| {"event": "hub", "repo": "AdaptiveChunking/ali_ali-surface-only_s2", "keep_steps": [0, 1, 2, 4, 6, 8, 16, 32, 64, 128, 256, 512, 1000, 1024, 2048, 3000, 4096, 6000], "t": 3.4} |
| {"event": "start", "run": "ali_ali-surface-only_s2", "arch": "ali", "objective": "ratio", "condition": "A", "params": 22785088, "langs": ["en", "tr", "fi"], "premium": {"en": 1.0, "tr": 1.12, "fi": 1.104}, "schedule": {"tau1": 0.4, "tau2": 0.7, "p1_objective": "global", "p2_objective": "global", "p3_objective": "global", "freeze_in_p3": false, "freeze_scope": "router"}, "corpus": "/local/scratch/sas245/beetle-hnet/scratch/data/tier1_corpus", "premium_kind": "flores_per_sent", "doc_filtered": {}, "weights": {"en": 0.74, "tr": 0.1536, "fi": 0.1064}, "device": "cuda", "ali_params": 268096, "ali_config": {"enabled": true, "budget_B": 65536, "staging_capacity": 262144, "policy": "utility", "candidate_types": ["surface"], "gate": "vector", "retrieval": true, "memory_scope": "persistent", "read_dropout": 0.1, "write_rule": "mean", "eta_min": 0.0, "consolidate_every": 50, "probe_seqs": 2, "probe_reps": 8, "horizon": 32, "predictor": "mlp", "lambda_util": 0.1, "max_span": 64, "byte_classes_path": "/local/scratch/sas245/beetle-hnet/scratch/data/byte_classes.json", "gate_hidden": 0, "type_dim": 16, "gate_bias_init": -2.0}, "t": 3.4} |
| {"event": "probe_schedule", "n": 45, "steps": [0, 1, 2, 4, 6, 8, 16, 32, 64, 128, 200, 256, 400, 512, 600, 800, 1000, 1024, 1200, 1400, 1600, 1800, 2000, 2048, 2200, 2400, 2600, 2800, 3000, 3200, 3400, 3600, 3800, 4000, 4096, 4200, 4400, 4600, 4800, 5000, 5200, 5400, 5600, 5800, 6000], "t": 3.4} |
| {"event": "probe", "step": 0, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 10933.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.628225, "cps": 85.414062, "mean_p": 0.513249, "boundary_entropy": 0.686295, "bpb": 8.12598}, "tr": {"chunks": 8187.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.452245, "cps": 63.960938, "mean_p": 0.498002, "boundary_entropy": 0.686826, "bpb": 8.097536}, "fi": {"chunks": 13540.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.719792, "cps": 105.78125, "mean_p": 0.517169, "boundary_entropy": 0.686661, "bpb": 8.065117}}, "bpb": {"en": 8.12598, "tr": 8.09754, "fi": 8.06512}, "gini_cpc": 0.099077, "gini_cps": 0.109267, "bpic": 1.737048, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.7198781817505053, "fi": 1.1457544508645294}, "inventory": {"size": 0, "admitted": 0, "staged": 0, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 0.0, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 4.9} |
| {"event": "train", "step": 0, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 5.62735, "bpb": 8.11854, "chunk_loss": 0.062086, "F": 0.59833, "G": 0.50836, "chunks": 612.69, "bytes_per_chunk": 1.671, "boundary_entropy": 0.69105, "router_noise": 0.0, "grad_norm": 11.4935, "lr": 1e-06, "router_trainable": true, "sec_per_step": 0.032, "probe_sec": 0.026, "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}, "mean_gate": {"surface": null}, "n_reads": 0, "n_cand": 19574, "read_suppressed_frac": 0.0, "dropout_reads_logged": 0, "inventory": {"size": 2006, "admitted": 0, "staged": 2006, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 2006, "admitted_split": 0, "staged_split": 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}, "probe_overhead_sec_total": 0.026, "probe_overhead_seq_frac": 0.0, "util_loss": -0.0, "alpha": 0.03, "t": 8.1} |
| {"event": "probe", "step": 1, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 10908.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.626788, "cps": 85.21875, "mean_p": 0.513143, "boundary_entropy": 0.686304, "bpb": 8.094346}, "tr": {"chunks": 8204.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.453185, "cps": 64.09375, "mean_p": 0.498082, "boundary_entropy": 0.686828, "bpb": 8.076221}, "fi": {"chunks": 13505.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.717931, "cps": 105.507812, "mean_p": 0.517038, "boundary_entropy": 0.686672, "bpb": 8.036135}}, "bpb": {"en": 8.09435, "tr": 8.07622, "fi": 8.03614}, "gini_cpc": 0.098169, "gini_cps": 0.108348, "bpic": 1.739338, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.723026293680575, "fi": 1.1454119137393513}, "inventory": {"size": 2006, "admitted": 0, "staged": 2006, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 2006, "admitted_split": 0, "staged_split": 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": 9.4} |
| {"event": "probe", "step": 2, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 10881.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.625237, "cps": 85.007812, "mean_p": 0.512978, "boundary_entropy": 0.686318, "bpb": 8.031132}, "tr": {"chunks": 8248.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.455615, "cps": 64.4375, "mean_p": 0.498245, "boundary_entropy": 0.686832, "bpb": 8.031932}, "fi": {"chunks": 13467.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.715911, "cps": 105.210938, "mean_p": 0.516799, "boundary_entropy": 0.686692, "bpb": 7.97627}}, "bpb": {"en": 8.03113, "tr": 8.03193, "fi": 7.97627}, "gini_cpc": 0.09658, "gini_cps": 0.106741, "bpic": 1.740459, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.7287077875199965, "fi": 1.1450232008364603}, "inventory": {"size": 3532, "admitted": 0, "staged": 3532, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 3532, "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": 13.0} |
| {"event": "probe", "step": 4, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 10712.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.615526, "cps": 83.6875, "mean_p": 0.512102, "boundary_entropy": 0.68638, "bpb": 7.819743}, "tr": {"chunks": 8384.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.463128, "cps": 65.5, "mean_p": 0.498671, "boundary_entropy": 0.686846, "bpb": 7.878809}, "fi": {"chunks": 13216.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.702568, "cps": 103.25, "mean_p": 0.515685, "boundary_entropy": 0.686774, "bpb": 7.77363}}, "bpb": {"en": 7.81974, "tr": 7.87881, "fi": 7.77363}, "gini_cpc": 0.089616, "gini_cps": 0.099695, "bpic": 1.755756, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.7524095070127509, "fi": 1.141410077819419}, "inventory": {"size": 5985, "admitted": 0, "staged": 5985, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 5985, "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": 16.7} |
| {"event": "probe", "step": 6, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 10323.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.593174, "cps": 80.648438, "mean_p": 0.510352, "boundary_entropy": 0.686483, "bpb": 7.542832}, "tr": {"chunks": 8506.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.469867, "cps": 66.453125, "mean_p": 0.499187, "boundary_entropy": 0.686869, "bpb": 7.674262}, "fi": {"chunks": 12732.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.676838, "cps": 99.46875, "mean_p": 0.513643, "boundary_entropy": 0.686908, "bpb": 7.51819}}, "bpb": {"en": 7.54283, "tr": 7.67426, "fi": 7.51819}, "gini_cpc": 0.079305, "gini_cps": 0.089266, "bpic": 1.797535, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.792123722655643, "fi": 1.141045439059127}, "inventory": {"size": 7731, "admitted": 0, "staged": 7731, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 7731, "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": 22.0} |
| {"event": "probe", "step": 8, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 9633.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.553525, "cps": 75.257812, "mean_p": 0.507287, "boundary_entropy": 0.686614, "bpb": 7.293101}, "tr": {"chunks": 8640.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.477269, "cps": 67.5, "mean_p": 0.499595, "boundary_entropy": 0.686902, "bpb": 7.478103}, "fi": {"chunks": 11870.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.631014, "cps": 92.734375, "mean_p": 0.510288, "boundary_entropy": 0.687082, "bpb": 7.295775}}, "bpb": {"en": 7.2931, "tr": 7.4781, "fi": 7.29578}, "gini_cpc": 0.061678, "gini_cps": 0.071437, "bpic": 1.882095, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.8622352047480546, "fi": 1.1399909284642504}, "inventory": {"size": 9831, "admitted": 0, "staged": 9831, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 9831, "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": 27.0} |
| {"event": "probe", "step": 16, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3754.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.21571, "cps": 29.328125, "mean_p": 0.484398, "boundary_entropy": 0.686162, "bpb": 6.724672}, "tr": {"chunks": 7278.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.402033, "cps": 56.859375, "mean_p": 0.494386, "boundary_entropy": 0.686997, "bpb": 7.013514}, "fi": {"chunks": 4585.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.24374, "cps": 35.820312, "mean_p": 0.486539, "boundary_entropy": 0.686903, "bpb": 6.787241}}, "bpb": {"en": 6.72467, "tr": 7.01351, "fi": 6.78724}, "gini_cpc": 0.144188, "gini_cps": 0.150434, "bpic": 3.632708, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.8637658581329928, "fi": 1.1299450097310129}, "inventory": {"size": 23794, "admitted": 0, "staged": 23794, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 23794, "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": 31.8} |
| {"event": "probe", "step": 32, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 2600.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.1494, "cps": 20.3125, "mean_p": 0.486725, "boundary_entropy": 0.686946, "bpb": 6.304662}, "tr": {"chunks": 4641.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.256366, "cps": 36.257812, "mean_p": 0.49017, "boundary_entropy": 0.687559, "bpb": 6.651807}, "fi": {"chunks": 1390.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.073893, "cps": 10.859375, "mean_p": 0.479393, "boundary_entropy": 0.686972, "bpb": 6.420818}}, "bpb": {"en": 6.30466, "tr": 6.65181, "fi": 6.42082}, "gini_cpc": 0.253616, "gini_cps": 0.25111, "bpic": 6.573051, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.7159782908910126, "fi": 0.49459951828512777}, "inventory": {"size": 50318, "admitted": 0, "staged": 50318, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 50318, "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": 42.3} |
| {"event": "consolidate", "step": 50, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 5, "swapped": 0, "uncertain_staged": 82335, "t": 61.0} |
| {"event": "train", "step": 50, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 4.21233, "bpb": 6.0771, "chunk_loss": 0.034173, "F": 0.22531, "G": 0.48846, "chunks": 230.72, "bytes_per_chunk": 4.438, "boundary_entropy": 0.6916, "router_noise": 0.0, "grad_norm": 2.4788, "lr": 5.1e-05, "router_trainable": true, "sec_per_step": 1.083, "probe_sec": 0.0764, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0419, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": -0.00214, "n_shadow": 5, "mean_delta_shadow": -0.00022, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": -0.1558, "trunc_abs_frac": 0.47395, "mean_delta_full": -0.02246, "read_hit_rate": {"surface": 0.04369}, "mean_gate": {"surface": 0.11939}, "n_reads": 321, "n_cand": 7347, "read_suppressed_frac": 0.1153, "dropout_reads_logged": 321, "inventory": {"size": 84434, "admitted": 5, "staged": 84429, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 5, "staged_surface": 84429, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 12838.2001953125, "n_util_obs": 5, "util_count_spearman": -0.6}, "probe_overhead_sec_total": 3.01, "probe_overhead_seq_frac": 0.4871, "util_loss": 0.044549, "alpha": 0.03, "t": 62.2} |
| {"event": "probe", "step": 64, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 4806.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.276159, "cps": 37.546875, "mean_p": 0.485169, "boundary_entropy": 0.685508, "bpb": 5.552216}, "tr": {"chunks": 5417.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.299232, "cps": 42.320312, "mean_p": 0.486635, "boundary_entropy": 0.686488, "bpb": 5.956406}, "fi": {"chunks": 4729.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.251395, "cps": 36.945312, "mean_p": 0.484103, "boundary_entropy": 0.68634, "bpb": 5.717783}}, "bpb": {"en": 5.55222, "tr": 5.95641, "fi": 5.71778}, "gini_cpc": 0.038572, "gini_cps": 0.030676, "bpic": 3.794275, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0835492051551592, "fi": 0.9103277553422371}, "inventory": {"size": 110880, "admitted": 5, "staged": 110875, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 5, "staged_surface": 110875, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 13338.2001953125, "n_util_obs": 5, "util_count_spearman": -0.6}, "t": 74.1} |
| {"event": "consolidate", "step": 100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 65531, "swapped": 5, "uncertain_staged": 185435, "t": 93.8} |
| {"event": "train", "step": 100, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 3.24927, "bpb": 4.6877, "chunk_loss": 0.029627, "F": 0.16046, "G": 0.45229, "chunks": 164.31, "bytes_per_chunk": 6.232, "boundary_entropy": 0.68188, "router_noise": 0.0, "grad_norm": 1.5519, "lr": 0.000101, "router_trainable": true, "sec_per_step": 0.654, "probe_sec": 0.0547, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03847, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 2, "mean_delta_loo": -0.02032, "n_shadow": 12, "mean_delta_shadow": 0.00756, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.6359, "trunc_abs_frac": 0.48795, "mean_delta_full": -0.0119, "read_hit_rate": {"surface": 0.00613}, "mean_gate": {"surface": 0.11929}, "n_reads": 32, "n_cand": 5224, "read_suppressed_frac": 0.1562, "dropout_reads_logged": 32, "inventory": {"size": 187434, "admitted": 65536, "staged": 121898, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 121898, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 1.07568359375, "n_util_obs": 0, "util_count_spearman": NaN}, "probe_overhead_sec_total": 5.692, "probe_overhead_seq_frac": 0.4935, "util_loss": 0.000968, "alpha": 0.03, "t": 95.0} |
| {"event": "probe", "step": 128, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3556.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.204333, "cps": 27.78125, "mean_p": 0.411481, "boundary_entropy": 0.641534, "bpb": 4.162483}, "tr": {"chunks": 4698.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.259515, "cps": 36.703125, "mean_p": 0.434012, "boundary_entropy": 0.65825, "bpb": 4.432998}, "fi": {"chunks": 3859.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.205146, "cps": 30.148438, "mean_p": 0.427292, "boundary_entropy": 0.658929, "bpb": 4.358537}}, "bpb": {"en": 4.16248, "tr": 4.433, "fi": 4.35854}, "gini_cpc": 0.05499, "gini_cps": 0.062853, "bpic": 4.683563, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.2700617271484937, "fi": 1.0039804660401714}, "inventory": {"size": 237574, "admitted": 65536, "staged": 172038, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 172038, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 1.0928497314453125, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 117.5} |
| {"event": "consolidate", "step": 150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 210550, "t": 136.6} |
| {"event": "train", "step": 150, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.79147, "bpb": 4.02724, "chunk_loss": 0.031626, "F": 0.20102, "G": 0.38156, "chunks": 205.84, "bytes_per_chunk": 4.975, "boundary_entropy": 0.61787, "router_noise": 0.0, "grad_norm": 1.6745, "lr": 0.000151, "router_trainable": true, "sec_per_step": 0.854, "probe_sec": 0.0431, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02945, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 1, "mean_delta_loo": 0.05115, "n_shadow": 13, "mean_delta_shadow": -0.0093, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.24975, "trunc_abs_frac": 1.27956, "mean_delta_full": -0.0481, "read_hit_rate": {"surface": 0.0032}, "mean_gate": {"surface": 0.11987}, "n_reads": 21, "n_cand": 6555, "read_suppressed_frac": 0.2381, "dropout_reads_logged": 21, "inventory": {"size": 277840, "admitted": 65536, "staged": 212304, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 212304, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 1.1043853759765625, "n_util_obs": 0, "util_count_spearman": NaN}, "probe_overhead_sec_total": 8.805, "probe_overhead_seq_frac": 0.4957, "util_loss": 0.000352, "alpha": 0.03, "t": 137.6} |
| {"event": "consolidate", "step": 200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 262128, "t": 177.5} |
| {"event": "probe", "step": 200, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3555.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.204275, "cps": 27.773438, "mean_p": 0.375825, "boundary_entropy": 0.586906, "bpb": 3.63459}, "tr": {"chunks": 4166.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.230128, "cps": 32.546875, "mean_p": 0.366681, "boundary_entropy": 0.586107, "bpb": 3.82995}, "fi": {"chunks": 3567.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.189623, "cps": 27.867188, "mean_p": 0.388793, "boundary_entropy": 0.62306, "bpb": 3.879231}}, "bpb": {"en": 3.63459, "tr": 3.82995, "fi": 3.87923}, "gini_cpc": 0.043272, "gini_cps": 0.036086, "bpic": 5.025868, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.1265571526830414, "fi": 0.9282730478867401}, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 1.1140594482421875, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 178.0} |
| {"event": "eval", "step": 200, "phase": 1, "bpb": {"en": 3.6415, "tr": 3.84096, "fi": 3.87788}, "bytes_per_chunk": {"en": 5.461, "tr": 5.171, "fi": 6.237}, "bpb_mean": 3.78678, "frozen_inventory": true, "t": 179.0} |
| {"event": "train", "step": 200, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.5352, "bpb": 3.65752, "chunk_loss": 0.030558, "F": 0.17929, "G": 0.36345, "chunks": 183.59, "bytes_per_chunk": 5.578, "boundary_entropy": 0.58584, "router_noise": 0.0, "grad_norm": 1.2758, "lr": 0.000201, "router_trainable": true, "sec_per_step": 0.909, "probe_sec": 0.0617, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0352, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 0, "mean_delta_loo": NaN, "n_shadow": 14, "mean_delta_shadow": 0.00092, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 1.34189, "trunc_abs_frac": 2.27483, "mean_delta_full": -0.00903, "read_hit_rate": {"surface": 0.00137}, "mean_gate": {"surface": 0.12166}, "n_reads": 8, "n_cand": 5843, "read_suppressed_frac": 0.125, "dropout_reads_logged": 8, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 1.1141815185546875, "n_util_obs": 0, "util_count_spearman": NaN}, "probe_overhead_sec_total": 11.781, "probe_overhead_seq_frac": 0.4967, "util_loss": 0.000389, "alpha": 0.03, "t": 183.1} |
| {"event": "consolidate", "step": 250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262123, "t": 224.3} |
| {"event": "train", "step": 250, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.38004, "bpb": 3.43368, "chunk_loss": 0.032033, "F": 0.21838, "G": 0.3479, "chunks": 223.62, "bytes_per_chunk": 4.579, "boundary_entropy": 0.54111, "router_noise": 0.0, "grad_norm": 1.7862, "lr": 0.000251, "router_trainable": true, "sec_per_step": 0.845, "probe_sec": 0.105, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06354, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": 0.00064, "n_shadow": 9, "mean_delta_shadow": -0.01421, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.40428, "trunc_abs_frac": 0.45196, "mean_delta_full": 0.0542, "read_hit_rate": {"surface": 0.07159}, "mean_gate": {"surface": 0.12097}, "n_reads": 510, "n_cand": 7124, "read_suppressed_frac": 0.0843, "dropout_reads_logged": 510, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 1.56951904296875, "n_util_obs": 2, "util_count_spearman": NaN}, "probe_overhead_sec_total": 15.042, "probe_overhead_seq_frac": 0.4974, "util_loss": 0.000579, "alpha": 0.03, "t": 225.3} |
| {"event": "probe", "step": 256, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3161.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.181635, "cps": 24.695312, "mean_p": 0.342611, "boundary_entropy": 0.531507, "bpb": 3.344207}, "tr": {"chunks": 4673.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.258134, "cps": 36.507812, "mean_p": 0.328607, "boundary_entropy": 0.518022, "bpb": 3.530362}, "fi": {"chunks": 4719.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.250864, "cps": 36.867188, "mean_p": 0.36248, "boundary_entropy": 0.567214, "bpb": 3.652668}}, "bpb": {"en": 3.34421, "tr": 3.53036, "fi": 3.65267}, "gini_cpc": 0.073844, "gini_cps": 0.082743, "bpic": 4.519398, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.4211661475304682, "fi": 1.3811400475779558}, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 1.6121673583984375, "n_util_obs": 2, "util_count_spearman": NaN}, "t": 228.3} |
| {"event": "consolidate", "step": 300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262124, "t": 255.1} |
| {"event": "train", "step": 300, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.26781, "bpb": 3.27175, "chunk_loss": 0.030779, "F": 0.19092, "G": 0.30501, "chunks": 195.5, "bytes_per_chunk": 5.238, "boundary_entropy": 0.471, "router_noise": 0.0, "grad_norm": 0.9937, "lr": 0.0003, "router_trainable": true, "sec_per_step": 0.613, "probe_sec": 0.0618, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04026, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.01864, "n_shadow": 8, "mean_delta_shadow": -0.00109, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.14414, "trunc_abs_frac": 0.14233, "mean_delta_full": -0.06909, "read_hit_rate": {"surface": 0.11311}, "mean_gate": {"surface": 0.12684}, "n_reads": 704, "n_cand": 6224, "read_suppressed_frac": 0.1094, "dropout_reads_logged": 704, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 2.2884521484375, "n_util_obs": 8, "util_count_spearman": -0.07142857142857142}, "probe_overhead_sec_total": 17.481, "probe_overhead_seq_frac": 0.4978, "util_loss": 0.000852, "alpha": 0.03, "t": 256.0} |
| {"event": "consolidate", "step": 350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262123, "t": 301.4} |
| {"event": "train", "step": 350, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.17205, "bpb": 3.13361, "chunk_loss": 0.031342, "F": 0.21213, "G": 0.2945, "chunks": 217.22, "bytes_per_chunk": 4.714, "boundary_entropy": 0.4292, "router_noise": 0.0, "grad_norm": 1.4736, "lr": 0.0002999, "router_trainable": true, "sec_per_step": 0.926, "probe_sec": 0.06, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03651, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.04335, "n_shadow": 8, "mean_delta_shadow": 0.0181, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.1386, "trunc_abs_frac": 11.31314, "mean_delta_full": 0.16081, "read_hit_rate": {"surface": 0.21434}, "mean_gate": {"surface": 0.15996}, "n_reads": 1483, "n_cand": 6919, "read_suppressed_frac": 0.1005, "dropout_reads_logged": 1483, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 3.9331207275390625, "n_util_obs": 15, "util_count_spearman": -0.35714285714285715}, "probe_overhead_sec_total": 21.087, "probe_overhead_seq_frac": 0.4981, "util_loss": 0.002734, "alpha": 0.03, "t": 302.3} |
| {"event": "consolidate", "step": 400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262122, "t": 341.6} |
| {"event": "probe", "step": 400, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3009.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.172901, "cps": 23.507812, "mean_p": 0.266743, "boundary_entropy": 0.391975, "bpb": 2.911612}, "tr": {"chunks": 3080.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.170138, "cps": 24.0625, "mean_p": 0.26728, "boundary_entropy": 0.421972, "bpb": 3.183065}, "fi": {"chunks": 3438.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.182765, "cps": 26.859375, "mean_p": 0.271416, "boundary_entropy": 0.439286, "bpb": 3.3772}}, "bpb": {"en": 2.91161, "tr": 3.18307, "fi": 3.3772}, "gini_cpc": 0.016011, "gini_cps": 0.03002, "bpic": 5.954865, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9840158582970638, "fi": 1.0570509512343307}, "inventory": {"size": 326656, "admitted": 65536, "staged": 261120, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 261120, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 4.279693603515625, "n_util_obs": 17, "util_count_spearman": -0.056372549019607844}, "t": 342.0} |
| {"event": "eval", "step": 400, "phase": 1, "bpb": {"en": 2.92311, "tr": 3.18976, "fi": 3.36989}, "bytes_per_chunk": {"en": 5.559, "tr": 6.233, "fi": 5.927}, "bpb_mean": 3.16092, "frozen_inventory": true, "t": 342.9} |
| {"event": "train", "step": 400, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.08884, "bpb": 3.01356, "chunk_loss": 0.030234, "F": 0.17484, "G": 0.26871, "chunks": 179.03, "bytes_per_chunk": 5.72, "boundary_entropy": 0.39108, "router_noise": 0.0, "grad_norm": 1.5315, "lr": 0.0002998, "router_trainable": true, "sec_per_step": 0.898, "probe_sec": 0.0856, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06767, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 11, "mean_delta_loo": 0.04486, "n_shadow": 3, "mean_delta_shadow": 0.09221, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 1.83647, "trunc_abs_frac": 0.5875, "mean_delta_full": 0.15564, "read_hit_rate": {"surface": 0.1485}, "mean_gate": {"surface": 0.1845}, "n_reads": 846, "n_cand": 5697, "read_suppressed_frac": 0.1005, "dropout_reads_logged": 846, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 4.2926025390625, "n_util_obs": 17, "util_count_spearman": -0.0784313725490196}, "probe_overhead_sec_total": 24.333, "probe_overhead_seq_frac": 0.4984, "util_loss": 0.004201, "alpha": 0.03, "t": 347.2} |
| {"event": "consolidate", "step": 450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262120, "t": 378.1} |
| {"event": "train", "step": 450, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.00231, "bpb": 2.88873, "chunk_loss": 0.03058, "F": 0.19208, "G": 0.26529, "chunks": 196.69, "bytes_per_chunk": 5.206, "boundary_entropy": 0.35487, "router_noise": 0.0, "grad_norm": 1.648, "lr": 0.0002995, "router_trainable": true, "sec_per_step": 0.633, "probe_sec": 0.0511, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03684, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 11, "mean_delta_loo": 0.16087, "n_shadow": 3, "mean_delta_shadow": 0.054, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 1.08392, "trunc_abs_frac": 0.86239, "mean_delta_full": 0.30356, "read_hit_rate": {"surface": 0.17822}, "mean_gate": {"surface": 0.214}, "n_reads": 1116, "n_cand": 6262, "read_suppressed_frac": 0.103, "dropout_reads_logged": 1116, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 5.4947967529296875, "n_util_obs": 35, "util_count_spearman": 0.010924369747899159}, "probe_overhead_sec_total": 26.954, "probe_overhead_seq_frac": 0.4985, "util_loss": 0.00562, "alpha": 0.03, "t": 378.9} |
| {"event": "consolidate", "step": 500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262122, "t": 409.7} |
| {"event": "train", "step": 500, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.90076, "bpb": 2.74221, "chunk_loss": 0.031196, "F": 0.21735, "G": 0.26904, "chunks": 222.56, "bytes_per_chunk": 4.601, "boundary_entropy": 0.32193, "router_noise": 0.0, "grad_norm": 1.3318, "lr": 0.0002992, "router_trainable": true, "sec_per_step": 0.627, "probe_sec": 0.0294, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02205, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 11, "mean_delta_loo": -0.0158, "n_shadow": 3, "mean_delta_shadow": 0.00214, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 3.25818, "trunc_abs_frac": 0.65867, "mean_delta_full": 0.03357, "read_hit_rate": {"surface": 0.12158}, "mean_gate": {"surface": 0.23236}, "n_reads": 862, "n_cand": 7090, "read_suppressed_frac": 0.1056, "dropout_reads_logged": 862, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 4.91497802734375, "n_util_obs": 41, "util_count_spearman": -0.08310104529616724}, "probe_overhead_sec_total": 29.595, "probe_overhead_seq_frac": 0.4987, "util_loss": 0.009112, "alpha": 0.03, "t": 410.2} |
| {"event": "probe", "step": 512, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3362.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.193185, "cps": 26.265625, "mean_p": 0.256279, "boundary_entropy": 0.331594, "bpb": 2.692087}, "tr": {"chunks": 3243.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.179142, "cps": 25.335938, "mean_p": 0.262095, "boundary_entropy": 0.373246, "bpb": 3.027173}, "fi": {"chunks": 4007.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.213014, "cps": 31.304688, "mean_p": 0.275039, "boundary_entropy": 0.371553, "bpb": 3.207562}}, "bpb": {"en": 2.69209, "tr": 3.02717, "fi": 3.20756}, "gini_cpc": 0.038578, "gini_cps": 0.047996, "bpic": 5.346023, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9273054416654675, "fi": 1.1026403223068912}, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 5.061431884765625, "n_util_obs": 44, "util_count_spearman": -0.05130373502466526}, "t": 419.1} |
| {"event": "consolidate", "step": 550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262120, "t": 451.7} |
| {"event": "train", "step": 550, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.90246, "bpb": 2.74466, "chunk_loss": 0.030779, "F": 0.20056, "G": 0.26507, "chunks": 205.38, "bytes_per_chunk": 4.986, "boundary_entropy": 0.32196, "router_noise": 0.0, "grad_norm": 1.7, "lr": 0.0002987, "router_trainable": true, "sec_per_step": 0.86, "probe_sec": 0.0923, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05673, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.10886, "n_shadow": 6, "mean_delta_shadow": -0.07982, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 2.3396, "trunc_abs_frac": 0.80993, "mean_delta_full": 0.0505, "read_hit_rate": {"surface": 0.13394}, "mean_gate": {"surface": 0.25261}, "n_reads": 876, "n_cand": 6540, "read_suppressed_frac": 0.1039, "dropout_reads_logged": 876, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 5.9677276611328125, "n_util_obs": 60, "util_count_spearman": -0.12497916087802168}, "probe_overhead_sec_total": 32.775, "probe_overhead_seq_frac": 0.4988, "util_loss": 0.008691, "alpha": 0.03, "t": 453.2} |
| {"event": "consolidate", "step": 600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262123, "t": 500.8} |
| {"event": "probe", "step": 600, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3397.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.195196, "cps": 26.539062, "mean_p": 0.25071, "boundary_entropy": 0.304036, "bpb": 2.592043}, "tr": {"chunks": 3437.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.189858, "cps": 26.851562, "mean_p": 0.25998, "boundary_entropy": 0.339278, "bpb": 2.91714}, "fi": {"chunks": 4186.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.222529, "cps": 32.703125, "mean_p": 0.272115, "boundary_entropy": 0.331031, "bpb": 3.13884}}, "bpb": {"en": 2.59204, "tr": 2.91714, "fi": 3.13884}, "gini_cpc": 0.035848, "gini_cps": 0.047731, "bpic": 5.148094, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9726521565481505, "fi": 1.1400290389971721}, "inventory": {"size": 326656, "admitted": 65536, "staged": 261120, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 261120, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 8.520980834960938, "n_util_obs": 80, "util_count_spearman": -0.18285386107827514}, "t": 501.3} |
| {"event": "eval", "step": 600, "phase": 1, "bpb": {"en": 2.60583, "tr": 2.9242, "fi": 3.13919}, "bytes_per_chunk": {"en": 4.938, "tr": 5.836, "fi": 4.757}, "bpb_mean": 2.88974, "frozen_inventory": true, "t": 502.3} |
| {"event": "train", "step": 600, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.88984, "bpb": 2.72646, "chunk_loss": 0.030691, "F": 0.1998, "G": 0.25561, "chunks": 204.59, "bytes_per_chunk": 5.005, "boundary_entropy": 0.30186, "router_noise": 0.0, "grad_norm": 1.2805, "lr": 0.0002982, "router_trainable": true, "sec_per_step": 1.063, "probe_sec": 0.0807, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07033, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.39233, "n_shadow": 4, "mean_delta_shadow": -0.11682, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.81846, "trunc_abs_frac": 0.89184, "mean_delta_full": 0.44246, "read_hit_rate": {"surface": 0.22394}, "mean_gate": {"surface": 0.22994}, "n_reads": 1459, "n_cand": 6515, "read_suppressed_frac": 0.0994, "dropout_reads_logged": 1459, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 8.543243408203125, "n_util_obs": 80, "util_count_spearman": -0.17671224046389042}, "probe_overhead_sec_total": 36.726, "probe_overhead_seq_frac": 0.4989, "util_loss": 0.011021, "alpha": 0.03, "t": 506.4} |
| {"event": "consolidate", "step": 650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262122, "t": 543.8} |
| {"event": "train", "step": 650, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.8404, "bpb": 2.65513, "chunk_loss": 0.030699, "F": 0.19916, "G": 0.2576, "chunks": 203.94, "bytes_per_chunk": 5.021, "boundary_entropy": 0.29836, "router_noise": 0.0, "grad_norm": 1.0709, "lr": 0.0002975, "router_trainable": true, "sec_per_step": 0.752, "probe_sec": 0.0523, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04364, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": -0.11965, "n_shadow": 7, "mean_delta_shadow": -0.26966, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 1.40924, "trunc_abs_frac": 1.16006, "mean_delta_full": -0.3053, "read_hit_rate": {"surface": 0.24808}, "mean_gate": {"surface": 0.2075}, "n_reads": 1611, "n_cand": 6494, "read_suppressed_frac": 0.0875, "dropout_reads_logged": 1611, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 9.901504516601562, "n_util_obs": 94, "util_count_spearman": -0.171304306992076}, "probe_overhead_sec_total": 39.879, "probe_overhead_seq_frac": 0.499, "util_loss": 0.011001, "alpha": 0.03, "t": 544.0} |
| {"event": "consolidate", "step": 700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262124, "t": 570.6} |
| {"event": "train", "step": 700, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.98833, "bpb": 2.86855, "chunk_loss": 0.031131, "F": 0.20959, "G": 0.26683, "chunks": 214.62, "bytes_per_chunk": 4.771, "boundary_entropy": 0.29519, "router_noise": 0.0, "grad_norm": 1.1323, "lr": 0.0002967, "router_trainable": true, "sec_per_step": 0.536, "probe_sec": 0.0274, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02101, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.29, "n_shadow": 8, "mean_delta_shadow": -0.02179, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.52875, "trunc_abs_frac": 0.64769, "mean_delta_full": 0.29089, "read_hit_rate": {"surface": 0.27882}, "mean_gate": {"surface": 0.21554}, "n_reads": 1906, "n_cand": 6836, "read_suppressed_frac": 0.086, "dropout_reads_logged": 1906, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 12.564544677734375, "n_util_obs": 106, "util_count_spearman": -0.25852900352502495}, "probe_overhead_sec_total": 42.842, "probe_overhead_seq_frac": 0.4991, "util_loss": 0.011464, "alpha": 0.03, "t": 570.8} |
| {"event": "consolidate", "step": 750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262125, "t": 609.7} |
| {"event": "train", "step": 750, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.78231, "bpb": 2.57134, "chunk_loss": 0.030781, "F": 0.20401, "G": 0.25915, "chunks": 208.91, "bytes_per_chunk": 4.902, "boundary_entropy": 0.28199, "router_noise": 0.0, "grad_norm": 1.1563, "lr": 0.0002959, "router_trainable": true, "sec_per_step": 0.794, "probe_sec": 0.0665, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03376, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": 0.06551, "n_shadow": 5, "mean_delta_shadow": -0.05524, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -1.35766, "trunc_abs_frac": 1.83109, "mean_delta_full": -0.05721, "read_hit_rate": {"surface": 0.27281}, "mean_gate": {"surface": 0.21684}, "n_reads": 1815, "n_cand": 6653, "read_suppressed_frac": 0.0942, "dropout_reads_logged": 1815, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 13.887069702148438, "n_util_obs": 113, "util_count_spearman": -0.24367057023088695}, "probe_overhead_sec_total": 46.15, "probe_overhead_seq_frac": 0.4991, "util_loss": 0.011683, "alpha": 0.03, "t": 610.5} |
| {"event": "consolidate", "step": 800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262125, "t": 647.2} |
| {"event": "probe", "step": 800, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3462.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.198931, "cps": 27.046875, "mean_p": 0.246752, "boundary_entropy": 0.280153, "bpb": 2.435587}, "tr": {"chunks": 3433.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.189637, "cps": 26.820312, "mean_p": 0.259802, "boundary_entropy": 0.323382, "bpb": 2.778206}, "fi": {"chunks": 4280.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.227526, "cps": 33.4375, "mean_p": 0.266496, "boundary_entropy": 0.298819, "bpb": 2.98712}}, "bpb": {"en": 2.43559, "tr": 2.77821, "fi": 2.98712}, "gini_cpc": 0.041, "gini_cps": 0.050529, "bpic": 5.076689, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9532796205345667, "fi": 1.1437442987329862}, "inventory": {"size": 326656, "admitted": 65536, "staged": 261120, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 261120, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 15.175125122070312, "n_util_obs": 128, "util_count_spearman": -0.19611220472440946}, "t": 647.7} |
| {"event": "eval", "step": 800, "phase": 1, "bpb": {"en": 2.44562, "tr": 2.78721, "fi": 2.98911}, "bytes_per_chunk": {"en": 4.884, "tr": 5.945, "fi": 4.657}, "bpb_mean": 2.74065, "frozen_inventory": true, "t": 648.7} |
| {"event": "train", "step": 800, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.86081, "bpb": 2.68458, "chunk_loss": 0.03101, "F": 0.20905, "G": 0.26169, "chunks": 214.06, "bytes_per_chunk": 4.784, "boundary_entropy": 0.28257, "router_noise": 0.0, "grad_norm": 1.3089, "lr": 0.0002949, "router_trainable": true, "sec_per_step": 0.851, "probe_sec": 0.036, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02058, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.03901, "n_shadow": 4, "mean_delta_shadow": -0.04677, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.99394, "trunc_abs_frac": 3.56638, "mean_delta_full": -0.07053, "read_hit_rate": {"surface": 0.29011}, "mean_gate": {"surface": 0.21737}, "n_reads": 1978, "n_cand": 6818, "read_suppressed_frac": 0.095, "dropout_reads_logged": 1978, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 15.205307006835938, "n_util_obs": 130, "util_count_spearman": -0.18592437118389718}, "probe_overhead_sec_total": 49.297, "probe_overhead_seq_frac": 0.4992, "util_loss": 0.012781, "alpha": 0.03, "t": 653.1} |
| {"event": "consolidate", "step": 850, "retired": 1, "split_retired": 0, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262125, "t": 689.7} |
| {"event": "train", "step": 850, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.81668, "bpb": 2.62092, "chunk_loss": 0.030714, "F": 0.20074, "G": 0.25745, "chunks": 205.56, "bytes_per_chunk": 4.981, "boundary_entropy": 0.2767, "router_noise": 0.0, "grad_norm": 1.1248, "lr": 0.0002938, "router_trainable": true, "sec_per_step": 0.749, "probe_sec": 0.0527, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03545, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": 0.20514, "n_shadow": 10, "mean_delta_shadow": -0.15238, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 0.82555, "trunc_abs_frac": 2.27255, "mean_delta_full": -0.14091, "read_hit_rate": {"surface": 0.24335}, "mean_gate": {"surface": 0.22651}, "n_reads": 1593, "n_cand": 6546, "read_suppressed_frac": 0.0954, "dropout_reads_logged": 1593, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 15.55853271484375, "n_util_obs": 126, "util_count_spearman": -0.2602872449572864}, "probe_overhead_sec_total": 52.339, "probe_overhead_seq_frac": 0.4992, "util_loss": 0.014053, "alpha": 0.03, "t": 690.5} |
| {"event": "consolidate", "step": 900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262125, "t": 726.7} |
| {"event": "train", "step": 900, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.61248, "bpb": 2.32632, "chunk_loss": 0.030632, "F": 0.19748, "G": 0.25342, "chunks": 202.22, "bytes_per_chunk": 5.064, "boundary_entropy": 0.2685, "router_noise": 0.0, "grad_norm": 1.078, "lr": 0.0002927, "router_trainable": true, "sec_per_step": 0.739, "probe_sec": 0.0556, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02933, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": 0.01427, "n_shadow": 5, "mean_delta_shadow": 0.0746, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -0.32762, "trunc_abs_frac": 0.82363, "mean_delta_full": -0.09253, "read_hit_rate": {"surface": 0.29088}, "mean_gate": {"surface": 0.21931}, "n_reads": 1873, "n_cand": 6439, "read_suppressed_frac": 0.0982, "dropout_reads_logged": 1873, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 17.043594360351562, "n_util_obs": 141, "util_count_spearman": -0.20125005351256475}, "probe_overhead_sec_total": 55.339, "probe_overhead_seq_frac": 0.4993, "util_loss": 0.012542, "alpha": 0.03, "t": 727.5} |
| {"event": "consolidate", "step": 950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262126, "t": 772.9} |
| {"event": "train", "step": 950, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.68819, "bpb": 2.43554, "chunk_loss": 0.030672, "F": 0.19995, "G": 0.2501, "chunks": 204.75, "bytes_per_chunk": 5.001, "boundary_entropy": 0.26174, "router_noise": 0.0, "grad_norm": 1.1792, "lr": 0.0002914, "router_trainable": true, "sec_per_step": 0.917, "probe_sec": 0.0513, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04354, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.31676, "n_shadow": 7, "mean_delta_shadow": -0.07049, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.54468, "trunc_abs_frac": 0.54468, "mean_delta_full": -0.44263, "read_hit_rate": {"surface": 0.29494}, "mean_gate": {"surface": 0.21692}, "n_reads": 1923, "n_cand": 6520, "read_suppressed_frac": 0.0952, "dropout_reads_logged": 1923, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 18.50213623046875, "n_util_obs": 148, "util_count_spearman": -0.15444399919676927}, "probe_overhead_sec_total": 58.397, "probe_overhead_seq_frac": 0.4993, "util_loss": 0.012927, "alpha": 0.03, "t": 773.3} |
| {"event": "consolidate", "step": 1000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 55, "uncertain_staged": 262126, "t": 814.0} |
| {"event": "probe", "step": 1000, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3290.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.189048, "cps": 25.703125, "mean_p": 0.238503, "boundary_entropy": 0.261372, "bpb": 2.344654}, "tr": {"chunks": 3439.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.189969, "cps": 26.867188, "mean_p": 0.25803, "boundary_entropy": 0.311534, "bpb": 2.659303}, "fi": {"chunks": 4150.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.220616, "cps": 32.421875, "mean_p": 0.26114, "boundary_entropy": 0.27794, "bpb": 2.848379}}, "bpb": {"en": 2.34465, "tr": 2.6593, "fi": 2.84838}, "gini_cpc": 0.035097, "gini_cps": 0.052701, "bpic": 5.214818, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0048699211385306, "fi": 1.1669827474353347}, "inventory": {"size": 327625, "admitted": 65536, "staged": 262089, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262089, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 20.565139770507812, "n_util_obs": 165, "util_count_spearman": -0.02710713422268096}, "t": 814.6} |
| {"event": "eval", "step": 1000, "phase": 1, "bpb": {"en": 2.35576, "tr": 2.66678, "fi": 2.8525}, "bytes_per_chunk": {"en": 5.115, "tr": 5.907, "fi": 4.822}, "bpb_mean": 2.62501, "frozen_inventory": true, "t": 815.9} |
| {"event": "train", "step": 1000, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.74865, "bpb": 2.52277, "chunk_loss": 0.03058, "F": 0.19641, "G": 0.25018, "chunks": 201.12, "bytes_per_chunk": 5.091, "boundary_entropy": 0.26486, "router_noise": 0.0, "grad_norm": 1.0985, "lr": 0.0002901, "router_trainable": true, "sec_per_step": 0.949, "probe_sec": 0.0282, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01955, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.03422, "n_shadow": 4, "mean_delta_shadow": 0.01644, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.97611, "trunc_abs_frac": 1.48857, "mean_delta_full": 0.16864, "read_hit_rate": {"surface": 0.29872}, "mean_gate": {"surface": 0.2189}, "n_reads": 1913, "n_cand": 6404, "read_suppressed_frac": 0.1093, "dropout_reads_logged": 1913, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 20.594329833984375, "n_util_obs": 165, "util_count_spearman": -0.02267785109395453}, "probe_overhead_sec_total": 61.434, "probe_overhead_seq_frac": 0.4993, "util_loss": 0.013373, "alpha": 0.03, "t": 820.7} |
| {"event": "probe", "step": 1024, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3228.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.185485, "cps": 25.21875, "mean_p": 0.236457, "boundary_entropy": 0.26258, "bpb": 2.333773}, "tr": {"chunks": 3384.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.18693, "cps": 26.4375, "mean_p": 0.260169, "boundary_entropy": 0.313762, "bpb": 2.641372}, "fi": {"chunks": 4257.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.226304, "cps": 33.257812, "mean_p": 0.261264, "boundary_entropy": 0.271507, "bpb": 2.83022}}, "bpb": {"en": 2.33377, "tr": 2.64137, "fi": 2.83022}, "gini_cpc": 0.045451, "gini_cps": 0.063115, "bpic": 5.219615, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0077908178048496, "fi": 1.2200632924775685}, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 21.312103271484375, "n_util_obs": 171, "util_count_spearman": 0.03449256545057977}, "t": 829.9} |
| {"event": "consolidate", "step": 1050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262129, "t": 856.6} |
| {"event": "train", "step": 1050, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.6405, "bpb": 2.36674, "chunk_loss": 0.030765, "F": 0.19943, "G": 0.24755, "chunks": 204.22, "bytes_per_chunk": 5.014, "boundary_entropy": 0.25545, "router_noise": 0.0, "grad_norm": 1.0459, "lr": 0.0002886, "router_trainable": true, "sec_per_step": 0.74, "probe_sec": 0.0295, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01939, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": 0.36022, "n_shadow": 5, "mean_delta_shadow": -0.22599, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 1.07896, "trunc_abs_frac": 2.97896, "mean_delta_full": 0.33085, "read_hit_rate": {"surface": 0.31924}, "mean_gate": {"surface": 0.21506}, "n_reads": 2076, "n_cand": 6503, "read_suppressed_frac": 0.1089, "dropout_reads_logged": 2076, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 22.051300048828125, "n_util_obs": 173, "util_count_spearman": -0.05301979435016881}, "probe_overhead_sec_total": 64.084, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.013809, "alpha": 0.03, "t": 857.7} |
| {"event": "consolidate", "step": 1100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 31, "uncertain_staged": 262129, "t": 897.7} |
| {"event": "train", "step": 1100, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.75143, "bpb": 2.52679, "chunk_loss": 0.030819, "F": 0.20584, "G": 0.25111, "chunks": 210.78, "bytes_per_chunk": 4.858, "boundary_entropy": 0.26164, "router_noise": 0.0, "grad_norm": 1.0638, "lr": 0.0002871, "router_trainable": true, "sec_per_step": 0.804, "probe_sec": 0.0297, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01905, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.35107, "n_shadow": 4, "mean_delta_shadow": 0.09239, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -2.17876, "trunc_abs_frac": 1.22481, "mean_delta_full": -0.05109, "read_hit_rate": {"surface": 0.33904}, "mean_gate": {"surface": 0.21666}, "n_reads": 2276, "n_cand": 6713, "read_suppressed_frac": 0.1015, "dropout_reads_logged": 2276, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 24.778411865234375, "n_util_obs": 189, "util_count_spearman": -0.04660945209187083}, "probe_overhead_sec_total": 66.867, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.014726, "alpha": 0.03, "t": 897.9} |
| {"event": "consolidate", "step": 1150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262131, "t": 936.1} |
| {"event": "train", "step": 1150, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.64311, "bpb": 2.37051, "chunk_loss": 0.03057, "F": 0.19647, "G": 0.24551, "chunks": 201.19, "bytes_per_chunk": 5.09, "boundary_entropy": 0.25224, "router_noise": 0.0, "grad_norm": 0.8687, "lr": 0.0002855, "router_trainable": true, "sec_per_step": 0.787, "probe_sec": 0.0669, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02916, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.43588, "n_shadow": 8, "mean_delta_shadow": -0.21589, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 6.39908, "trunc_abs_frac": 2.12593, "mean_delta_full": 0.00887, "read_hit_rate": {"surface": 0.31814}, "mean_gate": {"surface": 0.21127}, "n_reads": 2038, "n_cand": 6406, "read_suppressed_frac": 0.1006, "dropout_reads_logged": 2038, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 26.068267822265625, "n_util_obs": 185, "util_count_spearman": -0.11632809423833526}, "probe_overhead_sec_total": 69.493, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.014289, "alpha": 0.03, "t": 937.2} |
| {"event": "consolidate", "step": 1200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 876, "uncertain_staged": 262131, "t": 971.1} |
| {"event": "probe", "step": 1200, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3313.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.190369, "cps": 25.882812, "mean_p": 0.234137, "boundary_entropy": 0.241139, "bpb": 2.271076}, "tr": {"chunks": 3367.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.185991, "cps": 26.304688, "mean_p": 0.250389, "boundary_entropy": 0.289718, "bpb": 2.561697}, "fi": {"chunks": 4353.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.231407, "cps": 34.007812, "mean_p": 0.26404, "boundary_entropy": 0.251132, "bpb": 2.723371}}, "bpb": {"en": 2.27108, "tr": 2.5617, "fi": 2.72337}, "gini_cpc": 0.049817, "gini_cps": 0.062842, "bpic": 5.142028, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9770015422531289, "fi": 1.2155685859384155}, "inventory": {"size": 326804, "admitted": 65536, "staged": 261268, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 261268, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 27.697677612304688, "n_util_obs": 193, "util_count_spearman": -0.08567918380428396}, "t": 971.7} |
| {"event": "eval", "step": 1200, "phase": 1, "bpb": {"en": 2.28127, "tr": 2.57139, "fi": 2.73079}, "bytes_per_chunk": {"en": 5.028, "tr": 5.936, "fi": 4.591}, "bpb_mean": 2.52782, "frozen_inventory": true, "t": 972.5} |
| {"event": "train", "step": 1200, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.6216, "bpb": 2.33947, "chunk_loss": 0.030534, "F": 0.19537, "G": 0.24705, "chunks": 200.06, "bytes_per_chunk": 5.118, "boundary_entropy": 0.24854, "router_noise": 0.0, "grad_norm": 0.8844, "lr": 0.0002837, "router_trainable": true, "sec_per_step": 0.809, "probe_sec": 0.0864, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03997, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": 0.15231, "n_shadow": 5, "mean_delta_shadow": 0.01696, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.22187, "trunc_abs_frac": 0.22187, "mean_delta_full": 0.32764, "read_hit_rate": {"surface": 0.32433}, "mean_gate": {"surface": 0.21769}, "n_reads": 2066, "n_cand": 6370, "read_suppressed_frac": 0.1041, "dropout_reads_logged": 2066, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 27.729202270507812, "n_util_obs": 193, "util_count_spearman": -0.084442411667516}, "probe_overhead_sec_total": 72.153, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.014779, "alpha": 0.03, "t": 977.7} |
| {"event": "consolidate", "step": 1250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262131, "t": 1025.1} |
| {"event": "train", "step": 1250, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.6441, "bpb": 2.37193, "chunk_loss": 0.030528, "F": 0.19489, "G": 0.24556, "chunks": 199.56, "bytes_per_chunk": 5.131, "boundary_entropy": 0.25059, "router_noise": 0.0, "grad_norm": 0.9775, "lr": 0.0002819, "router_trainable": true, "sec_per_step": 0.971, "probe_sec": 0.1132, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06019, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.00379, "n_shadow": 4, "mean_delta_shadow": -0.11499, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.74477, "trunc_abs_frac": 1.4458, "mean_delta_full": 0.14193, "read_hit_rate": {"surface": 0.32594}, "mean_gate": {"surface": 0.21548}, "n_reads": 2071, "n_cand": 6354, "read_suppressed_frac": 0.0985, "dropout_reads_logged": 2071, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 29.299484252929688, "n_util_obs": 200, "util_count_spearman": -0.028922255593945696}, "probe_overhead_sec_total": 75.382, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.015441, "alpha": 0.03, "t": 1026.3} |
| {"event": "consolidate", "step": 1300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262131, "t": 1073.1} |
| {"event": "train", "step": 1300, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.68051, "bpb": 2.42446, "chunk_loss": 0.0307, "F": 0.20175, "G": 0.2468, "chunks": 206.59, "bytes_per_chunk": 4.957, "boundary_entropy": 0.23824, "router_noise": 0.0, "grad_norm": 0.9613, "lr": 0.00028, "router_trainable": true, "sec_per_step": 0.942, "probe_sec": 0.0374, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02535, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.41631, "n_shadow": 6, "mean_delta_shadow": 0.02691, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.68615, "trunc_abs_frac": 1.00663, "mean_delta_full": 0.22559, "read_hit_rate": {"surface": 0.34747}, "mean_gate": {"surface": 0.21316}, "n_reads": 2286, "n_cand": 6579, "read_suppressed_frac": 0.1037, "dropout_reads_logged": 2286, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 30.968643188476562, "n_util_obs": 211, "util_count_spearman": -0.0032926273889120226}, "probe_overhead_sec_total": 78.764, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.014729, "alpha": 0.03, "t": 1073.4} |
| {"event": "consolidate", "step": 1350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262131, "t": 1117.1} |
| {"event": "train", "step": 1350, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.60202, "bpb": 2.31122, "chunk_loss": 0.030632, "F": 0.19916, "G": 0.24408, "chunks": 203.94, "bytes_per_chunk": 5.021, "boundary_entropy": 0.23798, "router_noise": 0.0, "grad_norm": 0.9115, "lr": 0.000278, "router_trainable": true, "sec_per_step": 0.892, "probe_sec": 0.0358, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02172, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": 0.13671, "n_shadow": 2, "mean_delta_shadow": -0.14169, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.82946, "trunc_abs_frac": 0.73406, "mean_delta_full": -0.2559, "read_hit_rate": {"surface": 0.35679}, "mean_gate": {"surface": 0.20824}, "n_reads": 2317, "n_cand": 6494, "read_suppressed_frac": 0.104, "dropout_reads_logged": 2317, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 32.65925598144531, "n_util_obs": 220, "util_count_spearman": -0.0015828678364546542}, "probe_overhead_sec_total": 82.09, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.014294, "alpha": 0.03, "t": 1118.0} |
| {"event": "consolidate", "step": 1400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262132, "t": 1144.8} |
| {"event": "probe", "step": 1400, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3379.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.194162, "cps": 26.398438, "mean_p": 0.232848, "boundary_entropy": 0.230151, "bpb": 2.220065}, "tr": {"chunks": 3631.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.200574, "cps": 28.367188, "mean_p": 0.249418, "boundary_entropy": 0.28176, "bpb": 2.493764}, "fi": {"chunks": 4303.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.228749, "cps": 33.617188, "mean_p": 0.25434, "boundary_entropy": 0.235495, "bpb": 2.638133}}, "bpb": {"en": 2.22007, "tr": 2.49376, "fi": 2.63813}, "gini_cpc": 0.036983, "gini_cps": 0.054451, "bpic": 5.014762, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.033026888638305, "fi": 1.1781359030214873}, "inventory": {"size": 326656, "admitted": 65536, "staged": 261120, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 261120, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 34.01264953613281, "n_util_obs": 219, "util_count_spearman": -0.054103305265803696}, "t": 1145.2} |
| {"event": "eval", "step": 1400, "phase": 1, "bpb": {"en": 2.23033, "tr": 2.50099, "fi": 2.64481}, "bytes_per_chunk": {"en": 5.002, "tr": 5.48, "fi": 4.685}, "bpb_mean": 2.45871, "frozen_inventory": true, "t": 1146.2} |
| {"event": "train", "step": 1400, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.60318, "bpb": 2.31289, "chunk_loss": 0.030587, "F": 0.19644, "G": 0.24269, "chunks": 201.16, "bytes_per_chunk": 5.091, "boundary_entropy": 0.23745, "router_noise": 0.0, "grad_norm": 0.8621, "lr": 0.0002759, "router_trainable": true, "sec_per_step": 0.681, "probe_sec": 0.1262, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0861, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.0481, "n_shadow": 8, "mean_delta_shadow": 0.02076, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -3.64615, "trunc_abs_frac": 1.21332, "mean_delta_full": -0.04761, "read_hit_rate": {"surface": 0.34239}, "mean_gate": {"surface": 0.20921}, "n_reads": 2193, "n_cand": 6405, "read_suppressed_frac": 0.1067, "dropout_reads_logged": 2193, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 34.046112060546875, "n_util_obs": 220, "util_count_spearman": -0.047665724028594904}, "probe_overhead_sec_total": 85.176, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.014785, "alpha": 0.03, "t": 1152.0} |
| {"event": "consolidate", "step": 1450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 62, "uncertain_staged": 262132, "t": 1204.2} |
| {"event": "train", "step": 1450, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.65027, "bpb": 2.38083, "chunk_loss": 0.030526, "F": 0.19629, "G": 0.24277, "chunks": 201.0, "bytes_per_chunk": 5.095, "boundary_entropy": 0.23397, "router_noise": 0.0, "grad_norm": 0.8714, "lr": 0.0002738, "router_trainable": true, "sec_per_step": 1.059, "probe_sec": 0.055, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02801, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.26677, "n_shadow": 6, "mean_delta_shadow": -0.20389, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.91941, "trunc_abs_frac": 0.70816, "mean_delta_full": 0.57382, "read_hit_rate": {"surface": 0.33734}, "mean_gate": {"surface": 0.21199}, "n_reads": 2159, "n_cand": 6400, "read_suppressed_frac": 0.0982, "dropout_reads_logged": 2159, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 35.8343505859375, "n_util_obs": 226, "util_count_spearman": -0.05106816070736953}, "probe_overhead_sec_total": 88.858, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.013669, "alpha": 0.03, "t": 1205.0} |
| {"event": "consolidate", "step": 1500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 298, "uncertain_staged": 262132, "t": 1245.6} |
| {"event": "train", "step": 1500, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.66659, "bpb": 2.40438, "chunk_loss": 0.030895, "F": 0.20993, "G": 0.25112, "chunks": 214.97, "bytes_per_chunk": 4.763, "boundary_entropy": 0.23507, "router_noise": 0.0, "grad_norm": 0.8249, "lr": 0.0002715, "router_trainable": true, "sec_per_step": 0.834, "probe_sec": 0.0763, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04577, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.07034, "n_shadow": 6, "mean_delta_shadow": 0.14427, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 1.33069, "trunc_abs_frac": 17.03169, "mean_delta_full": 0.08346, "read_hit_rate": {"surface": 0.37024}, "mean_gate": {"surface": 0.20277}, "n_reads": 2535, "n_cand": 6847, "read_suppressed_frac": 0.0907, "dropout_reads_logged": 2535, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 37.55560302734375, "n_util_obs": 234, "util_count_spearman": -0.03482462421147865}, "probe_overhead_sec_total": 91.953, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.013749, "alpha": 0.03, "t": 1246.7} |
| {"event": "consolidate", "step": 1550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 819, "uncertain_staged": 262132, "t": 1293.4} |
| {"event": "train", "step": 1550, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.59215, "bpb": 2.29698, "chunk_loss": 0.030758, "F": 0.2045, "G": 0.24507, "chunks": 209.41, "bytes_per_chunk": 4.89, "boundary_entropy": 0.2299, "router_noise": 0.0, "grad_norm": 0.8898, "lr": 0.0002692, "router_trainable": true, "sec_per_step": 0.951, "probe_sec": 0.0862, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0437, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 11, "mean_delta_loo": 0.02808, "n_shadow": 3, "mean_delta_shadow": 0.04582, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.37004, "trunc_abs_frac": 0.37004, "mean_delta_full": -0.36914, "read_hit_rate": {"surface": 0.37067}, "mean_gate": {"surface": 0.20277}, "n_reads": 2472, "n_cand": 6669, "read_suppressed_frac": 0.1036, "dropout_reads_logged": 2472, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 39.28160095214844, "n_util_obs": 239, "util_count_spearman": -0.03735720148423711}, "probe_overhead_sec_total": 95.528, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.013363, "alpha": 0.03, "t": 1294.2} |
| {"event": "consolidate", "step": 1600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 351, "uncertain_staged": 262132, "t": 1335.8} |
| {"event": "probe", "step": 1600, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3275.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.188186, "cps": 25.585938, "mean_p": 0.227308, "boundary_entropy": 0.216402, "bpb": 2.176535}, "tr": {"chunks": 3406.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.188146, "cps": 26.609375, "mean_p": 0.23738, "boundary_entropy": 0.260305, "bpb": 2.428394}, "fi": {"chunks": 4218.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.224231, "cps": 32.953125, "mean_p": 0.25154, "boundary_entropy": 0.220918, "bpb": 2.580525}}, "bpb": {"en": 2.17653, "tr": 2.42839, "fi": 2.58053}, "gini_cpc": 0.040057, "gini_cps": 0.057681, "bpic": 5.205248, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.999785670883279, "fi": 1.1915369316556295}, "inventory": {"size": 327329, "admitted": 65536, "staged": 261793, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 261793, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 41.08940124511719, "n_util_obs": 245, "util_count_spearman": -0.0247121710778007}, "t": 1336.4} |
| {"event": "eval", "step": 1600, "phase": 1, "bpb": {"en": 2.18966, "tr": 2.43821, "fi": 2.58907}, "bytes_per_chunk": {"en": 5.044, "tr": 5.756, "fi": 4.717}, "bpb_mean": 2.40565, "frozen_inventory": true, "t": 1337.2} |
| {"event": "train", "step": 1600, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.52083, "bpb": 2.19409, "chunk_loss": 0.030533, "F": 0.1983, "G": 0.23926, "chunks": 203.06, "bytes_per_chunk": 5.043, "boundary_entropy": 0.21876, "router_noise": 0.0, "grad_norm": 0.9103, "lr": 0.0002668, "router_trainable": true, "sec_per_step": 0.968, "probe_sec": 0.0763, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04547, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.1095, "n_shadow": 8, "mean_delta_shadow": -0.09979, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 1.89367, "trunc_abs_frac": 3.22389, "mean_delta_full": -0.06171, "read_hit_rate": {"surface": 0.36313}, "mean_gate": {"surface": 0.20022}, "n_reads": 2348, "n_cand": 6466, "read_suppressed_frac": 0.1009, "dropout_reads_logged": 2348, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 41.12522888183594, "n_util_obs": 245, "util_count_spearman": -0.02421398787423806}, "probe_overhead_sec_total": 98.518, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.01346, "alpha": 0.03, "t": 1342.7} |
| {"event": "consolidate", "step": 1650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 241, "uncertain_staged": 262132, "t": 1385.5} |
| {"event": "train", "step": 1650, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.58476, "bpb": 2.28632, "chunk_loss": 0.030509, "F": 0.19742, "G": 0.23576, "chunks": 202.16, "bytes_per_chunk": 5.065, "boundary_entropy": 0.21369, "router_noise": 0.0, "grad_norm": 0.8311, "lr": 0.0002643, "router_trainable": true, "sec_per_step": 0.88, "probe_sec": 0.0727, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04759, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.0834, "n_shadow": 6, "mean_delta_shadow": -0.25068, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.76452, "trunc_abs_frac": 1.51341, "mean_delta_full": -0.45463, "read_hit_rate": {"surface": 0.36601}, "mean_gate": {"surface": 0.20263}, "n_reads": 2356, "n_cand": 6437, "read_suppressed_frac": 0.0896, "dropout_reads_logged": 2356, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 43.84724426269531, "n_util_obs": 251, "util_count_spearman": 0.003710873618573346}, "probe_overhead_sec_total": 101.675, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.013486, "alpha": 0.03, "t": 1386.7} |
| {"event": "consolidate", "step": 1700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262133, "t": 1435.9} |
| {"event": "train", "step": 1700, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.5418, "bpb": 2.22435, "chunk_loss": 0.03048, "F": 0.19614, "G": 0.23468, "chunks": 200.84, "bytes_per_chunk": 5.098, "boundary_entropy": 0.21868, "router_noise": 0.0, "grad_norm": 0.7875, "lr": 0.0002618, "router_trainable": true, "sec_per_step": 1.01, "probe_sec": 0.0729, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03863, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": 0.17171, "n_shadow": 5, "mean_delta_shadow": -0.06552, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.58423, "trunc_abs_frac": 0.43465, "mean_delta_full": 0.28235, "read_hit_rate": {"surface": 0.36482}, "mean_gate": {"surface": 0.20302}, "n_reads": 2333, "n_cand": 6395, "read_suppressed_frac": 0.096, "dropout_reads_logged": 2333, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 45.71504211425781, "n_util_obs": 262, "util_count_spearman": 0.0031113548772821954}, "probe_overhead_sec_total": 105.15, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.014087, "alpha": 0.03, "t": 1437.2} |
| {"event": "consolidate", "step": 1750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262133, "t": 1472.1} |
| {"event": "train", "step": 1750, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.57005, "bpb": 2.2651, "chunk_loss": 0.03064, "F": 0.20325, "G": 0.23966, "chunks": 208.12, "bytes_per_chunk": 4.92, "boundary_entropy": 0.21497, "router_noise": 0.0, "grad_norm": 0.8061, "lr": 0.0002591, "router_trainable": true, "sec_per_step": 0.722, "probe_sec": 0.061, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03004, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.02335, "n_shadow": 6, "mean_delta_shadow": -0.07837, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 0.92593, "trunc_abs_frac": 1.31787, "mean_delta_full": -0.06167, "read_hit_rate": {"surface": 0.3873}, "mean_gate": {"surface": 0.20141}, "n_reads": 2567, "n_cand": 6628, "read_suppressed_frac": 0.0966, "dropout_reads_logged": 2567, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 47.55279541015625, "n_util_obs": 271, "util_count_spearman": 0.054894775065841735}, "probe_overhead_sec_total": 107.905, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.014063, "alpha": 0.03, "t": 1473.3} |
| {"event": "consolidate", "step": 1800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 98, "uncertain_staged": 262133, "t": 1511.1} |
| {"event": "probe", "step": 1800, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3364.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.1933, "cps": 26.28125, "mean_p": 0.224591, "boundary_entropy": 0.203007, "bpb": 2.139344}, "tr": {"chunks": 3637.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.200906, "cps": 28.414062, "mean_p": 0.232764, "boundary_entropy": 0.243204, "bpb": 2.366942}, "fi": {"chunks": 4404.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.234118, "cps": 34.40625, "mean_p": 0.253479, "boundary_entropy": 0.209073, "bpb": 2.532807}}, "bpb": {"en": 2.13934, "tr": 2.36694, "fi": 2.53281}, "gini_cpc": 0.043309, "gini_cps": 0.060792, "bpic": 4.97431, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0393477559345803, "fi": 1.2111656909323492}, "inventory": {"size": 327582, "admitted": 65536, "staged": 262046, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262046, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 49.40849304199219, "n_util_obs": 280, "util_count_spearman": 0.06281769159346971}, "t": 1511.7} |
| {"event": "eval", "step": 1800, "phase": 1, "bpb": {"en": 2.15327, "tr": 2.37629, "fi": 2.53945}, "bytes_per_chunk": {"en": 5.016, "tr": 5.519, "fi": 4.458}, "bpb_mean": 2.35634, "frozen_inventory": true, "t": 1512.5} |
| {"event": "train", "step": 1800, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.54824, "bpb": 2.23364, "chunk_loss": 0.030504, "F": 0.19559, "G": 0.23234, "chunks": 200.28, "bytes_per_chunk": 5.113, "boundary_entropy": 0.21055, "router_noise": 0.0, "grad_norm": 0.8105, "lr": 0.0002564, "router_trainable": true, "sec_per_step": 0.924, "probe_sec": 0.053, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02336, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.44265, "n_shadow": 7, "mean_delta_shadow": 0.01855, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.55967, "trunc_abs_frac": 1.23042, "mean_delta_full": 0.07684, "read_hit_rate": {"surface": 0.3762}, "mean_gate": {"surface": 0.19907}, "n_reads": 2399, "n_cand": 6377, "read_suppressed_frac": 0.1021, "dropout_reads_logged": 2399, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 49.445098876953125, "n_util_obs": 280, "util_count_spearman": 0.06272124138713539}, "probe_overhead_sec_total": 111.047, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.014904, "alpha": 0.03, "t": 1519.5} |
| {"event": "consolidate", "step": 1850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 302, "uncertain_staged": 262133, "t": 1559.8} |
| {"event": "train", "step": 1850, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.55274, "bpb": 2.24013, "chunk_loss": 0.030623, "F": 0.20126, "G": 0.23412, "chunks": 206.09, "bytes_per_chunk": 4.969, "boundary_entropy": 0.19931, "router_noise": 0.0, "grad_norm": 0.7449, "lr": 0.0002537, "router_trainable": true, "sec_per_step": 0.813, "probe_sec": 0.0293, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01948, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.11904, "n_shadow": 8, "mean_delta_shadow": -0.09321, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.85249, "trunc_abs_frac": 0.85249, "mean_delta_full": 0.23047, "read_hit_rate": {"surface": 0.38854}, "mean_gate": {"surface": 0.19722}, "n_reads": 2550, "n_cand": 6563, "read_suppressed_frac": 0.1012, "dropout_reads_logged": 2550, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 51.36427307128906, "n_util_obs": 289, "util_count_spearman": 0.054910432915141576}, "probe_overhead_sec_total": 113.923, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.014851, "alpha": 0.03, "t": 1560.2} |
| {"event": "consolidate", "step": 1900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262133, "t": 1604.2} |
| {"event": "train", "step": 1900, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.5365, "bpb": 2.2167, "chunk_loss": 0.030776, "F": 0.20514, "G": 0.23784, "chunks": 210.06, "bytes_per_chunk": 4.875, "boundary_entropy": 0.20566, "router_noise": 0.0, "grad_norm": 0.8527, "lr": 0.0002508, "router_trainable": true, "sec_per_step": 0.897, "probe_sec": 0.0503, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0251, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 11, "mean_delta_loo": 0.14388, "n_shadow": 3, "mean_delta_shadow": 0.02574, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 8.30229, "trunc_abs_frac": 8.57239, "mean_delta_full": 0.0213, "read_hit_rate": {"surface": 0.38072}, "mean_gate": {"surface": 0.19222}, "n_reads": 2547, "n_cand": 6690, "read_suppressed_frac": 0.1037, "dropout_reads_logged": 2547, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 53.25975036621094, "n_util_obs": 296, "util_count_spearman": 0.07114067330467633}, "probe_overhead_sec_total": 116.883, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.014691, "alpha": 0.03, "t": 1605.0} |
| {"event": "consolidate", "step": 1950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 76, "uncertain_staged": 262133, "t": 1647.9} |
| {"event": "train", "step": 1950, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.49165, "bpb": 2.152, "chunk_loss": 0.030548, "F": 0.19733, "G": 0.2325, "chunks": 202.06, "bytes_per_chunk": 5.068, "boundary_entropy": 0.20222, "router_noise": 0.0, "grad_norm": 0.7616, "lr": 0.0002479, "router_trainable": true, "sec_per_step": 0.869, "probe_sec": 0.0403, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02597, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.16188, "n_shadow": 7, "mean_delta_shadow": -0.03557, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.80235, "trunc_abs_frac": 28.20652, "mean_delta_full": 0.1947, "read_hit_rate": {"surface": 0.38545}, "mean_gate": {"surface": 0.20481}, "n_reads": 2480, "n_cand": 6434, "read_suppressed_frac": 0.1032, "dropout_reads_logged": 2480, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 55.155487060546875, "n_util_obs": 306, "util_count_spearman": 0.11433844901268922}, "probe_overhead_sec_total": 119.943, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.014843, "alpha": 0.03, "t": 1648.4} |
| {"event": "consolidate", "step": 2000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 87, "uncertain_staged": 262133, "t": 1691.5} |
| {"event": "probe", "step": 2000, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3347.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.192323, "cps": 26.148438, "mean_p": 0.221222, "boundary_entropy": 0.187883, "bpb": 2.110459}, "tr": {"chunks": 3501.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.193393, "cps": 27.351562, "mean_p": 0.22508, "boundary_entropy": 0.235075, "bpb": 2.316816}, "fi": {"chunks": 4406.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.234225, "cps": 34.421875, "mean_p": 0.249177, "boundary_entropy": 0.19437, "bpb": 2.474273}}, "bpb": {"en": 2.11046, "tr": 2.31682, "fi": 2.47427}, "gini_cpc": 0.04506, "gini_cps": 0.062733, "bpic": 5.041052, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.005564579578256, "fi": 1.2178702374223473}, "inventory": {"size": 327593, "admitted": 65536, "staged": 262057, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262057, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 57.042724609375, "n_util_obs": 314, "util_count_spearman": 0.12750272369868887}, "t": 1692.3} |
| {"event": "eval", "step": 2000, "phase": 1, "bpb": {"en": 2.12351, "tr": 2.32939, "fi": 2.4843}, "bytes_per_chunk": {"en": 4.994, "tr": 5.603, "fi": 4.488}, "bpb_mean": 2.3124, "frozen_inventory": true, "t": 1693.3} |
| {"event": "train", "step": 2000, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.64912, "bpb": 2.37918, "chunk_loss": 0.030887, "F": 0.21201, "G": 0.24252, "chunks": 217.09, "bytes_per_chunk": 4.717, "boundary_entropy": 0.19952, "router_noise": 0.0, "grad_norm": 0.8373, "lr": 0.000245, "router_trainable": true, "sec_per_step": 1.02, "probe_sec": 0.0594, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04265, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.01933, "n_shadow": 4, "mean_delta_shadow": 0.16599, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.01858, "trunc_abs_frac": 0.78268, "mean_delta_full": -0.12976, "read_hit_rate": {"surface": 0.39132}, "mean_gate": {"surface": 0.1952}, "n_reads": 2706, "n_cand": 6915, "read_suppressed_frac": 0.0946, "dropout_reads_logged": 2706, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 57.084014892578125, "n_util_obs": 314, "util_count_spearman": 0.12835951439633092}, "probe_overhead_sec_total": 123.246, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.014982, "alpha": 0.03, "t": 1699.4} |
| {"event": "probe", "step": 2048, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3340.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.191921, "cps": 26.09375, "mean_p": 0.219643, "boundary_entropy": 0.187015, "bpb": 2.098679}, "tr": {"chunks": 3598.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.198752, "cps": 28.109375, "mean_p": 0.229558, "boundary_entropy": 0.240084, "bpb": 2.314748}, "fi": {"chunks": 4226.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.224656, "cps": 33.015625, "mean_p": 0.245375, "boundary_entropy": 0.197867, "bpb": 2.469209}}, "bpb": {"en": 2.09868, "tr": 2.31475, "fi": 2.46921}, "gini_cpc": 0.035466, "gini_cps": 0.052908, "bpic": 5.081691, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0355909845226965, "fi": 1.1705642672445764}, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 58.89544677734375, "n_util_obs": 319, "util_count_spearman": 0.12714377141485442}, "t": 1742.6} |
| {"event": "consolidate", "step": 2050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 65, "uncertain_staged": 262133, "t": 1756.7} |
| {"event": "train", "step": 2050, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.48876, "bpb": 2.14782, "chunk_loss": 0.030488, "F": 0.19342, "G": 0.22436, "chunks": 198.06, "bytes_per_chunk": 5.17, "boundary_entropy": 0.19803, "router_noise": 0.0, "grad_norm": 0.823, "lr": 0.0002419, "router_trainable": true, "sec_per_step": 1.166, "probe_sec": 0.0399, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02312, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": 0.09771, "n_shadow": 2, "mean_delta_shadow": 0.123, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.2977, "trunc_abs_frac": 0.18731, "mean_delta_full": 0.2652, "read_hit_rate": {"surface": 0.37996}, "mean_gate": {"surface": 0.19821}, "n_reads": 2396, "n_cand": 6306, "read_suppressed_frac": 0.0906, "dropout_reads_logged": 2396, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 59.02056884765625, "n_util_obs": 319, "util_count_spearman": 0.12853727637762755}, "probe_overhead_sec_total": 126.365, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.014391, "alpha": 0.03, "t": 1757.8} |
| {"event": "consolidate", "step": 2100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262133, "t": 1791.6} |
| {"event": "train", "step": 2100, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.54923, "bpb": 2.23507, "chunk_loss": 0.030398, "F": 0.19312, "G": 0.22384, "chunks": 197.75, "bytes_per_chunk": 5.178, "boundary_entropy": 0.19252, "router_noise": 0.0, "grad_norm": 0.8281, "lr": 0.0002388, "router_trainable": true, "sec_per_step": 0.68, "probe_sec": 0.0743, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06842, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.05864, "n_shadow": 4, "mean_delta_shadow": -0.15797, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.52775, "trunc_abs_frac": 0.5431, "mean_delta_full": -0.33267, "read_hit_rate": {"surface": 0.37341}, "mean_gate": {"surface": 0.19055}, "n_reads": 2351, "n_cand": 6296, "read_suppressed_frac": 0.1008, "dropout_reads_logged": 2351, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 60.751312255859375, "n_util_obs": 323, "util_count_spearman": 0.12677049795484951}, "probe_overhead_sec_total": 129.248, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.014197, "alpha": 0.03, "t": 1791.8} |
| {"event": "consolidate", "step": 2150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262133, "t": 1814.4} |
| {"event": "train", "step": 2150, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.50482, "bpb": 2.171, "chunk_loss": 0.030547, "F": 0.2005, "G": 0.23152, "chunks": 205.31, "bytes_per_chunk": 4.988, "boundary_entropy": 0.19583, "router_noise": 0.0, "grad_norm": 0.7374, "lr": 0.0002357, "router_trainable": true, "sec_per_step": 0.462, "probe_sec": 0.0807, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06946, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.04262, "n_shadow": 6, "mean_delta_shadow": 0.10113, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.80623, "trunc_abs_frac": 0.72562, "mean_delta_full": 0.28757, "read_hit_rate": {"surface": 0.39768}, "mean_gate": {"surface": 0.20467}, "n_reads": 2600, "n_cand": 6538, "read_suppressed_frac": 0.1031, "dropout_reads_logged": 2600, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 62.21888732910156, "n_util_obs": 328, "util_count_spearman": 0.07228419083646766}, "probe_overhead_sec_total": 131.765, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.014463, "alpha": 0.03, "t": 1814.9} |
| {"event": "consolidate", "step": 2200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262133, "t": 1853.7} |
| {"event": "probe", "step": 2200, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3253.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.186922, "cps": 25.414062, "mean_p": 0.216177, "boundary_entropy": 0.187532, "bpb": 2.088477}, "tr": {"chunks": 3641.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.201127, "cps": 28.445312, "mean_p": 0.22733, "boundary_entropy": 0.232501, "bpb": 2.294765}, "fi": {"chunks": 4157.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.220988, "cps": 32.476562, "mean_p": 0.243885, "boundary_entropy": 0.199894, "bpb": 2.441941}}, "bpb": {"en": 2.08848, "tr": 2.29477, "fi": 2.44194}, "gini_cpc": 0.037289, "gini_cps": 0.054535, "bpic": 5.133653, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0759948295319168, "fi": 1.182246938305829}, "inventory": {"size": 326656, "admitted": 65536, "staged": 261120, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 261120, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 63.44615173339844, "n_util_obs": 329, "util_count_spearman": 0.04465054589120959}, "t": 1854.3} |
| {"event": "eval", "step": 2200, "phase": 1, "bpb": {"en": 2.09864, "tr": 2.30208, "fi": 2.45023}, "bytes_per_chunk": {"en": 5.138, "tr": 5.347, "fi": 4.799}, "bpb_mean": 2.28365, "frozen_inventory": true, "t": 1855.1} |
| {"event": "train", "step": 2200, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.48362, "bpb": 2.14042, "chunk_loss": 0.030424, "F": 0.19583, "G": 0.22471, "chunks": 200.53, "bytes_per_chunk": 5.106, "boundary_entropy": 0.19145, "router_noise": 0.0, "grad_norm": 0.6867, "lr": 0.0002325, "router_trainable": true, "sec_per_step": 0.905, "probe_sec": 0.0633, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03648, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.21969, "n_shadow": 7, "mean_delta_shadow": -0.09267, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -0.73747, "trunc_abs_frac": 3.95629, "mean_delta_full": -0.13347, "read_hit_rate": {"surface": 0.36899}, "mean_gate": {"surface": 0.19705}, "n_reads": 2356, "n_cand": 6385, "read_suppressed_frac": 0.0993, "dropout_reads_logged": 2356, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 63.48210144042969, "n_util_obs": 330, "util_count_spearman": 0.04935466876012926}, "probe_overhead_sec_total": 134.813, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.014751, "alpha": 0.03, "t": 1860.1} |
| {"event": "consolidate", "step": 2250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262133, "t": 1902.7} |
| {"event": "train", "step": 2250, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.50628, "bpb": 2.1731, "chunk_loss": 0.030503, "F": 0.19827, "G": 0.22793, "chunks": 203.03, "bytes_per_chunk": 5.044, "boundary_entropy": 0.18736, "router_noise": 0.0, "grad_norm": 0.8084, "lr": 0.0002293, "router_trainable": true, "sec_per_step": 0.871, "probe_sec": 0.0472, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03037, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.16118, "n_shadow": 8, "mean_delta_shadow": -0.05367, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.79579, "trunc_abs_frac": 0.9048, "mean_delta_full": -0.21233, "read_hit_rate": {"surface": 0.36164}, "mean_gate": {"surface": 0.19939}, "n_reads": 2338, "n_cand": 6465, "read_suppressed_frac": 0.1065, "dropout_reads_logged": 2338, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 63.59532165527344, "n_util_obs": 319, "util_count_spearman": 0.03130454069715388}, "probe_overhead_sec_total": 138.094, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.014384, "alpha": 0.03, "t": 1903.7} |
| {"event": "consolidate", "step": 2300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262133, "t": 1942.5} |
| {"event": "train", "step": 2300, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.51564, "bpb": 2.18661, "chunk_loss": 0.0304, "F": 0.19501, "G": 0.22449, "chunks": 199.69, "bytes_per_chunk": 5.128, "boundary_entropy": 0.18356, "router_noise": 0.0, "grad_norm": 0.7427, "lr": 0.000226, "router_trainable": true, "sec_per_step": 0.789, "probe_sec": 0.0678, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05587, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.54118, "n_shadow": 4, "mean_delta_shadow": -0.01001, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.8016, "trunc_abs_frac": 1.06087, "mean_delta_full": 0.50845, "read_hit_rate": {"surface": 0.35042}, "mean_gate": {"surface": 0.19604}, "n_reads": 2228, "n_cand": 6358, "read_suppressed_frac": 0.0866, "dropout_reads_logged": 2228, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 65.26704406738281, "n_util_obs": 325, "util_count_spearman": 0.024088414901654906}, "probe_overhead_sec_total": 140.956, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.014218, "alpha": 0.03, "t": 1943.1} |
| {"event": "consolidate", "step": 2350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 74, "uncertain_staged": 262133, "t": 1962.9} |
| {"event": "train", "step": 2350, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.56469, "bpb": 2.25737, "chunk_loss": 0.030501, "F": 0.19968, "G": 0.23, "chunks": 204.47, "bytes_per_chunk": 5.008, "boundary_entropy": 0.18691, "router_noise": 0.0, "grad_norm": 0.7653, "lr": 0.0002226, "router_trainable": true, "sec_per_step": 0.414, "probe_sec": 0.0423, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02421, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.02861, "n_shadow": 4, "mean_delta_shadow": 0.0668, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.86766, "trunc_abs_frac": 0.85493, "mean_delta_full": 0.27557, "read_hit_rate": {"surface": 0.36769}, "mean_gate": {"surface": 0.1917}, "n_reads": 2394, "n_cand": 6511, "read_suppressed_frac": 0.0944, "dropout_reads_logged": 2394, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 67.1214599609375, "n_util_obs": 333, "util_count_spearman": 0.066065074544468}, "probe_overhead_sec_total": 143.493, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.014503, "alpha": 0.03, "t": 1963.8} |
| {"event": "consolidate", "step": 2400, "retired": 1, "split_retired": 0, "merge_admitted": 0, "admitted": 1, "swapped": 6, "uncertain_staged": 262133, "t": 2001.1} |
| {"event": "probe", "step": 2400, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3268.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.187784, "cps": 25.53125, "mean_p": 0.214983, "boundary_entropy": 0.175701, "bpb": 2.06226}, "tr": {"chunks": 3552.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.196211, "cps": 27.75, "mean_p": 0.222824, "boundary_entropy": 0.216715, "bpb": 2.260085}, "fi": {"chunks": 4340.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.230716, "cps": 33.90625, "mean_p": 0.245029, "boundary_entropy": 0.178499, "bpb": 2.39663}}, "bpb": {"en": 2.06226, "tr": 2.26009, "fi": 2.39663}, "gini_cpc": 0.046561, "gini_cps": 0.064038, "bpic": 5.083513, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0448753362964314, "fi": 1.2286266134941357}, "inventory": {"size": 327673, "admitted": 65536, "staged": 262137, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262137, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 68.87753295898438, "n_util_obs": 341, "util_count_spearman": 0.06125252541149471}, "t": 2001.9} |
| {"event": "eval", "step": 2400, "phase": 1, "bpb": {"en": 2.07386, "tr": 2.26773, "fi": 2.40742}, "bytes_per_chunk": {"en": 5.019, "tr": 5.696, "fi": 4.6}, "bpb_mean": 2.24967, "frozen_inventory": true, "t": 2003.0} |
| {"event": "phase_transition", "step": 2400, "phase": 2, "objective": "global", "router_trainable": true, "encoder_trainable": true, "frozen_router_params": 0, "t": 2007.9} |
| {"event": "train", "step": 2400, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.50374, "bpb": 2.16944, "chunk_loss": 0.03059, "F": 0.20316, "G": 0.23121, "chunks": 208.03, "bytes_per_chunk": 4.922, "boundary_entropy": 0.17993, "router_noise": 0.0, "grad_norm": 0.7546, "lr": 0.0002192, "router_trainable": true, "sec_per_step": 0.909, "probe_sec": 0.1037, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05877, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.44623, "n_shadow": 8, "mean_delta_shadow": -0.04607, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 8, "trunc_ratio_sum": 1.65842, "trunc_abs_frac": 1.50519, "mean_delta_full": 0.06546, "read_hit_rate": {"surface": 0.35789}, "mean_gate": {"surface": 0.19881}, "n_reads": 2371, "n_cand": 6625, "read_suppressed_frac": 0.0966, "dropout_reads_logged": 2371, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 68.91371154785156, "n_util_obs": 341, "util_count_spearman": 0.06174308136008531}, "probe_overhead_sec_total": 146.904, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.014878, "alpha": 0.03, "t": 2009.3} |
| {"event": "consolidate", "step": 2450, "retired": 1, "split_retired": 0, "merge_admitted": 0, "admitted": 1, "swapped": 11, "uncertain_staged": 262133, "t": 2054.1} |
| {"event": "train", "step": 2450, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.4623, "bpb": 2.10965, "chunk_loss": 0.030435, "F": 0.19611, "G": 0.22518, "chunks": 200.81, "bytes_per_chunk": 5.099, "boundary_entropy": 0.18224, "router_noise": 0.0, "grad_norm": 0.7039, "lr": 0.0002158, "router_trainable": true, "sec_per_step": 0.911, "probe_sec": 0.049, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02896, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": -0.06321, "n_shadow": 5, "mean_delta_shadow": -0.08491, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.30889, "trunc_abs_frac": 1.51632, "mean_delta_full": -0.13681, "read_hit_rate": {"surface": 0.36159}, "mean_gate": {"surface": 0.19793}, "n_reads": 2312, "n_cand": 6394, "read_suppressed_frac": 0.1008, "dropout_reads_logged": 2312, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 70.72555541992188, "n_util_obs": 343, "util_count_spearman": 0.05638819012215821}, "probe_overhead_sec_total": 150.251, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.014636, "alpha": 0.03, "t": 2054.8} |
| {"event": "consolidate", "step": 2500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 37, "uncertain_staged": 262133, "t": 2092.1} |
| {"event": "train", "step": 2500, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.43903, "bpb": 2.07608, "chunk_loss": 0.030569, "F": 0.20407, "G": 0.22911, "chunks": 208.97, "bytes_per_chunk": 4.9, "boundary_entropy": 0.17375, "router_noise": 0.0, "grad_norm": 0.7462, "lr": 0.0002123, "router_trainable": true, "sec_per_step": 0.755, "probe_sec": 0.3911, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05564, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": 0.22504, "n_shadow": 5, "mean_delta_shadow": -0.11158, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.14824, "trunc_abs_frac": 1.31142, "mean_delta_full": 0.11127, "read_hit_rate": {"surface": 0.38918}, "mean_gate": {"surface": 0.19248}, "n_reads": 2590, "n_cand": 6655, "read_suppressed_frac": 0.1012, "dropout_reads_logged": 2590, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 72.68185424804688, "n_util_obs": 351, "util_count_spearman": 0.08873767586238433}, "probe_overhead_sec_total": 153.983, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.014401, "alpha": 0.03, "t": 2092.6} |
| {"event": "consolidate", "step": 2550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262133, "t": 2109.3} |
| {"event": "train", "step": 2550, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.46653, "bpb": 2.11575, "chunk_loss": 0.030469, "F": 0.19922, "G": 0.22442, "chunks": 204.0, "bytes_per_chunk": 5.02, "boundary_entropy": 0.17449, "router_noise": 0.0, "grad_norm": 0.6976, "lr": 0.0002088, "router_trainable": true, "sec_per_step": 0.339, "probe_sec": 0.0609, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05392, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.23313, "n_shadow": 8, "mean_delta_shadow": 0.14498, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.58696, "trunc_abs_frac": 0.54877, "mean_delta_full": 0.46674, "read_hit_rate": {"surface": 0.40702}, "mean_gate": {"surface": 0.18736}, "n_reads": 2644, "n_cand": 6496, "read_suppressed_frac": 0.0998, "dropout_reads_logged": 2644, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 74.90078735351562, "n_util_obs": 357, "util_count_spearman": 0.07239031775033014}, "probe_overhead_sec_total": 156.445, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.014298, "alpha": 0.03, "t": 2109.5} |
| {"event": "consolidate", "step": 2600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 920, "uncertain_staged": 262133, "t": 2150.8} |
| {"event": "probe", "step": 2600, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3283.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.188646, "cps": 25.648438, "mean_p": 0.210677, "boundary_entropy": 0.161172, "bpb": 2.043115}, "tr": {"chunks": 3551.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.196155, "cps": 27.742188, "mean_p": 0.216714, "boundary_entropy": 0.200398, "bpb": 2.246011}, "fi": {"chunks": 4323.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.229812, "cps": 33.773438, "mean_p": 0.241565, "boundary_entropy": 0.166997, "bpb": 2.371998}}, "bpb": {"en": 2.04311, "tr": 2.24601, "fi": 2.372}, "gini_cpc": 0.044653, "gini_cps": 0.062143, "bpic": 5.084879, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0398084881635359, "fi": 1.2182224245724496}, "inventory": {"size": 326760, "admitted": 65536, "staged": 261224, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 261224, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 77.27220153808594, "n_util_obs": 361, "util_count_spearman": 0.08403989379306244}, "t": 2151.5} |
| {"event": "eval", "step": 2600, "phase": 2, "bpb": {"en": 2.056, "tr": 2.25714, "fi": 2.3793}, "bytes_per_chunk": {"en": 5.04, "tr": 5.553, "fi": 4.598}, "bpb_mean": 2.23081, "frozen_inventory": true, "t": 2152.6} |
| {"event": "train", "step": 2600, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.42226, "bpb": 2.05188, "chunk_loss": 0.030392, "F": 0.19693, "G": 0.21985, "chunks": 201.66, "bytes_per_chunk": 5.078, "boundary_entropy": 0.16591, "router_noise": 0.0, "grad_norm": 0.7405, "lr": 0.0002053, "router_trainable": true, "sec_per_step": 0.961, "probe_sec": 0.0551, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02546, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.04525, "n_shadow": 4, "mean_delta_shadow": -0.29495, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.20167, "trunc_abs_frac": 0.20167, "mean_delta_full": -0.2627, "read_hit_rate": {"surface": 0.43794}, "mean_gate": {"surface": 0.18137}, "n_reads": 2812, "n_cand": 6421, "read_suppressed_frac": 0.0928, "dropout_reads_logged": 2812, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 77.31510925292969, "n_util_obs": 361, "util_count_spearman": 0.08400279150987441}, "probe_overhead_sec_total": 159.562, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.01426, "alpha": 0.03, "t": 2157.6} |
| {"event": "consolidate", "step": 2650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 602, "uncertain_staged": 262133, "t": 2189.4} |
| {"event": "train", "step": 2650, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.47367, "bpb": 2.12605, "chunk_loss": 0.030373, "F": 0.19443, "G": 0.22007, "chunks": 199.09, "bytes_per_chunk": 5.143, "boundary_entropy": 0.17888, "router_noise": 0.0, "grad_norm": 0.7442, "lr": 0.0002017, "router_trainable": true, "sec_per_step": 0.654, "probe_sec": 0.0468, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02698, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.23063, "n_shadow": 6, "mean_delta_shadow": 0.11948, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.75772, "trunc_abs_frac": 0.75772, "mean_delta_full": 1.98035, "read_hit_rate": {"surface": 0.44013}, "mean_gate": {"surface": 0.17906}, "n_reads": 2790, "n_cand": 6339, "read_suppressed_frac": 0.0957, "dropout_reads_logged": 2790, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 79.65852355957031, "n_util_obs": 367, "util_count_spearman": 0.10337183204365526}, "probe_overhead_sec_total": 162.858, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.015397, "alpha": 0.03, "t": 2190.3} |
| {"event": "consolidate", "step": 2700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262133, "t": 2228.9} |
| {"event": "train", "step": 2700, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.44975, "bpb": 2.09155, "chunk_loss": 0.030494, "F": 0.20059, "G": 0.22548, "chunks": 205.41, "bytes_per_chunk": 4.985, "boundary_entropy": 0.16911, "router_noise": 0.0, "grad_norm": 0.6982, "lr": 0.0001981, "router_trainable": true, "sec_per_step": 0.786, "probe_sec": 0.0432, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02391, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.00363, "n_shadow": 4, "mean_delta_shadow": -0.17068, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.57538, "trunc_abs_frac": 1.76933, "mean_delta_full": -0.3085, "read_hit_rate": {"surface": 0.46705}, "mean_gate": {"surface": 0.17156}, "n_reads": 3055, "n_cand": 6541, "read_suppressed_frac": 0.0992, "dropout_reads_logged": 3055, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 82.06648254394531, "n_util_obs": 372, "util_count_spearman": 0.10125456576985677}, "probe_overhead_sec_total": 165.953, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.015266, "alpha": 0.03, "t": 2229.6} |
| {"event": "consolidate", "step": 2750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262133, "t": 2267.6} |
| {"event": "train", "step": 2750, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.4623, "bpb": 2.10965, "chunk_loss": 0.030577, "F": 0.20279, "G": 0.2265, "chunks": 207.66, "bytes_per_chunk": 4.931, "boundary_entropy": 0.16975, "router_noise": 0.0, "grad_norm": 0.7147, "lr": 0.0001945, "router_trainable": true, "sec_per_step": 0.776, "probe_sec": 0.0592, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02517, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.10486, "n_shadow": 6, "mean_delta_shadow": -0.47061, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 1.11922, "trunc_abs_frac": 1.29963, "mean_delta_full": -0.23674, "read_hit_rate": {"surface": 0.48132}, "mean_gate": {"surface": 0.17138}, "n_reads": 3183, "n_cand": 6613, "read_suppressed_frac": 0.0911, "dropout_reads_logged": 3183, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 84.49424743652344, "n_util_obs": 379, "util_count_spearman": 0.11343810240568825}, "probe_overhead_sec_total": 169.351, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.014979, "alpha": 0.03, "t": 2268.4} |
| {"event": "consolidate", "step": 2800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262133, "t": 2308.7} |
| {"event": "probe", "step": 2800, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3310.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.190197, "cps": 25.859375, "mean_p": 0.212142, "boundary_entropy": 0.16036, "bpb": 2.024954}, "tr": {"chunks": 3553.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.196266, "cps": 27.757812, "mean_p": 0.214393, "boundary_entropy": 0.19837, "bpb": 2.200203}, "fi": {"chunks": 4315.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.229387, "cps": 33.710938, "mean_p": 0.241913, "boundary_entropy": 0.165026, "bpb": 2.328486}}, "bpb": {"en": 2.02495, "tr": 2.2002, "fi": 2.32849}, "gini_cpc": 0.042424, "gini_cps": 0.059939, "bpic": 5.075327, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0319075321427755, "fi": 1.206049250285998}, "inventory": {"size": 326656, "admitted": 65536, "staged": 261120, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 261120, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 86.81578063964844, "n_util_obs": 380, "util_count_spearman": 0.10075571322787062}, "t": 2309.3} |
| {"event": "eval", "step": 2800, "phase": 2, "bpb": {"en": 2.03822, "tr": 2.20999, "fi": 2.33673}, "bytes_per_chunk": {"en": 4.993, "tr": 5.575, "fi": 4.654}, "bpb_mean": 2.19498, "frozen_inventory": true, "t": 2310.2} |
| {"event": "train", "step": 2800, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.41126, "bpb": 2.03602, "chunk_loss": 0.030459, "F": 0.19977, "G": 0.22346, "chunks": 204.56, "bytes_per_chunk": 5.006, "boundary_entropy": 0.16582, "router_noise": 0.0, "grad_norm": 0.7122, "lr": 0.0001909, "router_trainable": true, "sec_per_step": 0.943, "probe_sec": 0.0561, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03208, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": -0.01345, "n_shadow": 7, "mean_delta_shadow": -0.12257, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 1.2197, "trunc_abs_frac": 1.2197, "mean_delta_full": 0.23169, "read_hit_rate": {"surface": 0.49232}, "mean_gate": {"surface": 0.17424}, "n_reads": 3207, "n_cand": 6514, "read_suppressed_frac": 0.0839, "dropout_reads_logged": 3207, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 86.86471557617188, "n_util_obs": 380, "util_count_spearman": 0.10103783186222103}, "probe_overhead_sec_total": 172.701, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.014794, "alpha": 0.03, "t": 2315.5} |
| {"event": "consolidate", "step": 2850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262133, "t": 2356.9} |
| {"event": "train", "step": 2850, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.46351, "bpb": 2.1114, "chunk_loss": 0.030643, "F": 0.20825, "G": 0.23105, "chunks": 213.25, "bytes_per_chunk": 4.802, "boundary_entropy": 0.16159, "router_noise": 0.0, "grad_norm": 0.7255, "lr": 0.0001872, "router_trainable": true, "sec_per_step": 0.849, "probe_sec": 0.1036, "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": 13, "mean_delta_loo": -0.069, "n_shadow": 1, "mean_delta_shadow": -0.87488, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 1.33443, "trunc_abs_frac": 2.06878, "mean_delta_full": 0.03714, "read_hit_rate": {"surface": 0.48778}, "mean_gate": {"surface": 0.16724}, "n_reads": 3313, "n_cand": 6792, "read_suppressed_frac": 0.0996, "dropout_reads_logged": 3313, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 89.3717041015625, "n_util_obs": 381, "util_count_spearman": 0.09764530738965221}, "probe_overhead_sec_total": 175.805, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.014517, "alpha": 0.03, "t": 2358.0} |
| {"event": "consolidate", "step": 2900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262133, "t": 2387.5} |
| {"event": "train", "step": 2900, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.38994, "bpb": 2.00525, "chunk_loss": 0.030332, "F": 0.19373, "G": 0.2172, "chunks": 198.38, "bytes_per_chunk": 5.162, "boundary_entropy": 0.16272, "router_noise": 0.0, "grad_norm": 0.6439, "lr": 0.0001835, "router_trainable": true, "sec_per_step": 0.593, "probe_sec": 0.0465, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02524, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": -0.06359, "n_shadow": 5, "mean_delta_shadow": -0.08241, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.73212, "trunc_abs_frac": 1.88028, "mean_delta_full": -0.16132, "read_hit_rate": {"surface": 0.49509}, "mean_gate": {"surface": 0.16845}, "n_reads": 3127, "n_cand": 6316, "read_suppressed_frac": 0.0982, "dropout_reads_logged": 3127, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 91.88484191894531, "n_util_obs": 385, "util_count_spearman": 0.10014870092504607}, "probe_overhead_sec_total": 178.593, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.014923, "alpha": 0.03, "t": 2387.7} |
| {"event": "consolidate", "step": 2950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262133, "t": 2411.1} |
| {"event": "train", "step": 2950, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.54497, "bpb": 2.22892, "chunk_loss": 0.030625, "F": 0.20435, "G": 0.22709, "chunks": 209.25, "bytes_per_chunk": 4.894, "boundary_entropy": 0.16489, "router_noise": 0.0, "grad_norm": 0.7197, "lr": 0.0001799, "router_trainable": true, "sec_per_step": 0.479, "probe_sec": 0.0303, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01898, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.07122, "n_shadow": 6, "mean_delta_shadow": -0.04867, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.90905, "trunc_abs_frac": 1.23917, "mean_delta_full": -0.18286, "read_hit_rate": {"surface": 0.47524}, "mean_gate": {"surface": 0.16885}, "n_reads": 3167, "n_cand": 6664, "read_suppressed_frac": 0.1007, "dropout_reads_logged": 3167, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 94.31953430175781, "n_util_obs": 388, "util_count_spearman": 0.092742024950511}, "probe_overhead_sec_total": 181.038, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.014787, "alpha": 0.03, "t": 2411.6} |
| {"event": "consolidate", "step": 3000, "retired": 1, "split_retired": 0, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262133, "t": 2440.9} |
| {"event": "probe", "step": 3000, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3346.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.192266, "cps": 26.140625, "mean_p": 0.210484, "boundary_entropy": 0.148836, "bpb": 2.003389}, "tr": {"chunks": 3651.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.201679, "cps": 28.523438, "mean_p": 0.212178, "boundary_entropy": 0.183049, "bpb": 2.175503}, "fi": {"chunks": 4273.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.227154, "cps": 33.382812, "mean_p": 0.237332, "boundary_entropy": 0.157596, "bpb": 2.310823}}, "bpb": {"en": 2.00339, "tr": 2.1755, "fi": 2.31082}, "gini_cpc": 0.037448, "gini_cps": 0.054836, "bpic": 5.033895, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0489612983340764, "fi": 1.1814604635286872}, "inventory": {"size": 326655, "admitted": 65536, "staged": 261119, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 261119, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 96.63943481445312, "n_util_obs": 388, "util_count_spearman": 0.0925022837438672}, "t": 2441.6} |
| {"event": "eval", "step": 3000, "phase": 2, "bpb": {"en": 2.01742, "tr": 2.1861, "fi": 2.3163}, "bytes_per_chunk": {"en": 4.953, "tr": 5.374, "fi": 4.674}, "bpb_mean": 2.17327, "frozen_inventory": true, "t": 2442.5} |
| {"event": "train", "step": 3000, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.49089, "bpb": 2.1509, "chunk_loss": 0.030492, "F": 0.20053, "G": 0.21947, "chunks": 205.34, "bytes_per_chunk": 4.987, "boundary_entropy": 0.15357, "router_noise": 0.0, "grad_norm": 0.7312, "lr": 0.0001761, "router_trainable": true, "sec_per_step": 0.783, "probe_sec": 0.0652, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02759, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": -0.08373, "n_shadow": 5, "mean_delta_shadow": 0.0793, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 0.5485, "trunc_abs_frac": 1.69962, "mean_delta_full": -0.36287, "read_hit_rate": {"surface": 0.48096}, "mean_gate": {"surface": 0.16616}, "n_reads": 3145, "n_cand": 6539, "read_suppressed_frac": 0.0995, "dropout_reads_logged": 3145, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 96.68742370605469, "n_util_obs": 388, "util_count_spearman": 0.0929602138075306}, "probe_overhead_sec_total": 183.586, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.014565, "alpha": 0.03, "t": 2450.7} |
| {"event": "consolidate", "step": 3050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 595, "uncertain_staged": 262133, "t": 2489.2} |
| {"event": "train", "step": 3050, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.44129, "bpb": 2.07935, "chunk_loss": 0.030493, "F": 0.19925, "G": 0.2203, "chunks": 204.03, "bytes_per_chunk": 5.019, "boundary_entropy": 0.15424, "router_noise": 0.0, "grad_norm": 0.667, "lr": 0.0001724, "router_trainable": true, "sec_per_step": 0.794, "probe_sec": 0.0551, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03278, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.09561, "n_shadow": 7, "mean_delta_shadow": 0.02071, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.6574, "trunc_abs_frac": 0.70088, "mean_delta_full": 0.53992, "read_hit_rate": {"surface": 0.49823}, "mean_gate": {"surface": 0.15949}, "n_reads": 3237, "n_cand": 6497, "read_suppressed_frac": 0.0989, "dropout_reads_logged": 3237, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 99.23139953613281, "n_util_obs": 392, "util_count_spearman": 0.10289357391415702}, "probe_overhead_sec_total": 186.842, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.014682, "alpha": 0.03, "t": 2490.4} |
| {"event": "consolidate", "step": 3100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 690, "uncertain_staged": 262133, "t": 2534.0} |
| {"event": "train", "step": 3100, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.48762, "bpb": 2.14618, "chunk_loss": 0.030403, "F": 0.19638, "G": 0.21875, "chunks": 201.09, "bytes_per_chunk": 5.092, "boundary_entropy": 0.15674, "router_noise": 0.0, "grad_norm": 0.6983, "lr": 0.0001687, "router_trainable": true, "sec_per_step": 0.895, "probe_sec": 0.0608, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02493, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.02912, "n_shadow": 8, "mean_delta_shadow": 0.03837, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.05689, "trunc_abs_frac": 0.57056, "mean_delta_full": -0.29998, "read_hit_rate": {"surface": 0.4868}, "mean_gate": {"surface": 0.16393}, "n_reads": 3117, "n_cand": 6403, "read_suppressed_frac": 0.1007, "dropout_reads_logged": 3117, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 101.76332092285156, "n_util_obs": 394, "util_count_spearman": 0.11494617419230979}, "probe_overhead_sec_total": 189.863, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.014472, "alpha": 0.03, "t": 2535.2} |
| {"event": "consolidate", "step": 3150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262133, "t": 2576.3} |
| {"event": "train", "step": 3150, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.45239, "bpb": 2.09536, "chunk_loss": 0.030393, "F": 0.19635, "G": 0.21662, "chunks": 201.06, "bytes_per_chunk": 5.093, "boundary_entropy": 0.15336, "router_noise": 0.0, "grad_norm": 0.6953, "lr": 0.000165, "router_trainable": true, "sec_per_step": 0.838, "probe_sec": 0.0519, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02283, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.27368, "n_shadow": 4, "mean_delta_shadow": -0.02833, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 1.30794, "trunc_abs_frac": 17.54698, "mean_delta_full": 0.60483, "read_hit_rate": {"surface": 0.49922}, "mean_gate": {"surface": 0.15629}, "n_reads": 3196, "n_cand": 6402, "read_suppressed_frac": 0.1001, "dropout_reads_logged": 3196, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 104.45606994628906, "n_util_obs": 401, "util_count_spearman": 0.1360215386945977}, "probe_overhead_sec_total": 192.812, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.013286, "alpha": 0.03, "t": 2577.1} |
| {"event": "consolidate", "step": 3200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262133, "t": 2614.1} |
| {"event": "probe", "step": 3200, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3289.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.18899, "cps": 25.695312, "mean_p": 0.208271, "boundary_entropy": 0.145697, "bpb": 1.98855}, "tr": {"chunks": 3529.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.19494, "cps": 27.570312, "mean_p": 0.211439, "boundary_entropy": 0.179533, "bpb": 2.156106}, "fi": {"chunks": 4357.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.23162, "cps": 34.039062, "mean_p": 0.238716, "boundary_entropy": 0.145665, "bpb": 2.283631}}, "bpb": {"en": 1.98855, "tr": 2.15611, "fi": 2.28363}, "gini_cpc": 0.046169, "gini_cps": 0.063714, "bpic": 5.076689, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0314812874345405, "fi": 1.2255637962682637}, "inventory": {"size": 326656, "admitted": 65536, "staged": 261120, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 261120, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 106.8275146484375, "n_util_obs": 399, "util_count_spearman": 0.12854999891846497}, "t": 2614.8} |
| {"event": "eval", "step": 3200, "phase": 2, "bpb": {"en": 2.00084, "tr": 2.16371, "fi": 2.29187}, "bytes_per_chunk": {"en": 5.09, "tr": 5.595, "fi": 4.57}, "bpb_mean": 2.15214, "frozen_inventory": true, "t": 2615.7} |
| {"event": "train", "step": 3200, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.43215, "bpb": 2.06616, "chunk_loss": 0.030406, "F": 0.19754, "G": 0.21899, "chunks": 202.28, "bytes_per_chunk": 5.062, "boundary_entropy": 0.14752, "router_noise": 0.0, "grad_norm": 0.6326, "lr": 0.0001613, "router_trainable": true, "sec_per_step": 0.884, "probe_sec": 0.0293, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01996, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.18531, "n_shadow": 4, "mean_delta_shadow": -0.24803, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.62179, "trunc_abs_frac": 1.15351, "mean_delta_full": -0.19202, "read_hit_rate": {"surface": 0.50862}, "mean_gate": {"surface": 0.15692}, "n_reads": 3276, "n_cand": 6441, "read_suppressed_frac": 0.1056, "dropout_reads_logged": 3276, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 106.87750244140625, "n_util_obs": 400, "util_count_spearman": 0.1267332905361843}, "probe_overhead_sec_total": 195.441, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.013187, "alpha": 0.03, "t": 2621.3} |
| {"event": "consolidate", "step": 3250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262133, "t": 2665.6} |
| {"event": "train", "step": 3250, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.43451, "bpb": 2.06957, "chunk_loss": 0.03048, "F": 0.19907, "G": 0.21953, "chunks": 203.84, "bytes_per_chunk": 5.023, "boundary_entropy": 0.15147, "router_noise": 0.0, "grad_norm": 0.6998, "lr": 0.0001576, "router_trainable": true, "sec_per_step": 0.9, "probe_sec": 0.0659, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02603, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.14481, "n_shadow": 8, "mean_delta_shadow": 0.03739, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.2722, "trunc_abs_frac": 1.03403, "mean_delta_full": -0.13883, "read_hit_rate": {"surface": 0.52257}, "mean_gate": {"surface": 0.16497}, "n_reads": 3392, "n_cand": 6491, "read_suppressed_frac": 0.0985, "dropout_reads_logged": 3392, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 109.47886657714844, "n_util_obs": 403, "util_count_spearman": 0.12871408072813717}, "probe_overhead_sec_total": 198.736, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.013627, "alpha": 0.03, "t": 2666.3} |
| {"event": "consolidate", "step": 3300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262133, "t": 2700.4} |
| {"event": "train", "step": 3300, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.41449, "bpb": 2.04068, "chunk_loss": 0.030534, "F": 0.2009, "G": 0.22145, "chunks": 205.72, "bytes_per_chunk": 4.978, "boundary_entropy": 0.14872, "router_noise": 0.0, "grad_norm": 0.6765, "lr": 0.0001539, "router_trainable": true, "sec_per_step": 0.698, "probe_sec": 0.0621, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03058, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": -0.24959, "n_shadow": 7, "mean_delta_shadow": -0.04311, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.63005, "trunc_abs_frac": 0.78578, "mean_delta_full": -0.41988, "read_hit_rate": {"surface": 0.51977}, "mean_gate": {"surface": 0.15744}, "n_reads": 3405, "n_cand": 6551, "read_suppressed_frac": 0.0993, "dropout_reads_logged": 3405, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 112.10551452636719, "n_util_obs": 405, "util_count_spearman": 0.12878643912506477}, "probe_overhead_sec_total": 201.691, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.013055, "alpha": 0.03, "t": 2701.2} |
| {"event": "consolidate", "step": 3350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262133, "t": 2733.1} |
| {"event": "train", "step": 3350, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.35585, "bpb": 1.95608, "chunk_loss": 0.030411, "F": 0.19766, "G": 0.21682, "chunks": 202.41, "bytes_per_chunk": 5.059, "boundary_entropy": 0.14588, "router_noise": 0.0, "grad_norm": 0.7155, "lr": 0.0001501, "router_trainable": true, "sec_per_step": 0.655, "probe_sec": 0.0544, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02536, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": 0.25281, "n_shadow": 9, "mean_delta_shadow": 0.01033, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 4.41053, "trunc_abs_frac": 0.44842, "mean_delta_full": -0.0174, "read_hit_rate": {"surface": 0.5294}, "mean_gate": {"surface": 0.15349}, "n_reads": 3412, "n_cand": 6445, "read_suppressed_frac": 0.0988, "dropout_reads_logged": 3412, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 114.69361877441406, "n_util_obs": 407, "util_count_spearman": 0.13021287097278}, "probe_overhead_sec_total": 204.922, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.013634, "alpha": 0.03, "t": 2734.0} |
| {"event": "consolidate", "step": 3400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262133, "t": 2773.8} |
| {"event": "probe", "step": 3400, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3283.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.188646, "cps": 25.648438, "mean_p": 0.20718, "boundary_entropy": 0.139438, "bpb": 1.974896}, "tr": {"chunks": 3537.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.195382, "cps": 27.632812, "mean_p": 0.208079, "boundary_entropy": 0.173487, "bpb": 2.141629}, "fi": {"chunks": 4289.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.228005, "cps": 33.507812, "mean_p": 0.235123, "boundary_entropy": 0.140767, "bpb": 2.260131}}, "bpb": {"en": 1.9749, "tr": 2.14163, "fi": 2.26013}, "gini_cpc": 0.042873, "gini_cps": 0.060371, "bpic": 5.10685, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0357089897590612, "fi": 1.2086412165142808}, "inventory": {"size": 326656, "admitted": 65536, "staged": 261120, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 261120, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 117.30026245117188, "n_util_obs": 412, "util_count_spearman": 0.13869928182679145}, "t": 2774.3} |
| {"event": "eval", "step": 3400, "phase": 2, "bpb": {"en": 1.98833, "tr": 2.15166, "fi": 2.26585}, "bytes_per_chunk": {"en": 5.024, "tr": 5.484, "fi": 4.623}, "bpb_mean": 2.13528, "frozen_inventory": true, "t": 2774.9} |
| {"event": "train", "step": 3400, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.42221, "bpb": 2.05182, "chunk_loss": 0.030388, "F": 0.19647, "G": 0.21539, "chunks": 201.19, "bytes_per_chunk": 5.09, "boundary_entropy": 0.14405, "router_noise": 0.0, "grad_norm": 0.7284, "lr": 0.0001465, "router_trainable": true, "sec_per_step": 0.914, "probe_sec": 0.0626, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02953, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.03148, "n_shadow": 7, "mean_delta_shadow": 0.1283, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 1656.53125, "trunc_abs_frac": 2.99787, "mean_delta_full": 6e-05, "read_hit_rate": {"surface": 0.51561}, "mean_gate": {"surface": 0.16013}, "n_reads": 3303, "n_cand": 6406, "read_suppressed_frac": 0.0999, "dropout_reads_logged": 3303, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 117.35066223144531, "n_util_obs": 412, "util_count_spearman": 0.1388070344560213}, "probe_overhead_sec_total": 207.692, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.013236, "alpha": 0.03, "t": 2779.7} |
| {"event": "consolidate", "step": 3450, "retired": 1, "split_retired": 0, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262133, "t": 2816.4} |
| {"event": "train", "step": 3450, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.3883, "bpb": 2.00289, "chunk_loss": 0.030382, "F": 0.19742, "G": 0.21668, "chunks": 202.16, "bytes_per_chunk": 5.065, "boundary_entropy": 0.14565, "router_noise": 0.0, "grad_norm": 0.7688, "lr": 0.0001428, "router_trainable": true, "sec_per_step": 0.743, "probe_sec": 0.0523, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04137, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.1675, "n_shadow": 7, "mean_delta_shadow": -0.48965, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.64786, "trunc_abs_frac": 0.53616, "mean_delta_full": 0.04606, "read_hit_rate": {"surface": 0.53394}, "mean_gate": {"surface": 0.15856}, "n_reads": 3437, "n_cand": 6437, "read_suppressed_frac": 0.0989, "dropout_reads_logged": 3437, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 120.00294494628906, "n_util_obs": 413, "util_count_spearman": 0.14828723623622497}, "probe_overhead_sec_total": 210.715, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.013082, "alpha": 0.03, "t": 2816.8} |
| {"event": "consolidate", "step": 3500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262133, "t": 2855.1} |
| {"event": "train", "step": 3500, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.36734, "bpb": 1.97265, "chunk_loss": 0.030319, "F": 0.19199, "G": 0.21017, "chunks": 196.59, "bytes_per_chunk": 5.209, "boundary_entropy": 0.14047, "router_noise": 0.0, "grad_norm": 0.6605, "lr": 0.0001391, "router_trainable": true, "sec_per_step": 0.788, "probe_sec": 0.1134, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05503, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.07058, "n_shadow": 7, "mean_delta_shadow": -0.08322, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 1.17924, "trunc_abs_frac": 0.59536, "mean_delta_full": -0.08683, "read_hit_rate": {"surface": 0.5568}, "mean_gate": {"surface": 0.15256}, "n_reads": 3485, "n_cand": 6259, "read_suppressed_frac": 0.1059, "dropout_reads_logged": 3485, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 122.87368774414062, "n_util_obs": 415, "util_count_spearman": 0.13468910777169502}, "probe_overhead_sec_total": 213.904, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.012795, "alpha": 0.03, "t": 2856.2} |
| {"event": "consolidate", "step": 3550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262133, "t": 2896.2} |
| {"event": "train", "step": 3550, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.40464, "bpb": 2.02647, "chunk_loss": 0.030377, "F": 0.19611, "G": 0.21499, "chunks": 200.81, "bytes_per_chunk": 5.099, "boundary_entropy": 0.14158, "router_noise": 0.0, "grad_norm": 0.6643, "lr": 0.0001355, "router_trainable": true, "sec_per_step": 0.811, "probe_sec": 0.0475, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02739, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.07849, "n_shadow": 7, "mean_delta_shadow": 0.12705, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 1.15712, "trunc_abs_frac": 1.20195, "mean_delta_full": -0.12897, "read_hit_rate": {"surface": 0.53957}, "mean_gate": {"surface": 0.15596}, "n_reads": 3450, "n_cand": 6394, "read_suppressed_frac": 0.1035, "dropout_reads_logged": 3450, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 125.700927734375, "n_util_obs": 424, "util_count_spearman": 0.141957298020846}, "probe_overhead_sec_total": 217.011, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.012809, "alpha": 0.03, "t": 2896.7} |
| {"event": "consolidate", "step": 3600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262133, "t": 2935.1} |
| {"event": "probe", "step": 3600, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3266.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.187669, "cps": 25.515625, "mean_p": 0.205389, "boundary_entropy": 0.137281, "bpb": 1.960693}, "tr": {"chunks": 3508.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.19378, "cps": 27.40625, "mean_p": 0.206878, "boundary_entropy": 0.170644, "bpb": 2.110938}, "fi": {"chunks": 4270.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.226995, "cps": 33.359375, "mean_p": 0.234232, "boundary_entropy": 0.141583, "bpb": 2.234551}}, "bpb": {"en": 1.96069, "tr": 2.11094, "fi": 2.23455}, "gini_cpc": 0.043089, "gini_cps": 0.060606, "bpic": 5.136907, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0325639848375285, "fi": 1.2095502940700322}, "inventory": {"size": 326656, "admitted": 65536, "staged": 261120, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 261120, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 128.52537536621094, "n_util_obs": 426, "util_count_spearman": 0.1385920472323771}, "t": 2935.7} |
| {"event": "eval", "step": 3600, "phase": 2, "bpb": {"en": 1.97441, "tr": 2.11949, "fi": 2.24139}, "bytes_per_chunk": {"en": 5.064, "tr": 5.532, "fi": 4.682}, "bpb_mean": 2.11176, "frozen_inventory": true, "t": 2937.0} |
| {"event": "train", "step": 3600, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.40483, "bpb": 2.02674, "chunk_loss": 0.03043, "F": 0.19833, "G": 0.21609, "chunks": 203.09, "bytes_per_chunk": 5.042, "boundary_entropy": 0.13941, "router_noise": 0.0, "grad_norm": 0.632, "lr": 0.0001319, "router_trainable": true, "sec_per_step": 0.933, "probe_sec": 0.036, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02447, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": -0.01116, "n_shadow": 7, "mean_delta_shadow": -0.20631, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -1.01203, "trunc_abs_frac": 0.74325, "mean_delta_full": 0.0387, "read_hit_rate": {"surface": 0.54894}, "mean_gate": {"surface": 0.14947}, "n_reads": 3550, "n_cand": 6467, "read_suppressed_frac": 0.0986, "dropout_reads_logged": 3550, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 128.5795440673828, "n_util_obs": 426, "util_count_spearman": 0.1384389362578807}, "probe_overhead_sec_total": 220.543, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.017637, "alpha": 0.03, "t": 2943.4} |
| {"event": "consolidate", "step": 3650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 91, "uncertain_staged": 262133, "t": 2987.7} |
| {"event": "train", "step": 3650, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.44246, "bpb": 2.08103, "chunk_loss": 0.030509, "F": 0.20096, "G": 0.21942, "chunks": 205.78, "bytes_per_chunk": 4.976, "boundary_entropy": 0.1398, "router_noise": 0.0, "grad_norm": 0.6804, "lr": 0.0001283, "router_trainable": true, "sec_per_step": 0.913, "probe_sec": 0.0803, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05381, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 11, "mean_delta_loo": 0.19571, "n_shadow": 3, "mean_delta_shadow": 0.0698, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 1.08237, "trunc_abs_frac": 0.66418, "mean_delta_full": 0.14227, "read_hit_rate": {"surface": 0.55028}, "mean_gate": {"surface": 0.14958}, "n_reads": 3606, "n_cand": 6553, "read_suppressed_frac": 0.1034, "dropout_reads_logged": 3606, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 131.3769073486328, "n_util_obs": 430, "util_count_spearman": 0.13923793763745132}, "probe_overhead_sec_total": 224.148, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.016423, "alpha": 0.03, "t": 2989.0} |
| {"event": "consolidate", "step": 3700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 330, "uncertain_staged": 262133, "t": 3018.5} |
| {"event": "train", "step": 3700, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.39438, "bpb": 2.01167, "chunk_loss": 0.030443, "F": 0.2002, "G": 0.21948, "chunks": 205.0, "bytes_per_chunk": 4.995, "boundary_entropy": 0.13966, "router_noise": 0.0, "grad_norm": 0.6685, "lr": 0.0001247, "router_trainable": true, "sec_per_step": 0.611, "probe_sec": 0.0692, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04103, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": -0.00944, "n_shadow": 7, "mean_delta_shadow": -0.02128, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 17.94667, "trunc_abs_frac": 1.24625, "mean_delta_full": 0.00887, "read_hit_rate": {"surface": 0.57276}, "mean_gate": {"surface": 0.14956}, "n_reads": 3739, "n_cand": 6528, "read_suppressed_frac": 0.0955, "dropout_reads_logged": 3739, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 134.20452880859375, "n_util_obs": 439, "util_count_spearman": 0.1511666532866996}, "probe_overhead_sec_total": 226.831, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.01263, "alpha": 0.03, "t": 3019.6} |
| {"event": "consolidate", "step": 3750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 28, "uncertain_staged": 262133, "t": 3043.9} |
| {"event": "train", "step": 3750, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.37708, "bpb": 1.9867, "chunk_loss": 0.030369, "F": 0.19684, "G": 0.21453, "chunks": 201.56, "bytes_per_chunk": 5.08, "boundary_entropy": 0.13991, "router_noise": 0.0, "grad_norm": 0.6726, "lr": 0.0001212, "router_trainable": true, "sec_per_step": 0.507, "probe_sec": 0.0378, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02681, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.21654, "n_shadow": 6, "mean_delta_shadow": -0.06652, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.65677, "trunc_abs_frac": 0.68503, "mean_delta_full": -0.18445, "read_hit_rate": {"surface": 0.55157}, "mean_gate": {"surface": 0.15092}, "n_reads": 3540, "n_cand": 6418, "read_suppressed_frac": 0.0986, "dropout_reads_logged": 3540, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 136.98406982421875, "n_util_obs": 443, "util_count_spearman": 0.14080616550753486}, "probe_overhead_sec_total": 229.409, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.016739, "alpha": 0.03, "t": 3044.9} |
| {"event": "consolidate", "step": 3800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 35, "uncertain_staged": 262133, "t": 3091.5} |
| {"event": "probe", "step": 3800, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3228.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.185485, "cps": 25.21875, "mean_p": 0.203387, "boundary_entropy": 0.131719, "bpb": 1.947829}, "tr": {"chunks": 3522.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.194553, "cps": 27.515625, "mean_p": 0.204366, "boundary_entropy": 0.164524, "bpb": 2.104116}, "fi": {"chunks": 4253.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.226091, "cps": 33.226562, "mean_p": 0.232936, "boundary_entropy": 0.136835, "bpb": 2.214294}}, "bpb": {"en": 1.94783, "tr": 2.10412, "fi": 2.21429}, "gini_cpc": 0.044661, "gini_cps": 0.062104, "bpic": 5.156048, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0488886703039835, "fi": 1.218916885813272}, "inventory": {"size": 327645, "admitted": 65536, "staged": 262109, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262109, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 139.70816040039062, "n_util_obs": 449, "util_count_spearman": 0.15371093777835157}, "t": 3092.0} |
| {"event": "eval", "step": 3800, "phase": 2, "bpb": {"en": 1.96211, "tr": 2.11259, "fi": 2.22306}, "bytes_per_chunk": {"en": 5.068, "tr": 5.584, "fi": 4.699}, "bpb_mean": 2.09926, "frozen_inventory": true, "t": 3093.0} |
| {"event": "train", "step": 3800, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.44928, "bpb": 2.09087, "chunk_loss": 0.030397, "F": 0.19821, "G": 0.21534, "chunks": 202.97, "bytes_per_chunk": 5.045, "boundary_entropy": 0.13586, "router_noise": 0.0, "grad_norm": 0.6733, "lr": 0.0001177, "router_trainable": true, "sec_per_step": 1.094, "probe_sec": 0.0561, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02357, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": -0.03428, "n_shadow": 5, "mean_delta_shadow": 0.07972, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 1.9517, "trunc_abs_frac": 0.87924, "mean_delta_full": 0.05588, "read_hit_rate": {"surface": 0.54603}, "mean_gate": {"surface": 0.14865}, "n_reads": 3529, "n_cand": 6463, "read_suppressed_frac": 0.0992, "dropout_reads_logged": 3529, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 139.7620086669922, "n_util_obs": 449, "util_count_spearman": 0.15344833740417912}, "probe_overhead_sec_total": 233.233, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.016546, "alpha": 0.03, "t": 3099.6} |
| {"event": "consolidate", "step": 3850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 132, "uncertain_staged": 262133, "t": 3136.3} |
| {"event": "train", "step": 3850, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.54556, "bpb": 2.22976, "chunk_loss": 0.030663, "F": 0.20087, "G": 0.21833, "chunks": 205.69, "bytes_per_chunk": 4.978, "boundary_entropy": 0.14063, "router_noise": 0.0, "grad_norm": 0.7415, "lr": 0.0001142, "router_trainable": true, "sec_per_step": 0.754, "probe_sec": 0.0443, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02293, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.40701, "n_shadow": 6, "mean_delta_shadow": 0.09566, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 1.20655, "trunc_abs_frac": 2.43762, "mean_delta_full": 0.55375, "read_hit_rate": {"surface": 0.5371}, "mean_gate": {"surface": 0.15383}, "n_reads": 3518, "n_cand": 6550, "read_suppressed_frac": 0.0975, "dropout_reads_logged": 3518, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 142.57931518554688, "n_util_obs": 452, "util_count_spearman": 0.1581994621665327}, "probe_overhead_sec_total": 236.248, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.016537, "alpha": 0.03, "t": 3137.3} |
| {"event": "consolidate", "step": 3900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 44, "uncertain_staged": 262133, "t": 3178.3} |
| {"event": "train", "step": 3900, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.43666, "bpb": 2.07267, "chunk_loss": 0.030496, "F": 0.20062, "G": 0.2161, "chunks": 205.44, "bytes_per_chunk": 4.984, "boundary_entropy": 0.13382, "router_noise": 0.0, "grad_norm": 0.7246, "lr": 0.0001108, "router_trainable": true, "sec_per_step": 0.838, "probe_sec": 0.0948, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04377, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": -0.10094, "n_shadow": 5, "mean_delta_shadow": -0.04207, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.20409, "trunc_abs_frac": 1.95774, "mean_delta_full": -0.1405, "read_hit_rate": {"surface": 0.54509}, "mean_gate": {"surface": 0.14544}, "n_reads": 3566, "n_cand": 6542, "read_suppressed_frac": 0.099, "dropout_reads_logged": 3566, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 145.35765075683594, "n_util_obs": 453, "util_count_spearman": 0.1519471670186196}, "probe_overhead_sec_total": 239.675, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.01773, "alpha": 0.03, "t": 3179.2} |
| {"event": "consolidate", "step": 3950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 65, "uncertain_staged": 262133, "t": 3216.2} |
| {"event": "train", "step": 3950, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.48922, "bpb": 2.14849, "chunk_loss": 0.030426, "F": 0.19653, "G": 0.21319, "chunks": 201.25, "bytes_per_chunk": 5.088, "boundary_entropy": 0.13341, "router_noise": 0.0, "grad_norm": 0.7893, "lr": 0.0001074, "router_trainable": true, "sec_per_step": 0.762, "probe_sec": 0.0653, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02492, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.0245, "n_shadow": 6, "mean_delta_shadow": -0.0777, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 1.03267, "trunc_abs_frac": 1.03267, "mean_delta_full": 0.08594, "read_hit_rate": {"surface": 0.5387}, "mean_gate": {"surface": 0.14858}, "n_reads": 3452, "n_cand": 6408, "read_suppressed_frac": 0.0953, "dropout_reads_logged": 3452, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 148.1629180908203, "n_util_obs": 455, "util_count_spearman": 0.15661469636404987}, "probe_overhead_sec_total": 242.455, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.016598, "alpha": 0.03, "t": 3217.3} |
| {"event": "consolidate", "step": 4000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 19, "uncertain_staged": 262133, "t": 3256.9} |
| {"event": "probe", "step": 4000, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3258.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.187209, "cps": 25.453125, "mean_p": 0.203533, "boundary_entropy": 0.130238, "bpb": 1.933819}, "tr": {"chunks": 3548.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.19599, "cps": 27.71875, "mean_p": 0.205746, "boundary_entropy": 0.164317, "bpb": 2.086847}, "fi": {"chunks": 4357.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.23162, "cps": 34.039062, "mean_p": 0.234479, "boundary_entropy": 0.133261, "bpb": 2.19334}}, "bpb": {"en": 1.93382, "tr": 2.08685, "fi": 2.19334}, "gini_cpc": 0.048156, "gini_cps": 0.065633, "bpic": 5.082146, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0469021698935974, "fi": 1.2372250846919337}, "inventory": {"size": 327661, "admitted": 65536, "staged": 262125, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262125, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 150.8103485107422, "n_util_obs": 454, "util_count_spearman": 0.15654302826972755}, "t": 3257.8} |
| {"event": "eval", "step": 4000, "phase": 2, "bpb": {"en": 1.94883, "tr": 2.09503, "fi": 2.20061}, "bytes_per_chunk": {"en": 5.041, "tr": 5.551, "fi": 4.604}, "bpb_mean": 2.08149, "frozen_inventory": true, "t": 3258.8} |
| {"event": "train", "step": 4000, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.40974, "bpb": 2.03382, "chunk_loss": 0.030454, "F": 0.19659, "G": 0.21341, "chunks": 201.31, "bytes_per_chunk": 5.087, "boundary_entropy": 0.13406, "router_noise": 0.0, "grad_norm": 0.6489, "lr": 0.000104, "router_trainable": true, "sec_per_step": 0.939, "probe_sec": 0.0627, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02553, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": -0.09694, "n_shadow": 7, "mean_delta_shadow": -0.11372, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.50278, "trunc_abs_frac": 0.58589, "mean_delta_full": -0.14929, "read_hit_rate": {"surface": 0.5546}, "mean_gate": {"surface": 0.14841}, "n_reads": 3555, "n_cand": 6410, "read_suppressed_frac": 0.0993, "dropout_reads_logged": 3555, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 150.86459350585938, "n_util_obs": 454, "util_count_spearman": 0.15655822444319}, "probe_overhead_sec_total": 245.868, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.016815, "alpha": 0.03, "t": 3264.3} |
| {"event": "consolidate", "step": 4050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 8, "uncertain_staged": 262133, "t": 3302.7} |
| {"event": "train", "step": 4050, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.42214, "bpb": 2.05172, "chunk_loss": 0.030483, "F": 0.20139, "G": 0.21833, "chunks": 206.22, "bytes_per_chunk": 4.966, "boundary_entropy": 0.1362, "router_noise": 0.0, "grad_norm": 0.7096, "lr": 0.0001007, "router_trainable": true, "sec_per_step": 0.787, "probe_sec": 0.0602, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03241, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": -0.0488, "n_shadow": 7, "mean_delta_shadow": 0.01469, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.21772, "trunc_abs_frac": 0.22527, "mean_delta_full": -0.13741, "read_hit_rate": {"surface": 0.5683}, "mean_gate": {"surface": 0.15306}, "n_reads": 3732, "n_cand": 6567, "read_suppressed_frac": 0.09, "dropout_reads_logged": 3732, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 153.5802764892578, "n_util_obs": 456, "util_count_spearman": 0.15185055650941734}, "probe_overhead_sec_total": 248.886, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.012079, "alpha": 0.03, "t": 3303.6} |
| {"event": "probe", "step": 4096, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3292.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.189163, "cps": 25.71875, "mean_p": 0.20348, "boundary_entropy": 0.126003, "bpb": 1.929948}, "tr": {"chunks": 3596.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.198641, "cps": 28.09375, "mean_p": 0.204744, "boundary_entropy": 0.158323, "bpb": 2.079777}, "fi": {"chunks": 4264.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.226676, "cps": 33.3125, "mean_p": 0.232137, "boundary_entropy": 0.129274, "bpb": 2.189361}}, "bpb": {"en": 1.92995, "tr": 2.07978, "fi": 2.18936}, "gini_cpc": 0.040699, "gini_cps": 0.058106, "bpic": 5.087159, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0501066900225111, "fi": 1.1983111662710213}, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 156.1010284423828, "n_util_obs": 458, "util_count_spearman": 0.15757674320596463}, "t": 3337.2} |
| {"event": "consolidate", "step": 4100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 15, "uncertain_staged": 262133, "t": 3354.5} |
| {"event": "train", "step": 4100, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.36322, "bpb": 1.96671, "chunk_loss": 0.030432, "F": 0.19931, "G": 0.21447, "chunks": 204.09, "bytes_per_chunk": 5.017, "boundary_entropy": 0.12857, "router_noise": 0.0, "grad_norm": 0.7174, "lr": 9.75e-05, "router_trainable": true, "sec_per_step": 1.038, "probe_sec": 0.1029, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07997, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.15954, "n_shadow": 8, "mean_delta_shadow": -0.06221, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.65785, "trunc_abs_frac": 0.65584, "mean_delta_full": -0.07352, "read_hit_rate": {"surface": 0.58932}, "mean_gate": {"surface": 0.14886}, "n_reads": 3830, "n_cand": 6499, "read_suppressed_frac": 0.1068, "dropout_reads_logged": 3830, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 156.35000610351562, "n_util_obs": 457, "util_count_spearman": 0.1528215270327913}, "probe_overhead_sec_total": 251.935, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.01716, "alpha": 0.03, "t": 3355.5} |
| {"event": "consolidate", "step": 4150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 24, "uncertain_staged": 262133, "t": 3392.7} |
| {"event": "train", "step": 4150, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.42189, "bpb": 2.05136, "chunk_loss": 0.030634, "F": 0.20737, "G": 0.22245, "chunks": 212.34, "bytes_per_chunk": 4.822, "boundary_entropy": 0.13256, "router_noise": 0.0, "grad_norm": 0.7641, "lr": 9.43e-05, "router_trainable": true, "sec_per_step": 0.753, "probe_sec": 0.0278, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01963, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.15877, "n_shadow": 6, "mean_delta_shadow": 0.03563, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.05327, "trunc_abs_frac": 0.41755, "mean_delta_full": 0.06531, "read_hit_rate": {"surface": 0.5604}, "mean_gate": {"surface": 0.14912}, "n_reads": 3790, "n_cand": 6763, "read_suppressed_frac": 0.1045, "dropout_reads_logged": 3790, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 159.15415954589844, "n_util_obs": 463, "util_count_spearman": 0.1565984479382153}, "probe_overhead_sec_total": 254.891, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.012377, "alpha": 0.03, "t": 3393.1} |
| {"event": "consolidate", "step": 4200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 361, "uncertain_staged": 262133, "t": 3429.7} |
| {"event": "probe", "step": 4200, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3265.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.187611, "cps": 25.507812, "mean_p": 0.202528, "boundary_entropy": 0.126555, "bpb": 1.928043}, "tr": {"chunks": 3542.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.195658, "cps": 27.671875, "mean_p": 0.204952, "boundary_entropy": 0.16162, "bpb": 2.070828}, "fi": {"chunks": 4278.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.22742, "cps": 33.421875, "mean_p": 0.231478, "boundary_entropy": 0.130767, "bpb": 2.186197}}, "bpb": {"en": 1.92804, "tr": 2.07083, "fi": 2.1862}, "gini_cpc": 0.043458, "gini_cps": 0.060923, "bpic": 5.117907, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.042891049083689, "fi": 1.2121875840298388}, "inventory": {"size": 327319, "admitted": 65536, "staged": 261783, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 261783, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 161.9582061767578, "n_util_obs": 466, "util_count_spearman": 0.16784135849683823}, "t": 3430.3} |
| {"event": "eval", "step": 4200, "phase": 2, "bpb": {"en": 1.94165, "tr": 2.08017, "fi": 2.19533}, "bytes_per_chunk": {"en": 5.054, "tr": 5.526, "fi": 4.688}, "bpb_mean": 2.07238, "frozen_inventory": true, "t": 3431.1} |
| {"event": "phase_transition", "step": 4200, "phase": 3, "objective": "global", "router_trainable": true, "encoder_trainable": true, "frozen_router_params": 0, "t": 3434.9} |
| {"event": "train", "step": 4200, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35053, "bpb": 1.9484, "chunk_loss": 0.030286, "F": 0.19211, "G": 0.20753, "chunks": 196.72, "bytes_per_chunk": 5.205, "boundary_entropy": 0.12862, "router_noise": 0.0, "grad_norm": 0.692, "lr": 9.12e-05, "router_trainable": true, "sec_per_step": 0.839, "probe_sec": 0.0249, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01844, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": -0.07032, "n_shadow": 9, "mean_delta_shadow": 0.06278, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.85224, "trunc_abs_frac": 5.2767, "mean_delta_full": 0.0759, "read_hit_rate": {"surface": 0.56411}, "mean_gate": {"surface": 0.14794}, "n_reads": 3533, "n_cand": 6263, "read_suppressed_frac": 0.0971, "dropout_reads_logged": 3533, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 162.01211547851562, "n_util_obs": 466, "util_count_spearman": 0.16789120808286928}, "probe_overhead_sec_total": 258.061, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.015609, "alpha": 0.03, "t": 3435.1} |
| {"event": "consolidate", "step": 4250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 95, "uncertain_staged": 262133, "t": 3476.9} |
| {"event": "train", "step": 4250, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38569, "bpb": 1.99913, "chunk_loss": 0.030464, "F": 0.2019, "G": 0.21699, "chunks": 206.75, "bytes_per_chunk": 4.953, "boundary_entropy": 0.12952, "router_noise": 0.0, "grad_norm": 0.7023, "lr": 8.81e-05, "router_trainable": true, "sec_per_step": 0.853, "probe_sec": 0.0721, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05479, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": 0.31134, "n_shadow": 10, "mean_delta_shadow": -0.06705, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 1.20165, "trunc_abs_frac": 4.20196, "mean_delta_full": 0.10785, "read_hit_rate": {"surface": 0.55954}, "mean_gate": {"surface": 0.14487}, "n_reads": 3684, "n_cand": 6584, "read_suppressed_frac": 0.1018, "dropout_reads_logged": 3684, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 164.8869171142578, "n_util_obs": 469, "util_count_spearman": 0.1660947996359015}, "probe_overhead_sec_total": 261.375, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.017296, "alpha": 0.03, "t": 3477.7} |
| {"event": "consolidate", "step": 4300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 467, "uncertain_staged": 262133, "t": 3520.8} |
| {"event": "train", "step": 4300, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38801, "bpb": 2.00248, "chunk_loss": 0.030292, "F": 0.19293, "G": 0.20888, "chunks": 197.56, "bytes_per_chunk": 5.183, "boundary_entropy": 0.13154, "router_noise": 0.0, "grad_norm": 0.7505, "lr": 8.5e-05, "router_trainable": true, "sec_per_step": 0.882, "probe_sec": 0.109, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07494, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 11, "mean_delta_loo": 0.09739, "n_shadow": 3, "mean_delta_shadow": -0.07685, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.0332, "trunc_abs_frac": 0.44736, "mean_delta_full": 0.22457, "read_hit_rate": {"surface": 0.55517}, "mean_gate": {"surface": 0.14587}, "n_reads": 3492, "n_cand": 6290, "read_suppressed_frac": 0.0951, "dropout_reads_logged": 3492, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 167.78297424316406, "n_util_obs": 474, "util_count_spearman": 0.18820926826925552}, "probe_overhead_sec_total": 264.498, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.016442, "alpha": 0.03, "t": 3521.8} |
| {"event": "consolidate", "step": 4350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262133, "t": 3560.1} |
| {"event": "train", "step": 4350, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.40922, "bpb": 2.03308, "chunk_loss": 0.030388, "F": 0.19763, "G": 0.21282, "chunks": 202.38, "bytes_per_chunk": 5.06, "boundary_entropy": 0.12405, "router_noise": 0.0, "grad_norm": 0.6674, "lr": 8.21e-05, "router_trainable": true, "sec_per_step": 0.785, "probe_sec": 0.049, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02381, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": 0.00622, "n_shadow": 5, "mean_delta_shadow": -0.02054, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 1.06228, "trunc_abs_frac": 0.44974, "mean_delta_full": -0.17102, "read_hit_rate": {"surface": 0.57821}, "mean_gate": {"surface": 0.14234}, "n_reads": 3726, "n_cand": 6444, "read_suppressed_frac": 0.1031, "dropout_reads_logged": 3726, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 170.78407287597656, "n_util_obs": 475, "util_count_spearman": 0.1980031012383539}, "probe_overhead_sec_total": 267.647, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.016128, "alpha": 0.03, "t": 3561.0} |
| {"event": "consolidate", "step": 4400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 528, "uncertain_staged": 262133, "t": 3600.4} |
| {"event": "probe", "step": 4400, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3277.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.188301, "cps": 25.601562, "mean_p": 0.202365, "boundary_entropy": 0.123368, "bpb": 1.916788}, "tr": {"chunks": 3592.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.19842, "cps": 28.0625, "mean_p": 0.202366, "boundary_entropy": 0.155129, "bpb": 2.056413}, "fi": {"chunks": 4282.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.227633, "cps": 33.453125, "mean_p": 0.23156, "boundary_entropy": 0.129311, "bpb": 2.169328}}, "bpb": {"en": 1.91679, "tr": 2.05641, "fi": 2.16933}, "gini_cpc": 0.042681, "gini_cps": 0.060084, "bpic": 5.087615, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.053739973771959, "fi": 1.20887795623124}, "inventory": {"size": 327152, "admitted": 65536, "staged": 261616, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 261616, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 173.7914276123047, "n_util_obs": 478, "util_count_spearman": 0.193468596452024}, "t": 3600.9} |
| {"event": "eval", "step": 4400, "phase": 3, "bpb": {"en": 1.9311, "tr": 2.06524, "fi": 2.17811}, "bytes_per_chunk": {"en": 5.042, "tr": 5.423, "fi": 4.668}, "bpb_mean": 2.05815, "frozen_inventory": true, "t": 3601.8} |
| {"event": "train", "step": 4400, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35981, "bpb": 1.9618, "chunk_loss": 0.030312, "F": 0.19562, "G": 0.2102, "chunks": 200.31, "bytes_per_chunk": 5.112, "boundary_entropy": 0.12363, "router_noise": 0.0, "grad_norm": 0.6528, "lr": 7.92e-05, "router_trainable": true, "sec_per_step": 0.921, "probe_sec": 0.0728, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0303, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.14498, "n_shadow": 8, "mean_delta_shadow": -0.00581, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.04559, "trunc_abs_frac": 0.04559, "mean_delta_full": 0.50208, "read_hit_rate": {"surface": 0.59956}, "mean_gate": {"surface": 0.14078}, "n_reads": 3824, "n_cand": 6378, "read_suppressed_frac": 0.0988, "dropout_reads_logged": 3824, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 173.8497772216797, "n_util_obs": 478, "util_count_spearman": 0.19378741695429808}, "probe_overhead_sec_total": 270.709, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.012608, "alpha": 0.03, "t": 3607.1} |
| {"event": "consolidate", "step": 4450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 117, "uncertain_staged": 262133, "t": 3645.0} |
| {"event": "train", "step": 4450, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35789, "bpb": 1.95902, "chunk_loss": 0.030258, "F": 0.18982, "G": 0.205, "chunks": 194.38, "bytes_per_chunk": 5.268, "boundary_entropy": 0.12757, "router_noise": 0.0, "grad_norm": 0.7047, "lr": 7.63e-05, "router_trainable": true, "sec_per_step": 0.773, "probe_sec": 0.0607, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02376, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": 0.07601, "n_shadow": 5, "mean_delta_shadow": -0.09832, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 2.32428, "trunc_abs_frac": 0.91576, "mean_delta_full": 0.21692, "read_hit_rate": {"surface": 0.60094}, "mean_gate": {"surface": 0.13777}, "n_reads": 3718, "n_cand": 6187, "read_suppressed_frac": 0.0955, "dropout_reads_logged": 3718, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 176.89599609375, "n_util_obs": 480, "util_count_spearman": 0.19922507786968685}, "probe_overhead_sec_total": 273.606, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.012911, "alpha": 0.03, "t": 3645.7} |
| {"event": "consolidate", "step": 4500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 32, "uncertain_staged": 262133, "t": 3676.4} |
| {"event": "train", "step": 4500, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32886, "bpb": 1.91714, "chunk_loss": 0.03037, "F": 0.19888, "G": 0.21281, "chunks": 203.66, "bytes_per_chunk": 5.028, "boundary_entropy": 0.12274, "router_noise": 0.0, "grad_norm": 0.6537, "lr": 7.36e-05, "router_trainable": true, "sec_per_step": 0.629, "probe_sec": 0.0998, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07296, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": 0.18415, "n_shadow": 5, "mean_delta_shadow": 0.27852, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.10564, "trunc_abs_frac": 0.49373, "mean_delta_full": 0.16581, "read_hit_rate": {"surface": 0.61496}, "mean_gate": {"surface": 0.13585}, "n_reads": 3988, "n_cand": 6485, "read_suppressed_frac": 0.0963, "dropout_reads_logged": 3988, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 179.9103240966797, "n_util_obs": 484, "util_count_spearman": 0.18972906560223585}, "probe_overhead_sec_total": 276.13, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.013532, "alpha": 0.03, "t": 3677.2} |
| {"event": "consolidate", "step": 4550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 246, "uncertain_staged": 262133, "t": 3724.4} |
| {"event": "train", "step": 4550, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.36072, "bpb": 1.9631, "chunk_loss": 0.030329, "F": 0.1958, "G": 0.21077, "chunks": 200.5, "bytes_per_chunk": 5.107, "boundary_entropy": 0.12511, "router_noise": 0.0, "grad_norm": 0.6838, "lr": 7.09e-05, "router_trainable": true, "sec_per_step": 0.949, "probe_sec": 0.0465, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0341, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": 0.01692, "n_shadow": 9, "mean_delta_shadow": -0.03352, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.44396, "trunc_abs_frac": 0.7792, "mean_delta_full": 0.08441, "read_hit_rate": {"surface": 0.59931}, "mean_gate": {"surface": 0.13811}, "n_reads": 3826, "n_cand": 6384, "read_suppressed_frac": 0.1004, "dropout_reads_logged": 3826, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 183.04385375976562, "n_util_obs": 488, "util_count_spearman": 0.18245390834811367}, "probe_overhead_sec_total": 279.461, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.011935, "alpha": 0.03, "t": 3724.6} |
| {"event": "consolidate", "step": 4600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 27, "uncertain_staged": 262133, "t": 3765.1} |
| {"event": "probe", "step": 4600, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3267.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.187726, "cps": 25.523438, "mean_p": 0.203183, "boundary_entropy": 0.120502, "bpb": 1.906574}, "tr": {"chunks": 3586.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.198089, "cps": 28.015625, "mean_p": 0.202359, "boundary_entropy": 0.149928, "bpb": 2.047595}, "fi": {"chunks": 4347.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.231088, "cps": 33.960938, "mean_p": 0.233932, "boundary_entropy": 0.126332, "bpb": 2.153546}}, "bpb": {"en": 1.90657, "tr": 2.0476, "fi": 2.15355}, "gini_cpc": 0.04686, "gini_cps": 0.064286, "bpic": 5.065357, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0551998468918318, "fi": 1.2309849477029222}, "inventory": {"size": 327653, "admitted": 65536, "staged": 262117, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262117, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 186.0485382080078, "n_util_obs": 491, "util_count_spearman": 0.1761385687261217}, "t": 3765.7} |
| {"event": "eval", "step": 4600, "phase": 3, "bpb": {"en": 1.92128, "tr": 2.05569, "fi": 2.16144}, "bytes_per_chunk": {"en": 5.018, "tr": 5.43, "fi": 4.601}, "bpb_mean": 2.04614, "frozen_inventory": true, "t": 3766.6} |
| {"event": "train", "step": 4600, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.36181, "bpb": 1.96468, "chunk_loss": 0.030537, "F": 0.20694, "G": 0.21987, "chunks": 211.91, "bytes_per_chunk": 4.832, "boundary_entropy": 0.12561, "router_noise": 0.0, "grad_norm": 0.7264, "lr": 6.82e-05, "router_trainable": true, "sec_per_step": 0.942, "probe_sec": 0.0454, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02301, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": 0.01515, "n_shadow": 2, "mean_delta_shadow": 0.1088, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 2.09139, "trunc_abs_frac": 13.93799, "mean_delta_full": -0.08081, "read_hit_rate": {"surface": 0.62898}, "mean_gate": {"surface": 0.13861}, "n_reads": 4245, "n_cand": 6749, "read_suppressed_frac": 0.1011, "dropout_reads_logged": 4245, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 186.11331176757812, "n_util_obs": 491, "util_count_spearman": 0.17592759911269132}, "probe_overhead_sec_total": 282.796, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.016338, "alpha": 0.03, "t": 3771.7} |
| {"event": "consolidate", "step": 4650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 158, "uncertain_staged": 262133, "t": 3796.0} |
| {"event": "train", "step": 4650, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.39124, "bpb": 2.00714, "chunk_loss": 0.030466, "F": 0.20145, "G": 0.21598, "chunks": 206.28, "bytes_per_chunk": 4.964, "boundary_entropy": 0.12583, "router_noise": 0.0, "grad_norm": 0.6725, "lr": 6.57e-05, "router_trainable": true, "sec_per_step": 0.509, "probe_sec": 0.0692, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03355, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.02783, "n_shadow": 4, "mean_delta_shadow": 0.05346, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.39264, "trunc_abs_frac": 1.09719, "mean_delta_full": 0.63519, "read_hit_rate": {"surface": 0.60009}, "mean_gate": {"surface": 0.14051}, "n_reads": 3942, "n_cand": 6569, "read_suppressed_frac": 0.104, "dropout_reads_logged": 3942, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 189.18458557128906, "n_util_obs": 493, "util_count_spearman": 0.17516512215490518}, "probe_overhead_sec_total": 285.369, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.011613, "alpha": 0.03, "t": 3797.1} |
| {"event": "consolidate", "step": 4700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 269, "uncertain_staged": 262133, "t": 3843.6} |
| {"event": "train", "step": 4700, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.42145, "bpb": 2.05072, "chunk_loss": 0.03061, "F": 0.20703, "G": 0.21912, "chunks": 212.0, "bytes_per_chunk": 4.83, "boundary_entropy": 0.11943, "router_noise": 0.0, "grad_norm": 0.679, "lr": 6.32e-05, "router_trainable": true, "sec_per_step": 0.948, "probe_sec": 0.0569, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02962, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": 0.28519, "n_shadow": 5, "mean_delta_shadow": -0.01701, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 1.30744, "trunc_abs_frac": 1.16261, "mean_delta_full": -0.08527, "read_hit_rate": {"surface": 0.61641}, "mean_gate": {"surface": 0.13515}, "n_reads": 4162, "n_cand": 6752, "read_suppressed_frac": 0.0942, "dropout_reads_logged": 4162, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 192.22122192382812, "n_util_obs": 498, "util_count_spearman": 0.16001022918181748}, "probe_overhead_sec_total": 288.519, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.011794, "alpha": 0.03, "t": 3844.5} |
| {"event": "consolidate", "step": 4750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 658, "uncertain_staged": 262133, "t": 3882.4} |
| {"event": "train", "step": 4750, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.40616, "bpb": 2.02865, "chunk_loss": 0.030419, "F": 0.20093, "G": 0.21451, "chunks": 205.75, "bytes_per_chunk": 4.977, "boundary_entropy": 0.12385, "router_noise": 0.0, "grad_norm": 0.6627, "lr": 6.08e-05, "router_trainable": true, "sec_per_step": 0.769, "probe_sec": 0.0434, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02447, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.06819, "n_shadow": 4, "mean_delta_shadow": -0.01508, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 7, "trunc_ratio_sum": 5.36562, "trunc_abs_frac": 1.25056, "mean_delta_full": -0.01557, "read_hit_rate": {"surface": 0.62485}, "mean_gate": {"surface": 0.13503}, "n_reads": 4094, "n_cand": 6552, "read_suppressed_frac": 0.1006, "dropout_reads_logged": 4094, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 195.29132080078125, "n_util_obs": 502, "util_count_spearman": 0.16248090784545158}, "probe_overhead_sec_total": 291.422, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.011512, "alpha": 0.03, "t": 3883.0} |
| {"event": "consolidate", "step": 4800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 519, "uncertain_staged": 262133, "t": 3919.3} |
| {"event": "probe", "step": 4800, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3248.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.186634, "cps": 25.375, "mean_p": 0.200278, "boundary_entropy": 0.118111, "bpb": 1.898507}, "tr": {"chunks": 3602.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.198973, "cps": 28.140625, "mean_p": 0.202561, "boundary_entropy": 0.150465, "bpb": 2.034901}, "fi": {"chunks": 4285.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.227792, "cps": 33.476562, "mean_p": 0.230673, "boundary_entropy": 0.122758, "bpb": 2.142835}}, "bpb": {"en": 1.89851, "tr": 2.0349, "fi": 2.14283}, "gini_cpc": 0.044732, "gini_cps": 0.062087, "bpic": 5.094926, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0661081335619467, "fi": 1.2205260200443457}, "inventory": {"size": 327161, "admitted": 65536, "staged": 261625, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 261625, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 198.3743896484375, "n_util_obs": 506, "util_count_spearman": 0.17019562823420126}, "t": 3920.1} |
| {"event": "eval", "step": 4800, "phase": 3, "bpb": {"en": 1.91275, "tr": 2.04396, "fi": 2.15}, "bytes_per_chunk": {"en": 5.056, "tr": 5.462, "fi": 4.671}, "bpb_mean": 2.03557, "frozen_inventory": true, "t": 3921.1} |
| {"event": "train", "step": 4800, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.37799, "bpb": 1.98803, "chunk_loss": 0.030447, "F": 0.20157, "G": 0.21393, "chunks": 206.41, "bytes_per_chunk": 4.961, "boundary_entropy": 0.11953, "router_noise": 0.0, "grad_norm": 0.6657, "lr": 5.85e-05, "router_trainable": true, "sec_per_step": 0.875, "probe_sec": 0.1001, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08146, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.29102, "n_shadow": 6, "mean_delta_shadow": -0.1733, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -1.83497, "trunc_abs_frac": 0.53812, "mean_delta_full": 0.00869, "read_hit_rate": {"surface": 0.62209}, "mean_gate": {"surface": 0.13272}, "n_reads": 4089, "n_cand": 6573, "read_suppressed_frac": 0.1022, "dropout_reads_logged": 4089, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 198.43678283691406, "n_util_obs": 506, "util_count_spearman": 0.17044108236357247}, "probe_overhead_sec_total": 294.322, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.01594, "alpha": 0.03, "t": 3926.7} |
| {"event": "consolidate", "step": 4850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 83, "uncertain_staged": 262133, "t": 3965.5} |
| {"event": "train", "step": 4850, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35318, "bpb": 1.95223, "chunk_loss": 0.030401, "F": 0.19797, "G": 0.21269, "chunks": 202.72, "bytes_per_chunk": 5.051, "boundary_entropy": 0.11963, "router_noise": 0.0, "grad_norm": 0.6941, "lr": 5.62e-05, "router_trainable": true, "sec_per_step": 0.789, "probe_sec": 0.0471, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02202, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.03582, "n_shadow": 7, "mean_delta_shadow": -0.13842, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.7874, "trunc_abs_frac": 1.23697, "mean_delta_full": -0.37781, "read_hit_rate": {"surface": 0.62928}, "mean_gate": {"surface": 0.13749}, "n_reads": 4062, "n_cand": 6455, "read_suppressed_frac": 0.1012, "dropout_reads_logged": 4062, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 201.51785278320312, "n_util_obs": 508, "util_count_spearman": 0.17226733789394819}, "probe_overhead_sec_total": 297.241, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.016049, "alpha": 0.03, "t": 3966.2} |
| {"event": "consolidate", "step": 4900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262133, "t": 4002.4} |
| {"event": "train", "step": 4900, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38006, "bpb": 1.99101, "chunk_loss": 0.030391, "F": 0.19434, "G": 0.20711, "chunks": 199.0, "bytes_per_chunk": 5.146, "boundary_entropy": 0.11773, "router_noise": 0.0, "grad_norm": 0.6949, "lr": 5.41e-05, "router_trainable": true, "sec_per_step": 0.739, "probe_sec": 0.0402, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02128, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": 0.15536, "n_shadow": 2, "mean_delta_shadow": 0.07663, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -0.08943, "trunc_abs_frac": 1.17721, "mean_delta_full": 0.06381, "read_hit_rate": {"surface": 0.60874}, "mean_gate": {"surface": 0.13509}, "n_reads": 3857, "n_cand": 6336, "read_suppressed_frac": 0.0985, "dropout_reads_logged": 3857, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 204.58177185058594, "n_util_obs": 510, "util_count_spearman": 0.1747201511967537}, "probe_overhead_sec_total": 300.19, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.011931, "alpha": 0.03, "t": 4003.1} |
| {"event": "consolidate", "step": 4950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 58, "uncertain_staged": 262133, "t": 4045.2} |
| {"event": "train", "step": 4950, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.3396, "bpb": 1.93263, "chunk_loss": 0.030382, "F": 0.19742, "G": 0.21102, "chunks": 202.16, "bytes_per_chunk": 5.065, "boundary_entropy": 0.12176, "router_noise": 0.0, "grad_norm": 0.6896, "lr": 5.2e-05, "router_trainable": true, "sec_per_step": 0.86, "probe_sec": 0.0545, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03144, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 11, "mean_delta_loo": 0.23465, "n_shadow": 3, "mean_delta_shadow": -0.04832, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 1.03401, "trunc_abs_frac": 4.16888, "mean_delta_full": 0.09982, "read_hit_rate": {"surface": 0.63166}, "mean_gate": {"surface": 0.13424}, "n_reads": 4066, "n_cand": 6437, "read_suppressed_frac": 0.0954, "dropout_reads_logged": 4066, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 207.73080444335938, "n_util_obs": 517, "util_count_spearman": 0.17906202080859088}, "probe_overhead_sec_total": 303.884, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.016532, "alpha": 0.03, "t": 4046.1} |
| {"event": "consolidate", "step": 5000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 22, "uncertain_staged": 262133, "t": 4090.1} |
| {"event": "probe", "step": 5000, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3277.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.188301, "cps": 25.601562, "mean_p": 0.202168, "boundary_entropy": 0.117515, "bpb": 1.892437}, "tr": {"chunks": 3652.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.201735, "cps": 28.53125, "mean_p": 0.201546, "boundary_entropy": 0.142485, "bpb": 2.026794}, "fi": {"chunks": 4289.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.228005, "cps": 33.507812, "mean_p": 0.230994, "boundary_entropy": 0.120184, "bpb": 2.132438}}, "bpb": {"en": 1.89244, "tr": 2.02679, "fi": 2.13244}, "gini_cpc": 0.042828, "gini_cps": 0.060141, "bpic": 5.057229, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0713414209953214, "fi": 1.2108541696113468}, "inventory": {"size": 327658, "admitted": 65536, "staged": 262122, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262122, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 210.7989501953125, "n_util_obs": 523, "util_count_spearman": 0.1810330681243987}, "t": 4090.8} |
| {"event": "eval", "step": 5000, "phase": 3, "bpb": {"en": 1.90724, "tr": 2.0363, "fi": 2.1392}, "bytes_per_chunk": {"en": 5.022, "tr": 5.35, "fi": 4.662}, "bpb_mean": 2.02758, "frozen_inventory": true, "t": 4092.0} |
| {"event": "train", "step": 5000, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.33164, "bpb": 1.92114, "chunk_loss": 0.030374, "F": 0.19836, "G": 0.21221, "chunks": 203.12, "bytes_per_chunk": 5.041, "boundary_entropy": 0.1188, "router_noise": 0.0, "grad_norm": 0.6293, "lr": 5e-05, "router_trainable": true, "sec_per_step": 1.06, "probe_sec": 0.0628, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0239, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.00585, "n_shadow": 4, "mean_delta_shadow": 0.0915, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.67458, "trunc_abs_frac": 1.72169, "mean_delta_full": -0.16412, "read_hit_rate": {"surface": 0.62863}, "mean_gate": {"surface": 0.1312}, "n_reads": 4066, "n_cand": 6468, "read_suppressed_frac": 0.1003, "dropout_reads_logged": 4066, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 210.86099243164062, "n_util_obs": 523, "util_count_spearman": 0.18104275669657055}, "probe_overhead_sec_total": 307.14, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.011441, "alpha": 0.03, "t": 4099.1} |
| {"event": "consolidate", "step": 5050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 6, "uncertain_staged": 262133, "t": 4124.9} |
| {"event": "train", "step": 5050, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.34821, "bpb": 1.94505, "chunk_loss": 0.03042, "F": 0.20026, "G": 0.21374, "chunks": 205.06, "bytes_per_chunk": 4.994, "boundary_entropy": 0.11721, "router_noise": 0.0, "grad_norm": 0.6903, "lr": 4.81e-05, "router_trainable": true, "sec_per_step": 0.52, "probe_sec": 0.031, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0191, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 13, "mean_delta_loo": -0.12134, "n_shadow": 1, "mean_delta_shadow": -0.0365, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 1.96807, "trunc_abs_frac": 2.03647, "mean_delta_full": -0.18083, "read_hit_rate": {"surface": 0.62879}, "mean_gate": {"surface": 0.13053}, "n_reads": 4106, "n_cand": 6530, "read_suppressed_frac": 0.1011, "dropout_reads_logged": 4106, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 213.93003845214844, "n_util_obs": 525, "util_count_spearman": 0.18632199889219714}, "probe_overhead_sec_total": 309.68, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.015389, "alpha": 0.03, "t": 4125.1} |
| {"event": "consolidate", "step": 5100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1, "uncertain_staged": 262133, "t": 4148.1} |
| {"event": "train", "step": 5100, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.37367, "bpb": 1.98179, "chunk_loss": 0.030501, "F": 0.20227, "G": 0.21571, "chunks": 207.12, "bytes_per_chunk": 4.944, "boundary_entropy": 0.1211, "router_noise": 0.0, "grad_norm": 0.7014, "lr": 4.63e-05, "router_trainable": true, "sec_per_step": 0.469, "probe_sec": 0.0707, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05348, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.06154, "n_shadow": 7, "mean_delta_shadow": -0.17819, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 1.27664, "trunc_abs_frac": 7.38493, "mean_delta_full": 0.16113, "read_hit_rate": {"surface": 0.61613}, "mean_gate": {"surface": 0.13289}, "n_reads": 4064, "n_cand": 6596, "read_suppressed_frac": 0.0992, "dropout_reads_logged": 4064, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 217.059326171875, "n_util_obs": 527, "util_count_spearman": 0.19795109592007754}, "probe_overhead_sec_total": 312.56, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.01195, "alpha": 0.03, "t": 4148.6} |
| {"event": "consolidate", "step": 5150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 40, "uncertain_staged": 262133, "t": 4191.5} |
| {"event": "train", "step": 5150, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.36644, "bpb": 1.97135, "chunk_loss": 0.030545, "F": 0.2048, "G": 0.21817, "chunks": 209.72, "bytes_per_chunk": 4.883, "boundary_entropy": 0.11962, "router_noise": 0.0, "grad_norm": 0.7061, "lr": 4.45e-05, "router_trainable": true, "sec_per_step": 0.879, "probe_sec": 0.0635, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03441, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": 0.09293, "n_shadow": 5, "mean_delta_shadow": -0.0103, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.93397, "trunc_abs_frac": 0.64338, "mean_delta_full": 0.30644, "read_hit_rate": {"surface": 0.64291}, "mean_gate": {"surface": 0.13362}, "n_reads": 4294, "n_cand": 6679, "read_suppressed_frac": 0.1001, "dropout_reads_logged": 4294, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 220.2293243408203, "n_util_obs": 534, "util_count_spearman": 0.19829662115340546}, "probe_overhead_sec_total": 315.657, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.014328, "alpha": 0.03, "t": 4192.5} |
| {"event": "consolidate", "step": 5200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 10, "uncertain_staged": 262133, "t": 4234.7} |
| {"event": "probe", "step": 5200, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3261.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.187381, "cps": 25.476562, "mean_p": 0.200027, "boundary_entropy": 0.114633, "bpb": 1.88577}, "tr": {"chunks": 3601.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.198917, "cps": 28.132812, "mean_p": 0.20098, "boundary_entropy": 0.142239, "bpb": 2.019638}, "fi": {"chunks": 4314.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.229334, "cps": 33.703125, "mean_p": 0.23051, "boundary_entropy": 0.117877, "bpb": 2.122788}}, "bpb": {"en": 1.88577, "tr": 2.01964, "fi": 2.12279}, "gini_cpc": 0.04543, "gini_cps": 0.062813, "bpic": 5.076235, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0615632890013351, "fi": 1.2238877240933965}, "inventory": {"size": 327670, "admitted": 65536, "staged": 262134, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262134, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 223.3218994140625, "n_util_obs": 538, "util_count_spearman": 0.18684958025416967}, "t": 4235.3} |
| {"event": "eval", "step": 5200, "phase": 3, "bpb": {"en": 1.90144, "tr": 2.02857, "fi": 2.13016}, "bytes_per_chunk": {"en": 5.049, "tr": 5.439, "fi": 4.65}, "bpb_mean": 2.02006, "frozen_inventory": true, "t": 4236.2} |
| {"event": "train", "step": 5200, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32541, "bpb": 1.91216, "chunk_loss": 0.030345, "F": 0.19766, "G": 0.20898, "chunks": 202.41, "bytes_per_chunk": 5.059, "boundary_entropy": 0.11431, "router_noise": 0.0, "grad_norm": 0.6626, "lr": 4.29e-05, "router_trainable": true, "sec_per_step": 1.026, "probe_sec": 0.0737, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03933, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.02225, "n_shadow": 7, "mean_delta_shadow": 0.15771, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.6814, "trunc_abs_frac": 0.65241, "mean_delta_full": 0.33435, "read_hit_rate": {"surface": 0.645}, "mean_gate": {"surface": 0.12965}, "n_reads": 4157, "n_cand": 6445, "read_suppressed_frac": 0.1025, "dropout_reads_logged": 4157, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 223.3853302001953, "n_util_obs": 538, "util_count_spearman": 0.18675642760631112}, "probe_overhead_sec_total": 318.815, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.016155, "alpha": 0.03, "t": 4243.8} |
| {"event": "consolidate", "step": 5250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 93, "uncertain_staged": 262133, "t": 4281.6} |
| {"event": "train", "step": 5250, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32247, "bpb": 1.90792, "chunk_loss": 0.030379, "F": 0.19818, "G": 0.21044, "chunks": 202.94, "bytes_per_chunk": 5.046, "boundary_entropy": 0.11626, "router_noise": 0.0, "grad_norm": 0.6613, "lr": 4.14e-05, "router_trainable": true, "sec_per_step": 0.766, "probe_sec": 0.0839, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06389, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.0143, "n_shadow": 8, "mean_delta_shadow": -0.06159, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 8, "trunc_ratio_sum": 2.18461, "trunc_abs_frac": 2.10132, "mean_delta_full": -0.03964, "read_hit_rate": {"surface": 0.64639}, "mean_gate": {"surface": 0.12998}, "n_reads": 4177, "n_cand": 6462, "read_suppressed_frac": 0.1006, "dropout_reads_logged": 4177, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 226.5234832763672, "n_util_obs": 543, "util_count_spearman": 0.17676059077471268}, "probe_overhead_sec_total": 322.373, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.011369, "alpha": 0.03, "t": 4282.1} |
| {"event": "consolidate", "step": 5300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 11, "uncertain_staged": 262133, "t": 4321.4} |
| {"event": "train", "step": 5300, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38055, "bpb": 1.99171, "chunk_loss": 0.030399, "F": 0.19995, "G": 0.21207, "chunks": 204.75, "bytes_per_chunk": 5.001, "boundary_entropy": 0.11603, "router_noise": 0.0, "grad_norm": 0.6806, "lr": 3.99e-05, "router_trainable": true, "sec_per_step": 0.799, "probe_sec": 0.0502, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02598, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": 0.03426, "n_shadow": 5, "mean_delta_shadow": 0.07101, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.90849, "trunc_abs_frac": 0.94405, "mean_delta_full": -0.13602, "read_hit_rate": {"surface": 0.6161}, "mean_gate": {"surface": 0.13291}, "n_reads": 4017, "n_cand": 6520, "read_suppressed_frac": 0.0966, "dropout_reads_logged": 4017, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 229.540283203125, "n_util_obs": 546, "util_count_spearman": 0.18186309606978787}, "probe_overhead_sec_total": 325.545, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.015524, "alpha": 0.03, "t": 4322.1} |
| {"event": "consolidate", "step": 5350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 7, "uncertain_staged": 262133, "t": 4356.5} |
| {"event": "train", "step": 5350, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.31079, "bpb": 1.89106, "chunk_loss": 0.030287, "F": 0.19449, "G": 0.20683, "chunks": 199.16, "bytes_per_chunk": 5.142, "boundary_entropy": 0.11459, "router_noise": 0.0, "grad_norm": 0.6845, "lr": 3.86e-05, "router_trainable": true, "sec_per_step": 0.708, "probe_sec": 0.0893, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07597, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.17232, "n_shadow": 6, "mean_delta_shadow": 0.0856, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.70521, "trunc_abs_frac": 4.47071, "mean_delta_full": -0.61078, "read_hit_rate": {"surface": 0.63602}, "mean_gate": {"surface": 0.12679}, "n_reads": 4033, "n_cand": 6341, "read_suppressed_frac": 0.0972, "dropout_reads_logged": 4033, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 232.70663452148438, "n_util_obs": 552, "util_count_spearman": 0.1806453359713079}, "probe_overhead_sec_total": 328.245, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.011347, "alpha": 0.03, "t": 4357.5} |
| {"event": "consolidate", "step": 5400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 3, "uncertain_staged": 262133, "t": 4396.4} |
| {"event": "probe", "step": 5400, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3280.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.188473, "cps": 25.625, "mean_p": 0.20081, "boundary_entropy": 0.112464, "bpb": 1.879842}, "tr": {"chunks": 3617.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.199801, "cps": 28.257812, "mean_p": 0.202504, "boundary_entropy": 0.141716, "bpb": 2.013211}, "fi": {"chunks": 4290.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.228058, "cps": 33.515625, "mean_p": 0.229403, "boundary_entropy": 0.116511, "bpb": 2.11639}}, "bpb": {"en": 1.87984, "tr": 2.01321, "fi": 2.11639}, "gini_cpc": 0.042818, "gini_cps": 0.060189, "bpic": 5.071243, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0601034156850435, "fi": 1.2100287390227047}, "inventory": {"size": 327677, "admitted": 65536, "staged": 262141, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262141, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 235.79400634765625, "n_util_obs": 559, "util_count_spearman": 0.18386202469288768}, "t": 4397.1} |
| {"event": "eval", "step": 5400, "phase": 3, "bpb": {"en": 1.89445, "tr": 2.02298, "fi": 2.12375}, "bytes_per_chunk": {"en": 4.994, "tr": 5.404, "fi": 4.665}, "bpb_mean": 2.01373, "frozen_inventory": true, "t": 4398.2} |
| {"event": "train", "step": 5400, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38164, "bpb": 1.99328, "chunk_loss": 0.030689, "F": 0.20825, "G": 0.21975, "chunks": 213.25, "bytes_per_chunk": 4.802, "boundary_entropy": 0.11656, "router_noise": 0.0, "grad_norm": 0.7129, "lr": 3.73e-05, "router_trainable": true, "sec_per_step": 0.896, "probe_sec": 0.0349, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02704, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 11, "mean_delta_loo": 0.01806, "n_shadow": 3, "mean_delta_shadow": 0.04331, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -0.02179, "trunc_abs_frac": 0.27821, "mean_delta_full": 0.13586, "read_hit_rate": {"surface": 0.63987}, "mean_gate": {"surface": 0.13103}, "n_reads": 4346, "n_cand": 6792, "read_suppressed_frac": 0.1026, "dropout_reads_logged": 4346, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 235.86032104492188, "n_util_obs": 559, "util_count_spearman": 0.18363273713627112}, "probe_overhead_sec_total": 331.369, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.011854, "alpha": 0.03, "t": 4402.3} |
| {"event": "consolidate", "step": 5450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1, "uncertain_staged": 262133, "t": 4441.0} |
| {"event": "train", "step": 5450, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.31842, "bpb": 1.90207, "chunk_loss": 0.030585, "F": 0.20633, "G": 0.21744, "chunks": 211.28, "bytes_per_chunk": 4.847, "boundary_entropy": 0.1155, "router_noise": 0.0, "grad_norm": 0.6786, "lr": 3.62e-05, "router_trainable": true, "sec_per_step": 0.789, "probe_sec": 0.0969, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06008, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": -0.06479, "n_shadow": 7, "mean_delta_shadow": -0.0296, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -3.31565, "trunc_abs_frac": 0.54667, "mean_delta_full": -0.03149, "read_hit_rate": {"surface": 0.64868}, "mean_gate": {"surface": 0.13213}, "n_reads": 4365, "n_cand": 6729, "read_suppressed_frac": 0.0978, "dropout_reads_logged": 4365, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 238.95797729492188, "n_util_obs": 565, "util_count_spearman": 0.1730971705512369}, "probe_overhead_sec_total": 334.259, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.011019, "alpha": 0.03, "t": 4441.8} |
| {"event": "consolidate", "step": 5500, "retired": 1, "split_retired": 0, "merge_admitted": 0, "admitted": 1, "swapped": 1, "uncertain_staged": 262133, "t": 4485.4} |
| {"event": "train", "step": 5500, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.3076, "bpb": 1.88647, "chunk_loss": 0.030387, "F": 0.19827, "G": 0.21093, "chunks": 203.03, "bytes_per_chunk": 5.044, "boundary_entropy": 0.11597, "router_noise": 0.0, "grad_norm": 0.681, "lr": 3.51e-05, "router_trainable": true, "sec_per_step": 0.878, "probe_sec": 0.0463, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02283, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": 0.01857, "n_shadow": 5, "mean_delta_shadow": -0.38589, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -1.15653, "trunc_abs_frac": 1.11544, "mean_delta_full": 0.06453, "read_hit_rate": {"surface": 0.63542}, "mean_gate": {"surface": 0.12796}, "n_reads": 4108, "n_cand": 6465, "read_suppressed_frac": 0.0988, "dropout_reads_logged": 4108, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 242.07144165039062, "n_util_obs": 574, "util_count_spearman": 0.16748699300594616}, "probe_overhead_sec_total": 337.354, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.010343, "alpha": 0.03, "t": 4485.7} |
| {"event": "consolidate", "step": 5550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1, "uncertain_staged": 262133, "t": 4505.8} |
| {"event": "train", "step": 5550, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35792, "bpb": 1.95907, "chunk_loss": 0.030365, "F": 0.19662, "G": 0.20913, "chunks": 201.34, "bytes_per_chunk": 5.086, "boundary_entropy": 0.11715, "router_noise": 0.0, "grad_norm": 0.7116, "lr": 3.41e-05, "router_trainable": true, "sec_per_step": 0.409, "probe_sec": 0.0266, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02049, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": -0.07797, "n_shadow": 7, "mean_delta_shadow": -0.14493, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.98184, "trunc_abs_frac": 0.30877, "mean_delta_full": -0.08626, "read_hit_rate": {"surface": 0.63594}, "mean_gate": {"surface": 0.12974}, "n_reads": 4077, "n_cand": 6411, "read_suppressed_frac": 0.1065, "dropout_reads_logged": 4077, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 245.229736328125, "n_util_obs": 579, "util_count_spearman": 0.17548168843245357}, "probe_overhead_sec_total": 339.748, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.015141, "alpha": 0.03, "t": 4506.1} |
| {"event": "consolidate", "step": 5600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 262133, "t": 4543.3} |
| {"event": "probe", "step": 5600, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3275.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.188186, "cps": 25.585938, "mean_p": 0.200733, "boundary_entropy": 0.11318, "bpb": 1.875293}, "tr": {"chunks": 3606.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.199194, "cps": 28.171875, "mean_p": 0.200151, "boundary_entropy": 0.139723, "bpb": 2.00565}, "fi": {"chunks": 4323.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.229812, "cps": 33.773438, "mean_p": 0.231051, "boundary_entropy": 0.117788, "bpb": 2.108673}}, "bpb": {"en": 1.87529, "tr": 2.00565, "fi": 2.10867}, "gini_cpc": 0.044963, "gini_cps": 0.062359, "bpic": 5.063549, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.058492991545832, "fi": 1.221198235075222}, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 248.33517456054688, "n_util_obs": 586, "util_count_spearman": 0.18714146713052354}, "t": 4544.1} |
| {"event": "eval", "step": 5600, "phase": 3, "bpb": {"en": 1.89123, "tr": 2.01382, "fi": 2.11611}, "bytes_per_chunk": {"en": 5.023, "tr": 5.398, "fi": 4.642}, "bpb_mean": 2.00706, "frozen_inventory": true, "t": 4545.0} |
| {"event": "train", "step": 5600, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.34937, "bpb": 1.94672, "chunk_loss": 0.030445, "F": 0.20142, "G": 0.21243, "chunks": 206.25, "bytes_per_chunk": 4.965, "boundary_entropy": 0.11663, "router_noise": 0.0, "grad_norm": 0.7564, "lr": 3.33e-05, "router_trainable": true, "sec_per_step": 0.911, "probe_sec": 0.0767, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0465, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.01543, "n_shadow": 8, "mean_delta_shadow": -0.15407, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.34897, "trunc_abs_frac": 1.16577, "mean_delta_full": 0.2059, "read_hit_rate": {"surface": 0.63322}, "mean_gate": {"surface": 0.12846}, "n_reads": 4159, "n_cand": 6568, "read_suppressed_frac": 0.0995, "dropout_reads_logged": 4159, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 248.3986358642578, "n_util_obs": 586, "util_count_spearman": 0.18695459272758708}, "probe_overhead_sec_total": 343.071, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.010648, "alpha": 0.03, "t": 4551.7} |
| {"event": "consolidate", "step": 5650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 262133, "t": 4592.0} |
| {"event": "train", "step": 5650, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.33387, "bpb": 1.92437, "chunk_loss": 0.030472, "F": 0.20279, "G": 0.21378, "chunks": 207.66, "bytes_per_chunk": 4.931, "boundary_entropy": 0.11319, "router_noise": 0.0, "grad_norm": 0.7121, "lr": 3.25e-05, "router_trainable": true, "sec_per_step": 0.828, "probe_sec": 0.0604, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02595, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.21901, "n_shadow": 7, "mean_delta_shadow": -0.05539, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.60061, "trunc_abs_frac": 0.65846, "mean_delta_full": 0.14778, "read_hit_rate": {"surface": 0.63284}, "mean_gate": {"surface": 0.12499}, "n_reads": 4185, "n_cand": 6613, "read_suppressed_frac": 0.0922, "dropout_reads_logged": 4185, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 251.5631561279297, "n_util_obs": 590, "util_count_spearman": 0.19278889735927243}, "probe_overhead_sec_total": 346.022, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.011043, "alpha": 0.03, "t": 4593.1} |
| {"event": "consolidate", "step": 5700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 262133, "t": 4634.3} |
| {"event": "train", "step": 5700, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.33801, "bpb": 1.93035, "chunk_loss": 0.030431, "F": 0.19788, "G": 0.21026, "chunks": 202.62, "bytes_per_chunk": 5.054, "boundary_entropy": 0.11631, "router_noise": 0.0, "grad_norm": 0.7313, "lr": 3.18e-05, "router_trainable": true, "sec_per_step": 0.839, "probe_sec": 0.0975, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05886, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.17375, "n_shadow": 4, "mean_delta_shadow": -0.08228, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 1.44823, "trunc_abs_frac": 1.052, "mean_delta_full": 0.14325, "read_hit_rate": {"surface": 0.62988}, "mean_gate": {"surface": 0.12491}, "n_reads": 4064, "n_cand": 6452, "read_suppressed_frac": 0.0974, "dropout_reads_logged": 4064, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 254.7249298095703, "n_util_obs": 593, "util_count_spearman": 0.19186198959508766}, "probe_overhead_sec_total": 349.452, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.011511, "alpha": 0.03, "t": 4635.0} |
| {"event": "consolidate", "step": 5750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 3, "uncertain_staged": 262133, "t": 4667.2} |
| {"event": "train", "step": 5750, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38958, "bpb": 2.00474, "chunk_loss": 0.030646, "F": 0.20883, "G": 0.21946, "chunks": 213.84, "bytes_per_chunk": 4.789, "boundary_entropy": 0.11378, "router_noise": 0.0, "grad_norm": 0.7138, "lr": 3.13e-05, "router_trainable": true, "sec_per_step": 0.664, "probe_sec": 0.0942, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04612, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": -0.18358, "n_shadow": 10, "mean_delta_shadow": -0.04859, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 0.48143, "trunc_abs_frac": 0.60291, "mean_delta_full": -0.28682, "read_hit_rate": {"surface": 0.6328}, "mean_gate": {"surface": 0.13082}, "n_reads": 4310, "n_cand": 6811, "read_suppressed_frac": 0.1081, "dropout_reads_logged": 4310, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 257.8212890625, "n_util_obs": 595, "util_count_spearman": 0.18177280932490403}, "probe_overhead_sec_total": 352.021, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.010229, "alpha": 0.03, "t": 4668.2} |
| {"event": "consolidate", "step": 5800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 262133, "t": 4705.4} |
| {"event": "probe", "step": 5800, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3271.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.187956, "cps": 25.554688, "mean_p": 0.200498, "boundary_entropy": 0.111067, "bpb": 1.871674}, "tr": {"chunks": 3609.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.199359, "cps": 28.195312, "mean_p": 0.19917, "boundary_entropy": 0.135251, "bpb": 2.002143}, "fi": {"chunks": 4270.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.226995, "cps": 33.359375, "mean_p": 0.228882, "boundary_entropy": 0.114908, "bpb": 2.105127}}, "bpb": {"en": 1.87167, "tr": 2.00214, "fi": 2.10513}, "gini_cpc": 0.042366, "gini_cps": 0.059731, "bpic": 5.088072, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0606690750352028, "fi": 1.2077013942013834}, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 260.8974304199219, "n_util_obs": 602, "util_count_spearman": 0.1784045606548615}, "t": 4705.9} |
| {"event": "eval", "step": 5800, "phase": 3, "bpb": {"en": 1.88655, "tr": 2.00968, "fi": 2.11295}, "bytes_per_chunk": {"en": 5.015, "tr": 5.4, "fi": 4.675}, "bpb_mean": 2.00306, "frozen_inventory": true, "t": 4706.7} |
| {"event": "train", "step": 5800, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32083, "bpb": 1.90555, "chunk_loss": 0.030503, "F": 0.20004, "G": 0.21205, "chunks": 204.84, "bytes_per_chunk": 4.999, "boundary_entropy": 0.11609, "router_noise": 0.0, "grad_norm": 0.7227, "lr": 3.08e-05, "router_trainable": true, "sec_per_step": 0.874, "probe_sec": 0.0328, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02462, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.15698, "n_shadow": 8, "mean_delta_shadow": -0.02834, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.66822, "trunc_abs_frac": 0.27416, "mean_delta_full": 0.31453, "read_hit_rate": {"surface": 0.63882}, "mean_gate": {"surface": 0.12949}, "n_reads": 4167, "n_cand": 6523, "read_suppressed_frac": 0.1054, "dropout_reads_logged": 4167, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 260.96099853515625, "n_util_obs": 602, "util_count_spearman": 0.17853634250376407}, "probe_overhead_sec_total": 354.905, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.011525, "alpha": 0.03, "t": 4711.9} |
| {"event": "consolidate", "step": 5850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1, "uncertain_staged": 262133, "t": 4748.9} |
| {"event": "train", "step": 5850, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.33973, "bpb": 1.93283, "chunk_loss": 0.030454, "F": 0.20264, "G": 0.21346, "chunks": 207.5, "bytes_per_chunk": 4.935, "boundary_entropy": 0.10943, "router_noise": 0.0, "grad_norm": 0.7261, "lr": 3.05e-05, "router_trainable": true, "sec_per_step": 0.757, "probe_sec": 0.0685, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03798, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": 0.03725, "n_shadow": 5, "mean_delta_shadow": -0.04843, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.34545, "trunc_abs_frac": 0.34659, "mean_delta_full": 0.08118, "read_hit_rate": {"surface": 0.62031}, "mean_gate": {"surface": 0.12528}, "n_reads": 4099, "n_cand": 6608, "read_suppressed_frac": 0.0981, "dropout_reads_logged": 4099, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 264.06640625, "n_util_obs": 610, "util_count_spearman": 0.18800689932848252}, "probe_overhead_sec_total": 357.894, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.01553, "alpha": 0.03, "t": 4749.8} |
| {"event": "consolidate", "step": 5900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 262133, "t": 4790.6} |
| {"event": "train", "step": 5900, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.27076, "bpb": 1.83332, "chunk_loss": 0.030293, "F": 0.19495, "G": 0.20656, "chunks": 199.62, "bytes_per_chunk": 5.13, "boundary_entropy": 0.11002, "router_noise": 0.0, "grad_norm": 0.6993, "lr": 3.02e-05, "router_trainable": true, "sec_per_step": 0.844, "probe_sec": 0.0528, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02273, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.33193, "n_shadow": 6, "mean_delta_shadow": -0.09524, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.65626, "trunc_abs_frac": 0.82738, "mean_delta_full": 0.4068, "read_hit_rate": {"surface": 0.63955}, "mean_gate": {"surface": 0.12775}, "n_reads": 4065, "n_cand": 6356, "read_suppressed_frac": 0.1065, "dropout_reads_logged": 4065, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 267.23822021484375, "n_util_obs": 614, "util_count_spearman": 0.19195596336420237}, "probe_overhead_sec_total": 361.037, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.011969, "alpha": 0.03, "t": 4792.0} |
| {"event": "consolidate", "step": 5950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 3, "uncertain_staged": 262133, "t": 4834.7} |
| {"event": "train", "step": 5950, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.37403, "bpb": 1.98231, "chunk_loss": 0.030506, "F": 0.20435, "G": 0.21374, "chunks": 209.25, "bytes_per_chunk": 4.894, "boundary_entropy": 0.11149, "router_noise": 0.0, "grad_norm": 0.7358, "lr": 3.01e-05, "router_trainable": true, "sec_per_step": 0.86, "probe_sec": 0.0386, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02808, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 11, "mean_delta_loo": 0.0405, "n_shadow": 3, "mean_delta_shadow": -0.0427, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.17387, "trunc_abs_frac": 1.13934, "mean_delta_full": 0.17239, "read_hit_rate": {"surface": 0.6307}, "mean_gate": {"surface": 0.12849}, "n_reads": 4203, "n_cand": 6664, "read_suppressed_frac": 0.0997, "dropout_reads_logged": 4203, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 270.3709716796875, "n_util_obs": 619, "util_count_spearman": 0.1969297713087044}, "probe_overhead_sec_total": 364.151, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.011316, "alpha": 0.03, "t": 4835.0} |
| {"event": "consolidate", "step": 6000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 262133, "t": 4869.0} |
| {"event": "train", "step": 5999, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.37963, "bpb": 1.99038, "chunk_loss": 0.030528, "F": 0.20389, "G": 0.21462, "chunks": 208.78, "bytes_per_chunk": 4.905, "boundary_entropy": 0.10964, "router_noise": 0.0, "grad_norm": 0.7032, "lr": 3e-05, "router_trainable": true, "sec_per_step": 0.68, "probe_sec": 0.0393, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02216, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.04773, "n_shadow": 6, "mean_delta_shadow": 0.02965, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -0.81549, "trunc_abs_frac": 0.90791, "mean_delta_full": 0.03625, "read_hit_rate": {"surface": 0.62731}, "mean_gate": {"surface": 0.12782}, "n_reads": 4171, "n_cand": 6649, "read_suppressed_frac": 0.1036, "dropout_reads_logged": 4171, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 273.5026550292969, "n_util_obs": 634, "util_count_spearman": 0.20645851364294462}, "probe_overhead_sec_total": 367.551, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.011461, "alpha": 0.03, "t": 4869.0} |
| {"event": "probe", "step": 6000, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3289.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.18899, "cps": 25.695312, "mean_p": 0.200249, "boundary_entropy": 0.109269, "bpb": 1.869223}, "tr": {"chunks": 3625.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.200243, "cps": 28.320312, "mean_p": 0.199511, "boundary_entropy": 0.135421, "bpb": 1.997069}, "fi": {"chunks": 4302.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.228696, "cps": 33.609375, "mean_p": 0.229723, "boundary_entropy": 0.113161, "bpb": 2.101508}}, "bpb": {"en": 1.86922, "tr": 1.99707, "fi": 2.10151}, "gini_cpc": 0.042837, "gini_cps": 0.060212, "bpic": 5.058131, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.059540852068634, "fi": 1.2100930575042623}, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 273.5026550292969, "n_util_obs": 634, "util_count_spearman": 0.20645851364294462}, "t": 4869.6} |
| {"event": "eval", "step": 6000, "phase": 3, "bpb": {"en": 1.88388, "tr": 2.00572, "fi": 2.10909}, "bytes_per_chunk": {"en": 5.0, "tr": 5.405, "fi": 4.653}, "bpb_mean": 1.99956, "frozen_inventory": true, "t": 4870.5} |
| {"event": "reps_dumped", "dirs": ["/local/scratch/sas245/beetle-hnet/runs/ali_ali-surface-only_s2/probe/reps/th0.30", "/local/scratch/sas245/beetle-hnet/runs/ali_ali-surface-only_s2/probe/reps/th0.40", "/local/scratch/sas245/beetle-hnet/runs/ali_ali-surface-only_s2/probe/reps/th0.50", "/local/scratch/sas245/beetle-hnet/runs/ali_ali-surface-only_s2/probe/reps/th0.60", "/local/scratch/sas245/beetle-hnet/runs/ali_ali-surface-only_s2/probe/reps/th0.70"], "t": 4874.9} |
| {"event": "final", "run": "ali_ali-surface-only_s2", "arch": "ali", "objective": "ratio", "condition": "A", "step": 6000, "params": 22785088, "seed": 2, "per_lang": {"en": {"cpc": 0.18899040395334138, "cps": 25.6953125, "bpb": 1.8909751833701143, "bytes_per_chunk": 4.941684339507046, "reads_per_chunk": 0.6454404071822424}, "tr": {"cpc": 0.20024305363751865, "cps": 28.3203125, "bpb": 1.9969033396288134, "bytes_per_chunk": 5.424211386655907, "reads_per_chunk": 0.6042604260426042}, "fi": {"cpc": 0.2286959757588645, "cps": 33.609375, "bpb": 2.1269553840859663, "bytes_per_chunk": 4.6268598760965105, "reads_per_chunk": 0.5824434756517288}}, "bpb_mean": 2.0049446356949647, "gini_cpc": 0.042837223435351124, "gini_cps": 0.06021160247265822, "bpic": 5.0581312410841655, "mean_boundary_entropy": 0.10963739454746246, "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": 367.55076241493225, "probe_overhead_seq_frac": 0.499890625, "wall_sec": 4881.37065076828, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 262144, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 273.5026550292969, "n_util_obs": 634, "util_count_spearman": 0.20645851364294462}, "ali_params": 268096, "ali_config": {"enabled": true, "budget_B": 65536, "staging_capacity": 262144, "policy": "utility", "candidate_types": ["surface"], "gate": "vector", "retrieval": true, "memory_scope": "persistent", "read_dropout": 0.1, "write_rule": "mean", "eta_min": 0.0, "consolidate_every": 50, "probe_seqs": 2, "probe_reps": 8, "horizon": 32, "predictor": "mlp", "lambda_util": 0.1, "max_span": 64, "byte_classes_path": "/local/scratch/sas245/beetle-hnet/scratch/data/byte_classes.json", "gate_hidden": 0, "type_dim": 16, "gate_bias_init": -2.0}, "util_vs_frequency_spearman": 0.20645851364294462, "paired_vs_dropout": {"n": 1271, "pearson": 0.010131524760241336, "spearman": 0.057012994436028565, "mean_abs_discrepancy": 0.7288607954978943, "mean_paired": 0.05603782460093498, "mean_dropout": 0.03682003542780876}, "predictor_calibration": {"n": 3002, "all": {"n": 3002, "pearson": 0.25370888610082126, "spearman": 0.2509241538461091, "mae": 0.09459048509597778, "reliability": [{"pred": -0.01753804460167885, "obs": -0.01962674781680107, "n": 301}, {"pred": -0.003386806696653366, "obs": 0.013677716255187988, "n": 300}, {"pred": 0.004582475405186415, "obs": 0.00020681605383288115, "n": 300}, {"pred": 0.010856733657419682, "obs": 0.0157343577593565, "n": 300}, {"pred": 0.016129519790410995, "obs": 0.010498776100575924, "n": 300}, {"pred": 0.02049746923148632, "obs": 0.03469627723097801, "n": 300}, {"pred": 0.02548068016767502, "obs": 0.02551276609301567, "n": 300}, {"pred": 0.033579349517822266, "obs": 0.04783807694911957, "n": 300}, {"pred": 0.050531644374132156, "obs": 0.07529444992542267, "n": 300}, {"pred": 0.08998440951108932, "obs": 0.11622634530067444, "n": 301}]}, "admitted": {"n": 2980, "pearson": 0.25645960990978683, "spearman": 0.252463216133101, "mae": 0.09447316080331802, "reliability": [{"pred": -0.015881717205047607, "obs": -0.01660739816725254, "n": 298}, {"pred": -0.0031394371762871742, "obs": 0.013916200026869774, "n": 298}, {"pred": 0.004743920173496008, "obs": 0.0007726026233285666, "n": 298}, {"pred": 0.010953477583825588, "obs": 0.013493483886122704, "n": 298}, {"pred": 0.01621410623192787, "obs": 0.01386602595448494, "n": 298}, {"pred": 0.020552827045321465, "obs": 0.032228291034698486, "n": 298}, {"pred": 0.02552204392850399, "obs": 0.025296460837125778, "n": 298}, {"pred": 0.03361062332987785, "obs": 0.04926929250359535, "n": 298}, {"pred": 0.050478674471378326, "obs": 0.07597373425960541, "n": 298}, {"pred": 0.09004535526037216, "obs": 0.11916620284318924, "n": 298}]}, "staged": {"n": 22, "pearson": -0.2805988212074476, "spearman": -0.20609824957651046, "mae": 0.11048157513141632, "reliability": [{"pred": -0.08545661717653275, "obs": -0.1483001708984375, "n": 3}, {"pred": -0.06478649377822876, "obs": 0.14098358154296875, "n": 2}, {"pred": -0.04920625686645508, "obs": -0.0057373046875, "n": 2}, {"pred": -0.042590148746967316, "obs": -0.07889577746391296, "n": 2}, {"pred": -0.021269530057907104, "obs": -0.04656219482421875, "n": 2}, {"pred": -0.002642657607793808, "obs": -0.08152273297309875, "n": 2}, {"pred": 0.013415537774562836, "obs": -0.05863896757364273, "n": 2}, {"pred": 0.022900830954313278, "obs": -0.07929196953773499, "n": 2}, {"pred": 0.05842483043670654, "obs": -0.025918515399098396, "n": 2}, {"pred": 0.08392983675003052, "obs": -0.17579889297485352, "n": 3}]}}, "t": 4882.4} |
| {"event": "hub_done", "repo": "AdaptiveChunking/ali_ali-surface-only_s2", "pushed": [0, 1, 2, 4, 6, 8, 16, 32, 64, 128, 256, 512, 1000, 1024, 2048, 3000, 4096, 6000], "failed": [], "ok": true, "t": 4922.7} |
|
|
