| {"event": "hub", "repo": "AdaptiveChunking/ali_mem-static_s2", "keep_steps": [0, 1, 2, 4, 6, 8, 16, 32, 64, 128, 256, 512, 1000, 1024, 2048, 3000, 4096, 6000], "t": 3.7} |
| {"event": "start", "run": "ali_mem-static_s2", "arch": "ali", "objective": "ratio", "condition": "A", "params": 22785088, "langs": ["en", "tr", "fi"], "premium": {"en": 1.0, "tr": 1.12, "fi": 1.104}, "schedule": {"tau1": 0.4, "tau2": 0.7, "p1_objective": "global", "p2_objective": "global", "p3_objective": "global", "freeze_in_p3": false, "freeze_scope": "router"}, "corpus": "/local/scratch/sas245/beetle-hnet/scratch/data/tier1_corpus", "premium_kind": "flores_per_sent", "doc_filtered": {}, "weights": {"en": 0.74, "tr": 0.1536, "fi": 0.1064}, "device": "cuda", "ali_params": 268096, "ali_config": {"enabled": true, "budget_B": 65536, "staging_capacity": 262144, "policy": "frequency", "value_mode": "static", "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": 3.7} |
| {"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": 3.7} |
| {"event": "probe", "step": 0, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 10933.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.628225, "cps": 85.414062, "mean_p": 0.513249, "boundary_entropy": 0.686295, "bpb": 8.12598}, "tr": {"chunks": 8187.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.452245, "cps": 63.960938, "mean_p": 0.498002, "boundary_entropy": 0.686826, "bpb": 8.097536}, "fi": {"chunks": 13540.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.719792, "cps": 105.78125, "mean_p": 0.517169, "boundary_entropy": 0.686661, "bpb": 8.065117}}, "bpb": {"en": 8.12598, "tr": 8.09754, "fi": 8.06512}, "gini_cpc": 0.099077, "gini_cps": 0.109267, "bpic": 1.737048, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.7198781817505053, "fi": 1.1457544508645294}, "inventory": {"size": 0, "admitted": 0, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 0.0, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 5.6} |
| {"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": 0.017, "probe_sec": 0.024, "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": 2006, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 2006, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 9.757726669311523, "n_util_obs": 0, "util_count_spearman": NaN}, "probe_overhead_sec_total": 0.024, "probe_overhead_seq_frac": 0.0, "util_loss": -0.0, "alpha": 0.03, "t": 6.8} |
| {"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": 8.105501}, "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": 8.08287}, "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": 8.064965}}, "bpb": {"en": 8.1055, "tr": 8.08287, "fi": 8.06496}, "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": 2006, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 2006, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 9.757726669311523, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 7.7} |
| {"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": 0.51299, "boundary_entropy": 0.686317, "bpb": 8.043288}, "tr": {"chunks": 8249.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.45567, "cps": 64.445312, "mean_p": 0.498259, "boundary_entropy": 0.686832, "bpb": 8.039134}, "fi": {"chunks": 13469.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.716017, "cps": 105.226562, "mean_p": 0.516809, "boundary_entropy": 0.686691, "bpb": 8.006826}}, "bpb": {"en": 8.04329, "tr": 8.03913, "fi": 8.00683}, "gini_cpc": 0.09659, "gini_cps": 0.106752, "bpic": 1.740299, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.7287961371547589, "fi": 1.1451932495779524}, "inventory": {"size": 3538, "admitted": 3538, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 3538, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 10.44658088684082, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 9.2} |
| {"event": "probe", "step": 4, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 10723.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.616158, "cps": 83.773438, "mean_p": 0.512135, "boundary_entropy": 0.686378, "bpb": 7.832985}, "tr": {"chunks": 8397.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.463846, "cps": 65.601562, "mean_p": 0.498709, "boundary_entropy": 0.686846, "bpb": 7.885797}, "fi": {"chunks": 13218.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.702674, "cps": 103.265625, "mean_p": 0.51569, "boundary_entropy": 0.686774, "bpb": 7.806067}}, "bpb": {"en": 7.83299, "tr": 7.8858, "fi": 7.80607}, "gini_cpc": 0.089314, "gini_cps": 0.099388, "bpic": 1.754345, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.7528031297217783, "fi": 1.140411736823444}, "inventory": {"size": 5970, "admitted": 5970, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 5970, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 12.41859245300293, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 11.3} |
| {"event": "probe", "step": 6, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 10331.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.593633, "cps": 80.710938, "mean_p": 0.510418, "boundary_entropy": 0.686481, "bpb": 7.553844}, "tr": {"chunks": 8525.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.470916, "cps": 66.601562, "mean_p": 0.499242, "boundary_entropy": 0.686868, "bpb": 7.678905}, "fi": {"chunks": 12731.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.676785, "cps": 99.460938, "mean_p": 0.513617, "boundary_entropy": 0.686909, "bpb": 7.545371}}, "bpb": {"en": 7.55384, "tr": 7.67891, "fi": 7.54537}, "gini_cpc": 0.078816, "gini_cps": 0.088771, "bpic": 1.796055, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.7932783374421973, "fi": 1.1400722986380099}, "inventory": {"size": 7702, "admitted": 7702, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 7702, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 14.471695899963379, "n_util_obs": 1, "util_count_spearman": NaN}, "t": 13.3} |
| {"event": "probe", "step": 8, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 9671.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.555709, "cps": 75.554688, "mean_p": 0.507413, "boundary_entropy": 0.686612, "bpb": 7.298776}, "tr": {"chunks": 8649.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.477766, "cps": 67.570312, "mean_p": 0.499658, "boundary_entropy": 0.686901, "bpb": 7.478762}, "fi": {"chunks": 11846.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.629738, "cps": 92.546875, "mean_p": 0.510214, "boundary_entropy": 0.687083, "bpb": 7.311148}}, "bpb": {"en": 7.29878, "tr": 7.47876, "fi": 7.31115}, "gini_cpc": 0.060915, "gini_cps": 0.070653, "bpic": 1.88066, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.8597418797146783, "fi": 1.1332156974068897}, "inventory": {"size": 9776, "admitted": 9776, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 9776, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 14.903641700744629, "n_util_obs": 3, "util_count_spearman": -1.0}, "t": 15.6} |
| {"event": "probe", "step": 16, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3808.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.218813, "cps": 29.75, "mean_p": 0.484761, "boundary_entropy": 0.686194, "bpb": 6.725723}, "tr": {"chunks": 7350.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.40601, "cps": 57.421875, "mean_p": 0.494739, "boundary_entropy": 0.687008, "bpb": 7.013479}, "fi": {"chunks": 4534.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.241029, "cps": 35.421875, "mean_p": 0.486409, "boundary_entropy": 0.686896, "bpb": 6.787099}}, "bpb": {"en": 6.72572, "tr": 7.01348, "fi": 6.7871}, "gini_cpc": 0.144133, "gini_cps": 0.15048, "bpic": 3.615345, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.8555128577973752, "fi": 1.1015312256953178}, "inventory": {"size": 23736, "admitted": 23736, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 23736, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 9.627948760986328, "n_util_obs": 5, "util_count_spearman": -0.3}, "t": 19.4} |
| {"event": "probe", "step": 32, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 2615.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.150261, "cps": 20.429688, "mean_p": 0.486829, "boundary_entropy": 0.686957, "bpb": 6.305568}, "tr": {"chunks": 4603.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.254267, "cps": 35.960938, "mean_p": 0.490004, "boundary_entropy": 0.68755, "bpb": 6.652436}, "fi": {"chunks": 1354.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.071979, "cps": 10.578125, "mean_p": 0.479016, "boundary_entropy": 0.686935, "bpb": 6.421911}}, "bpb": {"en": 6.30557, "tr": 6.65244, "fi": 6.42191}, "gini_cpc": 0.255033, "gini_cps": 0.252683, "bpic": 6.618292, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.6921655548888561, "fi": 0.47902613427540713}, "inventory": {"size": 50391, "admitted": 50391, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 50391, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 6.432775974273682, "n_util_obs": 7, "util_count_spearman": -0.5357142857142857}, "t": 23.3} |
| {"event": "consolidate", "step": 50, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 30.7} |
| {"event": "train", "step": 50, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 4.21305, "bpb": 6.07814, "chunk_loss": 0.034282, "F": 0.22684, "G": 0.48857, "chunks": 232.28, "bytes_per_chunk": 4.408, "boundary_entropy": 0.69161, "router_noise": 0.0, "grad_norm": 2.4804, "lr": 5.1e-05, "router_trainable": true, "sec_per_step": 0.481, "probe_sec": 0.0355, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02811, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.00324, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.06707, "trunc_abs_frac": 0.29158, "mean_delta_full": 0.04004, "read_hit_rate": {"surface": 0.70556}, "mean_gate": {"surface": 0.11934}, "n_reads": 5219, "n_cand": 7397, "read_suppressed_frac": 0.1029, "dropout_reads_logged": 5219, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 6.1887359619140625, "n_util_obs": 9, "util_count_spearman": -0.5333333333333333}, "probe_overhead_sec_total": 2.718, "probe_overhead_seq_frac": 0.4804, "util_loss": 0.043012, "alpha": 0.03, "t": 30.9} |
| {"event": "probe", "step": 64, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 4913.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.282308, "cps": 38.382812, "mean_p": 0.485887, "boundary_entropy": 0.685571, "bpb": 5.552076}, "tr": {"chunks": 5574.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.307905, "cps": 43.546875, "mean_p": 0.487176, "boundary_entropy": 0.686518, "bpb": 5.954788}, "fi": {"chunks": 4803.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.255329, "cps": 37.523438, "mean_p": 0.484372, "boundary_entropy": 0.686348, "bpb": 5.717894}}, "bpb": {"en": 5.55208, "tr": 5.95479, "fi": 5.71789}, "gini_cpc": 0.041453, "gini_cps": 0.033617, "bpic": 3.710399, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0906710081384157, "fi": 0.9044364553945147}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 7.035858154296875, "n_util_obs": 12, "util_count_spearman": -0.23776223776223776}, "t": 37.3} |
| {"event": "consolidate", "step": 100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 50.7} |
| {"event": "train", "step": 100, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 3.24952, "bpb": 4.68807, "chunk_loss": 0.029614, "F": 0.16025, "G": 0.4532, "chunks": 164.09, "bytes_per_chunk": 6.24, "boundary_entropy": 0.6822, "router_noise": 0.0, "grad_norm": 1.4999, "lr": 0.000101, "router_trainable": true, "sec_per_step": 0.405, "probe_sec": 0.0316, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0184, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.01753, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.08264, "trunc_abs_frac": 0.56432, "mean_delta_full": -0.03693, "read_hit_rate": {"surface": 0.58904}, "mean_gate": {"surface": 0.11975}, "n_reads": 3073, "n_cand": 5217, "read_suppressed_frac": 0.1022, "dropout_reads_logged": 3073, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 9.414291381835938, "n_util_obs": 18, "util_count_spearman": -0.025799793601651185}, "probe_overhead_sec_total": 5.159, "probe_overhead_seq_frac": 0.4901, "util_loss": 0.000973, "alpha": 0.03, "t": 51.1} |
| {"event": "probe", "step": 128, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3515.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.201977, "cps": 27.460938, "mean_p": 0.414646, "boundary_entropy": 0.644736, "bpb": 4.14927}, "tr": {"chunks": 4558.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.251781, "cps": 35.609375, "mean_p": 0.435811, "boundary_entropy": 0.659486, "bpb": 4.402854}, "fi": {"chunks": 3715.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.197491, "cps": 29.023438, "mean_p": 0.426008, "boundary_entropy": 0.659605, "bpb": 4.342084}}, "bpb": {"en": 4.14927, "tr": 4.40285, "fi": 4.34208}, "gini_cpc": 0.055576, "gini_cps": 0.058987, "bpic": 4.812691, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.2465869044441367, "fi": 0.9777903020182874}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 11.288742065429688, "n_util_obs": 21, "util_count_spearman": -0.18961038961038962}, "t": 63.8} |
| {"event": "consolidate", "step": 150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 73.4} |
| {"event": "train", "step": 150, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.78927, "bpb": 4.02407, "chunk_loss": 0.032013, "F": 0.2081, "G": 0.38904, "chunks": 213.09, "bytes_per_chunk": 4.805, "boundary_entropy": 0.62334, "router_noise": 0.0, "grad_norm": 1.27, "lr": 0.000151, "router_trainable": true, "sec_per_step": 0.45, "probe_sec": 0.0342, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02743, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.00605, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -1.47195, "trunc_abs_frac": 0.88548, "mean_delta_full": 0.02466, "read_hit_rate": {"surface": 0.70797}, "mean_gate": {"surface": 0.12044}, "n_reads": 4805, "n_cand": 6787, "read_suppressed_frac": 0.1053, "dropout_reads_logged": 4805, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 12.905426025390625, "n_util_obs": 25, "util_count_spearman": -0.09230769230769231}, "probe_overhead_sec_total": 7.58, "probe_overhead_seq_frac": 0.4934, "util_loss": 0.000333, "alpha": 0.03, "t": 73.6} |
| {"event": "consolidate", "step": 200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 87.6} |
| {"event": "probe", "step": 200, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3431.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.19715, "cps": 26.804688, "mean_p": 0.368525, "boundary_entropy": 0.582, "bpb": 3.641577}, "tr": {"chunks": 4120.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.227587, "cps": 32.1875, "mean_p": 0.38236, "boundary_entropy": 0.596272, "bpb": 3.8107}, "fi": {"chunks": 4648.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.247089, "cps": 36.3125, "mean_p": 0.417327, "boundary_entropy": 0.639186, "bpb": 3.866947}}, "bpb": {"en": 3.64158, "tr": 3.8107, "fi": 3.86695}, "gini_cpc": 0.049556, "gini_cps": 0.066508, "bpic": 4.650545, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.154383383415664, "fi": 1.2533074986154826}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 15.91021728515625, "n_util_obs": 37, "util_count_spearman": 0.044570886676149835}, "t": 88.2} |
| {"event": "eval", "step": 200, "phase": 1, "bpb": {"en": 3.64748, "tr": 3.82176, "fi": 3.86485}, "bytes_per_chunk": {"en": 5.711, "tr": 5.242, "fi": 4.448}, "bpb_mean": 3.77803, "frozen_inventory": true, "t": 89.2} |
| {"event": "train", "step": 200, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.53462, "bpb": 3.65669, "chunk_loss": 0.030351, "F": 0.17416, "G": 0.36329, "chunks": 178.34, "bytes_per_chunk": 5.742, "boundary_entropy": 0.58556, "router_noise": 0.0, "grad_norm": 1.5273, "lr": 0.000201, "router_trainable": true, "sec_per_step": 0.332, "probe_sec": 0.0323, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02008, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.01202, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 0.07844, "trunc_abs_frac": 0.76574, "mean_delta_full": 0.17281, "read_hit_rate": {"surface": 0.62996}, "mean_gate": {"surface": 0.12205}, "n_reads": 3575, "n_cand": 5675, "read_suppressed_frac": 0.0959, "dropout_reads_logged": 3575, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 15.966079711914062, "n_util_obs": 37, "util_count_spearman": 0.04480796586059744}, "probe_overhead_sec_total": 10.058, "probe_overhead_seq_frac": 0.495, "util_loss": 0.000318, "alpha": 0.03, "t": 90.2} |
| {"event": "consolidate", "step": 250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 102.9} |
| {"event": "train", "step": 250, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.37767, "bpb": 3.43025, "chunk_loss": 0.03059, "F": 0.18094, "G": 0.34638, "chunks": 185.28, "bytes_per_chunk": 5.527, "boundary_entropy": 0.54812, "router_noise": 0.0, "grad_norm": 1.5779, "lr": 0.000251, "router_trainable": true, "sec_per_step": 0.259, "probe_sec": 0.0285, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0235, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.03199, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.3506, "trunc_abs_frac": 0.83232, "mean_delta_full": -0.14697, "read_hit_rate": {"surface": 0.64219}, "mean_gate": {"surface": 0.12485}, "n_reads": 3787, "n_cand": 5897, "read_suppressed_frac": 0.0911, "dropout_reads_logged": 3787, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 19.10980224609375, "n_util_obs": 45, "util_count_spearman": -0.11515530820832569}, "probe_overhead_sec_total": 12.711, "probe_overhead_seq_frac": 0.496, "util_loss": 0.000417, "alpha": 0.03, "t": 103.2} |
| {"event": "probe", "step": 256, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3637.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.208987, "cps": 28.414062, "mean_p": 0.353465, "boundary_entropy": 0.545168, "bpb": 3.341115}, "tr": {"chunks": 4607.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.254488, "cps": 35.992188, "mean_p": 0.333062, "boundary_entropy": 0.528435, "bpb": 3.520702}, "fi": {"chunks": 4985.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.265005, "cps": 38.945312, "mean_p": 0.380002, "boundary_entropy": 0.58409, "bpb": 3.652308}}, "bpb": {"en": 3.34111, "tr": 3.5207, "fi": 3.65231}, "gini_cpc": 0.051264, "gini_cps": 0.067932, "bpic": 4.288457, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.2177229187620997, "fi": 1.2680433428004227}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 19.474349975585938, "n_util_obs": 48, "util_count_spearman": -0.04711246200607903}, "t": 105.6} |
| {"event": "consolidate", "step": 300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 117.0} |
| {"event": "train", "step": 300, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.27809, "bpb": 3.28659, "chunk_loss": 0.03041, "F": 0.17734, "G": 0.3239, "chunks": 181.59, "bytes_per_chunk": 5.639, "boundary_entropy": 0.5034, "router_noise": 0.0, "grad_norm": 1.67, "lr": 0.0003, "router_trainable": true, "sec_per_step": 0.279, "probe_sec": 0.0247, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01927, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.03166, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.00142, "trunc_abs_frac": 0.95472, "mean_delta_full": 0.0689, "read_hit_rate": {"surface": 0.59197}, "mean_gate": {"surface": 0.13818}, "n_reads": 3421, "n_cand": 5779, "read_suppressed_frac": 0.0988, "dropout_reads_logged": 3421, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 22.158462524414062, "n_util_obs": 67, "util_count_spearman": -0.047888897757203286}, "probe_overhead_sec_total": 14.893, "probe_overhead_seq_frac": 0.4967, "util_loss": 0.00075, "alpha": 0.03, "t": 117.2} |
| {"event": "consolidate", "step": 350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 128.1} |
| {"event": "train", "step": 350, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.16791, "bpb": 3.12763, "chunk_loss": 0.03113, "F": 0.20285, "G": 0.30117, "chunks": 207.72, "bytes_per_chunk": 4.93, "boundary_entropy": 0.45157, "router_noise": 0.0, "grad_norm": 1.0664, "lr": 0.0002999, "router_trainable": true, "sec_per_step": 0.222, "probe_sec": 0.027, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02191, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.04585, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -0.07347, "trunc_abs_frac": 0.57511, "mean_delta_full": 0.04797, "read_hit_rate": {"surface": 0.65911}, "mean_gate": {"surface": 0.15783}, "n_reads": 4360, "n_cand": 6615, "read_suppressed_frac": 0.1021, "dropout_reads_logged": 4360, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 25.311660766601562, "n_util_obs": 86, "util_count_spearman": 0.0972786323829512}, "probe_overhead_sec_total": 17.417, "probe_overhead_seq_frac": 0.4972, "util_loss": 0.001031, "alpha": 0.03, "t": 128.3} |
| {"event": "consolidate", "step": 400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 138.7} |
| {"event": "probe", "step": 400, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3177.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.182555, "cps": 24.820312, "mean_p": 0.272976, "boundary_entropy": 0.39773, "bpb": 2.897627}, "tr": {"chunks": 3841.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.212175, "cps": 30.007812, "mean_p": 0.27489, "boundary_entropy": 0.435869, "bpb": 3.168954}, "fi": {"chunks": 4025.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.213971, "cps": 31.445312, "mean_p": 0.285011, "boundary_entropy": 0.455204, "bpb": 3.377762}}, "bpb": {"en": 2.89763, "tr": 3.16895, "fi": 3.37776}, "gini_cpc": 0.034408, "gini_cps": 0.051194, "bpic": 5.137372, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.1622529605405059, "fi": 1.1720898542024991}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 28.589736938476562, "n_util_obs": 104, "util_count_spearman": 0.20361677157793662}, "t": 139.4} |
| {"event": "eval", "step": 400, "phase": 1, "bpb": {"en": 2.9092, "tr": 3.17812, "fi": 3.37149}, "bytes_per_chunk": {"en": 5.119, "tr": 4.75, "fi": 5.06}, "bpb_mean": 3.15294, "frozen_inventory": true, "t": 140.4} |
| {"event": "train", "step": 400, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.08074, "bpb": 3.00188, "chunk_loss": 0.030877, "F": 0.20142, "G": 0.27767, "chunks": 206.25, "bytes_per_chunk": 4.965, "boundary_entropy": 0.40124, "router_noise": 0.0, "grad_norm": 1.1954, "lr": 0.0002998, "router_trainable": true, "sec_per_step": 0.262, "probe_sec": 0.03, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02485, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.03866, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 1.98158, "trunc_abs_frac": 0.4274, "mean_delta_full": 0.03247, "read_hit_rate": {"surface": 0.6591}, "mean_gate": {"surface": 0.20125}, "n_reads": 4329, "n_cand": 6568, "read_suppressed_frac": 0.104, "dropout_reads_logged": 4329, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 28.65484619140625, "n_util_obs": 105, "util_count_spearman": 0.2262388554841385}, "probe_overhead_sec_total": 19.854, "probe_overhead_seq_frac": 0.4975, "util_loss": 0.001704, "alpha": 0.03, "t": 141.3} |
| {"event": "consolidate", "step": 450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 151.9} |
| {"event": "train", "step": 450, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.99319, "bpb": 2.87556, "chunk_loss": 0.030872, "F": 0.20282, "G": 0.27229, "chunks": 207.69, "bytes_per_chunk": 4.93, "boundary_entropy": 0.36724, "router_noise": 0.0, "grad_norm": 1.5307, "lr": 0.0002995, "router_trainable": true, "sec_per_step": 0.216, "probe_sec": 0.0412, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03348, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.08889, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.93415, "trunc_abs_frac": 2.46237, "mean_delta_full": 0.12142, "read_hit_rate": {"surface": 0.67962}, "mean_gate": {"surface": 0.22605}, "n_reads": 4495, "n_cand": 6614, "read_suppressed_frac": 0.0974, "dropout_reads_logged": 4495, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 32.0809326171875, "n_util_obs": 116, "util_count_spearman": 0.24639218852112404}, "probe_overhead_sec_total": 22.674, "probe_overhead_seq_frac": 0.4978, "util_loss": 0.002326, "alpha": 0.03, "t": 152.2} |
| {"event": "consolidate", "step": 500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 163.3} |
| {"event": "train", "step": 500, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.90309, "bpb": 2.74557, "chunk_loss": 0.031266, "F": 0.21878, "G": 0.27094, "chunks": 224.03, "bytes_per_chunk": 4.571, "boundary_entropy": 0.33049, "router_noise": 0.0, "grad_norm": 1.6298, "lr": 0.0002992, "router_trainable": true, "sec_per_step": 0.229, "probe_sec": 0.0904, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08037, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.05447, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 5.33264, "trunc_abs_frac": 2.86395, "mean_delta_full": -0.03158, "read_hit_rate": {"surface": 0.71206}, "mean_gate": {"surface": 0.24283}, "n_reads": 5082, "n_cand": 7137, "read_suppressed_frac": 0.1004, "dropout_reads_logged": 5082, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 35.61328125, "n_util_obs": 133, "util_count_spearman": 0.18699564379061628}, "probe_overhead_sec_total": 25.007, "probe_overhead_seq_frac": 0.498, "util_loss": 0.002803, "alpha": 0.03, "t": 163.6} |
| {"event": "probe", "step": 512, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3479.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.199908, "cps": 27.179688, "mean_p": 0.264343, "boundary_entropy": 0.329677, "bpb": 2.692192}, "tr": {"chunks": 3494.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.193007, "cps": 27.296875, "mean_p": 0.259037, "boundary_entropy": 0.373017, "bpb": 3.030089}, "fi": {"chunks": 4308.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.229015, "cps": 33.65625, "mean_p": 0.267792, "boundary_entropy": 0.376814, "bpb": 3.22122}}, "bpb": {"en": 2.69219, "tr": 3.03009, "fi": 3.22122}, "gini_cpc": 0.038598, "gini_cps": 0.048991, "bpic": 5.028987, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9654772409363259, "fi": 1.1456013127910443}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 36.381103515625, "n_util_obs": 139, "util_count_spearman": 0.2260780195719201}, "t": 166.8} |
| {"event": "consolidate", "step": 550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 176.3} |
| {"event": "train", "step": 550, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.90391, "bpb": 2.74676, "chunk_loss": 0.030828, "F": 0.20349, "G": 0.26331, "chunks": 208.38, "bytes_per_chunk": 4.914, "boundary_entropy": 0.32351, "router_noise": 0.0, "grad_norm": 1.8179, "lr": 0.0002987, "router_trainable": true, "sec_per_step": 0.259, "probe_sec": 0.0371, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02813, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.04568, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 4.76428, "trunc_abs_frac": 1.58918, "mean_delta_full": -0.01603, "read_hit_rate": {"surface": 0.68791}, "mean_gate": {"surface": 0.2573}, "n_reads": 4565, "n_cand": 6636, "read_suppressed_frac": 0.0979, "dropout_reads_logged": 4565, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 39.130523681640625, "n_util_obs": 145, "util_count_spearman": 0.1649448215733426}, "probe_overhead_sec_total": 27.849, "probe_overhead_seq_frac": 0.4982, "util_loss": 0.003192, "alpha": 0.03, "t": 176.6} |
| {"event": "consolidate", "step": 600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 187.5} |
| {"event": "probe", "step": 600, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3488.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.200425, "cps": 27.25, "mean_p": 0.253984, "boundary_entropy": 0.297414, "bpb": 2.580902}, "tr": {"chunks": 3517.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.194277, "cps": 27.476562, "mean_p": 0.259831, "boundary_entropy": 0.342407, "bpb": 2.913992}, "fi": {"chunks": 4289.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.228005, "cps": 33.507812, "mean_p": 0.270436, "boundary_entropy": 0.359256, "bpb": 3.147049}}, "bpb": {"en": 2.5809, "tr": 2.91399, "fi": 3.14705}, "gini_cpc": 0.036109, "gini_cps": 0.047282, "bpic": 5.023198, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9693251048916318, "fi": 1.1376058239152478}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 42.701751708984375, "n_util_obs": 161, "util_count_spearman": 0.12863046974243442}, "t": 188.2} |
| {"event": "eval", "step": 600, "phase": 1, "bpb": {"en": 2.59525, "tr": 2.92278, "fi": 3.14822}, "bytes_per_chunk": {"en": 4.696, "tr": 5.576, "fi": 4.585}, "bpb_mean": 2.88875, "frozen_inventory": true, "t": 189.2} |
| {"event": "train", "step": 600, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.88726, "bpb": 2.72275, "chunk_loss": 0.030902, "F": 0.20877, "G": 0.25979, "chunks": 213.78, "bytes_per_chunk": 4.79, "boundary_entropy": 0.30204, "router_noise": 0.0, "grad_norm": 1.2541, "lr": 0.0002982, "router_trainable": true, "sec_per_step": 0.276, "probe_sec": 0.1114, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10365, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.17398, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 7, "trunc_ratio_sum": 4.05748, "trunc_abs_frac": 0.96282, "mean_delta_full": 0.03563, "read_hit_rate": {"surface": 0.69672}, "mean_gate": {"surface": 0.26876}, "n_reads": 4744, "n_cand": 6809, "read_suppressed_frac": 0.1031, "dropout_reads_logged": 4744, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 42.77313232421875, "n_util_obs": 161, "util_count_spearman": 0.1309108737505289}, "probe_overhead_sec_total": 30.663, "probe_overhead_seq_frac": 0.4983, "util_loss": 0.003678, "alpha": 0.03, "t": 190.3} |
| {"event": "consolidate", "step": 650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 201.6} |
| {"event": "train", "step": 650, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.83505, "bpb": 2.64741, "chunk_loss": 0.030864, "F": 0.20517, "G": 0.26101, "chunks": 210.09, "bytes_per_chunk": 4.874, "boundary_entropy": 0.29153, "router_noise": 0.0, "grad_norm": 1.1536, "lr": 0.0002975, "router_trainable": true, "sec_per_step": 0.229, "probe_sec": 0.0467, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03509, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.08559, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.79, "trunc_abs_frac": 0.59942, "mean_delta_full": -0.17041, "read_hit_rate": {"surface": 0.71873}, "mean_gate": {"surface": 0.27335}, "n_reads": 4809, "n_cand": 6691, "read_suppressed_frac": 0.0946, "dropout_reads_logged": 4809, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 46.37016296386719, "n_util_obs": 172, "util_count_spearman": 0.1413666523994068}, "probe_overhead_sec_total": 33.399, "probe_overhead_seq_frac": 0.4985, "util_loss": 0.004995, "alpha": 0.03, "t": 201.8} |
| {"event": "consolidate", "step": 700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 213.1} |
| {"event": "train", "step": 700, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.98639, "bpb": 2.86575, "chunk_loss": 0.031196, "F": 0.21185, "G": 0.26694, "chunks": 216.94, "bytes_per_chunk": 4.72, "boundary_entropy": 0.29507, "router_noise": 0.0, "grad_norm": 1.1306, "lr": 0.0002967, "router_trainable": true, "sec_per_step": 0.231, "probe_sec": 0.0901, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08187, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.00226, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.79631, "trunc_abs_frac": 0.69922, "mean_delta_full": 0.24382, "read_hit_rate": {"surface": 0.68509}, "mean_gate": {"surface": 0.28343}, "n_reads": 4734, "n_cand": 6910, "read_suppressed_frac": 0.0938, "dropout_reads_logged": 4734, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 50.059112548828125, "n_util_obs": 181, "util_count_spearman": 0.16655460084602317}, "probe_overhead_sec_total": 36.435, "probe_overhead_seq_frac": 0.4986, "util_loss": 0.006684, "alpha": 0.03, "t": 213.4} |
| {"event": "consolidate", "step": 750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 224.5} |
| {"event": "train", "step": 750, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.78054, "bpb": 2.56877, "chunk_loss": 0.03116, "F": 0.21982, "G": 0.26438, "chunks": 225.09, "bytes_per_chunk": 4.549, "boundary_entropy": 0.27485, "router_noise": 0.0, "grad_norm": 1.1913, "lr": 0.0002959, "router_trainable": true, "sec_per_step": 0.226, "probe_sec": 0.0273, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02109, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.05342, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -1.35791, "trunc_abs_frac": 1.07879, "mean_delta_full": -0.0494, "read_hit_rate": {"surface": 0.72737}, "mean_gate": {"surface": 0.2885}, "n_reads": 5216, "n_cand": 7171, "read_suppressed_frac": 0.1026, "dropout_reads_logged": 5216, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 53.774383544921875, "n_util_obs": 193, "util_count_spearman": 0.17884782390098872}, "probe_overhead_sec_total": 39.323, "probe_overhead_seq_frac": 0.4987, "util_loss": 0.007782, "alpha": 0.03, "t": 224.7} |
| {"event": "consolidate", "step": 800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 235.7} |
| {"event": "probe", "step": 800, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3420.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.196518, "cps": 26.71875, "mean_p": 0.243085, "boundary_entropy": 0.266513, "bpb": 2.43139}, "tr": {"chunks": 3729.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.205988, "cps": 29.132812, "mean_p": 0.253635, "boundary_entropy": 0.313246, "bpb": 2.764372}, "fi": {"chunks": 4124.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.219233, "cps": 32.21875, "mean_p": 0.263281, "boundary_entropy": 0.305083, "bpb": 2.986519}}, "bpb": {"en": 2.43139, "tr": 2.76437, "fi": 2.98652}, "gini_cpc": 0.024357, "gini_cps": 0.041633, "bpic": 5.032556, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0481895992813055, "fi": 1.115590448664322}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 57.42372131347656, "n_util_obs": 204, "util_count_spearman": 0.183490952421721}, "t": 236.4} |
| {"event": "eval", "step": 800, "phase": 1, "bpb": {"en": 2.44276, "tr": 2.77382, "fi": 2.98799}, "bytes_per_chunk": {"en": 4.801, "tr": 5.11, "fi": 4.755}, "bpb_mean": 2.73486, "frozen_inventory": true, "t": 237.5} |
| {"event": "train", "step": 800, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.85485, "bpb": 2.67599, "chunk_loss": 0.031012, "F": 0.21204, "G": 0.25977, "chunks": 217.12, "bytes_per_chunk": 4.716, "boundary_entropy": 0.27469, "router_noise": 0.0, "grad_norm": 1.2558, "lr": 0.0002949, "router_trainable": true, "sec_per_step": 0.279, "probe_sec": 0.0405, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03519, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.15555, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.80581, "trunc_abs_frac": 0.73104, "mean_delta_full": 0.16144, "read_hit_rate": {"surface": 0.69592}, "mean_gate": {"surface": 0.27167}, "n_reads": 4813, "n_cand": 6916, "read_suppressed_frac": 0.1001, "dropout_reads_logged": 4813, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 57.49676513671875, "n_util_obs": 204, "util_count_spearman": 0.18624384545018147}, "probe_overhead_sec_total": 41.892, "probe_overhead_seq_frac": 0.4988, "util_loss": 0.008575, "alpha": 0.03, "t": 238.6} |
| {"event": "consolidate", "step": 850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 249.6} |
| {"event": "train", "step": 850, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.80509, "bpb": 2.6042, "chunk_loss": 0.030883, "F": 0.21002, "G": 0.25727, "chunks": 215.06, "bytes_per_chunk": 4.761, "boundary_entropy": 0.2693, "router_noise": 0.0, "grad_norm": 0.9388, "lr": 0.0002938, "router_trainable": true, "sec_per_step": 0.224, "probe_sec": 0.0302, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01882, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.06918, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -12.80645, "trunc_abs_frac": 0.51615, "mean_delta_full": 0.01261, "read_hit_rate": {"surface": 0.69869}, "mean_gate": {"surface": 0.27033}, "n_reads": 4786, "n_cand": 6850, "read_suppressed_frac": 0.0928, "dropout_reads_logged": 4786, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 61.17291259765625, "n_util_obs": 221, "util_count_spearman": 0.16722069663246134}, "probe_overhead_sec_total": 44.382, "probe_overhead_seq_frac": 0.4988, "util_loss": 0.00872, "alpha": 0.03, "t": 249.8} |
| {"event": "consolidate", "step": 900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 260.5} |
| {"event": "train", "step": 900, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.60521, "bpb": 2.31584, "chunk_loss": 0.030946, "F": 0.21432, "G": 0.25294, "chunks": 219.47, "bytes_per_chunk": 4.666, "boundary_entropy": 0.24747, "router_noise": 0.0, "grad_norm": 1.2135, "lr": 0.0002927, "router_trainable": true, "sec_per_step": 0.218, "probe_sec": 0.029, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02286, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.01826, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 0.39576, "trunc_abs_frac": 0.78958, "mean_delta_full": -0.17468, "read_hit_rate": {"surface": 0.75297}, "mean_gate": {"surface": 0.26876}, "n_reads": 5264, "n_cand": 6991, "read_suppressed_frac": 0.0984, "dropout_reads_logged": 5264, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 64.90559387207031, "n_util_obs": 232, "util_count_spearman": 0.17068787027207252}, "probe_overhead_sec_total": 46.825, "probe_overhead_seq_frac": 0.4989, "util_loss": 0.008073, "alpha": 0.03, "t": 260.8} |
| {"event": "consolidate", "step": 950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 272.4} |
| {"event": "train", "step": 950, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.68459, "bpb": 2.43034, "chunk_loss": 0.030785, "F": 0.20618, "G": 0.24896, "chunks": 211.12, "bytes_per_chunk": 4.85, "boundary_entropy": 0.24715, "router_noise": 0.0, "grad_norm": 1.2035, "lr": 0.0002914, "router_trainable": true, "sec_per_step": 0.237, "probe_sec": 0.0731, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06606, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.093, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 1.11788, "trunc_abs_frac": 2.30466, "mean_delta_full": 0.42731, "read_hit_rate": {"surface": 0.71802}, "mean_gate": {"surface": 0.26391}, "n_reads": 4828, "n_cand": 6724, "read_suppressed_frac": 0.0942, "dropout_reads_logged": 4828, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 68.60150146484375, "n_util_obs": 245, "util_count_spearman": 0.17223770583103123}, "probe_overhead_sec_total": 49.805, "probe_overhead_seq_frac": 0.4989, "util_loss": 0.01038, "alpha": 0.03, "t": 272.6} |
| {"event": "consolidate", "step": 1000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 283.4} |
| {"event": "probe", "step": 1000, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3418.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.196403, "cps": 26.703125, "mean_p": 0.234809, "boundary_entropy": 0.241737, "bpb": 2.329496}, "tr": {"chunks": 3454.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.190797, "cps": 26.984375, "mean_p": 0.238452, "boundary_entropy": 0.282731, "bpb": 2.644541}, "fi": {"chunks": 4123.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.21918, "cps": 32.210938, "mean_p": 0.252459, "boundary_entropy": 0.272466, "bpb": 2.848404}}, "bpb": {"en": 2.3295, "tr": 2.64454, "fi": 2.8484}, "gini_cpc": 0.031205, "gini_cps": 0.042747, "bpic": 5.1598, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9714575851911954, "fi": 1.1159725524440884}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 72.2164306640625, "n_util_obs": 258, "util_count_spearman": 0.16673817328321655}, "t": 284.0} |
| {"event": "eval", "step": 1000, "phase": 1, "bpb": {"en": 2.33943, "tr": 2.65169, "fi": 2.85338}, "bytes_per_chunk": {"en": 4.826, "tr": 5.713, "fi": 4.751}, "bpb_mean": 2.61483, "frozen_inventory": true, "t": 285.1} |
| {"event": "train", "step": 1000, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.73632, "bpb": 2.50499, "chunk_loss": 0.03073, "F": 0.20654, "G": 0.24542, "chunks": 211.5, "bytes_per_chunk": 4.842, "boundary_entropy": 0.24934, "router_noise": 0.0, "grad_norm": 1.0063, "lr": 0.0002901, "router_trainable": true, "sec_per_step": 0.273, "probe_sec": 0.0425, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03482, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.20137, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 1.52906, "trunc_abs_frac": 0.56855, "mean_delta_full": 0.06091, "read_hit_rate": {"surface": 0.69448}, "mean_gate": {"surface": 0.26858}, "n_reads": 4678, "n_cand": 6736, "read_suppressed_frac": 0.096, "dropout_reads_logged": 4678, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 72.28791809082031, "n_util_obs": 259, "util_count_spearman": 0.1571143288090259}, "probe_overhead_sec_total": 52.417, "probe_overhead_seq_frac": 0.499, "util_loss": 0.010135, "alpha": 0.03, "t": 286.3} |
| {"event": "probe", "step": 1024, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3479.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.199908, "cps": 27.179688, "mean_p": 0.236061, "boundary_entropy": 0.245423, "bpb": 2.320256}, "tr": {"chunks": 3564.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.196873, "cps": 27.84375, "mean_p": 0.240465, "boundary_entropy": 0.27932, "bpb": 2.629687}, "fi": {"chunks": 4267.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.226835, "cps": 33.335938, "mean_p": 0.261208, "boundary_entropy": 0.266552, "bpb": 2.824695}}, "bpb": {"en": 2.32026, "tr": 2.62969, "fi": 2.82469}, "gini_cpc": 0.03203, "gini_cps": 0.046449, "bpic": 5.016092, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9848199446757486, "fi": 1.1346984219311478}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 74.00225830078125, "n_util_obs": 268, "util_count_spearman": 0.1897888215176243}, "t": 292.2} |
| {"event": "consolidate", "step": 1050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 299.0} |
| {"event": "train", "step": 1050, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.63688, "bpb": 2.36152, "chunk_loss": 0.030919, "F": 0.21106, "G": 0.24628, "chunks": 216.12, "bytes_per_chunk": 4.738, "boundary_entropy": 0.24091, "router_noise": 0.0, "grad_norm": 1.0912, "lr": 0.0002886, "router_trainable": true, "sec_per_step": 0.26, "probe_sec": 0.0361, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02649, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.00769, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.98527, "trunc_abs_frac": 0.5768, "mean_delta_full": 0.13395, "read_hit_rate": {"surface": 0.72865}, "mean_gate": {"surface": 0.2551}, "n_reads": 5016, "n_cand": 6884, "read_suppressed_frac": 0.0959, "dropout_reads_logged": 5016, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 76.04011535644531, "n_util_obs": 273, "util_count_spearman": 0.18761579294497716}, "probe_overhead_sec_total": 55.457, "probe_overhead_seq_frac": 0.499, "util_loss": 0.010308, "alpha": 0.03, "t": 299.3} |
| {"event": "consolidate", "step": 1100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 310.0} |
| {"event": "train", "step": 1100, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.74209, "bpb": 2.51331, "chunk_loss": 0.030854, "F": 0.21042, "G": 0.24563, "chunks": 215.47, "bytes_per_chunk": 4.752, "boundary_entropy": 0.23842, "router_noise": 0.0, "grad_norm": 1.0118, "lr": 0.0002871, "router_trainable": true, "sec_per_step": 0.219, "probe_sec": 0.0351, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02676, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.02055, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.33823, "trunc_abs_frac": 0.68263, "mean_delta_full": -0.32109, "read_hit_rate": {"surface": 0.70887}, "mean_gate": {"surface": 0.26863}, "n_reads": 4865, "n_cand": 6863, "read_suppressed_frac": 0.0948, "dropout_reads_logged": 4865, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 79.73275756835938, "n_util_obs": 285, "util_count_spearman": 0.16068997022530435}, "probe_overhead_sec_total": 57.797, "probe_overhead_seq_frac": 0.4991, "util_loss": 0.011428, "alpha": 0.03, "t": 310.2} |
| {"event": "consolidate", "step": 1150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 321.3} |
| {"event": "train", "step": 1150, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.6458, "bpb": 2.37438, "chunk_loss": 0.03073, "F": 0.20844, "G": 0.24106, "chunks": 213.44, "bytes_per_chunk": 4.798, "boundary_entropy": 0.22916, "router_noise": 0.0, "grad_norm": 0.8811, "lr": 0.0002855, "router_trainable": true, "sec_per_step": 0.226, "probe_sec": 0.0394, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03276, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.28709, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 8, "trunc_ratio_sum": 1.42283, "trunc_abs_frac": 0.92307, "mean_delta_full": 0.18741, "read_hit_rate": {"surface": 0.72315}, "mean_gate": {"surface": 0.26533}, "n_reads": 4916, "n_cand": 6798, "read_suppressed_frac": 0.0932, "dropout_reads_logged": 4916, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 83.44345092773438, "n_util_obs": 301, "util_count_spearman": 0.1346804547051703}, "probe_overhead_sec_total": 60.319, "probe_overhead_seq_frac": 0.4991, "util_loss": 0.011621, "alpha": 0.03, "t": 321.5} |
| {"event": "consolidate", "step": 1200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 332.8} |
| {"event": "probe", "step": 1200, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3454.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.198472, "cps": 26.984375, "mean_p": 0.229141, "boundary_entropy": 0.215173, "bpb": 2.261434}, "tr": {"chunks": 3537.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.195382, "cps": 27.632812, "mean_p": 0.232525, "boundary_entropy": 0.260171, "bpb": 2.547607}, "fi": {"chunks": 4398.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.233799, "cps": 34.359375, "mean_p": 0.262978, "boundary_entropy": 0.243318, "bpb": 2.729573}}, "bpb": {"en": 2.26143, "tr": 2.54761, "fi": 2.72957}, "gini_cpc": 0.040805, "gini_cps": 0.055258, "bpic": 4.981298, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9844332986042265, "fi": 1.1779995608650546}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 87.07058715820312, "n_util_obs": 314, "util_count_spearman": 0.15051258759790026}, "t": 333.4} |
| {"event": "eval", "step": 1200, "phase": 1, "bpb": {"en": 2.27257, "tr": 2.55946, "fi": 2.73807}, "bytes_per_chunk": {"en": 4.704, "tr": 5.599, "fi": 4.432}, "bpb_mean": 2.52337, "frozen_inventory": true, "t": 334.4} |
| {"event": "train", "step": 1200, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.61601, "bpb": 2.33141, "chunk_loss": 0.030749, "F": 0.21014, "G": 0.24184, "chunks": 215.19, "bytes_per_chunk": 4.759, "boundary_entropy": 0.22278, "router_noise": 0.0, "grad_norm": 0.9283, "lr": 0.0002837, "router_trainable": true, "sec_per_step": 0.276, "probe_sec": 0.05, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04387, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.05107, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 3.4311, "trunc_abs_frac": 1.69685, "mean_delta_full": 0.07707, "read_hit_rate": {"surface": 0.726}, "mean_gate": {"surface": 0.2753}, "n_reads": 4976, "n_cand": 6854, "read_suppressed_frac": 0.1069, "dropout_reads_logged": 4976, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 87.14640808105469, "n_util_obs": 315, "util_count_spearman": 0.1568421520422702}, "probe_overhead_sec_total": 62.935, "probe_overhead_seq_frac": 0.4992, "util_loss": 0.013993, "alpha": 0.03, "t": 335.3} |
| {"event": "consolidate", "step": 1250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 346.3} |
| {"event": "train", "step": 1250, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.63332, "bpb": 2.35639, "chunk_loss": 0.030551, "F": 0.19809, "G": 0.237, "chunks": 202.84, "bytes_per_chunk": 5.048, "boundary_entropy": 0.22867, "router_noise": 0.0, "grad_norm": 0.8609, "lr": 0.0002819, "router_trainable": true, "sec_per_step": 0.224, "probe_sec": 0.0905, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08381, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.09484, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 3.249, "trunc_abs_frac": 2.70051, "mean_delta_full": 0.01021, "read_hit_rate": {"surface": 0.69732}, "mean_gate": {"surface": 0.26337}, "n_reads": 4504, "n_cand": 6459, "read_suppressed_frac": 0.0964, "dropout_reads_logged": 4504, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 90.802490234375, "n_util_obs": 324, "util_count_spearman": 0.1652134980357978}, "probe_overhead_sec_total": 65.718, "probe_overhead_seq_frac": 0.4992, "util_loss": 0.014684, "alpha": 0.03, "t": 346.5} |
| {"event": "consolidate", "step": 1300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 357.0} |
| {"event": "train", "step": 1300, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.67327, "bpb": 2.41402, "chunk_loss": 0.030831, "F": 0.21069, "G": 0.24358, "chunks": 215.75, "bytes_per_chunk": 4.746, "boundary_entropy": 0.21482, "router_noise": 0.0, "grad_norm": 0.9306, "lr": 0.00028, "router_trainable": true, "sec_per_step": 0.213, "probe_sec": 0.0223, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01759, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.16494, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.68874, "trunc_abs_frac": 0.4937, "mean_delta_full": 0.86664, "read_hit_rate": {"surface": 0.70067}, "mean_gate": {"surface": 0.26456}, "n_reads": 4815, "n_cand": 6872, "read_suppressed_frac": 0.107, "dropout_reads_logged": 4815, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 94.43942260742188, "n_util_obs": 334, "util_count_spearman": 0.15552217691370313}, "probe_overhead_sec_total": 68.4, "probe_overhead_seq_frac": 0.4992, "util_loss": 0.013849, "alpha": 0.03, "t": 357.2} |
| {"event": "consolidate", "step": 1350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 367.4} |
| {"event": "train", "step": 1350, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.59875, "bpb": 2.30652, "chunk_loss": 0.030786, "F": 0.21265, "G": 0.23856, "chunks": 217.75, "bytes_per_chunk": 4.703, "boundary_entropy": 0.21116, "router_noise": 0.0, "grad_norm": 0.9803, "lr": 0.000278, "router_trainable": true, "sec_per_step": 0.211, "probe_sec": 0.0698, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0605, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.22398, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 1.14409, "trunc_abs_frac": 1.02023, "mean_delta_full": 0.45431, "read_hit_rate": {"surface": 0.72866}, "mean_gate": {"surface": 0.27368}, "n_reads": 5054, "n_cand": 6936, "read_suppressed_frac": 0.0987, "dropout_reads_logged": 5054, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 98.0986328125, "n_util_obs": 347, "util_count_spearman": 0.1357280255385343}, "probe_overhead_sec_total": 70.98, "probe_overhead_seq_frac": 0.4993, "util_loss": 0.015012, "alpha": 0.03, "t": 367.7} |
| {"event": "consolidate", "step": 1400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 379.3} |
| {"event": "probe", "step": 1400, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3432.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.197207, "cps": 26.8125, "mean_p": 0.223049, "boundary_entropy": 0.20335, "bpb": 2.209986}, "tr": {"chunks": 3390.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.187262, "cps": 26.484375, "mean_p": 0.225868, "boundary_entropy": 0.24451, "bpb": 2.491055}, "fi": {"chunks": 4470.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.237627, "cps": 34.921875, "mean_p": 0.254169, "boundary_entropy": 0.224011, "bpb": 2.632351}}, "bpb": {"en": 2.20999, "tr": 2.49106, "fi": 2.63235}, "gini_cpc": 0.053974, "gini_cps": 0.063762, "bpic": 5.024088, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9495678188021998, "fi": 1.2049595850962072}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 101.66990661621094, "n_util_obs": 360, "util_count_spearman": 0.1551625125587831}, "t": 379.8} |
| {"event": "eval", "step": 1400, "phase": 1, "bpb": {"en": 2.2212, "tr": 2.49967, "fi": 2.63865}, "bytes_per_chunk": {"en": 4.804, "tr": 5.987, "fi": 4.429}, "bpb_mean": 2.45317, "frozen_inventory": true, "t": 380.8} |
| {"event": "train", "step": 1400, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.59994, "bpb": 2.30823, "chunk_loss": 0.030606, "F": 0.20145, "G": 0.23266, "chunks": 206.28, "bytes_per_chunk": 4.964, "boundary_entropy": 0.20993, "router_noise": 0.0, "grad_norm": 0.8703, "lr": 0.0002759, "router_trainable": true, "sec_per_step": 0.28, "probe_sec": 0.0233, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01866, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.28494, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 1.45312, "trunc_abs_frac": 0.78973, "mean_delta_full": 0.04736, "read_hit_rate": {"surface": 0.70178}, "mean_gate": {"surface": 0.26034}, "n_reads": 4610, "n_cand": 6569, "read_suppressed_frac": 0.0987, "dropout_reads_logged": 4610, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 101.7410888671875, "n_util_obs": 360, "util_count_spearman": 0.15706475082528645}, "probe_overhead_sec_total": 73.806, "probe_overhead_seq_frac": 0.4993, "util_loss": 0.017205, "alpha": 0.03, "t": 381.8} |
| {"event": "consolidate", "step": 1450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 392.8} |
| {"event": "train", "step": 1450, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.6506, "bpb": 2.38131, "chunk_loss": 0.030506, "F": 0.19995, "G": 0.22984, "chunks": 204.75, "bytes_per_chunk": 5.001, "boundary_entropy": 0.20307, "router_noise": 0.0, "grad_norm": 0.9327, "lr": 0.0002738, "router_trainable": true, "sec_per_step": 0.226, "probe_sec": 0.086, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08078, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.02976, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 1.49393, "trunc_abs_frac": 1.33014, "mean_delta_full": 0.11655, "read_hit_rate": {"surface": 0.69003}, "mean_gate": {"surface": 0.26327}, "n_reads": 4499, "n_cand": 6520, "read_suppressed_frac": 0.1018, "dropout_reads_logged": 4499, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 105.34616088867188, "n_util_obs": 372, "util_count_spearman": 0.15305997272763763}, "probe_overhead_sec_total": 76.453, "probe_overhead_seq_frac": 0.4993, "util_loss": 0.016255, "alpha": 0.03, "t": 393.0} |
| {"event": "consolidate", "step": 1500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 403.9} |
| {"event": "train", "step": 1500, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.66693, "bpb": 2.40487, "chunk_loss": 0.03075, "F": 0.20743, "G": 0.23991, "chunks": 212.41, "bytes_per_chunk": 4.821, "boundary_entropy": 0.20654, "router_noise": 0.0, "grad_norm": 0.8864, "lr": 0.0002715, "router_trainable": true, "sec_per_step": 0.223, "probe_sec": 0.0264, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01928, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.17316, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.33753, "trunc_abs_frac": 0.33756, "mean_delta_full": 0.44208, "read_hit_rate": {"surface": 0.702}, "mean_gate": {"surface": 0.25616}, "n_reads": 4749, "n_cand": 6765, "read_suppressed_frac": 0.1009, "dropout_reads_logged": 4749, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 108.90460205078125, "n_util_obs": 381, "util_count_spearman": 0.13754177037160956}, "probe_overhead_sec_total": 79.2, "probe_overhead_seq_frac": 0.4993, "util_loss": 0.01637, "alpha": 0.03, "t": 404.2} |
| {"event": "consolidate", "step": 1550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 415.3} |
| {"event": "train", "step": 1550, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.58922, "bpb": 2.29276, "chunk_loss": 0.030538, "F": 0.19952, "G": 0.2282, "chunks": 204.31, "bytes_per_chunk": 5.012, "boundary_entropy": 0.19931, "router_noise": 0.0, "grad_norm": 0.9085, "lr": 0.0002692, "router_trainable": true, "sec_per_step": 0.225, "probe_sec": 0.024, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01925, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.00241, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.06913, "trunc_abs_frac": 0.27865, "mean_delta_full": 0.34387, "read_hit_rate": {"surface": 0.68076}, "mean_gate": {"surface": 0.2686}, "n_reads": 4429, "n_cand": 6506, "read_suppressed_frac": 0.0971, "dropout_reads_logged": 4429, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 112.39385986328125, "n_util_obs": 390, "util_count_spearman": 0.14389583216066226}, "probe_overhead_sec_total": 81.836, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.013225, "alpha": 0.03, "t": 415.4} |
| {"event": "consolidate", "step": 1600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 426.4} |
| {"event": "probe", "step": 1600, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3224.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.185255, "cps": 25.1875, "mean_p": 0.213398, "boundary_entropy": 0.189798, "bpb": 2.169735}, "tr": {"chunks": 3627.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.200354, "cps": 28.335938, "mean_p": 0.220997, "boundary_entropy": 0.227833, "bpb": 2.421376}, "fi": {"chunks": 4470.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.237627, "cps": 34.921875, "mean_p": 0.251151, "boundary_entropy": 0.205861, "bpb": 2.578023}}, "bpb": {"en": 2.16974, "tr": 2.42138, "fi": 2.57802}, "gini_cpc": 0.056021, "gini_cps": 0.073374, "bpic": 5.011218, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.08149892283047, "fi": 1.2826989131669302}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 115.73036193847656, "n_util_obs": 400, "util_count_spearman": 0.14477080696207237}, "t": 426.9} |
| {"event": "eval", "step": 1600, "phase": 1, "bpb": {"en": 2.18166, "tr": 2.43091, "fi": 2.58525}, "bytes_per_chunk": {"en": 5.033, "tr": 5.386, "fi": 4.426}, "bpb_mean": 2.39927, "frozen_inventory": true, "t": 427.9} |
| {"event": "train", "step": 1600, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.51523, "bpb": 2.18601, "chunk_loss": 0.030576, "F": 0.2038, "G": 0.23052, "chunks": 208.69, "bytes_per_chunk": 4.907, "boundary_entropy": 0.19387, "router_noise": 0.0, "grad_norm": 0.8363, "lr": 0.0002668, "router_trainable": true, "sec_per_step": 0.27, "probe_sec": 0.0256, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02045, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.25255, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.44195, "trunc_abs_frac": 0.36652, "mean_delta_full": 0.39636, "read_hit_rate": {"surface": 0.71727}, "mean_gate": {"surface": 0.25143}, "n_reads": 4767, "n_cand": 6646, "read_suppressed_frac": 0.1007, "dropout_reads_logged": 4767, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 115.80406188964844, "n_util_obs": 400, "util_count_spearman": 0.14542377991558578}, "probe_overhead_sec_total": 84.442, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.01399, "alpha": 0.03, "t": 428.9} |
| {"event": "consolidate", "step": 1650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 443.4} |
| {"event": "train", "step": 1650, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.58273, "bpb": 2.28339, "chunk_loss": 0.030443, "F": 0.1969, "G": 0.22491, "chunks": 201.62, "bytes_per_chunk": 5.079, "boundary_entropy": 0.19188, "router_noise": 0.0, "grad_norm": 0.8736, "lr": 0.0002643, "router_trainable": true, "sec_per_step": 0.296, "probe_sec": 0.1372, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.13199, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.23144, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -23.07095, "trunc_abs_frac": 1.32988, "mean_delta_full": -0.00452, "read_hit_rate": {"surface": 0.68738}, "mean_gate": {"surface": 0.26665}, "n_reads": 4413, "n_cand": 6420, "read_suppressed_frac": 0.097, "dropout_reads_logged": 4413, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 119.21922302246094, "n_util_obs": 410, "util_count_spearman": 0.14823184513478366}, "probe_overhead_sec_total": 87.02, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.017116, "alpha": 0.03, "t": 443.7} |
| {"event": "consolidate", "step": 1700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 455.9} |
| {"event": "train", "step": 1700, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.5437, "bpb": 2.22709, "chunk_loss": 0.030373, "F": 0.19177, "G": 0.22409, "chunks": 196.38, "bytes_per_chunk": 5.215, "boundary_entropy": 0.19651, "router_noise": 0.0, "grad_norm": 0.8056, "lr": 0.0002618, "router_trainable": true, "sec_per_step": 0.246, "probe_sec": 0.0607, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04935, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.17321, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.0198, "trunc_abs_frac": 22.84402, "mean_delta_full": -0.17572, "read_hit_rate": {"surface": 0.67722}, "mean_gate": {"surface": 0.24629}, "n_reads": 4234, "n_cand": 6252, "read_suppressed_frac": 0.1018, "dropout_reads_logged": 4234, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 122.6810302734375, "n_util_obs": 422, "util_count_spearman": 0.14833798224008468}, "probe_overhead_sec_total": 89.691, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.014359, "alpha": 0.03, "t": 456.0} |
| {"event": "consolidate", "step": 1750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 466.6} |
| {"event": "train", "step": 1750, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.57227, "bpb": 2.2683, "chunk_loss": 0.03053, "F": 0.20197, "G": 0.22788, "chunks": 206.81, "bytes_per_chunk": 4.951, "boundary_entropy": 0.189, "router_noise": 0.0, "grad_norm": 0.8578, "lr": 0.0002591, "router_trainable": true, "sec_per_step": 0.216, "probe_sec": 0.0343, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02683, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.13542, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -0.92613, "trunc_abs_frac": 1.3252, "mean_delta_full": 0.17228, "read_hit_rate": {"surface": 0.6898}, "mean_gate": {"surface": 0.25576}, "n_reads": 4543, "n_cand": 6586, "read_suppressed_frac": 0.1035, "dropout_reads_logged": 4543, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 126.10433959960938, "n_util_obs": 434, "util_count_spearman": 0.15195316207174336}, "probe_overhead_sec_total": 92.47, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.015677, "alpha": 0.03, "t": 466.9} |
| {"event": "consolidate", "step": 1800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 477.8} |
| {"event": "probe", "step": 1800, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3243.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.186347, "cps": 25.335938, "mean_p": 0.209591, "boundary_entropy": 0.175096, "bpb": 2.137008}, "tr": {"chunks": 3604.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.199083, "cps": 28.15625, "mean_p": 0.218181, "boundary_entropy": 0.214924, "bpb": 2.368196}, "fi": {"chunks": 4463.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.237255, "cps": 34.867188, "mean_p": 0.248865, "boundary_entropy": 0.190418, "bpb": 2.527422}}, "bpb": {"en": 2.13701, "tr": 2.3682, "fi": 2.52742}, "gini_cpc": 0.054503, "gini_cps": 0.071913, "bpic": 5.016092, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0683447063092513, "fi": 1.2731869395015643}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 129.4836883544922, "n_util_obs": 450, "util_count_spearman": 0.13657882192700802}, "t": 478.5} |
| {"event": "eval", "step": 1800, "phase": 1, "bpb": {"en": 2.15181, "tr": 2.37712, "fi": 2.53272}, "bytes_per_chunk": {"en": 5.027, "tr": 5.547, "fi": 4.466}, "bpb_mean": 2.35388, "frozen_inventory": true, "t": 479.4} |
| {"event": "train", "step": 1800, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.54198, "bpb": 2.2246, "chunk_loss": 0.030506, "F": 0.19934, "G": 0.22329, "chunks": 204.12, "bytes_per_chunk": 5.017, "boundary_entropy": 0.18406, "router_noise": 0.0, "grad_norm": 0.8201, "lr": 0.0002564, "router_trainable": true, "sec_per_step": 0.275, "probe_sec": 0.0316, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02572, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.05837, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.98398, "trunc_abs_frac": 0.92163, "mean_delta_full": -0.22473, "read_hit_rate": {"surface": 0.69046}, "mean_gate": {"surface": 0.25126}, "n_reads": 4488, "n_cand": 6500, "read_suppressed_frac": 0.0956, "dropout_reads_logged": 4488, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 129.5509796142578, "n_util_obs": 451, "util_count_spearman": 0.13716582731565474}, "probe_overhead_sec_total": 95.242, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.01626, "alpha": 0.03, "t": 480.6} |
| {"event": "consolidate", "step": 1850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 492.8} |
| {"event": "train", "step": 1850, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.54921, "bpb": 2.23504, "chunk_loss": 0.03052, "F": 0.19943, "G": 0.22267, "chunks": 204.22, "bytes_per_chunk": 5.014, "boundary_entropy": 0.17125, "router_noise": 0.0, "grad_norm": 0.7142, "lr": 0.0002537, "router_trainable": true, "sec_per_step": 0.258, "probe_sec": 0.0608, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03656, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.35201, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -2.55698, "trunc_abs_frac": 8.47466, "mean_delta_full": -0.08569, "read_hit_rate": {"surface": 0.69276}, "mean_gate": {"surface": 0.24048}, "n_reads": 4505, "n_cand": 6503, "read_suppressed_frac": 0.0915, "dropout_reads_logged": 4505, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 132.9830322265625, "n_util_obs": 458, "util_count_spearman": 0.13148000370350577}, "probe_overhead_sec_total": 97.841, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.015682, "alpha": 0.03, "t": 493.5} |
| {"event": "consolidate", "step": 1900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 510.0} |
| {"event": "train", "step": 1900, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.53395, "bpb": 2.21303, "chunk_loss": 0.030526, "F": 0.19763, "G": 0.22094, "chunks": 202.38, "bytes_per_chunk": 5.06, "boundary_entropy": 0.1793, "router_noise": 0.0, "grad_norm": 0.8231, "lr": 0.0002508, "router_trainable": true, "sec_per_step": 0.337, "probe_sec": 0.0578, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04378, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.1565, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.10434, "trunc_abs_frac": 0.6215, "mean_delta_full": -0.21175, "read_hit_rate": {"surface": 0.68824}, "mean_gate": {"surface": 0.2455}, "n_reads": 4435, "n_cand": 6444, "read_suppressed_frac": 0.0947, "dropout_reads_logged": 4435, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 136.42282104492188, "n_util_obs": 472, "util_count_spearman": 0.1378464665317334}, "probe_overhead_sec_total": 100.85, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.024995, "alpha": 0.03, "t": 510.4} |
| {"event": "consolidate", "step": 1950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 533.7} |
| {"event": "train", "step": 1950, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.48392, "bpb": 2.14084, "chunk_loss": 0.030488, "F": 0.19785, "G": 0.22187, "chunks": 202.59, "bytes_per_chunk": 5.054, "boundary_entropy": 0.17475, "router_noise": 0.0, "grad_norm": 0.756, "lr": 0.0002479, "router_trainable": true, "sec_per_step": 0.492, "probe_sec": 0.0673, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03674, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.09843, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -0.30636, "trunc_abs_frac": 0.19017, "mean_delta_full": 0.06462, "read_hit_rate": {"surface": 0.71074}, "mean_gate": {"surface": 0.25414}, "n_reads": 4585, "n_cand": 6451, "read_suppressed_frac": 0.0995, "dropout_reads_logged": 4585, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 139.8558807373047, "n_util_obs": 482, "util_count_spearman": 0.12271901331371407}, "probe_overhead_sec_total": 103.634, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.025076, "alpha": 0.03, "t": 535.0} |
| {"event": "consolidate", "step": 2000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 550.3} |
| {"event": "probe", "step": 2000, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3221.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.185083, "cps": 25.164062, "mean_p": 0.207446, "boundary_entropy": 0.156414, "bpb": 2.110183}, "tr": {"chunks": 3594.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.198531, "cps": 28.078125, "mean_p": 0.214088, "boundary_entropy": 0.199732, "bpb": 2.327583}, "fi": {"chunks": 4592.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.244112, "cps": 35.875, "mean_p": 0.248377, "boundary_entropy": 0.169728, "bpb": 2.473008}}, "bpb": {"en": 2.11018, "tr": 2.32758, "fi": 2.47301}, "gini_cpc": 0.062691, "gini_cps": 0.080126, "bpic": 4.973437, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0726571123261126, "fi": 1.3189349946528508}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 143.22044372558594, "n_util_obs": 498, "util_count_spearman": 0.12200516084049888}, "t": 550.9} |
| {"event": "eval", "step": 2000, "phase": 1, "bpb": {"en": 2.12124, "tr": 2.33694, "fi": 2.48405}, "bytes_per_chunk": {"en": 5.045, "tr": 5.752, "fi": 4.326}, "bpb_mean": 2.31408, "frozen_inventory": true, "t": 552.0} |
| {"event": "train", "step": 2000, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.64507, "bpb": 2.37333, "chunk_loss": 0.030635, "F": 0.20627, "G": 0.22888, "chunks": 211.22, "bytes_per_chunk": 4.848, "boundary_entropy": 0.17028, "router_noise": 0.0, "grad_norm": 0.8782, "lr": 0.000245, "router_trainable": true, "sec_per_step": 0.362, "probe_sec": 0.0263, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02091, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.06382, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 1.06046, "trunc_abs_frac": 1.03265, "mean_delta_full": 0.32572, "read_hit_rate": {"surface": 0.67578}, "mean_gate": {"surface": 0.24494}, "n_reads": 4546, "n_cand": 6727, "read_suppressed_frac": 0.099, "dropout_reads_logged": 4546, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 143.28707885742188, "n_util_obs": 499, "util_count_spearman": 0.12015276822386525}, "probe_overhead_sec_total": 106.272, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.017641, "alpha": 0.03, "t": 553.1} |
| {"event": "probe", "step": 2048, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3250.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.186749, "cps": 25.390625, "mean_p": 0.207806, "boundary_entropy": 0.158156, "bpb": 2.09228}, "tr": {"chunks": 3652.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.201735, "cps": 28.53125, "mean_p": 0.21448, "boundary_entropy": 0.201223, "bpb": 2.313997}, "fi": {"chunks": 4446.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.236351, "cps": 34.734375, "mean_p": 0.245599, "boundary_entropy": 0.173274, "bpb": 2.456344}}, "bpb": {"en": 2.09228, "tr": 2.314, "fi": 2.45634}, "gini_cpc": 0.052922, "gini_cps": 0.070262, "bpic": 4.999295, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0802417958774364, "fi": 1.2656054436234117}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 146.54800415039062, "n_util_obs": 508, "util_count_spearman": 0.13451731787815674}, "t": 563.3} |
| {"event": "consolidate", "step": 2050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 564.5} |
| {"event": "train", "step": 2050, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.48403, "bpb": 2.14101, "chunk_loss": 0.030381, "F": 0.19208, "G": 0.21445, "chunks": 196.69, "bytes_per_chunk": 5.206, "boundary_entropy": 0.1675, "router_noise": 0.0, "grad_norm": 0.783, "lr": 0.0002419, "router_trainable": true, "sec_per_step": 0.233, "probe_sec": 0.0241, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01817, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.04277, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 7, "trunc_ratio_sum": -14.73923, "trunc_abs_frac": 1.24214, "mean_delta_full": 0.01447, "read_hit_rate": {"surface": 0.68109}, "mean_gate": {"surface": 0.23853}, "n_reads": 4265, "n_cand": 6262, "read_suppressed_frac": 0.095, "dropout_reads_logged": 4265, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 146.75843811035156, "n_util_obs": 509, "util_count_spearman": 0.1337326821427461}, "probe_overhead_sec_total": 108.439, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.017329, "alpha": 0.03, "t": 564.7} |
| {"event": "consolidate", "step": 2100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 574.6} |
| {"event": "train", "step": 2100, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.54306, "bpb": 2.22617, "chunk_loss": 0.03043, "F": 0.19641, "G": 0.21693, "chunks": 201.12, "bytes_per_chunk": 5.091, "boundary_entropy": 0.16064, "router_noise": 0.0, "grad_norm": 0.7688, "lr": 0.0002388, "router_trainable": true, "sec_per_step": 0.202, "probe_sec": 0.0664, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05783, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.09254, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.63989, "trunc_abs_frac": 0.67751, "mean_delta_full": 0.52966, "read_hit_rate": {"surface": 0.69316}, "mean_gate": {"surface": 0.24575}, "n_reads": 4439, "n_cand": 6404, "read_suppressed_frac": 0.0978, "dropout_reads_logged": 4439, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 150.1815643310547, "n_util_obs": 524, "util_count_spearman": 0.14258924189718147}, "probe_overhead_sec_total": 110.246, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.022154, "alpha": 0.03, "t": 574.8} |
| {"event": "consolidate", "step": 2150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 589.9} |
| {"event": "train", "step": 2150, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.50151, "bpb": 2.16623, "chunk_loss": 0.030448, "F": 0.19943, "G": 0.21853, "chunks": 204.22, "bytes_per_chunk": 5.014, "boundary_entropy": 0.15994, "router_noise": 0.0, "grad_norm": 0.762, "lr": 0.0002357, "router_trainable": true, "sec_per_step": 0.306, "probe_sec": 0.0586, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04851, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.08495, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 1.17046, "trunc_abs_frac": 1.17046, "mean_delta_full": 0.14465, "read_hit_rate": {"surface": 0.70306}, "mean_gate": {"surface": 0.23645}, "n_reads": 4572, "n_cand": 6503, "read_suppressed_frac": 0.108, "dropout_reads_logged": 4572, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 153.66497802734375, "n_util_obs": 537, "util_count_spearman": 0.12899089328789023}, "probe_overhead_sec_total": 112.655, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.017153, "alpha": 0.03, "t": 590.1} |
| {"event": "consolidate", "step": 2200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 610.9} |
| {"event": "probe", "step": 2200, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3201.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.183934, "cps": 25.007812, "mean_p": 0.202715, "boundary_entropy": 0.148715, "bpb": 2.082117}, "tr": {"chunks": 3650.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.201624, "cps": 28.515625, "mean_p": 0.210576, "boundary_entropy": 0.189987, "bpb": 2.286767}, "fi": {"chunks": 4408.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.234331, "cps": 34.4375, "mean_p": 0.241105, "boundary_entropy": 0.162744, "bpb": 2.433458}}, "bpb": {"en": 2.08212, "tr": 2.28677, "fi": 2.43346}, "gini_cpc": 0.0542, "gini_cps": 0.071469, "bpic": 5.038813, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0961771858325429, "fi": 1.2739962464882135}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 157.04840087890625, "n_util_obs": 549, "util_count_spearman": 0.12327848610683426}, "t": 611.7} |
| {"event": "eval", "step": 2200, "phase": 1, "bpb": {"en": 2.09326, "tr": 2.29599, "fi": 2.44044}, "bytes_per_chunk": {"en": 5.035, "tr": 5.306, "fi": 4.387}, "bpb_mean": 2.27657, "frozen_inventory": true, "t": 612.5} |
| {"event": "train", "step": 2200, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.475, "bpb": 2.12797, "chunk_loss": 0.03038, "F": 0.19766, "G": 0.2157, "chunks": 202.41, "bytes_per_chunk": 5.059, "boundary_entropy": 0.15272, "router_noise": 0.0, "grad_norm": 0.6793, "lr": 0.0002325, "router_trainable": true, "sec_per_step": 0.472, "probe_sec": 0.0932, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08509, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.00391, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -0.82417, "trunc_abs_frac": 0.59623, "mean_delta_full": 0.08736, "read_hit_rate": {"surface": 0.70365}, "mean_gate": {"surface": 0.23788}, "n_reads": 4535, "n_cand": 6445, "read_suppressed_frac": 0.0977, "dropout_reads_logged": 4535, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 157.11843872070312, "n_util_obs": 549, "util_count_spearman": 0.12315630157683288}, "probe_overhead_sec_total": 116.178, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.02408, "alpha": 0.03, "t": 613.7} |
| {"event": "consolidate", "step": 2250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 631.4} |
| {"event": "train", "step": 2250, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.50018, "bpb": 2.1643, "chunk_loss": 0.030466, "F": 0.19974, "G": 0.2175, "chunks": 204.53, "bytes_per_chunk": 5.007, "boundary_entropy": 0.15349, "router_noise": 0.0, "grad_norm": 0.8047, "lr": 0.0002293, "router_trainable": true, "sec_per_step": 0.359, "probe_sec": 0.0576, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04686, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.09489, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.17475, "trunc_abs_frac": 0.44468, "mean_delta_full": -0.33634, "read_hit_rate": {"surface": 0.70352}, "mean_gate": {"surface": 0.24627}, "n_reads": 4582, "n_cand": 6513, "read_suppressed_frac": 0.1021, "dropout_reads_logged": 4582, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 160.57293701171875, "n_util_obs": 560, "util_count_spearman": 0.1259383081597756}, "probe_overhead_sec_total": 119.432, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.017923, "alpha": 0.03, "t": 631.7} |
| {"event": "consolidate", "step": 2300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 646.9} |
| {"event": "train", "step": 2300, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.5143, "bpb": 2.18467, "chunk_loss": 0.030381, "F": 0.19754, "G": 0.21301, "chunks": 202.28, "bytes_per_chunk": 5.062, "boundary_entropy": 0.14572, "router_noise": 0.0, "grad_norm": 0.7794, "lr": 0.000226, "router_trainable": true, "sec_per_step": 0.31, "probe_sec": 0.0282, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01998, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.00553, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.06346, "trunc_abs_frac": 0.49298, "mean_delta_full": -0.16398, "read_hit_rate": {"surface": 0.68576}, "mean_gate": {"surface": 0.23074}, "n_reads": 4417, "n_cand": 6441, "read_suppressed_frac": 0.094, "dropout_reads_logged": 4417, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 163.98739624023438, "n_util_obs": 570, "util_count_spearman": 0.1311694324389139}, "probe_overhead_sec_total": 122.77, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.022997, "alpha": 0.03, "t": 647.2} |
| {"event": "consolidate", "step": 2350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 661.3} |
| {"event": "train", "step": 2350, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.5624, "bpb": 2.25407, "chunk_loss": 0.030457, "F": 0.20264, "G": 0.21819, "chunks": 207.5, "bytes_per_chunk": 4.935, "boundary_entropy": 0.14814, "router_noise": 0.0, "grad_norm": 0.75, "lr": 0.0002226, "router_trainable": true, "sec_per_step": 0.289, "probe_sec": 0.0891, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07437, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.11206, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.83433, "trunc_abs_frac": 0.74499, "mean_delta_full": -0.35736, "read_hit_rate": {"surface": 0.6872}, "mean_gate": {"surface": 0.24698}, "n_reads": 4541, "n_cand": 6608, "read_suppressed_frac": 0.1033, "dropout_reads_logged": 4541, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 167.45045471191406, "n_util_obs": 590, "util_count_spearman": 0.1298887989930201}, "probe_overhead_sec_total": 125.878, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.014314, "alpha": 0.03, "t": 661.6} |
| {"event": "consolidate", "step": 2400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 674.1} |
| {"event": "probe", "step": 2400, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3167.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.18198, "cps": 24.742188, "mean_p": 0.197793, "boundary_entropy": 0.139083, "bpb": 2.064515}, "tr": {"chunks": 3587.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.198144, "cps": 28.023438, "mean_p": 0.209018, "boundary_entropy": 0.179932, "bpb": 2.264944}, "fi": {"chunks": 4510.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.239753, "cps": 35.234375, "mean_p": 0.244044, "boundary_entropy": 0.155802, "bpb": 2.385447}}, "bpb": {"en": 2.06452, "tr": 2.26494, "fi": 2.38545}, "gini_cpc": 0.062134, "gini_cps": 0.079486, "bpic": 5.036577, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0888219867923787, "fi": 1.3174699410084227}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 170.85174560546875, "n_util_obs": 592, "util_count_spearman": 0.13858957634918076}, "t": 674.7} |
| {"event": "eval", "step": 2400, "phase": 1, "bpb": {"en": 2.0747, "tr": 2.27227, "fi": 2.39364}, "bytes_per_chunk": {"en": 5.141, "tr": 5.709, "fi": 4.424}, "bpb_mean": 2.24687, "frozen_inventory": true, "t": 675.7} |
| {"event": "phase_transition", "step": 2400, "phase": 2, "objective": "global", "router_trainable": true, "encoder_trainable": true, "frozen_router_params": 0, "t": 676.7} |
| {"event": "train", "step": 2400, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.50419, "bpb": 2.17008, "chunk_loss": 0.030483, "F": 0.20203, "G": 0.21759, "chunks": 206.88, "bytes_per_chunk": 4.95, "boundary_entropy": 0.14574, "router_noise": 0.0, "grad_norm": 0.7508, "lr": 0.0002192, "router_trainable": true, "sec_per_step": 0.308, "probe_sec": 0.0291, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02032, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.00444, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 1.35704, "trunc_abs_frac": 0.74004, "mean_delta_full": 0.13171, "read_hit_rate": {"surface": 0.69824}, "mean_gate": {"surface": 0.23778}, "n_reads": 4600, "n_cand": 6588, "read_suppressed_frac": 0.0978, "dropout_reads_logged": 4600, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 170.92245483398438, "n_util_obs": 592, "util_count_spearman": 0.13950099234337832}, "probe_overhead_sec_total": 129.336, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.025664, "alpha": 0.03, "t": 677.0} |
| {"event": "consolidate", "step": 2450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 689.0} |
| {"event": "train", "step": 2450, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.46521, "bpb": 2.11385, "chunk_loss": 0.030366, "F": 0.19543, "G": 0.21209, "chunks": 200.12, "bytes_per_chunk": 5.117, "boundary_entropy": 0.14614, "router_noise": 0.0, "grad_norm": 0.7077, "lr": 0.0002158, "router_trainable": true, "sec_per_step": 0.245, "probe_sec": 0.0307, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01876, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.04557, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": -0.01473, "trunc_abs_frac": 0.01473, "mean_delta_full": -0.3606, "read_hit_rate": {"surface": 0.69083}, "mean_gate": {"surface": 0.24886}, "n_reads": 4402, "n_cand": 6372, "read_suppressed_frac": 0.1034, "dropout_reads_logged": 4402, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 174.3367462158203, "n_util_obs": 604, "util_count_spearman": 0.1399226931748612}, "probe_overhead_sec_total": 132.28, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.022664, "alpha": 0.03, "t": 689.3} |
| {"event": "consolidate", "step": 2500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 701.8} |
| {"event": "train", "step": 2500, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.43731, "bpb": 2.07361, "chunk_loss": 0.030486, "F": 0.20511, "G": 0.21851, "chunks": 210.03, "bytes_per_chunk": 4.875, "boundary_entropy": 0.13583, "router_noise": 0.0, "grad_norm": 0.7561, "lr": 0.0002123, "router_trainable": true, "sec_per_step": 0.257, "probe_sec": 0.0322, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02592, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.17903, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.81474, "trunc_abs_frac": 1.11862, "mean_delta_full": 0.1297, "read_hit_rate": {"surface": 0.717}, "mean_gate": {"surface": 0.23271}, "n_reads": 4796, "n_cand": 6689, "read_suppressed_frac": 0.0967, "dropout_reads_logged": 4796, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 177.74642944335938, "n_util_obs": 619, "util_count_spearman": 0.13905208239982544}, "probe_overhead_sec_total": 135.957, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.017311, "alpha": 0.03, "t": 702.1} |
| {"event": "consolidate", "step": 2550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 714.6} |
| {"event": "train", "step": 2550, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.46634, "bpb": 2.11547, "chunk_loss": 0.030395, "F": 0.19736, "G": 0.2124, "chunks": 202.09, "bytes_per_chunk": 5.067, "boundary_entropy": 0.13943, "router_noise": 0.0, "grad_norm": 0.7373, "lr": 0.0002088, "router_trainable": true, "sec_per_step": 0.253, "probe_sec": 0.0306, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02564, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.04992, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 1.04964, "trunc_abs_frac": 1.22034, "mean_delta_full": 0.2742, "read_hit_rate": {"surface": 0.69169}, "mean_gate": {"surface": 0.22801}, "n_reads": 4451, "n_cand": 6435, "read_suppressed_frac": 0.1056, "dropout_reads_logged": 4451, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 181.185791015625, "n_util_obs": 627, "util_count_spearman": 0.1449366160158976}, "probe_overhead_sec_total": 138.341, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.016104, "alpha": 0.03, "t": 714.8} |
| {"event": "consolidate", "step": 2600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 725.8} |
| {"event": "probe", "step": 2600, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3202.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.183991, "cps": 25.015625, "mean_p": 0.196829, "boundary_entropy": 0.125096, "bpb": 2.041415}, "tr": {"chunks": 3529.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.19494, "cps": 27.570312, "mean_p": 0.205474, "boundary_entropy": 0.169602, "bpb": 2.242588}, "fi": {"chunks": 4423.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.235128, "cps": 34.554688, "mean_p": 0.238538, "boundary_entropy": 0.137098, "bpb": 2.367742}}, "bpb": {"en": 2.04141, "tr": 2.24259, "fi": 2.36774}, "gini_cpc": 0.055518, "gini_cps": 0.072978, "bpic": 5.086247, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0595071687608382, "fi": 1.2779323040849722}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 184.54811096191406, "n_util_obs": 633, "util_count_spearman": 0.1363944370125871}, "t": 726.4} |
| {"event": "eval", "step": 2600, "phase": 2, "bpb": {"en": 2.05308, "tr": 2.25532, "fi": 2.37511}, "bytes_per_chunk": {"en": 5.1, "tr": 5.644, "fi": 4.463}, "bpb_mean": 2.22784, "frozen_inventory": true, "t": 727.6} |
| {"event": "train", "step": 2600, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.41703, "bpb": 2.04434, "chunk_loss": 0.030265, "F": 0.19278, "G": 0.20673, "chunks": 197.41, "bytes_per_chunk": 5.187, "boundary_entropy": 0.1292, "router_noise": 0.0, "grad_norm": 0.7774, "lr": 0.0002053, "router_trainable": true, "sec_per_step": 0.283, "probe_sec": 0.117, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10795, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.07126, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -1.61873, "trunc_abs_frac": 0.71406, "mean_delta_full": 0.06573, "read_hit_rate": {"surface": 0.6899}, "mean_gate": {"surface": 0.23374}, "n_reads": 4336, "n_cand": 6285, "read_suppressed_frac": 0.1019, "dropout_reads_logged": 4336, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 184.61465454101562, "n_util_obs": 633, "util_count_spearman": 0.13692676958824798}, "probe_overhead_sec_total": 141.313, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.015924, "alpha": 0.03, "t": 729.0} |
| {"event": "consolidate", "step": 2650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 743.8} |
| {"event": "train", "step": 2650, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.46977, "bpb": 2.12043, "chunk_loss": 0.030313, "F": 0.19434, "G": 0.20817, "chunks": 199.0, "bytes_per_chunk": 5.146, "boundary_entropy": 0.14015, "router_noise": 0.0, "grad_norm": 0.6935, "lr": 0.0002017, "router_trainable": true, "sec_per_step": 0.3, "probe_sec": 0.0365, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02646, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.20329, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 18.57533, "trunc_abs_frac": 9.67485, "mean_delta_full": 0.00911, "read_hit_rate": {"surface": 0.68687}, "mean_gate": {"surface": 0.23933}, "n_reads": 4352, "n_cand": 6336, "read_suppressed_frac": 0.1, "dropout_reads_logged": 4352, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 188.02886962890625, "n_util_obs": 642, "util_count_spearman": 0.1284435075638707}, "probe_overhead_sec_total": 145.277, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.020031, "alpha": 0.03, "t": 744.0} |
| {"event": "consolidate", "step": 2700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 757.9} |
| {"event": "train", "step": 2700, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.4491, "bpb": 2.09061, "chunk_loss": 0.030397, "F": 0.1994, "G": 0.21298, "chunks": 204.19, "bytes_per_chunk": 5.015, "boundary_entropy": 0.13529, "router_noise": 0.0, "grad_norm": 0.6771, "lr": 0.0001981, "router_trainable": true, "sec_per_step": 0.285, "probe_sec": 0.0725, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06564, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.0955, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 0.19378, "trunc_abs_frac": 0.62942, "mean_delta_full": 0.31637, "read_hit_rate": {"surface": 0.69209}, "mean_gate": {"surface": 0.2223}, "n_reads": 4500, "n_cand": 6502, "read_suppressed_frac": 0.0947, "dropout_reads_logged": 4500, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 191.46591186523438, "n_util_obs": 656, "util_count_spearman": 0.13297539460852972}, "probe_overhead_sec_total": 149.127, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.017518, "alpha": 0.03, "t": 758.2} |
| {"event": "consolidate", "step": 2750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 771.1} |
| {"event": "train", "step": 2750, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.45925, "bpb": 2.10525, "chunk_loss": 0.030476, "F": 0.20154, "G": 0.21508, "chunks": 206.38, "bytes_per_chunk": 4.962, "boundary_entropy": 0.13739, "router_noise": 0.0, "grad_norm": 0.6695, "lr": 0.0001945, "router_trainable": true, "sec_per_step": 0.263, "probe_sec": 0.0862, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07845, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.06734, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 2.50061, "trunc_abs_frac": 0.96411, "mean_delta_full": 0.11609, "read_hit_rate": {"surface": 0.70755}, "mean_gate": {"surface": 0.22589}, "n_reads": 4650, "n_cand": 6572, "read_suppressed_frac": 0.0985, "dropout_reads_logged": 4650, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 194.8795166015625, "n_util_obs": 667, "util_count_spearman": 0.1316705577972667}, "probe_overhead_sec_total": 152.488, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.017909, "alpha": 0.03, "t": 771.4} |
| {"event": "consolidate", "step": 2800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 784.4} |
| {"event": "probe", "step": 2800, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3187.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.183129, "cps": 24.898438, "mean_p": 0.195598, "boundary_entropy": 0.118892, "bpb": 2.025377}, "tr": {"chunks": 3484.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.192454, "cps": 27.21875, "mean_p": 0.201821, "boundary_entropy": 0.166329, "bpb": 2.195935}, "fi": {"chunks": 4435.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.235766, "cps": 34.648438, "mean_p": 0.23751, "boundary_entropy": 0.128929, "bpb": 2.319165}}, "bpb": {"en": 2.02538, "tr": 2.19593, "fi": 2.31916}, "gini_cpc": 0.0574, "gini_cps": 0.074914, "bpic": 5.10823, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0509199866517054, "fi": 1.2874305113953963}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 198.21234130859375, "n_util_obs": 675, "util_count_spearman": 0.1325272220627701}, "t": 785.1} |
| {"event": "eval", "step": 2800, "phase": 2, "bpb": {"en": 2.03663, "tr": 2.20623, "fi": 2.3264}, "bytes_per_chunk": {"en": 5.132, "tr": 5.739, "fi": 4.409}, "bpb_mean": 2.18975, "frozen_inventory": true, "t": 786.1} |
| {"event": "train", "step": 2800, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.41243, "bpb": 2.03771, "chunk_loss": 0.030287, "F": 0.19385, "G": 0.20708, "chunks": 198.5, "bytes_per_chunk": 5.159, "boundary_entropy": 0.12539, "router_noise": 0.0, "grad_norm": 0.6654, "lr": 0.0001909, "router_trainable": true, "sec_per_step": 0.321, "probe_sec": 0.0469, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03322, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.00267, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.86564, "trunc_abs_frac": 1.47232, "mean_delta_full": -0.16916, "read_hit_rate": {"surface": 0.70079}, "mean_gate": {"surface": 0.23212}, "n_reads": 4429, "n_cand": 6320, "read_suppressed_frac": 0.0946, "dropout_reads_logged": 4429, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 198.2810821533203, "n_util_obs": 675, "util_count_spearman": 0.1325187121436705}, "probe_overhead_sec_total": 155.93, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.01709, "alpha": 0.03, "t": 787.4} |
| {"event": "consolidate", "step": 2850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 800.5} |
| {"event": "train", "step": 2850, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.46326, "bpb": 2.11104, "chunk_loss": 0.030528, "F": 0.20721, "G": 0.21945, "chunks": 212.19, "bytes_per_chunk": 4.826, "boundary_entropy": 0.12684, "router_noise": 0.0, "grad_norm": 0.7849, "lr": 0.0001872, "router_trainable": true, "sec_per_step": 0.276, "probe_sec": 0.1236, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.11182, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.03318, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 6.42851, "trunc_abs_frac": 0.29974, "mean_delta_full": 0.0191, "read_hit_rate": {"surface": 0.70716}, "mean_gate": {"surface": 0.22023}, "n_reads": 4779, "n_cand": 6758, "read_suppressed_frac": 0.0908, "dropout_reads_logged": 4779, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 201.6824188232422, "n_util_obs": 687, "util_count_spearman": 0.13640982367546703}, "probe_overhead_sec_total": 159.429, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.019631, "alpha": 0.03, "t": 801.2} |
| {"event": "consolidate", "step": 2900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 814.1} |
| {"event": "train", "step": 2900, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.39439, "bpb": 2.01167, "chunk_loss": 0.030287, "F": 0.19363, "G": 0.20561, "chunks": 198.28, "bytes_per_chunk": 5.164, "boundary_entropy": 0.12973, "router_noise": 0.0, "grad_norm": 0.6528, "lr": 0.0001835, "router_trainable": true, "sec_per_step": 0.263, "probe_sec": 0.1152, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10505, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.09518, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -3.01088, "trunc_abs_frac": 2.49449, "mean_delta_full": -0.05375, "read_hit_rate": {"surface": 0.69983}, "mean_gate": {"surface": 0.22766}, "n_reads": 4418, "n_cand": 6313, "read_suppressed_frac": 0.0964, "dropout_reads_logged": 4418, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 205.12355041503906, "n_util_obs": 694, "util_count_spearman": 0.137372908682355}, "probe_overhead_sec_total": 163.499, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.018318, "alpha": 0.03, "t": 814.4} |
| {"event": "consolidate", "step": 2950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 827.7} |
| {"event": "train", "step": 2950, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.54246, "bpb": 2.2253, "chunk_loss": 0.030517, "F": 0.19998, "G": 0.21412, "chunks": 204.78, "bytes_per_chunk": 5.0, "boundary_entropy": 0.12879, "router_noise": 0.0, "grad_norm": 0.7294, "lr": 0.0001799, "router_trainable": true, "sec_per_step": 0.272, "probe_sec": 0.0262, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02094, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.17546, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.99283, "trunc_abs_frac": 0.99283, "mean_delta_full": 0.54443, "read_hit_rate": {"surface": 0.6867}, "mean_gate": {"surface": 0.22713}, "n_reads": 4478, "n_cand": 6521, "read_suppressed_frac": 0.0978, "dropout_reads_logged": 4478, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 208.5194854736328, "n_util_obs": 709, "util_count_spearman": 0.13255925489372355}, "probe_overhead_sec_total": 167.16, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.020471, "alpha": 0.03, "t": 828.0} |
| {"event": "consolidate", "step": 3000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 841.2} |
| {"event": "probe", "step": 3000, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3148.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.180888, "cps": 24.59375, "mean_p": 0.192946, "boundary_entropy": 0.117589, "bpb": 2.007737}, "tr": {"chunks": 3469.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.191626, "cps": 27.101562, "mean_p": 0.204328, "boundary_entropy": 0.164492, "bpb": 2.183013}, "fi": {"chunks": 4471.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.23768, "cps": 34.929688, "mean_p": 0.236854, "boundary_entropy": 0.125967, "bpb": 2.304395}}, "bpb": {"en": 2.00774, "tr": 2.18301, "fi": 2.3044}, "gini_cpc": 0.062048, "gini_cps": 0.079545, "bpic": 5.116522, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0593589618237755, "fi": 1.3139601163368078}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 211.8710479736328, "n_util_obs": 721, "util_count_spearman": 0.1417244611836218}, "t": 841.9} |
| {"event": "eval", "step": 3000, "phase": 2, "bpb": {"en": 2.01988, "tr": 2.19146, "fi": 2.30801}, "bytes_per_chunk": {"en": 5.191, "tr": 5.864, "fi": 4.366}, "bpb_mean": 2.17312, "frozen_inventory": true, "t": 843.1} |
| {"event": "train", "step": 3000, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.4921, "bpb": 2.15264, "chunk_loss": 0.030332, "F": 0.19302, "G": 0.20565, "chunks": 197.66, "bytes_per_chunk": 5.181, "boundary_entropy": 0.1228, "router_noise": 0.0, "grad_norm": 0.7427, "lr": 0.0001761, "router_trainable": true, "sec_per_step": 0.334, "probe_sec": 0.1197, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.11136, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.04911, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.31652, "trunc_abs_frac": 0.31652, "mean_delta_full": -0.79272, "read_hit_rate": {"surface": 0.67694}, "mean_gate": {"surface": 0.22272}, "n_reads": 4260, "n_cand": 6293, "read_suppressed_frac": 0.0962, "dropout_reads_logged": 4260, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 211.9343719482422, "n_util_obs": 721, "util_count_spearman": 0.14147601624520376}, "probe_overhead_sec_total": 171.24, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.016279, "alpha": 0.03, "t": 844.7} |
| {"event": "consolidate", "step": 3050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 857.5} |
| {"event": "train", "step": 3050, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.4408, "bpb": 2.07863, "chunk_loss": 0.030401, "F": 0.19708, "G": 0.20988, "chunks": 201.81, "bytes_per_chunk": 5.074, "boundary_entropy": 0.12275, "router_noise": 0.0, "grad_norm": 0.6922, "lr": 0.0001724, "router_trainable": true, "sec_per_step": 0.261, "probe_sec": 0.1119, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09893, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.04474, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 1.63398, "trunc_abs_frac": 1.03214, "mean_delta_full": 0.0623, "read_hit_rate": {"surface": 0.68861}, "mean_gate": {"surface": 0.2108}, "n_reads": 4425, "n_cand": 6426, "read_suppressed_frac": 0.0979, "dropout_reads_logged": 4425, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 215.32423400878906, "n_util_obs": 736, "util_count_spearman": 0.1542650835167406}, "probe_overhead_sec_total": 174.447, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.019111, "alpha": 0.03, "t": 857.8} |
| {"event": "consolidate", "step": 3100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 870.4} |
| {"event": "train", "step": 3100, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.48944, "bpb": 2.14881, "chunk_loss": 0.030285, "F": 0.19147, "G": 0.20465, "chunks": 196.06, "bytes_per_chunk": 5.223, "boundary_entropy": 0.11989, "router_noise": 0.0, "grad_norm": 0.6748, "lr": 0.0001687, "router_trainable": true, "sec_per_step": 0.258, "probe_sec": 0.0455, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03911, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.00354, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.64887, "trunc_abs_frac": 0.87802, "mean_delta_full": 0.09184, "read_hit_rate": {"surface": 0.67895}, "mean_gate": {"surface": 0.23046}, "n_reads": 4238, "n_cand": 6242, "read_suppressed_frac": 0.0953, "dropout_reads_logged": 4238, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 218.7040557861328, "n_util_obs": 748, "util_count_spearman": 0.15572570244971573}, "probe_overhead_sec_total": 178.492, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.018123, "alpha": 0.03, "t": 870.7} |
| {"event": "consolidate", "step": 3150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 884.0} |
| {"event": "train", "step": 3150, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.44857, "bpb": 2.08984, "chunk_loss": 0.03031, "F": 0.19357, "G": 0.20441, "chunks": 198.22, "bytes_per_chunk": 5.166, "boundary_entropy": 0.12123, "router_noise": 0.0, "grad_norm": 0.7218, "lr": 0.000165, "router_trainable": true, "sec_per_step": 0.272, "probe_sec": 0.0846, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07368, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.01042, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.18241, "trunc_abs_frac": 0.34302, "mean_delta_full": 0.29354, "read_hit_rate": {"surface": 0.68293}, "mean_gate": {"surface": 0.21233}, "n_reads": 4310, "n_cand": 6311, "read_suppressed_frac": 0.0965, "dropout_reads_logged": 4310, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 222.14599609375, "n_util_obs": 759, "util_count_spearman": 0.16233562199269375}, "probe_overhead_sec_total": 182.016, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.016641, "alpha": 0.03, "t": 884.3} |
| {"event": "consolidate", "step": 3200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 897.4} |
| {"event": "probe", "step": 3200, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3211.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.184508, "cps": 25.085938, "mean_p": 0.193777, "boundary_entropy": 0.108129, "bpb": 1.988849}, "tr": {"chunks": 3492.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.192896, "cps": 27.28125, "mean_p": 0.19871, "boundary_entropy": 0.152275, "bpb": 2.157392}, "fi": {"chunks": 4444.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.236245, "cps": 34.71875, "mean_p": 0.237131, "boundary_entropy": 0.123714, "bpb": 2.281436}}, "bpb": {"en": 1.98885, "tr": 2.15739, "fi": 2.28144}, "gini_cpc": 0.056206, "gini_cps": 0.073742, "bpic": 5.089441, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0454601857570163, "fi": 1.280400931616292}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 225.48924255371094, "n_util_obs": 773, "util_count_spearman": 0.15947830745669977}, "t": 898.1} |
| {"event": "eval", "step": 3200, "phase": 2, "bpb": {"en": 2.00114, "tr": 2.16653, "fi": 2.28726}, "bytes_per_chunk": {"en": 5.121, "tr": 5.728, "fi": 4.439}, "bpb_mean": 2.15164, "frozen_inventory": true, "t": 899.1} |
| {"event": "train", "step": 3200, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.43239, "bpb": 2.0665, "chunk_loss": 0.030339, "F": 0.1976, "G": 0.20806, "chunks": 202.34, "bytes_per_chunk": 5.061, "boundary_entropy": 0.11318, "router_noise": 0.0, "grad_norm": 0.6409, "lr": 0.0001613, "router_trainable": true, "sec_per_step": 0.32, "probe_sec": 0.0402, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03373, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.00992, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.19977, "trunc_abs_frac": 1.322, "mean_delta_full": 0.30789, "read_hit_rate": {"surface": 0.69533}, "mean_gate": {"surface": 0.21573}, "n_reads": 4480, "n_cand": 6443, "read_suppressed_frac": 0.1025, "dropout_reads_logged": 4480, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 225.55825805664062, "n_util_obs": 773, "util_count_spearman": 0.15938488920460225}, "probe_overhead_sec_total": 185.358, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.01916, "alpha": 0.03, "t": 900.3} |
| {"event": "consolidate", "step": 3250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 912.0} |
| {"event": "train", "step": 3250, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.43748, "bpb": 2.07384, "chunk_loss": 0.030304, "F": 0.19205, "G": 0.2044, "chunks": 196.66, "bytes_per_chunk": 5.207, "boundary_entropy": 0.11926, "router_noise": 0.0, "grad_norm": 0.7137, "lr": 0.0001576, "router_trainable": true, "sec_per_step": 0.239, "probe_sec": 0.1026, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08906, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.10973, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.54328, "trunc_abs_frac": 0.96144, "mean_delta_full": -0.20028, "read_hit_rate": {"surface": 0.68727}, "mean_gate": {"surface": 0.22177}, "n_reads": 4303, "n_cand": 6261, "read_suppressed_frac": 0.0978, "dropout_reads_logged": 4303, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 228.9490203857422, "n_util_obs": 788, "util_count_spearman": 0.17040223765163043}, "probe_overhead_sec_total": 188.742, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.016166, "alpha": 0.03, "t": 912.2} |
| {"event": "consolidate", "step": 3300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 925.1} |
| {"event": "train", "step": 3300, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.41361, "bpb": 2.03941, "chunk_loss": 0.030459, "F": 0.20068, "G": 0.2109, "chunks": 205.5, "bytes_per_chunk": 4.983, "boundary_entropy": 0.11304, "router_noise": 0.0, "grad_norm": 0.669, "lr": 0.0001539, "router_trainable": true, "sec_per_step": 0.263, "probe_sec": 0.033, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02398, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.10235, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.23708, "trunc_abs_frac": 0.84658, "mean_delta_full": -0.16482, "read_hit_rate": {"surface": 0.70293}, "mean_gate": {"surface": 0.21228}, "n_reads": 4600, "n_cand": 6544, "read_suppressed_frac": 0.0987, "dropout_reads_logged": 4600, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 232.3617401123047, "n_util_obs": 804, "util_count_spearman": 0.17748984268560425}, "probe_overhead_sec_total": 191.784, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.019165, "alpha": 0.03, "t": 925.4} |
| {"event": "consolidate", "step": 3350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 937.8} |
| {"event": "train", "step": 3350, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.35574, "bpb": 1.95592, "chunk_loss": 0.030325, "F": 0.19672, "G": 0.20583, "chunks": 201.44, "bytes_per_chunk": 5.083, "boundary_entropy": 0.10906, "router_noise": 0.0, "grad_norm": 0.752, "lr": 0.0001501, "router_trainable": true, "sec_per_step": 0.252, "probe_sec": 0.0363, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03134, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.0765, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.88682, "trunc_abs_frac": 0.88682, "mean_delta_full": 0.49182, "read_hit_rate": {"surface": 0.71204}, "mean_gate": {"surface": 0.20675}, "n_reads": 4567, "n_cand": 6414, "read_suppressed_frac": 0.1027, "dropout_reads_logged": 4567, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 235.78402709960938, "n_util_obs": 813, "util_count_spearman": 0.16174745986130962}, "probe_overhead_sec_total": 194.81, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.018356, "alpha": 0.03, "t": 938.0} |
| {"event": "consolidate", "step": 3400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 950.6} |
| {"event": "probe", "step": 3400, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3206.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.184221, "cps": 25.046875, "mean_p": 0.192794, "boundary_entropy": 0.10483, "bpb": 1.978183}, "tr": {"chunks": 3502.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.193449, "cps": 27.359375, "mean_p": 0.198128, "boundary_entropy": 0.149019, "bpb": 2.151134}, "fi": {"chunks": 4405.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.234171, "cps": 34.414062, "mean_p": 0.23659, "boundary_entropy": 0.117852, "bpb": 2.24893}}, "bpb": {"en": 1.97818, "tr": 2.15113, "fi": 2.24893}, "gini_cpc": 0.054426, "gini_cps": 0.071928, "bpic": 5.105012, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0500892015671466, "fi": 1.2711436476838769}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 239.15170288085938, "n_util_obs": 823, "util_count_spearman": 0.16116570915971934}, "t": 951.3} |
| {"event": "eval", "step": 3400, "phase": 2, "bpb": {"en": 1.98943, "tr": 2.16247, "fi": 2.25422}, "bytes_per_chunk": {"en": 5.106, "tr": 5.731, "fi": 4.505}, "bpb_mean": 2.13537, "frozen_inventory": true, "t": 952.3} |
| {"event": "train", "step": 3400, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.42138, "bpb": 2.05062, "chunk_loss": 0.030319, "F": 0.19379, "G": 0.20419, "chunks": 198.44, "bytes_per_chunk": 5.16, "boundary_entropy": 0.11314, "router_noise": 0.0, "grad_norm": 0.721, "lr": 0.0001465, "router_trainable": true, "sec_per_step": 0.311, "probe_sec": 0.0257, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01991, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.0197, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.89176, "trunc_abs_frac": 0.89035, "mean_delta_full": 0.27911, "read_hit_rate": {"surface": 0.68582}, "mean_gate": {"surface": 0.21589}, "n_reads": 4333, "n_cand": 6318, "read_suppressed_frac": 0.0999, "dropout_reads_logged": 4333, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 239.2177276611328, "n_util_obs": 823, "util_count_spearman": 0.16052292614502822}, "probe_overhead_sec_total": 197.701, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.016554, "alpha": 0.03, "t": 953.5} |
| {"event": "consolidate", "step": 3450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 966.0} |
| {"event": "train", "step": 3450, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.38572, "bpb": 1.99917, "chunk_loss": 0.030308, "F": 0.19571, "G": 0.20527, "chunks": 200.41, "bytes_per_chunk": 5.11, "boundary_entropy": 0.10964, "router_noise": 0.0, "grad_norm": 0.7881, "lr": 0.0001428, "router_trainable": true, "sec_per_step": 0.254, "probe_sec": 0.0624, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05465, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.14, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.94726, "trunc_abs_frac": 2.88176, "mean_delta_full": -0.09001, "read_hit_rate": {"surface": 0.69519}, "mean_gate": {"surface": 0.21536}, "n_reads": 4436, "n_cand": 6381, "read_suppressed_frac": 0.1014, "dropout_reads_logged": 4436, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 242.61825561523438, "n_util_obs": 839, "util_count_spearman": 0.15980967305235913}, "probe_overhead_sec_total": 200.975, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.017599, "alpha": 0.03, "t": 966.2} |
| {"event": "consolidate", "step": 3500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 982.9} |
| {"event": "train", "step": 3500, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.36462, "bpb": 1.96874, "chunk_loss": 0.03026, "F": 0.1918, "G": 0.2, "chunks": 196.41, "bytes_per_chunk": 5.214, "boundary_entropy": 0.10457, "router_noise": 0.0, "grad_norm": 0.6641, "lr": 0.0001391, "router_trainable": true, "sec_per_step": 0.346, "probe_sec": 0.4068, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04387, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.21456, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.80335, "trunc_abs_frac": 1.03254, "mean_delta_full": 0.70646, "read_hit_rate": {"surface": 0.70494}, "mean_gate": {"surface": 0.21488}, "n_reads": 4408, "n_cand": 6253, "read_suppressed_frac": 0.0973, "dropout_reads_logged": 4408, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 246.04295349121094, "n_util_obs": 851, "util_count_spearman": 0.14579648810010537}, "probe_overhead_sec_total": 204.558, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.015473, "alpha": 0.03, "t": 983.5} |
| {"event": "consolidate", "step": 3550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 997.2} |
| {"event": "train", "step": 3550, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.40619, "bpb": 2.0287, "chunk_loss": 0.030346, "F": 0.19583, "G": 0.2081, "chunks": 200.53, "bytes_per_chunk": 5.106, "boundary_entropy": 0.11259, "router_noise": 0.0, "grad_norm": 0.6583, "lr": 0.0001355, "router_trainable": true, "sec_per_step": 0.279, "probe_sec": 0.0508, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04208, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.02074, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.92419, "trunc_abs_frac": 0.67338, "mean_delta_full": 0.14612, "read_hit_rate": {"surface": 0.69193}, "mean_gate": {"surface": 0.21324}, "n_reads": 4418, "n_cand": 6385, "read_suppressed_frac": 0.0903, "dropout_reads_logged": 4418, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 249.47650146484375, "n_util_obs": 861, "util_count_spearman": 0.13331845764673958}, "probe_overhead_sec_total": 208.076, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.017673, "alpha": 0.03, "t": 997.5} |
| {"event": "consolidate", "step": 3600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1015.2} |
| {"event": "probe", "step": 3600, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3235.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.185887, "cps": 25.273438, "mean_p": 0.193217, "boundary_entropy": 0.097801, "bpb": 1.961006}, "tr": {"chunks": 3548.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.19599, "cps": 27.71875, "mean_p": 0.196708, "boundary_entropy": 0.136458, "bpb": 2.115893}, "fi": {"chunks": 4416.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.234756, "cps": 34.5, "mean_p": 0.235085, "boundary_entropy": 0.111067, "bpb": 2.226112}}, "bpb": {"en": 1.96101, "tr": 2.11589, "fi": 2.22611}, "gini_cpc": 0.052834, "gini_cps": 0.070304, "bpic": 5.065809, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0543453692467823, "fi": 1.2628943389284297}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 252.85740661621094, "n_util_obs": 869, "util_count_spearman": 0.14226083828952177}, "t": 1015.8} |
| {"event": "eval", "step": 3600, "phase": 2, "bpb": {"en": 1.97289, "tr": 2.12549, "fi": 2.23229}, "bytes_per_chunk": {"en": 5.076, "tr": 5.682, "fi": 4.459}, "bpb_mean": 2.11022, "frozen_inventory": true, "t": 1017.0} |
| {"event": "train", "step": 3600, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.4064, "bpb": 2.029, "chunk_loss": 0.030396, "F": 0.19888, "G": 0.20761, "chunks": 203.66, "bytes_per_chunk": 5.028, "boundary_entropy": 0.10367, "router_noise": 0.0, "grad_norm": 0.6469, "lr": 0.0001319, "router_trainable": true, "sec_per_step": 0.464, "probe_sec": 0.1288, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07517, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.06671, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 3.44343, "trunc_abs_frac": 2.59808, "mean_delta_full": -0.02881, "read_hit_rate": {"surface": 0.69499}, "mean_gate": {"surface": 0.2074}, "n_reads": 4507, "n_cand": 6485, "read_suppressed_frac": 0.1056, "dropout_reads_logged": 4507, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 252.9254913330078, "n_util_obs": 869, "util_count_spearman": 0.1420340161361482}, "probe_overhead_sec_total": 211.735, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.015933, "alpha": 0.03, "t": 1020.7} |
| {"event": "consolidate", "step": 3650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1043.6} |
| {"event": "train", "step": 3650, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.43687, "bpb": 2.07297, "chunk_loss": 0.030412, "F": 0.19913, "G": 0.20743, "chunks": 203.91, "bytes_per_chunk": 5.022, "boundary_entropy": 0.10489, "router_noise": 0.0, "grad_norm": 0.6854, "lr": 0.0001283, "router_trainable": true, "sec_per_step": 0.472, "probe_sec": 0.1091, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09503, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.04584, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 9, "trunc_ratio_sum": -1.10192, "trunc_abs_frac": 0.62282, "mean_delta_full": -0.04187, "read_hit_rate": {"surface": 0.69536}, "mean_gate": {"surface": 0.20622}, "n_reads": 4515, "n_cand": 6493, "read_suppressed_frac": 0.0986, "dropout_reads_logged": 4515, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 256.3682861328125, "n_util_obs": 877, "util_count_spearman": 0.13632816009413337}, "probe_overhead_sec_total": 215.325, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.014721, "alpha": 0.03, "t": 1044.3} |
| {"event": "consolidate", "step": 3700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1057.6} |
| {"event": "train", "step": 3700, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.39365, "bpb": 2.01062, "chunk_loss": 0.030351, "F": 0.1991, "G": 0.20765, "chunks": 203.88, "bytes_per_chunk": 5.023, "boundary_entropy": 0.10316, "router_noise": 0.0, "grad_norm": 0.6496, "lr": 0.0001247, "router_trainable": true, "sec_per_step": 0.27, "probe_sec": 0.0583, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05371, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.00147, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -0.4643, "trunc_abs_frac": 1.12702, "mean_delta_full": 0.07556, "read_hit_rate": {"surface": 0.70456}, "mean_gate": {"surface": 0.21285}, "n_reads": 4574, "n_cand": 6492, "read_suppressed_frac": 0.1034, "dropout_reads_logged": 4574, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 259.7848815917969, "n_util_obs": 887, "util_count_spearman": 0.1291450110225217}, "probe_overhead_sec_total": 219.03, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.015899, "alpha": 0.03, "t": 1057.8} |
| {"event": "consolidate", "step": 3750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1071.0} |
| {"event": "train", "step": 3750, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.37856, "bpb": 1.98884, "chunk_loss": 0.030298, "F": 0.19443, "G": 0.20369, "chunks": 199.09, "bytes_per_chunk": 5.143, "boundary_entropy": 0.10726, "router_noise": 0.0, "grad_norm": 0.6605, "lr": 0.0001212, "router_trainable": true, "sec_per_step": 0.268, "probe_sec": 0.117, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10829, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.07172, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.2423, "trunc_abs_frac": 0.2423, "mean_delta_full": -0.14661, "read_hit_rate": {"surface": 0.69585}, "mean_gate": {"surface": 0.21291}, "n_reads": 4411, "n_cand": 6339, "read_suppressed_frac": 0.1011, "dropout_reads_logged": 4411, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 263.1944580078125, "n_util_obs": 898, "util_count_spearman": 0.12582076790406876}, "probe_overhead_sec_total": 222.767, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.015396, "alpha": 0.03, "t": 1071.2} |
| {"event": "consolidate", "step": 3800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1089.4} |
| {"event": "probe", "step": 3800, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3160.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.181578, "cps": 24.6875, "mean_p": 0.188678, "boundary_entropy": 0.09627, "bpb": 1.95123}, "tr": {"chunks": 3498.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.193228, "cps": 27.328125, "mean_p": 0.195623, "boundary_entropy": 0.137079, "bpb": 2.110804}, "fi": {"chunks": 4353.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.231407, "cps": 34.007812, "mean_p": 0.231581, "boundary_entropy": 0.105461, "bpb": 2.210965}}, "bpb": {"en": 1.95123, "tr": 2.1108, "fi": 2.21096}, "gini_cpc": 0.054798, "gini_cps": 0.072231, "bpic": 5.152302, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0641584337724288, "fi": 1.2744236472196109}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 266.53228759765625, "n_util_obs": 912, "util_count_spearman": 0.12901223572500797}, "t": 1090.0} |
| {"event": "eval", "step": 3800, "phase": 2, "bpb": {"en": 1.96426, "tr": 2.12021, "fi": 2.21663}, "bytes_per_chunk": {"en": 5.162, "tr": 5.737, "fi": 4.545}, "bpb_mean": 2.10037, "frozen_inventory": true, "t": 1091.0} |
| {"event": "train", "step": 3800, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.45051, "bpb": 2.09264, "chunk_loss": 0.030368, "F": 0.19861, "G": 0.20676, "chunks": 203.38, "bytes_per_chunk": 5.035, "boundary_entropy": 0.10163, "router_noise": 0.0, "grad_norm": 0.6884, "lr": 0.0001177, "router_trainable": true, "sec_per_step": 0.421, "probe_sec": 0.0596, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05309, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.00759, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.66471, "trunc_abs_frac": 0.56552, "mean_delta_full": -0.08051, "read_hit_rate": {"surface": 0.67989}, "mean_gate": {"surface": 0.20029}, "n_reads": 4403, "n_cand": 6476, "read_suppressed_frac": 0.1079, "dropout_reads_logged": 4403, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 266.59783935546875, "n_util_obs": 912, "util_count_spearman": 0.12903503824329332}, "probe_overhead_sec_total": 225.532, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.013225, "alpha": 0.03, "t": 1092.3} |
| {"event": "consolidate", "step": 3850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1106.1} |
| {"event": "train", "step": 3850, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.54682, "bpb": 2.23159, "chunk_loss": 0.030575, "F": 0.19812, "G": 0.20785, "chunks": 202.88, "bytes_per_chunk": 5.047, "boundary_entropy": 0.10434, "router_noise": 0.0, "grad_norm": 0.7188, "lr": 0.0001142, "router_trainable": true, "sec_per_step": 0.29, "probe_sec": 0.0523, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03268, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.10197, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.22023, "trunc_abs_frac": 0.5708, "mean_delta_full": 0.3121, "read_hit_rate": {"surface": 0.67121}, "mean_gate": {"surface": 0.20802}, "n_reads": 4336, "n_cand": 6460, "read_suppressed_frac": 0.1012, "dropout_reads_logged": 4336, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 270.02105712890625, "n_util_obs": 926, "util_count_spearman": 0.12522874045026536}, "probe_overhead_sec_total": 228.204, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.01771, "alpha": 0.03, "t": 1106.8} |
| {"event": "consolidate", "step": 3900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1126.9} |
| {"event": "train", "step": 3900, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.43754, "bpb": 2.07394, "chunk_loss": 0.030356, "F": 0.19604, "G": 0.20424, "chunks": 200.75, "bytes_per_chunk": 5.101, "boundary_entropy": 0.10339, "router_noise": 0.0, "grad_norm": 0.772, "lr": 0.0001108, "router_trainable": true, "sec_per_step": 0.409, "probe_sec": 0.0514, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0413, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.18166, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 1.09499, "trunc_abs_frac": 1.54514, "mean_delta_full": 0.24456, "read_hit_rate": {"surface": 0.68508}, "mean_gate": {"surface": 0.19654}, "n_reads": 4379, "n_cand": 6392, "read_suppressed_frac": 0.0977, "dropout_reads_logged": 4379, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 273.41192626953125, "n_util_obs": 937, "util_count_spearman": 0.12141133364406353}, "probe_overhead_sec_total": 231.4, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.023869, "alpha": 0.03, "t": 1127.2} |
| {"event": "consolidate", "step": 3950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1140.5} |
| {"event": "train", "step": 3950, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.48894, "bpb": 2.14809, "chunk_loss": 0.030317, "F": 0.1926, "G": 0.20108, "chunks": 197.22, "bytes_per_chunk": 5.192, "boundary_entropy": 0.1002, "router_noise": 0.0, "grad_norm": 0.7703, "lr": 0.0001074, "router_trainable": true, "sec_per_step": 0.268, "probe_sec": 0.0236, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01824, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.05881, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": -0.08634, "trunc_abs_frac": 0.47287, "mean_delta_full": 0.19393, "read_hit_rate": {"surface": 0.67081}, "mean_gate": {"surface": 0.19629}, "n_reads": 4212, "n_cand": 6279, "read_suppressed_frac": 0.1038, "dropout_reads_logged": 4212, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 276.8309326171875, "n_util_obs": 947, "util_count_spearman": 0.1105290788082932}, "probe_overhead_sec_total": 234.444, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.015722, "alpha": 0.03, "t": 1140.6} |
| {"event": "consolidate", "step": 4000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1154.7} |
| {"event": "probe", "step": 4000, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3165.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.181865, "cps": 24.726562, "mean_p": 0.188944, "boundary_entropy": 0.094581, "bpb": 1.939558}, "tr": {"chunks": 3526.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.194774, "cps": 27.546875, "mean_p": 0.196679, "boundary_entropy": 0.133509, "bpb": 2.09056}, "fi": {"chunks": 4399.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.233853, "cps": 34.367188, "mean_p": 0.233282, "boundary_entropy": 0.103913, "bpb": 2.188089}}, "bpb": {"en": 1.93956, "tr": 2.09056, "fi": 2.18809}, "gini_cpc": 0.056771, "gini_cps": 0.074181, "bpic": 5.1156, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0709819770125295, "fi": 1.285856440254656}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 280.18035888671875, "n_util_obs": 956, "util_count_spearman": 0.10780263741686072}, "t": 1155.3} |
| {"event": "eval", "step": 4000, "phase": 2, "bpb": {"en": 1.95213, "tr": 2.10014, "fi": 2.19346}, "bytes_per_chunk": {"en": 5.135, "tr": 5.727, "fi": 4.517}, "bpb_mean": 2.08191, "frozen_inventory": true, "t": 1156.5} |
| {"event": "train", "step": 4000, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.41005, "bpb": 2.03427, "chunk_loss": 0.030388, "F": 0.19559, "G": 0.20384, "chunks": 200.28, "bytes_per_chunk": 5.113, "boundary_entropy": 0.10019, "router_noise": 0.0, "grad_norm": 0.6539, "lr": 0.000104, "router_trainable": true, "sec_per_step": 0.339, "probe_sec": 0.0567, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04844, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.23456, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": -1.02428, "trunc_abs_frac": 0.82526, "mean_delta_full": -0.09898, "read_hit_rate": {"surface": 0.69515}, "mean_gate": {"surface": 0.20433}, "n_reads": 4433, "n_cand": 6377, "read_suppressed_frac": 0.0999, "dropout_reads_logged": 4433, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 280.248291015625, "n_util_obs": 956, "util_count_spearman": 0.10779879644519419}, "probe_overhead_sec_total": 237.743, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.017922, "alpha": 0.03, "t": 1157.6} |
| {"event": "consolidate", "step": 4050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1169.1} |
| {"event": "train", "step": 4050, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.42239, "bpb": 2.05207, "chunk_loss": 0.030417, "F": 0.20013, "G": 0.20834, "chunks": 204.94, "bytes_per_chunk": 4.997, "boundary_entropy": 0.10157, "router_noise": 0.0, "grad_norm": 0.6929, "lr": 0.0001007, "router_trainable": true, "sec_per_step": 0.235, "probe_sec": 0.0355, "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": 14, "mean_delta_loo": 0.08352, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -4.6501, "trunc_abs_frac": 1.48212, "mean_delta_full": 0.03951, "read_hit_rate": {"surface": 0.69246}, "mean_gate": {"surface": 0.20785}, "n_reads": 4519, "n_cand": 6526, "read_suppressed_frac": 0.1087, "dropout_reads_logged": 4519, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 283.6505126953125, "n_util_obs": 969, "util_count_spearman": 0.11227891557626843}, "probe_overhead_sec_total": 240.925, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.016059, "alpha": 0.03, "t": 1169.3} |
| {"event": "probe", "step": 4096, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3177.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.182555, "cps": 24.820312, "mean_p": 0.188839, "boundary_entropy": 0.091962, "bpb": 1.934371}, "tr": {"chunks": 3544.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.195769, "cps": 27.6875, "mean_p": 0.196193, "boundary_entropy": 0.133859, "bpb": 2.087667}, "fi": {"chunks": 4396.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.233693, "cps": 34.34375, "mean_p": 0.232748, "boundary_entropy": 0.103501, "bpb": 2.186596}}, "bpb": {"en": 1.93437, "tr": 2.08767, "fi": 2.1866}, "gini_cpc": 0.055705, "gini_cps": 0.073101, "bpic": 5.103175, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0723833616650749, "fi": 1.2801259625029031}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 286.71875, "n_util_obs": 978, "util_count_spearman": 0.10955368328060545}, "t": 1181.6} |
| {"event": "consolidate", "step": 4100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1183.5} |
| {"event": "train", "step": 4100, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.36411, "bpb": 1.96799, "chunk_loss": 0.030304, "F": 0.19461, "G": 0.20288, "chunks": 199.28, "bytes_per_chunk": 5.138, "boundary_entropy": 0.09844, "router_noise": 0.0, "grad_norm": 0.7188, "lr": 9.75e-05, "router_trainable": true, "sec_per_step": 0.29, "probe_sec": 0.0365, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02731, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.14702, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 1.22268, "trunc_abs_frac": 30.73621, "mean_delta_full": 0.26697, "read_hit_rate": {"surface": 0.70701}, "mean_gate": {"surface": 0.2046}, "n_reads": 4486, "n_cand": 6345, "read_suppressed_frac": 0.1072, "dropout_reads_logged": 4486, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 287.05487060546875, "n_util_obs": 980, "util_count_spearman": 0.11328342715373194}, "probe_overhead_sec_total": 244.077, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.014656, "alpha": 0.03, "t": 1183.8} |
| {"event": "consolidate", "step": 4150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1195.5} |
| {"event": "train", "step": 4150, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.42396, "bpb": 2.05434, "chunk_loss": 0.030512, "F": 0.20499, "G": 0.21272, "chunks": 209.91, "bytes_per_chunk": 4.878, "boundary_entropy": 0.09916, "router_noise": 0.0, "grad_norm": 0.7496, "lr": 9.43e-05, "router_trainable": true, "sec_per_step": 0.237, "probe_sec": 0.0784, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06931, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.00739, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.17276, "trunc_abs_frac": 0.17276, "mean_delta_full": 0.78149, "read_hit_rate": {"surface": 0.70471}, "mean_gate": {"surface": 0.1987}, "n_reads": 4711, "n_cand": 6685, "read_suppressed_frac": 0.0955, "dropout_reads_logged": 4711, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 290.46514892578125, "n_util_obs": 992, "util_count_spearman": 0.1104135531472742}, "probe_overhead_sec_total": 248.035, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.020069, "alpha": 0.03, "t": 1195.7} |
| {"event": "consolidate", "step": 4200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1207.6} |
| {"event": "probe", "step": 4200, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3182.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.182842, "cps": 24.859375, "mean_p": 0.189364, "boundary_entropy": 0.088915, "bpb": 1.931904}, "tr": {"chunks": 3520.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.194443, "cps": 27.5, "mean_p": 0.193896, "boundary_entropy": 0.127117, "bpb": 2.074578}, "fi": {"chunks": 4380.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.232842, "cps": 34.21875, "mean_p": 0.23212, "boundary_entropy": 0.098199, "bpb": 2.180019}}, "bpb": {"en": 1.9319, "tr": 2.07458, "fi": 2.18002}, "gini_cpc": 0.054634, "gini_cps": 0.072069, "bpic": 5.119293, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0634475056535386, "fi": 1.2734625330051312}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 293.81805419921875, "n_util_obs": 1000, "util_count_spearman": 0.10245393259257626}, "t": 1208.2} |
| {"event": "eval", "step": 4200, "phase": 2, "bpb": {"en": 1.94395, "tr": 2.08604, "fi": 2.18648}, "bytes_per_chunk": {"en": 5.124, "tr": 5.688, "fi": 4.516}, "bpb_mean": 2.07216, "frozen_inventory": true, "t": 1209.4} |
| {"event": "phase_transition", "step": 4200, "phase": 3, "objective": "global", "router_trainable": true, "encoder_trainable": true, "frozen_router_params": 0, "t": 1210.3} |
| {"event": "train", "step": 4200, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35347, "bpb": 1.95265, "chunk_loss": 0.030233, "F": 0.19086, "G": 0.19843, "chunks": 195.44, "bytes_per_chunk": 5.24, "boundary_entropy": 0.09539, "router_noise": 0.0, "grad_norm": 0.7208, "lr": 9.12e-05, "router_trainable": true, "sec_per_step": 0.297, "probe_sec": 0.0324, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02741, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.0457, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.27497, "trunc_abs_frac": 2.43633, "mean_delta_full": 0.39508, "read_hit_rate": {"surface": 0.69142}, "mean_gate": {"surface": 0.2014}, "n_reads": 4302, "n_cand": 6222, "read_suppressed_frac": 0.103, "dropout_reads_logged": 4302, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 293.8831787109375, "n_util_obs": 1000, "util_count_spearman": 0.10203997797865609}, "probe_overhead_sec_total": 251.407, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.018215, "alpha": 0.03, "t": 1210.6} |
| {"event": "consolidate", "step": 4250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1222.1} |
| {"event": "train", "step": 4250, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38491, "bpb": 1.99801, "chunk_loss": 0.030351, "F": 0.19714, "G": 0.20509, "chunks": 201.88, "bytes_per_chunk": 5.072, "boundary_entropy": 0.09625, "router_noise": 0.0, "grad_norm": 0.6474, "lr": 8.81e-05, "router_trainable": true, "sec_per_step": 0.235, "probe_sec": 0.0496, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04367, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.15167, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.72789, "trunc_abs_frac": 1.45117, "mean_delta_full": 0.27702, "read_hit_rate": {"surface": 0.69042}, "mean_gate": {"surface": 0.20793}, "n_reads": 4438, "n_cand": 6428, "read_suppressed_frac": 0.0994, "dropout_reads_logged": 4438, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 297.3067626953125, "n_util_obs": 1014, "util_count_spearman": 0.09356906455874399}, "probe_overhead_sec_total": 254.883, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.016593, "alpha": 0.03, "t": 1222.3} |
| {"event": "consolidate", "step": 4300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1234.6} |
| {"event": "train", "step": 4300, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38637, "bpb": 2.00011, "chunk_loss": 0.030247, "F": 0.19159, "G": 0.19939, "chunks": 196.19, "bytes_per_chunk": 5.219, "boundary_entropy": 0.09636, "router_noise": 0.0, "grad_norm": 0.7313, "lr": 8.5e-05, "router_trainable": true, "sec_per_step": 0.249, "probe_sec": 0.0237, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01831, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.05506, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -1.14058, "trunc_abs_frac": 1.72497, "mean_delta_full": 0.02677, "read_hit_rate": {"surface": 0.68956}, "mean_gate": {"surface": 0.20225}, "n_reads": 4307, "n_cand": 6246, "read_suppressed_frac": 0.0987, "dropout_reads_logged": 4307, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 300.74658203125, "n_util_obs": 1023, "util_count_spearman": 0.09576700125640833}, "probe_overhead_sec_total": 258.273, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.018518, "alpha": 0.03, "t": 1234.8} |
| {"event": "consolidate", "step": 4350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1247.0} |
| {"event": "train", "step": 4350, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.40904, "bpb": 2.03282, "chunk_loss": 0.030323, "F": 0.19693, "G": 0.20398, "chunks": 201.66, "bytes_per_chunk": 5.078, "boundary_entropy": 0.09187, "router_noise": 0.0, "grad_norm": 0.6746, "lr": 8.21e-05, "router_trainable": true, "sec_per_step": 0.252, "probe_sec": 0.0791, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06134, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.06461, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 1.91297, "trunc_abs_frac": 0.65128, "mean_delta_full": -0.17078, "read_hit_rate": {"surface": 0.6851}, "mean_gate": {"surface": 0.19964}, "n_reads": 4399, "n_cand": 6421, "read_suppressed_frac": 0.09, "dropout_reads_logged": 4399, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 304.12969970703125, "n_util_obs": 1032, "util_count_spearman": 0.09440397367605437}, "probe_overhead_sec_total": 261.719, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.015212, "alpha": 0.03, "t": 1247.4} |
| {"event": "consolidate", "step": 4400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1259.0} |
| {"event": "probe", "step": 4400, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3177.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.182555, "cps": 24.820312, "mean_p": 0.188901, "boundary_entropy": 0.087589, "bpb": 1.920838}, "tr": {"chunks": 3498.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.193228, "cps": 27.328125, "mean_p": 0.193422, "boundary_entropy": 0.126374, "bpb": 2.064365}, "fi": {"chunks": 4367.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.232151, "cps": 34.117188, "mean_p": 0.23177, "boundary_entropy": 0.101301, "bpb": 2.162666}}, "bpb": {"en": 1.92084, "tr": 2.06436, "fi": 2.16267}, "gini_cpc": 0.054388, "gini_cps": 0.071847, "bpic": 5.137837, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0584641645328532, "fi": 1.2716810915036805}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 307.4659423828125, "n_util_obs": 1040, "util_count_spearman": 0.09853152221820463}, "t": 1259.8} |
| {"event": "eval", "step": 4400, "phase": 3, "bpb": {"en": 1.93311, "tr": 2.07405, "fi": 2.17}, "bytes_per_chunk": {"en": 5.154, "tr": 5.761, "fi": 4.529}, "bpb_mean": 2.05905, "frozen_inventory": true, "t": 1260.7} |
| {"event": "train", "step": 4400, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.3585, "bpb": 1.9599, "chunk_loss": 0.030238, "F": 0.19266, "G": 0.19936, "chunks": 197.28, "bytes_per_chunk": 5.191, "boundary_entropy": 0.09092, "router_noise": 0.0, "grad_norm": 0.6421, "lr": 7.92e-05, "router_trainable": true, "sec_per_step": 0.295, "probe_sec": 0.0462, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04113, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.05576, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 0.21709, "trunc_abs_frac": 0.70435, "mean_delta_full": 0.21971, "read_hit_rate": {"surface": 0.69193}, "mean_gate": {"surface": 0.19651}, "n_reads": 4346, "n_cand": 6281, "read_suppressed_frac": 0.1015, "dropout_reads_logged": 4346, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 307.53155517578125, "n_util_obs": 1040, "util_count_spearman": 0.09847993647161288}, "probe_overhead_sec_total": 265.666, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.017057, "alpha": 0.03, "t": 1262.1} |
| {"event": "consolidate", "step": 4450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1273.9} |
| {"event": "train", "step": 4450, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.3539, "bpb": 1.95326, "chunk_loss": 0.030195, "F": 0.1875, "G": 0.19468, "chunks": 192.0, "bytes_per_chunk": 5.333, "boundary_entropy": 0.09315, "router_noise": 0.0, "grad_norm": 0.6911, "lr": 7.63e-05, "router_trainable": true, "sec_per_step": 0.238, "probe_sec": 0.0239, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0183, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.0004, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -15.51696, "trunc_abs_frac": 0.66707, "mean_delta_full": 0.00285, "read_hit_rate": {"surface": 0.68745}, "mean_gate": {"surface": 0.19952}, "n_reads": 4201, "n_cand": 6111, "read_suppressed_frac": 0.0988, "dropout_reads_logged": 4201, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 310.93212890625, "n_util_obs": 1052, "util_count_spearman": 0.09656492057655024}, "probe_overhead_sec_total": 269.104, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.017008, "alpha": 0.03, "t": 1274.0} |
| {"event": "consolidate", "step": 4500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1285.2} |
| {"event": "train", "step": 4500, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.3374, "bpb": 1.92947, "chunk_loss": 0.030324, "F": 0.19757, "G": 0.20394, "chunks": 202.31, "bytes_per_chunk": 5.061, "boundary_entropy": 0.0895, "router_noise": 0.0, "grad_norm": 0.6669, "lr": 7.36e-05, "router_trainable": true, "sec_per_step": 0.227, "probe_sec": 0.03, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01865, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.04415, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.14478, "trunc_abs_frac": 0.53253, "mean_delta_full": 0.30511, "read_hit_rate": {"surface": 0.70289}, "mean_gate": {"surface": 0.18961}, "n_reads": 4528, "n_cand": 6442, "read_suppressed_frac": 0.0908, "dropout_reads_logged": 4528, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 314.3165283203125, "n_util_obs": 1060, "util_count_spearman": 0.09741985442624011}, "probe_overhead_sec_total": 271.971, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.015719, "alpha": 0.03, "t": 1285.4} |
| {"event": "consolidate", "step": 4550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1297.3} |
| {"event": "train", "step": 4550, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35854, "bpb": 1.95995, "chunk_loss": 0.030273, "F": 0.19415, "G": 0.20028, "chunks": 198.81, "bytes_per_chunk": 5.151, "boundary_entropy": 0.093, "router_noise": 0.0, "grad_norm": 0.6592, "lr": 7.09e-05, "router_trainable": true, "sec_per_step": 0.242, "probe_sec": 0.0401, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03507, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.02593, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.49616, "trunc_abs_frac": 0.47496, "mean_delta_full": 0.29431, "read_hit_rate": {"surface": 0.68073}, "mean_gate": {"surface": 0.19936}, "n_reads": 4309, "n_cand": 6330, "read_suppressed_frac": 0.1016, "dropout_reads_logged": 4309, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 317.7494812011719, "n_util_obs": 1074, "util_count_spearman": 0.08594886473710785}, "probe_overhead_sec_total": 275.702, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.014519, "alpha": 0.03, "t": 1297.5} |
| {"event": "consolidate", "step": 4600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1309.4} |
| {"event": "probe", "step": 4600, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3186.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.183072, "cps": 24.890625, "mean_p": 0.189112, "boundary_entropy": 0.086362, "bpb": 1.91138}, "tr": {"chunks": 3519.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.194388, "cps": 27.492188, "mean_p": 0.194619, "boundary_entropy": 0.124228, "bpb": 2.052798}, "fi": {"chunks": 4378.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.232736, "cps": 34.203125, "mean_p": 0.231729, "boundary_entropy": 0.095165, "bpb": 2.151471}}, "bpb": {"en": 1.91138, "tr": 2.0528, "fi": 2.15147}, "gini_cpc": 0.05426, "gini_cps": 0.071701, "bpic": 5.118831, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0618106185228218, "fi": 1.2712829503032494}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 321.0794677734375, "n_util_obs": 1088, "util_count_spearman": 0.080880714311045}, "t": 1310.1} |
| {"event": "eval", "step": 4600, "phase": 3, "bpb": {"en": 1.92355, "tr": 2.06135, "fi": 2.15702}, "bytes_per_chunk": {"en": 5.124, "tr": 5.726, "fi": 4.522}, "bpb_mean": 2.04731, "frozen_inventory": true, "t": 1311.0} |
| {"event": "train", "step": 4600, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.36208, "bpb": 1.96506, "chunk_loss": 0.030391, "F": 0.20233, "G": 0.2084, "chunks": 207.19, "bytes_per_chunk": 4.942, "boundary_entropy": 0.09236, "router_noise": 0.0, "grad_norm": 0.6668, "lr": 6.82e-05, "router_trainable": true, "sec_per_step": 0.298, "probe_sec": 0.1769, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.16935, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.08568, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.09097, "trunc_abs_frac": 0.28, "mean_delta_full": 0.12891, "read_hit_rate": {"surface": 0.70142}, "mean_gate": {"surface": 0.19396}, "n_reads": 4628, "n_cand": 6598, "read_suppressed_frac": 0.0994, "dropout_reads_logged": 4628, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 321.1499938964844, "n_util_obs": 1088, "util_count_spearman": 0.0809950340256853}, "probe_overhead_sec_total": 279.211, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.01956, "alpha": 0.03, "t": 1312.4} |
| {"event": "consolidate", "step": 4650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1323.7} |
| {"event": "train", "step": 4650, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.39358, "bpb": 2.0105, "chunk_loss": 0.030321, "F": 0.19504, "G": 0.20354, "chunks": 199.72, "bytes_per_chunk": 5.127, "boundary_entropy": 0.09474, "router_noise": 0.0, "grad_norm": 0.6747, "lr": 6.57e-05, "router_trainable": true, "sec_per_step": 0.233, "probe_sec": 0.041, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03362, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.14654, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.3867, "trunc_abs_frac": 1.0014, "mean_delta_full": 0.09732, "read_hit_rate": {"surface": 0.68454}, "mean_gate": {"surface": 0.19648}, "n_reads": 4353, "n_cand": 6359, "read_suppressed_frac": 0.1011, "dropout_reads_logged": 4353, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 324.5137023925781, "n_util_obs": 1099, "util_count_spearman": 0.07713397425928677}, "probe_overhead_sec_total": 282.451, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.01681, "alpha": 0.03, "t": 1324.0} |
| {"event": "consolidate", "step": 4700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1336.0} |
| {"event": "train", "step": 4700, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.4234, "bpb": 2.05353, "chunk_loss": 0.030402, "F": 0.20129, "G": 0.20669, "chunks": 206.12, "bytes_per_chunk": 4.968, "boundary_entropy": 0.08826, "router_noise": 0.0, "grad_norm": 0.6645, "lr": 6.32e-05, "router_trainable": true, "sec_per_step": 0.243, "probe_sec": 0.0362, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.028, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.08895, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 0.54367, "trunc_abs_frac": 7.30552, "mean_delta_full": 0.12486, "read_hit_rate": {"surface": 0.69805}, "mean_gate": {"surface": 0.19108}, "n_reads": 4582, "n_cand": 6564, "read_suppressed_frac": 0.0895, "dropout_reads_logged": 4582, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 327.9302978515625, "n_util_obs": 1115, "util_count_spearman": 0.08279225957135761}, "probe_overhead_sec_total": 285.849, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.01711, "alpha": 0.03, "t": 1336.2} |
| {"event": "consolidate", "step": 4750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1347.6} |
| {"event": "train", "step": 4750, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.40865, "bpb": 2.03225, "chunk_loss": 0.030359, "F": 0.19968, "G": 0.20601, "chunks": 204.47, "bytes_per_chunk": 5.008, "boundary_entropy": 0.09045, "router_noise": 0.0, "grad_norm": 0.6595, "lr": 6.08e-05, "router_trainable": true, "sec_per_step": 0.233, "probe_sec": 0.0441, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03541, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.04509, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.55094, "trunc_abs_frac": 0.89557, "mean_delta_full": -0.33297, "read_hit_rate": {"surface": 0.69544}, "mean_gate": {"surface": 0.19025}, "n_reads": 4528, "n_cand": 6511, "read_suppressed_frac": 0.1049, "dropout_reads_logged": 4528, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 331.302978515625, "n_util_obs": 1131, "util_count_spearman": 0.0756854851465925}, "probe_overhead_sec_total": 289.029, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.017073, "alpha": 0.03, "t": 1347.8} |
| {"event": "consolidate", "step": 4800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1360.0} |
| {"event": "probe", "step": 4800, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3167.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.18198, "cps": 24.742188, "mean_p": 0.18777, "boundary_entropy": 0.084507, "bpb": 1.903077}, "tr": {"chunks": 3480.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.192233, "cps": 27.1875, "mean_p": 0.192931, "boundary_entropy": 0.122968, "bpb": 2.040877}, "fi": {"chunks": 4358.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.231673, "cps": 34.046875, "mean_p": 0.230232, "boundary_entropy": 0.090059, "bpb": 2.137966}}, "bpb": {"en": 1.90308, "tr": 2.04088, "fi": 2.13797}, "gini_cpc": 0.054678, "gini_cps": 0.072149, "bpic": 5.155111, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0563424906711676, "fi": 1.273067406411243}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 334.6233825683594, "n_util_obs": 1140, "util_count_spearman": 0.08388266665051117}, "t": 1360.6} |
| {"event": "eval", "step": 4800, "phase": 3, "bpb": {"en": 1.91608, "tr": 2.05052, "fi": 2.14247}, "bytes_per_chunk": {"en": 5.15, "tr": 5.79, "fi": 4.539}, "bpb_mean": 2.03636, "frozen_inventory": true, "t": 1361.8} |
| {"event": "train", "step": 4800, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.37878, "bpb": 1.98916, "chunk_loss": 0.03033, "F": 0.19907, "G": 0.20468, "chunks": 203.84, "bytes_per_chunk": 5.023, "boundary_entropy": 0.08694, "router_noise": 0.0, "grad_norm": 0.6468, "lr": 5.85e-05, "router_trainable": true, "sec_per_step": 0.304, "probe_sec": 0.0805, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07395, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.03409, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.95605, "trunc_abs_frac": 1.71965, "mean_delta_full": 0.20889, "read_hit_rate": {"surface": 0.69928}, "mean_gate": {"surface": 0.18665}, "n_reads": 4539, "n_cand": 6491, "read_suppressed_frac": 0.1051, "dropout_reads_logged": 4539, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 334.6918640136719, "n_util_obs": 1140, "util_count_spearman": 0.08290193198285663}, "probe_overhead_sec_total": 292.623, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.016035, "alpha": 0.03, "t": 1363.0} |
| {"event": "consolidate", "step": 4850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1374.8} |
| {"event": "train", "step": 4850, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35316, "bpb": 1.95219, "chunk_loss": 0.030308, "F": 0.19373, "G": 0.20138, "chunks": 198.38, "bytes_per_chunk": 5.162, "boundary_entropy": 0.0907, "router_noise": 0.0, "grad_norm": 0.6793, "lr": 5.62e-05, "router_trainable": true, "sec_per_step": 0.241, "probe_sec": 0.1451, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.13683, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 7e-05, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": -0.13368, "trunc_abs_frac": 0.76733, "mean_delta_full": 0.21948, "read_hit_rate": {"surface": 0.6911}, "mean_gate": {"surface": 0.19404}, "n_reads": 4365, "n_cand": 6316, "read_suppressed_frac": 0.104, "dropout_reads_logged": 4365, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 338.11553955078125, "n_util_obs": 1147, "util_count_spearman": 0.08848031616050062}, "probe_overhead_sec_total": 295.986, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.015817, "alpha": 0.03, "t": 1375.1} |
| {"event": "consolidate", "step": 4900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1386.5} |
| {"event": "train", "step": 4900, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38579, "bpb": 1.99928, "chunk_loss": 0.030323, "F": 0.19022, "G": 0.19815, "chunks": 194.78, "bytes_per_chunk": 5.257, "boundary_entropy": 0.08816, "router_noise": 0.0, "grad_norm": 0.6807, "lr": 5.41e-05, "router_trainable": true, "sec_per_step": 0.233, "probe_sec": 0.0647, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05845, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.07104, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.63995, "trunc_abs_frac": 1.34162, "mean_delta_full": 0.34941, "read_hit_rate": {"surface": 0.67086}, "mean_gate": {"surface": 0.19077}, "n_reads": 4160, "n_cand": 6201, "read_suppressed_frac": 0.095, "dropout_reads_logged": 4160, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 341.5111083984375, "n_util_obs": 1154, "util_count_spearman": 0.09606148002186367}, "probe_overhead_sec_total": 299.002, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.018893, "alpha": 0.03, "t": 1386.7} |
| {"event": "consolidate", "step": 4950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1398.4} |
| {"event": "train", "step": 4950, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.33855, "bpb": 1.93113, "chunk_loss": 0.030299, "F": 0.19345, "G": 0.20076, "chunks": 198.09, "bytes_per_chunk": 5.169, "boundary_entropy": 0.09121, "router_noise": 0.0, "grad_norm": 0.6715, "lr": 5.2e-05, "router_trainable": true, "sec_per_step": 0.237, "probe_sec": 0.0239, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01816, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.18967, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -0.93965, "trunc_abs_frac": 0.78217, "mean_delta_full": 0.0709, "read_hit_rate": {"surface": 0.69558}, "mean_gate": {"surface": 0.19154}, "n_reads": 4387, "n_cand": 6307, "read_suppressed_frac": 0.0966, "dropout_reads_logged": 4387, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 344.8926696777344, "n_util_obs": 1162, "util_count_spearman": 0.0990629671527858}, "probe_overhead_sec_total": 302.361, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.018963, "alpha": 0.03, "t": 1398.6} |
| {"event": "consolidate", "step": 5000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1410.9} |
| {"event": "probe", "step": 5000, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3212.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.184566, "cps": 25.09375, "mean_p": 0.189616, "boundary_entropy": 0.080572, "bpb": 1.895778}, "tr": {"chunks": 3524.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.194664, "cps": 27.53125, "mean_p": 0.193309, "boundary_entropy": 0.117575, "bpb": 2.030009}, "fi": {"chunks": 4398.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.233799, "cps": 34.359375, "mean_p": 0.231942, "boundary_entropy": 0.089374, "bpb": 2.125291}}, "bpb": {"en": 1.89578, "tr": 2.03001, "fi": 2.12529}, "gini_cpc": 0.053541, "gini_cps": 0.071014, "bpic": 5.095384, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0547121084971844, "fi": 1.266752952437079}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 348.2156982421875, "n_util_obs": 1173, "util_count_spearman": 0.1096549789367717}, "t": 1411.5} |
| {"event": "eval", "step": 5000, "phase": 3, "bpb": {"en": 1.90907, "tr": 2.04046, "fi": 2.13075}, "bytes_per_chunk": {"en": 5.109, "tr": 5.716, "fi": 4.513}, "bpb_mean": 2.02676, "frozen_inventory": true, "t": 1412.6} |
| {"event": "train", "step": 5000, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.33134, "bpb": 1.92071, "chunk_loss": 0.030332, "F": 0.19788, "G": 0.20282, "chunks": 202.62, "bytes_per_chunk": 5.054, "boundary_entropy": 0.08229, "router_noise": 0.0, "grad_norm": 0.6209, "lr": 5e-05, "router_trainable": true, "sec_per_step": 0.304, "probe_sec": 0.0566, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04843, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.03398, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.09539, "trunc_abs_frac": 2.99761, "mean_delta_full": 0.28824, "read_hit_rate": {"surface": 0.70691}, "mean_gate": {"surface": 0.182}, "n_reads": 4561, "n_cand": 6452, "read_suppressed_frac": 0.1026, "dropout_reads_logged": 4561, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 348.2850341796875, "n_util_obs": 1173, "util_count_spearman": 0.11021665090367014}, "probe_overhead_sec_total": 306.355, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.018718, "alpha": 0.03, "t": 1413.8} |
| {"event": "consolidate", "step": 5050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1425.8} |
| {"event": "train", "step": 5050, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.34616, "bpb": 1.9421, "chunk_loss": 0.030368, "F": 0.19974, "G": 0.20513, "chunks": 204.53, "bytes_per_chunk": 5.007, "boundary_entropy": 0.08505, "router_noise": 0.0, "grad_norm": 0.7024, "lr": 4.81e-05, "router_trainable": true, "sec_per_step": 0.244, "probe_sec": 0.0448, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03689, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.0683, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 1.61364, "trunc_abs_frac": 1.33264, "mean_delta_full": -0.22235, "read_hit_rate": {"surface": 0.69446}, "mean_gate": {"surface": 0.1875}, "n_reads": 4523, "n_cand": 6513, "read_suppressed_frac": 0.0971, "dropout_reads_logged": 4523, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 351.67987060546875, "n_util_obs": 1179, "util_count_spearman": 0.11718230003910904}, "probe_overhead_sec_total": 309.272, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.015639, "alpha": 0.03, "t": 1426.0} |
| {"event": "consolidate", "step": 5100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1438.2} |
| {"event": "train", "step": 5100, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38103, "bpb": 1.99241, "chunk_loss": 0.030341, "F": 0.19681, "G": 0.20375, "chunks": 201.53, "bytes_per_chunk": 5.081, "boundary_entropy": 0.09054, "router_noise": 0.0, "grad_norm": 0.7143, "lr": 4.63e-05, "router_trainable": true, "sec_per_step": 0.248, "probe_sec": 0.0743, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06615, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.0066, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.10806, "trunc_abs_frac": 0.74832, "mean_delta_full": -0.27747, "read_hit_rate": {"surface": 0.68973}, "mean_gate": {"surface": 0.18454}, "n_reads": 4426, "n_cand": 6417, "read_suppressed_frac": 0.1055, "dropout_reads_logged": 4426, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 355.09930419921875, "n_util_obs": 1189, "util_count_spearman": 0.10781335804100328}, "probe_overhead_sec_total": 312.685, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.021508, "alpha": 0.03, "t": 1438.4} |
| {"event": "consolidate", "step": 5150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1450.6} |
| {"event": "train", "step": 5150, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.37256, "bpb": 1.98019, "chunk_loss": 0.030394, "F": 0.2023, "G": 0.20717, "chunks": 207.16, "bytes_per_chunk": 4.943, "boundary_entropy": 0.08771, "router_noise": 0.0, "grad_norm": 0.7026, "lr": 4.45e-05, "router_trainable": true, "sec_per_step": 0.249, "probe_sec": 0.0802, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0755, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.12594, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": -0.16012, "trunc_abs_frac": 0.16012, "mean_delta_full": -0.31714, "read_hit_rate": {"surface": 0.70653}, "mean_gate": {"surface": 0.17917}, "n_reads": 4661, "n_cand": 6597, "read_suppressed_frac": 0.0959, "dropout_reads_logged": 4661, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 358.5105895996094, "n_util_obs": 1197, "util_count_spearman": 0.11116268584744798}, "probe_overhead_sec_total": 316.951, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.016169, "alpha": 0.03, "t": 1450.9} |
| {"event": "consolidate", "step": 5200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1463.1} |
| {"event": "probe", "step": 5200, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3190.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.183302, "cps": 24.921875, "mean_p": 0.188729, "boundary_entropy": 0.081044, "bpb": 1.889973}, "tr": {"chunks": 3529.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.19494, "cps": 27.570312, "mean_p": 0.193808, "boundary_entropy": 0.118455, "bpb": 2.02513}, "fi": {"chunks": 4393.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.233534, "cps": 34.320312, "mean_p": 0.230973, "boundary_entropy": 0.087183, "bpb": 2.116455}}, "bpb": {"en": 1.88997, "tr": 2.02513, "fi": 2.11645}, "gini_cpc": 0.054739, "gini_cps": 0.072174, "bpic": 5.105472, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0634927756652677, "fi": 1.2740391010462264}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 361.82745361328125, "n_util_obs": 1208, "util_count_spearman": 0.11576735291855698}, "t": 1463.8} |
| {"event": "eval", "step": 5200, "phase": 3, "bpb": {"en": 1.90359, "tr": 2.03433, "fi": 2.12217}, "bytes_per_chunk": {"en": 5.112, "tr": 5.704, "fi": 4.525}, "bpb_mean": 2.02003, "frozen_inventory": true, "t": 1464.7} |
| {"event": "train", "step": 5200, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32177, "bpb": 1.9069, "chunk_loss": 0.03029, "F": 0.19598, "G": 0.20121, "chunks": 200.69, "bytes_per_chunk": 5.102, "boundary_entropy": 0.08488, "router_noise": 0.0, "grad_norm": 0.6564, "lr": 4.29e-05, "router_trainable": true, "sec_per_step": 0.301, "probe_sec": 0.0254, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01885, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.01786, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 2.69242, "trunc_abs_frac": 0.49974, "mean_delta_full": -0.05817, "read_hit_rate": {"surface": 0.70047}, "mean_gate": {"surface": 0.18156}, "n_reads": 4476, "n_cand": 6390, "read_suppressed_frac": 0.1052, "dropout_reads_logged": 4476, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 361.8963623046875, "n_util_obs": 1208, "util_count_spearman": 0.11566372901641124}, "probe_overhead_sec_total": 320.811, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.018521, "alpha": 0.03, "t": 1465.9} |
| {"event": "consolidate", "step": 5250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1477.4} |
| {"event": "train", "step": 5250, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32563, "bpb": 1.91248, "chunk_loss": 0.030316, "F": 0.19577, "G": 0.20225, "chunks": 200.47, "bytes_per_chunk": 5.108, "boundary_entropy": 0.085, "router_noise": 0.0, "grad_norm": 0.66, "lr": 4.14e-05, "router_trainable": true, "sec_per_step": 0.236, "probe_sec": 0.053, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04596, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.19275, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.90612, "trunc_abs_frac": 0.84068, "mean_delta_full": -0.1842, "read_hit_rate": {"surface": 0.70014}, "mean_gate": {"surface": 0.18078}, "n_reads": 4469, "n_cand": 6383, "read_suppressed_frac": 0.1047, "dropout_reads_logged": 4469, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 365.289794921875, "n_util_obs": 1212, "util_count_spearman": 0.11546960605862505}, "probe_overhead_sec_total": 324.619, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.01721, "alpha": 0.03, "t": 1477.7} |
| {"event": "consolidate", "step": 5300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1494.2} |
| {"event": "train", "step": 5300, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.37877, "bpb": 1.98914, "chunk_loss": 0.030347, "F": 0.19763, "G": 0.20436, "chunks": 202.38, "bytes_per_chunk": 5.06, "boundary_entropy": 0.08291, "router_noise": 0.0, "grad_norm": 0.7071, "lr": 3.99e-05, "router_trainable": true, "sec_per_step": 0.335, "probe_sec": 0.0374, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03158, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.00679, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 1.16761, "trunc_abs_frac": 1.16761, "mean_delta_full": 0.21558, "read_hit_rate": {"surface": 0.6977}, "mean_gate": {"surface": 0.18407}, "n_reads": 4496, "n_cand": 6444, "read_suppressed_frac": 0.0981, "dropout_reads_logged": 4496, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 368.69122314453125, "n_util_obs": 1220, "util_count_spearman": 0.11463099125194058}, "probe_overhead_sec_total": 327.598, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.018028, "alpha": 0.03, "t": 1494.4} |
| {"event": "consolidate", "step": 5350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1506.1} |
| {"event": "train", "step": 5350, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.31225, "bpb": 1.89318, "chunk_loss": 0.030226, "F": 0.19128, "G": 0.19787, "chunks": 195.88, "bytes_per_chunk": 5.228, "boundary_entropy": 0.08237, "router_noise": 0.0, "grad_norm": 0.6891, "lr": 3.86e-05, "router_trainable": true, "sec_per_step": 0.238, "probe_sec": 0.0843, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07572, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.09591, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 1.0, "trunc_abs_frac": 1.0, "mean_delta_full": 0.08435, "read_hit_rate": {"surface": 0.6966}, "mean_gate": {"surface": 0.17824}, "n_reads": 4344, "n_cand": 6236, "read_suppressed_frac": 0.1043, "dropout_reads_logged": 4344, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 372.10308837890625, "n_util_obs": 1229, "util_count_spearman": 0.11607407171793321}, "probe_overhead_sec_total": 330.583, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.017001, "alpha": 0.03, "t": 1506.3} |
| {"event": "consolidate", "step": 5400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1518.3} |
| {"event": "probe", "step": 5400, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3179.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.18267, "cps": 24.835938, "mean_p": 0.187883, "boundary_entropy": 0.0798, "bpb": 1.884922}, "tr": {"chunks": 3526.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.194774, "cps": 27.546875, "mean_p": 0.192684, "boundary_entropy": 0.117399, "bpb": 2.017462}, "fi": {"chunks": 4389.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.233321, "cps": 34.289062, "mean_p": 0.230554, "boundary_entropy": 0.086592, "bpb": 2.112034}}, "bpb": {"en": 1.88492, "tr": 2.01746, "fi": 2.11203}, "gini_cpc": 0.055287, "gini_cps": 0.072712, "bpic": 5.113755, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.066265478843868, "fi": 1.2772834638644583}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 375.4354248046875, "n_util_obs": 1239, "util_count_spearman": 0.12063018537232927}, "t": 1519.0} |
| {"event": "eval", "step": 5400, "phase": 3, "bpb": {"en": 1.89789, "tr": 2.02772, "fi": 2.11725}, "bytes_per_chunk": {"en": 5.127, "tr": 5.727, "fi": 4.532}, "bpb_mean": 2.01429, "frozen_inventory": true, "t": 1520.2} |
| {"event": "train", "step": 5400, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.37978, "bpb": 1.9906, "chunk_loss": 0.030524, "F": 0.20352, "G": 0.209, "chunks": 208.41, "bytes_per_chunk": 4.913, "boundary_entropy": 0.08414, "router_noise": 0.0, "grad_norm": 0.7319, "lr": 3.73e-05, "router_trainable": true, "sec_per_step": 0.301, "probe_sec": 0.0501, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0376, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.25088, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.89972, "trunc_abs_frac": 0.47113, "mean_delta_full": 0.42635, "read_hit_rate": {"surface": 0.70408}, "mean_gate": {"surface": 0.17602}, "n_reads": 4673, "n_cand": 6637, "read_suppressed_frac": 0.0929, "dropout_reads_logged": 4673, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 375.5056457519531, "n_util_obs": 1240, "util_count_spearman": 0.11885055288853208}, "probe_overhead_sec_total": 333.95, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.015263, "alpha": 0.03, "t": 1521.4} |
| {"event": "consolidate", "step": 5450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1535.6} |
| {"event": "train", "step": 5450, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32263, "bpb": 1.90815, "chunk_loss": 0.030468, "F": 0.20337, "G": 0.20881, "chunks": 208.25, "bytes_per_chunk": 4.917, "boundary_entropy": 0.08473, "router_noise": 0.0, "grad_norm": 0.7048, "lr": 3.62e-05, "router_trainable": true, "sec_per_step": 0.289, "probe_sec": 0.1111, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09199, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.0191, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -1.83027, "trunc_abs_frac": 2.36846, "mean_delta_full": -0.04443, "read_hit_rate": {"surface": 0.71653}, "mean_gate": {"surface": 0.1839}, "n_reads": 4752, "n_cand": 6632, "read_suppressed_frac": 0.0985, "dropout_reads_logged": 4752, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 378.9190368652344, "n_util_obs": 1257, "util_count_spearman": 0.11606635488713522}, "probe_overhead_sec_total": 336.996, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.018586, "alpha": 0.03, "t": 1535.9} |
| {"event": "consolidate", "step": 5500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1550.7} |
| {"event": "train", "step": 5500, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.30795, "bpb": 1.88698, "chunk_loss": 0.030285, "F": 0.19583, "G": 0.20171, "chunks": 200.53, "bytes_per_chunk": 5.106, "boundary_entropy": 0.08301, "router_noise": 0.0, "grad_norm": 0.6748, "lr": 3.51e-05, "router_trainable": true, "sec_per_step": 0.306, "probe_sec": 0.0373, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0228, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.02199, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.43783, "trunc_abs_frac": 1.29586, "mean_delta_full": -0.75635, "read_hit_rate": {"surface": 0.69569}, "mean_gate": {"surface": 0.1778}, "n_reads": 4442, "n_cand": 6385, "read_suppressed_frac": 0.0966, "dropout_reads_logged": 4442, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 382.34381103515625, "n_util_obs": 1269, "util_count_spearman": 0.11059804672438867}, "probe_overhead_sec_total": 339.901, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.017273, "alpha": 0.03, "t": 1551.2} |
| {"event": "consolidate", "step": 5550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1578.1} |
| {"event": "train", "step": 5550, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.36395, "bpb": 1.96776, "chunk_loss": 0.030235, "F": 0.19147, "G": 0.19718, "chunks": 196.06, "bytes_per_chunk": 5.223, "boundary_entropy": 0.08752, "router_noise": 0.0, "grad_norm": 0.7298, "lr": 3.41e-05, "router_trainable": true, "sec_per_step": 0.545, "probe_sec": 0.037, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03114, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.27832, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 0, "read_hit_rate": {"surface": 0.69289}, "mean_gate": {"surface": 0.1815}, "n_reads": 4325, "n_cand": 6242, "read_suppressed_frac": 0.1047, "dropout_reads_logged": 4325, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 385.73504638671875, "n_util_obs": 1276, "util_count_spearman": 0.10606873383365618}, "probe_overhead_sec_total": 343.217, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.024312, "alpha": 0.03, "t": 1578.4} |
| {"event": "consolidate", "step": 5600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1593.5} |
| {"event": "probe", "step": 5600, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3172.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.182267, "cps": 24.78125, "mean_p": 0.188399, "boundary_entropy": 0.079662, "bpb": 1.879517}, "tr": {"chunks": 3521.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.194498, "cps": 27.507812, "mean_p": 0.191957, "boundary_entropy": 0.112916, "bpb": 2.010927}, "fi": {"chunks": 4371.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.232364, "cps": 34.148438, "mean_p": 0.230184, "boundary_entropy": 0.086293, "bpb": 2.103566}}, "bpb": {"en": 1.87952, "tr": 2.01093, "fi": 2.10357}, "gini_cpc": 0.054829, "gini_cps": 0.072246, "bpic": 5.127621, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0671031826498767, "fi": 1.2748522788207888}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 389.07366943359375, "n_util_obs": 1281, "util_count_spearman": 0.10455129252584984}, "t": 1594.1} |
| {"event": "eval", "step": 5600, "phase": 3, "bpb": {"en": 1.8935, "tr": 2.02143, "fi": 2.10962}, "bytes_per_chunk": {"en": 5.15, "tr": 5.644, "fi": 4.543}, "bpb_mean": 2.00818, "frozen_inventory": true, "t": 1595.2} |
| {"event": "train", "step": 5600, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35125, "bpb": 1.94944, "chunk_loss": 0.030341, "F": 0.19711, "G": 0.20347, "chunks": 201.84, "bytes_per_chunk": 5.073, "boundary_entropy": 0.08548, "router_noise": 0.0, "grad_norm": 0.7319, "lr": 3.33e-05, "router_trainable": true, "sec_per_step": 0.359, "probe_sec": 0.0246, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01876, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.06399, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.15461, "trunc_abs_frac": 1.87521, "mean_delta_full": 0.24402, "read_hit_rate": {"surface": 0.69395}, "mean_gate": {"surface": 0.17818}, "n_reads": 4460, "n_cand": 6427, "read_suppressed_frac": 0.0948, "dropout_reads_logged": 4460, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 389.14208984375, "n_util_obs": 1282, "util_count_spearman": 0.10580803763589457}, "probe_overhead_sec_total": 346.658, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.01694, "alpha": 0.03, "t": 1596.4} |
| {"event": "consolidate", "step": 5650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1610.6} |
| {"event": "train", "step": 5650, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.33655, "bpb": 1.92823, "chunk_loss": 0.030382, "F": 0.20056, "G": 0.20482, "chunks": 205.38, "bytes_per_chunk": 4.986, "boundary_entropy": 0.08114, "router_noise": 0.0, "grad_norm": 0.6854, "lr": 3.25e-05, "router_trainable": true, "sec_per_step": 0.291, "probe_sec": 0.0719, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06248, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.0572, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 1.82748, "trunc_abs_frac": 1.794, "mean_delta_full": -0.1424, "read_hit_rate": {"surface": 0.70474}, "mean_gate": {"surface": 0.17365}, "n_reads": 4609, "n_cand": 6540, "read_suppressed_frac": 0.0929, "dropout_reads_logged": 4609, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 392.5546875, "n_util_obs": 1290, "util_count_spearman": 0.10230976845917863}, "probe_overhead_sec_total": 349.822, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.018882, "alpha": 0.03, "t": 1610.9} |
| {"event": "consolidate", "step": 5700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1625.8} |
| {"event": "train", "step": 5700, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.34529, "bpb": 1.94085, "chunk_loss": 0.030267, "F": 0.19385, "G": 0.19937, "chunks": 198.5, "bytes_per_chunk": 5.159, "boundary_entropy": 0.08329, "router_noise": 0.0, "grad_norm": 0.7434, "lr": 3.18e-05, "router_trainable": true, "sec_per_step": 0.301, "probe_sec": 0.048, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03699, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.02318, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -0.59473, "trunc_abs_frac": 0.66982, "mean_delta_full": -0.09257, "read_hit_rate": {"surface": 0.68718}, "mean_gate": {"surface": 0.17921}, "n_reads": 4343, "n_cand": 6320, "read_suppressed_frac": 0.096, "dropout_reads_logged": 4343, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 395.96844482421875, "n_util_obs": 1301, "util_count_spearman": 0.09941754776620869}, "probe_overhead_sec_total": 352.967, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.014158, "alpha": 0.03, "t": 1626.0} |
| {"event": "consolidate", "step": 5750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1648.5} |
| {"event": "train", "step": 5750, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.39826, "bpb": 2.01726, "chunk_loss": 0.030498, "F": 0.20401, "G": 0.20905, "chunks": 208.91, "bytes_per_chunk": 4.902, "boundary_entropy": 0.08034, "router_noise": 0.0, "grad_norm": 0.7616, "lr": 3.13e-05, "router_trainable": true, "sec_per_step": 0.455, "probe_sec": 0.0402, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03216, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.0244, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.56815, "trunc_abs_frac": 107.74136, "mean_delta_full": -0.41467, "read_hit_rate": {"surface": 0.69878}, "mean_gate": {"surface": 0.17579}, "n_reads": 4649, "n_cand": 6653, "read_suppressed_frac": 0.0977, "dropout_reads_logged": 4649, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 399.4006042480469, "n_util_obs": 1314, "util_count_spearman": 0.10443501550429483}, "probe_overhead_sec_total": 356.001, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.017002, "alpha": 0.03, "t": 1648.7} |
| {"event": "consolidate", "step": 5800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1661.5} |
| {"event": "probe", "step": 5800, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3197.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.183704, "cps": 24.976562, "mean_p": 0.188685, "boundary_entropy": 0.077009, "bpb": 1.876026}, "tr": {"chunks": 3528.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.194885, "cps": 27.5625, "mean_p": 0.191559, "boundary_entropy": 0.111337, "bpb": 2.007678}, "fi": {"chunks": 4383.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.233002, "cps": 34.242188, "mean_p": 0.230774, "boundary_entropy": 0.083973, "bpb": 2.098495}}, "bpb": {"en": 1.87603, "tr": 2.00768, "fi": 2.0985}, "gini_cpc": 0.053737, "gini_cps": 0.07118, "bpic": 5.10731, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0608635039087753, "fi": 1.2683557180345917}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 402.73370361328125, "n_util_obs": 1323, "util_count_spearman": 0.1052516950530546}, "t": 1662.1} |
| {"event": "eval", "step": 5800, "phase": 3, "bpb": {"en": 1.88984, "tr": 2.01747, "fi": 2.10506}, "bytes_per_chunk": {"en": 5.114, "tr": 5.671, "fi": 4.532}, "bpb_mean": 2.00412, "frozen_inventory": true, "t": 1663.0} |
| {"event": "train", "step": 5800, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32081, "bpb": 1.90553, "chunk_loss": 0.030378, "F": 0.19815, "G": 0.2034, "chunks": 202.91, "bytes_per_chunk": 5.047, "boundary_entropy": 0.08437, "router_noise": 0.0, "grad_norm": 0.7127, "lr": 3.08e-05, "router_trainable": true, "sec_per_step": 0.319, "probe_sec": 0.0487, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04134, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.0538, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.28135, "trunc_abs_frac": 0.63306, "mean_delta_full": 0.39635, "read_hit_rate": {"surface": 0.7067}, "mean_gate": {"surface": 0.17978}, "n_reads": 4566, "n_cand": 6461, "read_suppressed_frac": 0.1053, "dropout_reads_logged": 4566, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 402.80474853515625, "n_util_obs": 1323, "util_count_spearman": 0.10528698375591677}, "probe_overhead_sec_total": 358.748, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.015995, "alpha": 0.03, "t": 1664.6} |
| {"event": "consolidate", "step": 5850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1682.0} |
| {"event": "train", "step": 5850, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.34493, "bpb": 1.94032, "chunk_loss": 0.030383, "F": 0.20175, "G": 0.2064, "chunks": 206.59, "bytes_per_chunk": 4.957, "boundary_entropy": 0.07861, "router_noise": 0.0, "grad_norm": 0.725, "lr": 3.05e-05, "router_trainable": true, "sec_per_step": 0.352, "probe_sec": 0.0662, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05872, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.0656, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.37282, "trunc_abs_frac": 0.97895, "mean_delta_full": 0.03851, "read_hit_rate": {"surface": 0.69509}, "mean_gate": {"surface": 0.16929}, "n_reads": 4573, "n_cand": 6579, "read_suppressed_frac": 0.0991, "dropout_reads_logged": 4573, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 406.20770263671875, "n_util_obs": 1333, "util_count_spearman": 0.09996683218655733}, "probe_overhead_sec_total": 361.575, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.01889, "alpha": 0.03, "t": 1682.2} |
| {"event": "consolidate", "step": 5900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1695.2} |
| {"event": "train", "step": 5900, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.27718, "bpb": 1.84259, "chunk_loss": 0.030278, "F": 0.19406, "G": 0.19961, "chunks": 198.72, "bytes_per_chunk": 5.153, "boundary_entropy": 0.08095, "router_noise": 0.0, "grad_norm": 0.7024, "lr": 3.02e-05, "router_trainable": true, "sec_per_step": 0.268, "probe_sec": 0.1019, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09248, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.16371, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.42828, "trunc_abs_frac": 3.56583, "mean_delta_full": 0.19147, "read_hit_rate": {"surface": 0.6967}, "mean_gate": {"surface": 0.17639}, "n_reads": 4408, "n_cand": 6327, "read_suppressed_frac": 0.1012, "dropout_reads_logged": 4408, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 409.6495361328125, "n_util_obs": 1347, "util_count_spearman": 0.09780908508170598}, "probe_overhead_sec_total": 364.384, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.016659, "alpha": 0.03, "t": 1695.6} |
| {"event": "consolidate", "step": 5950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1712.5} |
| {"event": "train", "step": 5950, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.37896, "bpb": 1.98942, "chunk_loss": 0.030364, "F": 0.19986, "G": 0.20533, "chunks": 204.66, "bytes_per_chunk": 5.004, "boundary_entropy": 0.08342, "router_noise": 0.0, "grad_norm": 0.7231, "lr": 3.01e-05, "router_trainable": true, "sec_per_step": 0.342, "probe_sec": 0.0398, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.029, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.04131, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 1.94148, "trunc_abs_frac": 2.0459, "mean_delta_full": -0.01199, "read_hit_rate": {"surface": 0.68513}, "mean_gate": {"surface": 0.17801}, "n_reads": 4465, "n_cand": 6517, "read_suppressed_frac": 0.1006, "dropout_reads_logged": 4465, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 413.0550842285156, "n_util_obs": 1361, "util_count_spearman": 0.10117960872321163}, "probe_overhead_sec_total": 369.002, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.017227, "alpha": 0.03, "t": 1712.7} |
| {"event": "consolidate", "step": 6000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1726.8} |
| {"event": "train", "step": 5999, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38539, "bpb": 1.99869, "chunk_loss": 0.030402, "F": 0.19955, "G": 0.20413, "chunks": 204.34, "bytes_per_chunk": 5.011, "boundary_entropy": 0.07919, "router_noise": 0.0, "grad_norm": 0.7232, "lr": 3e-05, "router_trainable": true, "sec_per_step": 0.283, "probe_sec": 0.036, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02921, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.03919, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -0.05618, "trunc_abs_frac": 1.36765, "mean_delta_full": 0.01304, "read_hit_rate": {"surface": 0.70078}, "mean_gate": {"surface": 0.1735}, "n_reads": 4560, "n_cand": 6507, "read_suppressed_frac": 0.1031, "dropout_reads_logged": 4560, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 416.4355163574219, "n_util_obs": 1370, "util_count_spearman": 0.10343058858381184}, "probe_overhead_sec_total": 372.436, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.015332, "alpha": 0.03, "t": 1726.9} |
| {"event": "probe", "step": 6000, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3189.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.183244, "cps": 24.914062, "mean_p": 0.18757, "boundary_entropy": 0.076848, "bpb": 1.874307}, "tr": {"chunks": 3551.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.196155, "cps": 27.742188, "mean_p": 0.192424, "boundary_entropy": 0.111726, "bpb": 2.0035}, "fi": {"chunks": 4368.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.232205, "cps": 34.125, "mean_p": 0.230056, "boundary_entropy": 0.082232, "bpb": 2.096162}}, "bpb": {"en": 1.87431, "tr": 2.0035, "fi": 2.09616}, "gini_cpc": 0.053368, "gini_cps": 0.07076, "bpic": 5.10731, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.070458220959827, "fi": 1.2671859447743117}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 416.4355163574219, "n_util_obs": 1370, "util_count_spearman": 0.10343058858381184}, "t": 1727.7} |
| {"event": "eval", "step": 6000, "phase": 3, "bpb": {"en": 1.88853, "tr": 2.01289, "fi": 2.10216}, "bytes_per_chunk": {"en": 5.144, "tr": 5.642, "fi": 4.532}, "bpb_mean": 2.00119, "frozen_inventory": true, "t": 1728.9} |
| {"event": "reps_dumped", "dirs": ["/local/scratch/sas245/beetle-hnet/runs/ali_mem-static_s2/probe/reps/th0.30", "/local/scratch/sas245/beetle-hnet/runs/ali_mem-static_s2/probe/reps/th0.40", "/local/scratch/sas245/beetle-hnet/runs/ali_mem-static_s2/probe/reps/th0.50", "/local/scratch/sas245/beetle-hnet/runs/ali_mem-static_s2/probe/reps/th0.60", "/local/scratch/sas245/beetle-hnet/runs/ali_mem-static_s2/probe/reps/th0.70"], "t": 1730.5} |
| {"event": "final", "run": "ali_mem-static_s2", "arch": "ali", "objective": "ratio", "condition": "A", "step": 6000, "params": 22785088, "seed": 2, "per_lang": {"en": {"cpc": 0.1832442682296156, "cps": 24.9140625, "bpb": 1.8941970848535468, "bytes_per_chunk": 5.072494896428951, "reads_per_chunk": 0.7075629601679583}, "tr": {"cpc": 0.19615533337015964, "cps": 27.7421875, "bpb": 2.00349994918097, "bytes_per_chunk": 5.670246476969166, "reads_per_chunk": 0.6428733655624412}, "fi": {"cpc": 0.2322045611610228, "cps": 34.125, "bpb": 2.1211019545638505, "bytes_per_chunk": 4.510198288098413, "reads_per_chunk": 0.6642866359843435}}, "bpb_mean": 2.006266329532789, "gini_cpc": 0.05336817058318566, "gini_cps": 0.07075981274756948, "bpic": 5.107310046813108, "mean_boundary_entropy": 0.07919356226921082, "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": 372.43626976013184, "probe_overhead_seq_frac": 0.49983333333333335, "wall_sec": 1733.330031633377, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 416.4355163574219, "n_util_obs": 1370, "util_count_spearman": 0.10343058858381184}, "ali_params": 268096, "ali_config": {"enabled": true, "budget_B": 65536, "staging_capacity": 262144, "policy": "frequency", "value_mode": "static", "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": 0.10343058858381184, "paired_vs_dropout": {"n": 2905, "pearson": 0.027819030713503265, "spearman": 0.03131138480854185, "mean_abs_discrepancy": 0.9906601309776306, "mean_paired": 0.06966576725244522, "mean_dropout": 0.10589868575334549}, "predictor_calibration": {"n": 6762, "all": {"n": 6762, "pearson": 0.10358355946451712, "spearman": 0.14512896940135786, "mae": 0.12370336800813675, "reliability": [{"pred": 0.006918847095221281, "obs": 0.01674928516149521, "n": 677}, {"pred": 0.01974477618932724, "obs": 0.039461299777030945, "n": 676}, {"pred": 0.024014290422201157, "obs": 0.033669184893369675, "n": 676}, {"pred": 0.0270242802798748, "obs": 0.039033617824316025, "n": 676}, {"pred": 0.029993468895554543, "obs": 0.051629163324832916, "n": 676}, {"pred": 0.03322902321815491, "obs": 0.06770214438438416, "n": 676}, {"pred": 0.036821749061346054, "obs": 0.06338225305080414, "n": 676}, {"pred": 0.041446734219789505, "obs": 0.06890181452035904, "n": 676}, {"pred": 0.049227867275476456, "obs": 0.071042999625206, "n": 676}, {"pred": 0.07038162648677826, "obs": 0.10020823031663895, "n": 677}]}, "admitted": {"n": 6762, "pearson": 0.10358355946451712, "spearman": 0.14512896940135786, "mae": 0.12370336800813675, "reliability": [{"pred": 0.006918847095221281, "obs": 0.01674928516149521, "n": 677}, {"pred": 0.01974477618932724, "obs": 0.039461299777030945, "n": 676}, {"pred": 0.024014290422201157, "obs": 0.033669184893369675, "n": 676}, {"pred": 0.0270242802798748, "obs": 0.039033617824316025, "n": 676}, {"pred": 0.029993468895554543, "obs": 0.051629163324832916, "n": 676}, {"pred": 0.03322902321815491, "obs": 0.06770214438438416, "n": 676}, {"pred": 0.036821749061346054, "obs": 0.06338225305080414, "n": 676}, {"pred": 0.041446734219789505, "obs": 0.06890181452035904, "n": 676}, {"pred": 0.049227867275476456, "obs": 0.071042999625206, "n": 676}, {"pred": 0.07038162648677826, "obs": 0.10020823031663895, "n": 677}]}, "staged": {"n": 0, "pearson": NaN, "spearman": NaN, "mae": NaN}}, "t": 1733.4} |
| {"event": "hub_done", "repo": "AdaptiveChunking/ali_mem-static_s2", "pushed": [0, 1, 2, 4, 6, 8, 16, 32, 64, 128, 256, 512, 1000, 1024, 2048, 3000, 4096, 6000], "failed": [], "ok": true, "t": 1748.0} |
|
|
