| {"event": "start", "run": "ali_ali-ema_s0", "arch": "ali", "objective": "ratio", "condition": "A", "params": 22785088, "langs": ["en", "tr", "fi"], "premium": {"en": 1.0, "tr": 1.12, "fi": 1.104}, "schedule": {"tau1": 0.4, "tau2": 0.7, "p1_objective": "global", "p2_objective": "global", "p3_objective": "global", "freeze_in_p3": false, "freeze_scope": "router"}, "corpus": "/local/scratch/sas245/beetle-hnet/scratch/data/tier1_corpus", "premium_kind": "flores_per_sent", "doc_filtered": {}, "weights": {"en": 0.74, "tr": 0.1536, "fi": 0.1064}, "device": "cuda", "ali_params": 268096, "ali_config": {"enabled": true, "budget_B": 65536, "staging_capacity": 262144, "policy": "utility", "candidate_types": ["surface", "split", "merge", "skel"], "gate": "vector", "retrieval": true, "memory_scope": "persistent", "read_dropout": 0.1, "write_rule": "ema_floor", "eta_min": 0.02, "consolidate_every": 50, "probe_seqs": 2, "probe_reps": 8, "horizon": 32, "predictor": "mlp", "lambda_util": 0.1, "max_span": 64, "byte_classes_path": "/local/scratch/sas245/beetle-hnet/scratch/data/byte_classes.json", "gate_hidden": 0, "type_dim": 16, "gate_bias_init": -2.0}, "t": 8.8} |
| {"event": "probe_schedule", "n": 40, "steps": [0, 1, 2, 4, 8, 16, 32, 64, 128, 200, 256, 400, 600, 800, 1000, 1200, 1400, 1600, 1800, 2000, 2200, 2400, 2600, 2800, 3000, 3200, 3400, 3600, 3800, 4000, 4200, 4400, 4600, 4800, 5000, 5200, 5400, 5600, 5800, 6000], "t": 8.8} |
| {"event": "probe", "step": 0, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 6384.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.366833, "cps": 49.875, "mean_p": 0.496725, "boundary_entropy": 0.686995, "bpb": 8.160238}, "tr": {"chunks": 5290.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.292217, "cps": 41.328125, "mean_p": 0.489567, "boundary_entropy": 0.687288, "bpb": 8.171434}, "fi": {"chunks": 8895.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.472862, "cps": 69.492188, "mean_p": 0.500688, "boundary_entropy": 0.687512, "bpb": 8.214758}}, "bpb": {"en": 8.16024, "tr": 8.17143, "fi": 8.21476}, "gini_cpc": 0.106395, "gini_cps": 0.116842, "bpic": 2.758131, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.7965927738475616, "fi": 1.2890367847879758}, "inventory": {"size": 0, "admitted": 0, "staged": 0, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 0.0, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 12.0} |
| {"event": "train", "step": 0, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 5.66332, "bpb": 8.17044, "chunk_loss": 0.038537, "F": 0.28732, "G": 0.4892, "chunks": 294.22, "bytes_per_chunk": 3.48, "boundary_entropy": 0.69137, "router_noise": 0.0, "grad_norm": 11.0317, "lr": 1e-06, "router_trainable": true, "sec_per_step": 0.031, "probe_sec": 0.0168, "probe_extra_seqs": 0, "n_rows": 0, "extra_fwd_seqs": 0, "wall_s": 0.0, "overhead_frac_est": 0.0, "chunk_mismatch": 0, "n_loo": 0, "mean_delta_loo": NaN, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "read_hit_rate": {"surface": 0.0, "split": 0.0, "merge": 0.0, "skel": 0.0}, "mean_gate": {"surface": null, "split": null, "merge": null, "skel": null}, "n_reads": 0, "n_cand": 34933, "read_suppressed_frac": 0.0, "dropout_reads_logged": 0, "inventory": {"size": 9451, "admitted": 0, "staged": 9451, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 3696, "admitted_split": 0, "staged_split": 3051, "admitted_merge": 0, "staged_merge": 203, "admitted_skel": 0, "staged_skel": 2501, "mean_count": 0.0, "n_util_obs": 0, "util_count_spearman": NaN}, "probe_overhead_sec_total": 0.017, "probe_overhead_seq_frac": 0.0, "util_loss": 0.0, "alpha": 0.03, "t": 15.0} |
| {"event": "probe", "step": 1, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 6302.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.362121, "cps": 49.234375, "mean_p": 0.496454, "boundary_entropy": 0.686986, "bpb": 8.128356}, "tr": {"chunks": 5227.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.288737, "cps": 40.835938, "mean_p": 0.489372, "boundary_entropy": 0.687277, "bpb": 8.149326}, "fi": {"chunks": 8821.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.468928, "cps": 68.914062, "mean_p": 0.500418, "boundary_entropy": 0.687508, "bpb": 8.187192}}, "bpb": {"en": 8.12836, "tr": 8.14933, "fi": 8.18719}, "gini_cpc": 0.107277, "gini_cps": 0.11774, "bpic": 2.787813, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.7973475608639046, "fi": 1.2949460045871117}, "inventory": {"size": 9451, "admitted": 0, "staged": 9451, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 3696, "admitted_split": 0, "staged_split": 3051, "admitted_merge": 0, "staged_merge": 203, "admitted_skel": 0, "staged_skel": 2501, "mean_count": 0.0, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 17.1} |
| {"event": "probe", "step": 2, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 6171.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.354594, "cps": 48.210938, "mean_p": 0.495924, "boundary_entropy": 0.686967, "bpb": 8.064216}, "tr": {"chunks": 5127.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.283213, "cps": 40.054688, "mean_p": 0.488972, "boundary_entropy": 0.687254, "bpb": 8.10525}, "fi": {"chunks": 8679.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.461379, "cps": 67.804688, "mean_p": 0.499889, "boundary_entropy": 0.687499, "bpb": 8.130951}}, "bpb": {"en": 8.06422, "tr": 8.10525, "fi": 8.13095}, "gini_cpc": 0.10806, "gini_cps": 0.118536, "bpic": 2.839866, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.7986956880537021, "fi": 1.3011470419088613}, "inventory": {"size": 16573, "admitted": 0, "staged": 16573, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2, "admitted_surface": 0, "staged_surface": 6587, "admitted_split": 0, "staged_split": 5117, "admitted_merge": 0, "staged_merge": 496, "admitted_skel": 0, "staged_skel": 4373, "mean_count": 0.0, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 21.5} |
| {"event": "probe", "step": 4, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 5667.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.325634, "cps": 44.273438, "mean_p": 0.494022, "boundary_entropy": 0.686888, "bpb": 7.849059}, "tr": {"chunks": 4748.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.262277, "cps": 37.09375, "mean_p": 0.487542, "boundary_entropy": 0.687166, "bpb": 7.95192}, "fi": {"chunks": 8073.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.429164, "cps": 63.070312, "mean_p": 0.497939, "boundary_entropy": 0.687458, "bpb": 7.941799}}, "bpb": {"en": 7.84906, "tr": 7.95192, "fi": 7.9418}, "gini_cpc": 0.10939, "gini_cps": 0.119898, "bpic": 3.068585, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.8054360544437107, "fi": 1.3179349545965569}, "inventory": {"size": 30162, "admitted": 0, "staged": 30162, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 7, "admitted_surface": 0, "staged_surface": 11946, "admitted_split": 0, "staged_split": 9217, "admitted_merge": 0, "staged_merge": 927, "admitted_skel": 0, "staged_skel": 8072, "mean_count": 0.0, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 27.2} |
| {"event": "probe", "step": 8, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 4368.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.250991, "cps": 34.125, "mean_p": 0.488767, "boundary_entropy": 0.686599, "bpb": 7.288755}, "tr": {"chunks": 3764.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.207921, "cps": 29.40625, "mean_p": 0.483657, "boundary_entropy": 0.686884, "bpb": 7.525045}, "fi": {"chunks": 6361.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.338153, "cps": 49.695312, "mean_p": 0.49219, "boundary_entropy": 0.687252, "bpb": 7.435079}}, "bpb": {"en": 7.28876, "tr": 7.52504, "fi": 7.43508}, "gini_cpc": 0.108926, "gini_cps": 0.11946, "bpic": 3.914441, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.8284008843170309, "fi": 1.347271127017685}, "inventory": {"size": 55831, "admitted": 0, "staged": 55831, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 14, "admitted_surface": 0, "staged_surface": 21436, "admitted_split": 0, "staged_split": 17828, "admitted_merge": 0, "staged_merge": 1406, "admitted_skel": 0, "staged_skel": 15161, "mean_count": 0.0, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 34.9} |
| {"event": "probe", "step": 16, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 4437.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.254956, "cps": 34.664062, "mean_p": 0.490983, "boundary_entropy": 0.686931, "bpb": 6.660738}, "tr": {"chunks": 4544.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.251008, "cps": 35.5, "mean_p": 0.488391, "boundary_entropy": 0.687352, "bpb": 6.978511}, "fi": {"chunks": 5614.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.298442, "cps": 43.859375, "mean_p": 0.491247, "boundary_entropy": 0.687412, "bpb": 6.839393}}, "bpb": {"en": 6.66074, "tr": 6.97851, "fi": 6.83939}, "gini_cpc": 0.039312, "gini_cps": 0.053763, "bpic": 3.887085, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9845152841693148, "fi": 1.1705641423894617}, "inventory": {"size": 102485, "admitted": 0, "staged": 102485, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 24, "admitted_surface": 0, "staged_surface": 37514, "admitted_split": 0, "staged_split": 34860, "admitted_merge": 0, "staged_merge": 1839, "admitted_skel": 0, "staged_skel": 28272, "mean_count": 0.0, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 46.5} |
| {"event": "probe", "step": 32, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 4007.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.230248, "cps": 31.304688, "mean_p": 0.49458, "boundary_entropy": 0.68755, "bpb": 6.30724}, "tr": {"chunks": 4078.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.225267, "cps": 31.859375, "mean_p": 0.491916, "boundary_entropy": 0.687957, "bpb": 6.636367}, "fi": {"chunks": 4310.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.229121, "cps": 33.671875, "mean_p": 0.492878, "boundary_entropy": 0.687996, "bpb": 6.431573}}, "bpb": {"en": 6.30724, "tr": 6.63637, "fi": 6.43157}, "gini_cpc": 0.00485, "gini_cps": 0.016297, "bpic": 4.577007, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9783662162998435, "fi": 0.9951078781023751}, "inventory": {"size": 191393, "admitted": 0, "staged": 191393, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 48, "admitted_surface": 0, "staged_surface": 69436, "admitted_split": 0, "staged_split": 65810, "admitted_merge": 0, "staged_merge": 3284, "admitted_skel": 0, "staged_skel": 52863, "mean_count": 0.0, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 66.7} |
| {"event": "consolidate", "step": 50, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 5, "swapped": 0, "uncertain_staged": 262137, "t": 90.8} |
| {"event": "train", "step": 50, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 4.2256, "bpb": 6.09626, "chunk_loss": 0.030389, "F": 0.17184, "G": 0.49116, "chunks": 175.97, "bytes_per_chunk": 5.819, "boundary_entropy": 0.69209, "router_noise": 0.0, "grad_norm": 2.4391, "lr": 5.1e-05, "router_trainable": true, "sec_per_step": 1.556, "probe_sec": 0.2305, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.19715, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.00425, "n_shadow": 4, "mean_delta_shadow": 0.00256, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.0033, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -0.18677, "trunc_abs_frac": 0.66223, "mean_delta_full": -0.0105, "read_hit_rate": {"surface": 0.03699, "split": 0.09252, "merge": 0.0, "skel": 0.0}, "mean_gate": {"surface": 0.12097, "split": 0.12109, "merge": null, "skel": null}, "n_reads": 988, "n_cand": 22855, "read_suppressed_frac": 0.1083, "dropout_reads_logged": 988, "inventory": {"size": 262149, "admitted": 5, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 77, "admitted_surface": 1, "staged_surface": 93830, "admitted_split": 4, "staged_split": 90086, "admitted_merge": 0, "staged_merge": 5073, "admitted_skel": 0, "staged_skel": 73155, "mean_count": 11372.0, "n_util_obs": 5, "util_count_spearman": -0.3}, "probe_overhead_sec_total": 17.267, "probe_overhead_seq_frac": 0.4902, "util_loss": 0.085767, "alpha": 0.03, "t": 92.8} |
| {"event": "probe", "step": 64, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 4078.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.234327, "cps": 31.859375, "mean_p": 0.488409, "boundary_entropy": 0.686626, "bpb": 5.623195}, "tr": {"chunks": 3919.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.216483, "cps": 30.617188, "mean_p": 0.488237, "boundary_entropy": 0.687521, "bpb": 6.00986}, "fi": {"chunks": 5365.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.285205, "cps": 41.914062, "mean_p": 0.493012, "boundary_entropy": 0.687702, "bpb": 5.817154}}, "bpb": {"en": 5.62319, "tr": 6.00986, "fi": 5.81715}, "gini_cpc": 0.062247, "gini_cps": 0.072145, "bpic": 4.245772, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9238503141131518, "fi": 1.2171237629814387}, "inventory": {"size": 262149, "admitted": 5, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 84, "admitted_surface": 1, "staged_surface": 94397, "admitted_split": 4, "staged_split": 88371, "admitted_merge": 0, "staged_merge": 6114, "admitted_skel": 0, "staged_skel": 73262, "mean_count": 13748.400390625, "n_util_obs": 5, "util_count_spearman": 0.6}, "t": 115.1} |
| {"event": "consolidate", "step": 100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 65531, "swapped": 5, "uncertain_staged": 262139, "t": 174.0} |
| {"event": "train", "step": 100, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 3.21776, "bpb": 4.64225, "chunk_loss": 0.032169, "F": 0.20639, "G": 0.41949, "chunks": 211.34, "bytes_per_chunk": 4.845, "boundary_entropy": 0.66508, "router_noise": 0.0, "grad_norm": 4.069, "lr": 0.000101, "router_trainable": true, "sec_per_step": 1.654, "probe_sec": 0.4531, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.43393, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": 0.00255, "n_shadow": 7, "mean_delta_shadow": -0.00317, "n_split": 1, "mean_delta_split": 0.01123, "n_merge": 1, "mean_delta_merge": -0.01053, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 0.26811, "trunc_abs_frac": 0.49821, "mean_delta_full": -0.02303, "read_hit_rate": {"surface": 0.63824, "split": 0.39809, "merge": 0.09479, "skel": 0.43967}, "mean_gate": {"surface": 0.12006, "split": 0.12008, "merge": 0.12028, "skel": 0.12024}, "n_reads": 11828, "n_cand": 30247, "read_suppressed_frac": 0.1016, "dropout_reads_logged": 11828, "inventory": {"size": 266045, "admitted": 65536, "staged": 200509, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 115, "admitted_surface": 62562, "staged_surface": 35063, "admitted_split": 2222, "staged_split": 80194, "admitted_merge": 214, "staged_merge": 11294, "admitted_skel": 538, "staged_skel": 73958, "mean_count": 13.821853637695312, "n_util_obs": 0, "util_count_spearman": NaN}, "probe_overhead_sec_total": 40.237, "probe_overhead_seq_frac": 0.495, "util_loss": 0.00215, "alpha": 0.03, "t": 175.5} |
| {"event": "probe", "step": 128, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3605.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.207148, "cps": 28.164062, "mean_p": 0.376841, "boundary_entropy": 0.609541, "bpb": 4.176834}, "tr": {"chunks": 3025.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.167099, "cps": 23.632812, "mean_p": 0.411489, "boundary_entropy": 0.650821, "bpb": 4.500012}, "fi": {"chunks": 3378.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.179576, "cps": 26.390625, "mean_p": 0.39088, "boundary_entropy": 0.633929, "bpb": 4.423607}}, "bpb": {"en": 4.17683, "tr": 4.50001, "fi": 4.42361}, "gini_cpc": 0.048209, "gini_cps": 0.038636, "bpic": 5.668665, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.8066658632300008, "fi": 0.8668952292867861}, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 160, "admitted_surface": 62562, "staged_surface": 65211, "admitted_split": 2222, "staged_split": 92635, "admitted_merge": 214, "staged_merge": 15754, "admitted_skel": 538, "staged_skel": 88544, "mean_count": 18.217666625976562, "n_util_obs": 2, "util_count_spearman": NaN}, "t": 214.9} |
| {"event": "consolidate", "step": 150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 2, "uncertain_staged": 262136, "t": 252.6} |
| {"event": "train", "step": 150, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.81761, "bpb": 4.06495, "chunk_loss": 0.031674, "F": 0.20605, "G": 0.36362, "chunks": 211.0, "bytes_per_chunk": 4.853, "boundary_entropy": 0.59167, "router_noise": 0.0, "grad_norm": 2.5687, "lr": 0.000151, "router_trainable": true, "sec_per_step": 1.581, "probe_sec": 0.4494, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.4268, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.03959, "n_shadow": 6, "mean_delta_shadow": 0.02929, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.00793, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -0.7057, "trunc_abs_frac": 1.10444, "mean_delta_full": -0.07397, "read_hit_rate": {"surface": 0.54167, "split": 0.32072, "merge": 0.09181, "skel": 0.35916}, "mean_gate": {"surface": 0.11978, "split": 0.11974, "merge": 0.11992, "skel": 0.11975}, "n_reads": 9312, "n_cand": 28639, "read_suppressed_frac": 0.1038, "dropout_reads_logged": 9312, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 196, "admitted_surface": 62564, "staged_surface": 75877, "admitted_split": 2220, "staged_split": 85482, "admitted_merge": 214, "staged_merge": 19386, "admitted_skel": 538, "staged_skel": 81399, "mean_count": 20.833602905273438, "n_util_obs": 1, "util_count_spearman": NaN}, "probe_overhead_sec_total": 63.379, "probe_overhead_seq_frac": 0.4967, "util_loss": 0.00062, "alpha": 0.03, "t": 254.5} |
| {"event": "consolidate", "step": 200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 7, "uncertain_staged": 262131, "t": 331.3} |
| {"event": "probe", "step": 200, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3102.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.178245, "cps": 24.234375, "mean_p": 0.365786, "boundary_entropy": 0.56925, "bpb": 3.633861}, "tr": {"chunks": 3470.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.191681, "cps": 27.109375, "mean_p": 0.370537, "boundary_entropy": 0.590986, "bpb": 3.768857}, "fi": {"chunks": 3203.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.170273, "cps": 25.023438, "mean_p": 0.367985, "boundary_entropy": 0.601892, "bpb": 3.871867}}, "bpb": {"en": 3.63386, "tr": 3.76886, "fi": 3.87187}, "gini_cpc": 0.02642, "gini_cps": 0.025098, "bpic": 5.803785, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.075378254093196, "fi": 0.9552727338535099}, "inventory": {"size": 327673, "admitted": 65536, "staged": 262137, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 271, "admitted_surface": 62570, "staged_surface": 70909, "admitted_split": 2218, "staged_split": 86575, "admitted_merge": 214, "staged_merge": 24333, "admitted_skel": 534, "staged_skel": 80320, "mean_count": 24.599319458007812, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 333.9} |
| {"event": "eval", "step": 200, "phase": 1, "bpb": {"en": 3.6407, "tr": 3.77862, "fi": 3.87195}, "bytes_per_chunk": {"en": 5.431, "tr": 5.377, "fi": 6.191}, "bpb_mean": 3.76376, "frozen_inventory": true, "t": 338.8} |
| {"event": "train", "step": 200, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.54763, "bpb": 3.67546, "chunk_loss": 0.030795, "F": 0.1835, "G": 0.37047, "chunks": 187.91, "bytes_per_chunk": 5.45, "boundary_entropy": 0.57744, "router_noise": 0.0, "grad_norm": 1.4495, "lr": 0.000201, "router_trainable": true, "sec_per_step": 1.78, "probe_sec": 0.471, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.44611, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": -0.01, "n_shadow": 7, "mean_delta_shadow": -0.03048, "n_split": 1, "mean_delta_split": -0.01685, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.601, "trunc_abs_frac": 1.02431, "mean_delta_full": 0.01823, "read_hit_rate": {"surface": 0.47317, "split": 0.2877, "merge": 0.04034, "skel": 0.23342}, "mean_gate": {"surface": 0.12452, "split": 0.12434, "merge": 0.12528, "skel": 0.12395}, "n_reads": 6927, "n_cand": 26147, "read_suppressed_frac": 0.0951, "dropout_reads_logged": 6927, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 266, "admitted_surface": 62570, "staged_surface": 71561, "admitted_split": 2218, "staged_split": 86200, "admitted_merge": 214, "staged_merge": 24406, "admitted_skel": 534, "staged_skel": 79977, "mean_count": 24.70501708984375, "n_util_obs": 0, "util_count_spearman": NaN}, "probe_overhead_sec_total": 86.502, "probe_overhead_seq_frac": 0.4975, "util_loss": 0.000605, "alpha": 0.03, "t": 343.5} |
| {"event": "consolidate", "step": 250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262129, "t": 422.5} |
| {"event": "train", "step": 250, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.38506, "bpb": 3.44091, "chunk_loss": 0.031147, "F": 0.19626, "G": 0.34564, "chunks": 200.97, "bytes_per_chunk": 5.095, "boundary_entropy": 0.53487, "router_noise": 0.0, "grad_norm": 1.3079, "lr": 0.000251, "router_trainable": true, "sec_per_step": 1.613, "probe_sec": 0.4669, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.43939, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": -0.01028, "n_shadow": 2, "mean_delta_shadow": -0.01514, "n_split": 2, "mean_delta_split": -0.00117, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.54946, "trunc_abs_frac": 0.59882, "mean_delta_full": -0.04134, "read_hit_rate": {"surface": 0.44007, "split": 0.34777, "merge": 0.01775, "skel": 0.31803}, "mean_gate": {"surface": 0.13091, "split": 0.13011, "merge": 0.13251, "skel": 0.12756}, "n_reads": 8199, "n_cand": 28322, "read_suppressed_frac": 0.1001, "dropout_reads_logged": 8199, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 271, "admitted_surface": 61881, "staged_surface": 77878, "admitted_split": 2868, "staged_split": 78651, "admitted_merge": 245, "staged_merge": 30150, "admitted_skel": 542, "staged_skel": 75465, "mean_count": 32.66120910644531, "n_util_obs": 12, "util_count_spearman": -0.06293706293706294}, "probe_overhead_sec_total": 109.49, "probe_overhead_seq_frac": 0.498, "util_loss": 0.000994, "alpha": 0.03, "t": 424.2} |
| {"event": "probe", "step": 256, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3386.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.194564, "cps": 26.453125, "mean_p": 0.324698, "boundary_entropy": 0.50868, "bpb": 3.342057}, "tr": {"chunks": 3505.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.193614, "cps": 27.382812, "mean_p": 0.345244, "boundary_entropy": 0.526392, "bpb": 3.516772}, "fi": {"chunks": 4053.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.215459, "cps": 31.664062, "mean_p": 0.357405, "boundary_entropy": 0.556423, "bpb": 3.657437}}, "bpb": {"en": 3.34206, "tr": 3.51677, "fi": 3.65744}, "gini_cpc": 0.024126, "gini_cps": 0.040631, "bpic": 5.183845, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9951181268020733, "fi": 1.1073932875923758}, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 275, "admitted_surface": 61881, "staged_surface": 78589, "admitted_split": 2868, "staged_split": 77420, "admitted_merge": 245, "staged_merge": 31192, "admitted_skel": 542, "staged_skel": 74943, "mean_count": 33.34613037109375, "n_util_obs": 12, "util_count_spearman": -0.07692307692307693}, "t": 434.7} |
| {"event": "consolidate", "step": 300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262133, "t": 508.5} |
| {"event": "train", "step": 300, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.29828, "bpb": 3.31571, "chunk_loss": 0.031281, "F": 0.21094, "G": 0.2966, "chunks": 216.0, "bytes_per_chunk": 4.741, "boundary_entropy": 0.42423, "router_noise": 0.0, "grad_norm": 2.4649, "lr": 0.0003, "router_trainable": true, "sec_per_step": 1.722, "probe_sec": 0.4939, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.46461, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.00606, "n_shadow": 6, "mean_delta_shadow": -0.02524, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.0499, "trunc_abs_frac": 0.27037, "mean_delta_full": -0.04159, "read_hit_rate": {"surface": 0.48663, "split": 0.37285, "merge": 0.00701, "skel": 0.32786}, "mean_gate": {"surface": 0.14497, "split": 0.14327, "merge": 0.14206, "skel": 0.1435}, "n_reads": 9477, "n_cand": 30646, "read_suppressed_frac": 0.0944, "dropout_reads_logged": 9477, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 298, "admitted_surface": 60890, "staged_surface": 86393, "admitted_split": 3859, "staged_split": 68660, "admitted_merge": 245, "staged_merge": 38282, "admitted_skel": 542, "staged_skel": 68809, "mean_count": 39.415283203125, "n_util_obs": 21, "util_count_spearman": -0.06623376623376623}, "probe_overhead_sec_total": 132.081, "probe_overhead_seq_frac": 0.4983, "util_loss": 0.001177, "alpha": 0.03, "t": 510.3} |
| {"event": "consolidate", "step": 350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262130, "t": 592.6} |
| {"event": "train", "step": 350, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.19935, "bpb": 3.173, "chunk_loss": 0.030837, "F": 0.19724, "G": 0.28341, "chunks": 201.97, "bytes_per_chunk": 5.07, "boundary_entropy": 0.39181, "router_noise": 0.0, "grad_norm": 1.1537, "lr": 0.0002999, "router_trainable": true, "sec_per_step": 1.686, "probe_sec": 0.4822, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.45576, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": -0.04907, "n_shadow": 5, "mean_delta_shadow": 0.031, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.05457, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 1.37973, "trunc_abs_frac": 1.33295, "mean_delta_full": -0.09998, "read_hit_rate": {"surface": 0.44876, "split": 0.34962, "merge": 0.00188, "skel": 0.32917}, "mean_gate": {"surface": 0.16932, "split": 0.16559, "merge": 0.17182, "skel": 0.17477}, "n_reads": 8351, "n_cand": 28708, "read_suppressed_frac": 0.1007, "dropout_reads_logged": 8351, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 299, "admitted_surface": 60700, "staged_surface": 82595, "admitted_split": 4125, "staged_split": 64182, "admitted_merge": 170, "staged_merge": 49931, "admitted_skel": 541, "staged_skel": 65436, "mean_count": 42.297637939453125, "n_util_obs": 26, "util_count_spearman": -0.25333333333333335}, "probe_overhead_sec_total": 156.475, "probe_overhead_seq_frac": 0.4986, "util_loss": 0.001467, "alpha": 0.03, "t": 594.6} |
| {"event": "consolidate", "step": 400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262128, "t": 672.5} |
| {"event": "probe", "step": 400, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3587.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.206114, "cps": 28.023438, "mean_p": 0.260473, "boundary_entropy": 0.322032, "bpb": 2.895085}, "tr": {"chunks": 3721.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.205546, "cps": 29.070312, "mean_p": 0.291061, "boundary_entropy": 0.419631, "bpb": 3.201843}, "fi": {"chunks": 4208.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.223699, "cps": 32.875, "mean_p": 0.301995, "boundary_entropy": 0.41701, "bpb": 3.377908}}, "bpb": {"en": 2.89508, "tr": 3.20184, "fi": 3.37791}, "gini_cpc": 0.019047, "gini_cps": 0.03595, "bpic": 4.926363, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9972449876039226, "fi": 1.0853169638671072}, "inventory": {"size": 326656, "admitted": 65536, "staged": 261120, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 333, "admitted_surface": 59996, "staged_surface": 77901, "admitted_split": 4819, "staged_split": 58367, "admitted_merge": 182, "staged_merge": 63199, "admitted_skel": 539, "staged_skel": 61653, "mean_count": 49.354522705078125, "n_util_obs": 30, "util_count_spearman": -0.0932146829810901}, "t": 675.5} |
| {"event": "eval", "step": 400, "phase": 1, "bpb": {"en": 2.91007, "tr": 3.20817, "fi": 3.37043}, "bytes_per_chunk": {"en": 4.565, "tr": 5.741, "fi": 4.836}, "bpb_mean": 3.16289, "frozen_inventory": true, "t": 680.9} |
| {"event": "train", "step": 400, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.04397, "bpb": 2.94882, "chunk_loss": 0.031173, "F": 0.216, "G": 0.2738, "chunks": 221.19, "bytes_per_chunk": 4.63, "boundary_entropy": 0.33156, "router_noise": 0.0, "grad_norm": 1.095, "lr": 0.0002998, "router_trainable": true, "sec_per_step": 1.828, "probe_sec": 0.4873, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.45346, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 11, "mean_delta_loo": -0.02629, "n_shadow": 2, "mean_delta_shadow": 0.03778, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.12854, "n_dropped_unpaired": 0, "n_full": 8, "trunc_ratio_sum": -1.37908, "trunc_abs_frac": 4.91264, "mean_delta_full": 0.01918, "read_hit_rate": {"surface": 0.44905, "split": 0.36144, "merge": 0.00185, "skel": 0.31535}, "mean_gate": {"surface": 0.20504, "split": 0.19731, "merge": 0.24396, "skel": 0.21443}, "n_reads": 9190, "n_cand": 31346, "read_suppressed_frac": 0.102, "dropout_reads_logged": 9190, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 334, "admitted_surface": 59996, "staged_surface": 78165, "admitted_split": 4819, "staged_split": 58479, "admitted_merge": 182, "staged_merge": 63641, "admitted_skel": 539, "staged_skel": 61859, "mean_count": 49.4947509765625, "n_util_obs": 30, "util_count_spearman": -0.08387096774193549}, "probe_overhead_sec_total": 180.109, "probe_overhead_seq_frac": 0.4988, "util_loss": 0.001967, "alpha": 0.03, "t": 686.0} |
| {"event": "consolidate", "step": 450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262127, "t": 766.0} |
| {"event": "train", "step": 450, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.01463, "bpb": 2.9065, "chunk_loss": 0.031032, "F": 0.2117, "G": 0.26588, "chunks": 216.78, "bytes_per_chunk": 4.724, "boundary_entropy": 0.31939, "router_noise": 0.0, "grad_norm": 1.133, "lr": 0.0002995, "router_trainable": true, "sec_per_step": 1.634, "probe_sec": 0.4784, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.44914, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.28751, "n_shadow": 2, "mean_delta_shadow": -0.02048, "n_split": 1, "mean_delta_split": -0.01471, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.20285, "trunc_abs_frac": 0.61261, "mean_delta_full": -0.18793, "read_hit_rate": {"surface": 0.42187, "split": 0.35732, "merge": 0.00116, "skel": 0.27028}, "mean_gate": {"surface": 0.22039, "split": 0.20749, "merge": 0.24432, "skel": 0.22625}, "n_reads": 8607, "n_cand": 30948, "read_suppressed_frac": 0.1026, "dropout_reads_logged": 8607, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 331, "admitted_surface": 60162, "staged_surface": 74071, "admitted_split": 5036, "staged_split": 54193, "admitted_merge": 191, "staged_merge": 75761, "admitted_skel": 147, "staged_skel": 58119, "mean_count": 53.19801330566406, "n_util_obs": 38, "util_count_spearman": -0.04234598971441077}, "probe_overhead_sec_total": 203.548, "probe_overhead_seq_frac": 0.4989, "util_loss": 0.00249, "alpha": 0.03, "t": 767.7} |
| {"event": "consolidate", "step": 500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262126, "t": 850.3} |
| {"event": "train", "step": 500, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.99598, "bpb": 2.87959, "chunk_loss": 0.031164, "F": 0.21725, "G": 0.26638, "chunks": 222.47, "bytes_per_chunk": 4.603, "boundary_entropy": 0.3065, "router_noise": 0.0, "grad_norm": 1.1235, "lr": 0.0002992, "router_trainable": true, "sec_per_step": 1.685, "probe_sec": 0.4716, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.44639, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": -0.00721, "n_shadow": 4, "mean_delta_shadow": -0.18049, "n_split": 1, "mean_delta_split": 0.09436, "n_merge": 2, "mean_delta_merge": 0.25283, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.13776, "trunc_abs_frac": 0.17869, "mean_delta_full": -0.33252, "read_hit_rate": {"surface": 0.33484, "split": 0.31765, "merge": 0.00283, "skel": 0.1805}, "mean_gate": {"surface": 0.22708, "split": 0.2075, "merge": 0.27009, "skel": 0.23179}, "n_reads": 7250, "n_cand": 31700, "read_suppressed_frac": 0.1026, "dropout_reads_logged": 7250, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 258, "admitted_surface": 60212, "staged_surface": 70431, "admitted_split": 5098, "staged_split": 50200, "admitted_merge": 161, "staged_merge": 85995, "admitted_skel": 65, "staged_skel": 55518, "mean_count": 49.28556823730469, "n_util_obs": 36, "util_count_spearman": -0.2193050193050193}, "probe_overhead_sec_total": 226.806, "probe_overhead_seq_frac": 0.499, "util_loss": 0.003168, "alpha": 0.03, "t": 852.0} |
| {"event": "consolidate", "step": 550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262126, "t": 933.8} |
| {"event": "train", "step": 550, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.97187, "bpb": 2.8448, "chunk_loss": 0.031382, "F": 0.22763, "G": 0.26752, "chunks": 233.09, "bytes_per_chunk": 4.393, "boundary_entropy": 0.28881, "router_noise": 0.0, "grad_norm": 1.1834, "lr": 0.0002987, "router_trainable": true, "sec_per_step": 1.683, "probe_sec": 0.4711, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.43066, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": 0.08762, "n_shadow": 4, "mean_delta_shadow": -0.09889, "n_split": 1, "mean_delta_split": -0.25403, "n_merge": 3, "mean_delta_merge": -0.08099, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.54723, "trunc_abs_frac": 0.54723, "mean_delta_full": -0.1499, "read_hit_rate": {"surface": 0.35331, "split": 0.29801, "merge": 0.00176, "skel": 0.18164}, "mean_gate": {"surface": 0.23925, "split": 0.20579, "merge": 0.339, "skel": 0.2264}, "n_reads": 7444, "n_cand": 33100, "read_suppressed_frac": 0.1055, "dropout_reads_logged": 7444, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 268, "admitted_surface": 60129, "staged_surface": 67087, "admitted_split": 5277, "staged_split": 47493, "admitted_merge": 73, "staged_merge": 94334, "admitted_skel": 57, "staged_skel": 53230, "mean_count": 51.27653503417969, "n_util_obs": 41, "util_count_spearman": -0.21062717770034844}, "probe_overhead_sec_total": 250.516, "probe_overhead_seq_frac": 0.4991, "util_loss": 0.00537, "alpha": 0.03, "t": 936.1} |
| {"event": "consolidate", "step": 600, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262124, "t": 1012.6} |
| {"event": "probe", "step": 600, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3657.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.210136, "cps": 28.570312, "mean_p": 0.246444, "boundary_entropy": 0.272045, "bpb": 2.59244}, "tr": {"chunks": 3751.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.207203, "cps": 29.304688, "mean_p": 0.259079, "boundary_entropy": 0.329903, "bpb": 2.933932}, "fi": {"chunks": 4473.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.237786, "cps": 34.945312, "mean_p": 0.272305, "boundary_entropy": 0.311691, "bpb": 3.15558}}, "bpb": {"en": 2.59244, "tr": 2.93393, "fi": 3.15558}, "gini_cpc": 0.031122, "gini_cps": 0.045787, "bpic": 4.775019, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9860425873149439, "fi": 1.1315823753307206}, "inventory": {"size": 326655, "admitted": 65536, "staged": 261119, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 282, "admitted_surface": 59900, "staged_surface": 64366, "admitted_split": 5501, "staged_split": 45657, "admitted_merge": 73, "staged_merge": 99529, "admitted_skel": 62, "staged_skel": 51567, "mean_count": 63.811187744140625, "n_util_obs": 57, "util_count_spearman": -0.10740212600466684}, "t": 1015.3} |
| {"event": "eval", "step": 600, "phase": 1, "bpb": {"en": 2.60489, "tr": 2.94317, "fi": 3.15553}, "bytes_per_chunk": {"en": 4.475, "tr": 5.413, "fi": 4.413}, "bpb_mean": 2.9012, "frozen_inventory": true, "t": 1020.0} |
| {"event": "train", "step": 600, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.87703, "bpb": 2.70799, "chunk_loss": 0.031066, "F": 0.21912, "G": 0.25744, "chunks": 224.38, "bytes_per_chunk": 4.564, "boundary_entropy": 0.27901, "router_noise": 0.0, "grad_norm": 1.2002, "lr": 0.0002982, "router_trainable": true, "sec_per_step": 1.787, "probe_sec": 0.3659, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.34212, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.16239, "n_shadow": 6, "mean_delta_shadow": -0.08891, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 1.30849, "trunc_abs_frac": 1.30849, "mean_delta_full": 0.43726, "read_hit_rate": {"surface": 0.32666, "split": 0.41245, "merge": 0.00084, "skel": 0.22741}, "mean_gate": {"surface": 0.25797, "split": 0.21967, "merge": 0.26078, "skel": 0.2413}, "n_reads": 8718, "n_cand": 32148, "read_suppressed_frac": 0.0977, "dropout_reads_logged": 8718, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 282, "admitted_surface": 59900, "staged_surface": 64614, "admitted_split": 5501, "staged_split": 45797, "admitted_merge": 73, "staged_merge": 100008, "admitted_skel": 62, "staged_skel": 51725, "mean_count": 63.944244384765625, "n_util_obs": 57, "util_count_spearman": -0.12218045112781954}, "probe_overhead_sec_total": 271.163, "probe_overhead_seq_frac": 0.4992, "util_loss": 0.005385, "alpha": 0.03, "t": 1025.5} |
| {"event": "consolidate", "step": 650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262127, "t": 1089.0} |
| {"event": "train", "step": 650, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.90783, "bpb": 2.75242, "chunk_loss": 0.031185, "F": 0.21948, "G": 0.26241, "chunks": 224.75, "bytes_per_chunk": 4.556, "boundary_entropy": 0.27766, "router_noise": 0.0, "grad_norm": 1.1612, "lr": 0.0002975, "router_trainable": true, "sec_per_step": 1.283, "probe_sec": 0.1646, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.13743, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 2, "mean_delta_loo": -0.09332, "n_shadow": 8, "mean_delta_shadow": -0.03486, "n_split": 1, "mean_delta_split": -0.17517, "n_merge": 1, "mean_delta_merge": 0.02179, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.78679, "trunc_abs_frac": 0.41889, "mean_delta_full": -0.24943, "read_hit_rate": {"surface": 0.35922, "split": 0.41041, "merge": 0.00014, "skel": 0.19629}, "mean_gate": {"surface": 0.23017, "split": 0.18591, "merge": 0.21057, "skel": 0.19941}, "n_reads": 8750, "n_cand": 32180, "read_suppressed_frac": 0.0938, "dropout_reads_logged": 8750, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 311, "admitted_surface": 59912, "staged_surface": 62385, "admitted_split": 5527, "staged_split": 44218, "admitted_merge": 56, "staged_merge": 105083, "admitted_skel": 41, "staged_skel": 50458, "mean_count": 68.01463317871094, "n_util_obs": 63, "util_count_spearman": -0.05976382488479263}, "probe_overhead_sec_total": 283.723, "probe_overhead_seq_frac": 0.4992, "util_loss": 0.006137, "alpha": 0.03, "t": 1089.6} |
| {"event": "consolidate", "step": 700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 529, "uncertain_staged": 262127, "t": 1149.9} |
| {"event": "train", "step": 700, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.85076, "bpb": 2.67008, "chunk_loss": 0.031089, "F": 0.22015, "G": 0.25583, "chunks": 225.44, "bytes_per_chunk": 4.542, "boundary_entropy": 0.25916, "router_noise": 0.0, "grad_norm": 1.2089, "lr": 0.0002967, "router_trainable": true, "sec_per_step": 1.217, "probe_sec": 0.0884, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06703, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": 0.04799, "n_shadow": 2, "mean_delta_shadow": 0.09094, "n_split": 3, "mean_delta_split": -0.08516, "n_merge": 1, "mean_delta_merge": 0.06714, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.71207, "trunc_abs_frac": 0.96875, "mean_delta_full": 0.03943, "read_hit_rate": {"surface": 0.47814, "split": 0.46434, "merge": 0.03427, "skel": 0.26098}, "mean_gate": {"surface": 0.22804, "split": 0.18751, "merge": 0.22901, "skel": 0.20581}, "n_reads": 10833, "n_cand": 32043, "read_suppressed_frac": 0.1018, "dropout_reads_logged": 10833, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 317, "admitted_surface": 59699, "staged_surface": 61229, "admitted_split": 5593, "staged_split": 43551, "admitted_merge": 194, "staged_merge": 107923, "admitted_skel": 50, "staged_skel": 49441, "mean_count": 84.41200256347656, "n_util_obs": 79, "util_count_spearman": -0.047200584225900684}, "probe_overhead_sec_total": 290.678, "probe_overhead_seq_frac": 0.4993, "util_loss": 0.008135, "alpha": 0.03, "t": 1150.5} |
| {"event": "consolidate", "step": 750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 627, "uncertain_staged": 262130, "t": 1207.1} |
| {"event": "train", "step": 750, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.81018, "bpb": 2.61154, "chunk_loss": 0.030855, "F": 0.21027, "G": 0.25109, "chunks": 215.31, "bytes_per_chunk": 4.756, "boundary_entropy": 0.25969, "router_noise": 0.0, "grad_norm": 1.1414, "lr": 0.0002959, "router_trainable": true, "sec_per_step": 1.165, "probe_sec": 0.1482, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09883, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.0598, "n_shadow": 5, "mean_delta_shadow": 0.03932, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.07812, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 1.05568, "trunc_abs_frac": 1.313, "mean_delta_full": 0.0981, "read_hit_rate": {"surface": 0.52129, "split": 0.50446, "merge": 0.0208, "skel": 0.23891}, "mean_gate": {"surface": 0.2129, "split": 0.17432, "merge": 0.20283, "skel": 0.1868}, "n_reads": 11096, "n_cand": 31125, "read_suppressed_frac": 0.1002, "dropout_reads_logged": 11096, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 371, "admitted_surface": 59649, "staged_surface": 60224, "admitted_split": 5625, "staged_split": 43022, "admitted_merge": 208, "staged_merge": 110000, "admitted_skel": 54, "staged_skel": 48898, "mean_count": 96.05172729492188, "n_util_obs": 87, "util_count_spearman": -0.12659473645840927}, "probe_overhead_sec_total": 297.021, "probe_overhead_seq_frac": 0.4993, "util_loss": 0.008592, "alpha": 0.03, "t": 1208.7} |
| {"event": "consolidate", "step": 800, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262130, "t": 1272.0} |
| {"event": "probe", "step": 800, "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": 0.236412, "boundary_entropy": 0.235777, "bpb": 2.432506}, "tr": {"chunks": 3622.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.200077, "cps": 28.296875, "mean_p": 0.250407, "boundary_entropy": 0.287293, "bpb": 2.769373}, "fi": {"chunks": 4698.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.249747, "cps": 36.703125, "mean_p": 0.279164, "boundary_entropy": 0.266366, "bpb": 2.953107}}, "bpb": {"en": 2.43251, "tr": 2.76937, "fi": 2.95311}, "gini_cpc": 0.053623, "gini_cps": 0.071315, "bpic": 4.82415, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.012193565501799, "fi": 1.2634754467017688}, "inventory": {"size": 326655, "admitted": 65536, "staged": 261119, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 396, "admitted_surface": 58819, "staged_surface": 59301, "admitted_split": 5626, "staged_split": 42748, "admitted_merge": 1037, "staged_merge": 110810, "admitted_skel": 54, "staged_skel": 48260, "mean_count": 104.47586059570312, "n_util_obs": 93, "util_count_spearman": -0.1707200620691713}, "t": 1274.3} |
| {"event": "eval", "step": 800, "phase": 1, "bpb": {"en": 2.44465, "tr": 2.78035, "fi": 2.95789}, "bytes_per_chunk": {"en": 4.755, "tr": 5.594, "fi": 4.253}, "bpb_mean": 2.72763, "frozen_inventory": true, "t": 1277.4} |
| {"event": "train", "step": 800, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.80054, "bpb": 2.59763, "chunk_loss": 0.030906, "F": 0.21118, "G": 0.25219, "chunks": 216.25, "bytes_per_chunk": 4.735, "boundary_entropy": 0.24505, "router_noise": 0.0, "grad_norm": 0.9883, "lr": 0.0002949, "router_trainable": true, "sec_per_step": 1.459, "probe_sec": 0.111, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08527, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": -0.01623, "n_shadow": 4, "mean_delta_shadow": -0.07588, "n_split": 1, "mean_delta_split": 0.09192, "n_merge": 1, "mean_delta_merge": -0.15952, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 1.31687, "trunc_abs_frac": 3.31701, "mean_delta_full": -0.09116, "read_hit_rate": {"surface": 0.50523, "split": 0.48608, "merge": 0.0544, "skel": 0.23312}, "mean_gate": {"surface": 0.21487, "split": 0.17024, "merge": 0.1999, "skel": 0.18586}, "n_reads": 10898, "n_cand": 30943, "read_suppressed_frac": 0.1021, "dropout_reads_logged": 10898, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 395, "admitted_surface": 58819, "staged_surface": 59503, "admitted_split": 5626, "staged_split": 42892, "admitted_merge": 1037, "staged_merge": 111337, "admitted_skel": 54, "staged_skel": 48412, "mean_count": 104.64218139648438, "n_util_obs": 93, "util_count_spearman": -0.17152576765838082}, "probe_overhead_sec_total": 303.934, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.008234, "alpha": 0.03, "t": 1281.7} |
| {"event": "consolidate", "step": 850, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262129, "t": 1342.6} |
| {"event": "train", "step": 850, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.76907, "bpb": 2.55223, "chunk_loss": 0.030891, "F": 0.21283, "G": 0.25004, "chunks": 217.94, "bytes_per_chunk": 4.699, "boundary_entropy": 0.23548, "router_noise": 0.0, "grad_norm": 0.9283, "lr": 0.0002938, "router_trainable": true, "sec_per_step": 1.244, "probe_sec": 0.1232, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05916, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": -0.03555, "n_shadow": 3, "mean_delta_shadow": -0.0048, "n_split": 2, "mean_delta_split": -0.07034, "n_merge": 1, "mean_delta_merge": -0.29114, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.72007, "trunc_abs_frac": 0.54563, "mean_delta_full": -0.17537, "read_hit_rate": {"surface": 0.50677, "split": 0.43542, "merge": 0.06194, "skel": 0.23039}, "mean_gate": {"surface": 0.22692, "split": 0.16632, "merge": 0.20272, "skel": 0.17522}, "n_reads": 10298, "n_cand": 30854, "read_suppressed_frac": 0.0999, "dropout_reads_logged": 10298, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 435, "admitted_surface": 59270, "staged_surface": 58550, "admitted_split": 4923, "staged_split": 43163, "admitted_merge": 1292, "staged_merge": 112069, "admitted_skel": 51, "staged_skel": 48362, "mean_count": 103.88438415527344, "n_util_obs": 82, "util_count_spearman": -0.11359258170894962}, "probe_overhead_sec_total": 310.639, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.008632, "alpha": 0.03, "t": 1343.9} |
| {"event": "consolidate", "step": 900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262129, "t": 1399.0} |
| {"event": "train", "step": 900, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.62063, "bpb": 2.33807, "chunk_loss": 0.030767, "F": 0.20782, "G": 0.24535, "chunks": 212.81, "bytes_per_chunk": 4.812, "boundary_entropy": 0.22475, "router_noise": 0.0, "grad_norm": 0.9824, "lr": 0.0002927, "router_trainable": true, "sec_per_step": 1.115, "probe_sec": 0.1554, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.13327, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": 0.05809, "n_shadow": 3, "mean_delta_shadow": -0.57279, "n_split": 1, "mean_delta_split": 0.22107, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 7, "trunc_ratio_sum": -0.43729, "trunc_abs_frac": 2.79404, "mean_delta_full": -0.04742, "read_hit_rate": {"surface": 0.4792, "split": 0.36288, "merge": 0.06508, "skel": 0.19696}, "mean_gate": {"surface": 0.23604, "split": 0.16697, "merge": 0.21055, "skel": 0.19232}, "n_reads": 9023, "n_cand": 30608, "read_suppressed_frac": 0.1024, "dropout_reads_logged": 9023, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 438, "admitted_surface": 59897, "staged_surface": 57236, "admitted_split": 4305, "staged_split": 43139, "admitted_merge": 1294, "staged_merge": 113813, "admitted_skel": 40, "staged_skel": 47956, "mean_count": 95.14935302734375, "n_util_obs": 76, "util_count_spearman": -0.07838687628161313}, "probe_overhead_sec_total": 317.573, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.009144, "alpha": 0.03, "t": 1399.6} |
| {"event": "consolidate", "step": 950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262129, "t": 1469.9} |
| {"event": "train", "step": 950, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.77573, "bpb": 2.56184, "chunk_loss": 0.030839, "F": 0.21259, "G": 0.2431, "chunks": 217.69, "bytes_per_chunk": 4.704, "boundary_entropy": 0.21597, "router_noise": 0.0, "grad_norm": 0.9955, "lr": 0.0002914, "router_trainable": true, "sec_per_step": 1.432, "probe_sec": 0.1101, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07163, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.16594, "n_shadow": 4, "mean_delta_shadow": -0.24385, "n_split": 1, "mean_delta_split": 0.07803, "n_merge": 1, "mean_delta_merge": 0.76965, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.39422, "trunc_abs_frac": 8.33662, "mean_delta_full": 0.10982, "read_hit_rate": {"surface": 0.47361, "split": 0.27947, "merge": 0.0581, "skel": 0.12767}, "mean_gate": {"surface": 0.24433, "split": 0.15807, "merge": 0.21454, "skel": 0.19875}, "n_reads": 7680, "n_cand": 30926, "read_suppressed_frac": 0.1033, "dropout_reads_logged": 7680, "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": 60543, "staged_surface": 56766, "admitted_split": 3675, "staged_split": 43755, "admitted_merge": 1287, "staged_merge": 113859, "admitted_skel": 31, "staged_skel": 47764, "mean_count": 88.54171752929688, "n_util_obs": 71, "util_count_spearman": -0.1460093896713615}, "probe_overhead_sec_total": 324.451, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.008194, "alpha": 0.03, "t": 1471.2} |
| {"event": "consolidate", "step": 1000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262126, "t": 1528.5} |
| {"event": "probe", "step": 1000, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3383.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.194392, "cps": 26.429688, "mean_p": 0.225973, "boundary_entropy": 0.209594, "bpb": 2.350988}, "tr": {"chunks": 3575.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.197481, "cps": 27.929688, "mean_p": 0.241869, "boundary_entropy": 0.266874, "bpb": 2.655519}, "fi": {"chunks": 4732.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.251555, "cps": 36.96875, "mean_p": 0.277003, "boundary_entropy": 0.238423, "bpb": 2.829298}}, "bpb": {"en": 2.35099, "tr": 2.65552, "fi": 2.8293}, "gini_cpc": 0.059228, "gini_cps": 0.076932, "bpic": 4.853037, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0158921796220135, "fi": 1.294061673872023}, "inventory": {"size": 326656, "admitted": 65536, "staged": 261120, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 384, "admitted_surface": 60307, "staged_surface": 56661, "admitted_split": 3980, "staged_split": 42540, "admitted_merge": 1236, "staged_merge": 114878, "admitted_skel": 13, "staged_skel": 47041, "mean_count": 51.79864501953125, "n_util_obs": 46, "util_count_spearman": -0.2630280604378662}, "t": 1530.7} |
| {"event": "eval", "step": 1000, "phase": 1, "bpb": {"en": 2.36088, "tr": 2.66517, "fi": 2.83632}, "bytes_per_chunk": {"en": 4.767, "tr": 5.583, "fi": 4.249}, "bpb_mean": 2.62079, "frozen_inventory": true, "t": 1534.2} |
| {"event": "train", "step": 1000, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.72522, "bpb": 2.48897, "chunk_loss": 0.031237, "F": 0.22357, "G": 0.25144, "chunks": 228.94, "bytes_per_chunk": 4.473, "boundary_entropy": 0.21476, "router_noise": 0.0, "grad_norm": 1.0731, "lr": 0.0002901, "router_trainable": true, "sec_per_step": 1.414, "probe_sec": 0.097, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05445, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.1265, "n_shadow": 6, "mean_delta_shadow": 0.07832, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.29619, "trunc_abs_frac": 0.51525, "mean_delta_full": 0.36748, "read_hit_rate": {"surface": 0.36126, "split": 0.07943, "merge": 0.06871, "skel": 0.07256}, "mean_gate": {"surface": 0.25784, "split": 0.14947, "merge": 0.23866, "skel": 0.19296}, "n_reads": 4493, "n_cand": 32124, "read_suppressed_frac": 0.1033, "dropout_reads_logged": 4493, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 384, "admitted_surface": 60307, "staged_surface": 57025, "admitted_split": 3980, "staged_split": 42888, "admitted_merge": 1236, "staged_merge": 115092, "admitted_skel": 13, "staged_skel": 47139, "mean_count": 51.86720275878906, "n_util_obs": 46, "util_count_spearman": -0.261054579093432}, "probe_overhead_sec_total": 331.522, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.010215, "alpha": 0.03, "t": 1542.0} |
| {"event": "consolidate", "step": 1050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262124, "t": 1613.1} |
| {"event": "train", "step": 1050, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.74454, "bpb": 2.51684, "chunk_loss": 0.030898, "F": 0.21686, "G": 0.24451, "chunks": 222.06, "bytes_per_chunk": 4.611, "boundary_entropy": 0.21106, "router_noise": 0.0, "grad_norm": 0.9166, "lr": 0.0002886, "router_trainable": true, "sec_per_step": 1.452, "probe_sec": 0.1358, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05551, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": 0.10784, "n_shadow": 4, "mean_delta_shadow": -0.01318, "n_split": 2, "mean_delta_split": 0.02334, "n_merge": 1, "mean_delta_merge": -0.34412, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.21655, "trunc_abs_frac": 37.13656, "mean_delta_full": -0.21506, "read_hit_rate": {"surface": 0.35835, "split": 0.0822, "merge": 0.04289, "skel": 0.08338}, "mean_gate": {"surface": 0.28038, "split": 0.16295, "merge": 0.24535, "skel": 0.23152}, "n_reads": 4282, "n_cand": 31619, "read_suppressed_frac": 0.1063, "dropout_reads_logged": 4282, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 407, "admitted_surface": 59283, "staged_surface": 57294, "admitted_split": 5004, "staged_split": 42625, "admitted_merge": 1236, "staged_merge": 115161, "admitted_skel": 13, "staged_skel": 47064, "mean_count": 54.16719055175781, "n_util_obs": 50, "util_count_spearman": -0.3428653750177947}, "probe_overhead_sec_total": 338.774, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.012577, "alpha": 0.03, "t": 1614.6} |
| {"event": "consolidate", "step": 1100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 321, "uncertain_staged": 262121, "t": 1675.5} |
| {"event": "train", "step": 1100, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.72723, "bpb": 2.49186, "chunk_loss": 0.030886, "F": 0.21436, "G": 0.24353, "chunks": 219.5, "bytes_per_chunk": 4.665, "boundary_entropy": 0.20771, "router_noise": 0.0, "grad_norm": 0.8963, "lr": 0.0002871, "router_trainable": true, "sec_per_step": 1.25, "probe_sec": 0.1564, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10154, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.13617, "n_shadow": 6, "mean_delta_shadow": 0.01282, "n_split": 1, "mean_delta_split": 0.05914, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 4.03364, "trunc_abs_frac": 0.85975, "mean_delta_full": 0.02835, "read_hit_rate": {"surface": 0.45195, "split": 0.31795, "merge": 0.04885, "skel": 0.23241}, "mean_gate": {"surface": 0.28904, "split": 0.18311, "merge": 0.22442, "skel": 0.24577}, "n_reads": 8604, "n_cand": 31439, "read_suppressed_frac": 0.1002, "dropout_reads_logged": 8604, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 441, "admitted_surface": 58989, "staged_surface": 57369, "admitted_split": 5276, "staged_split": 42879, "admitted_merge": 1237, "staged_merge": 115242, "admitted_skel": 34, "staged_skel": 46654, "mean_count": 67.80436706542969, "n_util_obs": 63, "util_count_spearman": -0.13263248847926268}, "probe_overhead_sec_total": 345.458, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.015591, "alpha": 0.03, "t": 1677.1} |
| {"event": "consolidate", "step": 1150, "retired": 1, "split_retired": 1, "merge_admitted": 0, "admitted": 2, "swapped": 109, "uncertain_staged": 262122, "t": 1737.8} |
| {"event": "train", "step": 1150, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.58654, "bpb": 2.2889, "chunk_loss": 0.030737, "F": 0.21143, "G": 0.23594, "chunks": 216.5, "bytes_per_chunk": 4.73, "boundary_entropy": 0.19545, "router_noise": 0.0, "grad_norm": 0.844, "lr": 0.0002855, "router_trainable": true, "sec_per_step": 1.251, "probe_sec": 0.1236, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06983, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": 0.09911, "n_shadow": 5, "mean_delta_shadow": 0.07249, "n_split": 1, "mean_delta_split": 0.00696, "n_merge": 2, "mean_delta_merge": -0.2467, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.29197, "trunc_abs_frac": 5.08318, "mean_delta_full": -0.08307, "read_hit_rate": {"surface": 0.44664, "split": 0.35692, "merge": 0.04079, "skel": 0.24389}, "mean_gate": {"surface": 0.23936, "split": 0.14662, "merge": 0.20353, "skel": 0.17887}, "n_reads": 8926, "n_cand": 31027, "read_suppressed_frac": 0.1027, "dropout_reads_logged": 8926, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 470, "admitted_surface": 58882, "staged_surface": 57247, "admitted_split": 5372, "staged_split": 43275, "admitted_merge": 1237, "staged_merge": 114820, "admitted_skel": 45, "staged_skel": 46802, "mean_count": 75.60813903808594, "n_util_obs": 71, "util_count_spearman": -0.16894701542588866}, "probe_overhead_sec_total": 352.243, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.014363, "alpha": 0.03, "t": 1739.6} |
| {"event": "consolidate", "step": 1200, "retired": 1, "split_retired": 0, "merge_admitted": 0, "admitted": 1, "swapped": 95, "uncertain_staged": 262120, "t": 1806.8} |
| {"event": "probe", "step": 1200, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3432.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.197207, "cps": 26.8125, "mean_p": 0.22054, "boundary_entropy": 0.184571, "bpb": 2.271527}, "tr": {"chunks": 3841.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.212175, "cps": 30.007812, "mean_p": 0.237568, "boundary_entropy": 0.240118, "bpb": 2.548958}, "fi": {"chunks": 4755.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.252778, "cps": 37.148438, "mean_p": 0.271804, "boundary_entropy": 0.217513, "bpb": 2.723232}}, "bpb": {"en": 2.27153, "tr": 2.54896, "fi": 2.72323}, "gini_cpc": 0.055948, "gini_cps": 0.073329, "bpic": 4.716661, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0758967528080383, "fi": 1.281785867367442}, "inventory": {"size": 327584, "admitted": 65536, "staged": 262048, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 487, "admitted_surface": 58801, "staged_surface": 56877, "admitted_split": 5429, "staged_split": 43095, "admitted_merge": 1238, "staged_merge": 115553, "admitted_skel": 68, "staged_skel": 46523, "mean_count": 84.02143859863281, "n_util_obs": 82, "util_count_spearman": -0.20494988082410945}, "t": 1808.6} |
| {"event": "eval", "step": 1200, "phase": 1, "bpb": {"en": 2.28268, "tr": 2.55818, "fi": 2.73371}, "bytes_per_chunk": {"en": 4.724, "tr": 5.328, "fi": 4.222}, "bpb_mean": 2.52486, "frozen_inventory": true, "t": 1812.3} |
| {"event": "train", "step": 1200, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.6259, "bpb": 2.34567, "chunk_loss": 0.030804, "F": 0.21439, "G": 0.23797, "chunks": 219.53, "bytes_per_chunk": 4.664, "boundary_entropy": 0.19156, "router_noise": 0.0, "grad_norm": 0.8174, "lr": 0.0002837, "router_trainable": true, "sec_per_step": 1.534, "probe_sec": 0.0957, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07006, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": -0.05974, "n_shadow": 5, "mean_delta_shadow": -0.04051, "n_split": 1, "mean_delta_split": 0.02051, "n_merge": 2, "mean_delta_merge": -0.0294, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -0.33035, "trunc_abs_frac": 0.80158, "mean_delta_full": -0.02112, "read_hit_rate": {"surface": 0.4709, "split": 0.37715, "merge": 0.05172, "skel": 0.28665}, "mean_gate": {"surface": 0.25276, "split": 0.14231, "merge": 0.1954, "skel": 0.18204}, "n_reads": 9664, "n_cand": 31141, "read_suppressed_frac": 0.1011, "dropout_reads_logged": 9664, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 488, "admitted_surface": 58801, "staged_surface": 56794, "admitted_split": 5429, "staged_split": 43113, "admitted_merge": 1238, "staged_merge": 115674, "admitted_skel": 68, "staged_skel": 46563, "mean_count": 84.16893005371094, "n_util_obs": 82, "util_count_spearman": -0.2047975098224878}, "probe_overhead_sec_total": 359.228, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.013704, "alpha": 0.03, "t": 1816.3} |
| {"event": "consolidate", "step": 1250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 3, "uncertain_staged": 262121, "t": 1873.6} |
| {"event": "train", "step": 1250, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.67723, "bpb": 2.41974, "chunk_loss": 0.030898, "F": 0.21805, "G": 0.24011, "chunks": 223.28, "bytes_per_chunk": 4.586, "boundary_entropy": 0.18678, "router_noise": 0.0, "grad_norm": 0.8265, "lr": 0.0002819, "router_trainable": true, "sec_per_step": 1.169, "probe_sec": 0.1821, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.12718, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.45102, "n_shadow": 4, "mean_delta_shadow": 0.1134, "n_split": 2, "mean_delta_split": -0.16589, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.8955, "trunc_abs_frac": 0.28811, "mean_delta_full": 0.03349, "read_hit_rate": {"surface": 0.46352, "split": 0.40905, "merge": 0.0418, "skel": 0.28635}, "mean_gate": {"surface": 0.24173, "split": 0.13608, "merge": 0.20673, "skel": 0.1655}, "n_reads": 10174, "n_cand": 31904, "read_suppressed_frac": 0.0987, "dropout_reads_logged": 10174, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 512, "admitted_surface": 58798, "staged_surface": 56913, "admitted_split": 5432, "staged_split": 42870, "admitted_merge": 1238, "staged_merge": 116146, "admitted_skel": 68, "staged_skel": 46215, "mean_count": 92.76985168457031, "n_util_obs": 92, "util_count_spearman": -0.18161922229243407}, "probe_overhead_sec_total": 366.253, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.013695, "alpha": 0.03, "t": 1874.8} |
| {"event": "consolidate", "step": 1300, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 16, "uncertain_staged": 262120, "t": 1937.9} |
| {"event": "train", "step": 1300, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.62285, "bpb": 2.34128, "chunk_loss": 0.030775, "F": 0.21329, "G": 0.2347, "chunks": 218.41, "bytes_per_chunk": 4.689, "boundary_entropy": 0.18214, "router_noise": 0.0, "grad_norm": 0.9142, "lr": 0.00028, "router_trainable": true, "sec_per_step": 1.297, "probe_sec": 0.1718, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09765, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.14349, "n_shadow": 3, "mean_delta_shadow": 0.0577, "n_split": 2, "mean_delta_split": 0.01519, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 11.79869, "trunc_abs_frac": 1.54609, "mean_delta_full": 0.01865, "read_hit_rate": {"surface": 0.44761, "split": 0.41436, "merge": 0.03523, "skel": 0.29539}, "mean_gate": {"surface": 0.24601, "split": 0.13417, "merge": 0.1891, "skel": 0.16497}, "n_reads": 9989, "n_cand": 31416, "read_suppressed_frac": 0.1008, "dropout_reads_logged": 9989, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 512, "admitted_surface": 58785, "staged_surface": 56735, "admitted_split": 5442, "staged_split": 43192, "admitted_merge": 1240, "staged_merge": 116105, "admitted_skel": 69, "staged_skel": 46112, "mean_count": 101.49745178222656, "n_util_obs": 97, "util_count_spearman": -0.21604189473178073}, "probe_overhead_sec_total": 372.862, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.013316, "alpha": 0.03, "t": 1939.6} |
| {"event": "consolidate", "step": 1350, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 33, "uncertain_staged": 262120, "t": 2000.5} |
| {"event": "train", "step": 1350, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.60861, "bpb": 2.32073, "chunk_loss": 0.030991, "F": 0.22046, "G": 0.24061, "chunks": 225.75, "bytes_per_chunk": 4.536, "boundary_entropy": 0.18098, "router_noise": 0.0, "grad_norm": 0.9646, "lr": 0.000278, "router_trainable": true, "sec_per_step": 1.245, "probe_sec": 0.1274, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06669, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": 0.14001, "n_shadow": 4, "mean_delta_shadow": 0.14664, "n_split": 1, "mean_delta_split": -0.09277, "n_merge": 3, "mean_delta_merge": -0.33492, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.93744, "trunc_abs_frac": 1.33357, "mean_delta_full": -0.16, "read_hit_rate": {"surface": 0.46885, "split": 0.40933, "merge": 0.05503, "skel": 0.30712}, "mean_gate": {"surface": 0.24893, "split": 0.12198, "merge": 0.20848, "skel": 0.15816}, "n_reads": 10524, "n_cand": 32205, "read_suppressed_frac": 0.1022, "dropout_reads_logged": 10524, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 529, "admitted_surface": 58758, "staged_surface": 56597, "admitted_split": 5448, "staged_split": 43483, "admitted_merge": 1261, "staged_merge": 116078, "admitted_skel": 69, "staged_skel": 45986, "mean_count": 108.99539184570312, "n_util_obs": 103, "util_count_spearman": -0.1872336686728463}, "probe_overhead_sec_total": 378.798, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.01351, "alpha": 0.03, "t": 2001.9} |
| {"event": "consolidate", "step": 1400, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 55, "uncertain_staged": 262119, "t": 2066.7} |
| {"event": "probe", "step": 1400, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3473.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.199563, "cps": 27.132812, "mean_p": 0.217574, "boundary_entropy": 0.162836, "bpb": 2.219015}, "tr": {"chunks": 3958.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.218638, "cps": 30.921875, "mean_p": 0.231976, "boundary_entropy": 0.211606, "bpb": 2.483847}, "fi": {"chunks": 4834.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.256977, "cps": 37.765625, "mean_p": 0.267638, "boundary_entropy": 0.182611, "bpb": 2.636756}}, "bpb": {"en": 2.21902, "tr": 2.48385, "fi": 2.63676}, "gini_cpc": 0.05669, "gini_cps": 0.073977, "bpic": 4.62552, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0955812103690055, "fi": 1.2876982323276582}, "inventory": {"size": 327624, "admitted": 65536, "staged": 262088, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 559, "admitted_surface": 58707, "staged_surface": 56460, "admitted_split": 5465, "staged_split": 43624, "admitted_merge": 1289, "staged_merge": 115962, "admitted_skel": 75, "staged_skel": 46042, "mean_count": 116.73808288574219, "n_util_obs": 112, "util_count_spearman": -0.106826801517067}, "t": 2069.2} |
| {"event": "eval", "step": 1400, "phase": 1, "bpb": {"en": 2.23029, "tr": 2.49451, "fi": 2.6467}, "bytes_per_chunk": {"en": 4.677, "tr": 5.101, "fi": 4.139}, "bpb_mean": 2.45717, "frozen_inventory": true, "t": 2073.4} |
| {"event": "train", "step": 1400, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.688, "bpb": 2.43527, "chunk_loss": 0.030898, "F": 0.21915, "G": 0.23663, "chunks": 224.41, "bytes_per_chunk": 4.563, "boundary_entropy": 0.1712, "router_noise": 0.0, "grad_norm": 0.8456, "lr": 0.0002759, "router_trainable": true, "sec_per_step": 1.524, "probe_sec": 0.1017, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05633, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.01708, "n_shadow": 6, "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": 3, "trunc_ratio_sum": 0.95071, "trunc_abs_frac": 2.5851, "mean_delta_full": -0.14075, "read_hit_rate": {"surface": 0.46944, "split": 0.42401, "merge": 0.05255, "skel": 0.30552}, "mean_gate": {"surface": 0.26167, "split": 0.12873, "merge": 0.2083, "skel": 0.17243}, "n_reads": 10501, "n_cand": 31682, "read_suppressed_frac": 0.0938, "dropout_reads_logged": 10501, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 556, "admitted_surface": 58707, "staged_surface": 56427, "admitted_split": 5465, "staged_split": 43613, "admitted_merge": 1289, "staged_merge": 116115, "admitted_skel": 75, "staged_skel": 45989, "mean_count": 116.89834594726562, "n_util_obs": 112, "util_count_spearman": -0.10644240953975467}, "probe_overhead_sec_total": 385.193, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.013592, "alpha": 0.03, "t": 2078.1} |
| {"event": "consolidate", "step": 1450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 40, "uncertain_staged": 262119, "t": 2163.8} |
| {"event": "train", "step": 1450, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.64197, "bpb": 2.36887, "chunk_loss": 0.030888, "F": 0.21976, "G": 0.23674, "chunks": 225.03, "bytes_per_chunk": 4.55, "boundary_entropy": 0.16929, "router_noise": 0.0, "grad_norm": 0.79, "lr": 0.0002738, "router_trainable": true, "sec_per_step": 1.772, "probe_sec": 0.0917, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05157, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": 0.11587, "n_shadow": 7, "mean_delta_shadow": 0.12753, "n_split": 1, "mean_delta_split": -0.06046, "n_merge": 1, "mean_delta_merge": -0.22333, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.90948, "trunc_abs_frac": 0.87197, "mean_delta_full": 0.02974, "read_hit_rate": {"surface": 0.47998, "split": 0.40568, "merge": 0.06109, "skel": 0.32488}, "mean_gate": {"surface": 0.24736, "split": 0.11582, "merge": 0.19822, "skel": 0.15586}, "n_reads": 10566, "n_cand": 31839, "read_suppressed_frac": 0.0963, "dropout_reads_logged": 10566, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 579, "admitted_surface": 58671, "staged_surface": 56405, "admitted_split": 5474, "staged_split": 43798, "admitted_merge": 1315, "staged_merge": 116335, "admitted_skel": 76, "staged_skel": 45606, "mean_count": 124.87446594238281, "n_util_obs": 119, "util_count_spearman": -0.0930209371884347}, "probe_overhead_sec_total": 392.71, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.014207, "alpha": 0.03, "t": 2166.7} |
| {"event": "consolidate", "step": 1500, "retired": 1, "split_retired": 1, "merge_admitted": 0, "admitted": 2, "swapped": 1024, "uncertain_staged": 262119, "t": 2236.1} |
| {"event": "train", "step": 1500, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.60325, "bpb": 2.313, "chunk_loss": 0.030657, "F": 0.20932, "G": 0.22905, "chunks": 214.34, "bytes_per_chunk": 4.777, "boundary_entropy": 0.16936, "router_noise": 0.0, "grad_norm": 0.8614, "lr": 0.0002715, "router_trainable": true, "sec_per_step": 1.414, "probe_sec": 0.0968, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05056, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.2073, "n_shadow": 6, "mean_delta_shadow": -0.04816, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.87113, "trunc_abs_frac": 1.07428, "mean_delta_full": -0.2576, "read_hit_rate": {"surface": 0.44427, "split": 0.42045, "merge": 0.06358, "skel": 0.29166}, "mean_gate": {"surface": 0.24514, "split": 0.12028, "merge": 0.18572, "skel": 0.16169}, "n_reads": 10049, "n_cand": 30901, "read_suppressed_frac": 0.1017, "dropout_reads_logged": 10049, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 582, "admitted_surface": 57693, "staged_surface": 56310, "admitted_split": 5491, "staged_split": 43721, "admitted_merge": 2269, "staged_merge": 116425, "admitted_skel": 83, "staged_skel": 45688, "mean_count": 132.41439819335938, "n_util_obs": 126, "util_count_spearman": -0.09356130483689538}, "probe_overhead_sec_total": 398.993, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.012394, "alpha": 0.03, "t": 2237.4} |
| {"event": "consolidate", "step": 1550, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 188, "uncertain_staged": 262119, "t": 2307.8} |
| {"event": "train", "step": 1550, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.56862, "bpb": 2.26304, "chunk_loss": 0.030674, "F": 0.21091, "G": 0.22512, "chunks": 215.97, "bytes_per_chunk": 4.741, "boundary_entropy": 0.15789, "router_noise": 0.0, "grad_norm": 0.706, "lr": 0.0002692, "router_trainable": true, "sec_per_step": 1.454, "probe_sec": 0.1642, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.12129, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": -0.04625, "n_shadow": 8, "mean_delta_shadow": 0.12904, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.2688, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 1.63336, "trunc_abs_frac": 0.98912, "mean_delta_full": 0.08362, "read_hit_rate": {"surface": 0.47318, "split": 0.44778, "merge": 0.07669, "skel": 0.32129}, "mean_gate": {"surface": 0.23125, "split": 0.11038, "merge": 0.18284, "skel": 0.15979}, "n_reads": 10836, "n_cand": 30960, "read_suppressed_frac": 0.1012, "dropout_reads_logged": 10836, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 593, "admitted_surface": 57531, "staged_surface": 56355, "admitted_split": 5497, "staged_split": 44123, "admitted_merge": 2422, "staged_merge": 115656, "admitted_skel": 86, "staged_skel": 46010, "mean_count": 146.89602661132812, "n_util_obs": 138, "util_count_spearman": -0.06719575149147582}, "probe_overhead_sec_total": 405.157, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.011039, "alpha": 0.03, "t": 2310.1} |
| {"event": "consolidate", "step": 1600, "retired": 1, "split_retired": 2, "merge_admitted": 0, "admitted": 3, "swapped": 39, "uncertain_staged": 262120, "t": 2376.9} |
| {"event": "probe", "step": 1600, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3425.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.196805, "cps": 26.757812, "mean_p": 0.215622, "boundary_entropy": 0.15441, "bpb": 2.1784}, "tr": {"chunks": 3989.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.22035, "cps": 31.164062, "mean_p": 0.223204, "boundary_entropy": 0.182977, "bpb": 2.416695}, "fi": {"chunks": 4901.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.260539, "cps": 38.289062, "mean_p": 0.268507, "boundary_entropy": 0.172369, "bpb": 2.574997}}, "bpb": {"en": 2.1784, "tr": 2.41669, "fi": 2.575}, "gini_cpc": 0.062697, "gini_cps": 0.079903, "bpic": 4.60674, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.1196364517555866, "fi": 1.3238426343958556}, "inventory": {"size": 327638, "admitted": 65536, "staged": 262102, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 621, "admitted_surface": 57496, "staged_surface": 56424, "admitted_split": 5502, "staged_split": 44099, "admitted_merge": 2451, "staged_merge": 115705, "admitted_skel": 87, "staged_skel": 45874, "mean_count": 154.3081512451172, "n_util_obs": 142, "util_count_spearman": -0.10079611727798722}, "t": 2379.9} |
| {"event": "eval", "step": 1600, "phase": 1, "bpb": {"en": 2.19018, "tr": 2.4291, "fi": 2.58408}, "bytes_per_chunk": {"en": 4.735, "tr": 5.063, "fi": 4.117}, "bpb_mean": 2.40112, "frozen_inventory": true, "t": 2384.6} |
| {"event": "train", "step": 1600, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.57267, "bpb": 2.26888, "chunk_loss": 0.030772, "F": 0.21396, "G": 0.23086, "chunks": 219.09, "bytes_per_chunk": 4.674, "boundary_entropy": 0.15485, "router_noise": 0.0, "grad_norm": 0.8339, "lr": 0.0002668, "router_trainable": true, "sec_per_step": 1.645, "probe_sec": 0.1106, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06058, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": -0.1031, "n_shadow": 1, "mean_delta_shadow": -0.31219, "n_split": 1, "mean_delta_split": -0.16333, "n_merge": 1, "mean_delta_merge": 0.08435, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -0.71271, "trunc_abs_frac": 0.71502, "mean_delta_full": 0.09656, "read_hit_rate": {"surface": 0.45952, "split": 0.44518, "merge": 0.07917, "skel": 0.31943}, "mean_gate": {"surface": 0.24439, "split": 0.11443, "merge": 0.17109, "skel": 0.16485}, "n_reads": 10938, "n_cand": 31567, "read_suppressed_frac": 0.1012, "dropout_reads_logged": 10938, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 620, "admitted_surface": 57496, "staged_surface": 56471, "admitted_split": 5502, "staged_split": 44047, "admitted_merge": 2451, "staged_merge": 115697, "admitted_skel": 87, "staged_skel": 45929, "mean_count": 154.47506713867188, "n_util_obs": 143, "util_count_spearman": -0.09236186348862405}, "probe_overhead_sec_total": 411.613, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.010725, "alpha": 0.03, "t": 2392.4} |
| {"event": "consolidate", "step": 1650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 72, "uncertain_staged": 262120, "t": 2507.3} |
| {"event": "train", "step": 1650, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.59029, "bpb": 2.29431, "chunk_loss": 0.030655, "F": 0.21252, "G": 0.22502, "chunks": 217.62, "bytes_per_chunk": 4.705, "boundary_entropy": 0.15533, "router_noise": 0.0, "grad_norm": 0.8369, "lr": 0.0002643, "router_trainable": true, "sec_per_step": 2.334, "probe_sec": 0.1663, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.11155, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.18676, "n_shadow": 5, "mean_delta_shadow": -0.11878, "n_split": 1, "mean_delta_split": 0.03638, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.72705, "trunc_abs_frac": 1.00069, "mean_delta_full": 0.60529, "read_hit_rate": {"surface": 0.46624, "split": 0.44621, "merge": 0.07797, "skel": 0.30521}, "mean_gate": {"surface": 0.24855, "split": 0.10933, "merge": 0.17211, "skel": 0.16577}, "n_reads": 10810, "n_cand": 31272, "read_suppressed_frac": 0.0984, "dropout_reads_logged": 10810, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 637, "admitted_surface": 57434, "staged_surface": 56266, "admitted_split": 5508, "staged_split": 44228, "admitted_merge": 2503, "staged_merge": 115855, "admitted_skel": 91, "staged_skel": 45795, "mean_count": 162.9438934326172, "n_util_obs": 156, "util_count_spearman": -0.06630525833320162}, "probe_overhead_sec_total": 420.458, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.010665, "alpha": 0.03, "t": 2509.1} |
| {"event": "consolidate", "step": 1700, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 18, "uncertain_staged": 262120, "t": 2618.0} |
| {"event": "train", "step": 1700, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.60626, "bpb": 2.31734, "chunk_loss": 0.030705, "F": 0.21338, "G": 0.22654, "chunks": 218.5, "bytes_per_chunk": 4.686, "boundary_entropy": 0.15189, "router_noise": 0.0, "grad_norm": 0.7567, "lr": 0.0002618, "router_trainable": true, "sec_per_step": 2.227, "probe_sec": 0.219, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.14597, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.15568, "n_shadow": 3, "mean_delta_shadow": 0.01534, "n_split": 1, "mean_delta_split": 0.02121, "n_merge": 1, "mean_delta_merge": -0.17212, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 1.08365, "trunc_abs_frac": 1.23618, "mean_delta_full": -0.04502, "read_hit_rate": {"surface": 0.46667, "split": 0.46007, "merge": 0.08011, "skel": 0.30759}, "mean_gate": {"surface": 0.25265, "split": 0.10868, "merge": 0.16737, "skel": 0.16873}, "n_reads": 10981, "n_cand": 31228, "read_suppressed_frac": 0.1, "dropout_reads_logged": 10981, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 658, "admitted_surface": 57420, "staged_surface": 56370, "admitted_split": 5511, "staged_split": 44100, "admitted_merge": 2513, "staged_merge": 115935, "admitted_skel": 92, "staged_skel": 45739, "mean_count": 174.12330627441406, "n_util_obs": 167, "util_count_spearman": -0.06400670947695622}, "probe_overhead_sec_total": 429.641, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.010452, "alpha": 0.03, "t": 2620.4} |
| {"event": "consolidate", "step": 1750, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 28, "uncertain_staged": 262121, "t": 2734.3} |
| {"event": "train", "step": 1750, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.55617, "bpb": 2.24507, "chunk_loss": 0.030552, "F": 0.20737, "G": 0.22195, "chunks": 212.34, "bytes_per_chunk": 4.822, "boundary_entropy": 0.14839, "router_noise": 0.0, "grad_norm": 0.7524, "lr": 0.0002591, "router_trainable": true, "sec_per_step": 2.335, "probe_sec": 0.1476, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06308, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": 0.1079, "n_shadow": 5, "mean_delta_shadow": -0.08348, "n_split": 1, "mean_delta_split": -0.03906, "n_merge": 2, "mean_delta_merge": 0.0361, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.001, "trunc_abs_frac": 0.29843, "mean_delta_full": -0.05062, "read_hit_rate": {"surface": 0.45601, "split": 0.46782, "merge": 0.07919, "skel": 0.32236}, "mean_gate": {"surface": 0.24311, "split": 0.102, "merge": 0.16977, "skel": 0.16306}, "n_reads": 10893, "n_cand": 30653, "read_suppressed_frac": 0.0987, "dropout_reads_logged": 10893, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 674, "admitted_surface": 57409, "staged_surface": 56369, "admitted_split": 5512, "staged_split": 44696, "admitted_merge": 2523, "staged_merge": 114786, "admitted_skel": 92, "staged_skel": 46293, "mean_count": 181.85557556152344, "n_util_obs": 175, "util_count_spearman": -0.056036721898790866}, "probe_overhead_sec_total": 437.628, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.010443, "alpha": 0.03, "t": 2737.2} |
| {"event": "consolidate", "step": 1800, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 68, "uncertain_staged": 262120, "t": 2857.8} |
| {"event": "probe", "step": 1800, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3389.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.194737, "cps": 26.476562, "mean_p": 0.211495, "boundary_entropy": 0.148667, "bpb": 2.141582}, "tr": {"chunks": 3906.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.215765, "cps": 30.515625, "mean_p": 0.217788, "boundary_entropy": 0.170208, "bpb": 2.377915}, "fi": {"chunks": 4818.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.256127, "cps": 37.640625, "mean_p": 0.264714, "boundary_entropy": 0.158862, "bpb": 2.510124}}, "bpb": {"en": 2.14158, "tr": 2.37792, "fi": 2.51012}, "gini_cpc": 0.061394, "gini_cps": 0.078648, "bpic": 4.683563, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.1079859132686316, "fi": 1.3152474352388788}, "inventory": {"size": 327611, "admitted": 65536, "staged": 262075, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 704, "admitted_surface": 57347, "staged_surface": 56182, "admitted_split": 5527, "staged_split": 44521, "admitted_merge": 2552, "staged_merge": 115521, "admitted_skel": 110, "staged_skel": 45851, "mean_count": 195.35475158691406, "n_util_obs": 186, "util_count_spearman": -0.08216645142641348}, "t": 2860.8} |
| {"event": "eval", "step": 1800, "phase": 1, "bpb": {"en": 2.15382, "tr": 2.3873, "fi": 2.52114}, "bytes_per_chunk": {"en": 4.776, "tr": 5.125, "fi": 4.153}, "bpb_mean": 2.35409, "frozen_inventory": true, "t": 2868.4} |
| {"event": "train", "step": 1800, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.52648, "bpb": 2.20225, "chunk_loss": 0.030582, "F": 0.20816, "G": 0.22256, "chunks": 213.16, "bytes_per_chunk": 4.804, "boundary_entropy": 0.14705, "router_noise": 0.0, "grad_norm": 0.6977, "lr": 0.0002564, "router_trainable": true, "sec_per_step": 2.788, "probe_sec": 0.1757, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06736, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 1, "mean_delta_loo": 0.02417, "n_shadow": 7, "mean_delta_shadow": 0.10126, "n_split": 1, "mean_delta_split": -0.28027, "n_merge": 3, "mean_delta_merge": 0.09547, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.84677, "trunc_abs_frac": 0.88207, "mean_delta_full": -0.2382, "read_hit_rate": {"surface": 0.46516, "split": 0.48786, "merge": 0.07711, "skel": 0.331}, "mean_gate": {"surface": 0.25161, "split": 0.10723, "merge": 0.16944, "skel": 0.16712}, "n_reads": 11297, "n_cand": 30864, "read_suppressed_frac": 0.0974, "dropout_reads_logged": 11297, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 706, "admitted_surface": 57347, "staged_surface": 56193, "admitted_split": 5527, "staged_split": 44520, "admitted_merge": 2552, "staged_merge": 115555, "admitted_skel": 110, "staged_skel": 45876, "mean_count": 195.5271453857422, "n_util_obs": 186, "util_count_spearman": -0.08199864888113984}, "probe_overhead_sec_total": 447.318, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.010508, "alpha": 0.03, "t": 2876.6} |
| {"event": "consolidate", "step": 1850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 49, "uncertain_staged": 262121, "t": 3020.3} |
| {"event": "train", "step": 1850, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.56054, "bpb": 2.25138, "chunk_loss": 0.030739, "F": 0.21683, "G": 0.22819, "chunks": 222.03, "bytes_per_chunk": 4.612, "boundary_entropy": 0.13773, "router_noise": 0.0, "grad_norm": 0.8466, "lr": 0.0002537, "router_trainable": true, "sec_per_step": 2.922, "probe_sec": 0.2125, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.15058, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": 0.26845, "n_shadow": 4, "mean_delta_shadow": 0.0168, "n_split": 2, "mean_delta_split": 0.06824, "n_merge": 2, "mean_delta_merge": 0.30951, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 1.09929, "trunc_abs_frac": 0.83043, "mean_delta_full": 0.38574, "read_hit_rate": {"surface": 0.47547, "split": 0.49106, "merge": 0.07968, "skel": 0.32478}, "mean_gate": {"surface": 0.22679, "split": 0.09381, "merge": 0.15001, "skel": 0.1465}, "n_reads": 11781, "n_cand": 31949, "read_suppressed_frac": 0.0934, "dropout_reads_logged": 11781, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 731, "admitted_surface": 57304, "staged_surface": 56085, "admitted_split": 5535, "staged_split": 44458, "admitted_merge": 2572, "staged_merge": 116028, "admitted_skel": 125, "staged_skel": 45573, "mean_count": 204.46656799316406, "n_util_obs": 194, "util_count_spearman": -0.036604717861061}, "probe_overhead_sec_total": 457.834, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.010303, "alpha": 0.03, "t": 3022.7} |
| {"event": "consolidate", "step": 1900, "retired": 1, "split_retired": 0, "merge_admitted": 0, "admitted": 1, "swapped": 39, "uncertain_staged": 262121, "t": 3155.2} |
| {"event": "train", "step": 1900, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.49404, "bpb": 2.15544, "chunk_loss": 0.03067, "F": 0.21304, "G": 0.22481, "chunks": 218.16, "bytes_per_chunk": 4.694, "boundary_entropy": 0.13434, "router_noise": 0.0, "grad_norm": 0.7047, "lr": 0.0002508, "router_trainable": true, "sec_per_step": 2.706, "probe_sec": 0.2383, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08735, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.01815, "n_shadow": 6, "mean_delta_shadow": -0.05428, "n_split": 1, "mean_delta_split": -0.04004, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -0.2074, "trunc_abs_frac": 1.70806, "mean_delta_full": -0.12346, "read_hit_rate": {"surface": 0.4782, "split": 0.49338, "merge": 0.0759, "skel": 0.31652}, "mean_gate": {"surface": 0.24319, "split": 0.10116, "merge": 0.16958, "skel": 0.15652}, "n_reads": 11589, "n_cand": 31482, "read_suppressed_frac": 0.1016, "dropout_reads_logged": 11589, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 752, "admitted_surface": 57284, "staged_surface": 56256, "admitted_split": 5537, "staged_split": 44331, "admitted_merge": 2585, "staged_merge": 115769, "admitted_skel": 130, "staged_skel": 45788, "mean_count": 212.59219360351562, "n_util_obs": 196, "util_count_spearman": -0.0341575755643921}, "probe_overhead_sec_total": 466.937, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.010474, "alpha": 0.03, "t": 3158.0} |
| {"event": "consolidate", "step": 1950, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 30, "uncertain_staged": 262121, "t": 3267.4} |
| {"event": "train", "step": 1950, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.57282, "bpb": 2.26909, "chunk_loss": 0.030794, "F": 0.21741, "G": 0.22893, "chunks": 222.62, "bytes_per_chunk": 4.6, "boundary_entropy": 0.13767, "router_noise": 0.0, "grad_norm": 0.7643, "lr": 0.0002479, "router_trainable": true, "sec_per_step": 2.229, "probe_sec": 0.1225, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06011, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.00751, "n_shadow": 5, "mean_delta_shadow": 0.01973, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.23877, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": -3.04081, "trunc_abs_frac": 1.45274, "mean_delta_full": -0.02468, "read_hit_rate": {"surface": 0.47786, "split": 0.49323, "merge": 0.08144, "skel": 0.33903}, "mean_gate": {"surface": 0.23216, "split": 0.09892, "merge": 0.16713, "skel": 0.15165}, "n_reads": 11948, "n_cand": 32058, "read_suppressed_frac": 0.0974, "dropout_reads_logged": 11948, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 764, "admitted_surface": 57259, "staged_surface": 56329, "admitted_split": 5542, "staged_split": 44546, "admitted_merge": 2599, "staged_merge": 115446, "admitted_skel": 136, "staged_skel": 45823, "mean_count": 221.4238739013672, "n_util_obs": 200, "util_count_spearman": -0.028118202955073876}, "probe_overhead_sec_total": 474.788, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.010979, "alpha": 0.03, "t": 3269.5} |
| {"event": "consolidate", "step": 2000, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 42, "uncertain_staged": 262122, "t": 3403.6} |
| {"event": "probe", "step": 2000, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3438.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.197552, "cps": 26.859375, "mean_p": 0.209518, "boundary_entropy": 0.135294, "bpb": 2.119439}, "tr": {"chunks": 3826.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.211346, "cps": 29.890625, "mean_p": 0.214138, "boundary_entropy": 0.161901, "bpb": 2.332101}, "fi": {"chunks": 4850.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.257828, "cps": 37.890625, "mean_p": 0.262757, "boundary_entropy": 0.147219, "bpb": 2.470786}}, "bpb": {"en": 2.11944, "tr": 2.3321, "fi": 2.47079}, "gini_cpc": 0.06027, "gini_cps": 0.077706, "bpic": 4.683176, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0698248022104269, "fi": 1.305112962106851}, "inventory": {"size": 327637, "admitted": 65536, "staged": 262101, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 776, "admitted_surface": 57226, "staged_surface": 56348, "admitted_split": 5544, "staged_split": 44392, "admitted_merge": 2620, "staged_merge": 115672, "admitted_skel": 146, "staged_skel": 45689, "mean_count": 229.79286193847656, "n_util_obs": 210, "util_count_spearman": 0.02526330384123516}, "t": 3405.6} |
| {"event": "eval", "step": 2000, "phase": 1, "bpb": {"en": 2.13037, "tr": 2.34338, "fi": 2.48162}, "bytes_per_chunk": {"en": 4.744, "tr": 5.272, "fi": 4.125}, "bpb_mean": 2.31846, "frozen_inventory": true, "t": 3410.5} |
| {"event": "train", "step": 2000, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.53965, "bpb": 2.22124, "chunk_loss": 0.03081, "F": 0.2204, "G": 0.23009, "chunks": 225.69, "bytes_per_chunk": 4.537, "boundary_entropy": 0.134, "router_noise": 0.0, "grad_norm": 0.8238, "lr": 0.000245, "router_trainable": true, "sec_per_step": 2.953, "probe_sec": 0.0967, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05944, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": -0.00481, "n_shadow": 6, "mean_delta_shadow": -0.03394, "n_split": 1, "mean_delta_split": 0.04456, "n_merge": 1, "mean_delta_merge": 0.32739, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.41568, "trunc_abs_frac": 1.12474, "mean_delta_full": 0.14658, "read_hit_rate": {"surface": 0.50848, "split": 0.49332, "merge": 0.10562, "skel": 0.34268}, "mean_gate": {"surface": 0.2369, "split": 0.09301, "merge": 0.16063, "skel": 0.14835}, "n_reads": 12386, "n_cand": 32157, "read_suppressed_frac": 0.1, "dropout_reads_logged": 12386, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 777, "admitted_surface": 57226, "staged_surface": 56336, "admitted_split": 5544, "staged_split": 44332, "admitted_merge": 2620, "staged_merge": 115772, "admitted_skel": 146, "staged_skel": 45704, "mean_count": 229.98194885253906, "n_util_obs": 210, "util_count_spearman": 0.025929977657407798}, "probe_overhead_sec_total": 484.751, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.01055, "alpha": 0.03, "t": 3417.1} |
| {"event": "consolidate", "step": 2050, "retired": 1, "split_retired": 2, "merge_admitted": 0, "admitted": 3, "swapped": 125, "uncertain_staged": 262121, "t": 3526.7} |
| {"event": "train", "step": 2050, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.54945, "bpb": 2.23538, "chunk_loss": 0.030501, "F": 0.20493, "G": 0.21704, "chunks": 209.84, "bytes_per_chunk": 4.88, "boundary_entropy": 0.13474, "router_noise": 0.0, "grad_norm": 0.799, "lr": 0.0002419, "router_trainable": true, "sec_per_step": 2.228, "probe_sec": 0.1286, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05341, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": 0.08708, "n_shadow": 4, "mean_delta_shadow": 0.26402, "n_split": 1, "mean_delta_split": 0.06813, "n_merge": 2, "mean_delta_merge": -1.14262, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.99125, "trunc_abs_frac": 0.75049, "mean_delta_full": -0.57882, "read_hit_rate": {"surface": 0.47075, "split": 0.49237, "merge": 0.08119, "skel": 0.33649}, "mean_gate": {"surface": 0.24974, "split": 0.10888, "merge": 0.16199, "skel": 0.16633}, "n_reads": 11231, "n_cand": 30294, "read_suppressed_frac": 0.1012, "dropout_reads_logged": 11231, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 788, "admitted_surface": 57105, "staged_surface": 56449, "admitted_split": 5574, "staged_split": 44434, "admitted_merge": 2667, "staged_merge": 115611, "admitted_skel": 190, "staged_skel": 45650, "mean_count": 238.18008422851562, "n_util_obs": 213, "util_count_spearman": 0.01625518025212993}, "probe_overhead_sec_total": 492.556, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.01095, "alpha": 0.03, "t": 3528.5} |
| {"event": "consolidate", "step": 2100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 467, "uncertain_staged": 262121, "t": 3638.5} |
| {"event": "train", "step": 2100, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.46962, "bpb": 2.12021, "chunk_loss": 0.030524, "F": 0.20798, "G": 0.2164, "chunks": 212.97, "bytes_per_chunk": 4.808, "boundary_entropy": 0.12655, "router_noise": 0.0, "grad_norm": 0.7213, "lr": 0.0002388, "router_trainable": true, "sec_per_step": 2.238, "probe_sec": 0.1252, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05813, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.07571, "n_shadow": 6, "mean_delta_shadow": -0.02492, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -0.05995, "trunc_abs_frac": 14.54147, "mean_delta_full": -0.0509, "read_hit_rate": {"surface": 0.48047, "split": 0.504, "merge": 0.10769, "skel": 0.33814}, "mean_gate": {"surface": 0.23528, "split": 0.09633, "merge": 0.15604, "skel": 0.15911}, "n_reads": 11812, "n_cand": 30797, "read_suppressed_frac": 0.1041, "dropout_reads_logged": 11812, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 811, "admitted_surface": 56654, "staged_surface": 56298, "admitted_split": 5599, "staged_split": 44321, "admitted_merge": 3062, "staged_merge": 115976, "admitted_skel": 221, "staged_skel": 45549, "mean_count": 247.79302978515625, "n_util_obs": 221, "util_count_spearman": 0.009492259189372772}, "probe_overhead_sec_total": 501.464, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.012487, "alpha": 0.03, "t": 3640.4} |
| {"event": "consolidate", "step": 2150, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 88, "uncertain_staged": 262121, "t": 3743.5} |
| {"event": "train", "step": 2150, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.4855, "bpb": 2.14312, "chunk_loss": 0.030696, "F": 0.21613, "G": 0.22475, "chunks": 221.31, "bytes_per_chunk": 4.627, "boundary_entropy": 0.12669, "router_noise": 0.0, "grad_norm": 0.7436, "lr": 0.0002357, "router_trainable": true, "sec_per_step": 2.142, "probe_sec": 0.1572, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07824, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": -0.04472, "n_shadow": 7, "mean_delta_shadow": 0.04026, "n_split": 1, "mean_delta_split": 0.02374, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.04058, "trunc_abs_frac": 0.14082, "mean_delta_full": -0.36194, "read_hit_rate": {"surface": 0.4905, "split": 0.4902, "merge": 0.11941, "skel": 0.34752}, "mean_gate": {"surface": 0.22479, "split": 0.08694, "merge": 0.14862, "skel": 0.14202}, "n_reads": 12227, "n_cand": 31820, "read_suppressed_frac": 0.1041, "dropout_reads_logged": 12227, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 825, "admitted_surface": 56574, "staged_surface": 56306, "admitted_split": 5617, "staged_split": 44493, "admitted_merge": 3101, "staged_merge": 115687, "admitted_skel": 244, "staged_skel": 45658, "mean_count": 256.54656982421875, "n_util_obs": 224, "util_count_spearman": -0.006473414477898782}, "probe_overhead_sec_total": 511.674, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.012139, "alpha": 0.03, "t": 3747.5} |
| {"event": "consolidate", "step": 2200, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 85, "uncertain_staged": 262121, "t": 3878.4} |
| {"event": "probe", "step": 2200, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3412.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.196058, "cps": 26.65625, "mean_p": 0.205899, "boundary_entropy": 0.127176, "bpb": 2.087716}, "tr": {"chunks": 3929.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.217036, "cps": 30.695312, "mean_p": 0.212842, "boundary_entropy": 0.151112, "bpb": 2.294121}, "fi": {"chunks": 4791.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.254691, "cps": 37.429688, "mean_p": 0.257729, "boundary_entropy": 0.13189, "bpb": 2.448747}}, "bpb": {"en": 2.08772, "tr": 2.29412, "fi": 2.44875}, "gini_cpc": 0.058535, "gini_cps": 0.075778, "bpic": 4.676228, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.1069973343235422, "fi": 1.2990605226307748}, "inventory": {"size": 327594, "admitted": 65536, "staged": 262058, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 848, "admitted_surface": 56499, "staged_surface": 56263, "admitted_split": 5628, "staged_split": 44638, "admitted_merge": 3137, "staged_merge": 115650, "admitted_skel": 272, "staged_skel": 45507, "mean_count": 265.4309997558594, "n_util_obs": 228, "util_count_spearman": 0.032500503707183624}, "t": 3881.3} |
| {"event": "eval", "step": 2200, "phase": 1, "bpb": {"en": 2.09957, "tr": 2.30255, "fi": 2.45808}, "bytes_per_chunk": {"en": 4.755, "tr": 5.147, "fi": 4.167}, "bpb_mean": 2.28673, "frozen_inventory": true, "t": 3887.2} |
| {"event": "train", "step": 2200, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.52594, "bpb": 2.20147, "chunk_loss": 0.030761, "F": 0.21652, "G": 0.2241, "chunks": 221.72, "bytes_per_chunk": 4.618, "boundary_entropy": 0.12449, "router_noise": 0.0, "grad_norm": 0.7027, "lr": 0.0002325, "router_trainable": true, "sec_per_step": 3.028, "probe_sec": 0.1402, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07588, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 3, "mean_delta_loo": 0.0201, "n_shadow": 4, "mean_delta_shadow": -0.10226, "n_split": 3, "mean_delta_split": 0.05416, "n_merge": 1, "mean_delta_merge": 0.54175, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 1.5517, "trunc_abs_frac": 0.39771, "mean_delta_full": 0.02155, "read_hit_rate": {"surface": 0.50757, "split": 0.49562, "merge": 0.12672, "skel": 0.35734}, "mean_gate": {"surface": 0.21933, "split": 0.0869, "merge": 0.14653, "skel": 0.1399}, "n_reads": 12530, "n_cand": 31863, "read_suppressed_frac": 0.1045, "dropout_reads_logged": 12530, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 849, "admitted_surface": 56499, "staged_surface": 56307, "admitted_split": 5628, "staged_split": 44619, "admitted_merge": 3137, "staged_merge": 115639, "admitted_skel": 272, "staged_skel": 45579, "mean_count": 265.6221923828125, "n_util_obs": 228, "util_count_spearman": 0.03247114188140455}, "probe_overhead_sec_total": 522.531, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.01245, "alpha": 0.03, "t": 3898.9} |
| {"event": "consolidate", "step": 2250, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 122, "uncertain_staged": 262121, "t": 4023.9} |
| {"event": "train", "step": 2250, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.49963, "bpb": 2.16351, "chunk_loss": 0.030586, "F": 0.21115, "G": 0.21967, "chunks": 216.22, "bytes_per_chunk": 4.736, "boundary_entropy": 0.11569, "router_noise": 0.0, "grad_norm": 0.7717, "lr": 0.0002293, "router_trainable": true, "sec_per_step": 2.535, "probe_sec": 0.1197, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06207, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": 0.0254, "n_shadow": 6, "mean_delta_shadow": 0.02283, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": -0.02844, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.76866, "trunc_abs_frac": 0.76866, "mean_delta_full": 0.1145, "read_hit_rate": {"surface": 0.49078, "split": 0.49951, "merge": 0.12677, "skel": 0.3552}, "mean_gate": {"surface": 0.22265, "split": 0.08524, "merge": 0.14355, "skel": 0.14622}, "n_reads": 12173, "n_cand": 31125, "read_suppressed_frac": 0.1027, "dropout_reads_logged": 12173, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 868, "admitted_surface": 56382, "staged_surface": 56034, "admitted_split": 5644, "staged_split": 44196, "admitted_merge": 3200, "staged_merge": 116522, "admitted_skel": 310, "staged_skel": 45392, "mean_count": 275.14141845703125, "n_util_obs": 231, "util_count_spearman": 0.04539126540625791}, "probe_overhead_sec_total": 532.021, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.012165, "alpha": 0.03, "t": 4025.6} |
| {"event": "consolidate", "step": 2300, "retired": 2, "split_retired": 0, "merge_admitted": 0, "admitted": 2, "swapped": 68, "uncertain_staged": 262122, "t": 4144.0} |
| {"event": "train", "step": 2300, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.53891, "bpb": 2.22018, "chunk_loss": 0.030658, "F": 0.21234, "G": 0.22153, "chunks": 217.44, "bytes_per_chunk": 4.709, "boundary_entropy": 0.11723, "router_noise": 0.0, "grad_norm": 0.6918, "lr": 0.000226, "router_trainable": true, "sec_per_step": 2.419, "probe_sec": 0.1816, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07361, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.05552, "n_shadow": 4, "mean_delta_shadow": 0.03576, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.00525, "n_dropped_unpaired": 0, "n_full": 7, "trunc_ratio_sum": 0.27401, "trunc_abs_frac": 0.76407, "mean_delta_full": 0.04998, "read_hit_rate": {"surface": 0.50202, "split": 0.48848, "merge": 0.13621, "skel": 0.33703}, "mean_gate": {"surface": 0.22699, "split": 0.08727, "merge": 0.14883, "skel": 0.13958}, "n_reads": 12003, "n_cand": 30947, "read_suppressed_frac": 0.1015, "dropout_reads_logged": 12003, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 875, "admitted_surface": 56322, "staged_surface": 56404, "admitted_split": 5652, "staged_split": 44632, "admitted_merge": 3221, "staged_merge": 115661, "admitted_skel": 341, "staged_skel": 45447, "mean_count": 283.4376220703125, "n_util_obs": 237, "util_count_spearman": 0.07413114952861981}, "probe_overhead_sec_total": 541.264, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.011762, "alpha": 0.03, "t": 4146.6} |
| {"event": "consolidate", "step": 2350, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 60, "uncertain_staged": 262122, "t": 4260.6} |
| {"event": "train", "step": 2350, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.45715, "bpb": 2.10222, "chunk_loss": 0.030625, "F": 0.2124, "G": 0.22107, "chunks": 217.5, "bytes_per_chunk": 4.708, "boundary_entropy": 0.11355, "router_noise": 0.0, "grad_norm": 0.7546, "lr": 0.0002226, "router_trainable": true, "sec_per_step": 2.323, "probe_sec": 0.3261, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.19548, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.11005, "n_shadow": 2, "mean_delta_shadow": 0.03174, "n_split": 2, "mean_delta_split": 0.02676, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.00019, "trunc_abs_frac": 0.13869, "mean_delta_full": -0.06265, "read_hit_rate": {"surface": 0.50318, "split": 0.51054, "merge": 0.13399, "skel": 0.36509}, "mean_gate": {"surface": 0.20982, "split": 0.08066, "merge": 0.13365, "skel": 0.13131}, "n_reads": 12628, "n_cand": 31458, "read_suppressed_frac": 0.0968, "dropout_reads_logged": 12628, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 906, "admitted_surface": 56274, "staged_surface": 56437, "admitted_split": 5658, "staged_split": 44960, "admitted_merge": 3244, "staged_merge": 114988, "admitted_skel": 360, "staged_skel": 45759, "mean_count": 292.496826171875, "n_util_obs": 245, "util_count_spearman": 0.058032309123761734}, "probe_overhead_sec_total": 550.677, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.011686, "alpha": 0.03, "t": 4262.7} |
| {"event": "consolidate", "step": 2400, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 51, "uncertain_staged": 262121, "t": 4381.4} |
| {"event": "probe", "step": 2400, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3410.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.195943, "cps": 26.640625, "mean_p": 0.205995, "boundary_entropy": 0.119707, "bpb": 2.065197}, "tr": {"chunks": 3854.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.212893, "cps": 30.109375, "mean_p": 0.208869, "boundary_entropy": 0.140609, "bpb": 2.26284}, "fi": {"chunks": 4879.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.25937, "cps": 38.117188, "mean_p": 0.261225, "boundary_entropy": 0.131161, "bpb": 2.394544}}, "bpb": {"en": 2.0652, "tr": 2.26284, "fi": 2.39454}, "gini_cpc": 0.06328, "gini_cps": 0.08065, "bpic": 4.671992, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0865029256666359, "fi": 1.323697278266242}, "inventory": {"size": 327628, "admitted": 65536, "staged": 262092, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 933, "admitted_surface": 56233, "staged_surface": 56262, "admitted_split": 5666, "staged_split": 45192, "admitted_merge": 3268, "staged_merge": 114701, "admitted_skel": 369, "staged_skel": 45937, "mean_count": 301.41595458984375, "n_util_obs": 251, "util_count_spearman": 0.06024398790438126}, "t": 4384.2} |
| {"event": "eval", "step": 2400, "phase": 1, "bpb": {"en": 2.07806, "tr": 2.2711, "fi": 2.4072}, "bytes_per_chunk": {"en": 4.766, "tr": 5.236, "fi": 4.119}, "bpb_mean": 2.25212, "frozen_inventory": true, "t": 4388.8} |
| {"event": "phase_transition", "step": 2400, "phase": 2, "objective": "global", "router_trainable": true, "encoder_trainable": true, "frozen_router_params": 0, "t": 4397.4} |
| {"event": "train", "step": 2400, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.461, "bpb": 2.10777, "chunk_loss": 0.030541, "F": 0.20862, "G": 0.21719, "chunks": 213.62, "bytes_per_chunk": 4.793, "boundary_entropy": 0.11896, "router_noise": 0.0, "grad_norm": 0.6802, "lr": 0.0002192, "router_trainable": true, "sec_per_step": 2.724, "probe_sec": 0.0823, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04743, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.10121, "n_shadow": 3, "mean_delta_shadow": 0.07233, "n_split": 2, "mean_delta_split": 0.35144, "n_merge": 1, "mean_delta_merge": 0.5824, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 2.09102, "trunc_abs_frac": 1.40811, "mean_delta_full": 0.07242, "read_hit_rate": {"surface": 0.49442, "split": 0.49507, "merge": 0.12434, "skel": 0.34924}, "mean_gate": {"surface": 0.21626, "split": 0.08378, "merge": 0.13308, "skel": 0.13762}, "n_reads": 11979, "n_cand": 30832, "read_suppressed_frac": 0.101, "dropout_reads_logged": 11979, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 931, "admitted_surface": 56233, "staged_surface": 56277, "admitted_split": 5666, "staged_split": 45198, "admitted_merge": 3268, "staged_merge": 114688, "admitted_skel": 369, "staged_skel": 45981, "mean_count": 301.5987548828125, "n_util_obs": 252, "util_count_spearman": 0.0666571558031308}, "probe_overhead_sec_total": 560.025, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.011674, "alpha": 0.03, "t": 4398.9} |
| {"event": "consolidate", "step": 2450, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 67, "uncertain_staged": 262121, "t": 4503.2} |
| {"event": "train", "step": 2450, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.47004, "bpb": 2.12082, "chunk_loss": 0.030798, "F": 0.22003, "G": 0.22757, "chunks": 225.31, "bytes_per_chunk": 4.545, "boundary_entropy": 0.11541, "router_noise": 0.0, "grad_norm": 0.7583, "lr": 0.0002158, "router_trainable": true, "sec_per_step": 2.116, "probe_sec": 0.1106, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06575, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.01859, "n_shadow": 6, "mean_delta_shadow": -0.05898, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.39429, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.75394, "trunc_abs_frac": 0.45483, "mean_delta_full": -0.19613, "read_hit_rate": {"surface": 0.51895, "split": 0.50024, "merge": 0.12692, "skel": 0.38296}, "mean_gate": {"surface": 0.21997, "split": 0.08535, "merge": 0.13643, "skel": 0.13792}, "n_reads": 12945, "n_cand": 32213, "read_suppressed_frac": 0.0996, "dropout_reads_logged": 12945, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 973, "admitted_surface": 56177, "staged_surface": 56325, "admitted_split": 5676, "staged_split": 45150, "admitted_merge": 3301, "staged_merge": 114886, "admitted_skel": 382, "staged_skel": 45783, "mean_count": 310.72503662109375, "n_util_obs": 262, "util_count_spearman": 0.10400946150176897}, "probe_overhead_sec_total": 568.54, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.011639, "alpha": 0.03, "t": 4504.7} |
| {"event": "consolidate", "step": 2500, "retired": 1, "split_retired": 1, "merge_admitted": 0, "admitted": 2, "swapped": 120, "uncertain_staged": 262121, "t": 4607.8} |
| {"event": "train", "step": 2500, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.56893, "bpb": 2.26349, "chunk_loss": 0.030724, "F": 0.21765, "G": 0.22381, "chunks": 222.88, "bytes_per_chunk": 4.595, "boundary_entropy": 0.11406, "router_noise": 0.0, "grad_norm": 0.7331, "lr": 0.0002123, "router_trainable": true, "sec_per_step": 2.101, "probe_sec": 0.1105, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06327, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": -0.05588, "n_shadow": 5, "mean_delta_shadow": -0.12761, "n_split": 1, "mean_delta_split": -0.08215, "n_merge": 1, "mean_delta_merge": -0.96631, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.12795, "trunc_abs_frac": 0.12795, "mean_delta_full": -0.10876, "read_hit_rate": {"surface": 0.51761, "split": 0.4945, "merge": 0.12606, "skel": 0.35559}, "mean_gate": {"surface": 0.22077, "split": 0.08265, "merge": 0.1414, "skel": 0.13043}, "n_reads": 12439, "n_cand": 31549, "read_suppressed_frac": 0.0984, "dropout_reads_logged": 12439, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 990, "admitted_surface": 56084, "staged_surface": 56344, "admitted_split": 5696, "staged_split": 45045, "admitted_merge": 3353, "staged_merge": 114833, "admitted_skel": 403, "staged_skel": 45922, "mean_count": 319.74755859375, "n_util_obs": 266, "util_count_spearman": 0.08084691270257476}, "probe_overhead_sec_total": 576.363, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.011113, "alpha": 0.03, "t": 4609.8} |
| {"event": "consolidate", "step": 2550, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 104, "uncertain_staged": 262121, "t": 4711.1} |
| {"event": "train", "step": 2550, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.49213, "bpb": 2.15269, "chunk_loss": 0.03059, "F": 0.21341, "G": 0.22047, "chunks": 218.53, "bytes_per_chunk": 4.686, "boundary_entropy": 0.11088, "router_noise": 0.0, "grad_norm": 0.7095, "lr": 0.0002088, "router_trainable": true, "sec_per_step": 2.054, "probe_sec": 0.2014, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06138, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.09384, "n_shadow": 4, "mean_delta_shadow": 0.08835, "n_split": 1, "mean_delta_split": 0.0289, "n_merge": 1, "mean_delta_merge": 0.00146, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.89591, "trunc_abs_frac": 1.01654, "mean_delta_full": -0.1806, "read_hit_rate": {"surface": 0.50596, "split": 0.50824, "merge": 0.12614, "skel": 0.37412}, "mean_gate": {"surface": 0.21897, "split": 0.08403, "merge": 0.13467, "skel": 0.1378}, "n_reads": 12652, "n_cand": 31564, "read_suppressed_frac": 0.1028, "dropout_reads_logged": 12652, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1008, "admitted_surface": 56040, "staged_surface": 56369, "admitted_split": 5710, "staged_split": 44998, "admitted_merge": 3371, "staged_merge": 114987, "admitted_skel": 415, "staged_skel": 45790, "mean_count": 328.53009033203125, "n_util_obs": 276, "util_count_spearman": 0.09526349442266824}, "probe_overhead_sec_total": 585.007, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.011427, "alpha": 0.03, "t": 4712.5} |
| {"event": "consolidate", "step": 2600, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 93, "uncertain_staged": 262120, "t": 4815.7} |
| {"event": "probe", "step": 2600, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3414.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.196173, "cps": 26.671875, "mean_p": 0.203296, "boundary_entropy": 0.112072, "bpb": 2.046682}, "tr": {"chunks": 3827.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.211401, "cps": 29.898438, "mean_p": 0.209143, "boundary_entropy": 0.135894, "bpb": 2.243993}, "fi": {"chunks": 4862.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.258466, "cps": 37.984375, "mean_p": 0.258107, "boundary_entropy": 0.117664, "bpb": 2.354815}}, "bpb": {"en": 2.04668, "tr": 2.24399, "fi": 2.35482}, "gini_cpc": 0.062351, "gini_cps": 0.07976, "bpic": 4.687433, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.077627124304422, "fi": 1.317539591640422}, "inventory": {"size": 327586, "admitted": 65536, "staged": 262050, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1040, "admitted_surface": 56004, "staged_surface": 56299, "admitted_split": 5724, "staged_split": 44890, "admitted_merge": 3386, "staged_merge": 115371, "admitted_skel": 422, "staged_skel": 45490, "mean_count": 337.93878173828125, "n_util_obs": 282, "util_count_spearman": 0.10130537656647219}, "t": 4819.2} |
| {"event": "eval", "step": 2600, "phase": 2, "bpb": {"en": 2.05731, "tr": 2.25239, "fi": 2.3667}, "bytes_per_chunk": {"en": 4.768, "tr": 5.315, "fi": 4.115}, "bpb_mean": 2.22547, "frozen_inventory": true, "t": 4823.5} |
| {"event": "train", "step": 2600, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.50134, "bpb": 2.16598, "chunk_loss": 0.030687, "F": 0.21423, "G": 0.22121, "chunks": 219.38, "bytes_per_chunk": 4.668, "boundary_entropy": 0.10936, "router_noise": 0.0, "grad_norm": 0.6971, "lr": 0.0002053, "router_trainable": true, "sec_per_step": 2.408, "probe_sec": 0.1064, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07036, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.0071, "n_shadow": 1, "mean_delta_shadow": -0.0336, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": -0.20218, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -50.39437, "trunc_abs_frac": 0.80437, "mean_delta_full": 0.00217, "read_hit_rate": {"surface": 0.53334, "split": 0.50425, "merge": 0.11961, "skel": 0.37132}, "mean_gate": {"surface": 0.20664, "split": 0.08007, "merge": 0.13707, "skel": 0.12967}, "n_reads": 12760, "n_cand": 31649, "read_suppressed_frac": 0.0998, "dropout_reads_logged": 12760, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1042, "admitted_surface": 56004, "staged_surface": 56334, "admitted_split": 5724, "staged_split": 44821, "admitted_merge": 3386, "staged_merge": 115501, "admitted_skel": 422, "staged_skel": 45488, "mean_count": 338.133544921875, "n_util_obs": 282, "util_count_spearman": 0.0976425661410415}, "probe_overhead_sec_total": 593.742, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.010297, "alpha": 0.03, "t": 4832.9} |
| {"event": "consolidate", "step": 2650, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 33, "uncertain_staged": 262121, "t": 4941.2} |
| {"event": "train", "step": 2650, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.43303, "bpb": 2.06742, "chunk_loss": 0.030479, "F": 0.20822, "G": 0.21302, "chunks": 213.22, "bytes_per_chunk": 4.803, "boundary_entropy": 0.10661, "router_noise": 0.0, "grad_norm": 0.6361, "lr": 0.0002017, "router_trainable": true, "sec_per_step": 2.206, "probe_sec": 0.1422, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06477, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.0196, "n_shadow": 3, "mean_delta_shadow": -0.02219, "n_split": 1, "mean_delta_split": 0.13892, "n_merge": 1, "mean_delta_merge": -0.39099, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 6.50571, "trunc_abs_frac": 3.73359, "mean_delta_full": -0.02494, "read_hit_rate": {"surface": 0.52835, "split": 0.4919, "merge": 0.13005, "skel": 0.36785}, "mean_gate": {"surface": 0.20183, "split": 0.07707, "merge": 0.12602, "skel": 0.13019}, "n_reads": 12276, "n_cand": 30743, "read_suppressed_frac": 0.0994, "dropout_reads_logged": 12276, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1059, "admitted_surface": 55978, "staged_surface": 56384, "admitted_split": 5743, "staged_split": 45157, "admitted_merge": 3390, "staged_merge": 115117, "admitted_skel": 425, "staged_skel": 45486, "mean_count": 349.7290344238281, "n_util_obs": 288, "util_count_spearman": 0.09468249279625766}, "probe_overhead_sec_total": 604.772, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.011328, "alpha": 0.03, "t": 4943.2} |
| {"event": "consolidate", "step": 2700, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 20, "uncertain_staged": 262122, "t": 5046.6} |
| {"event": "train", "step": 2700, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.52213, "bpb": 2.19597, "chunk_loss": 0.030627, "F": 0.21048, "G": 0.21881, "chunks": 215.53, "bytes_per_chunk": 4.751, "boundary_entropy": 0.113, "router_noise": 0.0, "grad_norm": 0.785, "lr": 0.0001981, "router_trainable": true, "sec_per_step": 2.105, "probe_sec": 0.1634, "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": -0.08068, "n_shadow": 5, "mean_delta_shadow": -0.14951, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": -0.13602, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.84536, "trunc_abs_frac": 0.92922, "mean_delta_full": -0.15125, "read_hit_rate": {"surface": 0.51347, "split": 0.49399, "merge": 0.11444, "skel": 0.35223}, "mean_gate": {"surface": 0.21691, "split": 0.08314, "merge": 0.13822, "skel": 0.13451}, "n_reads": 12094, "n_cand": 30985, "read_suppressed_frac": 0.1023, "dropout_reads_logged": 12094, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1068, "admitted_surface": 55958, "staged_surface": 56151, "admitted_split": 5754, "staged_split": 45142, "admitted_merge": 3391, "staged_merge": 115270, "admitted_skel": 433, "staged_skel": 45581, "mean_count": 359.075439453125, "n_util_obs": 295, "util_count_spearman": 0.08832771886746893}, "probe_overhead_sec_total": 614.428, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.011798, "alpha": 0.03, "t": 5048.4} |
| {"event": "consolidate", "step": 2750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 39, "uncertain_staged": 262122, "t": 5153.5} |
| {"event": "train", "step": 2750, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.45537, "bpb": 2.09966, "chunk_loss": 0.030604, "F": 0.21378, "G": 0.21976, "chunks": 218.91, "bytes_per_chunk": 4.678, "boundary_entropy": 0.10533, "router_noise": 0.0, "grad_norm": 0.6951, "lr": 0.0001945, "router_trainable": true, "sec_per_step": 2.158, "probe_sec": 0.0958, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05595, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.15881, "n_shadow": 4, "mean_delta_shadow": -0.04582, "n_split": 1, "mean_delta_split": -0.2085, "n_merge": 2, "mean_delta_merge": 0.01171, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.06252, "trunc_abs_frac": 1.08566, "mean_delta_full": 0.15002, "read_hit_rate": {"surface": 0.52718, "split": 0.50157, "merge": 0.1324, "skel": 0.35594}, "mean_gate": {"surface": 0.20956, "split": 0.07771, "merge": 0.13533, "skel": 0.12834}, "n_reads": 12561, "n_cand": 31333, "read_suppressed_frac": 0.0971, "dropout_reads_logged": 12561, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1091, "admitted_surface": 55927, "staged_surface": 56083, "admitted_split": 5763, "staged_split": 44975, "admitted_merge": 3408, "staged_merge": 115700, "admitted_skel": 438, "staged_skel": 45386, "mean_count": 368.81158447265625, "n_util_obs": 305, "util_count_spearman": 0.10732039691933723}, "probe_overhead_sec_total": 623.519, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.011257, "alpha": 0.03, "t": 5156.3} |
| {"event": "consolidate", "step": 2800, "retired": 3, "split_retired": 2, "merge_admitted": 0, "admitted": 5, "swapped": 46, "uncertain_staged": 262122, "t": 5271.7} |
| {"event": "probe", "step": 2800, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3422.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.196633, "cps": 26.734375, "mean_p": 0.201337, "boundary_entropy": 0.104964, "bpb": 2.02083}, "tr": {"chunks": 3818.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.210904, "cps": 29.828125, "mean_p": 0.206307, "boundary_entropy": 0.124997, "bpb": 2.22717}, "fi": {"chunks": 4908.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.260911, "cps": 38.34375, "mean_p": 0.257904, "boundary_entropy": 0.110642, "bpb": 2.32903}}, "bpb": {"en": 2.02083, "tr": 2.22717, "fi": 2.32903}, "gini_cpc": 0.064107, "gini_cps": 0.08155, "bpic": 4.670069, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0725794888932358, "fi": 1.3268956966839323}, "inventory": {"size": 327629, "admitted": 65536, "staged": 262093, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1107, "admitted_surface": 55904, "staged_surface": 56091, "admitted_split": 5774, "staged_split": 44842, "admitted_merge": 3413, "staged_merge": 115629, "admitted_skel": 445, "staged_skel": 45531, "mean_count": 375.8685302734375, "n_util_obs": 309, "util_count_spearman": 0.10766896742468306}, "t": 5276.0} |
| {"event": "eval", "step": 2800, "phase": 2, "bpb": {"en": 2.03476, "tr": 2.23348, "fi": 2.33993}, "bytes_per_chunk": {"en": 4.758, "tr": 5.289, "fi": 4.086}, "bpb_mean": 2.20272, "frozen_inventory": true, "t": 5279.3} |
| {"event": "train", "step": 2800, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.44551, "bpb": 2.08543, "chunk_loss": 0.030597, "F": 0.21414, "G": 0.21812, "chunks": 219.28, "bytes_per_chunk": 4.67, "boundary_entropy": 0.10149, "router_noise": 0.0, "grad_norm": 0.6761, "lr": 0.0001909, "router_trainable": true, "sec_per_step": 2.73, "probe_sec": 0.1602, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09903, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": -0.01896, "n_shadow": 3, "mean_delta_shadow": 0.06157, "n_split": 1, "mean_delta_split": -0.02319, "n_merge": 1, "mean_delta_merge": 0.06921, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -0.00171, "trunc_abs_frac": 0.4181, "mean_delta_full": 0.07727, "read_hit_rate": {"surface": 0.54345, "split": 0.50181, "merge": 0.13232, "skel": 0.37712}, "mean_gate": {"surface": 0.20076, "split": 0.07322, "merge": 0.13177, "skel": 0.12928}, "n_reads": 12813, "n_cand": 31379, "read_suppressed_frac": 0.1054, "dropout_reads_logged": 12813, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1107, "admitted_surface": 55904, "staged_surface": 56061, "admitted_split": 5774, "staged_split": 44845, "admitted_merge": 3413, "staged_merge": 115696, "admitted_skel": 445, "staged_skel": 45542, "mean_count": 376.06402587890625, "n_util_obs": 309, "util_count_spearman": 0.10747251495763867}, "probe_overhead_sec_total": 634.376, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.010891, "alpha": 0.03, "t": 5292.8} |
| {"event": "consolidate", "step": 2850, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 33, "uncertain_staged": 262122, "t": 5381.1} |
| {"event": "train", "step": 2850, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.54929, "bpb": 2.23516, "chunk_loss": 0.030765, "F": 0.21667, "G": 0.22267, "chunks": 221.88, "bytes_per_chunk": 4.615, "boundary_entropy": 0.10167, "router_noise": 0.0, "grad_norm": 0.724, "lr": 0.0001872, "router_trainable": true, "sec_per_step": 1.794, "probe_sec": 0.2584, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.15082, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": 0.09239, "n_shadow": 4, "mean_delta_shadow": 0.01405, "n_split": 1, "mean_delta_split": 0.02069, "n_merge": 2, "mean_delta_merge": -0.22575, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.12498, "trunc_abs_frac": 0.15276, "mean_delta_full": 0.42902, "read_hit_rate": {"surface": 0.53834, "split": 0.506, "merge": 0.12166, "skel": 0.3762}, "mean_gate": {"surface": 0.20216, "split": 0.07801, "merge": 0.13336, "skel": 0.13067}, "n_reads": 12935, "n_cand": 31842, "read_suppressed_frac": 0.1024, "dropout_reads_logged": 12935, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1129, "admitted_surface": 55878, "staged_surface": 56227, "admitted_split": 5789, "staged_split": 45161, "admitted_merge": 3417, "staged_merge": 115037, "admitted_skel": 452, "staged_skel": 45719, "mean_count": 385.5700988769531, "n_util_obs": 319, "util_count_spearman": 0.1043059087951736}, "probe_overhead_sec_total": 642.209, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.009934, "alpha": 0.03, "t": 5382.5} |
| {"event": "consolidate", "step": 2900, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 33, "uncertain_staged": 262122, "t": 5487.0} |
| {"event": "train", "step": 2900, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.49506, "bpb": 2.15691, "chunk_loss": 0.030818, "F": 0.22076, "G": 0.22575, "chunks": 226.06, "bytes_per_chunk": 4.53, "boundary_entropy": 0.10156, "router_noise": 0.0, "grad_norm": 0.6463, "lr": 0.0001835, "router_trainable": true, "sec_per_step": 2.16, "probe_sec": 0.2132, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.11329, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.04498, "n_shadow": 2, "mean_delta_shadow": 0.01007, "n_split": 2, "mean_delta_split": 0.05002, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.82217, "trunc_abs_frac": 0.82217, "mean_delta_full": -0.30615, "read_hit_rate": {"surface": 0.55526, "split": 0.50192, "merge": 0.12399, "skel": 0.37207}, "mean_gate": {"surface": 0.21303, "split": 0.07796, "merge": 0.13109, "skel": 0.1349}, "n_reads": 13154, "n_cand": 32232, "read_suppressed_frac": 0.0963, "dropout_reads_logged": 13154, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1135, "admitted_surface": 55847, "staged_surface": 56094, "admitted_split": 5801, "staged_split": 45202, "admitted_merge": 3424, "staged_merge": 115458, "admitted_skel": 464, "staged_skel": 45390, "mean_count": 394.55816650390625, "n_util_obs": 331, "util_count_spearman": 0.11555860060267605}, "probe_overhead_sec_total": 651.972, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.009218, "alpha": 0.03, "t": 5490.5} |
| {"event": "consolidate", "step": 2950, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 21, "uncertain_staged": 262122, "t": 5589.1} |
| {"event": "train", "step": 2950, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.42516, "bpb": 2.05607, "chunk_loss": 0.030549, "F": 0.20871, "G": 0.21427, "chunks": 213.72, "bytes_per_chunk": 4.791, "boundary_entropy": 0.09855, "router_noise": 0.0, "grad_norm": 0.6565, "lr": 0.0001799, "router_trainable": true, "sec_per_step": 2.006, "probe_sec": 0.0883, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04836, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": 0.02103, "n_shadow": 4, "mean_delta_shadow": 0.07783, "n_split": 2, "mean_delta_split": 0.08862, "n_merge": 1, "mean_delta_merge": 0.0061, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": -0.20686, "trunc_abs_frac": 0.20686, "mean_delta_full": -0.11743, "read_hit_rate": {"surface": 0.5296, "split": 0.50259, "merge": 0.12679, "skel": 0.36925}, "mean_gate": {"surface": 0.21111, "split": 0.07825, "merge": 0.13648, "skel": 0.13572}, "n_reads": 12423, "n_cand": 30807, "read_suppressed_frac": 0.1007, "dropout_reads_logged": 12423, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1155, "admitted_surface": 55829, "staged_surface": 55982, "admitted_split": 5814, "staged_split": 45167, "admitted_merge": 3426, "staged_merge": 115335, "admitted_skel": 467, "staged_skel": 45660, "mean_count": 403.82464599609375, "n_util_obs": 341, "util_count_spearman": 0.1178422596079642}, "probe_overhead_sec_total": 660.76, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.010433, "alpha": 0.03, "t": 5590.8} |
| {"event": "consolidate", "step": 3000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 70, "uncertain_staged": 262122, "t": 5678.7} |
| {"event": "probe", "step": 3000, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3437.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.197495, "cps": 26.851562, "mean_p": 0.202801, "boundary_entropy": 0.096669, "bpb": 2.008372}, "tr": {"chunks": 3777.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.208639, "cps": 29.507812, "mean_p": 0.202821, "boundary_entropy": 0.116801, "bpb": 2.188091}, "fi": {"chunks": 4885.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.259688, "cps": 38.164062, "mean_p": 0.256421, "boundary_entropy": 0.105229, "bpb": 2.306829}}, "bpb": {"en": 2.00837, "tr": 2.18809, "fi": 2.30683}, "gini_cpc": 0.062273, "gini_cps": 0.079786, "bpic": 4.688983, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0564307196924154, "fi": 1.3149137678080907}, "inventory": {"size": 327610, "admitted": 65536, "staged": 262074, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1185, "admitted_surface": 55793, "staged_surface": 55918, "admitted_split": 5834, "staged_split": 45457, "admitted_merge": 3428, "staged_merge": 114971, "admitted_skel": 481, "staged_skel": 45728, "mean_count": 413.37921142578125, "n_util_obs": 345, "util_count_spearman": 0.12421785041656597}, "t": 5681.2} |
| {"event": "eval", "step": 3000, "phase": 2, "bpb": {"en": 2.02022, "tr": 2.19781, "fi": 2.31581}, "bytes_per_chunk": {"en": 4.749, "tr": 5.343, "fi": 4.103}, "bpb_mean": 2.17795, "frozen_inventory": true, "t": 5685.9} |
| {"event": "train", "step": 3000, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.45677, "bpb": 2.10168, "chunk_loss": 0.030525, "F": 0.21033, "G": 0.2163, "chunks": 215.38, "bytes_per_chunk": 4.754, "boundary_entropy": 0.09604, "router_noise": 0.0, "grad_norm": 0.693, "lr": 0.0001761, "router_trainable": true, "sec_per_step": 2.089, "probe_sec": 0.1545, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10929, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": -0.02912, "n_shadow": 4, "mean_delta_shadow": 0.02946, "n_split": 0, "mean_delta_split": NaN, "n_merge": 3, "mean_delta_merge": -0.3217, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": -0.02291, "trunc_abs_frac": 0.02291, "mean_delta_full": -0.42358, "read_hit_rate": {"surface": 0.54125, "split": 0.50913, "merge": 0.12258, "skel": 0.37738}, "mean_gate": {"surface": 0.20335, "split": 0.07561, "merge": 0.13506, "skel": 0.13099}, "n_reads": 12688, "n_cand": 31047, "read_suppressed_frac": 0.1021, "dropout_reads_logged": 12688, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1187, "admitted_surface": 55793, "staged_surface": 55971, "admitted_split": 5834, "staged_split": 45459, "admitted_merge": 3428, "staged_merge": 114910, "admitted_skel": 481, "staged_skel": 45804, "mean_count": 413.57281494140625, "n_util_obs": 345, "util_count_spearman": 0.12456501282898005}, "probe_overhead_sec_total": 669.072, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.011114, "alpha": 0.03, "t": 5695.3} |
| {"event": "consolidate", "step": 3050, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 58, "uncertain_staged": 262122, "t": 5778.8} |
| {"event": "train", "step": 3050, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.447, "bpb": 2.08758, "chunk_loss": 0.03058, "F": 0.21304, "G": 0.21875, "chunks": 218.16, "bytes_per_chunk": 4.694, "boundary_entropy": 0.09765, "router_noise": 0.0, "grad_norm": 0.7179, "lr": 0.0001724, "router_trainable": true, "sec_per_step": 1.696, "probe_sec": 0.1398, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08447, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": 0.03309, "n_shadow": 5, "mean_delta_shadow": -0.01865, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -1.26736, "trunc_abs_frac": 2.48106, "mean_delta_full": 0.02065, "read_hit_rate": {"surface": 0.5467, "split": 0.51472, "merge": 0.12708, "skel": 0.38365}, "mean_gate": {"surface": 0.21048, "split": 0.07696, "merge": 0.13305, "skel": 0.13909}, "n_reads": 12980, "n_cand": 31350, "read_suppressed_frac": 0.098, "dropout_reads_logged": 12980, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1208, "admitted_surface": 55745, "staged_surface": 56049, "admitted_split": 5859, "staged_split": 45325, "admitted_merge": 3439, "staged_merge": 115037, "admitted_skel": 493, "staged_skel": 45733, "mean_count": 424.378173828125, "n_util_obs": 356, "util_count_spearman": 0.12636641080673322}, "probe_overhead_sec_total": 677.199, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.010846, "alpha": 0.03, "t": 5780.1} |
| {"event": "consolidate", "step": 3100, "retired": 1, "split_retired": 2, "merge_admitted": 0, "admitted": 3, "swapped": 18, "uncertain_staged": 262123, "t": 5866.4} |
| {"event": "train", "step": 3100, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.54291, "bpb": 2.22595, "chunk_loss": 0.030594, "F": 0.20865, "G": 0.21416, "chunks": 213.66, "bytes_per_chunk": 4.793, "boundary_entropy": 0.10377, "router_noise": 0.0, "grad_norm": 0.7473, "lr": 0.0001687, "router_trainable": true, "sec_per_step": 1.765, "probe_sec": 0.1669, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.13994, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": 0.02704, "n_shadow": 5, "mean_delta_shadow": -0.01057, "n_split": 2, "mean_delta_split": -0.1165, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 80.29166, "trunc_abs_frac": 0.36915, "mean_delta_full": -0.00073, "read_hit_rate": {"surface": 0.52652, "split": 0.50338, "merge": 0.11236, "skel": 0.36088}, "mean_gate": {"surface": 0.20841, "split": 0.07636, "merge": 0.1368, "skel": 0.13213}, "n_reads": 12173, "n_cand": 30597, "read_suppressed_frac": 0.1012, "dropout_reads_logged": 12173, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1231, "admitted_surface": 55727, "staged_surface": 56304, "admitted_split": 5864, "staged_split": 45136, "admitted_merge": 3445, "staged_merge": 115012, "admitted_skel": 500, "staged_skel": 45692, "mean_count": 432.9212341308594, "n_util_obs": 357, "util_count_spearman": 0.1132602309645219}, "probe_overhead_sec_total": 687.055, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.010792, "alpha": 0.03, "t": 5868.3} |
| {"event": "consolidate", "step": 3150, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 38, "uncertain_staged": 262123, "t": 5943.5} |
| {"event": "train", "step": 3150, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.45738, "bpb": 2.10256, "chunk_loss": 0.03063, "F": 0.21341, "G": 0.21798, "chunks": 218.53, "bytes_per_chunk": 4.686, "boundary_entropy": 0.09683, "router_noise": 0.0, "grad_norm": 0.6709, "lr": 0.000165, "router_trainable": true, "sec_per_step": 1.525, "probe_sec": 0.1239, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07298, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": 0.08848, "n_shadow": 4, "mean_delta_shadow": -0.03599, "n_split": 2, "mean_delta_split": 0.06812, "n_merge": 1, "mean_delta_merge": 0.39032, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 44.01471, "trunc_abs_frac": 44.01471, "mean_delta_full": -0.00208, "read_hit_rate": {"surface": 0.53843, "split": 0.5254, "merge": 0.12931, "skel": 0.37276}, "mean_gate": {"surface": 0.20615, "split": 0.077, "merge": 0.13912, "skel": 0.12879}, "n_reads": 12995, "n_cand": 31327, "read_suppressed_frac": 0.1026, "dropout_reads_logged": 12995, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1257, "admitted_surface": 55695, "staged_surface": 56196, "admitted_split": 5879, "staged_split": 45679, "admitted_merge": 3453, "staged_merge": 114649, "admitted_skel": 509, "staged_skel": 45620, "mean_count": 441.6148681640625, "n_util_obs": 360, "util_count_spearman": 0.11023041010028457}, "probe_overhead_sec_total": 694.516, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.009388, "alpha": 0.03, "t": 5944.5} |
| {"event": "consolidate", "step": 3200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 39, "uncertain_staged": 262123, "t": 6012.0} |
| {"event": "probe", "step": 3200, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3438.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.197552, "cps": 26.859375, "mean_p": 0.202661, "boundary_entropy": 0.094126, "bpb": 1.989858}, "tr": {"chunks": 3775.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.208529, "cps": 29.492188, "mean_p": 0.200754, "boundary_entropy": 0.113725, "bpb": 2.166306}, "fi": {"chunks": 4885.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.259688, "cps": 38.164062, "mean_p": 0.257975, "boundary_entropy": 0.111714, "bpb": 2.277532}}, "bpb": {"en": 1.98986, "tr": 2.16631, "fi": 2.27753}, "gini_cpc": 0.06222, "gini_cps": 0.079738, "bpic": 4.68937, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.055564199776362, "fi": 1.3145313030705084}, "inventory": {"size": 327641, "admitted": 65536, "staged": 262105, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1278, "admitted_surface": 55659, "staged_surface": 56242, "admitted_split": 5902, "staged_split": 45311, "admitted_merge": 3458, "staged_merge": 114760, "admitted_skel": 517, "staged_skel": 45792, "mean_count": 451.2296142578125, "n_util_obs": 366, "util_count_spearman": 0.10684136689170574}, "t": 6016.0} |
| {"event": "eval", "step": 3200, "phase": 2, "bpb": {"en": 2.00371, "tr": 2.17544, "fi": 2.28787}, "bytes_per_chunk": {"en": 4.751, "tr": 5.353, "fi": 4.109}, "bpb_mean": 2.15567, "frozen_inventory": true, "t": 6020.4} |
| {"event": "train", "step": 3200, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.42466, "bpb": 2.05536, "chunk_loss": 0.030778, "F": 0.21887, "G": 0.22389, "chunks": 224.12, "bytes_per_chunk": 4.569, "boundary_entropy": 0.09531, "router_noise": 0.0, "grad_norm": 0.6529, "lr": 0.0001613, "router_trainable": true, "sec_per_step": 1.636, "probe_sec": 0.118, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08514, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": -0.16665, "n_shadow": 9, "mean_delta_shadow": -0.01143, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 5.47278, "trunc_abs_frac": 6.06513, "mean_delta_full": -0.03812, "read_hit_rate": {"surface": 0.5612, "split": 0.51814, "merge": 0.12929, "skel": 0.38667}, "mean_gate": {"surface": 0.20134, "split": 0.07365, "merge": 0.13917, "skel": 0.13371}, "n_reads": 13552, "n_cand": 32282, "read_suppressed_frac": 0.1017, "dropout_reads_logged": 13552, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1277, "admitted_surface": 55659, "staged_surface": 56205, "admitted_split": 5902, "staged_split": 45227, "admitted_merge": 3458, "staged_merge": 114932, "admitted_skel": 517, "staged_skel": 45780, "mean_count": 451.43646240234375, "n_util_obs": 366, "util_count_spearman": 0.106577515117171}, "probe_overhead_sec_total": 702.784, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.009015, "alpha": 0.03, "t": 6026.4} |
| {"event": "consolidate", "step": 3250, "retired": 2, "split_retired": 1, "merge_admitted": 0, "admitted": 3, "swapped": 135, "uncertain_staged": 262123, "t": 6079.6} |
| {"event": "train", "step": 3250, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.40027, "bpb": 2.02016, "chunk_loss": 0.030445, "F": 0.20691, "G": 0.21012, "chunks": 211.88, "bytes_per_chunk": 4.833, "boundary_entropy": 0.09253, "router_noise": 0.0, "grad_norm": 0.632, "lr": 0.0001576, "router_trainable": true, "sec_per_step": 1.081, "probe_sec": 0.0608, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04017, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.00376, "n_shadow": 6, "mean_delta_shadow": 0.01659, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.22215, "trunc_abs_frac": 1.40428, "mean_delta_full": -0.03616, "read_hit_rate": {"surface": 0.52608, "split": 0.51774, "merge": 0.11703, "skel": 0.37077}, "mean_gate": {"surface": 0.19556, "split": 0.06995, "merge": 0.12625, "skel": 0.12738}, "n_reads": 12451, "n_cand": 30614, "read_suppressed_frac": 0.1022, "dropout_reads_logged": 12451, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1297, "admitted_surface": 55533, "staged_surface": 56217, "admitted_split": 5956, "staged_split": 44932, "admitted_merge": 3513, "staged_merge": 115397, "admitted_skel": 534, "staged_skel": 45598, "mean_count": 460.05224609375, "n_util_obs": 368, "util_count_spearman": 0.09286544500310809}, "probe_overhead_sec_total": 708.995, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.009502, "alpha": 0.03, "t": 6080.4} |
| {"event": "consolidate", "step": 3300, "retired": 1, "split_retired": 2, "merge_admitted": 0, "admitted": 3, "swapped": 98, "uncertain_staged": 262123, "t": 6140.7} |
| {"event": "train", "step": 3300, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.44672, "bpb": 2.08717, "chunk_loss": 0.030624, "F": 0.21368, "G": 0.2176, "chunks": 218.81, "bytes_per_chunk": 4.68, "boundary_entropy": 0.0946, "router_noise": 0.0, "grad_norm": 0.6823, "lr": 0.0001539, "router_trainable": true, "sec_per_step": 1.241, "probe_sec": 0.2007, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.11734, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.01008, "n_shadow": 4, "mean_delta_shadow": 0.20132, "n_split": 1, "mean_delta_split": 0.08508, "n_merge": 1, "mean_delta_merge": 0.17719, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.20099, "trunc_abs_frac": 0.20149, "mean_delta_full": 0.40674, "read_hit_rate": {"surface": 0.54476, "split": 0.52631, "merge": 0.12756, "skel": 0.36833}, "mean_gate": {"surface": 0.19212, "split": 0.06851, "merge": 0.12792, "skel": 0.11965}, "n_reads": 13120, "n_cand": 31567, "read_suppressed_frac": 0.0967, "dropout_reads_logged": 13120, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1307, "admitted_surface": 55446, "staged_surface": 56106, "admitted_split": 6019, "staged_split": 44985, "admitted_merge": 3526, "staged_merge": 115538, "admitted_skel": 545, "staged_skel": 45515, "mean_count": 468.0172119140625, "n_util_obs": 371, "util_count_spearman": 0.08752096983369674}, "probe_overhead_sec_total": 715.869, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.010087, "alpha": 0.03, "t": 6142.5} |
| {"event": "consolidate", "step": 3350, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 105, "uncertain_staged": 262123, "t": 6199.8} |
| {"event": "train", "step": 3350, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.4749, "bpb": 2.12784, "chunk_loss": 0.030717, "F": 0.21902, "G": 0.22212, "chunks": 224.28, "bytes_per_chunk": 4.566, "boundary_entropy": 0.0953, "router_noise": 0.0, "grad_norm": 0.7514, "lr": 0.0001501, "router_trainable": true, "sec_per_step": 1.157, "probe_sec": 0.0754, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04955, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.08339, "n_shadow": 5, "mean_delta_shadow": -0.04323, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.1004, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -3.23599, "trunc_abs_frac": 0.6768, "mean_delta_full": 0.01379, "read_hit_rate": {"surface": 0.54738, "split": 0.52948, "merge": 0.13595, "skel": 0.36772}, "mean_gate": {"surface": 0.20934, "split": 0.07637, "merge": 0.13405, "skel": 0.1334}, "n_reads": 13391, "n_cand": 31977, "read_suppressed_frac": 0.1001, "dropout_reads_logged": 13391, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1327, "admitted_surface": 55352, "staged_surface": 56093, "admitted_split": 6082, "staged_split": 45239, "admitted_merge": 3536, "staged_merge": 115132, "admitted_skel": 566, "staged_skel": 45680, "mean_count": 477.0107421875, "n_util_obs": 378, "util_count_spearman": 0.09890510916469837}, "probe_overhead_sec_total": 723.285, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.00952, "alpha": 0.03, "t": 6200.3} |
| {"event": "consolidate", "step": 3400, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 105, "uncertain_staged": 262123, "t": 6260.9} |
| {"event": "probe", "step": 3400, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3425.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.196805, "cps": 26.757812, "mean_p": 0.200284, "boundary_entropy": 0.091579, "bpb": 1.97799}, "tr": {"chunks": 3783.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.208971, "cps": 29.554688, "mean_p": 0.199904, "boundary_entropy": 0.112663, "bpb": 2.145307}, "fi": {"chunks": 4847.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.257668, "cps": 37.867188, "mean_p": 0.253557, "boundary_entropy": 0.101997, "bpb": 2.260557}}, "bpb": {"en": 1.97799, "tr": 2.14531, "fi": 2.26056}, "gini_cpc": 0.061159, "gini_cps": 0.07864, "bpic": 4.706097, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0618161687118035, "fi": 1.3092563250187126}, "inventory": {"size": 327573, "admitted": 65536, "staged": 262037, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1339, "admitted_surface": 55257, "staged_surface": 55916, "admitted_split": 6137, "staged_split": 44865, "admitted_merge": 3551, "staged_merge": 115826, "admitted_skel": 591, "staged_skel": 45430, "mean_count": 486.0859375, "n_util_obs": 379, "util_count_spearman": 0.08956407447225395}, "t": 6263.4} |
| {"event": "eval", "step": 3400, "phase": 2, "bpb": {"en": 1.99054, "tr": 2.15499, "fi": 2.26833}, "bytes_per_chunk": {"en": 4.757, "tr": 5.337, "fi": 4.113}, "bpb_mean": 2.13795, "frozen_inventory": true, "t": 6266.9} |
| {"event": "train", "step": 3400, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.39201, "bpb": 2.00825, "chunk_loss": 0.030487, "F": 0.20755, "G": 0.2108, "chunks": 212.53, "bytes_per_chunk": 4.818, "boundary_entropy": 0.08959, "router_noise": 0.0, "grad_norm": 0.6866, "lr": 0.0001465, "router_trainable": true, "sec_per_step": 1.451, "probe_sec": 0.0926, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06894, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": -0.05389, "n_shadow": 3, "mean_delta_shadow": -0.13132, "n_split": 1, "mean_delta_split": 0.08136, "n_merge": 3, "mean_delta_merge": -0.45287, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 1.08306, "trunc_abs_frac": 0.55957, "mean_delta_full": -0.27116, "read_hit_rate": {"surface": 0.54794, "split": 0.52568, "merge": 0.12943, "skel": 0.38913}, "mean_gate": {"surface": 0.19207, "split": 0.07029, "merge": 0.12267, "skel": 0.12751}, "n_reads": 12906, "n_cand": 30705, "read_suppressed_frac": 0.0988, "dropout_reads_logged": 12906, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1341, "admitted_surface": 55257, "staged_surface": 55956, "admitted_split": 6137, "staged_split": 44950, "admitted_merge": 3551, "staged_merge": 115745, "admitted_skel": 591, "staged_skel": 45493, "mean_count": 486.28289794921875, "n_util_obs": 379, "util_count_spearman": 0.08924928856882752}, "probe_overhead_sec_total": 729.761, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.010179, "alpha": 0.03, "t": 6272.9} |
| {"event": "consolidate", "step": 3450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 58, "uncertain_staged": 262123, "t": 6327.8} |
| {"event": "train", "step": 3450, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.38636, "bpb": 2.0001, "chunk_loss": 0.030392, "F": 0.20453, "G": 0.20814, "chunks": 209.44, "bytes_per_chunk": 4.889, "boundary_entropy": 0.08772, "router_noise": 0.0, "grad_norm": 0.6114, "lr": 0.0001428, "router_trainable": true, "sec_per_step": 1.108, "probe_sec": 0.0927, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05504, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": 0.01245, "n_shadow": 3, "mean_delta_shadow": -0.06512, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -1.50639, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.31993, "trunc_abs_frac": 0.38357, "mean_delta_full": 0.06677, "read_hit_rate": {"surface": 0.53868, "split": 0.53337, "merge": 0.13197, "skel": 0.36935}, "mean_gate": {"surface": 0.18786, "split": 0.06543, "merge": 0.12621, "skel": 0.12099}, "n_reads": 12655, "n_cand": 30255, "read_suppressed_frac": 0.1026, "dropout_reads_logged": 12655, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1367, "admitted_surface": 55203, "staged_surface": 56059, "admitted_split": 6166, "staged_split": 44881, "admitted_merge": 3567, "staged_merge": 115583, "admitted_skel": 600, "staged_skel": 45621, "mean_count": 496.422607421875, "n_util_obs": 384, "util_count_spearman": 0.08707520203333363}, "probe_overhead_sec_total": 735.628, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009321, "alpha": 0.03, "t": 6328.2} |
| {"event": "consolidate", "step": 3500, "retired": 1, "split_retired": 0, "merge_admitted": 0, "admitted": 1, "swapped": 33, "uncertain_staged": 262123, "t": 6368.4} |
| {"event": "train", "step": 3500, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.47366, "bpb": 2.12604, "chunk_loss": 0.030689, "F": 0.21326, "G": 0.21724, "chunks": 218.38, "bytes_per_chunk": 4.689, "boundary_entropy": 0.08837, "router_noise": 0.0, "grad_norm": 0.6748, "lr": 0.0001391, "router_trainable": true, "sec_per_step": 0.83, "probe_sec": 0.0887, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06373, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 3, "mean_delta_loo": -0.07559, "n_shadow": 4, "mean_delta_shadow": -0.06186, "n_split": 1, "mean_delta_split": 0.06506, "n_merge": 3, "mean_delta_merge": -0.0616, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 0.15906, "trunc_abs_frac": 0.28061, "mean_delta_full": -0.34157, "read_hit_rate": {"surface": 0.54629, "split": 0.53438, "merge": 0.12883, "skel": 0.36868}, "mean_gate": {"surface": 0.18899, "split": 0.06757, "merge": 0.12609, "skel": 0.12022}, "n_reads": 13148, "n_cand": 31365, "read_suppressed_frac": 0.1015, "dropout_reads_logged": 13148, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1388, "admitted_surface": 55170, "staged_surface": 56183, "admitted_split": 6186, "staged_split": 44953, "admitted_merge": 3574, "staged_merge": 115262, "admitted_skel": 606, "staged_skel": 45746, "mean_count": 507.28973388671875, "n_util_obs": 393, "util_count_spearman": 0.08979331316695614}, "probe_overhead_sec_total": 740.773, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009664, "alpha": 0.03, "t": 6369.7} |
| {"event": "consolidate", "step": 3550, "retired": 0, "split_retired": 3, "merge_admitted": 0, "admitted": 3, "swapped": 35, "uncertain_staged": 262124, "t": 6434.2} |
| {"event": "train", "step": 3550, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.47448, "bpb": 2.12723, "chunk_loss": 0.030858, "F": 0.22202, "G": 0.22605, "chunks": 227.34, "bytes_per_chunk": 4.504, "boundary_entropy": 0.08937, "router_noise": 0.0, "grad_norm": 0.6725, "lr": 0.0001355, "router_trainable": true, "sec_per_step": 1.322, "probe_sec": 0.285, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.14389, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 3, "mean_delta_loo": 0.31264, "n_shadow": 4, "mean_delta_shadow": 0.31896, "n_split": 1, "mean_delta_split": -0.20941, "n_merge": 3, "mean_delta_merge": -0.52299, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 13.81824, "trunc_abs_frac": 13.81824, "mean_delta_full": -0.02258, "read_hit_rate": {"surface": 0.55847, "split": 0.53246, "merge": 0.13632, "skel": 0.38689}, "mean_gate": {"surface": 0.1905, "split": 0.06744, "merge": 0.12987, "skel": 0.11947}, "n_reads": 13892, "n_cand": 32624, "read_suppressed_frac": 0.101, "dropout_reads_logged": 13892, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1401, "admitted_surface": 55133, "staged_surface": 56240, "admitted_split": 6217, "staged_split": 44771, "admitted_merge": 3576, "staged_merge": 115335, "admitted_skel": 610, "staged_skel": 45798, "mean_count": 516.2760009765625, "n_util_obs": 397, "util_count_spearman": 0.09602618159236491}, "probe_overhead_sec_total": 747.415, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.00922, "alpha": 0.03, "t": 6435.8} |
| {"event": "consolidate", "step": 3600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 29, "uncertain_staged": 262124, "t": 6495.3} |
| {"event": "probe", "step": 3600, "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": 0.201068, "boundary_entropy": 0.090993, "bpb": 1.957244}, "tr": {"chunks": 3761.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.207756, "cps": 29.382812, "mean_p": 0.200156, "boundary_entropy": 0.109185, "bpb": 2.127231}, "fi": {"chunks": 4937.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.262453, "cps": 38.570312, "mean_p": 0.256693, "boundary_entropy": 0.096757, "bpb": 2.238112}}, "bpb": {"en": 1.95724, "tr": 2.12723, "fi": 2.23811}, "gini_cpc": 0.06492, "gini_cps": 0.082468, "bpic": 4.675457, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.052261660526555, "fi": 1.3292975638000428}, "inventory": {"size": 327651, "admitted": 65536, "staged": 262115, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1438, "admitted_surface": 55109, "staged_surface": 56114, "admitted_split": 6231, "staged_split": 44994, "admitted_merge": 3583, "staged_merge": 115306, "admitted_skel": 613, "staged_skel": 45701, "mean_count": 526.1041870117188, "n_util_obs": 404, "util_count_spearman": 0.10740950025523645}, "t": 6496.9} |
| {"event": "eval", "step": 3600, "phase": 2, "bpb": {"en": 1.97052, "tr": 2.13522, "fi": 2.24833}, "bytes_per_chunk": {"en": 4.754, "tr": 5.368, "fi": 4.062}, "bpb_mean": 2.11802, "frozen_inventory": true, "t": 6500.2} |
| {"event": "train", "step": 3600, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.45371, "bpb": 2.09726, "chunk_loss": 0.030511, "F": 0.21014, "G": 0.21296, "chunks": 215.19, "bytes_per_chunk": 4.759, "boundary_entropy": 0.08812, "router_noise": 0.0, "grad_norm": 0.6803, "lr": 0.0001319, "router_trainable": true, "sec_per_step": 1.424, "probe_sec": 0.1302, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05983, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 1, "mean_delta_loo": 0.16333, "n_shadow": 5, "mean_delta_shadow": -0.05391, "n_split": 0, "mean_delta_split": NaN, "n_merge": 4, "mean_delta_merge": -0.04537, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 1.91922, "trunc_abs_frac": 0.63573, "mean_delta_full": 0.04987, "read_hit_rate": {"surface": 0.53166, "split": 0.51271, "merge": 0.12811, "skel": 0.35631}, "mean_gate": {"surface": 0.18473, "split": 0.06425, "merge": 0.12474, "skel": 0.11454}, "n_reads": 12412, "n_cand": 30617, "read_suppressed_frac": 0.1023, "dropout_reads_logged": 12412, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1440, "admitted_surface": 55109, "staged_surface": 56153, "admitted_split": 6231, "staged_split": 44971, "admitted_merge": 3583, "staged_merge": 115346, "admitted_skel": 613, "staged_skel": 45674, "mean_count": 526.2935791015625, "n_util_obs": 404, "util_count_spearman": 0.10761496331601729}, "probe_overhead_sec_total": 753.809, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008988, "alpha": 0.03, "t": 6507.0} |
| {"event": "consolidate", "step": 3650, "retired": 1, "split_retired": 0, "merge_admitted": 0, "admitted": 1, "swapped": 125, "uncertain_staged": 262124, "t": 6570.8} |
| {"event": "train", "step": 3650, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.46809, "bpb": 2.11801, "chunk_loss": 0.030713, "F": 0.21671, "G": 0.22046, "chunks": 221.91, "bytes_per_chunk": 4.615, "boundary_entropy": 0.08927, "router_noise": 0.0, "grad_norm": 0.6797, "lr": 0.0001283, "router_trainable": true, "sec_per_step": 1.284, "probe_sec": 0.0768, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05424, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.08179, "n_shadow": 3, "mean_delta_shadow": -0.07445, "n_split": 2, "mean_delta_split": -0.0274, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 1.63916, "trunc_abs_frac": 1.63916, "mean_delta_full": 0.11853, "read_hit_rate": {"surface": 0.55439, "split": 0.54746, "merge": 0.12562, "skel": 0.36834}, "mean_gate": {"surface": 0.19942, "split": 0.07434, "merge": 0.13137, "skel": 0.13243}, "n_reads": 13492, "n_cand": 31722, "read_suppressed_frac": 0.1038, "dropout_reads_logged": 13492, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1472, "admitted_surface": 54987, "staged_surface": 56158, "admitted_split": 6327, "staged_split": 45095, "admitted_merge": 3595, "staged_merge": 115081, "admitted_skel": 627, "staged_skel": 45810, "mean_count": 536.6143798828125, "n_util_obs": 414, "util_count_spearman": 0.09236305934237536}, "probe_overhead_sec_total": 760.015, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009906, "alpha": 0.03, "t": 6571.2} |
| {"event": "consolidate", "step": 3700, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 84, "uncertain_staged": 262124, "t": 6624.2} |
| {"event": "train", "step": 3700, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.36916, "bpb": 1.97529, "chunk_loss": 0.030484, "F": 0.20856, "G": 0.21133, "chunks": 213.56, "bytes_per_chunk": 4.795, "boundary_entropy": 0.08999, "router_noise": 0.0, "grad_norm": 0.6947, "lr": 0.0001247, "router_trainable": true, "sec_per_step": 1.097, "probe_sec": 0.0982, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0543, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.03699, "n_shadow": 4, "mean_delta_shadow": -0.0052, "n_split": 2, "mean_delta_split": -0.03381, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.59797, "trunc_abs_frac": 0.63126, "mean_delta_full": -0.20569, "read_hit_rate": {"surface": 0.54058, "split": 0.54791, "merge": 0.13146, "skel": 0.36987}, "mean_gate": {"surface": 0.19269, "split": 0.06872, "merge": 0.13089, "skel": 0.12429}, "n_reads": 13036, "n_cand": 30722, "read_suppressed_frac": 0.0973, "dropout_reads_logged": 13036, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1493, "admitted_surface": 54905, "staged_surface": 56202, "admitted_split": 6394, "staged_split": 44790, "admitted_merge": 3604, "staged_merge": 115350, "admitted_skel": 633, "staged_skel": 45802, "mean_count": 546.0914306640625, "n_util_obs": 419, "util_count_spearman": 0.08347006932619862}, "probe_overhead_sec_total": 765.29, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009266, "alpha": 0.03, "t": 6626.1} |
| {"event": "consolidate", "step": 3750, "retired": 2, "split_retired": 0, "merge_admitted": 0, "admitted": 2, "swapped": 57, "uncertain_staged": 262124, "t": 6681.8} |
| {"event": "train", "step": 3750, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.39601, "bpb": 2.01402, "chunk_loss": 0.030767, "F": 0.22107, "G": 0.22213, "chunks": 226.38, "bytes_per_chunk": 4.523, "boundary_entropy": 0.0804, "router_noise": 0.0, "grad_norm": 0.648, "lr": 0.0001212, "router_trainable": true, "sec_per_step": 1.132, "probe_sec": 0.1183, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08796, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": -0.23363, "n_shadow": 7, "mean_delta_shadow": 0.01825, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.57312, "trunc_abs_frac": 1.10171, "mean_delta_full": -0.07902, "read_hit_rate": {"surface": 0.57099, "split": 0.55267, "merge": 0.14178, "skel": 0.4011}, "mean_gate": {"surface": 0.1738, "split": 0.05952, "merge": 0.11763, "skel": 0.11271}, "n_reads": 14266, "n_cand": 32412, "read_suppressed_frac": 0.1, "dropout_reads_logged": 14266, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1501, "admitted_surface": 54849, "staged_surface": 56251, "admitted_split": 6439, "staged_split": 44713, "admitted_merge": 3609, "staged_merge": 115351, "admitted_skel": 639, "staged_skel": 45829, "mean_count": 555.849853515625, "n_util_obs": 422, "util_count_spearman": 0.08532387525773664}, "probe_overhead_sec_total": 772.333, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008937, "alpha": 0.03, "t": 6682.7} |
| {"event": "consolidate", "step": 3800, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 58, "uncertain_staged": 262124, "t": 6745.1} |
| {"event": "probe", "step": 3800, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3429.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.197035, "cps": 26.789062, "mean_p": 0.199978, "boundary_entropy": 0.087083, "bpb": 1.945219}, "tr": {"chunks": 3762.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.207811, "cps": 29.390625, "mean_p": 0.198945, "boundary_entropy": 0.104559, "bpb": 2.108041}, "fi": {"chunks": 4893.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.260114, "cps": 38.226562, "mean_p": 0.254189, "boundary_entropy": 0.095146, "bpb": 2.212778}}, "bpb": {"en": 1.94522, "tr": 2.10804, "fi": 2.21278}, "gini_cpc": 0.063241, "gini_cps": 0.080768, "bpic": 4.694803, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0546901131365098, "fi": 1.3201399304115216}, "inventory": {"size": 327621, "admitted": 65536, "staged": 262085, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1519, "admitted_surface": 54796, "staged_surface": 56169, "admitted_split": 6481, "staged_split": 44477, "admitted_merge": 3614, "staged_merge": 115894, "admitted_skel": 645, "staged_skel": 45545, "mean_count": 565.4228515625, "n_util_obs": 430, "util_count_spearman": 0.06754347470907121}, "t": 6746.7} |
| {"event": "eval", "step": 3800, "phase": 2, "bpb": {"en": 1.95883, "tr": 2.11682, "fi": 2.22157}, "bytes_per_chunk": {"en": 4.761, "tr": 5.352, "fi": 4.093}, "bpb_mean": 2.09907, "frozen_inventory": true, "t": 6750.1} |
| {"event": "train", "step": 3800, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.3743, "bpb": 1.9827, "chunk_loss": 0.030537, "F": 0.20755, "G": 0.21007, "chunks": 212.53, "bytes_per_chunk": 4.818, "boundary_entropy": 0.08659, "router_noise": 0.0, "grad_norm": 0.6334, "lr": 0.0001177, "router_trainable": true, "sec_per_step": 1.48, "probe_sec": 0.0767, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04279, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.03526, "n_shadow": 3, "mean_delta_shadow": 0.04037, "n_split": 1, "mean_delta_split": 0.00232, "n_merge": 2, "mean_delta_merge": -0.62288, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 1.87128, "trunc_abs_frac": 2.15704, "mean_delta_full": -0.148, "read_hit_rate": {"surface": 0.56183, "split": 0.54413, "merge": 0.13537, "skel": 0.40027}, "mean_gate": {"surface": 0.19125, "split": 0.06536, "merge": 0.11948, "skel": 0.1258}, "n_reads": 13286, "n_cand": 30635, "read_suppressed_frac": 0.1, "dropout_reads_logged": 13286, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1520, "admitted_surface": 54796, "staged_surface": 56166, "admitted_split": 6481, "staged_split": 44510, "admitted_merge": 3614, "staged_merge": 115914, "admitted_skel": 645, "staged_skel": 45554, "mean_count": 565.6256103515625, "n_util_obs": 430, "util_count_spearman": 0.06619068084562069}, "probe_overhead_sec_total": 778.099, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008427, "alpha": 0.03, "t": 6756.7} |
| {"event": "consolidate", "step": 3850, "retired": 1, "split_retired": 0, "merge_admitted": 0, "admitted": 1, "swapped": 31, "uncertain_staged": 262125, "t": 6812.7} |
| {"event": "train", "step": 3850, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.36569, "bpb": 1.97027, "chunk_loss": 0.030351, "F": 0.20105, "G": 0.20389, "chunks": 205.88, "bytes_per_chunk": 4.974, "boundary_entropy": 0.08667, "router_noise": 0.0, "grad_norm": 0.7544, "lr": 0.0001142, "router_trainable": true, "sec_per_step": 1.162, "probe_sec": 0.0832, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0524, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 3, "mean_delta_loo": 0.04431, "n_shadow": 5, "mean_delta_shadow": 0.01452, "n_split": 2, "mean_delta_split": -0.10089, "n_merge": 1, "mean_delta_merge": -0.07681, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.36143, "trunc_abs_frac": 0.36143, "mean_delta_full": -0.21252, "read_hit_rate": {"surface": 0.52837, "split": 0.55611, "merge": 0.11987, "skel": 0.36034}, "mean_gate": {"surface": 0.19089, "split": 0.06806, "merge": 0.11477, "skel": 0.12864}, "n_reads": 12557, "n_cand": 29841, "read_suppressed_frac": 0.098, "dropout_reads_logged": 12557, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1540, "admitted_surface": 54767, "staged_surface": 56212, "admitted_split": 6503, "staged_split": 44726, "admitted_merge": 3619, "staged_merge": 115702, "admitted_skel": 647, "staged_skel": 45504, "mean_count": 575.7031860351562, "n_util_obs": 432, "util_count_spearman": 0.05803351833321423}, "probe_overhead_sec_total": 783.971, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009144, "alpha": 0.03, "t": 6814.8} |
| {"event": "consolidate", "step": 3900, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 83, "uncertain_staged": 262125, "t": 6868.0} |
| {"event": "train", "step": 3900, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.39834, "bpb": 2.01738, "chunk_loss": 0.030506, "F": 0.20798, "G": 0.21043, "chunks": 212.97, "bytes_per_chunk": 4.808, "boundary_entropy": 0.08795, "router_noise": 0.0, "grad_norm": 0.6314, "lr": 0.0001108, "router_trainable": true, "sec_per_step": 1.075, "probe_sec": 0.0786, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04656, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": -0.16616, "n_shadow": 4, "mean_delta_shadow": 0.00414, "n_split": 1, "mean_delta_split": -0.02344, "n_merge": 1, "mean_delta_merge": -0.10388, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 1.35069, "trunc_abs_frac": 0.56833, "mean_delta_full": -0.30249, "read_hit_rate": {"surface": 0.54725, "split": 0.55543, "merge": 0.13302, "skel": 0.37075}, "mean_gate": {"surface": 0.19735, "split": 0.06999, "merge": 0.12789, "skel": 0.12783}, "n_reads": 13185, "n_cand": 30714, "read_suppressed_frac": 0.0957, "dropout_reads_logged": 13185, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1552, "admitted_surface": 54688, "staged_surface": 56121, "admitted_split": 6560, "staged_split": 44660, "admitted_merge": 3624, "staged_merge": 115686, "admitted_skel": 664, "staged_skel": 45677, "mean_count": 586.1731567382812, "n_util_obs": 441, "util_count_spearman": 0.05997720976422775}, "probe_overhead_sec_total": 790.309, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009206, "alpha": 0.03, "t": 6868.5} |
| {"event": "consolidate", "step": 3950, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 42, "uncertain_staged": 262125, "t": 6936.5} |
| {"event": "train", "step": 3950, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.42874, "bpb": 2.06123, "chunk_loss": 0.030638, "F": 0.21582, "G": 0.21751, "chunks": 221.0, "bytes_per_chunk": 4.633, "boundary_entropy": 0.08238, "router_noise": 0.0, "grad_norm": 0.666, "lr": 0.0001074, "router_trainable": true, "sec_per_step": 1.39, "probe_sec": 0.1354, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08622, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 11, "mean_delta_loo": -0.05836, "n_shadow": 3, "mean_delta_shadow": 0.00647, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -23.93379, "trunc_abs_frac": 0.82252, "mean_delta_full": 0.00668, "read_hit_rate": {"surface": 0.54844, "split": 0.55578, "merge": 0.14326, "skel": 0.38164}, "mean_gate": {"surface": 0.17935, "split": 0.06055, "merge": 0.11925, "skel": 0.11835}, "n_reads": 13630, "n_cand": 31436, "read_suppressed_frac": 0.0984, "dropout_reads_logged": 13630, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1568, "admitted_surface": 54647, "staged_surface": 56233, "admitted_split": 6584, "staged_split": 44567, "admitted_merge": 3629, "staged_merge": 115629, "admitted_skel": 676, "staged_skel": 45715, "mean_count": 595.932373046875, "n_util_obs": 447, "util_count_spearman": 0.06805802483525306}, "probe_overhead_sec_total": 796.678, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008704, "alpha": 0.03, "t": 6938.0} |
| {"event": "consolidate", "step": 4000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 62, "uncertain_staged": 262125, "t": 7018.6} |
| {"event": "probe", "step": 4000, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3425.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.196805, "cps": 26.757812, "mean_p": 0.198562, "boundary_entropy": 0.084247, "bpb": 1.929727}, "tr": {"chunks": 3735.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.206319, "cps": 29.179688, "mean_p": 0.196174, "boundary_entropy": 0.102864, "bpb": 2.089805}, "fi": {"chunks": 4904.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.260699, "cps": 38.3125, "mean_p": 0.254955, "boundary_entropy": 0.091307, "bpb": 2.196161}}, "bpb": {"en": 1.92973, "tr": 2.08981, "fi": 2.19616}, "gini_cpc": 0.064167, "gini_cps": 0.081731, "bpic": 4.702586, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0483434814006307, "fi": 1.3246529849168078}, "inventory": {"size": 327618, "admitted": 65536, "staged": 262082, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1576, "admitted_surface": 54588, "staged_surface": 56214, "admitted_split": 6618, "staged_split": 44347, "admitted_merge": 3634, "staged_merge": 115945, "admitted_skel": 696, "staged_skel": 45576, "mean_count": 606.2931518554688, "n_util_obs": 457, "util_count_spearman": 0.06255391014980284}, "t": 7020.4} |
| {"event": "eval", "step": 4000, "phase": 2, "bpb": {"en": 1.94397, "tr": 2.09792, "fi": 2.2046}, "bytes_per_chunk": {"en": 4.754, "tr": 5.412, "fi": 4.103}, "bpb_mean": 2.08216, "frozen_inventory": true, "t": 7023.4} |
| {"event": "train", "step": 4000, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.44265, "bpb": 2.0813, "chunk_loss": 0.030605, "F": 0.21594, "G": 0.21655, "chunks": 221.12, "bytes_per_chunk": 4.631, "boundary_entropy": 0.08491, "router_noise": 0.0, "grad_norm": 0.6916, "lr": 0.000104, "router_trainable": true, "sec_per_step": 1.807, "probe_sec": 0.0721, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04694, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": -0.0064, "n_shadow": 3, "mean_delta_shadow": -0.07331, "n_split": 1, "mean_delta_split": -0.05281, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -0.09572, "trunc_abs_frac": 0.09787, "mean_delta_full": -0.5035, "read_hit_rate": {"surface": 0.56104, "split": 0.55627, "merge": 0.14119, "skel": 0.38654}, "mean_gate": {"surface": 0.18636, "split": 0.06295, "merge": 0.12455, "skel": 0.12454}, "n_reads": 13674, "n_cand": 31340, "read_suppressed_frac": 0.1002, "dropout_reads_logged": 13674, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1576, "admitted_surface": 54588, "staged_surface": 56195, "admitted_split": 6618, "staged_split": 44406, "admitted_merge": 3634, "staged_merge": 115956, "admitted_skel": 696, "staged_skel": 45587, "mean_count": 606.5018310546875, "n_util_obs": 457, "util_count_spearman": 0.063136598264861}, "probe_overhead_sec_total": 805.025, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009166, "alpha": 0.03, "t": 7028.4} |
| {"event": "consolidate", "step": 4050, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 33, "uncertain_staged": 262125, "t": 7096.4} |
| {"event": "train", "step": 4050, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.36312, "bpb": 1.96656, "chunk_loss": 0.030561, "F": 0.21075, "G": 0.213, "chunks": 215.81, "bytes_per_chunk": 4.745, "boundary_entropy": 0.07876, "router_noise": 0.0, "grad_norm": 0.6514, "lr": 0.0001007, "router_trainable": true, "sec_per_step": 1.378, "probe_sec": 0.1018, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05157, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.1069, "n_shadow": 4, "mean_delta_shadow": -0.01446, "n_split": 1, "mean_delta_split": 0.01947, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.4482, "trunc_abs_frac": 3.24045, "mean_delta_full": 0.11538, "read_hit_rate": {"surface": 0.56008, "split": 0.58043, "merge": 0.14543, "skel": 0.38313}, "mean_gate": {"surface": 0.17569, "split": 0.06062, "merge": 0.11764, "skel": 0.11828}, "n_reads": 13850, "n_cand": 31003, "read_suppressed_frac": 0.0981, "dropout_reads_logged": 13850, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1585, "admitted_surface": 54559, "staged_surface": 56188, "admitted_split": 6643, "staged_split": 44285, "admitted_merge": 3634, "staged_merge": 115931, "admitted_skel": 700, "staged_skel": 45740, "mean_count": 616.6220092773438, "n_util_obs": 459, "util_count_spearman": 0.05754568951163709}, "probe_overhead_sec_total": 810.8, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008596, "alpha": 0.03, "t": 7097.3} |
| {"event": "consolidate", "step": 4100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 14, "uncertain_staged": 262125, "t": 7167.3} |
| {"event": "train", "step": 4100, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.43671, "bpb": 2.07274, "chunk_loss": 0.030727, "F": 0.21802, "G": 0.22094, "chunks": 223.25, "bytes_per_chunk": 4.587, "boundary_entropy": 0.08309, "router_noise": 0.0, "grad_norm": 0.684, "lr": 9.75e-05, "router_trainable": true, "sec_per_step": 1.424, "probe_sec": 0.1265, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08422, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": 0.11391, "n_shadow": 2, "mean_delta_shadow": -0.07767, "n_split": 1, "mean_delta_split": 0.47351, "n_merge": 3, "mean_delta_merge": 0.11092, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 1.23684, "trunc_abs_frac": 0.94501, "mean_delta_full": -0.08118, "read_hit_rate": {"surface": 0.56032, "split": 0.58817, "merge": 0.13814, "skel": 0.39386}, "mean_gate": {"surface": 0.17473, "split": 0.05898, "merge": 0.12523, "skel": 0.10708}, "n_reads": 14415, "n_cand": 32058, "read_suppressed_frac": 0.1035, "dropout_reads_logged": 14415, "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": 54547, "staged_surface": 56094, "admitted_split": 6648, "staged_split": 44398, "admitted_merge": 3638, "staged_merge": 116210, "admitted_skel": 703, "staged_skel": 45442, "mean_count": 627.2415161132812, "n_util_obs": 465, "util_count_spearman": 0.052240509610120875}, "probe_overhead_sec_total": 816.845, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009647, "alpha": 0.03, "t": 7168.5} |
| {"event": "consolidate", "step": 4150, "retired": 1, "split_retired": 3, "merge_admitted": 0, "admitted": 4, "swapped": 38, "uncertain_staged": 262125, "t": 7234.4} |
| {"event": "train", "step": 4150, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.3844, "bpb": 1.99727, "chunk_loss": 0.030876, "F": 0.22562, "G": 0.22586, "chunks": 231.03, "bytes_per_chunk": 4.432, "boundary_entropy": 0.07894, "router_noise": 0.0, "grad_norm": 0.6526, "lr": 9.43e-05, "router_trainable": true, "sec_per_step": 1.344, "probe_sec": 0.1032, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05006, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.08283, "n_shadow": 5, "mean_delta_shadow": -0.02119, "n_split": 1, "mean_delta_split": 0.1257, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.28413, "trunc_abs_frac": 0.28413, "mean_delta_full": 0.5177, "read_hit_rate": {"surface": 0.56854, "split": 0.58625, "merge": 0.15241, "skel": 0.41304}, "mean_gate": {"surface": 0.17055, "split": 0.0562, "merge": 0.12323, "skel": 0.11041}, "n_reads": 15077, "n_cand": 33032, "read_suppressed_frac": 0.0966, "dropout_reads_logged": 15077, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1610, "admitted_surface": 54507, "staged_surface": 56181, "admitted_split": 6674, "staged_split": 44352, "admitted_merge": 3647, "staged_merge": 116126, "admitted_skel": 708, "staged_skel": 45485, "mean_count": 635.970947265625, "n_util_obs": 465, "util_count_spearman": 0.04861893665287782}, "probe_overhead_sec_total": 823.887, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.00854, "alpha": 0.03, "t": 7235.7} |
| {"event": "consolidate", "step": 4200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 23, "uncertain_staged": 262125, "t": 7307.9} |
| {"event": "probe", "step": 4200, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3433.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.197265, "cps": 26.820312, "mean_p": 0.199114, "boundary_entropy": 0.081048, "bpb": 1.919575}, "tr": {"chunks": 3739.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.20654, "cps": 29.210938, "mean_p": 0.195215, "boundary_entropy": 0.099394, "bpb": 2.067472}, "fi": {"chunks": 4902.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.260592, "cps": 38.296875, "mean_p": 0.254282, "boundary_entropy": 0.090438, "bpb": 2.171878}}, "bpb": {"en": 1.91958, "tr": 2.06747, "fi": 2.17188}, "gini_cpc": 0.063544, "gini_cps": 0.081111, "bpic": 4.698691, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0470206097583916, "fi": 1.3210271403972889}, "inventory": {"size": 327657, "admitted": 65536, "staged": 262121, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1628, "admitted_surface": 54489, "staged_surface": 56321, "admitted_split": 6683, "staged_split": 44389, "admitted_merge": 3649, "staged_merge": 116163, "admitted_skel": 715, "staged_skel": 45248, "mean_count": 646.4303588867188, "n_util_obs": 473, "util_count_spearman": 0.04569568447097132}, "t": 7310.1} |
| {"event": "eval", "step": 4200, "phase": 2, "bpb": {"en": 1.93255, "tr": 2.07821, "fi": 2.17986}, "bytes_per_chunk": {"en": 4.75, "tr": 5.388, "fi": 4.095}, "bpb_mean": 2.06354, "frozen_inventory": true, "t": 7313.4} |
| {"event": "phase_transition", "step": 4200, "phase": 3, "objective": "global", "router_trainable": true, "encoder_trainable": true, "frozen_router_params": 0, "t": 7319.7} |
| {"event": "train", "step": 4200, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.37931, "bpb": 1.98992, "chunk_loss": 0.030432, "F": 0.20764, "G": 0.20927, "chunks": 212.62, "bytes_per_chunk": 4.816, "boundary_entropy": 0.08016, "router_noise": 0.0, "grad_norm": 0.6467, "lr": 9.12e-05, "router_trainable": true, "sec_per_step": 1.706, "probe_sec": 0.1094, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05534, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": -0.22411, "n_shadow": 6, "mean_delta_shadow": -0.00806, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": -0.07663, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 3.46506, "trunc_abs_frac": 0.33281, "mean_delta_full": -0.02533, "read_hit_rate": {"surface": 0.55242, "split": 0.56503, "merge": 0.13858, "skel": 0.39843}, "mean_gate": {"surface": 0.17174, "split": 0.05841, "merge": 0.11327, "skel": 0.11679}, "n_reads": 13478, "n_cand": 30636, "read_suppressed_frac": 0.0991, "dropout_reads_logged": 13478, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1630, "admitted_surface": 54489, "staged_surface": 56315, "admitted_split": 6683, "staged_split": 44411, "admitted_merge": 3649, "staged_merge": 116087, "admitted_skel": 715, "staged_skel": 45331, "mean_count": 646.6360473632812, "n_util_obs": 473, "util_count_spearman": 0.04563893079993383}, "probe_overhead_sec_total": 831.281, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008475, "alpha": 0.03, "t": 7321.0} |
| {"event": "consolidate", "step": 4250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 21, "uncertain_staged": 262124, "t": 7392.0} |
| {"event": "train", "step": 4250, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.37667, "bpb": 1.98612, "chunk_loss": 0.030462, "F": 0.20325, "G": 0.20672, "chunks": 208.12, "bytes_per_chunk": 4.92, "boundary_entropy": 0.08343, "router_noise": 0.0, "grad_norm": 0.7017, "lr": 8.81e-05, "router_trainable": true, "sec_per_step": 1.456, "probe_sec": 0.1586, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10535, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": -0.11964, "n_shadow": 5, "mean_delta_shadow": -0.01417, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.39349, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -0.10191, "trunc_abs_frac": 0.96763, "mean_delta_full": 0.14034, "read_hit_rate": {"surface": 0.54119, "split": 0.57354, "merge": 0.1275, "skel": 0.38038}, "mean_gate": {"surface": 0.18189, "split": 0.06278, "merge": 0.12102, "skel": 0.12128}, "n_reads": 13129, "n_cand": 30146, "read_suppressed_frac": 0.1002, "dropout_reads_logged": 13129, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1667, "admitted_surface": 54471, "staged_surface": 56499, "admitted_split": 6692, "staged_split": 44182, "admitted_merge": 3657, "staged_merge": 115944, "admitted_skel": 716, "staged_skel": 45519, "mean_count": 657.3387451171875, "n_util_obs": 480, "util_count_spearman": 0.032158669095749916}, "probe_overhead_sec_total": 837.113, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009136, "alpha": 0.03, "t": 7393.8} |
| {"event": "consolidate", "step": 4300, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 22, "uncertain_staged": 262125, "t": 7451.8} |
| {"event": "train", "step": 4300, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.46123, "bpb": 2.1081, "chunk_loss": 0.030636, "F": 0.21286, "G": 0.21591, "chunks": 217.97, "bytes_per_chunk": 4.698, "boundary_entropy": 0.08408, "router_noise": 0.0, "grad_norm": 0.7876, "lr": 8.5e-05, "router_trainable": true, "sec_per_step": 1.177, "probe_sec": 0.1253, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07969, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.04382, "n_shadow": 6, "mean_delta_shadow": -0.07544, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.97638, "trunc_abs_frac": 1.17235, "mean_delta_full": -0.2375, "read_hit_rate": {"surface": 0.54386, "split": 0.57439, "merge": 0.12878, "skel": 0.37877}, "mean_gate": {"surface": 0.19192, "split": 0.06588, "merge": 0.13028, "skel": 0.12593}, "n_reads": 13559, "n_cand": 31133, "read_suppressed_frac": 0.1029, "dropout_reads_logged": 13559, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1685, "admitted_surface": 54449, "staged_surface": 56371, "admitted_split": 6707, "staged_split": 44293, "admitted_merge": 3661, "staged_merge": 115717, "admitted_skel": 719, "staged_skel": 45763, "mean_count": 666.851318359375, "n_util_obs": 483, "util_count_spearman": 0.03446245421381957}, "probe_overhead_sec_total": 843.669, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008804, "alpha": 0.03, "t": 7452.7} |
| {"event": "consolidate", "step": 4350, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 55, "uncertain_staged": 262125, "t": 7526.1} |
| {"event": "train", "step": 4350, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32073, "bpb": 1.90542, "chunk_loss": 0.030548, "F": 0.21103, "G": 0.21177, "chunks": 216.09, "bytes_per_chunk": 4.739, "boundary_entropy": 0.08177, "router_noise": 0.0, "grad_norm": 0.6979, "lr": 8.21e-05, "router_trainable": true, "sec_per_step": 1.498, "probe_sec": 0.1201, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05898, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 11, "mean_delta_loo": 0.00249, "n_shadow": 3, "mean_delta_shadow": 0.04116, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.23886, "trunc_abs_frac": 1.32574, "mean_delta_full": -0.10345, "read_hit_rate": {"surface": 0.56008, "split": 0.57837, "merge": 0.1429, "skel": 0.39865}, "mean_gate": {"surface": 0.16825, "split": 0.05614, "merge": 0.11485, "skel": 0.115}, "n_reads": 13950, "n_cand": 31148, "read_suppressed_frac": 0.0991, "dropout_reads_logged": 13950, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1709, "admitted_surface": 54398, "staged_surface": 56263, "admitted_split": 6739, "staged_split": 44447, "admitted_merge": 3671, "staged_merge": 115980, "admitted_skel": 728, "staged_skel": 45454, "mean_count": 676.7269897460938, "n_util_obs": 492, "util_count_spearman": 0.03380449590208202}, "probe_overhead_sec_total": 851.37, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.00874, "alpha": 0.03, "t": 7527.5} |
| {"event": "consolidate", "step": 4400, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 37, "uncertain_staged": 262125, "t": 7595.0} |
| {"event": "probe", "step": 4400, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3426.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.196863, "cps": 26.765625, "mean_p": 0.198639, "boundary_entropy": 0.080732, "bpb": 1.906628}, "tr": {"chunks": 3753.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.207314, "cps": 29.320312, "mean_p": 0.195465, "boundary_entropy": 0.099465, "bpb": 2.052375}, "fi": {"chunks": 4896.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.260273, "cps": 38.25, "mean_p": 0.253747, "boundary_entropy": 0.088373, "bpb": 2.154811}}, "bpb": {"en": 1.90663, "tr": 2.05237, "fi": 2.15481}, "gini_cpc": 0.063622, "gini_cps": 0.081159, "bpic": 4.698302, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0530882681151337, "fi": 1.3221060338340247}, "inventory": {"size": 327641, "admitted": 65536, "staged": 262105, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1719, "admitted_surface": 54361, "staged_surface": 56357, "admitted_split": 6769, "staged_split": 44125, "admitted_merge": 3674, "staged_merge": 116083, "admitted_skel": 732, "staged_skel": 45540, "mean_count": 685.9444580078125, "n_util_obs": 501, "util_count_spearman": 0.03767970305697242}, "t": 7596.8} |
| {"event": "eval", "step": 4400, "phase": 3, "bpb": {"en": 1.92142, "tr": 2.06257, "fi": 2.16306}, "bytes_per_chunk": {"en": 4.767, "tr": 5.383, "fi": 4.089}, "bpb_mean": 2.04902, "frozen_inventory": true, "t": 7599.6} |
| {"event": "train", "step": 4400, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.43129, "bpb": 2.06491, "chunk_loss": 0.030795, "F": 0.21716, "G": 0.21903, "chunks": 222.38, "bytes_per_chunk": 4.605, "boundary_entropy": 0.08099, "router_noise": 0.0, "grad_norm": 0.7048, "lr": 7.92e-05, "router_trainable": true, "sec_per_step": 1.581, "probe_sec": 0.2157, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06427, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.03528, "n_shadow": 2, "mean_delta_shadow": -0.02762, "n_split": 3, "mean_delta_split": -0.05431, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -0.37385, "trunc_abs_frac": 0.97272, "mean_delta_full": 0.04854, "read_hit_rate": {"surface": 0.55251, "split": 0.58053, "merge": 0.13727, "skel": 0.39175}, "mean_gate": {"surface": 0.17857, "split": 0.06138, "merge": 0.12421, "skel": 0.1151}, "n_reads": 14096, "n_cand": 31750, "read_suppressed_frac": 0.1003, "dropout_reads_logged": 14096, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1720, "admitted_surface": 54361, "staged_surface": 56400, "admitted_split": 6769, "staged_split": 44077, "admitted_merge": 3674, "staged_merge": 116106, "admitted_skel": 732, "staged_skel": 45561, "mean_count": 686.1596069335938, "n_util_obs": 502, "util_count_spearman": 0.03467502829481588}, "probe_overhead_sec_total": 858.032, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.007866, "alpha": 0.03, "t": 7606.6} |
| {"event": "consolidate", "step": 4450, "retired": 0, "split_retired": 3, "merge_admitted": 0, "admitted": 3, "swapped": 60, "uncertain_staged": 262125, "t": 7677.5} |
| {"event": "train", "step": 4450, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.371, "bpb": 1.97794, "chunk_loss": 0.030655, "F": 0.21478, "G": 0.21636, "chunks": 219.94, "bytes_per_chunk": 4.656, "boundary_entropy": 0.0784, "router_noise": 0.0, "grad_norm": 0.6571, "lr": 7.63e-05, "router_trainable": true, "sec_per_step": 1.446, "probe_sec": 0.0881, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05166, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 3, "mean_delta_loo": 0.09371, "n_shadow": 4, "mean_delta_shadow": 0.02726, "n_split": 2, "mean_delta_split": -0.03174, "n_merge": 2, "mean_delta_merge": 0.30319, "n_dropped_unpaired": 0, "n_full": 0, "read_hit_rate": {"surface": 0.56309, "split": 0.57559, "merge": 0.14654, "skel": 0.39578}, "mean_gate": {"surface": 0.17206, "split": 0.05877, "merge": 0.11751, "skel": 0.11703}, "n_reads": 14017, "n_cand": 31347, "read_suppressed_frac": 0.0977, "dropout_reads_logged": 14017, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1729, "admitted_surface": 54301, "staged_surface": 56467, "admitted_split": 6815, "staged_split": 44286, "admitted_merge": 3682, "staged_merge": 115667, "admitted_skel": 738, "staged_skel": 45724, "mean_count": 695.861572265625, "n_util_obs": 506, "util_count_spearman": 0.03951545656237865}, "probe_overhead_sec_total": 863.803, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.00829, "alpha": 0.03, "t": 7678.9} |
| {"event": "consolidate", "step": 4500, "retired": 0, "split_retired": 4, "merge_admitted": 0, "admitted": 4, "swapped": 52, "uncertain_staged": 262125, "t": 7753.1} |
| {"event": "train", "step": 4500, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.3817, "bpb": 1.99337, "chunk_loss": 0.030675, "F": 0.21619, "G": 0.21768, "chunks": 221.38, "bytes_per_chunk": 4.626, "boundary_entropy": 0.08251, "router_noise": 0.0, "grad_norm": 0.7132, "lr": 7.36e-05, "router_trainable": true, "sec_per_step": 1.494, "probe_sec": 0.0678, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04658, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.07217, "n_shadow": 3, "mean_delta_shadow": 0.07609, "n_split": 2, "mean_delta_split": -0.24179, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.40131, "trunc_abs_frac": 0.58818, "mean_delta_full": 0.22504, "read_hit_rate": {"surface": 0.54396, "split": 0.58072, "merge": 0.13048, "skel": 0.40301}, "mean_gate": {"surface": 0.17596, "split": 0.0581, "merge": 0.12475, "skel": 0.11599}, "n_reads": 14165, "n_cand": 31966, "read_suppressed_frac": 0.0957, "dropout_reads_logged": 14165, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1737, "admitted_surface": 54248, "staged_surface": 56324, "admitted_split": 6852, "staged_split": 44019, "admitted_merge": 3689, "staged_merge": 116063, "admitted_skel": 747, "staged_skel": 45738, "mean_count": 704.9398193359375, "n_util_obs": 507, "util_count_spearman": 0.020179457669449955}, "probe_overhead_sec_total": 870.855, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008874, "alpha": 0.03, "t": 7753.6} |
| {"event": "consolidate", "step": 4550, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 78, "uncertain_staged": 262125, "t": 7821.0} |
| {"event": "train", "step": 4550, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.36063, "bpb": 1.96298, "chunk_loss": 0.030619, "F": 0.21454, "G": 0.216, "chunks": 219.69, "bytes_per_chunk": 4.661, "boundary_entropy": 0.07941, "router_noise": 0.0, "grad_norm": 0.6453, "lr": 7.09e-05, "router_trainable": true, "sec_per_step": 1.375, "probe_sec": 0.1217, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07004, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": -0.02548, "n_shadow": 5, "mean_delta_shadow": 0.01598, "n_split": 2, "mean_delta_split": 0.05499, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 3.22839, "trunc_abs_frac": 0.82557, "mean_delta_full": 0.04944, "read_hit_rate": {"surface": 0.56402, "split": 0.59062, "merge": 0.14312, "skel": 0.40616}, "mean_gate": {"surface": 0.17659, "split": 0.05876, "merge": 0.12348, "skel": 0.11703}, "n_reads": 14334, "n_cand": 31545, "read_suppressed_frac": 0.0987, "dropout_reads_logged": 14334, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1746, "admitted_surface": 54177, "staged_surface": 56522, "admitted_split": 6899, "staged_split": 43933, "admitted_merge": 3704, "staged_merge": 115869, "admitted_skel": 756, "staged_skel": 45820, "mean_count": 715.3023071289062, "n_util_obs": 516, "util_count_spearman": -0.00851574811600324}, "probe_overhead_sec_total": 876.917, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008598, "alpha": 0.03, "t": 7822.3} |
| {"event": "consolidate", "step": 4600, "retired": 0, "split_retired": 3, "merge_admitted": 0, "admitted": 3, "swapped": 83, "uncertain_staged": 262125, "t": 7892.7} |
| {"event": "probe", "step": 4600, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3423.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.19669, "cps": 26.742188, "mean_p": 0.199121, "boundary_entropy": 0.078923, "bpb": 1.895673}, "tr": {"chunks": 3719.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.205436, "cps": 29.054688, "mean_p": 0.194003, "boundary_entropy": 0.1014, "bpb": 2.038801}, "fi": {"chunks": 4904.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.260699, "cps": 38.3125, "mean_p": 0.254105, "boundary_entropy": 0.086698, "bpb": 2.144833}}, "bpb": {"en": 1.89567, "tr": 2.0388, "fi": 2.14483}, "gini_cpc": 0.064379, "gini_cps": 0.081964, "bpic": 4.709613, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0444624907343183, "fi": 1.3254269568624208}, "inventory": {"size": 327594, "admitted": 65536, "staged": 262058, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1755, "admitted_surface": 54103, "staged_surface": 56420, "admitted_split": 6949, "staged_split": 43956, "admitted_merge": 3715, "staged_merge": 116032, "admitted_skel": 769, "staged_skel": 45650, "mean_count": 724.41015625, "n_util_obs": 518, "util_count_spearman": -0.003799662144333425}, "t": 7894.6} |
| {"event": "eval", "step": 4600, "phase": 3, "bpb": {"en": 1.9097, "tr": 2.04854, "fi": 2.15339}, "bytes_per_chunk": {"en": 4.769, "tr": 5.396, "fi": 4.084}, "bpb_mean": 2.03721, "frozen_inventory": true, "t": 7898.0} |
| {"event": "train", "step": 4600, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.41591, "bpb": 2.04272, "chunk_loss": 0.030522, "F": 0.20779, "G": 0.20986, "chunks": 212.78, "bytes_per_chunk": 4.812, "boundary_entropy": 0.08115, "router_noise": 0.0, "grad_norm": 0.7124, "lr": 6.82e-05, "router_trainable": true, "sec_per_step": 1.639, "probe_sec": 0.1402, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07677, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": -0.05688, "n_shadow": 1, "mean_delta_shadow": -0.21069, "n_split": 1, "mean_delta_split": 0.18622, "n_merge": 2, "mean_delta_merge": -0.9762, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.76053, "trunc_abs_frac": 4.01716, "mean_delta_full": -0.19706, "read_hit_rate": {"surface": 0.53372, "split": 0.58474, "merge": 0.12943, "skel": 0.38642}, "mean_gate": {"surface": 0.18093, "split": 0.06162, "merge": 0.11982, "skel": 0.12468}, "n_reads": 13528, "n_cand": 30776, "read_suppressed_frac": 0.1002, "dropout_reads_logged": 13528, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1756, "admitted_surface": 54103, "staged_surface": 56440, "admitted_split": 6949, "staged_split": 44023, "admitted_merge": 3715, "staged_merge": 116023, "admitted_skel": 769, "staged_skel": 45658, "mean_count": 724.6165771484375, "n_util_obs": 518, "util_count_spearman": -0.004169181662119089}, "probe_overhead_sec_total": 883.13, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008638, "alpha": 0.03, "t": 7904.3} |
| {"event": "consolidate", "step": 4650, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 285, "uncertain_staged": 262125, "t": 7961.2} |
| {"event": "train", "step": 4650, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.33612, "bpb": 1.92762, "chunk_loss": 0.030561, "F": 0.21231, "G": 0.21306, "chunks": 217.41, "bytes_per_chunk": 4.71, "boundary_entropy": 0.08162, "router_noise": 0.0, "grad_norm": 0.6512, "lr": 6.57e-05, "router_trainable": true, "sec_per_step": 1.167, "probe_sec": 0.1117, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06017, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.03981, "n_shadow": 3, "mean_delta_shadow": 0.1003, "n_split": 2, "mean_delta_split": 0.10034, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -0.97443, "trunc_abs_frac": 0.88561, "mean_delta_full": 0.01432, "read_hit_rate": {"surface": 0.56188, "split": 0.60091, "merge": 0.13942, "skel": 0.42296}, "mean_gate": {"surface": 0.17682, "split": 0.06062, "merge": 0.12249, "skel": 0.11989}, "n_reads": 14392, "n_cand": 31291, "read_suppressed_frac": 0.0957, "dropout_reads_logged": 14392, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1777, "admitted_surface": 53838, "staged_surface": 56625, "admitted_split": 7046, "staged_split": 43922, "admitted_merge": 3751, "staged_merge": 115886, "admitted_skel": 901, "staged_skel": 45711, "mean_count": 735.33203125, "n_util_obs": 518, "util_count_spearman": -0.001697286174609523}, "probe_overhead_sec_total": 888.463, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008259, "alpha": 0.03, "t": 7962.6} |
| {"event": "consolidate", "step": 4700, "retired": 1, "split_retired": 0, "merge_admitted": 0, "admitted": 1, "swapped": 617, "uncertain_staged": 262125, "t": 8035.9} |
| {"event": "train", "step": 4700, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35245, "bpb": 1.95118, "chunk_loss": 0.03055, "F": 0.20984, "G": 0.21105, "chunks": 214.88, "bytes_per_chunk": 4.766, "boundary_entropy": 0.07981, "router_noise": 0.0, "grad_norm": 0.6793, "lr": 6.32e-05, "router_trainable": true, "sec_per_step": 1.487, "probe_sec": 0.0802, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04632, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": -0.14385, "n_shadow": 1, "mean_delta_shadow": -0.06967, "n_split": 1, "mean_delta_split": -0.07678, "n_merge": 1, "mean_delta_merge": -0.1248, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 1.22666, "trunc_abs_frac": 0.66078, "mean_delta_full": -0.33185, "read_hit_rate": {"surface": 0.56049, "split": 0.60007, "merge": 0.13183, "skel": 0.43409}, "mean_gate": {"surface": 0.17279, "split": 0.05805, "merge": 0.11775, "skel": 0.11431}, "n_reads": 14232, "n_cand": 30951, "read_suppressed_frac": 0.0982, "dropout_reads_logged": 14232, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1802, "admitted_surface": 53237, "staged_surface": 56761, "admitted_split": 7225, "staged_split": 43638, "admitted_merge": 3860, "staged_merge": 116144, "admitted_skel": 1214, "staged_skel": 45601, "mean_count": 746.305419921875, "n_util_obs": 522, "util_count_spearman": 0.001889267025391317}, "probe_overhead_sec_total": 894.859, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008839, "alpha": 0.03, "t": 8037.0} |
| {"event": "consolidate", "step": 4750, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 424, "uncertain_staged": 262125, "t": 8095.3} |
| {"event": "train", "step": 4750, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.31774, "bpb": 1.90109, "chunk_loss": 0.030549, "F": 0.21179, "G": 0.21303, "chunks": 216.88, "bytes_per_chunk": 4.722, "boundary_entropy": 0.07965, "router_noise": 0.0, "grad_norm": 0.6671, "lr": 6.08e-05, "router_trainable": true, "sec_per_step": 1.192, "probe_sec": 0.1829, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.1336, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": 0.02798, "n_shadow": 6, "mean_delta_shadow": -0.09436, "n_split": 1, "mean_delta_split": -0.15033, "n_merge": 1, "mean_delta_merge": -0.28125, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.36034, "trunc_abs_frac": 0.12699, "mean_delta_full": -0.01639, "read_hit_rate": {"surface": 0.56659, "split": 0.60615, "merge": 0.15445, "skel": 0.46545}, "mean_gate": {"surface": 0.16894, "split": 0.05652, "merge": 0.10874, "skel": 0.1113}, "n_reads": 14747, "n_cand": 31140, "read_suppressed_frac": 0.1037, "dropout_reads_logged": 14747, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1818, "admitted_surface": 52827, "staged_surface": 56716, "admitted_split": 7324, "staged_split": 43537, "admitted_merge": 3918, "staged_merge": 116683, "admitted_skel": 1467, "staged_skel": 45208, "mean_count": 756.7747802734375, "n_util_obs": 527, "util_count_spearman": -0.004165775899794107}, "probe_overhead_sec_total": 902.05, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009167, "alpha": 0.03, "t": 8096.6} |
| {"event": "consolidate", "step": 4800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 160, "uncertain_staged": 262125, "t": 8154.4} |
| {"event": "probe", "step": 4800, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3433.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.197265, "cps": 26.820312, "mean_p": 0.199138, "boundary_entropy": 0.074319, "bpb": 1.884701}, "tr": {"chunks": 3769.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.208198, "cps": 29.445312, "mean_p": 0.194924, "boundary_entropy": 0.095932, "bpb": 2.025964}, "fi": {"chunks": 4904.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.260699, "cps": 38.3125, "mean_p": 0.254414, "boundary_entropy": 0.081698, "bpb": 2.130012}}, "bpb": {"en": 1.8847, "tr": 2.02596, "fi": 2.13001}, "gini_cpc": 0.063482, "gini_cps": 0.081007, "bpic": 4.686271, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0554214170043803, "fi": 1.3215661151587728}, "inventory": {"size": 327520, "admitted": 65536, "staged": 261984, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1851, "admitted_surface": 52677, "staged_surface": 56766, "admitted_split": 7352, "staged_split": 43647, "admitted_merge": 3950, "staged_merge": 116494, "admitted_skel": 1557, "staged_skel": 45077, "mean_count": 767.7769775390625, "n_util_obs": 530, "util_count_spearman": -0.013392997029933431}, "t": 8156.8} |
| {"event": "eval", "step": 4800, "phase": 3, "bpb": {"en": 1.89879, "tr": 2.03496, "fi": 2.1382}, "bytes_per_chunk": {"en": 4.743, "tr": 5.373, "fi": 4.078}, "bpb_mean": 2.02398, "frozen_inventory": true, "t": 8160.1} |
| {"event": "train", "step": 4800, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38166, "bpb": 1.99332, "chunk_loss": 0.030821, "F": 0.22134, "G": 0.22322, "chunks": 226.66, "bytes_per_chunk": 4.518, "boundary_entropy": 0.07549, "router_noise": 0.0, "grad_norm": 0.7112, "lr": 5.85e-05, "router_trainable": true, "sec_per_step": 1.373, "probe_sec": 0.0904, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04621, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.03208, "n_shadow": 4, "mean_delta_shadow": 0.03068, "n_split": 1, "mean_delta_split": -0.06404, "n_merge": 1, "mean_delta_merge": 0.5062, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -0.27629, "trunc_abs_frac": 0.34067, "mean_delta_full": 0.1121, "read_hit_rate": {"surface": 0.57513, "split": 0.6056, "merge": 0.13827, "skel": 0.4725}, "mean_gate": {"surface": 0.16932, "split": 0.05576, "merge": 0.11319, "skel": 0.10992}, "n_reads": 15372, "n_cand": 32489, "read_suppressed_frac": 0.1008, "dropout_reads_logged": 15372, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1852, "admitted_surface": 52677, "staged_surface": 56790, "admitted_split": 7352, "staged_split": 43600, "admitted_merge": 3950, "staged_merge": 116740, "admitted_skel": 1557, "staged_skel": 45014, "mean_count": 768.0115356445312, "n_util_obs": 530, "util_count_spearman": -0.013506890390957592}, "probe_overhead_sec_total": 908.579, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008874, "alpha": 0.03, "t": 8165.2} |
| {"event": "consolidate", "step": 4850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 271, "uncertain_staged": 262125, "t": 8224.1} |
| {"event": "train", "step": 4850, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.33809, "bpb": 1.93046, "chunk_loss": 0.030733, "F": 0.21899, "G": 0.2192, "chunks": 224.25, "bytes_per_chunk": 4.566, "boundary_entropy": 0.07962, "router_noise": 0.0, "grad_norm": 0.6856, "lr": 5.62e-05, "router_trainable": true, "sec_per_step": 1.207, "probe_sec": 0.1316, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05245, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": 0.08475, "n_shadow": 3, "mean_delta_shadow": -0.0705, "n_split": 3, "mean_delta_split": -0.03794, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.23974, "trunc_abs_frac": 2.50398, "mean_delta_full": 0.3418, "read_hit_rate": {"surface": 0.58189, "split": 0.59539, "merge": 0.15846, "skel": 0.48507}, "mean_gate": {"surface": 0.17243, "split": 0.05544, "merge": 0.11661, "skel": 0.11327}, "n_reads": 15159, "n_cand": 31822, "read_suppressed_frac": 0.0965, "dropout_reads_logged": 15159, "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": 52431, "staged_surface": 56823, "admitted_split": 7401, "staged_split": 43572, "admitted_merge": 4020, "staged_merge": 116683, "admitted_skel": 1684, "staged_skel": 45066, "mean_count": 779.3584594726562, "n_util_obs": 534, "util_count_spearman": -0.007914994067783405}, "probe_overhead_sec_total": 914.331, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008473, "alpha": 0.03, "t": 8225.6} |
| {"event": "consolidate", "step": 4900, "retired": 1, "split_retired": 3, "merge_admitted": 0, "admitted": 4, "swapped": 210, "uncertain_staged": 262125, "t": 8284.8} |
| {"event": "train", "step": 4900, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.34494, "bpb": 1.94034, "chunk_loss": 0.030458, "F": 0.20706, "G": 0.20817, "chunks": 212.03, "bytes_per_chunk": 4.829, "boundary_entropy": 0.07442, "router_noise": 0.0, "grad_norm": 0.6908, "lr": 5.41e-05, "router_trainable": true, "sec_per_step": 1.214, "probe_sec": 0.0857, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04475, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.04148, "n_shadow": 6, "mean_delta_shadow": 0.14602, "n_split": 1, "mean_delta_split": 0.14276, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.417, "trunc_abs_frac": 0.41134, "mean_delta_full": 0.18135, "read_hit_rate": {"surface": 0.5522, "split": 0.58991, "merge": 0.14165, "skel": 0.46105}, "mean_gate": {"surface": 0.15985, "split": 0.05328, "merge": 0.10156, "skel": 0.10475}, "n_reads": 14085, "n_cand": 30551, "read_suppressed_frac": 0.1021, "dropout_reads_logged": 14085, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1872, "admitted_surface": 52220, "staged_surface": 56787, "admitted_split": 7444, "staged_split": 43835, "admitted_merge": 4085, "staged_merge": 116644, "admitted_skel": 1787, "staged_skel": 44878, "mean_count": 789.4619140625, "n_util_obs": 535, "util_count_spearman": -0.005104296005440717}, "probe_overhead_sec_total": 919.841, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009404, "alpha": 0.03, "t": 8286.3} |
| {"event": "consolidate", "step": 4950, "retired": 1, "split_retired": 0, "merge_admitted": 0, "admitted": 1, "swapped": 554, "uncertain_staged": 262125, "t": 8340.8} |
| {"event": "train", "step": 4950, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32464, "bpb": 1.91106, "chunk_loss": 0.030653, "F": 0.21625, "G": 0.2174, "chunks": 221.44, "bytes_per_chunk": 4.624, "boundary_entropy": 0.07368, "router_noise": 0.0, "grad_norm": 0.6737, "lr": 5.2e-05, "router_trainable": true, "sec_per_step": 1.116, "probe_sec": 0.2258, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.12574, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.07848, "n_shadow": 5, "mean_delta_shadow": 0.02948, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": 0.1693, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.27602, "trunc_abs_frac": 2.94885, "mean_delta_full": 0.03885, "read_hit_rate": {"surface": 0.56975, "split": 0.60033, "merge": 0.15665, "skel": 0.48267}, "mean_gate": {"surface": 0.15445, "split": 0.04867, "merge": 0.10583, "skel": 0.09642}, "n_reads": 15005, "n_cand": 31584, "read_suppressed_frac": 0.1002, "dropout_reads_logged": 15005, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1883, "admitted_surface": 51677, "staged_surface": 56665, "admitted_split": 7503, "staged_split": 43900, "admitted_merge": 4237, "staged_merge": 116774, "admitted_skel": 2119, "staged_skel": 44805, "mean_count": 800.3907470703125, "n_util_obs": 538, "util_count_spearman": -0.01525450223510331}, "probe_overhead_sec_total": 926.164, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009454, "alpha": 0.03, "t": 8342.1} |
| {"event": "consolidate", "step": 5000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 318, "uncertain_staged": 262125, "t": 8395.1} |
| {"event": "probe", "step": 5000, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3428.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.196978, "cps": 26.78125, "mean_p": 0.198465, "boundary_entropy": 0.075894, "bpb": 1.875765}, "tr": {"chunks": 3742.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.206706, "cps": 29.234375, "mean_p": 0.19365, "boundary_entropy": 0.096643, "bpb": 2.015586}, "fi": {"chunks": 4904.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.260699, "cps": 38.3125, "mean_p": 0.25371, "boundary_entropy": 0.084743, "bpb": 2.117075}}, "bpb": {"en": 1.87577, "tr": 2.01559, "fi": 2.11707}, "gini_cpc": 0.06394, "gini_cps": 0.081497, "bpic": 4.698691, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0493890753874286, "fi": 1.3234937203442434}, "inventory": {"size": 327362, "admitted": 65536, "staged": 261826, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1898, "admitted_surface": 51361, "staged_surface": 56779, "admitted_split": 7538, "staged_split": 43707, "admitted_merge": 4344, "staged_merge": 116474, "admitted_skel": 2293, "staged_skel": 44866, "mean_count": 811.5027465820312, "n_util_obs": 546, "util_count_spearman": 0.005003600646170019}, "t": 8396.7} |
| {"event": "eval", "step": 5000, "phase": 3, "bpb": {"en": 1.89079, "tr": 2.02399, "fi": 2.12539}, "bytes_per_chunk": {"en": 4.756, "tr": 5.388, "fi": 4.087}, "bpb_mean": 2.01339, "frozen_inventory": true, "t": 8399.4} |
| {"event": "train", "step": 5000, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.33009, "bpb": 1.91892, "chunk_loss": 0.030609, "F": 0.2139, "G": 0.21465, "chunks": 219.03, "bytes_per_chunk": 4.675, "boundary_entropy": 0.07662, "router_noise": 0.0, "grad_norm": 0.6861, "lr": 5e-05, "router_trainable": true, "sec_per_step": 1.259, "probe_sec": 0.0906, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04459, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.02244, "n_shadow": 4, "mean_delta_shadow": 0.02126, "n_split": 1, "mean_delta_split": 0.09216, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 5.30537, "trunc_abs_frac": 0.70152, "mean_delta_full": 0.01797, "read_hit_rate": {"surface": 0.57145, "split": 0.60919, "merge": 0.1594, "skel": 0.49039}, "mean_gate": {"surface": 0.16742, "split": 0.05319, "merge": 0.11464, "skel": 0.10648}, "n_reads": 15066, "n_cand": 31326, "read_suppressed_frac": 0.1, "dropout_reads_logged": 15066, "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": 51361, "staged_surface": 56814, "admitted_split": 7538, "staged_split": 43718, "admitted_merge": 4344, "staged_merge": 116712, "admitted_skel": 2293, "staged_skel": 44900, "mean_count": 811.732666015625, "n_util_obs": 546, "util_count_spearman": 0.0037973770168508603}, "probe_overhead_sec_total": 931.393, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009283, "alpha": 0.03, "t": 8405.0} |
| {"event": "consolidate", "step": 5050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262125, "t": 8473.3} |
| {"event": "train", "step": 5050, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.34456, "bpb": 1.93979, "chunk_loss": 0.030573, "F": 0.21353, "G": 0.21444, "chunks": 218.66, "bytes_per_chunk": 4.683, "boundary_entropy": 0.07679, "router_noise": 0.0, "grad_norm": 0.657, "lr": 4.81e-05, "router_trainable": true, "sec_per_step": 1.377, "probe_sec": 0.0906, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06629, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.11226, "n_shadow": 1, "mean_delta_shadow": -0.09967, "n_split": 1, "mean_delta_split": -0.00024, "n_merge": 1, "mean_delta_merge": -0.09065, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.32795, "trunc_abs_frac": 0.25686, "mean_delta_full": -0.10492, "read_hit_rate": {"surface": 0.56726, "split": 0.59155, "merge": 0.16256, "skel": 0.50698}, "mean_gate": {"surface": 0.16387, "split": 0.05325, "merge": 0.10654, "skel": 0.10575}, "n_reads": 14931, "n_cand": 31312, "read_suppressed_frac": 0.0974, "dropout_reads_logged": 14931, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1935, "admitted_surface": 50349, "staged_surface": 56912, "admitted_split": 7637, "staged_split": 43873, "admitted_merge": 4486, "staged_merge": 117218, "admitted_skel": 3064, "staged_skel": 44141, "mean_count": 823.399169921875, "n_util_obs": 549, "util_count_spearman": 0.0027617567113672357}, "probe_overhead_sec_total": 939.924, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008525, "alpha": 0.03, "t": 8473.9} |
| {"event": "consolidate", "step": 5100, "retired": 1, "split_retired": 1, "merge_admitted": 0, "admitted": 2, "swapped": 637, "uncertain_staged": 262125, "t": 8541.9} |
| {"event": "train", "step": 5100, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.31355, "bpb": 1.89506, "chunk_loss": 0.030377, "F": 0.20438, "G": 0.205, "chunks": 209.28, "bytes_per_chunk": 4.893, "boundary_entropy": 0.07484, "router_noise": 0.0, "grad_norm": 0.6588, "lr": 4.63e-05, "router_trainable": true, "sec_per_step": 1.377, "probe_sec": 0.1418, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.11309, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": 0.11265, "n_shadow": 7, "mean_delta_shadow": -0.08083, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.69495, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 0.19898, "trunc_abs_frac": 3.96663, "mean_delta_full": 0.15456, "read_hit_rate": {"surface": 0.56324, "split": 0.60793, "merge": 0.1494, "skel": 0.51027}, "mean_gate": {"surface": 0.1516, "split": 0.048, "merge": 0.09889, "skel": 0.09377}, "n_reads": 14395, "n_cand": 29993, "read_suppressed_frac": 0.1005, "dropout_reads_logged": 14395, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1963, "admitted_surface": 49724, "staged_surface": 57219, "admitted_split": 7674, "staged_split": 44027, "admitted_merge": 4600, "staged_merge": 116843, "admitted_skel": 3538, "staged_skel": 44055, "mean_count": 834.3849487304688, "n_util_obs": 553, "util_count_spearman": -0.00488126720222709}, "probe_overhead_sec_total": 947.794, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.007551, "alpha": 0.03, "t": 8542.7} |
| {"event": "consolidate", "step": 5150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 267, "uncertain_staged": 262125, "t": 8620.4} |
| {"event": "train", "step": 5150, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.30267, "bpb": 1.87936, "chunk_loss": 0.030571, "F": 0.21255, "G": 0.21308, "chunks": 217.66, "bytes_per_chunk": 4.705, "boundary_entropy": 0.07673, "router_noise": 0.0, "grad_norm": 0.6788, "lr": 4.45e-05, "router_trainable": true, "sec_per_step": 1.586, "probe_sec": 0.2411, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.15155, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 3, "mean_delta_loo": -0.02909, "n_shadow": 5, "mean_delta_shadow": -0.02406, "n_split": 3, "mean_delta_split": 0.05341, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.24263, "trunc_abs_frac": 0.47428, "mean_delta_full": -0.04346, "read_hit_rate": {"surface": 0.56714, "split": 0.60154, "merge": 0.15447, "skel": 0.52848}, "mean_gate": {"surface": 0.15542, "split": 0.04849, "merge": 0.10338, "skel": 0.09802}, "n_reads": 15174, "n_cand": 31405, "read_suppressed_frac": 0.0996, "dropout_reads_logged": 15174, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1987, "admitted_surface": 49468, "staged_surface": 57202, "admitted_split": 7697, "staged_split": 43925, "admitted_merge": 4652, "staged_merge": 116993, "admitted_skel": 3719, "staged_skel": 44024, "mean_count": 845.9331665039062, "n_util_obs": 556, "util_count_spearman": -0.0023489075056231127}, "probe_overhead_sec_total": 956.635, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008525, "alpha": 0.03, "t": 8622.0} |
| {"event": "consolidate", "step": 5200, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 307, "uncertain_staged": 262125, "t": 8696.1} |
| {"event": "probe", "step": 5200, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3428.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.196978, "cps": 26.78125, "mean_p": 0.198047, "boundary_entropy": 0.075021, "bpb": 1.867763}, "tr": {"chunks": 3732.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.206154, "cps": 29.15625, "mean_p": 0.194105, "boundary_entropy": 0.097179, "bpb": 2.007725}, "fi": {"chunks": 4914.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.26123, "cps": 38.390625, "mean_p": 0.254841, "boundary_entropy": 0.084834, "bpb": 2.110275}}, "bpb": {"en": 1.86776, "tr": 2.00772, "fi": 2.11028}, "gini_cpc": 0.064476, "gini_cps": 0.08205, "bpic": 4.698691, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0465847218989535, "fi": 1.326192524831079}, "inventory": {"size": 327371, "admitted": 65536, "staged": 261835, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1999, "admitted_surface": 49173, "staged_surface": 57153, "admitted_split": 7723, "staged_split": 44174, "admitted_merge": 4699, "staged_merge": 116614, "admitted_skel": 3941, "staged_skel": 43894, "mean_count": 856.0443725585938, "n_util_obs": 558, "util_count_spearman": -0.0018667144666762341}, "t": 8698.3} |
| {"event": "eval", "step": 5200, "phase": 3, "bpb": {"en": 1.88316, "tr": 2.01631, "fi": 2.11791}, "bytes_per_chunk": {"en": 4.773, "tr": 5.385, "fi": 4.087}, "bpb_mean": 2.00579, "frozen_inventory": true, "t": 8701.7} |
| {"event": "train", "step": 5200, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.345, "bpb": 1.94042, "chunk_loss": 0.030775, "F": 0.21893, "G": 0.22023, "chunks": 224.19, "bytes_per_chunk": 4.568, "boundary_entropy": 0.07652, "router_noise": 0.0, "grad_norm": 0.6945, "lr": 4.29e-05, "router_trainable": true, "sec_per_step": 1.694, "probe_sec": 0.1052, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07411, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.00034, "n_shadow": 4, "mean_delta_shadow": -0.34414, "n_split": 1, "mean_delta_split": -0.13654, "n_merge": 1, "mean_delta_merge": 0.24408, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.21562, "trunc_abs_frac": 1.77111, "mean_delta_full": 0.09253, "read_hit_rate": {"surface": 0.57673, "split": 0.61858, "merge": 0.16934, "skel": 0.55437}, "mean_gate": {"surface": 0.16344, "split": 0.05257, "merge": 0.10809, "skel": 0.10376}, "n_reads": 16020, "n_cand": 32110, "read_suppressed_frac": 0.0973, "dropout_reads_logged": 16020, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2000, "admitted_surface": 49173, "staged_surface": 57265, "admitted_split": 7723, "staged_split": 44152, "admitted_merge": 4699, "staged_merge": 116840, "admitted_skel": 3941, "staged_skel": 43887, "mean_count": 856.2888793945312, "n_util_obs": 559, "util_count_spearman": -0.0002507151642331172}, "probe_overhead_sec_total": 965.221, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.00845, "alpha": 0.03, "t": 8706.7} |
| {"event": "consolidate", "step": 5250, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 155, "uncertain_staged": 262125, "t": 8777.8} |
| {"event": "train", "step": 5250, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.37769, "bpb": 1.98758, "chunk_loss": 0.030636, "F": 0.21677, "G": 0.21768, "chunks": 221.97, "bytes_per_chunk": 4.613, "boundary_entropy": 0.07466, "router_noise": 0.0, "grad_norm": 0.6963, "lr": 4.14e-05, "router_trainable": true, "sec_per_step": 1.454, "probe_sec": 0.1495, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07957, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.01332, "n_shadow": 3, "mean_delta_shadow": -0.05025, "n_split": 1, "mean_delta_split": -0.05231, "n_merge": 1, "mean_delta_merge": 0.10651, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.05528, "trunc_abs_frac": 0.23347, "mean_delta_full": 0.5979, "read_hit_rate": {"surface": 0.577, "split": 0.60958, "merge": 0.16536, "skel": 0.55226}, "mean_gate": {"surface": 0.15995, "split": 0.05167, "merge": 0.10234, "skel": 0.09958}, "n_reads": 15696, "n_cand": 31760, "read_suppressed_frac": 0.0977, "dropout_reads_logged": 15696, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2005, "admitted_surface": 49028, "staged_surface": 57433, "admitted_split": 7740, "staged_split": 43758, "admitted_merge": 4735, "staged_merge": 117111, "admitted_skel": 4033, "staged_skel": 43842, "mean_count": 868.4691162109375, "n_util_obs": 566, "util_count_spearman": -0.002565440311214835}, "probe_overhead_sec_total": 973.742, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008415, "alpha": 0.03, "t": 8779.4} |
| {"event": "consolidate", "step": 5300, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 220, "uncertain_staged": 262126, "t": 8859.3} |
| {"event": "train", "step": 5300, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.36803, "bpb": 1.97365, "chunk_loss": 0.030504, "F": 0.20847, "G": 0.21, "chunks": 213.47, "bytes_per_chunk": 4.797, "boundary_entropy": 0.07392, "router_noise": 0.0, "grad_norm": 0.661, "lr": 3.99e-05, "router_trainable": true, "sec_per_step": 1.627, "probe_sec": 0.2204, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.1435, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.07217, "n_shadow": 5, "mean_delta_shadow": -0.28182, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.08923, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 1.22293, "trunc_abs_frac": 6.18373, "mean_delta_full": -0.08344, "read_hit_rate": {"surface": 0.55817, "split": 0.60508, "merge": 0.14999, "skel": 0.54272}, "mean_gate": {"surface": 0.15413, "split": 0.04909, "merge": 0.1013, "skel": 0.09576}, "n_reads": 14854, "n_cand": 30707, "read_suppressed_frac": 0.1019, "dropout_reads_logged": 14854, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2021, "admitted_surface": 48815, "staged_surface": 57350, "admitted_split": 7750, "staged_split": 43725, "admitted_merge": 4765, "staged_merge": 117491, "admitted_skel": 4206, "staged_skel": 43578, "mean_count": 879.1728515625, "n_util_obs": 567, "util_count_spearman": 0.00027194975532422033}, "probe_overhead_sec_total": 983.894, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008942, "alpha": 0.03, "t": 8860.7} |
| {"event": "consolidate", "step": 5350, "retired": 1, "split_retired": 1, "merge_admitted": 0, "admitted": 2, "swapped": 364, "uncertain_staged": 262126, "t": 8930.7} |
| {"event": "train", "step": 5350, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32463, "bpb": 1.91104, "chunk_loss": 0.030775, "F": 0.21387, "G": 0.21521, "chunks": 219.0, "bytes_per_chunk": 4.676, "boundary_entropy": 0.07523, "router_noise": 0.0, "grad_norm": 0.7342, "lr": 3.86e-05, "router_trainable": true, "sec_per_step": 1.427, "probe_sec": 0.219, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.14827, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": -0.03099, "n_shadow": 5, "mean_delta_shadow": -0.0218, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.34812, "trunc_abs_frac": 0.38716, "mean_delta_full": -0.23474, "read_hit_rate": {"surface": 0.56709, "split": 0.61771, "merge": 0.17828, "skel": 0.54935}, "mean_gate": {"surface": 0.16742, "split": 0.05088, "merge": 0.12303, "skel": 0.10181}, "n_reads": 15457, "n_cand": 31121, "read_suppressed_frac": 0.0989, "dropout_reads_logged": 15457, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2039, "admitted_surface": 48460, "staged_surface": 57429, "admitted_split": 7771, "staged_split": 43802, "admitted_merge": 4794, "staged_merge": 117429, "admitted_skel": 4511, "staged_skel": 43484, "mean_count": 890.21533203125, "n_util_obs": 571, "util_count_spearman": 0.004598696612151045}, "probe_overhead_sec_total": 991.73, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008559, "alpha": 0.03, "t": 8932.1} |
| {"event": "consolidate", "step": 5400, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 152, "uncertain_staged": 262126, "t": 9007.0} |
| {"event": "probe", "step": 5400, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3423.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.19669, "cps": 26.742188, "mean_p": 0.197422, "boundary_entropy": 0.074684, "bpb": 1.865008}, "tr": {"chunks": 3742.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.206706, "cps": 29.234375, "mean_p": 0.193392, "boundary_entropy": 0.094858, "bpb": 1.998503}, "fi": {"chunks": 4900.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.260486, "cps": 38.28125, "mean_p": 0.252693, "boundary_entropy": 0.080715, "bpb": 2.10115}}, "bpb": {"en": 1.86501, "tr": 1.9985, "fi": 2.10115}, "gini_cpc": 0.064063, "gini_cps": 0.081613, "bpic": 4.702196, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.05092192533687, "fi": 1.3243458582026635}, "inventory": {"size": 327527, "admitted": 65536, "staged": 261991, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2051, "admitted_surface": 48320, "staged_surface": 57470, "admitted_split": 7777, "staged_split": 43948, "admitted_merge": 4810, "staged_merge": 117031, "admitted_skel": 4629, "staged_skel": 43542, "mean_count": 901.3544921875, "n_util_obs": 581, "util_count_spearman": -0.0031178087463021877}, "t": 9009.0} |
| {"event": "eval", "step": 5400, "phase": 3, "bpb": {"en": 1.87955, "tr": 2.0073, "fi": 2.10881}, "bytes_per_chunk": {"en": 4.773, "tr": 5.394, "fi": 4.088}, "bpb_mean": 1.99855, "frozen_inventory": true, "t": 9012.2} |
| {"event": "train", "step": 5400, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.2501, "bpb": 1.80352, "chunk_loss": 0.030489, "F": 0.20651, "G": 0.20679, "chunks": 211.47, "bytes_per_chunk": 4.842, "boundary_entropy": 0.07549, "router_noise": 0.0, "grad_norm": 0.639, "lr": 3.73e-05, "router_trainable": true, "sec_per_step": 1.714, "probe_sec": 0.0672, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04694, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.03407, "n_shadow": 3, "mean_delta_shadow": -0.04827, "n_split": 1, "mean_delta_split": -0.265, "n_merge": 1, "mean_delta_merge": 0.18689, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.31738, "trunc_abs_frac": 0.54641, "mean_delta_full": -0.16287, "read_hit_rate": {"surface": 0.57416, "split": 0.608, "merge": 0.17679, "skel": 0.5794}, "mean_gate": {"surface": 0.15676, "split": 0.04928, "merge": 0.10606, "skel": 0.098}, "n_reads": 15239, "n_cand": 30439, "read_suppressed_frac": 0.0988, "dropout_reads_logged": 15239, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2052, "admitted_surface": 48320, "staged_surface": 57494, "admitted_split": 7777, "staged_split": 43918, "admitted_merge": 4810, "staged_merge": 117162, "admitted_skel": 4629, "staged_skel": 43570, "mean_count": 901.5870361328125, "n_util_obs": 581, "util_count_spearman": -0.00136937987246543}, "probe_overhead_sec_total": 998.932, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008548, "alpha": 0.03, "t": 9017.8} |
| {"event": "consolidate", "step": 5450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 52, "uncertain_staged": 262126, "t": 9098.0} |
| {"event": "train", "step": 5450, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.30234, "bpb": 1.87887, "chunk_loss": 0.030582, "F": 0.21451, "G": 0.21562, "chunks": 219.66, "bytes_per_chunk": 4.662, "boundary_entropy": 0.07377, "router_noise": 0.0, "grad_norm": 0.6762, "lr": 3.62e-05, "router_trainable": true, "sec_per_step": 1.625, "probe_sec": 0.1108, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06975, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": -0.06187, "n_shadow": 4, "mean_delta_shadow": -0.1182, "n_split": 3, "mean_delta_split": 0.02896, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.70972, "trunc_abs_frac": 0.70972, "mean_delta_full": 0.18713, "read_hit_rate": {"surface": 0.56867, "split": 0.6202, "merge": 0.15434, "skel": 0.58683}, "mean_gate": {"surface": 0.152, "split": 0.04786, "merge": 0.10367, "skel": 0.09318}, "n_reads": 15875, "n_cand": 31702, "read_suppressed_frac": 0.0958, "dropout_reads_logged": 15875, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2073, "admitted_surface": 48273, "staged_surface": 57373, "admitted_split": 7784, "staged_split": 43542, "admitted_merge": 4830, "staged_merge": 118011, "admitted_skel": 4649, "staged_skel": 43218, "mean_count": 913.38623046875, "n_util_obs": 585, "util_count_spearman": 0.004641794138159235}, "probe_overhead_sec_total": 1005.95, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.007833, "alpha": 0.03, "t": 9099.0} |
| {"event": "consolidate", "step": 5500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 56, "uncertain_staged": 262126, "t": 9172.2} |
| {"event": "train", "step": 5500, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.31443, "bpb": 1.89631, "chunk_loss": 0.030528, "F": 0.21109, "G": 0.21205, "chunks": 216.16, "bytes_per_chunk": 4.737, "boundary_entropy": 0.07393, "router_noise": 0.0, "grad_norm": 0.6848, "lr": 3.51e-05, "router_trainable": true, "sec_per_step": 1.485, "probe_sec": 0.089, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05688, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.03685, "n_shadow": 5, "mean_delta_shadow": -0.19525, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.07874, "n_dropped_unpaired": 0, "n_full": 7, "trunc_ratio_sum": 0.46171, "trunc_abs_frac": 1.14918, "mean_delta_full": -0.12052, "read_hit_rate": {"surface": 0.57444, "split": 0.61915, "merge": 0.16912, "skel": 0.58425}, "mean_gate": {"surface": 0.15761, "split": 0.04887, "merge": 0.10143, "skel": 0.09528}, "n_reads": 15643, "n_cand": 31042, "read_suppressed_frac": 0.1002, "dropout_reads_logged": 15643, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2109, "admitted_surface": 48228, "staged_surface": 57470, "admitted_split": 7792, "staged_split": 43874, "admitted_merge": 4851, "staged_merge": 117553, "admitted_skel": 4665, "staged_skel": 43247, "mean_count": 925.2303466796875, "n_util_obs": 591, "util_count_spearman": 0.003919767883330582}, "probe_overhead_sec_total": 1012.502, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008597, "alpha": 0.03, "t": 9173.3} |
| {"event": "consolidate", "step": 5550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 48, "uncertain_staged": 262126, "t": 9250.4} |
| {"event": "train", "step": 5550, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.29718, "bpb": 1.87143, "chunk_loss": 0.030646, "F": 0.21912, "G": 0.21852, "chunks": 224.38, "bytes_per_chunk": 4.564, "boundary_entropy": 0.07064, "router_noise": 0.0, "grad_norm": 0.6596, "lr": 3.41e-05, "router_trainable": true, "sec_per_step": 1.558, "probe_sec": 0.0943, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05861, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.05924, "n_shadow": 4, "mean_delta_shadow": -0.09934, "n_split": 1, "mean_delta_split": 0.07483, "n_merge": 1, "mean_delta_merge": -0.10138, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": -159.14285, "trunc_abs_frac": 159.14285, "mean_delta_full": 0.00085, "read_hit_rate": {"surface": 0.58576, "split": 0.60989, "merge": 0.17805, "skel": 0.60219}, "mean_gate": {"surface": 0.14381, "split": 0.04409, "merge": 0.09621, "skel": 0.08496}, "n_reads": 16391, "n_cand": 32266, "read_suppressed_frac": 0.0977, "dropout_reads_logged": 16391, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2120, "admitted_surface": 48187, "staged_surface": 57335, "admitted_split": 7796, "staged_split": 43862, "admitted_merge": 4872, "staged_merge": 117657, "admitted_skel": 4681, "staged_skel": 43290, "mean_count": 937.0258178710938, "n_util_obs": 601, "util_count_spearman": -0.0008859265810008401}, "probe_overhead_sec_total": 1018.84, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008554, "alpha": 0.03, "t": 9251.1} |
| {"event": "consolidate", "step": 5600, "retired": 0, "split_retired": 4, "merge_admitted": 0, "admitted": 4, "swapped": 82, "uncertain_staged": 262126, "t": 9321.0} |
| {"event": "probe", "step": 5600, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3431.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.19715, "cps": 26.804688, "mean_p": 0.197381, "boundary_entropy": 0.073474, "bpb": 1.859451}, "tr": {"chunks": 3750.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.207148, "cps": 29.296875, "mean_p": 0.193767, "boundary_entropy": 0.09189, "bpb": 1.993239}, "fi": {"chunks": 4900.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.260486, "cps": 38.28125, "mean_p": 0.252871, "boundary_entropy": 0.079273, "bpb": 2.092873}}, "bpb": {"en": 1.85945, "tr": 1.99324, "fi": 2.09287}, "gini_cpc": 0.063515, "gini_cps": 0.081064, "bpic": 4.695969, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.050713031021539, "fi": 1.3212579051669242}, "inventory": {"size": 327594, "admitted": 65536, "staged": 262058, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2140, "admitted_surface": 48117, "staged_surface": 57333, "admitted_split": 7802, "staged_split": 43786, "admitted_merge": 4893, "staged_merge": 117893, "admitted_skel": 4724, "staged_skel": 43046, "mean_count": 945.3786010742188, "n_util_obs": 601, "util_count_spearman": 0.000986893410951112}, "t": 9323.2} |
| {"event": "eval", "step": 5600, "phase": 3, "bpb": {"en": 1.87496, "tr": 2.00219, "fi": 2.10106}, "bytes_per_chunk": {"en": 4.773, "tr": 5.395, "fi": 4.09}, "bpb_mean": 1.99274, "frozen_inventory": true, "t": 9326.5} |
| {"event": "train", "step": 5600, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32632, "bpb": 1.91348, "chunk_loss": 0.030629, "F": 0.21506, "G": 0.21622, "chunks": 220.22, "bytes_per_chunk": 4.65, "boundary_entropy": 0.07455, "router_noise": 0.0, "grad_norm": 0.675, "lr": 3.33e-05, "router_trainable": true, "sec_per_step": 1.641, "probe_sec": 0.201, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.12625, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.07748, "n_shadow": 5, "mean_delta_shadow": -0.00184, "n_split": 1, "mean_delta_split": 0.16583, "n_merge": 1, "mean_delta_merge": -0.08337, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.29604, "trunc_abs_frac": 0.67143, "mean_delta_full": 0.16968, "read_hit_rate": {"surface": 0.56992, "split": 0.61116, "merge": 0.1654, "skel": 0.59513}, "mean_gate": {"surface": 0.16391, "split": 0.05083, "merge": 0.10423, "skel": 0.09924}, "n_reads": 15726, "n_cand": 31437, "read_suppressed_frac": 0.1017, "dropout_reads_logged": 15726, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2140, "admitted_surface": 48117, "staged_surface": 57272, "admitted_split": 7802, "staged_split": 43858, "admitted_merge": 4893, "staged_merge": 118005, "admitted_skel": 4724, "staged_skel": 43009, "mean_count": 945.61865234375, "n_util_obs": 601, "util_count_spearman": 0.000947838912492424}, "probe_overhead_sec_total": 1025.36, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008928, "alpha": 0.03, "t": 9333.2} |
| {"event": "consolidate", "step": 5650, "retired": 1, "split_retired": 1, "merge_admitted": 0, "admitted": 2, "swapped": 89, "uncertain_staged": 262126, "t": 9417.9} |
| {"event": "train", "step": 5650, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.31052, "bpb": 1.89068, "chunk_loss": 0.030548, "F": 0.21201, "G": 0.21288, "chunks": 217.09, "bytes_per_chunk": 4.717, "boundary_entropy": 0.07216, "router_noise": 0.0, "grad_norm": 0.6632, "lr": 3.25e-05, "router_trainable": true, "sec_per_step": 1.724, "probe_sec": 0.1286, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05679, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": 0.0089, "n_shadow": 6, "mean_delta_shadow": 0.06715, "n_split": 1, "mean_delta_split": 0.1936, "n_merge": 1, "mean_delta_merge": 0.59943, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 1.82639, "trunc_abs_frac": 1.49865, "mean_delta_full": 0.04746, "read_hit_rate": {"surface": 0.56428, "split": 0.61847, "merge": 0.16287, "skel": 0.58761}, "mean_gate": {"surface": 0.15291, "split": 0.04747, "merge": 0.10315, "skel": 0.09198}, "n_reads": 15580, "n_cand": 31131, "read_suppressed_frac": 0.1004, "dropout_reads_logged": 15580, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2160, "admitted_surface": 48038, "staged_surface": 57328, "admitted_split": 7805, "staged_split": 44062, "admitted_merge": 4911, "staged_merge": 117501, "admitted_skel": 4782, "staged_skel": 43253, "mean_count": 956.3529663085938, "n_util_obs": 604, "util_count_spearman": 0.007047108256187955}, "probe_overhead_sec_total": 1033.086, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.007658, "alpha": 0.03, "t": 9419.4} |
| {"event": "consolidate", "step": 5700, "retired": 1, "split_retired": 1, "merge_admitted": 0, "admitted": 2, "swapped": 111, "uncertain_staged": 262126, "t": 9489.5} |
| {"event": "train", "step": 5700, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38518, "bpb": 1.99839, "chunk_loss": 0.030577, "F": 0.21344, "G": 0.21413, "chunks": 218.56, "bytes_per_chunk": 4.685, "boundary_entropy": 0.07399, "router_noise": 0.0, "grad_norm": 0.6968, "lr": 3.18e-05, "router_trainable": true, "sec_per_step": 1.441, "probe_sec": 0.2235, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.11284, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": 0.12794, "n_shadow": 4, "mean_delta_shadow": -0.10496, "n_split": 1, "mean_delta_split": 0.08423, "n_merge": 2, "mean_delta_merge": -0.25769, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": -0.01782, "trunc_abs_frac": 0.01782, "mean_delta_full": 0.16785, "read_hit_rate": {"surface": 0.5655, "split": 0.61964, "merge": 0.15931, "skel": 0.58453}, "mean_gate": {"surface": 0.15845, "split": 0.0485, "merge": 0.10734, "skel": 0.09566}, "n_reads": 15754, "n_cand": 31488, "read_suppressed_frac": 0.1004, "dropout_reads_logged": 15754, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2179, "admitted_surface": 47940, "staged_surface": 57509, "admitted_split": 7812, "staged_split": 43733, "admitted_merge": 4931, "staged_merge": 117818, "admitted_skel": 4853, "staged_skel": 43084, "mean_count": 967.8024291992188, "n_util_obs": 608, "util_count_spearman": 0.006372002378855977}, "probe_overhead_sec_total": 1039.978, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008858, "alpha": 0.03, "t": 9491.4} |
| {"event": "consolidate", "step": 5750, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 273, "uncertain_staged": 262126, "t": 9569.3} |
| {"event": "train", "step": 5750, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.37193, "bpb": 1.97927, "chunk_loss": 0.030624, "F": 0.21359, "G": 0.21471, "chunks": 218.72, "bytes_per_chunk": 4.682, "boundary_entropy": 0.07496, "router_noise": 0.0, "grad_norm": 0.7218, "lr": 3.13e-05, "router_trainable": true, "sec_per_step": 1.581, "probe_sec": 0.1149, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05173, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.07146, "n_shadow": 3, "mean_delta_shadow": 0.03078, "n_split": 1, "mean_delta_split": -0.00378, "n_merge": 1, "mean_delta_merge": 0.26105, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.41979, "trunc_abs_frac": 0.23423, "mean_delta_full": 0.10299, "read_hit_rate": {"surface": 0.55519, "split": 0.61118, "merge": 0.15068, "skel": 0.57627}, "mean_gate": {"surface": 0.16015, "split": 0.04949, "merge": 0.10789, "skel": 0.09622}, "n_reads": 15331, "n_cand": 31249, "read_suppressed_frac": 0.0987, "dropout_reads_logged": 15331, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2202, "admitted_surface": 47684, "staged_surface": 57413, "admitted_split": 7821, "staged_split": 43531, "admitted_merge": 4950, "staged_merge": 118340, "admitted_skel": 5081, "staged_skel": 42860, "mean_count": 977.39453125, "n_util_obs": 613, "util_count_spearman": -0.0005832622488979438}, "probe_overhead_sec_total": 1047.414, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008086, "alpha": 0.03, "t": 9570.5} |
| {"event": "consolidate", "step": 5800, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 714, "uncertain_staged": 262126, "t": 9648.3} |
| {"event": "probe", "step": 5800, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3426.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.196863, "cps": 26.765625, "mean_p": 0.19748, "boundary_entropy": 0.072568, "bpb": 1.855354}, "tr": {"chunks": 3752.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.207258, "cps": 29.3125, "mean_p": 0.193142, "boundary_entropy": 0.093324, "bpb": 1.986843}, "fi": {"chunks": 4908.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.260911, "cps": 38.34375, "mean_p": 0.253896, "boundary_entropy": 0.07993, "bpb": 2.086254}}, "bpb": {"en": 1.85535, "tr": 1.98684, "fi": 2.08625}, "gini_cpc": 0.064206, "gini_cps": 0.081747, "bpic": 4.694026, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0528076690562167, "fi": 1.3253464897993044}, "inventory": {"size": 326965, "admitted": 65536, "staged": 261429, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2220, "admitted_surface": 46993, "staged_surface": 57415, "admitted_split": 7825, "staged_split": 43849, "admitted_merge": 4976, "staged_merge": 117799, "admitted_skel": 5742, "staged_skel": 42366, "mean_count": 988.9061889648438, "n_util_obs": 618, "util_count_spearman": 0.0051325601450483045}, "t": 9650.2} |
| {"event": "eval", "step": 5800, "phase": 3, "bpb": {"en": 1.87073, "tr": 1.99551, "fi": 2.09433}, "bytes_per_chunk": {"en": 4.773, "tr": 5.388, "fi": 4.083}, "bpb_mean": 1.98685, "frozen_inventory": true, "t": 9654.0} |
| {"event": "train", "step": 5800, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.28827, "bpb": 1.85858, "chunk_loss": 0.030487, "F": 0.20929, "G": 0.21015, "chunks": 214.31, "bytes_per_chunk": 4.778, "boundary_entropy": 0.07364, "router_noise": 0.0, "grad_norm": 0.6751, "lr": 3.08e-05, "router_trainable": true, "sec_per_step": 1.836, "probe_sec": 0.1274, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07401, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": -0.10261, "n_shadow": 5, "mean_delta_shadow": 0.02728, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": 0.38452, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.01983, "trunc_abs_frac": 0.17979, "mean_delta_full": 0.21701, "read_hit_rate": {"surface": 0.57017, "split": 0.61874, "merge": 0.15602, "skel": 0.60123}, "mean_gate": {"surface": 0.14736, "split": 0.04634, "merge": 0.09361, "skel": 0.0906}, "n_reads": 15656, "n_cand": 31071, "read_suppressed_frac": 0.0975, "dropout_reads_logged": 15656, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2219, "admitted_surface": 46993, "staged_surface": 57583, "admitted_split": 7825, "staged_split": 43888, "admitted_merge": 4976, "staged_merge": 118210, "admitted_skel": 5742, "staged_skel": 42463, "mean_count": 989.1450805664062, "n_util_obs": 618, "util_count_spearman": 0.005017124951549797}, "probe_overhead_sec_total": 1054.144, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009037, "alpha": 0.03, "t": 9662.3} |
| {"event": "consolidate", "step": 5850, "retired": 1, "split_retired": 2, "merge_admitted": 0, "admitted": 3, "swapped": 863, "uncertain_staged": 262126, "t": 9747.2} |
| {"event": "train", "step": 5850, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.33973, "bpb": 1.93283, "chunk_loss": 0.030423, "F": 0.20502, "G": 0.20557, "chunks": 209.94, "bytes_per_chunk": 4.878, "boundary_entropy": 0.07171, "router_noise": 0.0, "grad_norm": 0.6863, "lr": 3.05e-05, "router_trainable": true, "sec_per_step": 1.707, "probe_sec": 0.0684, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04599, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.13079, "n_shadow": 6, "mean_delta_shadow": -0.01266, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.07593, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.964, "trunc_abs_frac": 0.91751, "mean_delta_full": 0.12717, "read_hit_rate": {"surface": 0.56626, "split": 0.60301, "merge": 0.16201, "skel": 0.58491}, "mean_gate": {"surface": 0.14925, "split": 0.04522, "merge": 0.10132, "skel": 0.08762}, "n_reads": 14907, "n_cand": 30150, "read_suppressed_frac": 0.1032, "dropout_reads_logged": 14907, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2232, "admitted_surface": 46140, "staged_surface": 57560, "admitted_split": 7832, "staged_split": 43874, "admitted_merge": 5001, "staged_merge": 118600, "admitted_skel": 6563, "staged_skel": 42110, "mean_count": 999.6614990234375, "n_util_obs": 622, "util_count_spearman": 0.011412010104653028}, "probe_overhead_sec_total": 1061.752, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008254, "alpha": 0.03, "t": 9747.6} |
| {"event": "consolidate", "step": 5900, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262126, "t": 9819.3} |
| {"event": "train", "step": 5900, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.37546, "bpb": 1.98437, "chunk_loss": 0.030615, "F": 0.21561, "G": 0.21621, "chunks": 220.78, "bytes_per_chunk": 4.638, "boundary_entropy": 0.07239, "router_noise": 0.0, "grad_norm": 0.7503, "lr": 3.02e-05, "router_trainable": true, "sec_per_step": 1.461, "probe_sec": 0.2279, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.12984, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.01994, "n_shadow": 4, "mean_delta_shadow": 0.07773, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.03296, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -3.31633, "trunc_abs_frac": 7.00331, "mean_delta_full": 0.01047, "read_hit_rate": {"surface": 0.57756, "split": 0.61569, "merge": 0.15726, "skel": 0.60194}, "mean_gate": {"surface": 0.15273, "split": 0.04729, "merge": 0.1059, "skel": 0.09328}, "n_reads": 16019, "n_cand": 31788, "read_suppressed_frac": 0.0966, "dropout_reads_logged": 16019, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2250, "admitted_surface": 45132, "staged_surface": 57606, "admitted_split": 7838, "staged_split": 43932, "admitted_merge": 5024, "staged_merge": 118988, "admitted_skel": 7542, "staged_skel": 41618, "mean_count": 1011.1588134765625, "n_util_obs": 630, "util_count_spearman": 0.02402369839069765}, "probe_overhead_sec_total": 1068.46, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008142, "alpha": 0.03, "t": 9820.7} |
| {"event": "consolidate", "step": 5950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262126, "t": 9894.1} |
| {"event": "train", "step": 5950, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.30546, "bpb": 1.88339, "chunk_loss": 0.030478, "F": 0.20773, "G": 0.2087, "chunks": 212.72, "bytes_per_chunk": 4.814, "boundary_entropy": 0.07564, "router_noise": 0.0, "grad_norm": 0.7743, "lr": 3.01e-05, "router_trainable": true, "sec_per_step": 1.48, "probe_sec": 0.0785, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04412, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.14448, "n_shadow": 4, "mean_delta_shadow": 0.02704, "n_split": 1, "mean_delta_split": 0.06836, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.2624, "trunc_abs_frac": 0.85616, "mean_delta_full": -0.17468, "read_hit_rate": {"surface": 0.57343, "split": 0.61102, "merge": 0.15824, "skel": 0.60389}, "mean_gate": {"surface": 0.15662, "split": 0.04931, "merge": 0.10216, "skel": 0.09373}, "n_reads": 15455, "n_cand": 30769, "read_suppressed_frac": 0.1013, "dropout_reads_logged": 15455, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2259, "admitted_surface": 44122, "staged_surface": 57881, "admitted_split": 7840, "staged_split": 43966, "admitted_merge": 5040, "staged_merge": 118841, "admitted_skel": 8534, "staged_skel": 41456, "mean_count": 1023.1349487304688, "n_util_obs": 637, "util_count_spearman": 0.02532870107234297}, "probe_overhead_sec_total": 1075.008, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.007627, "alpha": 0.03, "t": 9894.7} |
| {"event": "consolidate", "step": 6000, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262126, "t": 9976.0} |
| {"event": "train", "step": 5999, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32306, "bpb": 1.90877, "chunk_loss": 0.030703, "F": 0.21774, "G": 0.21778, "chunks": 222.97, "bytes_per_chunk": 4.593, "boundary_entropy": 0.07361, "router_noise": 0.0, "grad_norm": 0.7232, "lr": 3e-05, "router_trainable": true, "sec_per_step": 1.626, "probe_sec": 0.135, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08116, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": -0.01683, "n_shadow": 7, "mean_delta_shadow": 0.1548, "n_split": 1, "mean_delta_split": -0.11148, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 3.65044, "trunc_abs_frac": 2.78505, "mean_delta_full": 0.05579, "read_hit_rate": {"surface": 0.58285, "split": 0.62674, "merge": 0.16124, "skel": 0.62301}, "mean_gate": {"surface": 0.15374, "split": 0.04854, "merge": 0.09722, "skel": 0.09435}, "n_reads": 16532, "n_cand": 32159, "read_suppressed_frac": 0.097, "dropout_reads_logged": 16532, "inventory": {"size": 326655, "admitted": 65536, "staged": 261119, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2266, "admitted_surface": 43113, "staged_surface": 57651, "admitted_split": 7840, "staged_split": 43965, "admitted_merge": 5061, "staged_merge": 118478, "admitted_skel": 9522, "staged_skel": 41025, "mean_count": 1034.54833984375, "n_util_obs": 639, "util_count_spearman": 0.02287741325218124}, "probe_overhead_sec_total": 1082.319, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.00822, "alpha": 0.03, "t": 9976.0} |
| {"event": "probe", "step": 6000, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3424.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.196748, "cps": 26.75, "mean_p": 0.197029, "boundary_entropy": 0.072207, "bpb": 1.852647}, "tr": {"chunks": 3754.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.207369, "cps": 29.328125, "mean_p": 0.192836, "boundary_entropy": 0.092046, "bpb": 1.980494}, "fi": {"chunks": 4904.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.260699, "cps": 38.3125, "mean_p": 0.252953, "boundary_entropy": 0.077835, "bpb": 2.080541}}, "bpb": {"en": 1.85265, "tr": 1.98049, "fi": 2.08054}, "gini_cpc": 0.064129, "gini_cps": 0.081664, "bpic": 4.69558, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0539841527273066, "fi": 1.3250398578680103}, "inventory": {"size": 326655, "admitted": 65536, "staged": 261119, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2266, "admitted_surface": 43113, "staged_surface": 57651, "admitted_split": 7840, "staged_split": 43965, "admitted_merge": 5061, "staged_merge": 118478, "admitted_skel": 9522, "staged_skel": 41025, "mean_count": 1034.54833984375, "n_util_obs": 639, "util_count_spearman": 0.02287741325218124}, "t": 9978.1} |
| {"event": "eval", "step": 6000, "phase": 3, "bpb": {"en": 1.86797, "tr": 1.98958, "fi": 2.0881}, "bytes_per_chunk": {"en": 4.775, "tr": 5.389, "fi": 4.081}, "bpb_mean": 1.98188, "frozen_inventory": true, "t": 9981.8} |
| {"event": "reps_dumped", "dirs": ["/local/scratch/sas245/beetle-hnet/runs/ali_ali-ema_s0/probe/reps/th0.30", "/local/scratch/sas245/beetle-hnet/runs/ali_ali-ema_s0/probe/reps/th0.40", "/local/scratch/sas245/beetle-hnet/runs/ali_ali-ema_s0/probe/reps/th0.50", "/local/scratch/sas245/beetle-hnet/runs/ali_ali-ema_s0/probe/reps/th0.60", "/local/scratch/sas245/beetle-hnet/runs/ali_ali-ema_s0/probe/reps/th0.70"], "t": 9988.4} |
| {"event": "final", "run": "ali_ali-ema_s0", "arch": "ali", "objective": "ratio", "condition": "A", "step": 6000, "params": 22785088, "seed": 0, "per_lang": {"en": {"cpc": 0.1967476871803712, "cps": 26.75, "bpb": 1.8715781523392994, "bytes_per_chunk": 4.721103627129633, "reads_per_chunk": 2.2520260778734285}, "tr": {"cpc": 0.2073689443738607, "cps": 29.328125, "bpb": 1.9808109552343274, "bytes_per_chunk": 5.3862069674025825, "reads_per_chunk": 2.187971933140879}, "fi": {"cpc": 0.2606985274573388, "cps": 38.3125, "bpb": 2.107022143396508, "bytes_per_chunk": 4.066864726917319, "reads_per_chunk": 2.342949339497506}}, "bpb_mean": 1.986470416990045, "gini_cpc": 0.06412894312061912, "gini_cps": 0.08166418363405614, "bpic": 4.695580201953319, "mean_boundary_entropy": 0.07360895723104477, "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": 1082.318972826004, "probe_overhead_seq_frac": 0.4999166666666667, "wall_sec": 9998.564486980438, "inventory": {"size": 326655, "admitted": 65536, "staged": 261119, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2266, "admitted_surface": 43113, "staged_surface": 57651, "admitted_split": 7840, "staged_split": 43965, "admitted_merge": 5061, "staged_merge": 118478, "admitted_skel": 9522, "staged_skel": 41025, "mean_count": 1034.54833984375, "n_util_obs": 639, "util_count_spearman": 0.02287741325218124}, "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": "ema_floor", "eta_min": 0.02, "consolidate_every": 50, "probe_seqs": 2, "probe_reps": 8, "horizon": 32, "predictor": "mlp", "lambda_util": 0.1, "max_span": 64, "byte_classes_path": "/local/scratch/sas245/beetle-hnet/scratch/data/byte_classes.json", "gate_hidden": 0, "type_dim": 16, "gate_bias_init": -2.0}, "util_vs_frequency_spearman": 0.02287741325218124, "paired_vs_dropout": {"n": 1268, "pearson": -0.024898788949741606, "spearman": 0.004686638710315448, "mean_abs_discrepancy": 0.3598412275314331, "mean_paired": 0.026153046637773514, "mean_dropout": 0.03578787297010422}, "predictor_calibration": {"n": 2868, "all": {"n": 2868, "pearson": 0.20681618158294113, "spearman": 0.1483008085557181, "mae": 0.08303356170654297, "reliability": [{"pred": -0.02220221236348152, "obs": 0.012102223001420498, "n": 287}, {"pred": -0.009130437858402729, "obs": -0.004357097204774618, "n": 287}, {"pred": -0.00534175056964159, "obs": -0.004577903542667627, "n": 287}, {"pred": -0.0021076437551528215, "obs": 0.00418202904984355, "n": 286}, {"pred": 0.001443109242245555, "obs": -0.007365318015217781, "n": 287}, {"pred": 0.005746329668909311, "obs": 0.011148838326334953, "n": 287}, {"pred": 0.00961928628385067, "obs": 0.010647095739841461, "n": 286}, {"pred": 0.012873937375843525, "obs": 0.022422824054956436, "n": 287}, {"pred": 0.02463328279554844, "obs": 0.035355545580387115, "n": 287}, {"pred": 0.05612463131546974, "obs": 0.10583234578371048, "n": 287}]}, "admitted": {"n": 2847, "pearson": 0.20616254286090455, "spearman": 0.14666037252413738, "mae": 0.08345205336809158, "reliability": [{"pred": -0.022175243124365807, "obs": 0.011745003052055836, "n": 285}, {"pred": -0.009119470603764057, "obs": -0.003864285768941045, "n": 285}, {"pred": -0.005312324035912752, "obs": -0.002972489455714822, "n": 284}, {"pred": -0.002000086009502411, "obs": 0.0033795242197811604, "n": 285}, {"pred": 0.0015427975449711084, "obs": -0.007582270540297031, "n": 284}, {"pred": 0.0058468785136938095, "obs": 0.014041947200894356, "n": 285}, {"pred": 0.009695995599031448, "obs": 0.011420855298638344, "n": 285}, {"pred": 0.012964858673512936, "obs": 0.018299460411071777, "n": 284}, {"pred": 0.024859575554728508, "obs": 0.03732848912477493, "n": 285}, {"pred": 0.05624164268374443, "obs": 0.10654908418655396, "n": 285}]}, "staged": {"n": 21, "pearson": -0.23869206086032224, "spearman": 0.04025974025974026, "mae": 0.026298386976122856, "reliability": [{"pred": -0.02134764939546585, "obs": 0.0240020751953125, "n": 2}, {"pred": -0.014742600731551647, "obs": -0.04804294556379318, "n": 2}, {"pred": -0.007022621110081673, "obs": -0.03494574874639511, "n": 3}, {"pred": -0.005272498354315758, "obs": -0.07295365631580353, "n": 1}, {"pred": -0.004921686835587025, "obs": -0.016848638653755188, "n": 2}, {"pred": -0.004309409763664007, "obs": -0.020551763474941254, "n": 3}, {"pred": -0.0037094512954354286, "obs": -0.023353226482868195, "n": 2}, {"pred": -0.002791423350572586, "obs": -0.01817212998867035, "n": 1}, {"pred": 0.0006812354549765587, "obs": -0.00829547829926014, "n": 2}, {"pred": 0.00753486854955554, "obs": -0.025236064568161964, "n": 3}]}}, "t": 9999.9} |
|
|
