surface": 0.41299, "split": 0.18421, "merge": 0.02561, "skel": 0.07202}, "mean_gate": {"surface": 0.16629, "split": 0.10413, "merge": 0.15213, "skel": 0.1281}, "n_reads": 5231, "n_cand": 28852, "read_suppressed_frac": 0.0962, "dropout_reads_logged": 5231, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 739, "admitted_surface": 28122, "staged_surface": 54069, "admitted_split": 33779, "staged_split": 44203, "admitted_merge": 3627, "staged_merge": 116701, "admitted_skel": 8, "staged_skel": 47171, "mean_count": 109.295654296875, "n_util_obs": 160, "util_count_spearman": -0.0232948552677839}, "probe_overhead_sec_total": 310.454, "probe_overhead_seq_frac": 0.4997, "alpha": 0.03, "t": 2332.5}
{"event": "consolidate", "step": 2000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262069, "t": 2390.1}
{"event": "probe", "step": 2000, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3490.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.20054, "cps": 27.265625, "mean_p": 0.223587, "boundary_entropy": 0.169166, "bpb": 2.11044}, "tr": {"chunks": 4222.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.233221, "cps": 32.984375, "mean_p": 0.221799, "boundary_entropy": 0.154941, "bpb": 2.303469}, "fi": {"chunks": 4486.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.238477, "cps": 35.046875, "mean_p": 0.247558, "boundary_entropy": 0.178764, "bpb": 2.478164}}, "bpb": {"en": 2.11044, "tr": 2.30347, "fi": 2.47816}, "gini_cpc": 0.037623, "gini_cps": 0.054435, "bpic": 4.650926, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.1629642667151212, "fi": 1.1891758449568997}, "inventory": {"size": 326656, "admitted": 65536, "staged": 261120, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 770, "admitted_surface": 29131, "staged_surface": 52949, "admitted_split": 32762, "staged_split": 44404, "admitted_merge": 3635, "staged_merge": 116658, "admitted_skel": 8, "staged_skel": 47109, "mean_count": 111.14234924316406, "n_util_obs": 163, "util_count_spearman": -0.025310212202326522}, "t": 2392.3}
{"event": "eval", "step": 2000, "phase": 1, "bpb": {"en": 2.12184, "tr": 2.31288, "fi": 2.48789}, "bytes_per_chunk": {"en": 4.925, "tr": 4.73, "fi": 4.568}, "bpb_mean": 2.30753, "frozen_inventory": true, "t": 2395.6}
{"event": "train", "step": 2000, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.63951, "bpb": 2.36531, "chunk_loss": 0.030877, "F": 0.21756, "G": 0.23612, "chunks": 222.78, "bytes_per_chunk": 4.596, "boundary_entropy": 0.16977, "router_noise": 0.0, "grad_norm": 0.8639, "lr": 0.000245, "router_trainable": true, "sec_per_step": 1.377, "probe_sec": 0.1539, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05852, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 3, "mean_delta_loo": 0.01429, "n_shadow": 6, "mean_delta_shadow": -0.11251, "n_split": 2, "mean_delta_split": -0.13504, "n_merge": 1, "mean_delta_merge": -0.02188, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 1.50782, "trunc_abs_frac": 0.37186, "mean_delta_full": -0.08612, "read_hit_rate": {"surface": 0.40426, "split": 0.18337, "merge": 0.02109, "skel": 0.08182}, "mean_gate": {"surface": 0.15111, "split": 0.09628, "merge": 0.14132, "skel": 0.12122}, "n_reads": 5671, "n_cand": 31500, "read_suppressed_frac": 0.0972, "dropout_reads_logged": 5671, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 771, "admitted_surface": 29131, "staged_surface": 53145, "admitted_split": 32762, "staged_split": 44499, "admitted_merge": 3635, "staged_merge": 117319, "admitted_skel": 8, "staged_skel": 47181, "mean_count": 111.2288818359375, "n_util_obs": 163, "util_count_spearman": -0.024882111661235724}, "probe_overhead_sec_total": 319.514, "probe_overhead_seq_frac": 0.4998, "alpha": 0.03, "t": 2401.4}
{"event": "probe", "step": 2048, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3419.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.19646, "cps": 26.710938, "mean_p": 0.220659, "boundary_entropy": 0.173334, "bpb": 2.102966}, "tr": {"chunks": 3922.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.216649, "cps": 30.640625, "mean_p": 0.212021, "boundary_entropy": 0.171649, "bpb": 2.303122}, "fi": {"chunks": 4317.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.229493, "cps": 33.726562, "mean_p": 0.241865, "boundary_entropy": 0.184061, "bpb": 2.478513}}, "bpb": {"en": 2.10297, "tr": 2.30312, "fi": 2.47851}, "gini_cpc": 0.03427, "gini_cps": 0.051352, "bpic": 4.866358, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.1027626727350048, "fi": 1.168140777655112}, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 816, "admitted_surface": 29131, "staged_surface": 53519, "admitted_split": 32762, "staged_split": 44641, "admitted_merge": 3635, "staged_merge": 116516, "admitted_skel": 8, "staged_skel": 47468, "mean_count": 115.10115051269531, "n_util_obs": 168, "util_count_spearman": -0.0008731480403520938}, "t": 2454.9}
{"event": "consolidate", "step": 2050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262069, "t": 2472.5}
{"event": "train", "step": 2050, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.48684, "bpb": 2.14505, "chunk_loss": 0.030407, "F": 0.19156, "G": 0.21348, "chunks": 196.16, "bytes_per_chunk": 5.22, "boundary_entropy": 0.17114, "router_noise": 0.0, "grad_norm": 0.827, "lr": 0.0002419, "router_trainable": true, "sec_per_step": 1.435, "probe_sec": 0.1109, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05402, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": 0.03091, "n_shadow": 8, "mean_delta_shadow": -0.0742, "n_split": 1, "mean_delta_split": -0.0459, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.73196, "trunc_abs_frac": 0.65656, "mean_delta_full": -0.4952, "read_hit_rate": {"surface": 0.41745, "split": 0.19323, "merge": 0.02125, "skel": 0.0632}, "mean_gate": {"surface": 0.15739, "split": 0.09975, "merge": 0.14354, "skel": 0.1201}, "n_reads": 5176, "n_cand": 28296, "read_suppressed_frac": 0.0999, "dropout_reads_logged": 5176, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 817, "admitted_surface": 30147, "staged_surface": 52735, "admitted_split": 31744, "staged_split": 44818, "admitted_merge": 3637, "staged_merge": 116951, "admitted_skel": 8, "staged_skel": 47640, "mean_count": 117.11262512207031, "n_util_obs": 170, "util_count_spearman": -0.034558899860162064}, "probe_overhead_sec_total": 327.724, "probe_overhead_seq_frac": 0.4998, "alpha": 0.03, "t": 2473.2}
{"event": "consolidate", "step": 2100, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262071, "t": 2532.1}
{"event": "train", "step": 2100, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.54551, "bpb": 2.2297, "chunk_loss": 0.030409, "F": 0.1983, "G": 0.21772, "chunks": 203.06, "bytes_per_chunk": 5.043, "boundary_entropy": 0.16415, "router_noise": 0.0, "grad_norm": 0.8114, "lr": 0.0002388, "router_trainable": true, "sec_per_step": 1.198, "probe_sec": 0.2001, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06708, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.00393, "n_shadow": 6, "mean_delta_shadow": -0.05495, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.66235, "trunc_abs_frac": 1.0029, "mean_delta_full": -0.08023, "read_hit_rate": {"surface": 0.41262, "split": 0.1914, "merge": 0.0272, "skel": 0.08003}, "mean_gate": {"surface": 0.15754, "split": 0.09681, "merge": 0.14498, "skel": 0.12203}, "n_reads": 5398, "n_cand": 29128, "read_suppressed_frac": 0.1013, "dropout_reads_logged": 5398, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 828, "admitted_surface": 31121, "staged_surface": 52377, "admitted_split": 30723, "staged_split": 44986, "admitted_merge": 3683, "staged_merge": 117006, "admitted_skel": 9, "staged_skel": 47775, "mean_count": 121.29060363769531, "n_util_obs": 177, "util_count_spearman": -0.04087100292584962}, "probe_overhead_sec_total": 336.12, "probe_overhead_seq_frac": 0.4998, "alpha": 0.03, "t": 2533.0}
{"event": "consolidate", "step": 2150, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 1024, "uncertain_staged": 262071, "t": 2593.7}
{"event": "train", "step": 2150, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.49978, "bpb": 2.16373, "chunk_loss": 0.030549, "F": 0.20453, "G": 0.22283, "chunks": 209.44, "bytes_per_chunk": 4.889, "boundary_entropy": 0.16464, "router_noise": 0.0, "grad_norm": 0.703, "lr": 0.0002357, "router_trainable": true, "sec_per_step": 1.245, "probe_sec": 0.1834, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09867, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": -0.0776, "n_shadow": 4, "mean_delta_shadow": -0.07265, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": -0.25562, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 1.04726, "trunc_abs_frac": 0.79244, "mean_delta_full": -0.29277, "read_hit_rate": {"surface": 0.44513, "split": 0.19127, "merge": 0.06945, "skel": 0.08532}, "mean_gate": {"surface": 0.15996, "split": 0.09328, "merge": 0.14373, "skel": 0.11344}, "n_reads": 6096, "n_cand": 30038, "read_suppressed_frac": 0.0966, "dropout_reads_logged": 6096, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 860, "admitted_surface": 31133, "staged_surface": 52995, "admitted_split": 29702, "staged_split": 45073, "admitted_merge": 4692, "staged_merge": 116431, "admitted_skel": 9, "staged_skel": 47645, "mean_count": 126.25588989257812, "n_util_obs": 182, "util_count_spearman": -0.09232620472544392}, "probe_overhead_sec_total": 345.157, "probe_overhead_seq_frac": 0.4998, "alpha": 0.03, "t": 2595.3}
{"event": "consolidate", "step": 2200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262070, "t": 2651.5}
{"event": "probe", "step": 2200, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3456.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.198586, "cps": 27.0, "mean_p": 0.218971, "boundary_entropy": 0.158953, "bpb": 2.088025}, "tr": {"chunks": 4169.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.230293, "cps": 32.570312, "mean_p": 0.219184, "boundary_entropy": 0.144178, "bpb": 2.280584}, "fi": {"chunks": 4389.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.233321, "cps": 34.289062, "mean_p": 0.24141, "boundary_entropy": 0.168253, "bpb": 2.451821}}, "bpb": {"en": 2.08803, "tr": 2.28058, "fi": 2.45182}, "gini_cpc": 0.034969, "gini_cps": 0.051773, "bpic": 4.722157, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.1596628110288658, "fi": 1.1749086029007851}, "inventory": {"size": 326656, "admitted": 65536, "staged": 261120, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 887, "admitted_surface": 31453, "staged_surface": 53102, "admitted_split": 28684, "staged_split": 44961, "admitted_merge": 5390, "staged_merge": 115521, "admitted_skel": 9, "staged_skel": 47536, "mean_count": 131.30078125, "n_util_obs": 187, "util_count_spearman": -0.05946402002871702}, "t": 2653.8}
{"event": "eval", "step": 2200, "phase": 1, "bpb": {"en": 2.09789, "tr": 2.28705, "fi": 2.461}, "bytes_per_chunk": {"en": 5.065, "tr": 4.786, "fi": 4.679}, "bpb_mean": 2.28198, "frozen_inventory": true, "t": 2656.9}
{"event": "train", "step": 2200, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.47818, "bpb": 2.13256, "chunk_loss": 0.030463, "F": 0.2012, "G": 0.2197, "chunks": 206.03, "bytes_per_chunk": 4.97, "boundary_entropy": 0.15579, "router_noise": 0.0, "grad_norm": 0.7214, "lr": 0.0002325, "router_trainable": true, "sec_per_step": 1.351, "probe_sec": 0.0966, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04928, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.05755, "n_shadow": 5, "mean_delta_shadow": -0.00731, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.0741, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.82635, "trunc_abs_frac": 0.41496, "mean_delta_full": 0.07425, "read_hit_rate": {"surface": 0.46639, "split": 0.20439, "merge": 0.09205, "skel": 0.09295}, "mean_gate": {"surface": 0.14807, "split": 0.08905, "merge": 0.12771, "skel": 0.10766}, "n_reads": 6502, "n_cand": 29776, "read_suppressed_frac": 0.1015, "dropout_reads_logged": 6502, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 887, "admitted_surface": 31453, "staged_surface": 53283, "admitted_split": 28684, "staged_split": 45088, "admitted_merge": 5390, "staged_merge": 116067, "admitted_skel": 9, "staged_skel": 47706, "mean_count": 131.39999389648438, "n_util_obs": 187, "util_count_spearman": -0.05930711561516101}, "probe_overhead_sec_total": 352.532, "probe_overhead_seq_frac": 0.4998, "alpha": 0.03, "t": 2662.8}
{"event": "consolidate", "step": 2250, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262069, "t": 2723.4}
{"event": "train", "step": 2250, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.49479, "bpb": 2.15653, "chunk_loss": 0.030431, "F": 0.19962, "G": 0.21634, "chunks": 204.41, "bytes_per_chunk": 5.01, "boundary_entropy": 0.15371, "router_noise": 0.0, "grad_norm": 0.7353, "lr": 0.0002293, "router_trainable": true, "sec_per_step": 1.227, "probe_sec": 0.2022, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.11703, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.17903, "n_shadow": 5, "mean_delta_shadow": -0.17413, "n_split": 1, "mean_delta_split": 0.54791, "n_merge": 1, "mean_delta_merge": -0.21979, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.73558, "trunc_abs_frac": 1.38404, "mean_delta_full": -0.13965, "read_hit_rate": {"surface": 0.45383, "split": 0.20806, "merge": 0.08893, "skel": 0.08843}, "mean_gate": {"surface": 0.15114, "split": 0.0893, "merge": 0.12948, "skel": 0.10792}, "n_reads": 6350, "n_cand": 29427, "read_suppressed_frac": 0.1028, "dropout_reads_logged": 6350, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 912, "admitted_surface": 31936, "staged_surface": 53037, "admitted_split": 27665, "staged_split": 45074, "admitted_merge": 5926, "staged_merge": 116648, "admitted_skel": 9, "staged_skel": 47385, "mean_count": 137.38470458984375, "n_util_obs": 190, "util_count_spearman": -0.07292145977639708}, "probe_overhead_sec_total": 361.849, "probe_overhead_seq_frac": 0.4998, "alpha": 0.03, "t": 2724.2}
{"event": "consolidate", "step": 2300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262070, "t": 2778.8}
{"event": "train", "step": 2300, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.51481, "bpb": 2.18541, "chunk_loss": 0.030383, "F": 0.19788, "G": 0.21674, "chunks": 202.62, "bytes_per_chunk": 5.054, "boundary_entropy": 0.15873, "router_noise": 0.0, "grad_norm": 0.7308, "lr": 0.000226, "router_trainable": true, "sec_per_step": 1.105, "probe_sec": 0.1723, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.11103, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": 0.01924, "n_shadow": 9, "mean_delta_shadow": -0.00905, "n_split": 1, "mean_delta_split": 0.01477, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.0516, "trunc_abs_frac": 0.64467, "mean_delta_full": 0.35254, "read_hit_rate": {"surface": 0.44916, "split": 0.20953, "merge": 0.10343, "skel": 0.10042}, "mean_gate": {"surface": 0.14201, "split": 0.08428, "merge": 0.12191, "skel": 0.10136}, "n_reads": 6456, "n_cand": 29261, "read_suppressed_frac": 0.0987, "dropout_reads_logged": 6456, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 942, "admitted_surface": 31940, "staged_surface": 53580, "admitted_split": 26648, "staged_split": 45024, "admitted_merge": 6938, "staged_merge": 115906, "admitted_skel": 10, "staged_skel": 47634, "mean_count": 144.34779357910156, "n_util_obs": 196, "util_count_spearman": -0.08779733681836654}, "probe_overhead_sec_total": 369.641, "probe_overhead_seq_frac": 0.4998, "alpha": 0.03, "t": 2779.4}
{"event": "consolidate", "step": 2350, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262072, "t": 2837.9}
{"event": "train", "step": 2350, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.56233, "bpb": 2.25397, "chunk_loss": 0.03055, "F": 0.20837, "G": 0.22289, "chunks": 213.38, "bytes_per_chunk": 4.799, "boundary_entropy": 0.15276, "router_noise": 0.0, "grad_norm": 0.7886, "lr": 0.0002226, "router_trainable": true, "sec_per_step": 1.186, "probe_sec": 0.1281, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07233, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": 0.0358, "n_shadow": 3, "mean_delta_shadow": 0.02381, "n_split": 2, "mean_delta_split": 0.07718, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.69314, "trunc_abs_frac": 0.74913, "mean_delta_full": 0.11218, "read_hit_rate": {"surface": 0.45909, "split": 0.21325, "merge": 0.11221, "skel": 0.10684}, "mean_gate": {"surface": 0.14227, "split": 0.08477, "merge": 0.11906, "skel": 0.0983}, "n_reads": 6956, "n_cand": 30544, "read_suppressed_frac": 0.1009, "dropout_reads_logged": 6956, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 980, "admitted_surface": 32278, "staged_surface": 53583, "admitted_split": 25635, "staged_split": 45099, "admitted_merge": 7613, "staged_merge": 115816, "admitted_skel": 10, "staged_skel": 47646, "mean_count": 152.03518676757812, "n_util_obs": 201, "util_count_spearman": -0.1011881188118812}, "probe_overhead_sec_total": 378.674, "probe_overhead_seq_frac": 0.4998, "alpha": 0.03, "t": 2838.7}
{"event": "consolidate", "step": 2400, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262071, "t": 2894.3}
{"event": "probe", "step": 2400, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3436.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.197437, "cps": 26.84375, "mean_p": 0.216735, "boundary_entropy": 0.156846, "bpb": 2.061659}, "tr": {"chunks": 4154.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.229465, "cps": 32.453125, "mean_p": 0.217575, "boundary_entropy": 0.143632, "bpb": 2.243499}, "fi": {"chunks": 4346.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.231035, "cps": 33.953125, "mean_p": 0.239332, "boundary_entropy": 0.158708, "bpb": 2.400101}}, "bpb": {"en": 2.06166, "tr": 2.2435, "fi": 2.4001}, "gini_cpc": 0.034044, "gini_cps": 0.050827, "bpic": 4.753016, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.1622161493824275, "fi": 1.1701695791523166}, "inventory": {"size": 326655, "admitted": 65536, "staged": 261119, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1019, "admitted_surface": 33285, "staged_surface": 52470, "admitted_split": 24620, "staged_split": 44983, "admitted_merge": 7621, "staged_merge": 115684, "admitted_skel": 10, "staged_skel": 47982, "mean_count": 157.80636596679688, "n_util_obs": 205, "util_count_spearman": -0.12733768856820493}, "t": 2897.1}
{"event": "eval", "step": 2400, "phase": 1, "bpb": {"en": 2.07395, "tr": 2.25145, "fi": 2.40952}, "bytes_per_chunk": {"en": 5.048, "tr": 4.809, "fi": 4.741}, "bpb_mean": 2.24497, "frozen_inventory": true, "t": 2902.9}
{"event": "phase_transition", "step": 2400, "phase": 2, "objective": "global", "router_trainable": true, "encoder_trainable": true, "frozen_router_params": 0, "t": 2908.4}
{"event": "train", "step": 2400, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.5009, "bpb": 2.16534, "chunk_loss": 0.030531, "F": 0.20651, "G": 0.22119, "chunks": 211.47, "bytes_per_chunk": 4.842, "boundary_entropy": 0.14809, "router_noise": 0.0, "grad_norm": 0.7421, "lr": 0.0002192, "router_trainable": true, "sec_per_step": 1.404, "probe_sec": 0.1257, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07342, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 11, "mean_delta_loo": 0.02779, "n_shadow": 2, "mean_delta_shadow": -0.02591, "n_split": 1, "mean_delta_split": 0.00543, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": -0.21398, "trunc_abs_frac": 0.21398, "mean_delta_full": 0.45203, "read_hit_rate": {"surface": 0.45479, "split": 0.21271, "merge": 0.10264, "skel": 0.10019}, "mean_gate": {"surface": 0.14428, "split": 0.08568, "merge": 0.12228, "skel": 0.10103}, "n_reads": 6799, "n_cand": 30518, "read_suppressed_frac": 0.0984, "dropout_reads_logged": 6799, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1020, "admitted_surface": 33285, "staged_surface": 52693, "admitted_split": 24620, "staged_split": 45140, "admitted_merge": 7621, "staged_merge": 116265, "admitted_skel": 10, "staged_skel": 48046, "mean_count": 157.91012573242188, "n_util_obs": 205, "util_count_spearman": -0.12877175717422087}, "probe_overhead_sec_total": 387.134, "probe_overhead_seq_frac": 0.4998, "alpha": 0.03, "t": 2908.9}
{"event": "consolidate", "step": 2450, "retired": 0, "split_retired": 3, "merge_admitted": 0, "admitted": 3, "swapped": 1024, "uncertain_staged": 262071, "t": 2965.7}
{"event": "train", "step": 2450, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.46517, "bpb": 2.1138, "chunk_loss": 0.030415, "F": 0.20029, "G": 0.2151, "chunks": 205.09, "bytes_per_chunk": 4.993, "boundary_entropy": 0.14487, "router_noise": 0.0, "grad_norm": 0.731, "lr": 0.0002158, "router_trainable": true, "sec_per_step": 1.161, "probe_sec": 0.0829, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0402, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.02591, "n_shadow": 2, "mean_delta_shadow": -0.02588, "n_split": 1, "mean_delta_split": 0.19278, "n_merge": 2, "mean_delta_merge": 0.03091, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 6.59954, "trunc_abs_frac": 1.37949, "mean_delta_full": 0.01339, "read_hit_rate": {"surface": 0.47481, "split": 0.2364, "merge": 0.11371, "skel": 0.09834}, "mean_gate": {"surface": 0.14775, "split": 0.08487, "merge": 0.12218, "skel": 0.10283}, "n_reads": 7061, "n_cand": 29571, "read_suppressed_frac": 0.097, "dropout_reads_logged": 7061, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1051, "admitted_surface": 34248, "staged_surface": 52252, "admitted_split": 23663, "staged_split": 44875, "admitted_merge": 7615, "staged_merge": 117024, "admitted_skel": 10, "staged_skel": 47993, "mean_count": 167.04051208496094, "n_util_obs": 209, "util_count_spearman": -0.1571468005678532}, "probe_overhead_sec_total": 396.376, "probe_overhead_seq_frac": 0.4998, "alpha": 0.03, "t": 2966.9}
{"event": "consolidate", "step": 2500, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262072, "t": 3028.1}
{"event": "train", "step": 2500, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.439, "bpb": 2.07603, "chunk_loss": 0.030517, "F": 0.20532, "G": 0.22177, "chunks": 210.25, "bytes_per_chunk": 4.87, "boundary_entropy": 0.14526, "router_noise": 0.0, "grad_norm": 0.7205, "lr": 0.0002123, "router_trainable": true, "sec_per_step": 1.24, "probe_sec": 0.2189, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08459, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.02625, "n_shadow": 6, "mean_delta_shadow": -0.07662, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.07575, "trunc_abs_frac": 0.92448, "mean_delta_full": 0.14865, "read_hit_rate": {"surface": 0.50358, "split": 0.24835, "merge": 0.11525, "skel": 0.10108}, "mean_gate": {"surface": 0.14493, "split": 0.08181, "merge": 0.11495, "skel": 0.09933}, "n_reads": 7594, "n_cand": 30330, "read_suppressed_frac": 0.1023, "dropout_reads_logged": 7594, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1075, "admitted_surface": 35206, "staged_surface": 52213, "admitted_split": 22646, "staged_split": 44979, "admitted_merge": 7674, "staged_merge": 116892, "admitted_skel": 10, "staged_skel": 48060, "mean_count": 171.3274688720703, "n_util_obs": 210, "util_count_spearman": -0.14573836975195206}, "probe_overhead_sec_total": 404.842, "probe_overhead_seq_frac": 0.4998, "alpha": 0.03, "t": 3028.9}
{"event": "consolidate", "step": 2550, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262071, "t": 3079.0}
{"event": "train", "step": 2550, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.46063, "bpb": 2.10724, "chunk_loss": 0.030427, "F": 0.20343, "G": 0.21698, "chunks": 208.31, "bytes_per_chunk": 4.916, "boundary_entropy": 0.14261, "router_noise": 0.0, "grad_norm": 0.7241, "lr": 0.0002088, "router_trainable": true, "sec_per_step": 1.016, "probe_sec": 0.12, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05783, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": 0.09854, "n_shadow": 2, "mean_delta_shadow": 0.0742, "n_split": 1, "mean_delta_split": -0.15997, "n_merge": 1, "mean_delta_merge": -0.39117, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 13.04167, "trunc_abs_frac": 1.4271, "mean_delta_full": 0.00128, "read_hit_rate": {"surface": 0.48674, "split": 0.22354, "merge": 0.12299, "skel": 0.10888}, "mean_gate": {"surface": 0.14499, "split": 0.08126, "merge": 0.11513, "skel": 0.099}, "n_reads": 7248, "n_cand": 30235, "read_suppressed_frac": 0.1013, "dropout_reads_logged": 7248, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1102, "admitted_surface": 36044, "staged_surface": 51803, "admitted_split": 21628, "staged_split": 45009, "admitted_merge": 7854, "staged_merge": 117459, "admitted_skel": 10, "staged_skel": 47873, "mean_count": 175.6464385986328, "n_util_obs": 215, "util_count_spearman": -0.12013273688472162}, "probe_overhead_sec_total": 411.84, "probe_overhead_seq_frac": 0.4998, "alpha": 0.03, "t": 3079.7}
{"event": "consolidate", "step": 2600, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262070, "t": 3139.4}
{"event": "probe", "step": 2600, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3450.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.198242, "cps": 26.953125, "mean_p": 0.21599, "boundary_entropy": 0.148935, "bpb": 2.043422}, "tr": {"chunks": 4250.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.234768, "cps": 33.203125, "mean_p": 0.2206, "boundary_entropy": 0.135105, "bpb": 2.227445}, "fi": {"chunks": 4355.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.231513, "cps": 34.023438, "mean_p": 0.238528, "boundary_entropy": 0.155399, "bpb": 2.377082}}, "bpb": {"en": 2.04342, "tr": 2.22745, "fi": 2.37708}, "gini_cpc": 0.036644, "gini_cps": 0.050048, "bpic": 4.706097, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.1842500282201605, "fi": 1.1678345001652595}, "inventory": {"size": 326655, "admitted": 65536, "staged": 261119, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1133, "admitted_surface": 36930, "staged_surface": 51429, "admitted_split": 20615, "staged_split": 44714, "admitted_merge": 7981, "staged_merge": 117281, "admitted_skel": 10, "staged_skel": 47695, "mean_count": 181.3228302001953, "n_util_obs": 218, "util_count_spearman": -0.14589652098925734}, "t": 3141.9}
{"event": "eval", "step": 2600, "phase": 2, "bpb": {"en": 2.05586, "tr": 2.23787, "fi": 2.38346}, "bytes_per_chunk": {"en": 5.053, "tr": 4.714, "fi": 4.822}, "bpb_mean": 2.22573, "frozen_inventory": true, "t": 3145.6}
{"event": "train", "step": 2600, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.42306, "bpb": 2.05304, "chunk_loss": 0.030434, "F": 0.20169, "G": 0.21682, "chunks": 206.53, "bytes_per_chunk": 4.958, "boundary_entropy": 0.14382, "router_noise": 0.0, "grad_norm": 0.7115, "lr": 0.0002053, "router_trainable": true, "sec_per_step": 1.423, "probe_sec": 0.1091, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06013, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": 0.05819, "n_shadow": 5, "mean_delta_shadow": 0.00223, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -0.03725, "trunc_abs_frac": 14.57405, "mean_delta_full": -0.04865, "read_hit_rate": {"surface": 0.47894, "split": 0.23963, "merge": 0.11077, "skel": 0.09584}, "mean_gate": {"surface": 0.14671, "split": 0.08239, "merge": 0.11336, "skel": 0.10103}, "n_reads": 7105, "n_cand": 29600, "read_suppressed_frac": 0.1023, "dropout_reads_logged": 7105, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1133, "admitted_surface": 36930, "staged_surface": 51661, "admitted_split": 20615, "staged_split": 44801, "admitted_merge": 7981, "staged_merge": 117805, "admitted_skel": 10, "staged_skel": 47877, "mean_count": 181.43124389648438, "n_util_obs": 218, "util_count_spearman": -0.14473826330879985}, "probe_overhead_sec_total": 420.097, "probe_overhead_seq_frac": 0.4998, "alpha": 0.03, "t": 3150.9}
{"event": "consolidate", "step": 2650, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262069, "t": 3207.8}
{"event": "train", "step": 2650, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.46532, "bpb": 2.11401, "chunk_loss": 0.030392, "F": 0.20117, "G": 0.21284, "chunks": 206.0, "bytes_per_chunk": 4.971, "boundary_entropy": 0.13986, "router_noise": 0.0, "grad_norm": 0.6859, "lr": 0.0002017, "router_trainable": true, "sec_per_step": 1.175, "probe_sec": 0.1836, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.15141, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.13062, "n_shadow": 3, "mean_delta_shadow": 0.00502, "n_split": 1, "mean_delta_split": -0.02295, "n_merge": 1, "mean_delta_merge": 0.53491, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 4.42424, "trunc_abs_frac": 4.42424, "mean_delta_full": 0.00806, "read_hit_rate": {"surface": 0.47942, "split": 0.24331, "merge": 0.12056, "skel": 0.09354}, "mean_gate": {"surface": 0.14753, "split": 0.0818, "merge": 0.11644, "skel": 0.10457}, "n_reads": 7220, "n_cand": 29681, "read_suppressed_frac": 0.1012, "dropout_reads_logged": 7220, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1173, "admitted_surface": 37503, "staged_surface": 51984, "admitted_split": 19608, "staged_split": 45035, "admitted_merge": 8413, "staged_merge": 117125, "admitted_skel": 12, "staged_skel": 48000, "mean_count": 186.61227416992188, "n_util_obs": 224, "util_count_spearman": -0.13293618752301017}, "probe_overhead_sec_total": 428.2, "probe_overhead_seq_frac": 0.4998, "alpha": 0.03, "t": 3209.7}
{"event": "consolidate", "step": 2700, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262071, "t": 3265.9}
{"event": "train", "step": 2700, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.44841, "bpb": 2.08962, "chunk_loss": 0.030496, "F": 0.203, "G": 0.2178, "chunks": 207.88, "bytes_per_chunk": 4.926, "boundary_entropy": 0.14111, "router_noise": 0.0, "grad_norm": 0.7088, "lr": 0.0001981, "router_trainable": true, "sec_per_step": 1.151, "probe_sec": 0.5558, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.32753, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": 0.01757, "n_shadow": 8, "mean_delta_shadow": 0.05321, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.05103, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.25921, "trunc_abs_frac": 0.50515, "mean_delta_full": 0.19392, "read_hit_rate": {"surface": 0.47523, "split": 0.23997, "merge": 0.10959, "skel": 0.10797}, "mean_gate": {"surface": 0.14037, "split": 0.07618, "merge": 0.1105, "skel": 0.09817}, "n_reads": 7209, "n_cand": 29897, "read_suppressed_frac": 0.1026, "dropout_reads_logged": 7209, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1194, "admitted_surface": 38060, "staged_surface": 52219, "admitted_split": 18587, "staged_split": 45045, "admitted_merge": 8876, "staged_merge": 116847, "admitted_skel": 13, "staged_skel": 48033, "mean_count": 192.01742553710938, "n_util_obs": 225, "util_count_spearman": -0.11912557943531395}, "probe_overhead_sec_total": 437.344, "probe_overhead_seq_frac": 0.4998, "alpha": 0.03, "t": 3267.2}
{"event": "consolidate", "step": 2750, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262072, "t": 3335.0}
{"event": "train", "step": 2750, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.45648, "bpb": 2.10126, "chunk_loss": 0.030534, "F": 0.20578, "G": 0.21841, "chunks": 210.72, "bytes_per_chunk": 4.86, "boundary_entropy": 0.13996, "router_noise": 0.0, "grad_norm": 0.6832, "lr": 0.0001945, "router_trainable": true, "sec_per_step": 1.377, "probe_sec": 0.1951, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.11098, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.00677, "n_shadow": 8, "mean_delta_shadow": 0.01569, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 0.03562, "trunc_abs_frac": 0.62799, "mean_delta_full": -0.02384, "read_hit_rate": {"surface": 0.48875, "split": 0.25688, "merge": 0.11529, "skel": 0.10204}, "mean_gate": {"surface": 0.14818, "split": 0.07748, "merge": 0.11238, "skel": 0.10272}, "n_reads": 7543, "n_cand": 29971, "read_suppressed_frac": 0.1017, "dropout_reads_logged": 7543, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1227, "admitted_surface": 38201, "staged_surface": 52556, "admitted_split": 17587, "staged_split": 44828, "admitted_merge": 9735, "staged_merge": 116904, "admitted_skel": 13, "staged_skel": 47856, "mean_count": 196.8250274658203, "n_util_obs": 228, "util_count_spearman": -0.1215093905976868}, "probe_overhead_sec_total": 446.777, "probe_overhead_seq_frac": 0.4998, "alpha": 0.03, "t": 3336.1}
{"event": "consolidate", "step": 2800, "retired": 0, "split_retired": 3, "merge_admitted": 0, "admitted": 3, "swapped": 1024, "uncertain_staged": 262073, "t": 3394.7}
{"event": "probe", "step": 2800, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3445.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.197954, "cps": 26.914062, "mean_p": 0.21309, "boundary_entropy": 0.142993, "bpb": 2.025366}, "tr": {"chunks": 4138.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.228581, "cps": 32.328125, "mean_p": 0.217004, "boundary_entropy": 0.130829, "bpb": 2.185662}, "fi": {"chunks": 4314.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.229334, "cps": 33.703125, "mean_p": 0.237885, "boundary_entropy": 0.151486, "bpb": 2.333613}}, "bpb": {"en": 2.02537, "tr": 2.18566, "fi": 2.33361}, "gini_cpc": 0.031896, "gini_cps": 0.048696, "bpic": 4.768597, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.1547150569708073, "fi": 1.158518974823967}, "inventory": {"size": 326653, "admitted": 65536, "staged": 261117, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1258, "admitted_surface": 39099, "staged_surface": 51882, "admitted_split": 16594, "staged_split": 45024, "admitted_merge": 9830, "staged_merge": 116324, "admitted_skel": 13, "staged_skel": 47887, "mean_count": 202.2274932861328, "n_util_obs": 232, "util_count_spearman": -0.13072926777233423}, "t": 3396.8}
{"event": "eval", "step": 2800, "phase": 2, "bpb": {"en": 2.03705, "tr": 2.19449, "fi": 2.33986}, "bytes_per_chunk": {"en": 5.062, "tr": 4.845, "fi": 4.848}, "bpb_mean": 2.19047, "frozen_inventory": true, "t": 3400.2}
{"event": "train", "step": 2800, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.411, "bpb": 2.03564, "chunk_loss": 0.030419, "F": 0.20135, "G": 0.21444, "chunks": 206.19, "bytes_per_chunk": 4.966, "boundary_entropy": 0.13615, "router_noise": 0.0, "grad_norm": 0.6891, "lr": 0.0001909, "router_trainable": true, "sec_per_step": 1.413, "probe_sec": 0.2277, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.15815, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": -0.02832, "n_shadow": 10, "mean_delta_shadow": -0.02558, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.04138, "trunc_abs_frac": 0.68079, "mean_delta_full": -0.10116, "read_hit_rate": {"surface": 0.49558, "split": 0.26945, "merge": 0.11907, "skel": 0.103}, "mean_gate": {"surface": 0.14804, "split": 0.07807, "merge": 0.11547, "skel": 0.10321}, "n_reads": 7678, "n_cand": 29762, "read_suppressed_frac": 0.0966, "dropout_reads_logged": 7678, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1258, "admitted_surface": 39099, "staged_surface": 52024, "admitted_split": 16594, "staged_split": 45142, "admitted_merge": 9830, "staged_merge": 116936, "admitted_skel": 13, "staged_skel": 48042, "mean_count": 202.3446502685547, "n_util_obs": 232, "util_count_spearman": -0.13034922186257}, "probe_overhead_sec_total": 455.468, "probe_overhead_seq_frac": 0.4998, "alpha": 0.03, "t": 3406.7}
{"event": "consolidate", "step": 2850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262072, "t": 3463.3}
{"event": "train", "step": 2850, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.46539, "bpb": 2.11411, "chunk_loss": 0.030547, "F": 0.20734, "G": 0.22237, "chunks": 212.31, "bytes_per_chunk": 4.823, "boundary_entropy": 0.13654, "router_noise": 0.0, "grad_norm": 0.7365, "lr": 0.0001872, "router_trainable": true, "sec_per_step": 1.156, "probe_sec": 0.1836, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07591, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.06598, "n_shadow": 2, "mean_delta_shadow": 0.09564, "n_split": 1, "mean_delta_split": 0.0022, "n_merge": 2, "mean_delta_merge": -0.06754, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.18305, "trunc_abs_frac": 0.14034, "mean_delta_full": 0.12091, "read_hit_rate": {"surface": 0.47959, "split": 0.266, "merge": 0.11085, "skel": 0.13799}, "mean_gate": {"surface": 0.14205, "split": 0.07286, "merge": 0.10834, "skel": 0.10655}, "n_reads": 7909, "n_cand": 30799, "read_suppressed_frac": 0.0963, "dropout_reads_logged": 7909, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1283, "admitted_surface": 40085, "staged_surface": 51868, "admitted_split": 15573, "staged_split": 44810, "admitted_merge": 9864, "staged_merge": 117084, "admitted_skel": 14, "staged_skel": 48382, "mean_count": 207.36524963378906, "n_util_obs": 236, "util_count_spearman": -0.1327320255291853}, "probe_overhead_sec_total": 464.201, "probe_overhead_seq_frac": 0.4998, "alpha": 0.03, "t": 3464.5}
{"event": "consolidate", "step": 2900, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 1024, "uncertain_staged": 262073, "t": 3522.8}
{"event": "train", "step": 2900, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.39199, "bpb": 2.00821, "chunk_loss": 0.030338, "F": 0.19803, "G": 0.21201, "chunks": 202.78, "bytes_per_chunk": 5.05, "boundary_entropy": 0.13763, "router_noise": 0.0, "grad_norm": 0.6548, "lr": 0.0001835, "router_trainable": true, "sec_per_step": 1.18, "probe_sec": 0.1365, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05256, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": -0.0326, "n_shadow": 5, "mean_delta_shadow": -0.03382, "n_split": 1, "mean_delta_split": -0.04129, "n_merge": 1, "mean_delta_merge": -0.18274, "n_dropped_unpaired": 0, "n_full": 7, "trunc_ratio_sum": 0.10567, "trunc_abs_frac": 1.20634, "mean_delta_full": -0.0354, "read_hit_rate": {"surface": 0.49419, "split": 0.25429, "merge": 0.12685, "skel": 0.10276}, "mean_gate": {"surface": 0.14615, "split": 0.07404, "merge": 0.10965, "skel": 0.10499}, "n_reads": 7375, "n_cand": 29007, "read_suppressed_frac": 0.1021, "dropout_reads_logged": 7375, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1316, "admitted_surface": 41103, "staged_surface": 51533, "admitted_split": 14684, "staged_split": 44619, "admitted_merge": 9736, "staged_merge": 118077, "admitted_skel": 13, "staged_skel": 47915, "mean_count": 210.5399627685547, "n_util_obs": 235, "util_count_spearman": -0.12198538146813956}, "probe_overhead_sec_total": 472.472, "probe_overhead_seq_frac": 0.4998, "alpha": 0.03, "t": 3523.5}
{"event": "consolidate", "step": 2950, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 1024, "uncertain_staged": 262073, "t": 3581.0}
{"event": "train", "step": 2950, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.54235, "bpb": 2.22515, "chunk_loss": 0.03061, "F": 0.21152, "G": 0.22206, "chunks": 216.59, "bytes_per_chunk": 4.728, "boundary_entropy": 0.1338, "router_noise": 0.0, "grad_norm": 0.7232, "lr": 0.0001799, "router_trainable": true, "sec_per_step": 1.162, "probe_sec": 0.1118, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05219, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.08018, "n_shadow": 5, "mean_delta_shadow": 0.04203, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.14856, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.02574, "trunc_abs_frac": 0.68768, "mean_delta_full": 0.32776, "read_hit_rate": {"surface": 0.45209, "split": 0.23599, "merge": 0.10689, "skel": 0.10506}, "mean_gate": {"surface": 0.14114, "split": 0.07269, "merge": 0.11228, "skel": 0.09761}, "n_reads": 7197, "n_cand": 30698, "read_suppressed_frac": 0.1035, "dropout_reads_logged": 7197, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1339, "admitted_surface": 42123, "staged_surface": 51512, "admitted_split": 13974, "staged_split": 45451, "admitted_merge": 9427, "staged_merge": 117147, "admitted_skel": 12, "staged_skel": 48034, "mean_count": 214.20724487304688, "n_util_obs": 236, "util_count_spearman": -0.11288877669589432}, "probe_overhead_sec_total": 480.296, "probe_overhead_seq_frac": 0.4998, "alpha": 0.03, "t": 3581.6}
{"event": "consolidate", "step": 3000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262073, "t": 3637.8}
{"event": "probe", "step": 3000, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3481.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.200023, "cps": 27.195312, "mean_p": 0.214487, "boundary_entropy": 0.136194, "bpb": 2.008876}, "tr": {"chunks": 4180.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.230901, "cps": 32.65625, "mean_p": 0.216262, "boundary_entropy": 0.126369, "bpb": 2.163654}, "fi": {"chunks": 4276.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.227314, "cps": 33.40625, "mean_p": 0.235149, "boundary_entropy": 0.139664, "bpb": 2.316021}}, "bpb": {"en": 2.00888, "tr": 2.16365, "fi": 2.31602}, "gini_cpc": 0.031273, "gini_cps": 0.0444, "bpic": 4.752618, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.154372114636628, "fi": 1.1364384261709994}, "inventory": {"size": 326656, "admitted": 65536, "staged": 261120, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1357, "admitted_surface": 43138, "staged_surface": 51058, "admitted_split": 13035, "staged_split": 45164, "admitted_merge": 9351, "staged_merge": 117087, "admitted_skel": 12, "staged_skel": 47811, "mean_count": 219.17330932617188, "n_util_obs": 241, "util_count_spearman": -0.10885176642252913}, "t": 3640.1}
{"event": "eval", "step": 3000, "phase": 2, "bpb": {"en": 2.02178, "tr": 2.17245, "fi": 2.31836}, "bytes_per_chunk": {"en": 5.018, "tr": 4.782, "fi": 4.823}, "bpb_mean": 2.17086, "frozen_inventory": true, "t": 3643.9}
{"event": "train", "step": 3000, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.49014, "bpb": 2.14981, "chunk_loss": 0.030424, "F": 0.19897, "G": 0.21241, "chunks": 203.75, "bytes_per_chunk": 5.026, "boundary_entropy": 0.13145, "router_noise": 0.0, "grad_norm": 0.7358, "lr": 0.0001761, "router_trainable": true, "sec_per_step": 1.43, "probe_sec": 0.1614, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08442, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": 0.07024, "n_shadow": 7, "mean_delta_shadow": -0.07027, "n_split": 1, "mean_delta_split": 0.03339, "n_merge": 1, "mean_delta_merge": -0.14398, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.63944, "trunc_abs_frac": 0.65752, "mean_delta_full": 0.14926, "read_hit_rate": {"surface": 0.47796, "split": 0.25719, "merge": 0.1044, "skel": 0.10054}, "mean_gate": {"surface": 0.14434, "split": 0.0715, "merge": 0.10886, "skel": 0.10151}, "n_reads": 7230, "n_cand": 29431, "read_suppressed_frac": 0.1007, "dropout_reads_logged": 7230, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1358, "admitted_surface": 43138, "staged_surface": 51356, "admitted_split": 13035, "staged_split": 45357, "admitted_merge": 9351, "staged_merge": 117486, "admitted_skel": 12, "staged_skel": 47945, "mean_count": 219.2836456298828, "n_util_obs": 241, "util_count_spearman": -0.1093772504372278}, "probe_overhead_sec_total": 488.206, "probe_overhead_seq_frac": 0.4998, "alpha": 0.03, "t": 3653.1}
{"event": "consolidate", "step": 3050, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 1024, "uncertain_staged": 262073, "t": 3714.5}
{"event": "train", "step": 3050, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.4438, "bpb": 2.08296, "chunk_loss": 0.030496, "F": 0.20322, "G": 0.21507, "chunks": 208.09, "bytes_per_chunk": 4.921, "boundary_entropy": 0.1291, "router_noise": 0.0, "grad_norm": 0.6857, "lr": 0.0001724, "router_trainable": true, "sec_per_step": 1.24, "probe_sec": 0.1293, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09589, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.07855, "n_shadow": 1, "mean_delta_shadow": -0.17645, "n_split": 2, "mean_delta_split": 0.04369, "n_merge": 1, "mean_delta_merge": 0.07117, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 4.27904, "trunc_abs_frac": 0.47315, "mean_delta_full": 0.0134, "read_hit_rate": {"surface": 0.46145, "split": 0.26682, "merge": 0.10796, "skel": 0.11251}, "mean_gate": {"surface": 0.14869, "split": 0.07174, "merge": 0.1082, "skel": 0.10928}, "n_reads": 7458, "n_cand": 29979, "read_suppressed_frac": 0.1022, "dropout_reads_logged": 7458, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1389, "admitted_surface": 44149, "staged_surface": 51235, "admitted_split": 12348, "staged_split": 45251, "admitted_merge": 9026, "staged_merge": 117578, "admitted_skel": 13, "staged_skel": 48080, "mean_count": 222.90174865722656, "n_util_obs": 244, "util_count_spearman": -0.11113137287985667}, "probe_overhead_sec_total": 496.718, "probe_overhead_seq_frac": 0.4998, "alpha": 0.03, "t": 3715.1}
{"event": "consolidate", "step": 3100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262073, "t": 3765.4}
{"event": "train", "step": 3100, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.48384, "bpb": 2.14072, "chunk_loss": 0.03049, "F": 0.20605, "G": 0.21612, "chunks": 211.0, "bytes_per_chunk": 4.853, "boundary_entropy": 0.12258, "router_noise": 0.0, "grad_norm": 0.6897, "lr": 0.0001687, "router_trainable": true, "sec_per_step": 1.016, "probe_sec": 0.0732, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04828, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 3, "mean_delta_loo": 0.0717, "n_shadow": 7, "mean_delta_shadow": -0.01918, "n_split": 3, "mean_delta_split": -0.06396, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.56691, "trunc_abs_frac": 0.37597, "mean_delta_full": 0.2431, "read_hit_rate": {"surface": 0.50193, "split": 0.25115, "merge": 0.12664, "skel": 0.10309}, "mean_gate": {"surface": 0.14972, "split": 0.07175, "merge": 0.1183, "skel": 0.10593}, "n_reads": 7689, "n_cand": 30154, "read_suppressed_frac": 0.1001, "dropout_reads_logged": 7689, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1408, "admitted_surface": 45170, "staged_surface": 51228, "admitted_split": 11355, "staged_split": 45350, "admitted_merge": 8998, "staged_merge": 117302, "admitted_skel": 13, "staged_skel": 48264, "mean_count": 228.77505493164062, "n_util_obs": 248, "util_count_spearman": -0.11261763784929453}, "probe_overhead_sec_total": 504.099, "probe_overhead_seq_frac": 0.4998, "alpha": 0.03, "t": 3765.9}
{"event": "consolidate", "step": 3150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262075, "t": 3821.2}
{"event": "train", "step": 3150, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.45528, "bpb": 2.09952, "chunk_loss": 0.030362, "F": 0.19916, "G": 0.21062, "chunks": 203.94, "bytes_per_chunk": 5.021, "boundary_entropy": 0.13266, "router_noise": 0.0, "grad_norm": 0.7123, "lr": 0.000165, "router_trainable": true, "sec_per_step": 1.123, "probe_sec": 0.1389, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0568, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": -0.0533, "n_shadow": 4, "mean_delta_shadow": 0.04356, "n_split": 2, "mean_delta_split": -0.17224, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 12.4275, "trunc_abs_frac": 0.99994, "mean_delta_full": -0.0061, "read_hit_rate": {"surface": 0.50123, "split": 0.24991, "merge": 0.11204, "skel": 0.11109}, "mean_gate": {"surface": 0.14544, "split": 0.06964, "merge": 0.10974, "skel": 0.10476}, "n_reads": 7400, "n_cand": 29435, "read_suppressed_frac": 0.1001, "dropout_reads_logged": 7400, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1447, "admitted_surface": 46149, "staged_surface": 51153, "admitted_split": 10335, "staged_split": 45304, "admitted_merge": 9039, "staged_merge": 117496, "admitted_skel": 13, "staged_skel": 48191, "mean_count": 233.13768005371094, "n_util_obs": 251, "util_count_spearman": -0.12770027085947316}, "probe_overhead_sec_total": 512.172, "probe_overhead_seq_frac": 0.4998, "alpha": 0.03, "t": 3822.1}
{"event": "consolidate", "step": 3200, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 1024, "uncertain_staged": 262076, "t": 3876.8}
{"event": "probe", "step": 3200, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3446.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.198012, "cps": 26.921875, "mean_p": 0.211363, "boundary_entropy": 0.130186, "bpb": 1.992428}, "tr": {"chunks": 4161.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.229851, "cps": 32.507812, "mean_p": 0.215796, "boundary_entropy": 0.125167, "bpb": 2.141545}, "fi": {"chunks": 4367.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.232151, "cps": 34.117188, "mean_p": 0.236914, "boundary_entropy": 0.136321, "bpb": 2.292139}}, "bpb": {"en": 1.99243, "tr": 2.14154, "fi": 2.29214}, "gini_cpc": 0.034484, "gini_cps": 0.051278, "bpic": 4.737932, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.160796290165109, "fi": 1.1724117317780594}, "inventory": {"size": 326654, "admitted": 65536, "staged": 261118, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1485, "admitted_surface": 47107, "staged_surface": 51062, "admitted_split": 9321, "staged_split": 45188, "admitted_merge": 9095, "staged_merge": 116769, "admitted_skel": 13, "staged_skel": 48099, "mean_count": 238.1237030029297, "n_util_obs": 255, "util_count_spearman": -0.1434800803830067}, "t": 3879.4}
{"event": "eval", "step": 3200, "phase": 2, "bpb": {"en": 2.004, "tr": 2.1496, "fi": 2.29749}, "bytes_per_chunk": {"en": 5.041, "tr": 4.826, "fi": 4.763}, "bpb_mean": 2.15036, "frozen_inventory": true, "t": 3884.3}
{"event": "train", "step": 3200, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.43529, "bpb": 2.07068, "chunk_loss": 0.030394, "F": 0.20068, "G": 0.21251, "chunks": 205.5, "bytes_per_chunk": 4.983, "boundary_entropy": 0.12531, "router_noise": 0.0, "grad_norm": 0.6266, "lr": 0.0001613, "router_trainable": true, "sec_per_step": 1.381, "probe_sec": 0.1304, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07173, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.0172, "n_shadow": 4, "mean_delta_shadow": 0.09715, "n_split": 2, "mean_delta_split": 0.1889, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 2.30466, "trunc_abs_frac": 0.62011, "mean_delta_full": 0.012, "read_hit_rate": {"surface": 0.511, "split": 0.26634, "merge": 0.11425, "skel": 0.1062}, "mean_gate": {"surface": 0.145, "split": 0.06945, "merge": 0.10884, "skel": 0.10144}, "n_reads": 7702, "n_cand": 29643, "read_suppressed_frac": 0.0995, "dropout_reads_logged": 7702, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1483, "admitted_surface": 47107, "staged_surface": 51227, "admitted_split": 9321, "staged_split": 45323, "admitted_merge": 9095, "staged_merge": 117261, "admitted_skel": 13, "staged_skel": 48333, "mean_count": 238.24122619628906, "n_util_obs": 256, "util_count_spearman": -0.13383790654215555}, "probe_overhead_sec_total": 520.207, "probe_overhead_seq_frac": 0.4998, "alpha": 0.03, "t": 3891.1}
{"event": "consolidate", "step": 3250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262075, "t": 3945.1}
{"event": "train", "step": 3250, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.4332, "bpb": 2.06766, "chunk_loss": 0.030494, "F": 0.20444, "G": 0.21425, "chunks": 209.34, "bytes_per_chunk": 4.891, "boundary_entropy": 0.12305, "router_noise": 0.0, "grad_norm": 0.7039, "lr": 0.0001576, "router_trainable": true, "sec_per_step": 1.126, "probe_sec": 0.1211, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06008, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.06332, "n_shadow": 6, "mean_delta_shadow": -0.10751, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.09906, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.80377, "trunc_abs_frac": 0.80377, "mean_delta_full": -0.05505, "read_hit_rate": {"surface": 0.48043, "split": 0.25886, "merge": 0.12118, "skel": 0.09263}, "mean_gate": {"surface": 0.14542, "split": 0.0701, "merge": 0.1099, "skel": 0.10211}, "n_reads": 7446, "n_cand": 29769, "read_suppressed_frac": 0.0998, "dropout_reads_logged": 7446, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1517, "admitted_surface": 48098, "staged_surface": 51179, "admitted_split": 8311, "staged_split": 45497, "admitted_merge": 9114, "staged_merge": 117116, "admitted_skel": 13, "staged_skel": 48352, "mean_count": 243.48838806152344, "n_util_obs": 258, "util_count_spearman": -0.13591707242851314}, "probe_overhead_sec_total": 528.144, "probe_overhead_seq_frac": 0.4998, "alpha": 0.03, "t": 3947.4}
{"event": "consolidate", "step": 3300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262075, "t": 4000.7}
{"event": "train", "step": 3300, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.41436, "bpb": 2.04049, "chunk_loss": 0.030408, "F": 0.20047, "G": 0.21268, "chunks": 205.28, "bytes_per_chunk": 4.988, "boundary_entropy": 0.12859, "router_noise": 0.0, "grad_norm": 0.658, "lr": 0.0001539, "router_trainable": true, "sec_per_step": 1.079, "probe_sec": 0.0851, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04852, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": -0.00064, "n_shadow": 7, "mean_delta_shadow": -0.06368, "n_split": 1, "mean_delta_split": 0.02856, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.08848, "trunc_abs_frac": 0.19775, "mean_delta_full": -0.31549, "read_hit_rate": {"surface": 0.49151, "split": 0.27205, "merge": 0.11145, "skel": 0.11416}, "mean_gate": {"surface": 0.14683, "split": 0.0682, "merge": 0.11024, "skel": 0.10171}, "n_reads": 7641, "n_cand": 29577, "read_suppressed_frac": 0.0965, "dropout_reads_logged": 7641, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1541, "admitted_surface": 49121, "staged_surface": 51075, "admitted_split": 7319, "staged_split": 45296, "admitted_merge": 9083, "staged_merge": 117754, "admitted_skel": 13, "staged_skel": 48019, "mean_count": 249.05381774902344, "n_util_obs": 260, "util_count_spearman": -0.13418541694403763}, "probe_overhead_sec_total": 535.777, "probe_overhead_seq_frac": 0.4998, "alpha": 0.03, "t": 4001.4}
{"event": "consolidate", "step": 3350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262074, "t": 4055.5}
{"event": "train", "step": 3350, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.3555, "bpb": 1.95557, "chunk_loss": 0.030413, "F": 0.20081, "G": 0.21106, "chunks": 205.62, "bytes_per_chunk": 4.98, "boundary_entropy": 0.12336, "router_noise": 0.0, "grad_norm": 0.6988, "lr": 0.0001501, "router_trainable": true, "sec_per_step": 1.11, "probe_sec": 0.1497, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06063, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 11, "mean_delta_loo": 0.06948, "n_shadow": 3, "mean_delta_shadow": -0.02266, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 0.03797, "trunc_abs_frac": 24.82291, "mean_delta_full": -0.20253, "read_hit_rate": {"surface": 0.5029, "split": 0.2656, "merge": 0.12385, "skel": 0.1028}, "mean_gate": {"surface": 0.14951, "split": 0.06858, "merge": 0.10813, "skel": 0.10274}, "n_reads": 7699, "n_cand": 29745, "read_suppressed_frac": 0.1026, "dropout_reads_logged": 7699, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1577, "admitted_surface": 50138, "staged_surface": 51056, "admitted_split": 6422, "staged_split": 45260, "admitted_merge": 8963, "staged_merge": 117807, "admitted_skel": 13, "staged_skel": 48021, "mean_count": 254.1055145263672, "n_util_obs": 261, "util_count_spearman": -0.1872136033122109}, "probe_overhead_sec_total": 543.427, "probe_overhead_seq_frac": 0.4999, "alpha": 0.03, "t": 4056.9}
{"event": "consolidate", "step": 3400, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262075, "t": 4109.6}
{"event": "probe", "step": 3400, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3485.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.200253, "cps": 27.226562, "mean_p": 0.211718, "boundary_entropy": 0.124678, "bpb": 1.979631}, "tr": {"chunks": 4168.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.230238, "cps": 32.5625, "mean_p": 0.212168, "boundary_entropy": 0.120498, "bpb": 2.137393}, "fi": {"chunks": 4369.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.232258, "cps": 34.132812, "mean_p": 0.23466, "boundary_entropy": 0.129215, "bpb": 2.265164}}, "bpb": {"en": 1.97963, "tr": 2.13739, "fi": 2.26516}, "gini_cpc": 0.032194, "gini_cps": 0.049021, "bpic": 4.719015, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.1497369705996874, "fi": 1.1598224183826018}, "inventory": {"size": 326655, "admitted": 65536, "staged": 261119, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1599, "admitted_surface": 51159, "staged_surface": 50396, "admitted_split": 5825, "staged_split": 45482, "admitted_merge": 8539, "staged_merge": 117318, "admitted_skel": 13, "staged_skel": 47923, "mean_count": 258.13226318359375, "n_util_obs": 260, "util_count_spearman": -0.20831942211252555}, "t": 4111.8}
{"event": "eval", "step": 3400, "phase": 2, "bpb": {"en": 1.99078, "tr": 2.14487, "fi": 2.27093}, "bytes_per_chunk": {"en": 4.998, "tr": 4.813, "fi": 4.723}, "bpb_mean": 2.13553, "frozen_inventory": true, "t": 4115.8}
{"event": "train", "step": 3400, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.41873, "bpb": 2.04679, "chunk_loss": 0.030385, "F": 0.20071, "G": 0.21106, "chunks": 205.53, "bytes_per_chunk": 4.982, "boundary_entropy": 0.11994, "router_noise": 0.0, "grad_norm": 0.7387, "lr": 0.0001465, "router_trainable": true, "sec_per_step": 1.28, "probe_sec": 0.0811, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05008, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.02227, "n_shadow": 6, "mean_delta_shadow": -0.03696, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 0, "read_hit_rate": {"surface": 0.48235, "split": 0.24482, "merge": 0.11393, "skel": 0.11133}, "mean_gate": {"surface": 0.14287, "split": 0.06665, "merge": 0.10803, "skel": 0.09805}, "n_reads": 7293, "n_cand": 29654, "read_suppressed_frac": 0.1039, "dropout_reads_logged": 7293, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1598, "admitted_surface": 51159, "staged_surface": 50611, "admitted_split": 5825, "staged_split": 45635, "admitted_merge": 8539, "staged_merge": 117821, "admitted_skel": 13, "staged_skel": 48077, "mean_count": 258.2435302734375, "n_util_obs": 261, "util_count_spearman": -0.20779594626723022}, "probe_overhead_sec_total": 551.899, "probe_overhead_seq_frac": 0.4999, "alpha": 0.03, "t": 4120.9}
{"event": "consolidate", "step": 3450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262073, "t": 4181.4}
{"event": "train", "step": 3450, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.38866, "bpb": 2.00341, "chunk_loss": 0.030374, "F": 0.202, "G": 0.21085, "chunks": 206.84, "bytes_per_chunk": 4.951, "boundary_entropy": 0.1195, "router_noise": 0.0, "grad_norm": 0.7783, "lr": 0.0001428, "router_trainable": true, "sec_per_step": 1.229, "probe_sec": 0.1241, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05897, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": -0.01945, "n_shadow": 5, "mean_delta_shadow": 0.01307, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": -0.01318, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -1.92922, "trunc_abs_frac": 0.31747, "mean_delta_full": -0.02026, "read_hit_rate": {"surface": 0.49476, "split": 0.25168, "merge": 0.12418, "skel": 0.1058}, "mean_gate": {"surface": 0.14604, "split": 0.06685, "merge": 0.10881, "skel": 0.10257}, "n_reads": 7530, "n_cand": 29825, "read_suppressed_frac": 0.0965, "dropout_reads_logged": 7530, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1633, "admitted_surface": 51603, "staged_surface": 50772, "admitted_split": 5431, "staged_split": 45649, "admitted_merge": 8489, "staged_merge": 117883, "admitted_skel": 13, "staged_skel": 47840, "mean_count": 262.9710388183594, "n_util_obs": 261, "util_count_spearman": -0.20691072549456335}, "probe_overhead_sec_total": 560.297, "probe_overhead_seq_frac": 0.4999, "alpha": 0.03, "t": 4182.3}
{"event": "consolidate", "step": 3500, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 1024, "uncertain_staged": 262073, "t": 4237.9}
{"event": "train", "step": 3500, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.36537, "bpb": 1.96981, "chunk_loss": 0.030294, "F": 0.19495, "G": 0.20648, "chunks": 199.62, "bytes_per_chunk": 5.13, "boundary_entropy": 0.11331, "router_noise": 0.0, "grad_norm": 0.6808, "lr": 0.0001391, "router_trainable": true, "sec_per_step": 1.125, "probe_sec": 0.1176, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05689, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": -0.03033, "n_shadow": 7, "mean_delta_shadow": -0.03473, "n_split": 1, "mean_delta_split": 0.04881, "n_merge": 1, "mean_delta_merge": 0.21258, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.64259, "trunc_abs_frac": 0.8138, "mean_delta_full": 0.13318, "read_hit_rate": {"surface": 0.50677, "split": 0.25899, "merge": 0.12698, "skel": 0.13159}, "mean_gate": {"surface": 0.1467, "split": 0.06583, "merge": 0.10435, "skel": 0.10285}, "n_reads": 7622, "n_cand": 29036, "read_suppressed_frac": 0.0968, "dropout_reads_logged": 7622, "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": 52622, "staged_surface": 50593, "admitted_split": 4488, "staged_split": 45760, "admitted_merge": 8413, "staged_merge": 117692, "admitted_skel": 13, "staged_skel": 48099, "mean_count": 267.352294921875, "n_util_obs": 261, "util_count_spearman": -0.21736219382968527}, "probe_overhead_sec_total": 568.249, "probe_overhead_seq_frac": 0.4999, "alpha": 0.03, "t": 4238.6}
{"event": "consolidate", "step": 3550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 485, "uncertain_staged": 262074, "t": 4290.0}
{"event": "train", "step": 3550, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.40587, "bpb": 2.02824, "chunk_loss": 0.030393, "F": 0.20145, "G": 0.21068, "chunks": 206.28, "bytes_per_chunk": 4.964, "boundary_entropy": 0.1164, "router_noise": 0.0, "grad_norm": 0.6742, "lr": 0.0001355, "router_trainable": true, "sec_per_step": 1.044, "probe_sec": 0.1018, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07558, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 3, "mean_delta_loo": -0.02522, "n_shadow": 3, "mean_delta_shadow": -0.02222, "n_split": 2, "mean_delta_split": 0.17584, "n_merge": 2, "mean_delta_merge": -0.12872, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 1.13821, "trunc_abs_frac": 1.2844, "mean_delta_full": 0.0387, "read_hit_rate": {"surface": 0.48912, "split": 0.26495, "merge": 0.11917, "skel": 0.10613}, "mean_gate": {"surface": 0.14639, "split": 0.06659, "merge": 0.10932, "skel": 0.10282}, "n_reads": 7582, "n_cand": 29666, "read_suppressed_frac": 0.0956, "dropout_reads_logged": 7582, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1680, "admitted_surface": 53059, "staged_surface": 51181, "admitted_split": 4102, "staged_split": 45582, "admitted_merge": 8362, "staged_merge": 117406, "admitted_skel": 13, "staged_skel": 47975, "mean_count": 280.30047607421875, "n_util_obs": 265, "util_count_spearman": -0.22119643018352056}, "probe_overhead_sec_total": 575.809, "probe_overhead_seq_frac": 0.4999, "alpha": 0.03, "t": 4290.8}
{"event": "consolidate", "step": 3600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 957, "uncertain_staged": 262076, "t": 4346.1}
{"event": "probe", "step": 3600, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3525.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.202551, "cps": 27.539062, "mean_p": 0.213539, "boundary_entropy": 0.123166, "bpb": 1.963225}, "tr": {"chunks": 4146.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.229023, "cps": 32.390625, "mean_p": 0.210913, "boundary_entropy": 0.115637, "bpb": 2.101262}, "fi": {"chunks": 4288.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.227952, "cps": 33.5, "mean_p": 0.231985, "boundary_entropy": 0.127505, "bpb": 2.23953}}, "bpb": {"en": 1.96322, "tr": 2.10126, "fi": 2.23953}, "gini_cpc": 0.026758, "gini_cps": 0.042534, "bpic": 4.743875, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.1306905050414824, "fi": 1.1254025428761125}, "inventory": {"size": 326723, "admitted": 65536, "staged": 261187, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1711, "admitted_surface": 53902, "staged_surface": 50571, "admitted_split": 3429, "staged_split": 44987, "admitted_merge": 8192, "staged_merge": 117551, "admitted_skel": 13, "staged_skel": 48078, "mean_count": 284.92144775390625, "n_util_obs": 270, "util_count_spearman": -0.22210611364980323}, "t": 4348.3}
{"event": "eval", "step": 3600, "phase": 2, "bpb": {"en": 1.97474, "tr": 2.10812, "fi": 2.24459}, "bytes_per_chunk": {"en": 4.955, "tr": 4.825, "fi": 4.867}, "bpb_mean": 2.10915, "frozen_inventory": true, "t": 4352.5}
{"event": "train", "step": 3600, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.40537, "bpb": 2.02752, "chunk_loss": 0.030462, "F": 0.20361, "G": 0.2138, "chunks": 208.5, "bytes_per_chunk": 4.911, "boundary_entropy": 0.12006, "router_noise": 0.0, "grad_norm": 0.6395, "lr": 0.0001319, "router_trainable": true, "sec_per_step": 1.376, "probe_sec": 0.1577, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07308, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.07151, "n_shadow": 4, "mean_delta_shadow": -0.1266, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": -0.31812, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.86953, "trunc_abs_frac": 1.4345, "mean_delta_full": -0.23269, "read_hit_rate": {"surface": 0.48946, "split": 0.27385, "merge": 0.12515, "skel": 0.12245}, "mean_gate": {"surface": 0.14191, "split": 0.06475, "merge": 0.10623, "skel": 0.10188}, "n_reads": 7951, "n_cand": 30266, "read_suppressed_frac": 0.1035, "dropout_reads_logged": 7951, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1713, "admitted_surface": 53902, "staged_surface": 50790, "admitted_split": 3429, "staged_split": 45089, "admitted_merge": 8192, "staged_merge": 118032, "admitted_skel": 13, "staged_skel": 48233, "mean_count": 285.04278564453125, "n_util_obs": 270, "util_count_spearman": -0.22322912876859244}, "probe_overhead_sec_total": 584.576, "probe_overhead_seq_frac": 0.4999, "alpha": 0.03, "t": 4359.6}
{"event": "consolidate", "step": 3650, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262075, "t": 4421.0}
{"event": "train", "step": 3650, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.43974, "bpb": 2.07711, "chunk_loss": 0.030513, "F": 0.20593, "G": 0.21572, "chunks": 210.88, "bytes_per_chunk": 4.856, "boundary_entropy": 0.11758, "router_noise": 0.0, "grad_norm": 0.7082, "lr": 0.0001283, "router_trainable": true, "sec_per_step": 1.246, "probe_sec": 0.1613, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06337, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": 0.06578, "n_shadow": 9, "mean_delta_shadow": -0.05437, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 1.32308, "trunc_abs_frac": 1.32308, "mean_delta_full": -0.10315, "read_hit_rate": {"surface": 0.49181, "split": 0.27722, "merge": 0.10772, "skel": 0.1113}, "mean_gate": {"surface": 0.14134, "split": 0.06373, "merge": 0.1059, "skel": 0.09624}, "n_reads": 7895, "n_cand": 30470, "read_suppressed_frac": 0.1017, "dropout_reads_logged": 7895, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1727, "admitted_surface": 54841, "staged_surface": 50689, "admitted_split": 2647, "staged_split": 45386, "admitted_merge": 8035, "staged_merge": 117886, "admitted_skel": 13, "staged_skel": 48183, "mean_count": 289.78082275390625, "n_util_obs": 275, "util_count_spearman": -0.2367006486686284}, "probe_overhead_sec_total": 594.003, "probe_overhead_seq_frac": 0.4999, "alpha": 0.03, "t": 4421.9}
{"event": "consolidate", "step": 3700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 383, "uncertain_staged": 262077, "t": 4477.3}
{"event": "train", "step": 3700, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.39035, "bpb": 2.00584, "chunk_loss": 0.030459, "F": 0.20557, "G": 0.21414, "chunks": 210.5, "bytes_per_chunk": 4.865, "boundary_entropy": 0.11866, "router_noise": 0.0, "grad_norm": 0.6633, "lr": 0.0001247, "router_trainable": true, "sec_per_step": 1.12, "probe_sec": 0.1294, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05998, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": -0.04933, "n_shadow": 6, "mean_delta_shadow": 0.11952, "n_split": 1, "mean_delta_split": -0.03308, "n_merge": 2, "mean_delta_merge": -0.60172, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 0.59664, "trunc_abs_frac": 0.7628, "mean_delta_full": 0.13737, "read_hit_rate": {"surface": 0.50895, "split": 0.27958, "merge": 0.11826, "skel": 0.10697}, "mean_gate": {"surface": 0.14049, "split": 0.06387, "merge": 0.10393, "skel": 0.10064}, "n_reads": 8008, "n_cand": 30092, "read_suppressed_frac": 0.0935, "dropout_reads_logged": 8008, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1745, "admitted_surface": 55150, "staged_surface": 50976, "admitted_split": 2645, "staged_split": 45113, "admitted_merge": 7728, "staged_merge": 117995, "admitted_skel": 13, "staged_skel": 48060, "mean_count": 294.76812744140625, "n_util_obs": 281, "util_count_spearman": -0.21590855928906436}, "probe_overhead_sec_total": 602.463, "probe_overhead_seq_frac": 0.4999, "alpha": 0.03, "t": 4477.9}
{"event": "consolidate", "step": 3750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 182, "uncertain_staged": 262079, "t": 4524.6}
{"event": "train", "step": 3750, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.37525, "bpb": 1.98407, "chunk_loss": 0.030414, "F": 0.20181, "G": 0.2096, "chunks": 206.66, "bytes_per_chunk": 4.955, "boundary_entropy": 0.11225, "router_noise": 0.0, "grad_norm": 0.6785, "lr": 0.0001212, "router_trainable": true, "sec_per_step": 0.945, "probe_sec": 0.0971, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05565, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": 0.28141, "n_shadow": 7, "mean_delta_shadow": -0.03298, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.47009, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 3.16183, "trunc_abs_frac": 1.44465, "mean_delta_full": 0.102, "read_hit_rate": {"surface": 0.50828, "split": 0.27608, "merge": 0.11803, "skel": 0.11061}, "mean_gate": {"surface": 0.14336, "split": 0.06238, "merge": 0.10434, "skel": 0.1004}, "n_reads": 7885, "n_cand": 29836, "read_suppressed_frac": 0.0959, "dropout_reads_logged": 7885, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1769, "admitted_surface": 55216, "staged_surface": 51312, "admitted_split": 2641, "staged_split": 45263, "admitted_merge": 7666, "staged_merge": 117883, "admitted_skel": 13, "staged_skel": 47686, "mean_count": 299.91888427734375, "n_util_obs": 286, "util_count_spearman": -0.21966557191204175}, "probe_overhead_sec_total": 610.127, "probe_overhead_seq_frac": 0.4999, "alpha": 0.03, "t": 4525.2}
{"event": "consolidate", "step": 3800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 648, "uncertain_staged": 262079, "t": 4573.2}
{"event": "probe", "step": 3800, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3491.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.200598, "cps": 27.273438, "mean_p": 0.210571, "boundary_entropy": 0.11598, "bpb": 1.949895}, "tr": {"chunks": 4145.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.228968, "cps": 32.382812, "mean_p": 0.211537, "boundary_entropy": 0.1092, "bpb": 2.093661}, "fi": {"chunks": 4306.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.228909, "cps": 33.640625, "mean_p": 0.231243, "boundary_entropy": 0.121058, "bpb": 2.220642}}, "bpb": {"en": 1.9499, "tr": 2.09366, "fi": 2.22064}, "gini_cpc": 0.028723, "gini_cps": 0.045498, "bpic": 4.750628, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.14142729816539, "fi": 1.1411333906742676}, "inventory": {"size": 327032, "admitted": 65536, "staged": 261496, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1798, "admitted_surface": 55847, "staged_surface": 51159, "admitted_split": 2637, "staged_split": 45395, "admitted_merge": 7039, "staged_merge": 117107, "admitted_skel": 13, "staged_skel": 47835, "mean_count": 306.07379150390625, "n_util_obs": 288, "util_count_spearman": -0.21866723733969914}, "t": 4575.3}
{"event": "eval", "step": 3800, "phase": 2, "bpb": {"en": 1.96287, "tr": 2.10176, "fi": 2.22694}, "bytes_per_chunk": {"en": 4.975, "tr": 4.824, "fi": 4.865}, "bpb_mean": 2.09719, "frozen_inventory": true, "t": 4578.4}
{"event": "train", "step": 3800, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.45417, "bpb": 2.09793, "chunk_loss": 0.030421, "F": 0.20361, "G": 0.21224, "chunks": 208.5, "bytes_per_chunk": 4.911, "boundary_entropy": 0.11333, "router_noise": 0.0, "grad_norm": 0.6843, "lr": 0.0001177, "router_trainable": true, "sec_per_step": 1.185, "probe_sec": 0.1216, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05678, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": -0.07394, "n_shadow": 8, "mean_delta_shadow": -0.1258, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.03204, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.4742, "trunc_abs_frac": 0.39078, "mean_delta_full": -0.04022, "read_hit_rate": {"surface": 0.49849, "split": 0.26648, "merge": 0.10881, "skel": 0.11248}, "mean_gate": {"surface": 0.13914, "split": 0.06256, "merge": 0.10818, "skel": 0.0994}, "n_reads": 7768, "n_cand": 30290, "read_suppressed_frac": 0.0987, "dropout_reads_logged": 7768, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1797, "admitted_surface": 55847, "staged_surface": 51271, "admitted_split": 2637, "staged_split": 45449, "admitted_merge": 7039, "staged_merge": 117524, "admitted_skel": 13, "staged_skel": 47900, "mean_count": 306.19232177734375, "n_util_obs": 288, "util_count_spearman": -0.21882500367031893}, "probe_overhead_sec_total": 618.358, "probe_overhead_seq_frac": 0.4999, "alpha": 0.03, "t": 4584.4}
{"event": "consolidate", "step": 3850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 777, "uncertain_staged": 262079, "t": 4632.5}
{"event": "train", "step": 3850, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.53876, "bpb": 2.21996, "chunk_loss": 0.03078, "F": 0.21164, "G": 0.21784, "chunks": 216.72, "bytes_per_chunk": 4.725, "boundary_entropy": 0.11401, "router_noise": 0.0, "grad_norm": 0.7205, "lr": 0.0001142, "router_trainable": true, "sec_per_step": 0.981, "probe_sec": 0.0869, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04825, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.0154, "n_shadow": 5, "mean_delta_shadow": -0.05645, "n_split": 2, "mean_delta_split": 0.11665, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 1.70431, "trunc_abs_frac": 0.73228, "mean_delta_full": -0.08016, "read_hit_rate": {"surface": 0.46212, "split": 0.25569, "merge": 0.10915, "skel": 0.10319}, "mean_gate": {"surface": 0.14479, "split": 0.06417, "merge": 0.1106, "skel": 0.10152}, "n_reads": 7413, "n_cand": 30293, "read_suppressed_frac": 0.1008, "dropout_reads_logged": 7413, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1822, "admitted_surface": 56379, "staged_surface": 50893, "admitted_split": 2590, "staged_split": 45668, "admitted_merge": 6554, "staged_merge": 117712, "admitted_skel": 13, "staged_skel": 47871, "mean_count": 311.921875, "n_util_obs": 294, "util_count_spearman": -0.24445764299203926}, "probe_overhead_sec_total": 626.134, "probe_overhead_seq_frac": 0.4999, "alpha": 0.03, "t": 4633.4}
{"event": "consolidate", "step": 3900, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262080, "t": 4682.0}
{"event": "train", "step": 3900, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.43729, "bpb": 2.07356, "chunk_loss": 0.030464, "F": 0.20169, "G": 0.21212, "chunks": 206.53, "bytes_per_chunk": 4.958, "boundary_entropy": 0.11514, "router_noise": 0.0, "grad_norm": 0.7469, "lr": 0.0001108, "router_trainable": true, "sec_per_step": 0.986, "probe_sec": 0.2923, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.1486, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": -0.04418, "n_shadow": 4, "mean_delta_shadow": -0.00823, "n_split": 1, "mean_delta_split": -0.0127, "n_merge": 1, "mean_delta_merge": -0.08353, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 1.60189, "trunc_abs_frac": 0.30809, "mean_delta_full": -0.03654, "read_hit_rate": {"surface": 0.48335, "split": 0.26394, "merge": 0.09992, "skel": 0.09304}, "mean_gate": {"surface": 0.1364, "split": 0.06057, "merge": 0.10716, "skel": 0.09425}, "n_reads": 7348, "n_cand": 29752, "read_suppressed_frac": 0.1011, "dropout_reads_logged": 7348, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1848, "admitted_surface": 57398, "staged_surface": 50536, "admitted_split": 2386, "staged_split": 45302, "admitted_merge": 5741, "staged_merge": 118240, "admitted_skel": 11, "staged_skel": 48066, "mean_count": 312.34039306640625, "n_util_obs": 291, "util_count_spearman": -0.243080833493327}, "probe_overhead_sec_total": 633.489, "probe_overhead_seq_frac": 0.4999, "alpha": 0.03, "t": 4682.8}
{"event": "consolidate", "step": 3950, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 497, "uncertain_staged": 262080, "t": 4734.1}
{"event": "train", "step": 3950, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.48819, "bpb": 2.147, "chunk_loss": 0.030379, "F": 0.19855, "G": 0.20639, "chunks": 203.31, "bytes_per_chunk": 5.037, "boundary_entropy": 0.10989, "router_noise": 0.0, "grad_norm": 0.791, "lr": 0.0001074, "router_trainable": true, "sec_per_step": 1.043, "probe_sec": 0.1814, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.12683, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.04175, "n_shadow": 5, "mean_delta_shadow": 0.03188, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": 0.16202, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.78271, "trunc_abs_frac": 0.39732, "mean_delta_full": 0.01835, "read_hit_rate": {"surface": 0.50355, "split": 0.2759, "merge": 0.10214, "skel": 0.09928}, "mean_gate": {"surface": 0.13741, "split": 0.05989, "merge": 0.10582, "skel": 0.09423}, "n_reads": 7576, "n_cand": 29360, "read_suppressed_frac": 0.1061, "dropout_reads_logged": 7576, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1863, "admitted_surface": 57629, "staged_surface": 50843, "admitted_split": 2376, "staged_split": 45493, "admitted_merge": 5520, "staged_merge": 117712, "admitted_skel": 11, "staged_skel": 48096, "mean_count": 319.5177001953125, "n_util_obs": 296, "util_count_spearman": -0.2531559879963208}, "probe_overhead_sec_total": 641.247, "probe_overhead_seq_frac": 0.4999, "alpha": 0.03, "t": 4734.9}
{"event": "consolidate", "step": 4000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262081, "t": 4779.5}
{"event": "probe", "step": 4000, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3466.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.199161, "cps": 27.078125, "mean_p": 0.209168, "boundary_entropy": 0.117492, "bpb": 1.941268}, "tr": {"chunks": 4148.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.229133, "cps": 32.40625, "mean_p": 0.211489, "boundary_entropy": 0.110866, "bpb": 2.075419}, "fi": {"chunks": 4317.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.229493, "cps": 33.726562, "mean_p": 0.232229, "boundary_entropy": 0.122659, "bpb": 2.200464}}, "bpb": {"en": 1.94127, "tr": 2.07542, "fi": 2.20046}, "gini_cpc": 0.030742, "gini_cps": 0.047551, "bpic": 4.755008, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.150492410566337, "fi": 1.1523004382004696}, "inventory": {"size": 326656, "admitted": 65536, "staged": 261120, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1876, "admitted_surface": 58645, "staged_surface": 49682, "admitted_split": 2239, "staged_split": 45257, "admitted_merge": 4641, "staged_merge": 118292, "admitted_skel": 11, "staged_skel": 47889, "mean_count": 322.9200134277344, "n_util_obs": 298, "util_count_spearman": -0.25423161606412265}, "t": 4781.1}
{"event": "eval", "step": 4000, "phase": 2, "bpb": {"en": 1.95299, "tr": 2.08324, "fi": 2.20701}, "bytes_per_chunk": {"en": 5.032, "tr": 4.835, "fi": 4.846}, "bpb_mean": 2.08108, "frozen_inventory": true, "t": 4783.9}
{"event": "train", "step": 4000, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.41579, "bpb": 2.04256, "chunk_loss": 0.030422, "F": 0.20013, "G": 0.20863, "chunks": 204.94, "bytes_per_chunk": 4.997, "boundary_entropy": 0.11353, "router_noise": 0.0, "grad_norm": 0.6997, "lr": 0.000104, "router_trainable": true, "sec_per_step": 1.077, "probe_sec": 0.098, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06202, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": 0.01891, "n_shadow": 6, "mean_delta_shadow": -0.00255, "n_split": 2, "mean_delta_split": -0.09874, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 3.38643, "trunc_abs_frac": 1.80597, "mean_delta_full": -0.02938, "read_hit_rate": {"surface": 0.51042, "split": 0.27817, "merge": 0.11534, "skel": 0.11098}, "mean_gate": {"surface": 0.14225, "split": 0.06114, "merge": 0.10578, "skel": 0.09584}, "n_reads": 7832, "n_cand": 29530, "read_suppressed_frac": 0.0986, "dropout_reads_logged": 7832, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1875, "admitted_surface": 58645, "staged_surface": 49869, "admitted_split": 2239, "staged_split": 45528, "admitted_merge": 4641, "staged_merge": 118673, "admitted_skel": 11, "staged_skel": 48074, "mean_count": 323.0395202636719, "n_util_obs": 298, "util_count_spearman": -0.254289914281638}, "probe_overhead_sec_total": 648.401, "probe_overhead_seq_frac": 0.4999, "alpha": 0.03, "t": 4788.8}
{"event": "consolidate", "step": 4050, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262081, "t": 4831.6}
{"event": "train", "step": 4050, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.41915, "bpb": 2.0474, "chunk_loss": 0.030477, "F": 0.20639, "G": 0.21346, "chunks": 211.34, "bytes_per_chunk": 4.845, "boundary_entropy": 0.10998, "router_noise": 0.0, "grad_norm": 0.7124, "lr": 0.0001007, "router_trainable": true, "sec_per_step": 0.872, "probe_sec": 0.1217, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0937, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.09639, "n_shadow": 4, "mean_delta_shadow": 0.01366, "n_split": 2, "mean_delta_split": -0.07819, "n_merge": 1, "mean_delta_merge": 0.11063, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.31729, "trunc_abs_frac": 0.30584, "mean_delta_full": -0.37732, "read_hit_rate": {"surface": 0.50602, "split": 0.23417, "merge": 0.10688, "skel": 0.09032}, "mean_gate": {"surface": 0.14359, "split": 0.06134, "merge": 0.10771, "skel": 0.0929}, "n_reads": 7330, "n_cand": 30238, "read_suppressed_frac": 0.0962, "dropout_reads_logged": 7330, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1868, "admitted_surface": 59664, "staged_surface": 49740, "admitted_split": 1813, "staged_split": 45876, "admitted_merge": 4049, "staged_merge": 118428, "admitted_skel": 10, "staged_skel": 48100, "mean_count": 314.99420166015625, "n_util_obs": 290, "util_count_spearman": -0.2662544267408977}, "probe_overhead_sec_total": 654.958, "probe_overhead_seq_frac": 0.4999, "alpha": 0.03, "t": 4832.3}
{"event": "probe", "step": 4096, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3488.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.200425, "cps": 27.25, "mean_p": 0.211228, "boundary_entropy": 0.111939, "bpb": 1.934961}, "tr": {"chunks": 4149.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.229189, "cps": 32.414062, "mean_p": 0.213263, "boundary_entropy": 0.105537, "bpb": 2.067945}, "fi": {"chunks": 4325.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.229919, "cps": 33.789062, "mean_p": 0.232269, "boundary_entropy": 0.115727, "bpb": 2.194881}}, "bpb": {"en": 1.93496, "tr": 2.06794, "fi": 2.19488}, "gini_cpc": 0.029812, "gini_cps": 0.046648, "bpic": 4.742685, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.1435114757450613, "fi": 1.1471543922670662}, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1896, "admitted_surface": 59664, "staged_surface": 50650, "admitted_split": 1813, "staged_split": 45529, "admitted_merge": 4049, "staged_merge": 118068, "admitted_skel": 10, "staged_skel": 47897, "mean_count": 320.1725769042969, "n_util_obs": 296, "util_count_spearman": -0.24903362530481174}, "t": 4870.1}
{"event": "consolidate", "step": 4100, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 138, "uncertain_staged": 262081, "t": 4884.0}
{"event": "train", "step": 4100, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.35616, "bpb": 1.95652, "chunk_loss": 0.030504, "F": 0.2049, "G": 0.2131, "chunks": 209.81, "bytes_per_chunk": 4.881, "boundary_entropy": 0.11068, "router_noise": 0.0, "grad_norm": 0.7297, "lr": 9.75e-05, "router_trainable": true, "sec_per_step": 1.045, "probe_sec": 0.1176, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09392, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.02296, "n_shadow": 3, "mean_delta_shadow": -0.0677, "n_split": 1, "mean_delta_split": 0.08057, "n_merge": 2, "mean_delta_merge": -0.15307, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -4.54337, "trunc_abs_frac": 0.17097, "mean_delta_full": 0.01416, "read_hit_rate": {"surface": 0.52799, "split": 0.26855, "merge": 0.11489, "skel": 0.09676}, "mean_gate": {"surface": 0.14157, "split": 0.06008, "merge": 0.10652, "skel": 0.09746}, "n_reads": 7951, "n_cand": 30236, "read_suppressed_frac": 0.099, "dropout_reads_logged": 7951, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1898, "admitted_surface": 59663, "staged_surface": 50616, "admitted_split": 1814, "staged_split": 45724, "admitted_merge": 4049, "staged_merge": 117972, "admitted_skel": 10, "staged_skel": 47832, "mean_count": 322.89910888671875, "n_util_obs": 295, "util_count_spearman": -0.2632377275234418}, "probe_overhead_sec_total": 661.434, "probe_overhead_seq_frac": 0.4999, "alpha": 0.03, "t": 4884.6}
{"event": "consolidate", "step": 4150, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 4, "uncertain_staged": 262082, "t": 4928.5}
{"event": "train", "step": 4150, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.41864, "bpb": 2.04667, "chunk_loss": 0.0305, "F": 0.20667, "G": 0.21547, "chunks": 211.62, "bytes_per_chunk": 4.839, "boundary_entropy": 0.11049, "router_noise": 0.0, "grad_norm": 0.756, "lr": 9.43e-05, "router_trainable": true, "sec_per_step": 0.89, "probe_sec": 0.1082, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09054, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.13648, "n_shadow": 4, "mean_delta_shadow": -0.01688, "n_split": 1, "mean_delta_split": 0.00293, "n_merge": 3, "mean_delta_merge": 0.01232, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 0.17697, "trunc_abs_frac": 0.71947, "mean_delta_full": 0.12258, "read_hit_rate": {"surface": 0.5135, "split": 0.25096, "merge": 0.10256, "skel": 0.10848}, "mean_gate": {"surface": 0.13282, "split": 0.05838, "merge": 0.10151, "skel": 0.0916}, "n_reads": 7726, "n_cand": 30674, "read_suppressed_frac": 0.0953, "dropout_reads_logged": 7726, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1930, "admitted_surface": 59662, "staged_surface": 50752, "admitted_split": 1816, "staged_split": 45835, "admitted_merge": 4048, "staged_merge": 117913, "admitted_skel": 10, "staged_skel": 47644, "mean_count": 329.8004150390625, "n_util_obs": 300, "util_count_spearman": -0.25433352085395783}, "probe_overhead_sec_total": 667.608, "probe_overhead_seq_frac": 0.4999, "alpha": 0.03, "t": 4929.1}
{"event": "consolidate", "step": 4200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 191, "uncertain_staged": 262084, "t": 4966.0}
{"event": "probe", "step": 4200, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3507.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.201517, "cps": 27.398438, "mean_p": 0.210854, "boundary_entropy": 0.113453, "bpb": 1.932275}, "tr": {"chunks": 4110.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.227034, "cps": 32.109375, "mean_p": 0.210827, "boundary_entropy": 0.109574, "bpb": 2.061115}, "fi": {"chunks": 4287.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.227899, "cps": 33.492188, "mean_p": 0.231182, "boundary_entropy": 0.118017, "bpb": 2.191966}}, "bpb": {"en": 1.93228, "tr": 2.06112, "fi": 2.19197}, "gini_cpc": 0.026792, "gini_cps": 0.043683, "bpic": 4.765793, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.126625624391403, "fi": 1.1309149738979782}, "inventory": {"size": 327489, "admitted": 65536, "staged": 261953, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1955, "admitted_surface": 59657, "staged_surface": 50655, "admitted_split": 1820, "staged_split": 45608, "admitted_merge": 4048, "staged_merge": 118183, "admitted_skel": 11, "staged_skel": 47507, "mean_count": 338.9189453125, "n_util_obs": 306, "util_count_spearman": -0.24820755196053143}, "t": 4967.4}
{"event": "eval", "step": 4200, "phase": 2, "bpb": {"en": 1.94327, "tr": 2.06903, "fi": 2.20005}, "bytes_per_chunk": {"en": 4.972, "tr": 4.878, "fi": 4.864}, "bpb_mean": 2.07078, "frozen_inventory": true, "t": 4970.5}
{"event": "phase_transition", "step": 4200, "phase": 3, "objective": "global", "router_trainable": true, "encoder_trainable": true, "frozen_router_params": 0, "t": 4974.5}
{"event": "train", "step": 4200, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35192, "bpb": 1.9504, "chunk_loss": 0.030297, "F": 0.19733, "G": 0.20526, "chunks": 202.06, "bytes_per_chunk": 5.068, "boundary_entropy": 0.10607, "router_noise": 0.0, "grad_norm": 0.6941, "lr": 9.12e-05, "router_trainable": true, "sec_per_step": 0.92, "probe_sec": 0.1058, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07434, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": 0.11248, "n_shadow": 4, "mean_delta_shadow": 0.09563, "n_split": 2, "mean_delta_split": -0.03363, "n_merge": 1, "mean_delta_merge": -0.08701, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.39776, "trunc_abs_frac": 0.19557, "mean_delta_full": 0.13104, "read_hit_rate": {"surface": 0.51368, "split": 0.27004, "merge": 0.10934, "skel": 0.1123}, "mean_gate": {"surface": 0.14147, "split": 0.05959, "merge": 0.10665, "skel": 0.09909}, "n_reads": 7634, "n_cand": 29167, "read_suppressed_frac": 0.0925, "dropout_reads_logged": 7634, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1955, "admitted_surface": 59657, "staged_surface": 50756, "admitted_split": 1820, "staged_split": 45686, "admitted_merge": 4048, "staged_merge": 118144, "admitted_skel": 11, "staged_skel": 47558, "mean_count": 339.0354309082031, "n_util_obs": 307, "util_count_spearman": -0.24664085355788748}, "probe_overhead_sec_total": 672.47, "probe_overhead_seq_frac": 0.4999, "alpha": 0.03, "t": 4975.1}
{"event": "consolidate", "step": 4250, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 87, "uncertain_staged": 262085, "t": 5018.6}
{"event": "train", "step": 4250, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38325, "bpb": 1.99561, "chunk_loss": 0.03051, "F": 0.2092, "G": 0.21572, "chunks": 214.22, "bytes_per_chunk": 4.78, "boundary_entropy": 0.10798, "router_noise": 0.0, "grad_norm": 0.6622, "lr": 8.81e-05, "router_trainable": true, "sec_per_step": 0.882, "probe_sec": 0.1173, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09433, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 3, "mean_delta_loo": -0.07685, "n_shadow": 7, "mean_delta_shadow": -0.01616, "n_split": 2, "mean_delta_split": 0.08713, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": -25.18333, "trunc_abs_frac": 25.18333, "mean_delta_full": -0.00366, "read_hit_rate": {"surface": 0.5181, "split": 0.26994, "merge": 0.11648, "skel": 0.11779}, "mean_gate": {"surface": 0.14073, "split": 0.05858, "merge": 0.10528, "skel": 0.09948}, "n_reads": 8164, "n_cand": 30746, "read_suppressed_frac": 0.1007, "dropout_reads_logged": 8164, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1991, "admitted_surface": 59660, "staged_surface": 50886, "admitted_split": 1820, "staged_split": 45506, "admitted_merge": 4045, "staged_merge": 118168, "admitted_skel": 11, "staged_skel": 47584, "mean_count": 343.38787841796875, "n_util_obs": 305, "util_count_spearman": -0.23730020751920874}, "probe_overhead_sec_total": 678.886, "probe_overhead_seq_frac": 0.4999, "alpha": 0.03, "t": 5019.2}
{"event": "consolidate", "step": 4300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 13, "uncertain_staged": 262084, "t": 5064.1}
{"event": "train", "step": 4300, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.39204, "bpb": 2.00829, "chunk_loss": 0.030311, "F": 0.1973, "G": 0.2063, "chunks": 202.03, "bytes_per_chunk": 5.069, "boundary_entropy": 0.11465, "router_noise": 0.0, "grad_norm": 0.7597, "lr": 8.5e-05, "router_trainable": true, "sec_per_step": 0.91, "probe_sec": 0.0769, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0511, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.04653, "n_shadow": 4, "mean_delta_shadow": -0.1972, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": -0.25975, "trunc_abs_frac": 0.25975, "mean_delta_full": 0.19397, "read_hit_rate": {"surface": 0.50956, "split": 0.26419, "merge": 0.09952, "skel": 0.10453}, "mean_gate": {"surface": 0.13643, "split": 0.05787, "merge": 0.10332, "skel": 0.09582}, "n_reads": 7439, "n_cand": 29162, "read_suppressed_frac": 0.0962, "dropout_reads_logged": 7439, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2001, "admitted_surface": 59657, "staged_surface": 50870, "admitted_split": 1823, "staged_split": 45828, "admitted_merge": 4045, "staged_merge": 117684, "admitted_skel": 11, "staged_skel": 47762, "mean_count": 349.43865966796875, "n_util_obs": 309, "util_count_spearman": -0.21011632794879}, "probe_overhead_sec_total": 684.973, "probe_overhead_seq_frac": 0.4999, "alpha": 0.03, "t": 5064.6}
{"event": "consolidate", "step": 4350, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 20, "uncertain_staged": 262082, "t": 5113.2}
{"event": "train", "step": 4350, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.40842, "bpb": 2.03192, "chunk_loss": 0.030369, "F": 0.20023, "G": 0.20846, "chunks": 205.03, "bytes_per_chunk": 4.994, "boundary_entropy": 0.10761, "router_noise": 0.0, "grad_norm": 0.6646, "lr": 8.21e-05, "router_trainable": true, "sec_per_step": 0.989, "probe_sec": 0.2414, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.14895, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.17693, "n_shadow": 4, "mean_delta_shadow": 0.03188, "n_split": 1, "mean_delta_split": -0.10733, "n_merge": 1, "mean_delta_merge": 0.00775, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 2.55338, "trunc_abs_frac": 14.34255, "mean_delta_full": 0.05145, "read_hit_rate": {"surface": 0.5062, "split": 0.25878, "merge": 0.09851, "skel": 0.12475}, "mean_gate": {"surface": 0.1335, "split": 0.05542, "merge": 0.10159, "skel": 0.09633}, "n_reads": 7585, "n_cand": 29728, "read_suppressed_frac": 0.0955, "dropout_reads_logged": 7585, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2020, "admitted_surface": 59658, "staged_surface": 51008, "admitted_split": 1821, "staged_split": 45631, "admitted_merge": 4045, "staged_merge": 117869, "admitted_skel": 12, "staged_skel": 47636, "mean_count": 353.9112854003906, "n_util_obs": 309, "util_count_spearman": -0.21638446113861085}, "probe_overhead_sec_total": 691.705, "probe_overhead_seq_frac": 0.4999, "alpha": 0.03, "t": 5114.1}
{"event": "consolidate", "step": 4400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 326, "uncertain_staged": 262083, "t": 5163.5}
{"event": "probe", "step": 4400, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3523.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.202436, "cps": 27.523438, "mean_p": 0.211183, "boundary_entropy": 0.111753, "bpb": 1.919831}, "tr": {"chunks": 4191.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.231509, "cps": 32.742188, "mean_p": 0.21347, "boundary_entropy": 0.105252, "bpb": 2.047562}, "fi": {"chunks": 4301.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.228643, "cps": 33.601562, "mean_p": 0.231481, "boundary_entropy": 0.119596, "bpb": 2.172983}}, "bpb": {"en": 1.91983, "tr": 2.04756, "fi": 2.17298}, "gini_cpc": 0.029251, "gini_cps": 0.043168, "bpic": 4.721764, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.1436116909408018, "fi": 1.1294552699271734}, "inventory": {"size": 327354, "admitted": 65536, "staged": 261818, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2054, "admitted_surface": 59678, "staged_surface": 50279, "admitted_split": 1815, "staged_split": 45721, "admitted_merge": 4031, "staged_merge": 118124, "admitted_skel": 12, "staged_skel": 47694, "mean_count": 357.18896484375, "n_util_obs": 312, "util_count_spearman": -0.21368535323652807}, "t": 5166.1}
{"event": "eval", "step": 4400, "phase": 3, "bpb": {"en": 1.93278, "tr": 2.05487, "fi": 2.18001}, "bytes_per_chunk": {"en": 4.951, "tr": 4.8, "fi": 4.844}, "bpb_mean": 2.05589, "frozen_inventory": true, "t": 5170.6}
{"event": "train", "step": 4400, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35844, "bpb": 1.95982, "chunk_loss": 0.030377, "F": 0.20181, "G": 0.20976, "chunks": 206.66, "bytes_per_chunk": 4.955, "boundary_entropy": 0.10853, "router_noise": 0.0, "grad_norm": 0.6525, "lr": 7.92e-05, "router_trainable": true, "sec_per_step": 1.227, "probe_sec": 0.0738, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04294, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": 0.06651, "n_shadow": 2, "mean_delta_shadow": 0.34152, "n_split": 2, "mean_delta_split": -0.11557, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -0.25251, "trunc_abs_frac": 0.64361, "mean_delta_full": 0.07793, "read_hit_rate": {"surface": 0.53639, "split": 0.28373, "merge": 0.11452, "skel": 0.11973}, "mean_gate": {"surface": 0.137, "split": 0.05715, "merge": 0.10351, "skel": 0.10135}, "n_reads": 8205, "n_cand": 29913, "read_suppressed_frac": 0.0957, "dropout_reads_logged": 8205, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2055, "admitted_surface": 59678, "staged_surface": 50311, "admitted_split": 1815, "staged_split": 45692, "admitted_merge": 4031, "staged_merge": 118427, "admitted_skel": 12, "staged_skel": 47714, "mean_count": 357.31414794921875, "n_util_obs": 312, "util_count_spearman": -0.21394173971657715}, "probe_overhead_sec_total": 698.943, "probe_overhead_seq_frac": 0.4999, "alpha": 0.03, "t": 5175.4}
{"event": "consolidate", "step": 4450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 3, "uncertain_staged": 262082, "t": 5218.8}
{"event": "train", "step": 4450, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35348, "bpb": 1.95266, "chunk_loss": 0.030291, "F": 0.19556, "G": 0.20325, "chunks": 200.25, "bytes_per_chunk": 5.114, "boundary_entropy": 0.10469, "router_noise": 0.0, "grad_norm": 0.682, "lr": 7.63e-05, "router_trainable": true, "sec_per_step": 0.878, "probe_sec": 0.0913, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06733, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.00311, "n_shadow": 4, "mean_delta_shadow": -0.0309, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.08374, "n_dropped_unpaired": 0, "n_full": 0, "read_hit_rate": {"surface": 0.52321, "split": 0.28007, "merge": 0.1116, "skel": 0.11995}, "mean_gate": {"surface": 0.13795, "split": 0.05715, "merge": 0.10266, "skel": 0.09759}, "n_reads": 7768, "n_cand": 28805, "read_suppressed_frac": 0.095, "dropout_reads_logged": 7768, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2091, "admitted_surface": 59680, "staged_surface": 50699, "admitted_split": 1813, "staged_split": 45562, "admitted_merge": 4031, "staged_merge": 117981, "admitted_skel": 12, "staged_skel": 47902, "mean_count": 363.1158752441406, "n_util_obs": 317, "util_count_spearman": -0.21114910760558983}, "probe_overhead_sec_total": 705.382, "probe_overhead_seq_frac": 0.4999, "alpha": 0.03, "t": 5219.3}
{"event": "consolidate", "step": 4500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 4, "uncertain_staged": 262082, "t": 5260.4}
{"event": "train", "step": 4500, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.33217, "bpb": 1.92191, "chunk_loss": 0.030363, "F": 0.20114, "G": 0.20886, "chunks": 205.97, "bytes_per_chunk": 4.972, "boundary_entropy": 0.1062, "router_noise": 0.0, "grad_norm": 0.6803, "lr": 7.36e-05, "router_trainable": true, "sec_per_step": 0.832, "probe_sec": 0.0918, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06973, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": -0.11711, "n_shadow": 5, "mean_delta_shadow": 0.02031, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.00909, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": -0.11702, "trunc_abs_frac": 0.9085, "mean_delta_full": 0.08093, "read_hit_rate": {"surface": 0.53209, "split": 0.27531, "merge": 0.10081, "skel": 0.10986}, "mean_gate": {"surface": 0.13308, "split": 0.05528, "merge": 0.10085, "skel": 0.10024}, "n_reads": 7929, "n_cand": 29921, "read_suppressed_frac": 0.1092, "dropout_reads_logged": 7929, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2118, "admitted_surface": 59678, "staged_surface": 50637, "admitted_split": 1815, "staged_split": 45517, "admitted_merge": 4031, "staged_merge": 118261, "admitted_skel": 12, "staged_skel": 47729, "mean_count": 369.1893615722656, "n_util_obs": 321, "util_count_spearman": -0.21070340163696522}, "probe_overhead_sec_total": 711.771, "probe_overhead_seq_frac": 0.4999, "alpha": 0.03, "t": 5260.9}
{"event": "consolidate", "step": 4550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 6, "uncertain_staged": 262083, "t": 5300.1}
{"event": "train", "step": 4550, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.36138, "bpb": 1.96405, "chunk_loss": 0.030341, "F": 0.20087, "G": 0.20753, "chunks": 205.69, "bytes_per_chunk": 4.978, "boundary_entropy": 0.10278, "router_noise": 0.0, "grad_norm": 0.6648, "lr": 7.09e-05, "router_trainable": true, "sec_per_step": 0.81, "probe_sec": 0.1589, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10265, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 3, "mean_delta_loo": 0.07817, "n_shadow": 3, "mean_delta_shadow": 0.08273, "n_split": 3, "mean_delta_split": -0.01643, "n_merge": 2, "mean_delta_merge": -0.07404, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": -0.2342, "trunc_abs_frac": 0.2342, "mean_delta_full": -0.38049, "read_hit_rate": {"surface": 0.51863, "split": 0.26548, "merge": 0.10371, "skel": 0.11007}, "mean_gate": {"surface": 0.13678, "split": 0.05638, "merge": 0.10582, "skel": 0.09615}, "n_reads": 7710, "n_cand": 29693, "read_suppressed_frac": 0.1001, "dropout_reads_logged": 7710, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2152, "admitted_surface": 59680, "staged_surface": 50884, "admitted_split": 1813, "staged_split": 45605, "admitted_merge": 4031, "staged_merge": 117931, "admitted_skel": 12, "staged_skel": 47724, "mean_count": 375.38739013671875, "n_util_obs": 327, "util_count_spearman": -0.217991341339122}, "probe_overhead_sec_total": 717.327, "probe_overhead_seq_frac": 0.4999, "alpha": 0.03, "t": 5301.4}
{"event": "consolidate", "step": 4600, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 0, "uncertain_staged": 262081, "t": 5339.5}
{"event": "probe", "step": 4600, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3504.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.201345, "cps": 27.375, "mean_p": 0.209749, "boundary_entropy": 0.107188, "bpb": 1.911868}, "tr": {"chunks": 4071.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.22488, "cps": 31.804688, "mean_p": 0.207834, "boundary_entropy": 0.110614, "bpb": 2.039818}, "fi": {"chunks": 4291.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.228111, "cps": 33.523438, "mean_p": 0.230301, "boundary_entropy": 0.114089, "bpb": 2.164015}}, "bpb": {"en": 1.91187, "tr": 2.03982, "fi": 2.16401}, "gini_cpc": 0.027271, "gini_cps": 0.044216, "bpic": 4.781055, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.1168904400920456, "fi": 1.132939330528698}, "inventory": {"size": 327678, "admitted": 65536, "staged": 262142, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2158, "admitted_surface": 59679, "staged_surface": 50809, "admitted_split": 1814, "staged_split": 45297, "admitted_merge": 4031, "staged_merge": 118487, "admitted_skel": 12, "staged_skel": 47549, "mean_count": 380.95074462890625, "n_util_obs": 336, "util_count_spearman": -0.2283830785216188}, "t": 5341.3}
{"event": "eval", "step": 4600, "phase": 3, "bpb": {"en": 1.92463, "tr": 2.04673, "fi": 2.16926}, "bytes_per_chunk": {"en": 4.972, "tr": 4.934, "fi": 4.84}, "bpb_mean": 2.04687, "frozen_inventory": true, "t": 5344.2}
{"event": "train", "step": 4600, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.36067, "bpb": 1.96303, "chunk_loss": 0.030497, "F": 0.20813, "G": 0.21552, "chunks": 213.12, "bytes_per_chunk": 4.805, "boundary_entropy": 0.10891, "router_noise": 0.0, "grad_norm": 0.6732, "lr": 6.82e-05, "router_trainable": true, "sec_per_step": 0.952, "probe_sec": 0.095, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04874, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": 0.06789, "n_shadow": 2, "mean_delta_shadow": -0.05829, "n_split": 3, "mean_delta_split": -0.02661, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 2.58156, "trunc_abs_frac": 0.45422, "mean_delta_full": 0.02362, "read_hit_rate": {"surface": 0.52372, "split": 0.28004, "merge": 0.1159, "skel": 0.12989}, "mean_gate": {"surface": 0.13184, "split": 0.05391, "merge": 0.0965, "skel": 0.0967}, "n_reads": 8338, "n_cand": 30639, "read_suppressed_frac": 0.1041, "dropout_reads_logged": 8338, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2159, "admitted_surface": 59679, "staged_surface": 50859, "admitted_split": 1814, "staged_split": 45344, "admitted_merge": 4031, "staged_merge": 118401, "admitted_skel": 12, "staged_skel": 47540, "mean_count": 381.07794189453125, "n_util_obs": 336, "util_count_spearman": -0.2279067961306338}, "probe_overhead_sec_total": 723.143, "probe_overhead_seq_frac": 0.4999, "alpha": 0.03, "t": 5349.0}
{"event": "consolidate", "step": 4650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 9, "uncertain_staged": 262082, "t": 5381.0}
{"event": "train", "step": 4650, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.3893, "bpb": 2.00434, "chunk_loss": 0.030512, "F": 0.20706, "G": 0.21242, "chunks": 212.03, "bytes_per_chunk": 4.829, "boundary_entropy": 0.09925, "router_noise": 0.0, "grad_norm": 0.6644, "lr": 6.57e-05, "router_trainable": true, "sec_per_step": 0.651, "probe_sec": 0.0995, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0572, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": 0.23932, "n_shadow": 3, "mean_delta_shadow": -0.01407, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.77356, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 1.28774, "trunc_abs_frac": 0.29844, "mean_delta_full": 0.08459, "read_hit_rate": {"surface": 0.50215, "split": 0.2464, "merge": 0.09671, "skel": 0.10714}, "mean_gate": {"surface": 0.1396, "split": 0.05636, "merge": 0.10845, "skel": 0.09958}, "n_reads": 7484, "n_cand": 30280, "read_suppressed_frac": 0.102, "dropout_reads_logged": 7484, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2180, "admitted_surface": 59680, "staged_surface": 51028, "admitted_split": 1813, "staged_split": 45665, "admitted_merge": 4030, "staged_merge": 117725, "admitted_skel": 13, "staged_skel": 47726, "mean_count": 386.93896484375, "n_util_obs": 342, "util_count_spearman": -0.2279706300816953}, "probe_overhead_sec_total": 727.401, "probe_overhead_seq_frac": 0.4999, "alpha": 0.03, "t": 5381.6}
{"event": "consolidate", "step": 4700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 3, "uncertain_staged": 262083, "t": 5414.7}
{"event": "train", "step": 4700, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.41854, "bpb": 2.04653, "chunk_loss": 0.030662, "F": 0.21133, "G": 0.21961, "chunks": 216.41, "bytes_per_chunk": 4.732, "boundary_entropy": 0.10572, "router_noise": 0.0, "grad_norm": 0.6711, "lr": 6.32e-05, "router_trainable": true, "sec_per_step": 0.672, "probe_sec": 0.0603, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04158, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": -0.01848, "n_shadow": 5, "mean_delta_shadow": 0.04216, "n_split": 1, "mean_delta_split": -0.08618, "n_merge": 2, "mean_delta_merge": -0.1951, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.05746, "trunc_abs_frac": 0.60614, "mean_delta_full": -0.0308, "read_hit_rate": {"surface": 0.52561, "split": 0.28182, "merge": 0.10392, "skel": 0.128}, "mean_gate": {"surface": 0.13065, "split": 0.05379, "merge": 0.09927, "skel": 0.09975}, "n_reads": 8364, "n_cand": 30891, "read_suppressed_frac": 0.0998, "dropout_reads_logged": 8364, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2199, "admitted_surface": 59681, "staged_surface": 50876, "admitted_split": 1812, "staged_split": 45388, "admitted_merge": 4030, "staged_merge": 118307, "admitted_skel": 13, "staged_skel": 47573, "mean_count": 392.591064453125, "n_util_obs": 348, "util_count_spearman": -0.21594954651028037}, "probe_overhead_sec_total": 731.965, "probe_overhead_seq_frac": 0.4999, "alpha": 0.03, "t": 5415.2}
{"event": "consolidate", "step": 4750, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 8, "uncertain_staged": 262084, "t": 5446.6}
{"event": "train", "step": 4750, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.40641, "bpb": 2.02902, "chunk_loss": 0.030437, "F": 0.20435, "G": 0.21188, "chunks": 209.25, "bytes_per_chunk": 4.894, "boundary_entropy": 0.10583, "router_noise": 0.0, "grad_norm": 0.6838, "lr": 6.08e-05, "router_trainable": true, "sec_per_step": 0.643, "probe_sec": 0.094, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05349, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.00554, "n_shadow": 2, "mean_delta_shadow": 0.04427, "n_split": 1, "mean_delta_split": 0.11749, "n_merge": 1, "mean_delta_merge": -0.05688, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.37818, "trunc_abs_frac": 1.04393, "mean_delta_full": 0.17273, "read_hit_rate": {"surface": 0.51471, "split": 0.27572, "merge": 0.09967, "skel": 0.12279}, "mean_gate": {"surface": 0.1333, "split": 0.05375, "merge": 0.10109, "skel": 0.09892}, "n_reads": 7960, "n_cand": 30156, "read_suppressed_frac": 0.1074, "dropout_reads_logged": 7960, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2228, "admitted_surface": 59681, "staged_surface": 50803, "admitted_split": 1812, "staged_split": 45732, "admitted_merge": 4030, "staged_merge": 117723, "admitted_skel": 13, "staged_skel": 47886, "mean_count": 396.9588317871094, "n_util_obs": 348, "util_count_spearman": -0.2415172836287045}, "probe_overhead_sec_total": 736.115, "probe_overhead_seq_frac": 0.4999, "alpha": 0.03, "t": 5447.3}
{"event": "consolidate", "step": 4800, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 8, "uncertain_staged": 262084, "t": 5477.8}
{"event": "probe", "step": 4800, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3489.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.200483, "cps": 27.257812, "mean_p": 0.207494, "boundary_entropy": 0.109861, "bpb": 1.902548}, "tr": {"chunks": 4097.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.226316, "cps": 32.007812, "mean_p": 0.209186, "boundary_entropy": 0.106192, "bpb": 2.025485}, "fi": {"chunks": 4287.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.227899, "cps": 33.492188, "mean_p": 0.229704, "boundary_entropy": 0.113843, "bpb": 2.150558}}, "bpb": {"en": 1.90255, "tr": 2.02548, "fi": 2.15056}, "gini_cpc": 0.027917, "gini_cps": 0.044808, "bpic": 4.778236, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.1288560458167411, "fi": 1.136749444958501}, "inventory": {"size": 327671, "admitted": 65536, "staged": 262135, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2245, "admitted_surface": 59676, "staged_surface": 51096, "admitted_split": 1817, "staged_split": 45521, "admitted_merge": 4030, "staged_merge": 117848, "admitted_skel": 13, "staged_skel": 47670, "mean_count": 402.79669189453125, "n_util_obs": 353, "util_count_spearman": -0.23756849668928193}, "t": 5479.3}
{"event": "eval", "step": 4800, "phase": 3, "bpb": {"en": 1.91584, "tr": 2.03314, "fi": 2.15566}, "bytes_per_chunk": {"en": 5.002, "tr": 4.883, "fi": 4.893}, "bpb_mean": 2.03488, "frozen_inventory": true, "t": 5481.8}
{"event": "train", "step": 4800, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.37786, "bpb": 1.98784, "chunk_loss": 0.030426, "F": 0.20142, "G": 0.2094, "chunks": 206.25, "bytes_per_chunk": 4.965, "boundary_entropy": 0.10889, "router_noise": 0.0, "grad_norm": 0.6592, "lr": 5.85e-05, "router_trainable": true, "sec_per_step": 0.784, "probe_sec": 0.0616, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04134, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": -0.03253, "n_shadow": 4, "mean_delta_shadow": 0.11151, "n_split": 1, "mean_delta_split": 0.01868, "n_merge": 2, "mean_delta_merge": 0.22346, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 1.69574, "trunc_abs_frac": 1.85256, "mean_delta_full": 0.03555, "read_hit_rate": {"surface": 0.51644, "split": 0.29458, "merge": 0.1022, "skel": 0.13091}, "mean_gate": {"surface": 0.13133, "split": 0.05281, "merge": 0.09924, "skel": 0.0996}, "n_reads": 8108, "n_cand": 29668, "read_suppressed_frac": 0.0993, "dropout_reads_logged": 8108, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2247, "admitted_surface": 59676, "staged_surface": 51140, "admitted_split": 1817, "staged_split": 45507, "admitted_merge": 4030, "staged_merge": 117768, "admitted_skel": 13, "staged_skel": 47729, "mean_count": 402.9204406738281, "n_util_obs": 353, "util_count_spearman": -0.23765934250681076}, "probe_overhead_sec_total": 740.184, "probe_overhead_seq_frac": 0.4999, "alpha": 0.03, "t": 5486.5}
{"event": "consolidate", "step": 4850, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 43, "uncertain_staged": 262084, "t": 5517.8}
{"event": "train", "step": 4850, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35243, "bpb": 1.95115, "chunk_loss": 0.030396, "F": 0.20245, "G": 0.20915, "chunks": 207.31, "bytes_per_chunk": 4.939, "boundary_entropy": 0.09915, "router_noise": 0.0, "grad_norm": 0.6908, "lr": 5.62e-05, "router_trainable": true, "sec_per_step": 0.638, "probe_sec": 0.0756, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05362, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.03842, "n_shadow": 4, "mean_delta_shadow": 0.09335, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.27838, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -9.41063, "trunc_abs_frac": 0.5473, "mean_delta_full": -0.00653, "read_hit_rate": {"surface": 0.50969, "split": 0.27714, "merge": 0.10731, "skel": 0.12492}, "mean_gate": {"surface": 0.13351, "split": 0.05388, "merge": 0.10453, "skel": 0.09385}, "n_reads": 7905, "n_cand": 29753, "read_suppressed_frac": 0.1042, "dropout_reads_logged": 7905, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2273, "admitted_surface": 59670, "staged_surface": 51035, "admitted_split": 1820, "staged_split": 45350, "admitted_merge": 4033, "staged_merge": 118042, "admitted_skel": 13, "staged_skel": 47717, "mean_count": 408.77972412109375, "n_util_obs": 359, "util_count_spearman": -0.23234403867872866}, "probe_overhead_sec_total": 744.499, "probe_overhead_seq_frac": 0.4999, "alpha": 0.03, "t": 5518.4}
{"event": "consolidate", "step": 4900, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 78, "uncertain_staged": 262084, "t": 5550.9}
{"event": "train", "step": 4900, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38146, "bpb": 1.99303, "chunk_loss": 0.03039, "F": 0.20102, "G": 0.20683, "chunks": 205.84, "bytes_per_chunk": 4.975, "boundary_entropy": 0.09789, "router_noise": 0.0, "grad_norm": 0.6568, "lr": 5.41e-05, "router_trainable": true, "sec_per_step": 0.66, "probe_sec": 0.0868, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04905, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": 0.06548, "n_shadow": 8, "mean_delta_shadow": 0.03159, "n_split": 1, "mean_delta_split": -0.40149, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.72061, "trunc_abs_frac": 0.2762, "mean_delta_full": 0.11267, "read_hit_rate": {"surface": 0.48757, "split": 0.27236, "merge": 0.08984, "skel": 0.11201}, "mean_gate": {"surface": 0.13292, "split": 0.05342, "merge": 0.10498, "skel": 0.09348}, "n_reads": 7466, "n_cand": 29579, "read_suppressed_frac": 0.1031, "dropout_reads_logged": 7466, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2276, "admitted_surface": 59670, "staged_surface": 51128, "admitted_split": 1820, "staged_split": 45657, "admitted_merge": 4033, "staged_merge": 117886, "admitted_skel": 13, "staged_skel": 47473, "mean_count": 413.17767333984375, "n_util_obs": 360, "util_count_spearman": -0.24113174741574653}, "probe_overhead_sec_total": 749.008, "probe_overhead_seq_frac": 0.4999, "alpha": 0.03, "t": 5551.4}
{"event": "consolidate", "step": 4950, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 915, "uncertain_staged": 262084, "t": 5582.9}
{"event": "train", "step": 4950, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.34101, "bpb": 1.93467, "chunk_loss": 0.030374, "F": 0.20053, "G": 0.20696, "chunks": 205.34, "bytes_per_chunk": 4.987, "boundary_entropy": 0.10476, "router_noise": 0.0, "grad_norm": 0.6725, "lr": 5.2e-05, "router_trainable": true, "sec_per_step": 0.647, "probe_sec": 0.0715, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05069, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 3, "mean_delta_loo": -0.13108, "n_shadow": 4, "mean_delta_shadow": -0.00583, "n_split": 1, "mean_delta_split": 0.02014, "n_merge": 4, "mean_delta_merge": -0.15071, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -1.32507, "trunc_abs_frac": 1.16838, "mean_delta_full": -0.03728, "read_hit_rate": {"surface": 0.50879, "split": 0.27013, "merge": 0.11618, "skel": 0.11175}, "mean_gate": {"surface": 0.13578, "split": 0.05398, "merge": 0.1025, "skel": 0.09778}, "n_reads": 7748, "n_cand": 29473, "read_suppressed_frac": 0.1035, "dropout_reads_logged": 7748, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2294, "admitted_surface": 59806, "staged_surface": 50224, "admitted_split": 1688, "staged_split": 45661, "admitted_merge": 4029, "staged_merge": 118384, "admitted_skel": 13, "staged_skel": 47875, "mean_count": 411.0582275390625, "n_util_obs": 348, "util_count_spearman": -0.2783864432290362}, "probe_overhead_sec_total": 753.371, "probe_overhead_seq_frac": 0.4999, "alpha": 0.03, "t": 5583.8}
{"event": "consolidate", "step": 5000, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 1024, "uncertain_staged": 262084, "t": 5616.6}
{"event": "probe", "step": 5000, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3500.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.201115, "cps": 27.34375, "mean_p": 0.208981, "boundary_entropy": 0.103521, "bpb": 1.895447}, "tr": {"chunks": 4118.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.227476, "cps": 32.171875, "mean_p": 0.20888, "boundary_entropy": 0.103354, "bpb": 2.017399}, "fi": {"chunks": 4326.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.229972, "cps": 33.796875, "mean_p": 0.23022, "boundary_entropy": 0.107864, "bpb": 2.138498}}, "bpb": {"en": 1.89545, "tr": 2.0174, "fi": 2.1385}, "gini_cpc": 0.029212, "gini_cps": 0.046104, "bpic": 4.749833, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.1310762067849844, "fi": 1.1434856201158896}, "inventory": {"size": 326654, "admitted": 65536, "staged": 261118, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2330, "admitted_surface": 59938, "staged_surface": 49458, "admitted_split": 1559, "staged_split": 45235, "admitted_merge": 4025, "staged_merge": 118789, "admitted_skel": 14, "staged_skel": 47636, "mean_count": 412.64202880859375, "n_util_obs": 350, "util_count_spearman": -0.2813971674439343}, "t": 5618.0}
{"event": "eval", "step": 5000, "phase": 3, "bpb": {"en": 1.90904, "tr": 2.02584, "fi": 2.14394}, "bytes_per_chunk": {"en": 4.974, "tr": 4.862, "fi": 4.845}, "bpb_mean": 2.02627, "frozen_inventory": true, "t": 5620.4}
{"event": "train", "step": 5000, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.3313, "bpb": 1.92066, "chunk_loss": 0.030377, "F": 0.20184, "G": 0.20868, "chunks": 206.69, "bytes_per_chunk": 4.954, "boundary_entropy": 0.10102, "router_noise": 0.0, "grad_norm": 0.6337, "lr": 5e-05, "router_trainable": true, "sec_per_step": 0.821, "probe_sec": 0.0621, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03835, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 3, "mean_delta_loo": -0.00522, "n_shadow": 8, "mean_delta_shadow": 0.04009, "n_split": 1, "mean_delta_split": -0.02136, "n_merge": 1, "mean_delta_merge": 0.09515, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.30315, "trunc_abs_frac": 0.4788, "mean_delta_full": -0.12976, "read_hit_rate": {"surface": 0.53008, "split": 0.3063, "merge": 0.1055, "skel": 0.13091}, "mean_gate": {"surface": 0.12939, "split": 0.05158, "merge": 0.09686, "skel": 0.09856}, "n_reads": 8400, "n_cand": 29805, "read_suppressed_frac": 0.1, "dropout_reads_logged": 8400, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2332, "admitted_surface": 59938, "staged_surface": 49596, "admitted_split": 1559, "staged_split": 45366, "admitted_merge": 4025, "staged_merge": 119371, "admitted_skel": 14, "staged_skel": 47811, "mean_count": 412.77020263671875, "n_util_obs": 350, "util_count_spearman": -0.2814725178826002}, "probe_overhead_sec_total": 757.781, "probe_overhead_seq_frac": 0.4999, "alpha": 0.03, "t": 5624.8}
{"event": "consolidate", "step": 5050, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 24, "uncertain_staged": 262085, "t": 5654.2}
{"event": "train", "step": 5050, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35082, "bpb": 1.94882, "chunk_loss": 0.0304, "F": 0.20267, "G": 0.20891, "chunks": 207.53, "bytes_per_chunk": 4.934, "boundary_entropy": 0.10007, "router_noise": 0.0, "grad_norm": 0.6732, "lr": 4.81e-05, "router_trainable": true, "sec_per_step": 0.598, "probe_sec": 0.1185, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10033, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": -0.13415, "n_shadow": 7, "mean_delta_shadow": 0.07226, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.00382, "trunc_abs_frac": 0.05411, "mean_delta_full": -0.07983, "read_hit_rate": {"surface": 0.5255, "split": 0.29031, "merge": 0.10643, "skel": 0.13077}, "mean_gate": {"surface": 0.13, "split": 0.05152, "merge": 0.09705, "skel": 0.09834}, "n_reads": 8258, "n_cand": 30077, "read_suppressed_frac": 0.1069, "dropout_reads_logged": 8258, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2359, "admitted_surface": 59939, "staged_surface": 50395, "admitted_split": 1559, "staged_split": 45620, "admitted_merge": 4024, "staged_merge": 118093, "admitted_skel": 14, "staged_skel": 48036, "mean_count": 417.529541015625, "n_util_obs": 357, "util_count_spearman": -0.2854298959621675}, "probe_overhead_sec_total": 761.713, "probe_overhead_seq_frac": 0.4999, "alpha": 0.03, "t": 5654.7}
{"event": "consolidate", "step": 5100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 378, "uncertain_staged": 262085, "t": 5688.6}
{"event": "train", "step": 5100, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.37602, "bpb": 1.98518, "chunk_loss": 0.030478, "F": 0.20697, "G": 0.21209, "chunks": 211.94, "bytes_per_chunk": 4.832, "boundary_entropy": 0.10207, "router_noise": 0.0, "grad_norm": 0.7272, "lr": 4.63e-05, "router_trainable": true, "sec_per_step": 0.687, "probe_sec": 0.0689, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04944, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": 0.07289, "n_shadow": 5, "mean_delta_shadow": 0.01788, "n_split": 1, "mean_delta_split": 0.0564, "n_merge": 1, "mean_delta_merge": 0.005, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.32097, "trunc_abs_frac": 0.8821, "mean_delta_full": 0.21968, "read_hit_rate": {"surface": 0.50607, "split": 0.25514, "merge": 0.10375, "skel": 0.11866}, "mean_gate": {"surface": 0.12798, "split": 0.05123, "merge": 0.09769, "skel": 0.08993}, "n_reads": 7717, "n_cand": 30294, "read_suppressed_frac": 0.096, "dropout_reads_logged": 7717, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2380, "admitted_surface": 60007, "staged_surface": 50183, "admitted_split": 1509, "staged_split": 45471, "admitted_merge": 4007, "staged_merge": 118713, "admitted_skel": 13, "staged_skel": 47777, "mean_count": 414.46319580078125, "n_util_obs": 355, "util_count_spearman": -0.2667921066678111}, "probe_overhead_sec_total": 766.525, "probe_overhead_seq_frac": 0.4999, "alpha": 0.03, "t": 5689.1}
{"event": "consolidate", "step": 5150, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 164, "uncertain_staged": 262085, "t": 5719.3}
{"event": "train", "step": 5150, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.37152, "bpb": 1.97868, "chunk_loss": 0.03052, "F": 0.20743, "G": 0.2138, "chunks": 212.41, "bytes_per_chunk": 4.821, "boundary_entropy": 0.10154, "router_noise": 0.0, "grad_norm": 0.6862, "lr": 4.45e-05, "router_trainable": true, "sec_per_step": 0.615, "probe_sec": 0.0822, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05552, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": 0.05751, "n_shadow": 3, "mean_delta_shadow": -0.19248, "n_split": 2, "mean_delta_split": -0.23064, "n_merge": 2, "mean_delta_merge": -0.62646, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 1.39483, "trunc_abs_frac": 0.60668, "mean_delta_full": -0.12109, "read_hit_rate": {"surface": 0.50333, "split": 0.27347, "merge": 0.10723, "skel": 0.12073}, "mean_gate": {"surface": 0.13026, "split": 0.05197, "merge": 0.09682, "skel": 0.09528}, "n_reads": 7929, "n_cand": 30195, "read_suppressed_frac": 0.0974, "dropout_reads_logged": 7929, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2410, "admitted_surface": 60095, "staged_surface": 50274, "admitted_split": 1439, "staged_split": 45847, "admitted_merge": 3989, "staged_merge": 118290, "admitted_skel": 13, "staged_skel": 47733, "mean_count": 419.6251220703125, "n_util_obs": 357, "util_count_spearman": -0.2851495377022903}, "probe_overhead_sec_total": 770.614, "probe_overhead_seq_frac": 0.4999, "alpha": 0.03, "t": 5719.8}
{"event": "consolidate", "step": 5200, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 1024, "uncertain_staged": 262086, "t": 5750.3}
{"event": "probe", "step": 5200, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3518.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.202149, "cps": 27.484375, "mean_p": 0.20979, "boundary_entropy": 0.104415, "bpb": 1.88857}, "tr": {"chunks": 4168.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.230238, "cps": 32.5625, "mean_p": 0.211016, "boundary_entropy": 0.09854, "bpb": 2.008907}, "fi": {"chunks": 4323.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.229812, "cps": 33.773438, "mean_p": 0.230459, "boundary_entropy": 0.10753, "bpb": 2.129512}}, "bpb": {"en": 1.88857, "tr": 2.00891, "fi": 2.12951}, "gini_cpc": 0.028279, "gini_cps": 0.044689, "bpic": 4.724124, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.1389520587094686, "fi": 1.1368459976894123}, "inventory": {"size": 326654, "admitted": 65536, "staged": 261118, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2377, "admitted_surface": 61079, "staged_surface": 49504, "admitted_split": 1110, "staged_split": 45700, "admitted_merge": 3334, "staged_merge": 118256, "admitted_skel": 13, "staged_skel": 47658, "mean_count": 364.23004150390625, "n_util_obs": 341, "util_count_spearman": -0.3255816663144813}, "t": 5752.0}
{"event": "eval", "step": 5200, "phase": 3, "bpb": {"en": 1.9023, "tr": 2.01657, "fi": 2.13514}, "bytes_per_chunk": {"en": 4.966, "tr": 4.811, "fi": 4.855}, "bpb_mean": 2.018, "frozen_inventory": true, "t": 5755.1}
{"event": "train", "step": 5200, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32393, "bpb": 1.91003, "chunk_loss": 0.030293, "F": 0.19733, "G": 0.20439, "chunks": 202.06, "bytes_per_chunk": 5.068, "boundary_entropy": 0.09868, "router_noise": 0.0, "grad_norm": 0.6693, "lr": 4.29e-05, "router_trainable": true, "sec_per_step": 0.791, "probe_sec": 0.0784, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0467, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.12566, "n_shadow": 3, "mean_delta_shadow": -0.08635, "n_split": 1, "mean_delta_split": 0.02539, "n_merge": 2, "mean_delta_merge": -0.31071, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.12291, "trunc_abs_frac": 0.2474, "mean_delta_full": 0.24844, "read_hit_rate": {"surface": 0.52129, "split": 0.18876, "merge": 0.09981, "skel": 0.12531}, "mean_gate": {"surface": 0.12663, "split": 0.05492, "merge": 0.0958, "skel": 0.09235}, "n_reads": 6821, "n_cand": 29559, "read_suppressed_frac": 0.1028, "dropout_reads_logged": 6821, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2379, "admitted_surface": 61079, "staged_surface": 49721, "admitted_split": 1110, "staged_split": 45964, "admitted_merge": 3334, "staged_merge": 118647, "admitted_skel": 13, "staged_skel": 47812, "mean_count": 364.3341064453125, "n_util_obs": 341, "util_count_spearman": -0.3259692702517031}, "probe_overhead_sec_total": 774.652, "probe_overhead_seq_frac": 0.4999, "alpha": 0.03, "t": 5759.3}
{"event": "consolidate", "step": 5250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262083, "t": 5790.4}
{"event": "train", "step": 5250, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32022, "bpb": 1.90468, "chunk_loss": 0.030423, "F": 0.20389, "G": 0.21082, "chunks": 208.78, "bytes_per_chunk": 4.905, "boundary_entropy": 0.10196, "router_noise": 0.0, "grad_norm": 0.6539, "lr": 4.14e-05, "router_trainable": true, "sec_per_step": 0.638, "probe_sec": 0.0693, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04248, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.01413, "n_shadow": 3, "mean_delta_shadow": -0.06767, "n_split": 2, "mean_delta_split": 0.11462, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 0, "read_hit_rate": {"surface": 0.55557, "split": 0.2155, "merge": 0.11123, "skel": 0.12517}, "mean_gate": {"surface": 0.12773, "split": 0.05398, "merge": 0.09606, "skel": 0.09313}, "n_reads": 7613, "n_cand": 30193, "read_suppressed_frac": 0.1032, "dropout_reads_logged": 7613, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2390, "admitted_surface": 61075, "staged_surface": 49171, "admitted_split": 1114, "staged_split": 46065, "admitted_merge": 3334, "staged_merge": 118940, "admitted_skel": 13, "staged_skel": 47968, "mean_count": 369.36602783203125, "n_util_obs": 343, "util_count_spearman": -0.3175098278310054}, "probe_overhead_sec_total": 778.942, "probe_overhead_seq_frac": 0.4999, "alpha": 0.03, "t": 5791.2}
{"event": "consolidate", "step": 5300, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 42, "uncertain_staged": 262083, "t": 5821.5}
{"event": "train", "step": 5300, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.37861, "bpb": 1.98891, "chunk_loss": 0.030443, "F": 0.20551, "G": 0.21096, "chunks": 210.44, "bytes_per_chunk": 4.866, "boundary_entropy": 0.09555, "router_noise": 0.0, "grad_norm": 0.6829, "lr": 3.99e-05, "router_trainable": true, "sec_per_step": 0.615, "probe_sec": 0.0909, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06899, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": 0.05333, "n_shadow": 4, "mean_delta_shadow": 0.05312, "n_split": 2, "mean_delta_split": -0.04348, "n_merge": 2, "mean_delta_merge": 0.09979, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 9.03765, "trunc_abs_frac": 3.24381, "mean_delta_full": -0.01351, "read_hit_rate": {"surface": 0.53954, "split": 0.22775, "merge": 0.1003, "skel": 0.12531}, "mean_gate": {"surface": 0.12981, "split": 0.05386, "merge": 0.09987, "skel": 0.09266}, "n_reads": 7652, "n_cand": 30503, "read_suppressed_frac": 0.0983, "dropout_reads_logged": 7652, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2405, "admitted_surface": 61067, "staged_surface": 49471, "admitted_split": 1124, "staged_split": 45825, "admitted_merge": 3332, "staged_merge": 119099, "admitted_skel": 13, "staged_skel": 47749, "mean_count": 374.0858459472656, "n_util_obs": 354, "util_count_spearman": -0.31619401362046085}, "probe_overhead_sec_total": 782.98, "probe_overhead_seq_frac": 0.4999, "alpha": 0.03, "t": 5821.9}
{"event": "consolidate", "step": 5350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 13, "uncertain_staged": 262084, "t": 5852.2}
{"event": "train", "step": 5350, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.30618, "bpb": 1.88441, "chunk_loss": 0.030315, "F": 0.19907, "G": 0.20685, "chunks": 203.84, "bytes_per_chunk": 5.023, "boundary_entropy": 0.10021, "router_noise": 0.0, "grad_norm": 0.7008, "lr": 3.86e-05, "router_trainable": true, "sec_per_step": 0.617, "probe_sec": 0.1511, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08365, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": -0.02895, "n_shadow": 1, "mean_delta_shadow": -0.04419, "n_split": 2, "mean_delta_split": 0.11705, "n_merge": 1, "mean_delta_merge": 0.10742, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.19525, "trunc_abs_frac": 0.07374, "mean_delta_full": -0.12207, "read_hit_rate": {"surface": 0.54352, "split": 0.25422, "merge": 0.09707, "skel": 0.12585}, "mean_gate": {"surface": 0.12517, "split": 0.05257, "merge": 0.09631, "skel": 0.08963}, "n_reads": 7756, "n_cand": 29711, "read_suppressed_frac": 0.0998, "dropout_reads_logged": 7756, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2417, "admitted_surface": 61067, "staged_surface": 49975, "admitted_split": 1126, "staged_split": 45965, "admitted_merge": 3330, "staged_merge": 118543, "admitted_skel": 13, "staged_skel": 47661, "mean_count": 378.7220458984375, "n_util_obs": 357, "util_count_spearman": -0.32043256926486674}, "probe_overhead_sec_total": 787.087, "probe_overhead_seq_frac": 0.4999, "alpha": 0.03, "t": 5852.8}
{"event": "consolidate", "step": 5400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 9, "uncertain_staged": 262082, "t": 5884.6}
{"event": "probe", "step": 5400, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3509.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.201632, "cps": 27.414062, "mean_p": 0.209301, "boundary_entropy": 0.102444, "bpb": 1.884247}, "tr": {"chunks": 4167.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.230183, "cps": 32.554688, "mean_p": 0.212545, "boundary_entropy": 0.098177, "bpb": 2.001924}, "fi": {"chunks": 4298.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.228483, "cps": 33.578125, "mean_p": 0.229667, "boundary_entropy": 0.105613, "bpb": 2.122283}}, "bpb": {"en": 1.88425, "tr": 2.00192, "fi": 2.12228}, "gini_cpc": 0.028826, "gini_cps": 0.043929, "bpic": 4.737932, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.1415993189410263, "fi": 1.1331705515586121}, "inventory": {"size": 327671, "admitted": 65536, "staged": 262135, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2453, "admitted_surface": 61064, "staged_surface": 50071, "admitted_split": 1126, "staged_split": 45954, "admitted_merge": 3332, "staged_merge": 118189, "admitted_skel": 14, "staged_skel": 47921, "mean_count": 385.0972900390625, "n_util_obs": 367, "util_count_spearman": -0.31394058724035034}, "t": 5886.0}
{"event": "eval", "step": 5400, "phase": 3, "bpb": {"en": 1.89762, "tr": 2.00992, "fi": 2.12833}, "bytes_per_chunk": {"en": 4.969, "tr": 4.817, "fi": 4.867}, "bpb_mean": 2.01196, "frozen_inventory": true, "t": 5888.3}
{"event": "train", "step": 5400, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.37943, "bpb": 1.9901, "chunk_loss": 0.030572, "F": 0.20743, "G": 0.21489, "chunks": 212.41, "bytes_per_chunk": 4.821, "boundary_entropy": 0.0991, "router_noise": 0.0, "grad_norm": 0.7367, "lr": 3.73e-05, "router_trainable": true, "sec_per_step": 0.791, "probe_sec": 0.0799, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05881, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": 0.09364, "n_shadow": 8, "mean_delta_shadow": -0.01514, "n_split": 1, "mean_delta_split": 0.0733, "n_merge": 1, "mean_delta_merge": -0.26141, "n_dropped_unpaired": 0, "n_full": 7, "trunc_ratio_sum": 0.03599, "trunc_abs_frac": 2.55191, "mean_delta_full": -0.12308, "read_hit_rate": {"surface": 0.53659, "split": 0.23611, "merge": 0.08495, "skel": 0.13709}, "mean_gate": {"surface": 0.1258, "split": 0.05221, "merge": 0.09756, "skel": 0.09149}, "n_reads": 7716, "n_cand": 30574, "read_suppressed_frac": 0.0994, "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": 2454, "admitted_surface": 61064, "staged_surface": 50087, "admitted_split": 1126, "staged_split": 45964, "admitted_merge": 3332, "staged_merge": 118208, "admitted_skel": 14, "staged_skel": 47885, "mean_count": 385.21502685546875, "n_util_obs": 367, "util_count_spearman": -0.3139042748023406}, "probe_overhead_sec_total": 791.17, "probe_overhead_seq_frac": 0.4999, "alpha": 0.03, "t": 5892.4}
{"event": "consolidate", "step": 5450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 10, "uncertain_staged": 262081, "t": 5921.9}
{"event": "train", "step": 5450, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.31761, "bpb": 1.9009, "chunk_loss": 0.030589, "F": 0.21124, "G": 0.21762, "chunks": 216.31, "bytes_per_chunk": 4.734, "boundary_entropy": 0.09746, "router_noise": 0.0, "grad_norm": 0.6992, "lr": 3.62e-05, "router_trainable": true, "sec_per_step": 0.615, "probe_sec": 0.158, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06134, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": -0.00893, "n_shadow": 6, "mean_delta_shadow": 0.00869, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": -0.17319, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.29037, "trunc_abs_frac": 0.61831, "mean_delta_full": -0.05197, "read_hit_rate": {"surface": 0.5521, "split": 0.29606, "merge": 0.11855, "skel": 0.13005}, "mean_gate": {"surface": 0.12962, "split": 0.0521, "merge": 0.1032, "skel": 0.09394}, "n_reads": 8848, "n_cand": 30928, "read_suppressed_frac": 0.0986, "dropout_reads_logged": 8848, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2486, "admitted_surface": 61057, "staged_surface": 50283, "admitted_split": 1133, "staged_split": 45570, "admitted_merge": 3332, "staged_merge": 118563, "admitted_skel": 14, "staged_skel": 47728, "mean_count": 392.435791015625, "n_util_obs": 375, "util_count_spearman": -0.31912506082176983}, "probe_overhead_sec_total": 795.292, "probe_overhead_seq_frac": 0.4999, "alpha": 0.03, "t": 5923.1}
{"event": "consolidate", "step": 5500, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 63, "uncertain_staged": 262082, "t": 5954.7}
{"event": "train", "step": 5500, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.30415, "bpb": 1.8815, "chunk_loss": 0.030501, "F": 0.20685, "G": 0.21309, "chunks": 211.81, "bytes_per_chunk": 4.834, "boundary_entropy": 0.10089, "router_noise": 0.0, "grad_norm": 0.6712, "lr": 3.51e-05, "router_trainable": true, "sec_per_step": 0.645, "probe_sec": 0.1162, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05785, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": -0.00994, "n_shadow": 4, "mean_delta_shadow": 0.02186, "n_split": 0, "mean_delta_split": NaN, "n_merge": 4, "mean_delta_merge": 0.27154, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -1.86209, "trunc_abs_frac": 1.74651, "mean_delta_full": -0.06498, "read_hit_rate": {"surface": 0.54566, "split": 0.29489, "merge": 0.11096, "skel": 0.12811}, "mean_gate": {"surface": 0.12726, "split": 0.05079, "merge": 0.10044, "skel": 0.0926}, "n_reads": 8601, "n_cand": 30549, "read_suppressed_frac": 0.1009, "dropout_reads_logged": 8601, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2509, "admitted_surface": 61048, "staged_surface": 50377, "admitted_split": 1141, "staged_split": 45485, "admitted_merge": 3333, "staged_merge": 118712, "admitted_skel": 14, "staged_skel": 47570, "mean_count": 397.5195007324219, "n_util_obs": 377, "util_count_spearman": -0.33007841216074557}, "probe_overhead_sec_total": 799.458, "probe_overhead_seq_frac": 0.4999, "alpha": 0.03, "t": 5955.3}
{"event": "consolidate", "step": 5550, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 52, "uncertain_staged": 262083, "t": 5987.4}
{"event": "train", "step": 5550, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.36102, "bpb": 1.96354, "chunk_loss": 0.030411, "F": 0.203, "G": 0.20851, "chunks": 207.88, "bytes_per_chunk": 4.926, "boundary_entropy": 0.09879, "router_noise": 0.0, "grad_norm": 0.7496, "lr": 3.41e-05, "router_trainable": true, "sec_per_step": 0.659, "probe_sec": 0.0683, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04264, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": 0.01589, "n_shadow": 5, "mean_delta_shadow": -0.07275, "n_split": 0, "mean_delta_split": NaN, "n_merge": 4, "mean_delta_merge": 0.02422, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -0.34436, "trunc_abs_frac": 0.53697, "mean_delta_full": 0.02191, "read_hit_rate": {"surface": 0.52795, "split": 0.28326, "merge": 0.10656, "skel": 0.11052}, "mean_gate": {"surface": 0.1273, "split": 0.0522, "merge": 0.09736, "skel": 0.09258}, "n_reads": 7988, "n_cand": 29582, "read_suppressed_frac": 0.1043, "dropout_reads_logged": 7988, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2524, "admitted_surface": 61048, "staged_surface": 50327, "admitted_split": 1141, "staged_split": 45704, "admitted_merge": 3333, "staged_merge": 118518, "admitted_skel": 14, "staged_skel": 47595, "mean_count": 402.42034912109375, "n_util_obs": 382, "util_count_spearman": -0.336738988399893}, "probe_overhead_sec_total": 803.8, "probe_overhead_seq_frac": 0.4999, "alpha": 0.03, "t": 5988.3}
{"event": "consolidate", "step": 5600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 88, "uncertain_staged": 262084, "t": 6019.5}
{"event": "probe", "step": 5600, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3510.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.201689, "cps": 27.421875, "mean_p": 0.208989, "boundary_entropy": 0.102, "bpb": 1.87938}, "tr": {"chunks": 4137.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.228526, "cps": 32.320312, "mean_p": 0.210359, "boundary_entropy": 0.099509, "bpb": 1.995237}, "fi": {"chunks": 4307.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.228962, "cps": 33.648438, "mean_p": 0.229922, "boundary_entropy": 0.106052, "bpb": 2.113652}}, "bpb": {"en": 1.87938, "tr": 1.99524, "fi": 2.11365}, "gini_cpc": 0.027582, "gini_cps": 0.044448, "bpic": 4.745859, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.1330575609369178, "fi": 1.1352198908894457}, "inventory": {"size": 327592, "admitted": 65536, "staged": 262056, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2546, "admitted_surface": 61046, "staged_surface": 50193, "admitted_split": 1142, "staged_split": 45779, "admitted_merge": 3334, "staged_merge": 118577, "admitted_skel": 14, "staged_skel": 47507, "mean_count": 407.74774169921875, "n_util_obs": 389, "util_count_spearman": -0.32270134107273546}, "t": 6021.1}
{"event": "eval", "step": 5600, "phase": 3, "bpb": {"en": 1.89328, "tr": 2.00228, "fi": 2.12004}, "bytes_per_chunk": {"en": 4.974, "tr": 4.851, "fi": 4.859}, "bpb_mean": 2.0052, "frozen_inventory": true, "t": 6023.8}
{"event": "train", "step": 5600, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.34688, "bpb": 1.94313, "chunk_loss": 0.0304, "F": 0.20218, "G": 0.20824, "chunks": 207.03, "bytes_per_chunk": 4.946, "boundary_entropy": 0.09865, "router_noise": 0.0, "grad_norm": 0.7367, "lr": 3.33e-05, "router_trainable": true, "sec_per_step": 0.79, "probe_sec": 0.0642, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04139, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.09155, "n_shadow": 4, "mean_delta_shadow": -0.02567, "n_split": 3, "mean_delta_split": 0.03261, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.97212, "trunc_abs_frac": 2.26355, "mean_delta_full": 0.07608, "read_hit_rate": {"surface": 0.54194, "split": 0.26134, "merge": 0.09252, "skel": 0.12461}, "mean_gate": {"surface": 0.1252, "split": 0.05035, "merge": 0.0983, "skel": 0.08748}, "n_reads": 7912, "n_cand": 30142, "read_suppressed_frac": 0.1012, "dropout_reads_logged": 7912, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2548, "admitted_surface": 61046, "staged_surface": 50206, "admitted_split": 1142, "staged_split": 45804, "admitted_merge": 3334, "staged_merge": 118591, "admitted_skel": 14, "staged_skel": 47543, "mean_count": 407.86846923828125, "n_util_obs": 389, "util_count_spearman": -0.3229075470469191}, "probe_overhead_sec_total": 807.965, "probe_overhead_seq_frac": 0.4999, "alpha": 0.03, "t": 6027.8}
{"event": "consolidate", "step": 5650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 102, "uncertain_staged": 262084, "t": 6059.5}
{"event": "train", "step": 5650, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.33384, "bpb": 1.92433, "chunk_loss": 0.030477, "F": 0.2063, "G": 0.21233, "chunks": 211.25, "bytes_per_chunk": 4.847, "boundary_entropy": 0.09783, "router_noise": 0.0, "grad_norm": 0.6885, "lr": 3.25e-05, "router_trainable": true, "sec_per_step": 0.643, "probe_sec": 0.0824, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05356, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.01002, "n_shadow": 6, "mean_delta_shadow": 0.1027, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 0.51126, "trunc_abs_frac": 0.91436, "mean_delta_full": 0.13757, "read_hit_rate": {"surface": 0.55366, "split": 0.29071, "merge": 0.1135, "skel": 0.13514}, "mean_gate": {"surface": 0.12283, "split": 0.05036, "merge": 0.0913, "skel": 0.09069}, "n_reads": 8648, "n_cand": 30528, "read_suppressed_frac": 0.0957, "dropout_reads_logged": 8648, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2578, "admitted_surface": 61035, "staged_surface": 50051, "admitted_split": 1152, "staged_split": 45574, "admitted_merge": 3335, "staged_merge": 118669, "admitted_skel": 14, "staged_skel": 47850, "mean_count": 414.32879638671875, "n_util_obs": 398, "util_count_spearman": -0.3246376350911221}, "probe_overhead_sec_total": 812.327, "probe_overhead_seq_frac": 0.4999, "alpha": 0.03, "t": 6060.0}
{"event": "consolidate", "step": 5700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 372, "uncertain_staged": 262084, "t": 6091.7}
{"event": "train", "step": 5700, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.34088, "bpb": 1.93448, "chunk_loss": 0.030332, "F": 0.20004, "G": 0.20561, "chunks": 204.84, "bytes_per_chunk": 4.999, "boundary_entropy": 0.09968, "router_noise": 0.0, "grad_norm": 0.7288, "lr": 3.18e-05, "router_trainable": true, "sec_per_step": 0.654, "probe_sec": 0.0607, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03984, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.08808, "n_shadow": 3, "mean_delta_shadow": -0.0124, "n_split": 0, "mean_delta_split": NaN, "n_merge": 3, "mean_delta_merge": 0.57529, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.74597, "trunc_abs_frac": 0.74597, "mean_delta_full": -0.40894, "read_hit_rate": {"surface": 0.55143, "split": 0.28434, "merge": 0.11077, "skel": 0.13007}, "mean_gate": {"surface": 0.12427, "split": 0.05099, "merge": 0.09394, "skel": 0.09132}, "n_reads": 8261, "n_cand": 29616, "read_suppressed_frac": 0.102, "dropout_reads_logged": 8261, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2608, "admitted_surface": 61030, "staged_surface": 49607, "admitted_split": 1157, "staged_split": 46048, "admitted_merge": 3334, "staged_merge": 118837, "admitted_skel": 15, "staged_skel": 47652, "mean_count": 420.337158203125, "n_util_obs": 406, "util_count_spearman": -0.32306980720710154}, "probe_overhead_sec_total": 816.664, "probe_overhead_seq_frac": 0.4999, "alpha": 0.03, "t": 6092.6}
{"event": "consolidate", "step": 5750, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 363, "uncertain_staged": 262085, "t": 6125.5}
{"event": "train", "step": 5750, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.39506, "bpb": 2.01264, "chunk_loss": 0.03046, "F": 0.20517, "G": 0.21158, "chunks": 210.09, "bytes_per_chunk": 4.874, "boundary_entropy": 0.09859, "router_noise": 0.0, "grad_norm": 0.7461, "lr": 3.13e-05, "router_trainable": true, "sec_per_step": 0.666, "probe_sec": 0.0632, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04096, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.05048, "n_shadow": 6, "mean_delta_shadow": -0.0562, "n_split": 1, "mean_delta_split": -0.35645, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -1.96407, "trunc_abs_frac": 0.60171, "mean_delta_full": 0.02463, "read_hit_rate": {"surface": 0.53445, "split": 0.23759, "merge": 0.09551, "skel": 0.14123}, "mean_gate": {"surface": 0.11975, "split": 0.05071, "merge": 0.09291, "skel": 0.08671}, "n_reads": 7719, "n_cand": 30181, "read_suppressed_frac": 0.1021, "dropout_reads_logged": 7719, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2628, "admitted_surface": 61034, "staged_surface": 49541, "admitted_split": 1153, "staged_split": 46086, "admitted_merge": 3334, "staged_merge": 118734, "admitted_skel": 15, "staged_skel": 47783, "mean_count": 422.305908203125, "n_util_obs": 405, "util_count_spearman": -0.31222950826927715}, "probe_overhead_sec_total": 821.122, "probe_overhead_seq_frac": 0.4999, "alpha": 0.03, "t": 6125.9}
{"event": "consolidate", "step": 5800, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 9, "uncertain_staged": 262084, "t": 6157.5}
{"event": "probe", "step": 5800, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3504.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.201345, "cps": 27.375, "mean_p": 0.208435, "boundary_entropy": 0.101504, "bpb": 1.876041}, "tr": {"chunks": 4115.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.22731, "cps": 32.148438, "mean_p": 0.20899, "boundary_entropy": 0.100348, "bpb": 1.99206}, "fi": {"chunks": 4302.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.228696, "cps": 33.609375, "mean_p": 0.228968, "boundary_entropy": 0.102707, "bpb": 2.108552}}, "bpb": {"en": 1.87604, "tr": 1.99206, "fi": 2.10855}, "gini_cpc": 0.027739, "gini_cps": 0.044627, "bpic": 4.758997, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.1289619653595597, "fi": 1.1358436261790863}, "inventory": {"size": 327669, "admitted": 65536, "staged": 262133, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2659, "admitted_surface": 61035, "staged_surface": 49532, "admitted_split": 1152, "staged_split": 45868, "admitted_merge": 3334, "staged_merge": 119028, "admitted_skel": 15, "staged_skel": 47705, "mean_count": 426.70135498046875, "n_util_obs": 409, "util_count_spearman": -0.30335583371240415}, "t": 6159.1}
{"event": "eval", "step": 5800, "phase": 3, "bpb": {"en": 1.8895, "tr": 1.999, "fi": 2.11562}, "bytes_per_chunk": {"en": 4.983, "tr": 4.877, "fi": 4.859}, "bpb_mean": 2.00138, "frozen_inventory": true, "t": 6161.4}
{"event": "train", "step": 5800, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.31994, "bpb": 1.90427, "chunk_loss": 0.030517, "F": 0.20419, "G": 0.20934, "chunks": 209.09, "bytes_per_chunk": 4.897, "boundary_entropy": 0.09887, "router_noise": 0.0, "grad_norm": 0.7253, "lr": 3.08e-05, "router_trainable": true, "sec_per_step": 0.786, "probe_sec": 0.0607, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04012, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.23022, "n_shadow": 3, "mean_delta_shadow": 0.04712, "n_split": 2, "mean_delta_split": -0.14252, "n_merge": 1, "mean_delta_merge": 0.05536, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.63695, "trunc_abs_frac": 0.58075, "mean_delta_full": 0.46577, "read_hit_rate": {"surface": 0.54528, "split": 0.25149, "merge": 0.10442, "skel": 0.12547}, "mean_gate": {"surface": 0.12303, "split": 0.05084, "merge": 0.09569, "skel": 0.09454}, "n_reads": 7900, "n_cand": 30021, "read_suppressed_frac": 0.1027, "dropout_reads_logged": 7900, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2661, "admitted_surface": 61035, "staged_surface": 49538, "admitted_split": 1152, "staged_split": 45883, "admitted_merge": 3334, "staged_merge": 119029, "admitted_skel": 15, "staged_skel": 47694, "mean_count": 426.8218994140625, "n_util_obs": 409, "util_count_spearman": -0.30749021369002216}, "probe_overhead_sec_total": 825.581, "probe_overhead_seq_frac": 0.4999, "alpha": 0.03, "t": 6165.2}
{"event": "consolidate", "step": 5850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 8, "uncertain_staged": 262084, "t": 6195.1}
{"event": "train", "step": 5850, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.34334, "bpb": 1.93803, "chunk_loss": 0.030416, "F": 0.20273, "G": 0.20983, "chunks": 207.59, "bytes_per_chunk": 4.933, "boundary_entropy": 0.09614, "router_noise": 0.0, "grad_norm": 0.7281, "lr": 3.05e-05, "router_trainable": true, "sec_per_step": 0.614, "probe_sec": 0.0682, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04182, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 11, "mean_delta_loo": 0.03006, "n_shadow": 1, "mean_delta_shadow": 0.32608, "n_split": 1, "mean_delta_split": -0.09875, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.42036, "trunc_abs_frac": 0.58635, "mean_delta_full": 0.46277, "read_hit_rate": {"surface": 0.54288, "split": 0.25589, "merge": 0.09378, "skel": 0.13899}, "mean_gate": {"surface": 0.1219, "split": 0.05016, "merge": 0.09415, "skel": 0.08818}, "n_reads": 7965, "n_cand": 30312, "read_suppressed_frac": 0.0948, "dropout_reads_logged": 7965, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2684, "admitted_surface": 61037, "staged_surface": 49746, "admitted_split": 1150, "staged_split": 45935, "admitted_merge": 3334, "staged_merge": 118705, "admitted_skel": 15, "staged_skel": 47758, "mean_count": 431.3981628417969, "n_util_obs": 412, "util_count_spearman": -0.30241510054600645}, "probe_overhead_sec_total": 829.812, "probe_overhead_seq_frac": 0.4999, "alpha": 0.03, "t": 6195.9}
{"event": "consolidate", "step": 5900, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1, "uncertain_staged": 262082, "t": 6224.2}
{"event": "train", "step": 5900, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.27394, "bpb": 1.83791, "chunk_loss": 0.030315, "F": 0.19937, "G": 0.20586, "chunks": 204.16, "bytes_per_chunk": 5.016, "boundary_entropy": 0.09614, "router_noise": 0.0, "grad_norm": 0.7104, "lr": 3.02e-05, "router_trainable": true, "sec_per_step": 0.574, "probe_sec": 0.0644, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04316, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.03511, "n_shadow": 4, "mean_delta_shadow": -0.04348, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.34679, "trunc_abs_frac": 0.88632, "mean_delta_full": 0.34115, "read_hit_rate": {"surface": 0.55315, "split": 0.26198, "merge": 0.10218, "skel": 0.12094}, "mean_gate": {"surface": 0.12399, "split": 0.0508, "merge": 0.09539, "skel": 0.08953}, "n_reads": 7887, "n_cand": 29546, "read_suppressed_frac": 0.0942, "dropout_reads_logged": 7887, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2708, "admitted_surface": 61037, "staged_surface": 49826, "admitted_split": 1150, "staged_split": 45606, "admitted_merge": 3334, "staged_merge": 119024, "admitted_skel": 15, "staged_skel": 47688, "mean_count": 437.07269287109375, "n_util_obs": 419, "util_count_spearman": -0.29562527612799416}, "probe_overhead_sec_total": 833.499, "probe_overhead_seq_frac": 0.4999, "alpha": 0.03, "t": 6224.7}
{"event": "consolidate", "step": 5950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 452, "uncertain_staged": 262081, "t": 6253.9}
{"event": "train", "step": 5950, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38068, "bpb": 1.99189, "chunk_loss": 0.030415, "F": 0.20392, "G": 0.20963, "chunks": 208.81, "bytes_per_chunk": 4.904, "boundary_entropy": 0.09725, "router_noise": 0.0, "grad_norm": 0.7573, "lr": 3.01e-05, "router_trainable": true, "sec_per_step": 0.592, "probe_sec": 0.0639, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04245, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": -0.07175, "n_shadow": 5, "mean_delta_shadow": -0.04091, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.20337, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 2.87833, "trunc_abs_frac": 0.75633, "mean_delta_full": -0.03971, "read_hit_rate": {"surface": 0.53519, "split": 0.25201, "merge": 0.10517, "skel": 0.13557}, "mean_gate": {"surface": 0.12291, "split": 0.05034, "merge": 0.09247, "skel": 0.08604}, "n_reads": 7968, "n_cand": 30401, "read_suppressed_frac": 0.1003, "dropout_reads_logged": 7968, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2718, "admitted_surface": 61040, "staged_surface": 49505, "admitted_split": 1147, "staged_split": 45745, "admitted_merge": 3334, "staged_merge": 119107, "admitted_skel": 15, "staged_skel": 47787, "mean_count": 438.6988525390625, "n_util_obs": 417, "util_count_spearman": -0.3149737923570258}, "probe_overhead_sec_total": 837.703, "probe_overhead_seq_frac": 0.4999, "alpha": 0.03, "t": 6254.3}
{"event": "consolidate", "step": 6000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 20, "uncertain_staged": 262082, "t": 6282.4}
{"event": "train", "step": 5999, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.3799, "bpb": 1.99077, "chunk_loss": 0.030517, "F": 0.20569, "G": 0.21233, "chunks": 210.62, "bytes_per_chunk": 4.862, "boundary_entropy": 0.09782, "router_noise": 0.0, "grad_norm": 0.7235, "lr": 3e-05, "router_trainable": true, "sec_per_step": 0.562, "probe_sec": 0.0677, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04445, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": -0.06495, "n_shadow": 6, "mean_delta_shadow": 0.03154, "n_split": 2, "mean_delta_split": -0.15981, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": -0.94334, "trunc_abs_frac": 0.94334, "mean_delta_full": 0.07971, "read_hit_rate": {"surface": 0.5401, "split": 0.27844, "merge": 0.10141, "skel": 0.13906}, "mean_gate": {"surface": 0.12371, "split": 0.05007, "merge": 0.09529, "skel": 0.09518}, "n_reads": 8252, "n_cand": 30104, "read_suppressed_frac": 0.1046, "dropout_reads_logged": 8252, "inventory": {"size": 327660, "admitted": 65536, "staged": 262124, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2732, "admitted_surface": 61042, "staged_surface": 49526, "admitted_split": 1145, "staged_split": 45492, "admitted_merge": 3334, "staged_merge": 119545, "admitted_skel": 15, "staged_skel": 47561, "mean_count": 443.0903015136719, "n_util_obs": 423, "util_count_spearman": -0.2996263998609549}, "probe_overhead_sec_total": 841.422, "probe_overhead_seq_frac": 0.4999, "alpha": 0.03, "t": 6282.4}
{"event": "probe", "step": 6000, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3522.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.202379, "cps": 27.515625, "mean_p": 0.210077, "boundary_entropy": 0.101357, "bpb": 1.87366}, "tr": {"chunks": 4192.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.231564, "cps": 32.75, "mean_p": 0.212144, "boundary_entropy": 0.094209, "bpb": 1.985808}, "fi": {"chunks": 4343.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.230876, "cps": 33.929688, "mean_p": 0.230155, "boundary_entropy": 0.100512, "bpb": 2.105846}}, "bpb": {"en": 1.87366, "tr": 1.98581, "fi": 2.10585}, "gini_cpc": 0.029266, "gini_cps": 0.045395, "bpic": 4.705316, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.144209346837108, "fi": 1.1408084109690793}, "inventory": {"size": 327660, "admitted": 65536, "staged": 262124, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2732, "admitted_surface": 61042, "staged_surface": 49526, "admitted_split": 1145, "staged_split": 45492, "admitted_merge": 3334, "staged_merge": 119545, "admitted_skel": 15, "staged_skel": 47561, "mean_count": 443.0903015136719, "n_util_obs": 423, "util_count_spearman": -0.2996263998609549}, "t": 6284.0}
{"event": "eval", "step": 6000, "phase": 3, "bpb": {"en": 1.88734, "tr": 1.99294, "fi": 2.11238}, "bytes_per_chunk": {"en": 4.954, "tr": 4.795, "fi": 4.81}, "bpb_mean": 1.99755, "frozen_inventory": true, "t": 6286.4}
{"event": "reps_dumped", "dirs": ["/local/scratch/sas245/beetle-hnet/runs/ali_ali-ridge_s2/probe/reps/th0.30", "/local/scratch/sas245/beetle-hnet/runs/ali_ali-ridge_s2/probe/reps/th0.40", "/local/scratch/sas245/beetle-hnet/runs/ali_ali-ridge_s2/probe/reps/th0.50", "/local/scratch/sas245/beetle-hnet/runs/ali_ali-ridge_s2/probe/reps/th0.60", "/local/scratch/sas245/beetle-hnet/runs/ali_ali-ridge_s2/probe/reps/th0.70"], "t": 6290.9}
{"event": "final", "run": "ali_ali-ridge_s2", "arch": "ali", "objective": "ratio", "condition": "A", "step": 6000, "params": 22785088, "seed": 2, "per_lang": {"en": {"cpc": 0.20237890018962248, "cps": 27.515625, "bpb": 1.8927548705901582, "bytes_per_chunk": 4.895084263106297, "reads_per_chunk": 1.2687829699827273}, "tr": {"cpc": 0.23156382919958018, "cps": 32.75, "bpb": 1.984679511485004, "bytes_per_chunk": 4.801787774989467, "reads_per_chunk": 0.9359533273496602}, "fi": {"cpc": 0.23087555153899314, "cps": 33.9296875, "bpb": 2.1297983540449694, "bytes_per_chunk": 4.782166116350767, "reads_per_chunk": 1.0121951219512195}}, "bpb_mean": 2.002410912040044, "gini_cpc": 0.02926607149371252, "gini_cps": 0.04539548256890891, "bpic": 4.705316413701584, "mean_boundary_entropy": 0.09782059490680695, "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": 841.4218673706055, "probe_overhead_seq_frac": 0.4999166666666667, "wall_sec": 6298.026879310608, "inventory": {"size": 327660, "admitted": 65536, "staged": 262124, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2732, "admitted_surface": 61042, "staged_surface": 49526, "admitted_split": 1145, "staged_split": 45492, "admitted_merge": 3334, "staged_merge": 119545, "admitted_skel": 15, "staged_skel": 47561, "mean_count": 443.0903015136719, "n_util_obs": 423, "util_count_spearman": -0.2996263998609549}, "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": "ridge", "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": -0.2996263998609549, "paired_vs_dropout": {"n": 1197, "pearson": 0.06311483164143808, "spearman": 0.05936734392194348, "mean_abs_discrepancy": 0.513998806476593, "mean_paired": 0.034128960222005844, "mean_dropout": 0.01173090934753418}, "predictor_calibration": {"n": 2850, "all": {"n": 2850, "pearson": 0.10693757729266393, "spearman": 0.1513702442483709, "mae": 0.07775943726301193, "reliability": [{"pred": -0.0010448035318404436, "obs": -0.00256543792784214, "n": 285}, {"pred": 0.00745965214446187, "obs": 0.020744096487760544, "n": 285}, {"pred": 0.010926668532192707, "obs": 0.018378740176558495, "n": 285}, {"pred": 0.014226917177438736, "obs": 0.012157000601291656, "n": 285}, {"pred": 0.01737281307578087, "obs": 0.02708069048821926, "n": 285}, {"pred": 0.020948253571987152, "obs": 0.026223139837384224, "n": 285}, {"pred": 0.024084294214844704, "obs": 0.024542229250073433, "n": 285}, {"pred": 0.02729223482310772, "obs": 0.02946031466126442, "n": 285}, {"pred": 0.030381400138139725, "obs": 0.04704346880316734, "n": 285}, {"pred": 0.03494471311569214, "obs": 0.05386907607316971, "n": 285}]}, "admitted": {"n": 2761, "pearson": 0.09193664443449157, "spearman": 0.13158322929427427, "mae": 0.0788039043545723, "reliability": [{"pred": 0.0024519567377865314, "obs": 0.005883881822228432, "n": 276}, {"pred": 0.00810050405561924, "obs": 0.021960513666272163, "n": 276}, {"pred": 0.01138301007449627, "obs": 0.01796468161046505, "n": 276}, {"pred": 0.01461473386734724, "obs": 0.017634738236665726, "n": 276}, {"pred": 0.017808683216571808, "obs": 0.02444634959101677, "n": 276}, {"pred": 0.02135266736149788, "obs": 0.033270709216594696, "n": 277}, {"pred": 0.024406712502241135, "obs": 0.023831233382225037, "n": 276}, {"pred": 0.027539802715182304, "obs": 0.028549859300255775, "n": 275}, {"pred": 0.030519448220729828, "obs": 0.04644057899713516, "n": 276}, {"pred": 0.03502901270985603, "obs": 0.05469576269388199, "n": 277}]}, "staged": {"n": 89, "pearson": -0.03301777600228161, "spearman": 0.12887299965951654, "mae": 0.04535769298672676, "reliability": [{"pred": -0.035169947892427444, "obs": -0.0008894602651707828, "n": 9}, {"pred": -0.018305407837033272, "obs": -0.05489850044250488, "n": 9}, {"pred": -0.012138119898736477, "obs": -0.020654231309890747, "n": 9}, {"pred": -0.008852433413267136, "obs": -0.07377444207668304, "n": 9}, {"pred": -0.005165671929717064, "obs": -0.04526631161570549, "n": 8}, {"pred": -0.0008322611101903021, "obs": -0.011243396438658237, "n": 9}, {"pred": 0.006236651446670294, "obs": -0.015276860445737839, "n": 9}, {"pred": 0.014798343181610107, "obs": -0.03459211438894272, "n": 9}, {"pred": 0.017646243795752525, "obs": -0.02035728469491005, "n": 9}, {"pred": 0.021754957735538483, "obs": -0.02192522957921028, "n": 9}]}}, "t": 6298.2}
{"event": "hub_done", "repo": "AdaptiveChunking/ali_ali-ridge_s2", "pushed": [0, 1, 2, 4, 6, 8, 16, 32, 64, 128, 256, 512, 1000, 1024, 2048, 3000, 4096, 6000], "failed": [], "ok": true, "t": 6314.2}