| {"event": "hub", "repo": "AdaptiveChunking/ali_ali-ema_s2", "keep_steps": [0, 1, 2, 4, 6, 8, 16, 32, 64, 128, 256, 512, 1000, 1024, 2048, 3000, 4096, 6000], "t": 3.0} |
| {"event": "start", "run": "ali_ali-ema_s2", "arch": "ali", "objective": "ratio", "condition": "A", "params": 22785088, "langs": ["en", "tr", "fi"], "premium": {"en": 1.0, "tr": 1.12, "fi": 1.104}, "schedule": {"tau1": 0.4, "tau2": 0.7, "p1_objective": "global", "p2_objective": "global", "p3_objective": "global", "freeze_in_p3": false, "freeze_scope": "router"}, "corpus": "/local/scratch/sas245/beetle-hnet/scratch/data/tier1_corpus", "premium_kind": "flores_per_sent", "doc_filtered": {}, "weights": {"en": 0.74, "tr": 0.1536, "fi": 0.1064}, "device": "cuda", "ali_params": 268096, "ali_config": {"enabled": true, "budget_B": 65536, "staging_capacity": 262144, "policy": "utility", "candidate_types": ["surface", "split", "merge", "skel"], "gate": "vector", "retrieval": true, "memory_scope": "persistent", "read_dropout": 0.1, "write_rule": "ema_floor", "eta_min": 0.02, "consolidate_every": 50, "probe_seqs": 2, "probe_reps": 8, "horizon": 32, "predictor": "mlp", "lambda_util": 0.1, "max_span": 64, "byte_classes_path": "/local/scratch/sas245/beetle-hnet/scratch/data/byte_classes.json", "gate_hidden": 0, "type_dim": 16, "gate_bias_init": -2.0}, "t": 3.0} |
| {"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": 3.0} |
| {"event": "probe", "step": 0, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 10933.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.628225, "cps": 85.414062, "mean_p": 0.513249, "boundary_entropy": 0.686295, "bpb": 8.12598}, "tr": {"chunks": 8187.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.452245, "cps": 63.960938, "mean_p": 0.498002, "boundary_entropy": 0.686826, "bpb": 8.097536}, "fi": {"chunks": 13540.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.719792, "cps": 105.78125, "mean_p": 0.517169, "boundary_entropy": 0.686661, "bpb": 8.065117}}, "bpb": {"en": 8.12598, "tr": 8.09754, "fi": 8.06512}, "gini_cpc": 0.099077, "gini_cps": 0.109267, "bpic": 1.737048, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.7198781817505053, "fi": 1.1457544508645294}, "inventory": {"size": 0, "admitted": 0, "staged": 0, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 0.0, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 7.1} |
| {"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": 11.4936, "lr": 1e-06, "router_trainable": true, "sec_per_step": 0.051, "probe_sec": 0.048, "probe_extra_seqs": 0, "n_rows": 0, "extra_fwd_seqs": 0, "wall_s": 0.0, "overhead_frac_est": 0.0, "chunk_mismatch": 0, "n_loo": 0, "mean_delta_loo": NaN, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "read_hit_rate": {"surface": 0.0, "split": 0.0, "merge": 0.0, "skel": 0.0}, "mean_gate": {"surface": null, "split": null, "merge": null, "skel": null}, "n_reads": 0, "n_cand": 58765, "read_suppressed_frac": 0.0, "dropout_reads_logged": 0, "inventory": {"size": 4975, "admitted": 0, "staged": 4975, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 2006, "admitted_split": 0, "staged_split": 1198, "admitted_merge": 0, "staged_merge": 852, "admitted_skel": 0, "staged_skel": 919, "mean_count": 0.0, "n_util_obs": 0, "util_count_spearman": NaN}, "probe_overhead_sec_total": 0.048, "probe_overhead_seq_frac": 0.0, "util_loss": -0.0, "alpha": 0.03, "t": 11.7} |
| {"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": 8.094365}, "tr": {"chunks": 8206.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.453295, "cps": 64.109375, "mean_p": 0.498082, "boundary_entropy": 0.686828, "bpb": 8.076188}, "fi": {"chunks": 13503.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.717825, "cps": 105.492188, "mean_p": 0.517038, "boundary_entropy": 0.686672, "bpb": 8.036121}}, "bpb": {"en": 8.09436, "tr": 8.07619, "fi": 8.03612}, "gini_cpc": 0.098088, "gini_cps": 0.108267, "bpic": 1.739338, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.7232025555756703, "fi": 1.1452422859105857}, "inventory": {"size": 4975, "admitted": 0, "staged": 4975, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 2006, "admitted_split": 0, "staged_split": 1198, "admitted_merge": 0, "staged_merge": 852, "admitted_skel": 0, "staged_skel": 919, "mean_count": 0.0, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 15.4} |
| {"event": "probe", "step": 2, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 10881.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.625237, "cps": 85.007812, "mean_p": 0.512978, "boundary_entropy": 0.686318, "bpb": 8.031143}, "tr": {"chunks": 8246.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.455505, "cps": 64.421875, "mean_p": 0.498244, "boundary_entropy": 0.686832, "bpb": 8.031874}, "fi": {"chunks": 13466.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.715858, "cps": 105.203125, "mean_p": 0.516798, "boundary_entropy": 0.686692, "bpb": 7.976331}}, "bpb": {"en": 8.03114, "tr": 8.03187, "fi": 7.97633}, "gini_cpc": 0.09661, "gini_cps": 0.106771, "bpic": 1.740619, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.7285310882504717, "fi": 1.1449381764657145}, "inventory": {"size": 8946, "admitted": 0, "staged": 8946, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 3535, "admitted_split": 0, "staged_split": 2218, "admitted_merge": 0, "staged_merge": 1439, "admitted_skel": 0, "staged_skel": 1754, "mean_count": 0.0, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 22.8} |
| {"event": "probe", "step": 4, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 10712.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.615526, "cps": 83.6875, "mean_p": 0.512102, "boundary_entropy": 0.68638, "bpb": 7.8197}, "tr": {"chunks": 8385.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.463183, "cps": 65.507812, "mean_p": 0.498671, "boundary_entropy": 0.686846, "bpb": 7.87881}, "fi": {"chunks": 13216.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.702568, "cps": 103.25, "mean_p": 0.515685, "boundary_entropy": 0.686774, "bpb": 7.773627}}, "bpb": {"en": 7.8197, "tr": 7.87881, "fi": 7.77363}, "gini_cpc": 0.089593, "gini_cps": 0.099671, "bpic": 1.755702, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.7524992505131102, "fi": 1.141410077819419}, "inventory": {"size": 15210, "admitted": 0, "staged": 15210, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4, "admitted_surface": 0, "staged_surface": 5977, "admitted_split": 0, "staged_split": 3782, "admitted_merge": 0, "staged_merge": 2535, "admitted_skel": 0, "staged_skel": 2916, "mean_count": 0.0, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 35.1} |
| {"event": "probe", "step": 6, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 10327.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.593403, "cps": 80.679688, "mean_p": 0.510352, "boundary_entropy": 0.686483, "bpb": 7.542835}, "tr": {"chunks": 8504.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.469756, "cps": 66.4375, "mean_p": 0.499188, "boundary_entropy": 0.686869, "bpb": 7.674266}, "fi": {"chunks": 12732.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.676838, "cps": 99.46875, "mean_p": 0.513643, "boundary_entropy": 0.686908, "bpb": 7.518219}}, "bpb": {"en": 7.54283, "tr": 7.67427, "fi": 7.51822}, "gini_cpc": 0.079342, "gini_cps": 0.089303, "bpic": 1.797421, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.791630727628789, "fi": 1.1406034731681387}, "inventory": {"size": 19891, "admitted": 0, "staged": 19891, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 7, "admitted_surface": 0, "staged_surface": 7710, "admitted_split": 0, "staged_split": 4789, "admitted_merge": 0, "staged_merge": 3603, "admitted_skel": 0, "staged_skel": 3789, "mean_count": 0.0, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 44.4} |
| {"event": "probe", "step": 8, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 9630.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.553353, "cps": 75.234375, "mean_p": 0.507287, "boundary_entropy": 0.686614, "bpb": 7.293066}, "tr": {"chunks": 8638.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.477158, "cps": 67.484375, "mean_p": 0.499596, "boundary_entropy": 0.686902, "bpb": 7.47805}, "fi": {"chunks": 11870.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.631014, "cps": 92.734375, "mean_p": 0.51029, "boundary_entropy": 0.687081, "bpb": 7.295882}}, "bpb": {"en": 7.29307, "tr": 7.47805, "fi": 7.29588}, "gini_cpc": 0.061733, "gini_cps": 0.071493, "bpic": 1.882408, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.8623041601854946, "fi": 1.140346065825143}, "inventory": {"size": 25367, "admitted": 0, "staged": 25367, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 9, "admitted_surface": 0, "staged_surface": 9813, "admitted_split": 0, "staged_split": 6097, "admitted_merge": 0, "staged_merge": 4556, "admitted_skel": 0, "staged_skel": 4901, "mean_count": 0.0, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 52.1} |
| {"event": "probe", "step": 16, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3751.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.215538, "cps": 29.304688, "mean_p": 0.484386, "boundary_entropy": 0.686161, "bpb": 6.724762}, "tr": {"chunks": 7273.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.401757, "cps": 56.820312, "mean_p": 0.494384, "boundary_entropy": 0.686996, "bpb": 7.013501}, "fi": {"chunks": 4587.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.243847, "cps": 35.835938, "mean_p": 0.486533, "boundary_entropy": 0.686903, "bpb": 6.787305}}, "bpb": {"en": 6.72476, "tr": 7.0135, "fi": 6.78731}, "gini_cpc": 0.144165, "gini_cps": 0.150407, "bpic": 3.634104, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.8639750385369256, "fi": 1.1313420066346032}, "inventory": {"size": 63000, "admitted": 0, "staged": 63000, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 23, "admitted_surface": 0, "staged_surface": 23782, "admitted_split": 0, "staged_split": 17661, "admitted_merge": 0, "staged_merge": 6732, "admitted_skel": 0, "staged_skel": 14825, "mean_count": 0.0, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 62.6} |
| {"event": "probe", "step": 32, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 2604.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.149629, "cps": 20.34375, "mean_p": 0.486737, "boundary_entropy": 0.686947, "bpb": 6.304656}, "tr": {"chunks": 4649.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.256808, "cps": 36.320312, "mean_p": 0.4902, "boundary_entropy": 0.687561, "bpb": 6.651796}, "fi": {"chunks": 1396.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.074212, "cps": 10.90625, "mean_p": 0.479424, "boundary_entropy": 0.686976, "bpb": 6.420884}}, "bpb": {"en": 6.30466, "tr": 6.6518, "fi": 6.42088}, "gini_cpc": 0.253263, "gini_cps": 0.250742, "bpic": 6.559371, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.7162957816142193, "fi": 0.495971447238808}, "inventory": {"size": 138819, "admitted": 0, "staged": 138819, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 47, "admitted_surface": 0, "staged_surface": 50319, "admitted_split": 0, "staged_split": 43526, "admitted_merge": 0, "staged_merge": 8687, "admitted_skel": 0, "staged_skel": 36287, "mean_count": 0.0, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 89.4} |
| {"event": "consolidate", "step": 50, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 3037, "swapped": 0, "uncertain_staged": 230596, "t": 122.9} |
| {"event": "train", "step": 50, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 4.21236, "bpb": 6.07715, "chunk_loss": 0.034232, "F": 0.22614, "G": 0.48852, "chunks": 231.56, "bytes_per_chunk": 4.422, "boundary_entropy": 0.69161, "router_noise": 0.0, "grad_norm": 2.4795, "lr": 5.1e-05, "router_trainable": true, "sec_per_step": 2.25, "probe_sec": 0.1145, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06616, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": 0.00836, "n_shadow": 5, "mean_delta_shadow": 0.00779, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 1.02249, "trunc_abs_frac": 2.93977, "mean_delta_full": 0.01846, "read_hit_rate": {"surface": 0.23325, "split": 0.0676, "merge": 0.0, "skel": 0.0}, "mean_gate": {"surface": 0.11896, "split": 0.11928, "merge": null, "skel": null}, "n_reads": 2345, "n_cand": 27554, "read_suppressed_frac": 0.0994, "dropout_reads_logged": 2345, "inventory": {"size": 236152, "admitted": 3037, "staged": 233115, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 80, "admitted_surface": 1, "staged_surface": 84412, "admitted_split": 2, "staged_split": 76816, "admitted_merge": 0, "staged_merge": 10407, "admitted_skel": 3034, "staged_skel": 61480, "mean_count": 30.82383918762207, "n_util_obs": 3, "util_count_spearman": 0.5}, "probe_overhead_sec_total": 7.841, "probe_overhead_seq_frac": 0.4902, "util_loss": 0.073386, "alpha": 0.03, "t": 124.2} |
| {"event": "probe", "step": 64, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 4801.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.275872, "cps": 37.507812, "mean_p": 0.485152, "boundary_entropy": 0.685508, "bpb": 5.552647}, "tr": {"chunks": 5408.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.298735, "cps": 42.25, "mean_p": 0.486607, "boundary_entropy": 0.686484, "bpb": 5.956909}, "fi": {"chunks": 4713.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.250545, "cps": 36.820312, "mean_p": 0.484039, "boundary_entropy": 0.686334, "bpb": 5.718363}}, "bpb": {"en": 5.55265, "tr": 5.95691, "fi": 5.71836}, "gini_cpc": 0.038934, "gini_cps": 0.03105, "bpic": 3.801903, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0828755443851386, "fi": 0.9081926243127254}, "inventory": {"size": 265181, "admitted": 3037, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 82, "admitted_surface": 1, "staged_surface": 93127, "admitted_split": 2, "staged_split": 86866, "admitted_merge": 0, "staged_merge": 11380, "admitted_skel": 3034, "staged_skel": 70771, "mean_count": 38.370758056640625, "n_util_obs": 3, "util_count_spearman": -0.5}, "t": 146.6} |
| {"event": "consolidate", "step": 100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 62499, "swapped": 3, "uncertain_staged": 262139, "t": 211.1} |
| {"event": "train", "step": 100, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 3.24834, "bpb": 4.68637, "chunk_loss": 0.02962, "F": 0.16034, "G": 0.45215, "chunks": 164.19, "bytes_per_chunk": 6.237, "boundary_entropy": 0.68184, "router_noise": 0.0, "grad_norm": 1.5075, "lr": 0.000101, "router_trainable": true, "sec_per_step": 1.758, "probe_sec": 0.1134, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07083, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": 0.03218, "n_shadow": 10, "mean_delta_shadow": -0.0166, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.13732, "trunc_abs_frac": 1.07033, "mean_delta_full": -0.01733, "read_hit_rate": {"surface": 0.05824, "split": 0.0, "merge": 0.0, "skel": 0.33117}, "mean_gate": {"surface": 0.11911, "split": null, "merge": null, "skel": 0.11912}, "n_reads": 1426, "n_cand": 22600, "read_suppressed_frac": 0.1087, "dropout_reads_logged": 1426, "inventory": {"size": 270587, "admitted": 65536, "staged": 205051, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 100, "admitted_surface": 4684, "staged_surface": 91360, "admitted_split": 0, "staged_split": 85011, "admitted_merge": 1, "staged_merge": 14726, "admitted_skel": 60851, "staged_skel": 13954, "mean_count": 4.4248046875, "n_util_obs": 0, "util_count_spearman": NaN}, "probe_overhead_sec_total": 14.568, "probe_overhead_seq_frac": 0.495, "util_loss": 0.003194, "alpha": 0.03, "t": 212.1} |
| {"event": "probe", "step": 128, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 4055.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.233006, "cps": 31.679688, "mean_p": 0.420034, "boundary_entropy": 0.646606, "bpb": 4.162535}, "tr": {"chunks": 4845.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.267635, "cps": 37.851562, "mean_p": 0.437864, "boundary_entropy": 0.65965, "bpb": 4.4012}, "fi": {"chunks": 3893.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.206953, "cps": 30.414062, "mean_p": 0.430546, "boundary_entropy": 0.660571, "bpb": 4.354393}}, "bpb": {"en": 4.16254, "tr": 4.4012, "fi": 4.35439}, "gini_cpc": 0.057172, "gini_cps": 0.04961, "bpic": 4.434613, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.1486203109716133, "fi": 0.8881898010588946}, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 128, "admitted_surface": 4684, "staged_surface": 106631, "admitted_split": 0, "staged_split": 92490, "admitted_merge": 1, "staged_merge": 18066, "admitted_skel": 60851, "staged_skel": 44957, "mean_count": 5.196990966796875, "n_util_obs": 6, "util_count_spearman": -0.2571428571428571}, "t": 253.1} |
| {"event": "consolidate", "step": 150, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 8, "uncertain_staged": 262138, "t": 299.2} |
| {"event": "train", "step": 150, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.7933, "bpb": 4.02989, "chunk_loss": 0.031639, "F": 0.20129, "G": 0.38215, "chunks": 206.12, "bytes_per_chunk": 4.968, "boundary_entropy": 0.61855, "router_noise": 0.0, "grad_norm": 1.6236, "lr": 0.000151, "router_trainable": true, "sec_per_step": 1.758, "probe_sec": 0.0837, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04341, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": 0.00937, "n_shadow": 3, "mean_delta_shadow": -0.02669, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.00488, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.08768, "trunc_abs_frac": 0.36935, "mean_delta_full": 0.09119, "read_hit_rate": {"surface": 0.18221, "split": 0.0, "merge": 0.0, "skel": 0.21907}, "mean_gate": {"surface": 0.12034, "split": null, "merge": null, "skel": 0.12245}, "n_reads": 2046, "n_cand": 26862, "read_suppressed_frac": 0.0943, "dropout_reads_logged": 2046, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 139, "admitted_surface": 4687, "staged_surface": 100763, "admitted_split": 0, "staged_split": 85006, "admitted_merge": 1, "staged_merge": 21014, "admitted_skel": 60848, "staged_skel": 55361, "mean_count": 5.6654815673828125, "n_util_obs": 0, "util_count_spearman": NaN}, "probe_overhead_sec_total": 21.74, "probe_overhead_seq_frac": 0.4967, "util_loss": 0.001063, "alpha": 0.03, "t": 300.0} |
| {"event": "consolidate", "step": 200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 9, "uncertain_staged": 262133, "t": 377.9} |
| {"event": "probe", "step": 200, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3703.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.212779, "cps": 28.929688, "mean_p": 0.372739, "boundary_entropy": 0.584824, "bpb": 3.637026}, "tr": {"chunks": 4065.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.224548, "cps": 31.757812, "mean_p": 0.383076, "boundary_entropy": 0.604246, "bpb": 3.801495}, "fi": {"chunks": 3791.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.201531, "cps": 29.617188, "mean_p": 0.407756, "boundary_entropy": 0.639842, "bpb": 3.862243}}, "bpb": {"en": 3.63703, "tr": 3.80149, "fi": 3.86224}, "gini_cpc": 0.024019, "gini_cps": 0.020878, "bpic": 4.908037, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.05531085817646, "fi": 0.9471359236102928}, "inventory": {"size": 327671, "admitted": 65536, "staged": 262135, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 179, "admitted_surface": 4683, "staged_surface": 97185, "admitted_split": 0, "staged_split": 85561, "admitted_merge": 1, "staged_merge": 26597, "admitted_skel": 60852, "staged_skel": 52792, "mean_count": 5.1602630615234375, "n_util_obs": 2, "util_count_spearman": NaN}, "t": 379.6} |
| {"event": "eval", "step": 200, "phase": 1, "bpb": {"en": 3.64499, "tr": 3.81224, "fi": 3.85981}, "bytes_per_chunk": {"en": 5.287, "tr": 5.367, "fi": 6.068}, "bpb_mean": 3.77235, "frozen_inventory": true, "t": 382.7} |
| {"event": "train", "step": 200, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.53626, "bpb": 3.65905, "chunk_loss": 0.030565, "F": 0.17941, "G": 0.36426, "chunks": 183.72, "bytes_per_chunk": 5.574, "boundary_entropy": 0.58777, "router_noise": 0.0, "grad_norm": 1.5472, "lr": 0.000201, "router_trainable": true, "sec_per_step": 1.789, "probe_sec": 0.0613, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04766, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.00838, "n_shadow": 6, "mean_delta_shadow": 0.03514, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.48996, "trunc_abs_frac": 0.30086, "mean_delta_full": 0.07977, "read_hit_rate": {"surface": 0.0077, "split": 0.0, "merge": 0.0, "skel": 0.13465}, "mean_gate": {"surface": 0.12389, "split": null, "merge": null, "skel": 0.12682}, "n_reads": 542, "n_cand": 24817, "read_suppressed_frac": 0.1052, "dropout_reads_logged": 542, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 138, "admitted_surface": 4683, "staged_surface": 96980, "admitted_split": 0, "staged_split": 85798, "admitted_merge": 1, "staged_merge": 25987, "admitted_skel": 60852, "staged_skel": 53379, "mean_count": 5.1685333251953125, "n_util_obs": 3, "util_count_spearman": -0.5}, "probe_overhead_sec_total": 27.833, "probe_overhead_seq_frac": 0.4975, "util_loss": 0.000786, "alpha": 0.03, "t": 389.4} |
| {"event": "consolidate", "step": 250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262130, "t": 465.3} |
| {"event": "train", "step": 250, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.3979, "bpb": 3.45943, "chunk_loss": 0.031147, "F": 0.19177, "G": 0.3624, "chunks": 196.38, "bytes_per_chunk": 5.215, "boundary_entropy": 0.56125, "router_noise": 0.0, "grad_norm": 2.5705, "lr": 0.000251, "router_trainable": true, "sec_per_step": 1.541, "probe_sec": 0.1694, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07986, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": -0.00116, "n_shadow": 7, "mean_delta_shadow": 0.03383, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 1.37795, "trunc_abs_frac": 0.67553, "mean_delta_full": -0.0155, "read_hit_rate": {"surface": 0.02831, "split": 0.15002, "merge": 0.0, "skel": 0.12536}, "mean_gate": {"surface": 0.12656, "split": 0.1252, "merge": null, "skel": 0.12836}, "n_reads": 2202, "n_cand": 26652, "read_suppressed_frac": 0.0995, "dropout_reads_logged": 2202, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 73, "admitted_surface": 4627, "staged_surface": 94408, "admitted_split": 8, "staged_split": 82624, "admitted_merge": 3, "staged_merge": 27566, "admitted_skel": 60898, "staged_skel": 57546, "mean_count": 8.66729736328125, "n_util_obs": 14, "util_count_spearman": -0.3230769230769231}, "probe_overhead_sec_total": 34.129, "probe_overhead_seq_frac": 0.498, "util_loss": 0.000484, "alpha": 0.03, "t": 466.5} |
| {"event": "probe", "step": 256, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3121.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.179337, "cps": 24.382812, "mean_p": 0.35392, "boundary_entropy": 0.553782, "bpb": 3.367303}, "tr": {"chunks": 4274.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.236093, "cps": 33.390625, "mean_p": 0.339642, "boundary_entropy": 0.539786, "bpb": 3.543455}, "fi": {"chunks": 4070.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.216363, "cps": 31.796875, "mean_p": 0.373962, "boundary_entropy": 0.589615, "bpb": 3.67826}}, "bpb": {"en": 3.3673, "tr": 3.54345, "fi": 3.67826}, "gini_cpc": 0.059889, "gini_cps": 0.067045, "bpic": 4.948277, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.3164801584043375, "fi": 1.2064598605624055}, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 74, "admitted_surface": 4627, "staged_surface": 94042, "admitted_split": 8, "staged_split": 82070, "admitted_merge": 3, "staged_merge": 28288, "admitted_skel": 60898, "staged_skel": 57744, "mean_count": 8.815170288085938, "n_util_obs": 15, "util_count_spearman": -0.4928571428571429}, "t": 474.6} |
| {"event": "consolidate", "step": 300, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262132, "t": 552.7} |
| {"event": "train", "step": 300, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.28439, "bpb": 3.29568, "chunk_loss": 0.030637, "F": 0.18509, "G": 0.31689, "chunks": 189.53, "bytes_per_chunk": 5.403, "boundary_entropy": 0.49364, "router_noise": 0.0, "grad_norm": 1.4687, "lr": 0.0003, "router_trainable": true, "sec_per_step": 1.75, "probe_sec": 0.0809, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04949, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": 0.01504, "n_shadow": 7, "mean_delta_shadow": 0.02435, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.13727, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": -0.15478, "trunc_abs_frac": 0.15478, "mean_delta_full": -0.19727, "read_hit_rate": {"surface": 0.00431, "split": 0.0, "merge": 0.0, "skel": 0.02632}, "mean_gate": {"surface": 0.12918, "split": null, "merge": null, "skel": 0.12904}, "n_reads": 129, "n_cand": 26084, "read_suppressed_frac": 0.1395, "dropout_reads_logged": 129, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 90, "admitted_surface": 4608, "staged_surface": 90428, "admitted_split": 0, "staged_split": 77610, "admitted_merge": 29, "staged_merge": 34466, "admitted_skel": 60899, "staged_skel": 59640, "mean_count": 2.3400726318359375, "n_util_obs": 5, "util_count_spearman": 0.5}, "probe_overhead_sec_total": 39.838, "probe_overhead_seq_frac": 0.4983, "util_loss": 0.000789, "alpha": 0.03, "t": 554.0} |
| {"event": "consolidate", "step": 350, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262127, "t": 635.7} |
| {"event": "train", "step": 350, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.16789, "bpb": 3.12761, "chunk_loss": 0.030863, "F": 0.19275, "G": 0.30718, "chunks": 197.38, "bytes_per_chunk": 5.188, "boundary_entropy": 0.46278, "router_noise": 0.0, "grad_norm": 1.0375, "lr": 0.0002999, "router_trainable": true, "sec_per_step": 1.672, "probe_sec": 0.13, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0721, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 1, "mean_delta_loo": -0.00336, "n_shadow": 11, "mean_delta_shadow": -0.02621, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.06555, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.10777, "trunc_abs_frac": 0.28217, "mean_delta_full": 0.26221, "read_hit_rate": {"surface": 0.0, "split": 0.0, "merge": 0.0, "skel": 0.02386}, "mean_gate": {"surface": null, "split": null, "merge": null, "skel": 0.1279}, "n_reads": 95, "n_cand": 27121, "read_suppressed_frac": 0.0947, "dropout_reads_logged": 95, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 103, "admitted_surface": 3701, "staged_surface": 86364, "admitted_split": 1, "staged_split": 72199, "admitted_merge": 29, "staged_merge": 45267, "admitted_skel": 61805, "staged_skel": 58314, "mean_count": 2.3533782958984375, "n_util_obs": 6, "util_count_spearman": 0.08571428571428572}, "probe_overhead_sec_total": 47.226, "probe_overhead_seq_frac": 0.4986, "util_loss": 0.000933, "alpha": 0.03, "t": 637.6} |
| {"event": "consolidate", "step": 400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262123, "t": 712.5} |
| {"event": "probe", "step": 400, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 2965.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.170373, "cps": 23.164062, "mean_p": 0.27881, "boundary_entropy": 0.421015, "bpb": 2.915635}, "tr": {"chunks": 3061.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.169088, "cps": 23.914062, "mean_p": 0.280111, "boundary_entropy": 0.445877, "bpb": 3.188016}, "fi": {"chunks": 3896.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.207113, "cps": 30.4375, "mean_p": 0.294363, "boundary_entropy": 0.471191, "bpb": 3.384733}}, "bpb": {"en": 2.91563, "tr": 3.18802, "fi": 3.38473}, "gini_cpc": 0.04638, "gini_cps": 0.062555, "bpic": 5.717799, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.992458145859942, "fi": 1.2156442137700099}, "inventory": {"size": 326656, "admitted": 65536, "staged": 261120, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 119, "admitted_surface": 4169, "staged_surface": 81253, "admitted_split": 38, "staged_split": 68122, "admitted_merge": 342, "staged_merge": 54693, "admitted_skel": 60987, "staged_skel": 57052, "mean_count": 4.4640655517578125, "n_util_obs": 7, "util_count_spearman": -0.2857142857142857}, "t": 714.3} |
| {"event": "eval", "step": 400, "phase": 1, "bpb": {"en": 2.92718, "tr": 3.19439, "fi": 3.37576}, "bytes_per_chunk": {"en": 5.7, "tr": 6.487, "fi": 5.275}, "bpb_mean": 3.16578, "frozen_inventory": true, "t": 717.5} |
| {"event": "train", "step": 400, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.09128, "bpb": 3.01707, "chunk_loss": 0.030103, "F": 0.16898, "G": 0.2797, "chunks": 173.03, "bytes_per_chunk": 5.918, "boundary_entropy": 0.4189, "router_noise": 0.0, "grad_norm": 1.2904, "lr": 0.0002998, "router_trainable": true, "sec_per_step": 1.773, "probe_sec": 0.9991, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08761, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.03767, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": 0.06531, "n_merge": 2, "mean_delta_merge": -0.03339, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 1.35469, "trunc_abs_frac": 3.61334, "mean_delta_full": 0.01628, "read_hit_rate": {"surface": 0.0881, "split": 0.00372, "merge": 0.00822, "skel": 0.05728}, "mean_gate": {"surface": 0.14056, "split": 0.13984, "merge": 0.13905, "skel": 0.13908}, "n_reads": 809, "n_cand": 24899, "read_suppressed_frac": 0.0816, "dropout_reads_logged": 809, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 123, "admitted_surface": 4169, "staged_surface": 81438, "admitted_split": 38, "staged_split": 68453, "admitted_merge": 342, "staged_merge": 54857, "admitted_skel": 60987, "staged_skel": 57396, "mean_count": 4.476409912109375, "n_util_obs": 7, "util_count_spearman": -0.2857142857142857}, "probe_overhead_sec_total": 54.279, "probe_overhead_seq_frac": 0.4988, "util_loss": 0.001754, "alpha": 0.03, "t": 726.3} |
| {"event": "consolidate", "step": 450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262122, "t": 799.4} |
| {"event": "train", "step": 450, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.99671, "bpb": 2.88064, "chunk_loss": 0.030855, "F": 0.19989, "G": 0.27985, "chunks": 204.69, "bytes_per_chunk": 5.003, "boundary_entropy": 0.38493, "router_noise": 0.0, "grad_norm": 1.3143, "lr": 0.0002995, "router_trainable": true, "sec_per_step": 1.502, "probe_sec": 0.1881, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09438, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": -0.00723, "n_shadow": 3, "mean_delta_shadow": -0.03483, "n_split": 1, "mean_delta_split": -0.05157, "n_merge": 3, "mean_delta_merge": -0.08177, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 2.96818, "trunc_abs_frac": 20.05197, "mean_delta_full": -0.0187, "read_hit_rate": {"surface": 0.11322, "split": 0.00199, "merge": 0.01048, "skel": 0.05831}, "mean_gate": {"surface": 0.13174, "split": 0.12917, "merge": 0.13336, "skel": 0.12976}, "n_reads": 1111, "n_cand": 29420, "read_suppressed_frac": 0.1134, "dropout_reads_logged": 1111, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 139, "admitted_surface": 4429, "staged_surface": 77310, "admitted_split": 38, "staged_split": 62952, "admitted_merge": 1097, "staged_merge": 65992, "admitted_skel": 59972, "staged_skel": 55890, "mean_count": 5.392303466796875, "n_util_obs": 10, "util_count_spearman": -0.4909090909090909}, "probe_overhead_sec_total": 60.619, "probe_overhead_seq_frac": 0.4989, "util_loss": 0.002067, "alpha": 0.03, "t": 801.3} |
| {"event": "consolidate", "step": 500, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262118, "t": 871.6} |
| {"event": "train", "step": 500, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.90658, "bpb": 2.75061, "chunk_loss": 0.031136, "F": 0.21176, "G": 0.27564, "chunks": 216.84, "bytes_per_chunk": 4.722, "boundary_entropy": 0.35165, "router_noise": 0.0, "grad_norm": 1.3269, "lr": 0.0002992, "router_trainable": true, "sec_per_step": 1.432, "probe_sec": 0.0726, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05569, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.01171, "n_shadow": 5, "mean_delta_shadow": -0.11788, "n_split": 1, "mean_delta_split": 0.1972, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.50921, "trunc_abs_frac": 1.31095, "mean_delta_full": -0.17236, "read_hit_rate": {"surface": 0.11684, "split": 0.00123, "merge": 0.01236, "skel": 0.05832}, "mean_gate": {"surface": 0.15145, "split": 0.14621, "merge": 0.15472, "skel": 0.14408}, "n_reads": 1214, "n_cand": 31270, "read_suppressed_frac": 0.0832, "dropout_reads_logged": 1214, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 165, "admitted_surface": 5184, "staged_surface": 73327, "admitted_split": 39, "staged_split": 58441, "admitted_merge": 1364, "staged_merge": 76120, "admitted_skel": 58949, "staged_skel": 54256, "mean_count": 6.1021270751953125, "n_util_obs": 14, "util_count_spearman": 0.2835164835164835}, "probe_overhead_sec_total": 67.306, "probe_overhead_seq_frac": 0.499, "util_loss": 0.002856, "alpha": 0.03, "t": 872.9} |
| {"event": "probe", "step": 512, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3473.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.199563, "cps": 27.132812, "mean_p": 0.271219, "boundary_entropy": 0.355133, "bpb": 2.706286}, "tr": {"chunks": 3216.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.17765, "cps": 25.125, "mean_p": 0.274901, "boundary_entropy": 0.401001, "bpb": 3.036297}, "fi": {"chunks": 3874.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.205943, "cps": 30.265625, "mean_p": 0.275542, "boundary_entropy": 0.397278, "bpb": 3.220916}}, "bpb": {"en": 2.70629, "tr": 3.0363, "fi": 3.22092}, "gini_cpc": 0.032345, "gini_cps": 0.041529, "bpic": 5.370823, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.8901943336398994, "fi": 1.0319699942154215}, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 172, "admitted_surface": 5184, "staged_surface": 72482, "admitted_split": 39, "staged_split": 57936, "admitted_merge": 1364, "staged_merge": 78177, "admitted_skel": 58949, "staged_skel": 53549, "mean_count": 6.2899322509765625, "n_util_obs": 14, "util_count_spearman": 0.3142857142857143}, "t": 887.7} |
| {"event": "consolidate", "step": 550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262115, "t": 955.0} |
| {"event": "train", "step": 550, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.91162, "bpb": 2.75789, "chunk_loss": 0.030738, "F": 0.1969, "G": 0.27175, "chunks": 201.62, "bytes_per_chunk": 5.079, "boundary_entropy": 0.3428, "router_noise": 0.0, "grad_norm": 1.7023, "lr": 0.0002987, "router_trainable": true, "sec_per_step": 1.679, "probe_sec": 0.1187, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07511, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 3, "mean_delta_loo": -0.10844, "n_shadow": 9, "mean_delta_shadow": -0.13723, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.0437, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.95154, "trunc_abs_frac": 2.50684, "mean_delta_full": -0.20276, "read_hit_rate": {"surface": 0.09984, "split": 0.0484, "merge": 0.0025, "skel": 0.08005}, "mean_gate": {"surface": 0.16955, "split": 0.17111, "merge": 0.18084, "skel": 0.15321}, "n_reads": 1620, "n_cand": 29363, "read_suppressed_frac": 0.0963, "dropout_reads_logged": 1620, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 188, "admitted_surface": 5320, "staged_surface": 70550, "admitted_split": 36, "staged_split": 55985, "admitted_merge": 1244, "staged_merge": 83457, "admitted_skel": 58936, "staged_skel": 52152, "mean_count": 10.441482543945312, "n_util_obs": 26, "util_count_spearman": -0.003076923076923077}, "probe_overhead_sec_total": 73.947, "probe_overhead_seq_frac": 0.4991, "util_loss": 0.003421, "alpha": 0.03, "t": 956.9} |
| {"event": "consolidate", "step": 600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262116, "t": 1024.0} |
| {"event": "probe", "step": 600, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3429.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.197035, "cps": 26.789062, "mean_p": 0.259164, "boundary_entropy": 0.32326, "bpb": 2.594221}, "tr": {"chunks": 3323.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.183561, "cps": 25.960938, "mean_p": 0.26371, "boundary_entropy": 0.354085, "bpb": 2.922699}, "fi": {"chunks": 4131.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.219606, "cps": 32.273438, "mean_p": 0.279669, "boundary_entropy": 0.3575, "bpb": 3.154294}}, "bpb": {"en": 2.59422, "tr": 2.9227, "fi": 3.15429}, "gini_cpc": 0.040036, "gini_cps": 0.049496, "bpic": 5.212901, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.931614897914041, "fi": 1.1145510019476794}, "inventory": {"size": 326656, "admitted": 65536, "staged": 261120, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 225, "admitted_surface": 5340, "staged_surface": 68352, "admitted_split": 37, "staged_split": 53783, "admitted_merge": 1235, "staged_merge": 89813, "admitted_skel": 58924, "staged_skel": 49172, "mean_count": 13.280563354492188, "n_util_obs": 33, "util_count_spearman": 0.08088235294117647}, "t": 1025.3} |
| {"event": "eval", "step": 600, "phase": 1, "bpb": {"en": 2.60516, "tr": 2.93067, "fi": 3.15556}, "bytes_per_chunk": {"en": 4.846, "tr": 6.013, "fi": 4.784}, "bpb_mean": 2.89713, "frozen_inventory": true, "t": 1028.0} |
| {"event": "train", "step": 600, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.89409, "bpb": 2.7326, "chunk_loss": 0.030815, "F": 0.20276, "G": 0.26425, "chunks": 207.62, "bytes_per_chunk": 4.932, "boundary_entropy": 0.32288, "router_noise": 0.0, "grad_norm": 1.1752, "lr": 0.0002982, "router_trainable": true, "sec_per_step": 1.573, "probe_sec": 0.1788, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10297, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 3, "mean_delta_loo": 0.13074, "n_shadow": 6, "mean_delta_shadow": -0.10551, "n_split": 0, "mean_delta_split": NaN, "n_merge": 3, "mean_delta_merge": -0.01235, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 1.68732, "trunc_abs_frac": 2.04956, "mean_delta_full": -0.10315, "read_hit_rate": {"surface": 0.12492, "split": 0.06454, "merge": 0.00562, "skel": 0.1255}, "mean_gate": {"surface": 0.17092, "split": 0.14329, "merge": 0.15509, "skel": 0.15433}, "n_reads": 2271, "n_cand": 30079, "read_suppressed_frac": 0.1123, "dropout_reads_logged": 2271, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 228, "admitted_surface": 5340, "staged_surface": 68620, "admitted_split": 37, "staged_split": 53805, "admitted_merge": 1235, "staged_merge": 90326, "admitted_skel": 58924, "staged_skel": 49393, "mean_count": 13.315216064453125, "n_util_obs": 33, "util_count_spearman": 0.10394385026737968}, "probe_overhead_sec_total": 79.898, "probe_overhead_seq_frac": 0.4992, "util_loss": 0.003768, "alpha": 0.03, "t": 1035.5} |
| {"event": "consolidate", "step": 650, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262115, "t": 1112.4} |
| {"event": "train", "step": 650, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.8454, "bpb": 2.66235, "chunk_loss": 0.030733, "F": 0.19873, "G": 0.26376, "chunks": 203.5, "bytes_per_chunk": 5.032, "boundary_entropy": 0.31169, "router_noise": 0.0, "grad_norm": 1.2402, "lr": 0.0002975, "router_trainable": true, "sec_per_step": 1.569, "probe_sec": 0.1274, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08241, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": -0.07963, "n_shadow": 3, "mean_delta_shadow": -0.0819, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": 0.05904, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.55833, "trunc_abs_frac": 1.72568, "mean_delta_full": -0.24007, "read_hit_rate": {"surface": 0.13426, "split": 0.06278, "merge": 0.00527, "skel": 0.20837}, "mean_gate": {"surface": 0.17061, "split": 0.14008, "merge": 0.17686, "skel": 0.15184}, "n_reads": 2692, "n_cand": 29600, "read_suppressed_frac": 0.0944, "dropout_reads_logged": 2692, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 230, "admitted_surface": 4633, "staged_surface": 66912, "admitted_split": 38, "staged_split": 52103, "admitted_merge": 943, "staged_merge": 95348, "admitted_skel": 59922, "staged_skel": 47781, "mean_count": 15.79730224609375, "n_util_obs": 40, "util_count_spearman": -0.13302063789868668}, "probe_overhead_sec_total": 86.495, "probe_overhead_seq_frac": 0.4992, "util_loss": 0.003948, "alpha": 0.03, "t": 1114.0} |
| {"event": "consolidate", "step": 700, "retired": 1, "split_retired": 0, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262113, "t": 1194.2} |
| {"event": "train", "step": 700, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.9918, "bpb": 2.87356, "chunk_loss": 0.031017, "F": 0.20471, "G": 0.26826, "chunks": 209.62, "bytes_per_chunk": 4.885, "boundary_entropy": 0.31482, "router_noise": 0.0, "grad_norm": 1.1341, "lr": 0.0002967, "router_trainable": true, "sec_per_step": 1.634, "probe_sec": 0.119, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06722, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.07829, "n_shadow": 4, "mean_delta_shadow": -0.01621, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": -0.06505, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.66269, "trunc_abs_frac": 0.4697, "mean_delta_full": -0.13701, "read_hit_rate": {"surface": 0.1098, "split": 0.11355, "merge": 0.00015, "skel": 0.27783}, "mean_gate": {"surface": 0.16961, "split": 0.14128, "merge": 0.16627, "skel": 0.15033}, "n_reads": 3438, "n_cand": 29843, "read_suppressed_frac": 0.0971, "dropout_reads_logged": 3438, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 255, "admitted_surface": 4567, "staged_surface": 65659, "admitted_split": 42, "staged_split": 50486, "admitted_merge": 3, "staged_merge": 100006, "admitted_skel": 60924, "staged_skel": 45993, "mean_count": 23.44873046875, "n_util_obs": 49, "util_count_spearman": -0.1620408163265306}, "probe_overhead_sec_total": 93.245, "probe_overhead_seq_frac": 0.4993, "util_loss": 0.004683, "alpha": 0.03, "t": 1195.7} |
| {"event": "consolidate", "step": 750, "retired": 0, "split_retired": 3, "merge_admitted": 0, "admitted": 3, "swapped": 787, "uncertain_staged": 262117, "t": 1274.6} |
| {"event": "train", "step": 750, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.78301, "bpb": 2.57234, "chunk_loss": 0.030835, "F": 0.20493, "G": 0.26426, "chunks": 209.84, "bytes_per_chunk": 4.88, "boundary_entropy": 0.30022, "router_noise": 0.0, "grad_norm": 1.1488, "lr": 0.0002959, "router_trainable": true, "sec_per_step": 1.624, "probe_sec": 0.2121, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.11579, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 3, "mean_delta_loo": 0.02741, "n_shadow": 8, "mean_delta_shadow": -0.00621, "n_split": 3, "mean_delta_split": 0.06848, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.25489, "trunc_abs_frac": 2.17019, "mean_delta_full": -0.09363, "read_hit_rate": {"surface": 0.17447, "split": 0.12918, "merge": 0.0, "skel": 0.30037}, "mean_gate": {"surface": 0.16185, "split": 0.13586, "merge": null, "skel": 0.14432}, "n_reads": 4263, "n_cand": 30434, "read_suppressed_frac": 0.0971, "dropout_reads_logged": 4263, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 268, "admitted_surface": 3878, "staged_surface": 64560, "admitted_split": 28, "staged_split": 49248, "admitted_merge": 0, "staged_merge": 103762, "admitted_skel": 61630, "staged_skel": 44574, "mean_count": 28.638076782226562, "n_util_obs": 51, "util_count_spearman": -0.005429864253393665}, "probe_overhead_sec_total": 99.951, "probe_overhead_seq_frac": 0.4993, "util_loss": 0.005421, "alpha": 0.03, "t": 1277.0} |
| {"event": "consolidate", "step": 800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 392, "uncertain_staged": 262114, "t": 1342.5} |
| {"event": "probe", "step": 800, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3341.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.191978, "cps": 26.101562, "mean_p": 0.246338, "boundary_entropy": 0.292665, "bpb": 2.436759}, "tr": {"chunks": 3510.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.193891, "cps": 27.421875, "mean_p": 0.264591, "boundary_entropy": 0.341473, "bpb": 2.767386}, "fi": {"chunks": 4057.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.215672, "cps": 31.695312, "mean_p": 0.265775, "boundary_entropy": 0.324502, "bpb": 2.983161}}, "bpb": {"en": 2.43676, "tr": 2.76739, "fi": 2.98316}, "gini_cpc": 0.026258, "gini_cps": 0.04376, "bpic": 5.200953, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0099600835770928, "fi": 1.1234164239748976}, "inventory": {"size": 327288, "admitted": 65536, "staged": 261752, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 307, "admitted_surface": 3923, "staged_surface": 63713, "admitted_split": 31, "staged_split": 48896, "admitted_merge": 0, "staged_merge": 105334, "admitted_skel": 61582, "staged_skel": 43809, "mean_count": 34.516326904296875, "n_util_obs": 59, "util_count_spearman": 0.04821741671537113}, "t": 1344.4} |
| {"event": "eval", "step": 800, "phase": 1, "bpb": {"en": 2.4474, "tr": 2.77547, "fi": 2.98631}, "bytes_per_chunk": {"en": 5.012, "tr": 5.927, "fi": 4.883}, "bpb_mean": 2.73639, "frozen_inventory": true, "t": 1348.1} |
| {"event": "train", "step": 800, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.86019, "bpb": 2.68369, "chunk_loss": 0.030918, "F": 0.20352, "G": 0.26242, "chunks": 208.41, "bytes_per_chunk": 4.913, "boundary_entropy": 0.29904, "router_noise": 0.0, "grad_norm": 1.261, "lr": 0.0002949, "router_trainable": true, "sec_per_step": 1.563, "probe_sec": 0.1184, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06585, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.00266, "n_shadow": 5, "mean_delta_shadow": -0.10751, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.01855, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -0.08134, "trunc_abs_frac": 0.93381, "mean_delta_full": 0.14819, "read_hit_rate": {"surface": 0.16197, "split": 0.13817, "merge": 0.0, "skel": 0.30801}, "mean_gate": {"surface": 0.16169, "split": 0.13825, "merge": null, "skel": 0.14223}, "n_reads": 4165, "n_cand": 29588, "read_suppressed_frac": 0.0982, "dropout_reads_logged": 4165, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 306, "admitted_surface": 3923, "staged_surface": 63792, "admitted_split": 31, "staged_split": 49047, "admitted_merge": 0, "staged_merge": 105409, "admitted_skel": 61582, "staged_skel": 43896, "mean_count": 34.57989501953125, "n_util_obs": 59, "util_count_spearman": 0.04541203974284044}, "probe_overhead_sec_total": 106.242, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.005405, "alpha": 0.03, "t": 1355.1} |
| {"event": "consolidate", "step": 850, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262113, "t": 1434.0} |
| {"event": "train", "step": 850, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.81016, "bpb": 2.61151, "chunk_loss": 0.030608, "F": 0.19653, "G": 0.25523, "chunks": 201.25, "bytes_per_chunk": 5.088, "boundary_entropy": 0.29205, "router_noise": 0.0, "grad_norm": 0.9642, "lr": 0.0002938, "router_trainable": true, "sec_per_step": 1.603, "probe_sec": 0.0871, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05571, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": 0.14617, "n_shadow": 6, "mean_delta_shadow": 0.07983, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": -0.57123, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 1.07475, "trunc_abs_frac": 0.85149, "mean_delta_full": 0.12982, "read_hit_rate": {"surface": 0.17978, "split": 0.13701, "merge": 0.0, "skel": 0.31702}, "mean_gate": {"surface": 0.16025, "split": 0.13528, "merge": null, "skel": 0.1459}, "n_reads": 4391, "n_cand": 29607, "read_suppressed_frac": 0.102, "dropout_reads_logged": 4391, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 321, "admitted_surface": 4137, "staged_surface": 63273, "admitted_split": 29, "staged_split": 48571, "admitted_merge": 0, "staged_merge": 107155, "admitted_skel": 61370, "staged_skel": 43145, "mean_count": 38.9456787109375, "n_util_obs": 68, "util_count_spearman": 0.08451430044187988}, "probe_overhead_sec_total": 112.92, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.005361, "alpha": 0.03, "t": 1435.3} |
| {"event": "consolidate", "step": 900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262109, "t": 1516.0} |
| {"event": "train", "step": 900, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.60923, "bpb": 2.32163, "chunk_loss": 0.030719, "F": 0.20068, "G": 0.25752, "chunks": 205.5, "bytes_per_chunk": 4.983, "boundary_entropy": 0.27736, "router_noise": 0.0, "grad_norm": 1.1558, "lr": 0.0002927, "router_trainable": true, "sec_per_step": 1.642, "probe_sec": 0.2221, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.13287, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.04909, "n_shadow": 6, "mean_delta_shadow": -0.03081, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.14844, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 0.80022, "trunc_abs_frac": 2.2843, "mean_delta_full": -0.07784, "read_hit_rate": {"surface": 0.25061, "split": 0.1724, "merge": 0.0, "skel": 0.3485}, "mean_gate": {"surface": 0.1683, "split": 0.13852, "merge": null, "skel": 0.15098}, "n_reads": 5509, "n_cand": 30155, "read_suppressed_frac": 0.1002, "dropout_reads_logged": 5509, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 319, "admitted_surface": 4285, "staged_surface": 62655, "admitted_split": 55, "staged_split": 48081, "admitted_merge": 1, "staged_merge": 108872, "admitted_skel": 61195, "staged_skel": 42536, "mean_count": 45.10638427734375, "n_util_obs": 76, "util_count_spearman": -0.014299384825700616}, "probe_overhead_sec_total": 119.852, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.00544, "alpha": 0.03, "t": 1517.3} |
| {"event": "consolidate", "step": 950, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 721, "uncertain_staged": 262108, "t": 1597.1} |
| {"event": "train", "step": 950, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.68448, "bpb": 2.43018, "chunk_loss": 0.030737, "F": 0.20233, "G": 0.25404, "chunks": 207.19, "bytes_per_chunk": 4.942, "boundary_entropy": 0.26956, "router_noise": 0.0, "grad_norm": 1.1237, "lr": 0.0002914, "router_trainable": true, "sec_per_step": 1.635, "probe_sec": 0.1677, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10607, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.00284, "n_shadow": 4, "mean_delta_shadow": -0.03952, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": 0.2017, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": -5.34783, "trunc_abs_frac": 5.34783, "mean_delta_full": 0.00281, "read_hit_rate": {"surface": 0.27342, "split": 0.17906, "merge": 0.0, "skel": 0.41286}, "mean_gate": {"surface": 0.16167, "split": 0.1357, "merge": null, "skel": 0.14948}, "n_reads": 6000, "n_cand": 29937, "read_suppressed_frac": 0.1012, "dropout_reads_logged": 6000, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 342, "admitted_surface": 4312, "staged_surface": 62330, "admitted_split": 66, "staged_split": 47227, "admitted_merge": 1, "staged_merge": 110566, "admitted_skel": 61157, "staged_skel": 42021, "mean_count": 55.26673889160156, "n_util_obs": 82, "util_count_spearman": 0.04096625508193504}, "probe_overhead_sec_total": 126.215, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.005507, "alpha": 0.03, "t": 1599.1} |
| {"event": "consolidate", "step": 1000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 420, "uncertain_staged": 262104, "t": 1675.4} |
| {"event": "probe", "step": 1000, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3383.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.194392, "cps": 26.429688, "mean_p": 0.243673, "boundary_entropy": 0.271493, "bpb": 2.336966}, "tr": {"chunks": 3408.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.188256, "cps": 26.625, "mean_p": 0.259257, "boundary_entropy": 0.324041, "bpb": 2.642248}, "fi": {"chunks": 4127.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.219393, "cps": 32.242188, "mean_p": 0.257909, "boundary_entropy": 0.284083, "bpb": 2.857438}}, "bpb": {"en": 2.33697, "tr": 2.64225, "fi": 2.85744}, "gini_cpc": 0.034479, "gini_cps": 0.04543, "bpic": 5.19619, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9684365169655447, "fi": 1.1286121149767199}, "inventory": {"size": 327260, "admitted": 65536, "staged": 261724, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 345, "admitted_surface": 4195, "staged_surface": 62107, "admitted_split": 83, "staged_split": 47157, "admitted_merge": 1, "staged_merge": 110880, "admitted_skel": 61257, "staged_skel": 41580, "mean_count": 62.43830871582031, "n_util_obs": 91, "util_count_spearman": 0.05733397037744864}, "t": 1678.0} |
| {"event": "eval", "step": 1000, "phase": 1, "bpb": {"en": 2.34789, "tr": 2.65036, "fi": 2.86268}, "bytes_per_chunk": {"en": 4.959, "tr": 5.915, "fi": 4.762}, "bpb_mean": 2.62031, "frozen_inventory": true, "t": 1682.3} |
| {"event": "train", "step": 1000, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.74176, "bpb": 2.51283, "chunk_loss": 0.030672, "F": 0.20041, "G": 0.25307, "chunks": 205.22, "bytes_per_chunk": 4.99, "boundary_entropy": 0.27307, "router_noise": 0.0, "grad_norm": 0.9423, "lr": 0.0002901, "router_trainable": true, "sec_per_step": 1.885, "probe_sec": 0.2197, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.1757, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": -0.10673, "n_shadow": 7, "mean_delta_shadow": 0.0732, "n_split": 1, "mean_delta_split": 0.23364, "n_merge": 1, "mean_delta_merge": 0.24678, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.38121, "trunc_abs_frac": 0.47884, "mean_delta_full": 0.37817, "read_hit_rate": {"surface": 0.25233, "split": 0.1837, "merge": 0.0, "skel": 0.43098}, "mean_gate": {"surface": 0.15386, "split": 0.13234, "merge": null, "skel": 0.13969}, "n_reads": 6038, "n_cand": 29890, "read_suppressed_frac": 0.0979, "dropout_reads_logged": 6038, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 347, "admitted_surface": 4195, "staged_surface": 62241, "admitted_split": 83, "staged_split": 47210, "admitted_merge": 1, "staged_merge": 111034, "admitted_skel": 61257, "staged_skel": 41659, "mean_count": 62.53044128417969, "n_util_obs": 91, "util_count_spearman": 0.047396082178690874}, "probe_overhead_sec_total": 132.445, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.005599, "alpha": 0.03, "t": 1693.4} |
| {"event": "probe", "step": 1024, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3284.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.188703, "cps": 25.65625, "mean_p": 0.239703, "boundary_entropy": 0.272783, "bpb": 2.331469}, "tr": {"chunks": 3538.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.195437, "cps": 27.640625, "mean_p": 0.260526, "boundary_entropy": 0.324641, "bpb": 2.643004}, "fi": {"chunks": 4109.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.218436, "cps": 32.101562, "mean_p": 0.261011, "boundary_entropy": 0.289226, "bpb": 2.838052}}, "bpb": {"en": 2.33147, "tr": 2.643, "fi": 2.83805}, "gini_cpc": 0.032895, "gini_cps": 0.050316, "bpic": 5.19001, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0356863415818658, "fi": 1.1575645803168917}, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 358, "admitted_surface": 4195, "staged_surface": 62246, "admitted_split": 83, "staged_split": 47266, "admitted_merge": 1, "staged_merge": 110629, "admitted_skel": 61257, "staged_skel": 42003, "mean_count": 64.63471984863281, "n_util_obs": 96, "util_count_spearman": 0.029679970428719324}, "t": 1730.3} |
| {"event": "consolidate", "step": 1050, "retired": 1, "split_retired": 4, "merge_admitted": 0, "admitted": 5, "swapped": 184, "uncertain_staged": 262105, "t": 1781.8} |
| {"event": "train", "step": 1050, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.63998, "bpb": 2.36599, "chunk_loss": 0.03081, "F": 0.20242, "G": 0.24805, "chunks": 207.28, "bytes_per_chunk": 4.94, "boundary_entropy": 0.2619, "router_noise": 0.0, "grad_norm": 1.066, "lr": 0.0002886, "router_trainable": true, "sec_per_step": 1.795, "probe_sec": 0.1885, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.1322, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": 0.07976, "n_shadow": 3, "mean_delta_shadow": 0.0399, "n_split": 2, "mean_delta_split": -0.01154, "n_merge": 1, "mean_delta_merge": -0.1189, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.18493, "trunc_abs_frac": 0.23926, "mean_delta_full": -0.14703, "read_hit_rate": {"surface": 0.26405, "split": 0.22859, "merge": 0.0, "skel": 0.45206}, "mean_gate": {"surface": 0.15749, "split": 0.12964, "merge": null, "skel": 0.13976}, "n_reads": 6697, "n_cand": 29866, "read_suppressed_frac": 0.098, "dropout_reads_logged": 6697, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 358, "admitted_surface": 4020, "staged_surface": 62067, "admitted_split": 103, "staged_split": 46605, "admitted_merge": 1, "staged_merge": 111819, "admitted_skel": 61412, "staged_skel": 41653, "mean_count": 68.280517578125, "n_util_obs": 98, "util_count_spearman": -0.050041759909212044}, "probe_overhead_sec_total": 139.268, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.005733, "alpha": 0.03, "t": 1783.1} |
| {"event": "consolidate", "step": 1100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 552, "uncertain_staged": 262106, "t": 1862.2} |
| {"event": "train", "step": 1100, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.74825, "bpb": 2.52219, "chunk_loss": 0.030779, "F": 0.20407, "G": 0.25243, "chunks": 208.97, "bytes_per_chunk": 4.9, "boundary_entropy": 0.27066, "router_noise": 0.0, "grad_norm": 1.0187, "lr": 0.0002871, "router_trainable": true, "sec_per_step": 1.632, "probe_sec": 0.2167, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.13427, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.05954, "n_shadow": 3, "mean_delta_shadow": 0.13442, "n_split": 2, "mean_delta_split": -0.02881, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.25133, "trunc_abs_frac": 13.02298, "mean_delta_full": 0.03052, "read_hit_rate": {"surface": 0.26867, "split": 0.22185, "merge": 0.0003, "skel": 0.45152}, "mean_gate": {"surface": 0.15974, "split": 0.13178, "merge": 0.21125, "skel": 0.14084}, "n_reads": 6629, "n_cand": 29922, "read_suppressed_frac": 0.1089, "dropout_reads_logged": 6629, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 370, "admitted_surface": 3486, "staged_surface": 61784, "admitted_split": 140, "staged_split": 47456, "admitted_merge": 21, "staged_merge": 111727, "admitted_skel": 61889, "staged_skel": 41177, "mean_count": 75.20069885253906, "n_util_obs": 104, "util_count_spearman": -0.08874426544329457}, "probe_overhead_sec_total": 145.981, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.005923, "alpha": 0.03, "t": 1864.7} |
| {"event": "consolidate", "step": 1150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 749, "uncertain_staged": 262103, "t": 1941.2} |
| {"event": "train", "step": 1150, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.64624, "bpb": 2.37502, "chunk_loss": 0.030608, "F": 0.19864, "G": 0.24709, "chunks": 203.41, "bytes_per_chunk": 5.034, "boundary_entropy": 0.25983, "router_noise": 0.0, "grad_norm": 0.9099, "lr": 0.0002855, "router_trainable": true, "sec_per_step": 1.565, "probe_sec": 0.1313, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08261, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": 0.09981, "n_shadow": 2, "mean_delta_shadow": 0.03961, "n_split": 3, "mean_delta_split": -0.0189, "n_merge": 1, "mean_delta_merge": 0.00098, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.75924, "trunc_abs_frac": 1.8442, "mean_delta_full": 0.16852, "read_hit_rate": {"surface": 0.28208, "split": 0.23426, "merge": 0.00729, "skel": 0.51346}, "mean_gate": {"surface": 0.1605, "split": 0.13136, "merge": 0.14417, "skel": 0.13829}, "n_reads": 7265, "n_cand": 29779, "read_suppressed_frac": 0.1107, "dropout_reads_logged": 7265, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 379, "admitted_surface": 2745, "staged_surface": 61895, "admitted_split": 166, "staged_split": 46835, "admitted_merge": 483, "staged_merge": 112685, "admitted_skel": 62142, "staged_skel": 40729, "mean_count": 82.92973327636719, "n_util_obs": 115, "util_count_spearman": -0.05008285331018701}, "probe_overhead_sec_total": 152.385, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.006026, "alpha": 0.03, "t": 1943.0} |
| {"event": "consolidate", "step": 1200, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262103, "t": 2023.5} |
| {"event": "probe", "step": 1200, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3449.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.198184, "cps": 26.945312, "mean_p": 0.238894, "boundary_entropy": 0.247434, "bpb": 2.275269}, "tr": {"chunks": 3412.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.188477, "cps": 26.65625, "mean_p": 0.248361, "boundary_entropy": 0.295659, "bpb": 2.566708}, "fi": {"chunks": 4282.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.227633, "cps": 33.453125, "mean_p": 0.263627, "boundary_entropy": 0.263219, "bpb": 2.739822}}, "bpb": {"en": 2.27527, "tr": 2.56671, "fi": 2.73982}, "gini_cpc": 0.042494, "gini_cps": 0.052051, "bpic": 5.091268, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.951019445171886, "fi": 1.1485917838706214}, "inventory": {"size": 326655, "admitted": 65536, "staged": 261119, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 398, "admitted_surface": 1720, "staged_surface": 61748, "admitted_split": 168, "staged_split": 47161, "admitted_merge": 1506, "staged_merge": 111028, "admitted_skel": 62142, "staged_skel": 41182, "mean_count": 88.25614929199219, "n_util_obs": 119, "util_count_spearman": -0.042273180458624125}, "t": 2026.2} |
| {"event": "eval", "step": 1200, "phase": 1, "bpb": {"en": 2.28649, "tr": 2.57372, "fi": 2.74813}, "bytes_per_chunk": {"en": 4.939, "tr": 5.968, "fi": 4.699}, "bpb_mean": 2.53611, "frozen_inventory": true, "t": 2030.5} |
| {"event": "train", "step": 1200, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.6247, "bpb": 2.34394, "chunk_loss": 0.030598, "F": 0.19815, "G": 0.24948, "chunks": 202.91, "bytes_per_chunk": 5.047, "boundary_entropy": 0.2549, "router_noise": 0.0, "grad_norm": 0.9039, "lr": 0.0002837, "router_trainable": true, "sec_per_step": 1.922, "probe_sec": 0.139, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.102, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": 0.03825, "n_shadow": 5, "mean_delta_shadow": 0.00374, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -0.60939, "trunc_abs_frac": 1.08931, "mean_delta_full": 0.03292, "read_hit_rate": {"surface": 0.28989, "split": 0.22502, "merge": 0.03126, "skel": 0.51094}, "mean_gate": {"surface": 0.15878, "split": 0.127, "merge": 0.14957, "skel": 0.1387}, "n_reads": 7330, "n_cand": 29685, "read_suppressed_frac": 0.0954, "dropout_reads_logged": 7330, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 398, "admitted_surface": 1720, "staged_surface": 61956, "admitted_split": 168, "staged_split": 47295, "admitted_merge": 1506, "staged_merge": 111575, "admitted_skel": 62142, "staged_skel": 41318, "mean_count": 88.36799621582031, "n_util_obs": 119, "util_count_spearman": -0.04238000284859707}, "probe_overhead_sec_total": 159.316, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.006047, "alpha": 0.03, "t": 2039.1} |
| {"event": "consolidate", "step": 1250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262101, "t": 2119.5} |
| {"event": "train", "step": 1250, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.64248, "bpb": 2.3696, "chunk_loss": 0.030503, "F": 0.19464, "G": 0.24211, "chunks": 199.31, "bytes_per_chunk": 5.138, "boundary_entropy": 0.25521, "router_noise": 0.0, "grad_norm": 0.8878, "lr": 0.0002819, "router_trainable": true, "sec_per_step": 1.64, "probe_sec": 0.1437, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08954, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.01381, "n_shadow": 6, "mean_delta_shadow": 0.04169, "n_split": 1, "mean_delta_split": 0.06155, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -0.45041, "trunc_abs_frac": 2.61682, "mean_delta_full": -0.07163, "read_hit_rate": {"surface": 0.26883, "split": 0.2151, "merge": 0.05226, "skel": 0.49017}, "mean_gate": {"surface": 0.15473, "split": 0.1227, "merge": 0.13932, "skel": 0.13377}, "n_reads": 6877, "n_cand": 28855, "read_suppressed_frac": 0.0982, "dropout_reads_logged": 6877, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 404, "admitted_surface": 1341, "staged_surface": 62040, "admitted_split": 169, "staged_split": 47457, "admitted_merge": 2529, "staged_merge": 111341, "admitted_skel": 61497, "staged_skel": 41306, "mean_count": 94.44017028808594, "n_util_obs": 128, "util_count_spearman": -0.08937821795785153}, "probe_overhead_sec_total": 165.973, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.005874, "alpha": 0.03, "t": 2121.1} |
| {"event": "consolidate", "step": 1300, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262101, "t": 2200.0} |
| {"event": "train", "step": 1300, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.67634, "bpb": 2.41844, "chunk_loss": 0.030815, "F": 0.20767, "G": 0.25047, "chunks": 212.66, "bytes_per_chunk": 4.815, "boundary_entropy": 0.24321, "router_noise": 0.0, "grad_norm": 0.9443, "lr": 0.00028, "router_trainable": true, "sec_per_step": 1.634, "probe_sec": 0.1099, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05752, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.02595, "n_shadow": 4, "mean_delta_shadow": 0.06792, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.30652, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.36813, "trunc_abs_frac": 1.36839, "mean_delta_full": 0.08223, "read_hit_rate": {"surface": 0.26871, "split": 0.23195, "merge": 0.07061, "skel": 0.49559}, "mean_gate": {"surface": 0.15776, "split": 0.1203, "merge": 0.14454, "skel": 0.13072}, "n_reads": 7717, "n_cand": 30833, "read_suppressed_frac": 0.0973, "dropout_reads_logged": 7717, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 416, "admitted_surface": 1312, "staged_surface": 62463, "admitted_split": 185, "staged_split": 47111, "admitted_merge": 3529, "staged_merge": 111610, "admitted_skel": 60510, "staged_skel": 40960, "mean_count": 100.13107299804688, "n_util_obs": 140, "util_count_spearman": -0.10102993592967571}, "probe_overhead_sec_total": 172.749, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.005788, "alpha": 0.03, "t": 2202.7} |
| {"event": "consolidate", "step": 1350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262099, "t": 2284.2} |
| {"event": "train", "step": 1350, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.60516, "bpb": 2.31576, "chunk_loss": 0.030543, "F": 0.19406, "G": 0.24381, "chunks": 198.72, "bytes_per_chunk": 5.153, "boundary_entropy": 0.25025, "router_noise": 0.0, "grad_norm": 0.9941, "lr": 0.000278, "router_trainable": true, "sec_per_step": 1.648, "probe_sec": 0.0967, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06576, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.03107, "n_shadow": 4, "mean_delta_shadow": -0.04468, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.11256, "trunc_abs_frac": 0.09928, "mean_delta_full": -0.24725, "read_hit_rate": {"surface": 0.27296, "split": 0.23715, "merge": 0.08403, "skel": 0.47981}, "mean_gate": {"surface": 0.1591, "split": 0.12094, "merge": 0.1437, "skel": 0.1343}, "n_reads": 7322, "n_cand": 28866, "read_suppressed_frac": 0.105, "dropout_reads_logged": 7322, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 455, "admitted_surface": 1292, "staged_surface": 62632, "admitted_split": 364, "staged_split": 47561, "admitted_merge": 4374, "staged_merge": 110273, "admitted_skel": 59506, "staged_skel": 41678, "mean_count": 105.93804931640625, "n_util_obs": 149, "util_count_spearman": -0.09275530564121168}, "probe_overhead_sec_total": 179.579, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.005647, "alpha": 0.03, "t": 2285.1} |
| {"event": "consolidate", "step": 1400, "retired": 1, "split_retired": 0, "merge_admitted": 0, "admitted": 1, "swapped": 513, "uncertain_staged": 262096, "t": 2372.6} |
| {"event": "probe", "step": 1400, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3395.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.195081, "cps": 26.523438, "mean_p": 0.234672, "boundary_entropy": 0.237883, "bpb": 2.220228}, "tr": {"chunks": 3511.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.193946, "cps": 27.429688, "mean_p": 0.243619, "boundary_entropy": 0.285144, "bpb": 2.500251}, "fi": {"chunks": 4329.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.230131, "cps": 33.820312, "mean_p": 0.255188, "boundary_entropy": 0.239639, "bpb": 2.657452}}, "bpb": {"en": 2.22023, "tr": 2.50025, "fi": 2.65745}, "gini_cpc": 0.038962, "gini_cps": 0.055422, "bpic": 5.049577, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9941790785292831, "fi": 1.1796686659616717}, "inventory": {"size": 327166, "admitted": 65536, "staged": 261630, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 486, "admitted_surface": 1286, "staged_surface": 62616, "admitted_split": 770, "staged_split": 47338, "admitted_merge": 4399, "staged_merge": 109858, "admitted_skel": 59081, "staged_skel": 41818, "mean_count": 118.9541015625, "n_util_obs": 164, "util_count_spearman": -0.09034864581774651}, "t": 2374.5} |
| {"event": "eval", "step": 1400, "phase": 1, "bpb": {"en": 2.22974, "tr": 2.50619, "fi": 2.66293}, "bytes_per_chunk": {"en": 5.059, "tr": 5.66, "fi": 4.61}, "bpb_mean": 2.46629, "frozen_inventory": true, "t": 2379.1} |
| {"event": "train", "step": 1400, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.60521, "bpb": 2.31584, "chunk_loss": 0.030563, "F": 0.19485, "G": 0.24292, "chunks": 199.53, "bytes_per_chunk": 5.132, "boundary_entropy": 0.2453, "router_noise": 0.0, "grad_norm": 0.8989, "lr": 0.0002759, "router_trainable": true, "sec_per_step": 2.064, "probe_sec": 0.1265, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09015, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": 0.19705, "n_shadow": 1, "mean_delta_shadow": -0.06921, "n_split": 4, "mean_delta_split": -0.01178, "n_merge": 1, "mean_delta_merge": -0.08582, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 2.2674, "trunc_abs_frac": 0.65982, "mean_delta_full": 0.06702, "read_hit_rate": {"surface": 0.27546, "split": 0.28361, "merge": 0.0878, "skel": 0.51567}, "mean_gate": {"surface": 0.16083, "split": 0.12193, "merge": 0.14236, "skel": 0.13336}, "n_reads": 8033, "n_cand": 28823, "read_suppressed_frac": 0.1007, "dropout_reads_logged": 8033, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 486, "admitted_surface": 1286, "staged_surface": 62632, "admitted_split": 770, "staged_split": 47450, "admitted_merge": 4399, "staged_merge": 110177, "admitted_skel": 59081, "staged_skel": 41885, "mean_count": 119.07667541503906, "n_util_obs": 164, "util_count_spearman": -0.09083019329914437}, "probe_overhead_sec_total": 187.909, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.005962, "alpha": 0.03, "t": 2388.3} |
| {"event": "consolidate", "step": 1450, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 61, "uncertain_staged": 262102, "t": 2471.9} |
| {"event": "train", "step": 1450, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.65613, "bpb": 2.38928, "chunk_loss": 0.030435, "F": 0.19183, "G": 0.23956, "chunks": 196.44, "bytes_per_chunk": 5.213, "boundary_entropy": 0.24426, "router_noise": 0.0, "grad_norm": 0.8844, "lr": 0.0002738, "router_trainable": true, "sec_per_step": 1.701, "probe_sec": 0.1784, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0965, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": -0.01774, "n_shadow": 4, "mean_delta_shadow": -0.00257, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.16199, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -0.40342, "trunc_abs_frac": 1.22015, "mean_delta_full": 0.06421, "read_hit_rate": {"surface": 0.24976, "split": 0.29242, "merge": 0.07763, "skel": 0.52929}, "mean_gate": {"surface": 0.15878, "split": 0.11837, "merge": 0.13809, "skel": 0.12934}, "n_reads": 7845, "n_cand": 28356, "read_suppressed_frac": 0.1017, "dropout_reads_logged": 7845, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 490, "admitted_surface": 1268, "staged_surface": 62513, "admitted_split": 766, "staged_split": 46970, "admitted_merge": 4398, "staged_merge": 110449, "admitted_skel": 59104, "staged_skel": 42212, "mean_count": 129.24476623535156, "n_util_obs": 169, "util_count_spearman": -0.09085456395417445}, "probe_overhead_sec_total": 195.042, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.005905, "alpha": 0.03, "t": 2473.4} |
| {"event": "consolidate", "step": 1500, "retired": 0, "split_retired": 3, "merge_admitted": 0, "admitted": 3, "swapped": 55, "uncertain_staged": 262100, "t": 2554.2} |
| {"event": "train", "step": 1500, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.6713, "bpb": 2.41117, "chunk_loss": 0.030829, "F": 0.20605, "G": 0.24913, "chunks": 211.0, "bytes_per_chunk": 4.853, "boundary_entropy": 0.23957, "router_noise": 0.0, "grad_norm": 0.8473, "lr": 0.0002715, "router_trainable": true, "sec_per_step": 1.642, "probe_sec": 0.0939, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05312, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": -0.08382, "n_shadow": 8, "mean_delta_shadow": -0.01396, "n_split": 1, "mean_delta_split": -0.08167, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.72116, "trunc_abs_frac": 0.68307, "mean_delta_full": -0.2132, "read_hit_rate": {"surface": 0.2558, "split": 0.30286, "merge": 0.07401, "skel": 0.56059}, "mean_gate": {"surface": 0.14854, "split": 0.11003, "merge": 0.12813, "skel": 0.12121}, "n_reads": 8494, "n_cand": 29916, "read_suppressed_frac": 0.1044, "dropout_reads_logged": 8494, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 496, "admitted_surface": 1267, "staged_surface": 62689, "admitted_split": 775, "staged_split": 47290, "admitted_merge": 4398, "staged_merge": 110039, "admitted_skel": 59096, "staged_skel": 42126, "mean_count": 136.70704650878906, "n_util_obs": 176, "util_count_spearman": -0.09682808716707021}, "probe_overhead_sec_total": 202.097, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.005753, "alpha": 0.03, "t": 2555.4} |
| {"event": "consolidate", "step": 1550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 31, "uncertain_staged": 262102, "t": 2635.8} |
| {"event": "train", "step": 1550, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.5948, "bpb": 2.3008, "chunk_loss": 0.030655, "F": 0.19839, "G": 0.24305, "chunks": 203.16, "bytes_per_chunk": 5.04, "boundary_entropy": 0.23472, "router_noise": 0.0, "grad_norm": 0.9263, "lr": 0.0002692, "router_trainable": true, "sec_per_step": 1.638, "probe_sec": 0.0982, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0591, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": -0.02592, "n_shadow": 5, "mean_delta_shadow": -0.02006, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": 0.15904, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.62842, "trunc_abs_frac": 0.54614, "mean_delta_full": -0.0748, "read_hit_rate": {"surface": 0.27361, "split": 0.31101, "merge": 0.08451, "skel": 0.56275}, "mean_gate": {"surface": 0.15595, "split": 0.11231, "merge": 0.13514, "skel": 0.12755}, "n_reads": 8678, "n_cand": 29292, "read_suppressed_frac": 0.0996, "dropout_reads_logged": 8678, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 508, "admitted_surface": 1269, "staged_surface": 62402, "admitted_split": 784, "staged_split": 47227, "admitted_merge": 4399, "staged_merge": 110492, "admitted_skel": 59084, "staged_skel": 42023, "mean_count": 144.98280334472656, "n_util_obs": 186, "util_count_spearman": -0.0943516916951452}, "probe_overhead_sec_total": 210.093, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.005749, "alpha": 0.03, "t": 2637.3} |
| {"event": "consolidate", "step": 1600, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 36, "uncertain_staged": 262100, "t": 2711.1} |
| {"event": "probe", "step": 1600, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3240.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.186175, "cps": 25.3125, "mean_p": 0.227946, "boundary_entropy": 0.22366, "bpb": 2.175724}, "tr": {"chunks": 3408.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.188256, "cps": 26.625, "mean_p": 0.236864, "boundary_entropy": 0.268184, "bpb": 2.432851}, "fi": {"chunks": 4263.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.226623, "cps": 33.304688, "mean_p": 0.250809, "boundary_entropy": 0.226594, "bpb": 2.593194}}, "bpb": {"en": 2.17572, "tr": 2.43285, "fi": 2.59319}, "gini_cpc": 0.044863, "gini_cps": 0.062506, "bpic": 5.199523, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.011179239782234, "fi": 1.2172577806129983}, "inventory": {"size": 327642, "admitted": 65536, "staged": 262106, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 537, "admitted_surface": 1279, "staged_surface": 62707, "admitted_split": 799, "staged_split": 47585, "admitted_merge": 4402, "staged_merge": 109760, "admitted_skel": 59056, "staged_skel": 42054, "mean_count": 153.9337158203125, "n_util_obs": 191, "util_count_spearman": -0.11010527612001404}, "t": 2713.1} |
| {"event": "eval", "step": 1600, "phase": 1, "bpb": {"en": 2.18902, "tr": 2.44065, "fi": 2.59951}, "bytes_per_chunk": {"en": 5.172, "tr": 5.675, "fi": 4.675}, "bpb_mean": 2.40973, "frozen_inventory": true, "t": 2716.6} |
| {"event": "train", "step": 1600, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.51879, "bpb": 2.19115, "chunk_loss": 0.030519, "F": 0.1965, "G": 0.24042, "chunks": 201.22, "bytes_per_chunk": 5.089, "boundary_entropy": 0.22714, "router_noise": 0.0, "grad_norm": 0.9179, "lr": 0.0002668, "router_trainable": true, "sec_per_step": 1.729, "probe_sec": 0.118, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07172, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.02328, "n_shadow": 3, "mean_delta_shadow": -0.16512, "n_split": 1, "mean_delta_split": 0.11786, "n_merge": 1, "mean_delta_merge": -0.19781, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 1.52266, "trunc_abs_frac": 1.51047, "mean_delta_full": -0.08957, "read_hit_rate": {"surface": 0.2953, "split": 0.33166, "merge": 0.08643, "skel": 0.58438}, "mean_gate": {"surface": 0.155, "split": 0.10737, "merge": 0.13418, "skel": 0.12392}, "n_reads": 9248, "n_cand": 29426, "read_suppressed_frac": 0.0973, "dropout_reads_logged": 9248, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 539, "admitted_surface": 1279, "staged_surface": 62698, "admitted_split": 799, "staged_split": 47472, "admitted_merge": 4402, "staged_merge": 110016, "admitted_skel": 59056, "staged_skel": 41958, "mean_count": 154.0748291015625, "n_util_obs": 191, "util_count_spearman": -0.10681063430467126}, "probe_overhead_sec_total": 217.584, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.005673, "alpha": 0.03, "t": 2723.8} |
| {"event": "consolidate", "step": 1650, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 28, "uncertain_staged": 262101, "t": 2803.6} |
| {"event": "train", "step": 1650, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.59084, "bpb": 2.29509, "chunk_loss": 0.030452, "F": 0.19431, "G": 0.23493, "chunks": 198.97, "bytes_per_chunk": 5.147, "boundary_entropy": 0.22456, "router_noise": 0.0, "grad_norm": 0.9258, "lr": 0.0002643, "router_trainable": true, "sec_per_step": 1.613, "probe_sec": 0.0878, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06253, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": -0.12725, "n_shadow": 4, "mean_delta_shadow": 0.01899, "n_split": 1, "mean_delta_split": -0.07446, "n_merge": 1, "mean_delta_merge": -0.25192, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 1.6178, "trunc_abs_frac": 1.6178, "mean_delta_full": -0.02332, "read_hit_rate": {"surface": 0.2865, "split": 0.32178, "merge": 0.08345, "skel": 0.55588}, "mean_gate": {"surface": 0.15002, "split": 0.10183, "merge": 0.1274, "skel": 0.12143}, "n_reads": 8801, "n_cand": 29023, "read_suppressed_frac": 0.1023, "dropout_reads_logged": 8801, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 561, "admitted_surface": 1277, "staged_surface": 62648, "admitted_split": 807, "staged_split": 47196, "admitted_merge": 4405, "staged_merge": 110688, "admitted_skel": 59047, "staged_skel": 41612, "mean_count": 161.58010864257812, "n_util_obs": 195, "util_count_spearman": -0.10918776367467743}, "probe_overhead_sec_total": 224.564, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.005703, "alpha": 0.03, "t": 2804.4} |
| {"event": "consolidate", "step": 1700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 76, "uncertain_staged": 262102, "t": 2885.7} |
| {"event": "train", "step": 1700, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.54405, "bpb": 2.22759, "chunk_loss": 0.030423, "F": 0.19272, "G": 0.23234, "chunks": 197.34, "bytes_per_chunk": 5.189, "boundary_entropy": 0.22479, "router_noise": 0.0, "grad_norm": 0.7954, "lr": 0.0002618, "router_trainable": true, "sec_per_step": 1.651, "probe_sec": 0.0819, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04981, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": 0.0675, "n_shadow": 2, "mean_delta_shadow": 0.64154, "n_split": 3, "mean_delta_split": 0.08946, "n_merge": 1, "mean_delta_merge": -0.28375, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -2.59775, "trunc_abs_frac": 0.34776, "mean_delta_full": -0.02763, "read_hit_rate": {"surface": 0.31148, "split": 0.34702, "merge": 0.07487, "skel": 0.56017}, "mean_gate": {"surface": 0.15074, "split": 0.1045, "merge": 0.12898, "skel": 0.12428}, "n_reads": 9051, "n_cand": 28596, "read_suppressed_frac": 0.1051, "dropout_reads_logged": 9051, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 576, "admitted_surface": 1293, "staged_surface": 62558, "admitted_split": 843, "staged_split": 47324, "admitted_merge": 4407, "staged_merge": 110360, "admitted_skel": 58993, "staged_skel": 41902, "mean_count": 172.0213165283203, "n_util_obs": 210, "util_count_spearman": -0.10151768909563871}, "probe_overhead_sec_total": 231.19, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.005871, "alpha": 0.03, "t": 2886.9} |
| {"event": "consolidate", "step": 1750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 23, "uncertain_staged": 262105, "t": 2970.4} |
| {"event": "train", "step": 1750, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.5731, "bpb": 2.26951, "chunk_loss": 0.030515, "F": 0.19608, "G": 0.2379, "chunks": 200.78, "bytes_per_chunk": 5.1, "boundary_entropy": 0.22687, "router_noise": 0.0, "grad_norm": 0.8235, "lr": 0.0002591, "router_trainable": true, "sec_per_step": 1.705, "probe_sec": 0.1098, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05857, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.14476, "n_shadow": 4, "mean_delta_shadow": -0.03424, "n_split": 1, "mean_delta_split": -0.02554, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -3.54, "trunc_abs_frac": 30.2224, "mean_delta_full": -0.01221, "read_hit_rate": {"surface": 0.30471, "split": 0.34372, "merge": 0.07923, "skel": 0.56106}, "mean_gate": {"surface": 0.15712, "split": 0.10261, "merge": 0.13652, "skel": 0.12559}, "n_reads": 9154, "n_cand": 29089, "read_suppressed_frac": 0.101, "dropout_reads_logged": 9154, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 603, "admitted_surface": 1296, "staged_surface": 62707, "admitted_split": 846, "staged_split": 47278, "admitted_merge": 4407, "staged_merge": 110406, "admitted_skel": 58987, "staged_skel": 41753, "mean_count": 179.11578369140625, "n_util_obs": 223, "util_count_spearman": -0.09244011884926592}, "probe_overhead_sec_total": 238.614, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.00577, "alpha": 0.03, "t": 2972.2} |
| {"event": "consolidate", "step": 1800, "retired": 0, "split_retired": 3, "merge_admitted": 0, "admitted": 3, "swapped": 11, "uncertain_staged": 262105, "t": 3053.2} |
| {"event": "probe", "step": 1800, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3267.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.187726, "cps": 25.523438, "mean_p": 0.223741, "boundary_entropy": 0.213832, "bpb": 2.141554}, "tr": {"chunks": 3282.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.181296, "cps": 25.640625, "mean_p": 0.224494, "boundary_entropy": 0.254876, "bpb": 2.369386}, "fi": {"chunks": 4256.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.226251, "cps": 33.25, "mean_p": 0.248275, "boundary_entropy": 0.220329, "bpb": 2.540944}}, "bpb": {"en": 2.14155, "tr": 2.36939, "fi": 2.54094}, "gini_cpc": 0.050346, "gini_cps": 0.061021, "bpic": 5.250532, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9657462067760714, "fi": 1.2052155365593828}, "inventory": {"size": 327666, "admitted": 65536, "staged": 262130, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 627, "admitted_surface": 1293, "staged_surface": 62523, "admitted_split": 849, "staged_split": 46680, "admitted_merge": 4408, "staged_merge": 111539, "admitted_skel": 58986, "staged_skel": 41388, "mean_count": 190.9822540283203, "n_util_obs": 225, "util_count_spearman": -0.07768857985672145}, "t": 3054.8} |
| {"event": "eval", "step": 1800, "phase": 1, "bpb": {"en": 2.15839, "tr": 2.37768, "fi": 2.54727}, "bytes_per_chunk": {"en": 5.241, "tr": 5.993, "fi": 4.624}, "bpb_mean": 2.36111, "frozen_inventory": true, "t": 3058.4} |
| {"event": "train", "step": 1800, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.5499, "bpb": 2.23603, "chunk_loss": 0.030335, "F": 0.18607, "G": 0.22935, "chunks": 190.53, "bytes_per_chunk": 5.374, "boundary_entropy": 0.22313, "router_noise": 0.0, "grad_norm": 0.7962, "lr": 0.0002564, "router_trainable": true, "sec_per_step": 1.861, "probe_sec": 0.0978, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05567, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.02129, "n_shadow": 7, "mean_delta_shadow": 0.02589, "n_split": 1, "mean_delta_split": -0.04834, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.07207, "trunc_abs_frac": 0.31336, "mean_delta_full": 0.20917, "read_hit_rate": {"surface": 0.29827, "split": 0.35426, "merge": 0.0746, "skel": 0.54561}, "mean_gate": {"surface": 0.15342, "split": 0.09878, "merge": 0.12854, "skel": 0.12427}, "n_reads": 8686, "n_cand": 27681, "read_suppressed_frac": 0.0946, "dropout_reads_logged": 8686, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 627, "admitted_surface": 1293, "staged_surface": 62581, "admitted_split": 849, "staged_split": 46764, "admitted_merge": 4408, "staged_merge": 111379, "admitted_skel": 58986, "staged_skel": 41420, "mean_count": 191.1147918701172, "n_util_obs": 225, "util_count_spearman": -0.07685840707964602}, "probe_overhead_sec_total": 245.55, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.005738, "alpha": 0.03, "t": 3065.3} |
| {"event": "consolidate", "step": 1850, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 4, "uncertain_staged": 262106, "t": 3150.9} |
| {"event": "train", "step": 1850, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.55661, "bpb": 2.24572, "chunk_loss": 0.030517, "F": 0.19461, "G": 0.23276, "chunks": 199.28, "bytes_per_chunk": 5.138, "boundary_entropy": 0.20843, "router_noise": 0.0, "grad_norm": 0.7235, "lr": 0.0002537, "router_trainable": true, "sec_per_step": 1.731, "probe_sec": 0.0804, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04696, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.00231, "n_shadow": 5, "mean_delta_shadow": -0.00432, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": -0.00135, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.6611, "trunc_abs_frac": 0.66292, "mean_delta_full": -0.38013, "read_hit_rate": {"surface": 0.29835, "split": 0.37439, "merge": 0.0602, "skel": 0.56425}, "mean_gate": {"surface": 0.15297, "split": 0.09492, "merge": 0.12868, "skel": 0.12171}, "n_reads": 9234, "n_cand": 28736, "read_suppressed_frac": 0.0969, "dropout_reads_logged": 9234, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 626, "admitted_surface": 1292, "staged_surface": 62795, "admitted_split": 853, "staged_split": 47368, "admitted_merge": 4408, "staged_merge": 110327, "admitted_skel": 58983, "staged_skel": 41654, "mean_count": 199.34884643554688, "n_util_obs": 236, "util_count_spearman": -0.12341521087155984}, "probe_overhead_sec_total": 252.39, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.005963, "alpha": 0.03, "t": 3151.8} |
| {"event": "consolidate", "step": 1900, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 11, "uncertain_staged": 262107, "t": 3227.4} |
| {"event": "train", "step": 1900, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.53712, "bpb": 2.2176, "chunk_loss": 0.030595, "F": 0.19382, "G": 0.23316, "chunks": 198.47, "bytes_per_chunk": 5.16, "boundary_entropy": 0.21221, "router_noise": 0.0, "grad_norm": 0.8366, "lr": 0.0002508, "router_trainable": true, "sec_per_step": 1.568, "probe_sec": 0.1416, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05569, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": -0.01861, "n_shadow": 6, "mean_delta_shadow": -0.09936, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": -0.05579, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 1.21343, "trunc_abs_frac": 0.93084, "mean_delta_full": -0.08541, "read_hit_rate": {"surface": 0.3059, "split": 0.38542, "merge": 0.07158, "skel": 0.57605}, "mean_gate": {"surface": 0.15734, "split": 0.09524, "merge": 0.13084, "skel": 0.12595}, "n_reads": 9520, "n_cand": 28678, "read_suppressed_frac": 0.0993, "dropout_reads_logged": 9520, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 630, "admitted_surface": 1293, "staged_surface": 62698, "admitted_split": 863, "staged_split": 47050, "admitted_merge": 4408, "staged_merge": 110380, "admitted_skel": 58972, "staged_skel": 42016, "mean_count": 209.12548828125, "n_util_obs": 242, "util_count_spearman": -0.14474473559095713}, "probe_overhead_sec_total": 259.071, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.006039, "alpha": 0.03, "t": 3230.2} |
| {"event": "consolidate", "step": 1950, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 45, "uncertain_staged": 262108, "t": 3315.1} |
| {"event": "train", "step": 1950, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.48918, "bpb": 2.14844, "chunk_loss": 0.030472, "F": 0.19348, "G": 0.23172, "chunks": 198.12, "bytes_per_chunk": 5.168, "boundary_entropy": 0.2127, "router_noise": 0.0, "grad_norm": 0.7802, "lr": 0.0002479, "router_trainable": true, "sec_per_step": 1.712, "probe_sec": 0.1508, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.11558, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": 0.06562, "n_shadow": 6, "mean_delta_shadow": -0.04026, "n_split": 1, "mean_delta_split": 0.06152, "n_merge": 2, "mean_delta_merge": 0.04889, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 2.03697, "trunc_abs_frac": 1.70672, "mean_delta_full": 0.04788, "read_hit_rate": {"surface": 0.31531, "split": 0.38598, "merge": 0.07887, "skel": 0.57352}, "mean_gate": {"surface": 0.16182, "split": 0.09541, "merge": 0.13143, "skel": 0.13052}, "n_reads": 9488, "n_cand": 28407, "read_suppressed_frac": 0.1004, "dropout_reads_logged": 9488, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 645, "admitted_surface": 1305, "staged_surface": 62850, "admitted_split": 892, "staged_split": 47122, "admitted_merge": 4409, "staged_merge": 110464, "admitted_skel": 58930, "staged_skel": 41708, "mean_count": 216.3026123046875, "n_util_obs": 255, "util_count_spearman": -0.12390719082908754}, "probe_overhead_sec_total": 267.007, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.006091, "alpha": 0.03, "t": 3315.9} |
| {"event": "consolidate", "step": 2000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 33, "uncertain_staged": 262108, "t": 3402.1} |
| {"event": "probe", "step": 2000, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3302.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.189737, "cps": 25.796875, "mean_p": 0.223644, "boundary_entropy": 0.198602, "bpb": 2.111969}, "tr": {"chunks": 3307.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.182677, "cps": 25.835938, "mean_p": 0.224422, "boundary_entropy": 0.245905, "bpb": 2.32091}, "fi": {"chunks": 4297.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.22843, "cps": 33.570312, "mean_p": 0.245132, "boundary_entropy": 0.210624, "bpb": 2.495317}}, "bpb": {"en": 2.11197, "tr": 2.32091, "fi": 2.49532}, "gini_cpc": 0.050766, "gini_cps": 0.060823, "bpic": 5.201907, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9627880578236395, "fi": 1.20392801794097}, "inventory": {"size": 327647, "admitted": 65536, "staged": 262111, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 666, "admitted_surface": 1316, "staged_surface": 62661, "admitted_split": 898, "staged_split": 47564, "admitted_merge": 4409, "staged_merge": 110002, "admitted_skel": 58913, "staged_skel": 41884, "mean_count": 225.2870635986328, "n_util_obs": 271, "util_count_spearman": -0.11087748578212717}, "t": 3404.5} |
| {"event": "eval", "step": 2000, "phase": 1, "bpb": {"en": 2.12542, "tr": 2.32982, "fi": 2.5052}, "bytes_per_chunk": {"en": 5.128, "tr": 5.928, "fi": 4.653}, "bpb_mean": 2.32014, "frozen_inventory": true, "t": 3408.0} |
| {"event": "train", "step": 2000, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.65209, "bpb": 2.38346, "chunk_loss": 0.030777, "F": 0.20541, "G": 0.24092, "chunks": 210.34, "bytes_per_chunk": 4.868, "boundary_entropy": 0.21028, "router_noise": 0.0, "grad_norm": 0.8615, "lr": 0.000245, "router_trainable": true, "sec_per_step": 1.979, "probe_sec": 0.114, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0692, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.01953, "n_shadow": 5, "mean_delta_shadow": 0.01393, "n_split": 1, "mean_delta_split": 0.07495, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.26483, "trunc_abs_frac": 1.17229, "mean_delta_full": 0.09054, "read_hit_rate": {"surface": 0.31348, "split": 0.38639, "merge": 0.0573, "skel": 0.56767}, "mean_gate": {"surface": 0.15422, "split": 0.08977, "merge": 0.12358, "skel": 0.12169}, "n_reads": 9754, "n_cand": 29850, "read_suppressed_frac": 0.1001, "dropout_reads_logged": 9754, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 666, "admitted_surface": 1316, "staged_surface": 62672, "admitted_split": 898, "staged_split": 47597, "admitted_merge": 4409, "staged_merge": 110088, "admitted_skel": 58913, "staged_skel": 41787, "mean_count": 225.43589782714844, "n_util_obs": 271, "util_count_spearman": -0.11077828425343077}, "probe_overhead_sec_total": 274.221, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.006309, "alpha": 0.03, "t": 3414.8} |
| {"event": "probe", "step": 2048, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3248.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.186634, "cps": 25.375, "mean_p": 0.223133, "boundary_entropy": 0.199337, "bpb": 2.106618}, "tr": {"chunks": 3660.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.202176, "cps": 28.59375, "mean_p": 0.229066, "boundary_entropy": 0.24377, "bpb": 2.318486}, "fi": {"chunks": 4241.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.225453, "cps": 33.132812, "mean_p": 0.246744, "boundary_entropy": 0.203849, "bpb": 2.482326}}, "bpb": {"en": 2.10662, "tr": 2.31849, "fi": 2.48233}, "gini_cpc": 0.04213, "gini_cps": 0.059378, "bpic": 5.088528, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0832747831306841, "fi": 1.2079931974347888}, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 689, "admitted_surface": 1316, "staged_surface": 62467, "admitted_split": 898, "staged_split": 47136, "admitted_merge": 4409, "staged_merge": 110703, "admitted_skel": 58913, "staged_skel": 41838, "mean_count": 232.51454162597656, "n_util_obs": 280, "util_count_spearman": -0.0964850134750261}, "t": 3487.6} |
| {"event": "consolidate", "step": 2050, "retired": 1, "split_retired": 2, "merge_admitted": 0, "admitted": 3, "swapped": 13, "uncertain_staged": 262109, "t": 3502.5} |
| {"event": "train", "step": 2050, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.49054, "bpb": 2.1504, "chunk_loss": 0.030414, "F": 0.18887, "G": 0.22647, "chunks": 193.41, "bytes_per_chunk": 5.295, "boundary_entropy": 0.20615, "router_noise": 0.0, "grad_norm": 0.8185, "lr": 0.0002419, "router_trainable": true, "sec_per_step": 1.8, "probe_sec": 0.105, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0551, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.00245, "n_shadow": 3, "mean_delta_shadow": 0.05221, "n_split": 1, "mean_delta_split": -0.01093, "n_merge": 2, "mean_delta_merge": 0.00697, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.26574, "trunc_abs_frac": 1.11016, "mean_delta_full": -0.25004, "read_hit_rate": {"surface": 0.31395, "split": 0.39323, "merge": 0.06824, "skel": 0.56769}, "mean_gate": {"surface": 0.1477, "split": 0.08662, "merge": 0.11947, "skel": 0.11833}, "n_reads": 9362, "n_cand": 27991, "read_suppressed_frac": 0.0979, "dropout_reads_logged": 9362, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 688, "admitted_surface": 1323, "staged_surface": 62473, "admitted_split": 901, "staged_split": 47170, "admitted_merge": 4409, "staged_merge": 110750, "admitted_skel": 58903, "staged_skel": 41751, "mean_count": 232.11734008789062, "n_util_obs": 277, "util_count_spearman": -0.08776845388910019}, "probe_overhead_sec_total": 280.982, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.005969, "alpha": 0.03, "t": 3504.8} |
| {"event": "consolidate", "step": 2100, "retired": 1, "split_retired": 1, "merge_admitted": 0, "admitted": 2, "swapped": 21, "uncertain_staged": 262108, "t": 3585.7} |
| {"event": "train", "step": 2100, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.54502, "bpb": 2.22899, "chunk_loss": 0.030404, "F": 0.19281, "G": 0.22668, "chunks": 197.44, "bytes_per_chunk": 5.186, "boundary_entropy": 0.20036, "router_noise": 0.0, "grad_norm": 0.7868, "lr": 0.0002388, "router_trainable": true, "sec_per_step": 1.642, "probe_sec": 0.083, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05174, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.05835, "n_shadow": 2, "mean_delta_shadow": -0.10583, "n_split": 1, "mean_delta_split": -0.05069, "n_merge": 1, "mean_delta_merge": 0.06967, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.78261, "trunc_abs_frac": 0.7853, "mean_delta_full": 0.28259, "read_hit_rate": {"surface": 0.31292, "split": 0.38763, "merge": 0.06332, "skel": 0.5758}, "mean_gate": {"surface": 0.15277, "split": 0.08717, "merge": 0.12942, "skel": 0.12173}, "n_reads": 9495, "n_cand": 28576, "read_suppressed_frac": 0.0981, "dropout_reads_logged": 9495, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 700, "admitted_surface": 1325, "staged_surface": 62712, "admitted_split": 906, "staged_split": 46794, "admitted_merge": 4411, "staged_merge": 110992, "admitted_skel": 58894, "staged_skel": 41646, "mean_count": 239.9004669189453, "n_util_obs": 283, "util_count_spearman": -0.1071766428629016}, "probe_overhead_sec_total": 287.552, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.006512, "alpha": 0.03, "t": 3586.9} |
| {"event": "consolidate", "step": 2150, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 58, "uncertain_staged": 262109, "t": 3669.6} |
| {"event": "train", "step": 2150, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.50395, "bpb": 2.16975, "chunk_loss": 0.030533, "F": 0.19891, "G": 0.23305, "chunks": 203.69, "bytes_per_chunk": 5.027, "boundary_entropy": 0.1982, "router_noise": 0.0, "grad_norm": 0.7175, "lr": 0.0002357, "router_trainable": true, "sec_per_step": 1.703, "probe_sec": 0.1003, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05721, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.05999, "n_shadow": 6, "mean_delta_shadow": 0.02232, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.04468, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -0.72661, "trunc_abs_frac": 1.77769, "mean_delta_full": -0.06653, "read_hit_rate": {"surface": 0.34783, "split": 0.40143, "merge": 0.08336, "skel": 0.59747}, "mean_gate": {"surface": 0.16021, "split": 0.08795, "merge": 0.13224, "skel": 0.12763}, "n_reads": 10411, "n_cand": 29447, "read_suppressed_frac": 0.0968, "dropout_reads_logged": 10411, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 711, "admitted_surface": 1326, "staged_surface": 62584, "admitted_split": 922, "staged_split": 46652, "admitted_merge": 4412, "staged_merge": 111151, "admitted_skel": 58876, "staged_skel": 41757, "mean_count": 247.87124633789062, "n_util_obs": 286, "util_count_spearman": -0.12207383400602502}, "probe_overhead_sec_total": 294.201, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.006386, "alpha": 0.03, "t": 3672.0} |
| {"event": "consolidate", "step": 2200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 13, "uncertain_staged": 262109, "t": 3760.0} |
| {"event": "probe", "step": 2200, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3260.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.187324, "cps": 25.46875, "mean_p": 0.217277, "boundary_entropy": 0.186309, "bpb": 2.087707}, "tr": {"chunks": 3497.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.193172, "cps": 27.320312, "mean_p": 0.219776, "boundary_entropy": 0.227172, "bpb": 2.29353}, "fi": {"chunks": 4083.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.217054, "cps": 31.898438, "mean_p": 0.239572, "boundary_entropy": 0.202076, "bpb": 2.45715}}, "bpb": {"en": 2.08771, "tr": 2.29353, "fi": 2.45715}, "gini_cpc": 0.033169, "gini_cps": 0.050615, "bpic": 5.233579, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0312206497990877, "fi": 1.158708029794602}, "inventory": {"size": 327667, "admitted": 65536, "staged": 262131, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 730, "admitted_surface": 1329, "staged_surface": 62862, "admitted_split": 928, "staged_split": 46977, "admitted_merge": 4413, "staged_merge": 110234, "admitted_skel": 58866, "staged_skel": 42058, "mean_count": 255.32452392578125, "n_util_obs": 292, "util_count_spearman": -0.13737602317407582}, "t": 3762.2} |
| {"event": "eval", "step": 2200, "phase": 1, "bpb": {"en": 2.1004, "tr": 2.29929, "fi": 2.46551}, "bytes_per_chunk": {"en": 5.229, "tr": 5.505, "fi": 4.891}, "bpb_mean": 2.2884, "frozen_inventory": true, "t": 3766.3} |
| {"event": "train", "step": 2200, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.48033, "bpb": 2.13566, "chunk_loss": 0.03035, "F": 0.19101, "G": 0.22379, "chunks": 195.59, "bytes_per_chunk": 5.235, "boundary_entropy": 0.19169, "router_noise": 0.0, "grad_norm": 0.684, "lr": 0.0002325, "router_trainable": true, "sec_per_step": 2.044, "probe_sec": 0.1402, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07611, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": 0.02742, "n_shadow": 2, "mean_delta_shadow": -0.00513, "n_split": 1, "mean_delta_split": -0.16864, "n_merge": 2, "mean_delta_merge": 0.05276, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.83679, "trunc_abs_frac": 0.81617, "mean_delta_full": 0.09412, "read_hit_rate": {"surface": 0.34623, "split": 0.40377, "merge": 0.08507, "skel": 0.59352}, "mean_gate": {"surface": 0.15573, "split": 0.08572, "merge": 0.12859, "skel": 0.12534}, "n_reads": 10131, "n_cand": 28519, "read_suppressed_frac": 0.1036, "dropout_reads_logged": 10131, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 731, "admitted_surface": 1329, "staged_surface": 62824, "admitted_split": 928, "staged_split": 47004, "admitted_merge": 4413, "staged_merge": 110263, "admitted_skel": 58866, "staged_skel": 42053, "mean_count": 255.47911071777344, "n_util_obs": 292, "util_count_spearman": -0.13673470273042362}, "probe_overhead_sec_total": 300.549, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.006177, "alpha": 0.03, "t": 3774.2} |
| {"event": "consolidate", "step": 2250, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 23, "uncertain_staged": 262109, "t": 3831.7} |
| {"event": "train", "step": 2250, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.50266, "bpb": 2.16789, "chunk_loss": 0.030475, "F": 0.19699, "G": 0.22796, "chunks": 201.72, "bytes_per_chunk": 5.076, "boundary_entropy": 0.19542, "router_noise": 0.0, "grad_norm": 0.7976, "lr": 0.0002293, "router_trainable": true, "sec_per_step": 1.19, "probe_sec": 0.1558, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06723, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.11855, "n_shadow": 4, "mean_delta_shadow": -0.02774, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": -0.43793, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -0.5816, "trunc_abs_frac": 0.45289, "mean_delta_full": 0.19523, "read_hit_rate": {"surface": 0.32275, "split": 0.40761, "merge": 0.07213, "skel": 0.58527}, "mean_gate": {"surface": 0.15665, "split": 0.08547, "merge": 0.135, "skel": 0.12465}, "n_reads": 10090, "n_cand": 29161, "read_suppressed_frac": 0.1022, "dropout_reads_logged": 10090, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 735, "admitted_surface": 1335, "staged_surface": 62719, "admitted_split": 933, "staged_split": 47206, "admitted_merge": 4415, "staged_merge": 110243, "admitted_skel": 58853, "staged_skel": 41976, "mean_count": 262.85626220703125, "n_util_obs": 293, "util_count_spearman": -0.14066932665878784}, "probe_overhead_sec_total": 306.297, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.00617, "alpha": 0.03, "t": 3833.7} |
| {"event": "consolidate", "step": 2300, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 6, "uncertain_staged": 262110, "t": 3901.0} |
| {"event": "train", "step": 2300, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.52417, "bpb": 2.19891, "chunk_loss": 0.030384, "F": 0.1933, "G": 0.22522, "chunks": 197.94, "bytes_per_chunk": 5.173, "boundary_entropy": 0.18741, "router_noise": 0.0, "grad_norm": 0.76, "lr": 0.000226, "router_trainable": true, "sec_per_step": 1.359, "probe_sec": 0.0654, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04446, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.05873, "n_shadow": 6, "mean_delta_shadow": -0.13278, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": -0.51599, "trunc_abs_frac": 0.70004, "mean_delta_full": 0.03658, "read_hit_rate": {"surface": 0.32863, "split": 0.39536, "merge": 0.06667, "skel": 0.57253}, "mean_gate": {"surface": 0.14563, "split": 0.07899, "merge": 0.1288, "skel": 0.11623}, "n_reads": 9785, "n_cand": 28808, "read_suppressed_frac": 0.0965, "dropout_reads_logged": 9785, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 743, "admitted_surface": 1339, "staged_surface": 62712, "admitted_split": 933, "staged_split": 46801, "admitted_merge": 4415, "staged_merge": 110905, "admitted_skel": 58849, "staged_skel": 41726, "mean_count": 270.1070556640625, "n_util_obs": 297, "util_count_spearman": -0.14084936064801837}, "probe_overhead_sec_total": 313.886, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.00623, "alpha": 0.03, "t": 3901.7} |
| {"event": "consolidate", "step": 2350, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 0, "uncertain_staged": 262110, "t": 3975.3} |
| {"event": "train", "step": 2350, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.56812, "bpb": 2.26232, "chunk_loss": 0.03051, "F": 0.20001, "G": 0.2321, "chunks": 204.81, "bytes_per_chunk": 5.0, "boundary_entropy": 0.18947, "router_noise": 0.0, "grad_norm": 0.7482, "lr": 0.0002226, "router_trainable": true, "sec_per_step": 1.523, "probe_sec": 0.1567, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.11746, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.03934, "n_shadow": 4, "mean_delta_shadow": -0.12086, "n_split": 1, "mean_delta_split": 0.06494, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -0.02729, "trunc_abs_frac": 0.72697, "mean_delta_full": -0.08053, "read_hit_rate": {"surface": 0.3272, "split": 0.40653, "merge": 0.07319, "skel": 0.57809}, "mean_gate": {"surface": 0.15057, "split": 0.08005, "merge": 0.12828, "skel": 0.12022}, "n_reads": 10230, "n_cand": 29605, "read_suppressed_frac": 0.1017, "dropout_reads_logged": 10230, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 764, "admitted_surface": 1337, "staged_surface": 62752, "admitted_split": 933, "staged_split": 47023, "admitted_merge": 4416, "staged_merge": 110521, "admitted_skel": 58850, "staged_skel": 41848, "mean_count": 277.09771728515625, "n_util_obs": 307, "util_count_spearman": -0.139992567948115}, "probe_overhead_sec_total": 322.311, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.006352, "alpha": 0.03, "t": 3977.8} |
| {"event": "consolidate", "step": 2400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 20, "uncertain_staged": 262110, "t": 4052.8} |
| {"event": "probe", "step": 2400, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3258.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.187209, "cps": 25.453125, "mean_p": 0.216564, "boundary_entropy": 0.180216, "bpb": 2.066669}, "tr": {"chunks": 3267.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.180467, "cps": 25.523438, "mean_p": 0.216099, "boundary_entropy": 0.226208, "bpb": 2.267006}, "fi": {"chunks": 4210.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.223805, "cps": 32.890625, "mean_p": 0.240357, "boundary_entropy": 0.185915, "bpb": 2.407435}}, "bpb": {"en": 2.06667, "tr": 2.26701, "fi": 2.40744}, "gini_cpc": 0.048847, "gini_cps": 0.059121, "bpic": 5.284769, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9639879901472331, "fi": 1.1954825812607393}, "inventory": {"size": 327660, "admitted": 65536, "staged": 262124, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 792, "admitted_surface": 1342, "staged_surface": 62527, "admitted_split": 940, "staged_split": 46712, "admitted_merge": 4418, "staged_merge": 111055, "admitted_skel": 58836, "staged_skel": 41830, "mean_count": 285.8626708984375, "n_util_obs": 315, "util_count_spearman": -0.10687314513001847}, "t": 4055.6} |
| {"event": "eval", "step": 2400, "phase": 1, "bpb": {"en": 2.07992, "tr": 2.27318, "fi": 2.41703}, "bytes_per_chunk": {"en": 5.2, "tr": 6.021, "fi": 4.719}, "bpb_mean": 2.25671, "frozen_inventory": true, "t": 4060.6} |
| {"event": "phase_transition", "step": 2400, "phase": 2, "objective": "global", "router_trainable": true, "encoder_trainable": true, "frozen_router_params": 0, "t": 4067.9} |
| {"event": "train", "step": 2400, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.50618, "bpb": 2.17295, "chunk_loss": 0.030454, "F": 0.19577, "G": 0.2272, "chunks": 200.47, "bytes_per_chunk": 5.108, "boundary_entropy": 0.18649, "router_noise": 0.0, "grad_norm": 0.7619, "lr": 0.0002192, "router_trainable": true, "sec_per_step": 1.841, "probe_sec": 0.1931, "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": 4, "mean_delta_loo": 0.21339, "n_shadow": 2, "mean_delta_shadow": -0.04007, "n_split": 3, "mean_delta_split": 0.05464, "n_merge": 1, "mean_delta_merge": -0.17119, "n_dropped_unpaired": 0, "n_full": 0, "read_hit_rate": {"surface": 0.32305, "split": 0.40904, "merge": 0.06865, "skel": 0.59184}, "mean_gate": {"surface": 0.14666, "split": 0.07631, "merge": 0.12758, "skel": 0.11588}, "n_reads": 10180, "n_cand": 29235, "read_suppressed_frac": 0.0994, "dropout_reads_logged": 10180, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 793, "admitted_surface": 1342, "staged_surface": 62514, "admitted_split": 940, "staged_split": 46722, "admitted_merge": 4418, "staged_merge": 111141, "admitted_skel": 58836, "staged_skel": 41767, "mean_count": 286.01800537109375, "n_util_obs": 315, "util_count_spearman": -0.10583192238436938}, "probe_overhead_sec_total": 330.755, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.006281, "alpha": 0.03, "t": 4069.8} |
| {"event": "consolidate", "step": 2450, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 154, "uncertain_staged": 262111, "t": 4131.2} |
| {"event": "train", "step": 2450, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.46454, "bpb": 2.11289, "chunk_loss": 0.030354, "F": 0.19055, "G": 0.22309, "chunks": 195.12, "bytes_per_chunk": 5.248, "boundary_entropy": 0.18498, "router_noise": 0.0, "grad_norm": 0.697, "lr": 0.0002158, "router_trainable": true, "sec_per_step": 1.27, "probe_sec": 0.3525, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.32396, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 3, "mean_delta_loo": -0.05434, "n_shadow": 7, "mean_delta_shadow": -0.01014, "n_split": 1, "mean_delta_split": -0.05573, "n_merge": 1, "mean_delta_merge": 0.0694, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.1269, "trunc_abs_frac": 0.15168, "mean_delta_full": 0.18024, "read_hit_rate": {"surface": 0.33532, "split": 0.41398, "merge": 0.08366, "skel": 0.58878}, "mean_gate": {"surface": 0.1525, "split": 0.07852, "merge": 0.12886, "skel": 0.12253}, "n_reads": 10109, "n_cand": 28445, "read_suppressed_frac": 0.0988, "dropout_reads_logged": 10109, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 803, "admitted_surface": 1345, "staged_surface": 62696, "admitted_split": 1045, "staged_split": 46394, "admitted_merge": 4449, "staged_merge": 111197, "admitted_skel": 58697, "staged_skel": 41857, "mean_count": 293.7672119140625, "n_util_obs": 326, "util_count_spearman": -0.08609358630248733}, "probe_overhead_sec_total": 340.97, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.005898, "alpha": 0.03, "t": 4133.3} |
| {"event": "consolidate", "step": 2500, "retired": 1, "split_retired": 3, "merge_admitted": 0, "admitted": 4, "swapped": 130, "uncertain_staged": 262111, "t": 4218.8} |
| {"event": "train", "step": 2500, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.44091, "bpb": 2.07879, "chunk_loss": 0.030556, "F": 0.20248, "G": 0.231, "chunks": 207.34, "bytes_per_chunk": 4.939, "boundary_entropy": 0.1817, "router_noise": 0.0, "grad_norm": 0.752, "lr": 0.0002123, "router_trainable": true, "sec_per_step": 1.753, "probe_sec": 0.3474, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.29419, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 11, "mean_delta_loo": -0.04261, "n_shadow": 3, "mean_delta_shadow": -0.06468, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 7, "trunc_ratio_sum": -0.01348, "trunc_abs_frac": 0.87264, "mean_delta_full": 0.06614, "read_hit_rate": {"surface": 0.34939, "split": 0.42951, "merge": 0.10014, "skel": 0.61194}, "mean_gate": {"surface": 0.14742, "split": 0.07503, "merge": 0.12161, "skel": 0.11575}, "n_reads": 11178, "n_cand": 30097, "read_suppressed_frac": 0.102, "dropout_reads_logged": 11178, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 821, "admitted_surface": 1344, "staged_surface": 62780, "admitted_split": 1169, "staged_split": 46573, "admitted_merge": 4453, "staged_merge": 111095, "admitted_skel": 58570, "staged_skel": 41696, "mean_count": 300.85894775390625, "n_util_obs": 335, "util_count_spearman": -0.08522978091677877}, "probe_overhead_sec_total": 357.096, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.006207, "alpha": 0.03, "t": 4221.0} |
| {"event": "consolidate", "step": 2550, "retired": 1, "split_retired": 0, "merge_admitted": 0, "admitted": 1, "swapped": 116, "uncertain_staged": 262111, "t": 4304.0} |
| {"event": "train", "step": 2550, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.4673, "bpb": 2.11686, "chunk_loss": 0.030392, "F": 0.19348, "G": 0.2248, "chunks": 198.12, "bytes_per_chunk": 5.168, "boundary_entropy": 0.1835, "router_noise": 0.0, "grad_norm": 0.7456, "lr": 0.0002088, "router_trainable": true, "sec_per_step": 1.708, "probe_sec": 0.3207, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.26031, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 2, "mean_delta_loo": 0.0965, "n_shadow": 6, "mean_delta_shadow": 0.01754, "n_split": 3, "mean_delta_split": 0.15192, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.35596, "trunc_abs_frac": 0.26699, "mean_delta_full": 0.3654, "read_hit_rate": {"surface": 0.33687, "split": 0.41362, "merge": 0.08781, "skel": 0.59558}, "mean_gate": {"surface": 0.15543, "split": 0.07815, "merge": 0.13256, "skel": 0.12574}, "n_reads": 10359, "n_cand": 28928, "read_suppressed_frac": 0.1005, "dropout_reads_logged": 10359, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 846, "admitted_surface": 1349, "staged_surface": 62591, "admitted_split": 1279, "staged_split": 46422, "admitted_merge": 4453, "staged_merge": 111341, "admitted_skel": 58455, "staged_skel": 41790, "mean_count": 308.7119140625, "n_util_obs": 343, "util_count_spearman": -0.06229057456170088}, "probe_overhead_sec_total": 372.338, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.005992, "alpha": 0.03, "t": 4306.4} |
| {"event": "consolidate", "step": 2600, "retired": 2, "split_retired": 3, "merge_admitted": 0, "admitted": 5, "swapped": 20, "uncertain_staged": 262111, "t": 4389.0} |
| {"event": "probe", "step": 2600, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3295.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.189335, "cps": 25.742188, "mean_p": 0.215239, "boundary_entropy": 0.167681, "bpb": 2.045482}, "tr": {"chunks": 3563.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.196818, "cps": 27.835938, "mean_p": 0.218735, "boundary_entropy": 0.215192, "bpb": 2.247098}, "fi": {"chunks": 4154.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.220828, "cps": 32.453125, "mean_p": 0.236655, "boundary_entropy": 0.177663, "bpb": 2.381725}}, "bpb": {"en": 2.04548, "tr": 2.2471, "fi": 2.38173}, "gini_cpc": 0.03459, "gini_cps": 0.052004, "bpic": 5.151834, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0395226874510106, "fi": 1.1663350044839453}, "inventory": {"size": 327655, "admitted": 65536, "staged": 262119, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 854, "admitted_surface": 1346, "staged_surface": 62708, "admitted_split": 1294, "staged_split": 46412, "admitted_merge": 4455, "staged_merge": 111549, "admitted_skel": 58441, "staged_skel": 41450, "mean_count": 315.85675048828125, "n_util_obs": 344, "util_count_spearman": -0.0555860865722723}, "t": 4391.7} |
| {"event": "eval", "step": 2600, "phase": 2, "bpb": {"en": 2.05979, "tr": 2.25431, "fi": 2.38964}, "bytes_per_chunk": {"en": 5.149, "tr": 5.496, "fi": 4.784}, "bpb_mean": 2.23458, "frozen_inventory": true, "t": 4398.5} |
| {"event": "train", "step": 2600, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.42564, "bpb": 2.05676, "chunk_loss": 0.030349, "F": 0.19299, "G": 0.22097, "chunks": 197.62, "bytes_per_chunk": 5.182, "boundary_entropy": 0.17409, "router_noise": 0.0, "grad_norm": 0.7666, "lr": 0.0002053, "router_trainable": true, "sec_per_step": 2.026, "probe_sec": 0.3118, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.27496, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": -0.02655, "n_shadow": 5, "mean_delta_shadow": 0.08166, "n_split": 2, "mean_delta_split": -0.01028, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 0, "read_hit_rate": {"surface": 0.35172, "split": 0.42594, "merge": 0.08387, "skel": 0.58856}, "mean_gate": {"surface": 0.14811, "split": 0.07336, "merge": 0.11855, "skel": 0.11801}, "n_reads": 10328, "n_cand": 28522, "read_suppressed_frac": 0.1018, "dropout_reads_logged": 10328, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 854, "admitted_surface": 1346, "staged_surface": 62701, "admitted_split": 1294, "staged_split": 46425, "admitted_merge": 4455, "staged_merge": 111533, "admitted_skel": 58441, "staged_skel": 41485, "mean_count": 316.01434326171875, "n_util_obs": 344, "util_count_spearman": -0.05552182690673664}, "probe_overhead_sec_total": 387.686, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.006, "alpha": 0.03, "t": 4407.7} |
| {"event": "consolidate", "step": 2650, "retired": 0, "split_retired": 3, "merge_admitted": 0, "admitted": 3, "swapped": 46, "uncertain_staged": 262112, "t": 4502.5} |
| {"event": "train", "step": 2650, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.47137, "bpb": 2.12274, "chunk_loss": 0.030348, "F": 0.19183, "G": 0.22054, "chunks": 196.44, "bytes_per_chunk": 5.213, "boundary_entropy": 0.1808, "router_noise": 0.0, "grad_norm": 0.7079, "lr": 0.0002017, "router_trainable": true, "sec_per_step": 1.932, "probe_sec": 0.439, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.21273, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.20118, "n_shadow": 6, "mean_delta_shadow": 0.06102, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.07932, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 0.87439, "trunc_abs_frac": 1.33472, "mean_delta_full": 0.12132, "read_hit_rate": {"surface": 0.33866, "split": 0.43239, "merge": 0.08759, "skel": 0.57255}, "mean_gate": {"surface": 0.15744, "split": 0.0784, "merge": 0.13407, "skel": 0.12778}, "n_reads": 10347, "n_cand": 28634, "read_suppressed_frac": 0.0974, "dropout_reads_logged": 10347, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 870, "admitted_surface": 1345, "staged_surface": 62967, "admitted_split": 1340, "staged_split": 46047, "admitted_merge": 4453, "staged_merge": 111302, "admitted_skel": 58398, "staged_skel": 41828, "mean_count": 323.4963684082031, "n_util_obs": 346, "util_count_spearman": -0.04955295195692035}, "probe_overhead_sec_total": 403.899, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.005789, "alpha": 0.03, "t": 4504.3} |
| {"event": "consolidate", "step": 2700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 11, "uncertain_staged": 262112, "t": 4568.3} |
| {"event": "train", "step": 2700, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.45367, "bpb": 2.0972, "chunk_loss": 0.03046, "F": 0.19754, "G": 0.22486, "chunks": 202.28, "bytes_per_chunk": 5.062, "boundary_entropy": 0.17355, "router_noise": 0.0, "grad_norm": 0.7084, "lr": 0.0001981, "router_trainable": true, "sec_per_step": 1.299, "probe_sec": 0.1192, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08069, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": 0.05924, "n_shadow": 6, "mean_delta_shadow": 0.04427, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": 0.31012, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 1.57949, "trunc_abs_frac": 1.57949, "mean_delta_full": 0.04761, "read_hit_rate": {"surface": 0.34839, "split": 0.42781, "merge": 0.08831, "skel": 0.59302}, "mean_gate": {"surface": 0.14549, "split": 0.07392, "merge": 0.12191, "skel": 0.11825}, "n_reads": 10717, "n_cand": 29375, "read_suppressed_frac": 0.1025, "dropout_reads_logged": 10717, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 874, "admitted_surface": 1348, "staged_surface": 63010, "admitted_split": 1347, "staged_split": 46085, "admitted_merge": 4451, "staged_merge": 111004, "admitted_skel": 58390, "staged_skel": 42045, "mean_count": 332.19873046875, "n_util_obs": 352, "util_count_spearman": -0.05306483699112769}, "probe_overhead_sec_total": 411.402, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.006119, "alpha": 0.03, "t": 4569.3} |
| {"event": "consolidate", "step": 2750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 6, "uncertain_staged": 262113, "t": 4634.3} |
| {"event": "train", "step": 2750, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.46107, "bpb": 2.10788, "chunk_loss": 0.030508, "F": 0.19739, "G": 0.22609, "chunks": 202.12, "bytes_per_chunk": 5.066, "boundary_entropy": 0.17959, "router_noise": 0.0, "grad_norm": 0.6924, "lr": 0.0001945, "router_trainable": true, "sec_per_step": 1.315, "probe_sec": 0.1288, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.1042, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.05961, "n_shadow": 8, "mean_delta_shadow": -0.00102, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.4755, "trunc_abs_frac": 0.48481, "mean_delta_full": -0.16939, "read_hit_rate": {"surface": 0.35923, "split": 0.43425, "merge": 0.0812, "skel": 0.57966}, "mean_gate": {"surface": 0.1546, "split": 0.07525, "merge": 0.12908, "skel": 0.12432}, "n_reads": 10557, "n_cand": 28990, "read_suppressed_frac": 0.103, "dropout_reads_logged": 10557, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 896, "admitted_surface": 1351, "staged_surface": 62731, "admitted_split": 1350, "staged_split": 46091, "admitted_merge": 4451, "staged_merge": 111531, "admitted_skel": 58384, "staged_skel": 41791, "mean_count": 340.2481384277344, "n_util_obs": 356, "util_count_spearman": -0.06344637976214071}, "probe_overhead_sec_total": 420.001, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.006119, "alpha": 0.03, "t": 4635.0} |
| {"event": "consolidate", "step": 2800, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1, "uncertain_staged": 262113, "t": 4690.4} |
| {"event": "probe", "step": 2800, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3247.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.186577, "cps": 25.367188, "mean_p": 0.213202, "boundary_entropy": 0.163793, "bpb": 2.026264}, "tr": {"chunks": 3349.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.184997, "cps": 26.164062, "mean_p": 0.210382, "boundary_entropy": 0.210888, "bpb": 2.206452}, "fi": {"chunks": 4146.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.220403, "cps": 32.390625, "mean_p": 0.236077, "boundary_entropy": 0.169879, "bpb": 2.337889}}, "bpb": {"en": 2.02626, "tr": 2.20645, "fi": 2.33789}, "gini_cpc": 0.039873, "gini_cps": 0.055793, "bpic": 5.281326, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9915312986508555, "fi": 1.181297394009825}, "inventory": {"size": 327678, "admitted": 65536, "staged": 262142, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 916, "admitted_surface": 1349, "staged_surface": 62818, "admitted_split": 1352, "staged_split": 46635, "admitted_merge": 4451, "staged_merge": 110694, "admitted_skel": 58384, "staged_skel": 41995, "mean_count": 347.87591552734375, "n_util_obs": 366, "util_count_spearman": -0.06427067375460412}, "t": 4692.7} |
| {"event": "eval", "step": 2800, "phase": 2, "bpb": {"en": 2.04002, "tr": 2.21239, "fi": 2.34459}, "bytes_per_chunk": {"en": 5.212, "tr": 5.837, "fi": 4.751}, "bpb_mean": 2.199, "frozen_inventory": true, "t": 4696.0} |
| {"event": "train", "step": 2800, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.4125, "bpb": 2.03781, "chunk_loss": 0.030369, "F": 0.19351, "G": 0.22135, "chunks": 198.16, "bytes_per_chunk": 5.168, "boundary_entropy": 0.17011, "router_noise": 0.0, "grad_norm": 0.7129, "lr": 0.0001909, "router_trainable": true, "sec_per_step": 1.321, "probe_sec": 0.1359, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10876, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.06287, "n_shadow": 4, "mean_delta_shadow": 0.05103, "n_split": 1, "mean_delta_split": -0.1582, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.5875, "trunc_abs_frac": 1.20587, "mean_delta_full": -0.15982, "read_hit_rate": {"surface": 0.35869, "split": 0.42942, "merge": 0.10307, "skel": 0.5965}, "mean_gate": {"surface": 0.15761, "split": 0.07617, "merge": 0.13064, "skel": 0.1265}, "n_reads": 10720, "n_cand": 28838, "read_suppressed_frac": 0.1001, "dropout_reads_logged": 10720, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 916, "admitted_surface": 1349, "staged_surface": 62701, "admitted_split": 1352, "staged_split": 46661, "admitted_merge": 4451, "staged_merge": 110761, "admitted_skel": 58384, "staged_skel": 42021, "mean_count": 348.03948974609375, "n_util_obs": 366, "util_count_spearman": -0.06464649555056108}, "probe_overhead_sec_total": 427.703, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.0057, "alpha": 0.03, "t": 4701.1} |
| {"event": "consolidate", "step": 2850, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 27, "uncertain_staged": 262113, "t": 4746.0} |
| {"event": "train", "step": 2850, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.46638, "bpb": 2.11553, "chunk_loss": 0.030571, "F": 0.20453, "G": 0.22988, "chunks": 209.44, "bytes_per_chunk": 4.889, "boundary_entropy": 0.1659, "router_noise": 0.0, "grad_norm": 0.7396, "lr": 0.0001872, "router_trainable": true, "sec_per_step": 0.93, "probe_sec": 0.1029, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05575, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.05278, "n_shadow": 4, "mean_delta_shadow": -0.14712, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": -0.38678, "n_dropped_unpaired": 0, "n_full": 7, "trunc_ratio_sum": 1.31849, "trunc_abs_frac": 0.49739, "mean_delta_full": -0.17664, "read_hit_rate": {"surface": 0.37331, "split": 0.42985, "merge": 0.08813, "skel": 0.6216}, "mean_gate": {"surface": 0.14097, "split": 0.06898, "merge": 0.12144, "skel": 0.11241}, "n_reads": 11460, "n_cand": 30456, "read_suppressed_frac": 0.1003, "dropout_reads_logged": 11460, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 939, "admitted_surface": 1350, "staged_surface": 62623, "admitted_split": 1375, "staged_split": 46534, "admitted_merge": 4451, "staged_merge": 111125, "admitted_skel": 58360, "staged_skel": 41862, "mean_count": 355.55902099609375, "n_util_obs": 372, "util_count_spearman": -0.05501527143289261}, "probe_overhead_sec_total": 433.071, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.005811, "alpha": 0.03, "t": 4747.6} |
| {"event": "consolidate", "step": 2900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 21, "uncertain_staged": 262113, "t": 4786.9} |
| {"event": "train", "step": 2900, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.39793, "bpb": 2.01678, "chunk_loss": 0.030272, "F": 0.18851, "G": 0.21633, "chunks": 193.03, "bytes_per_chunk": 5.305, "boundary_entropy": 0.16789, "router_noise": 0.0, "grad_norm": 0.6936, "lr": 0.0001835, "router_trainable": true, "sec_per_step": 0.806, "probe_sec": 0.0606, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04094, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": -0.01129, "n_shadow": 5, "mean_delta_shadow": -0.05837, "n_split": 1, "mean_delta_split": 0.08398, "n_merge": 1, "mean_delta_merge": 0.05057, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 214.375, "trunc_abs_frac": 0.5527, "mean_delta_full": -5e-05, "read_hit_rate": {"surface": 0.35216, "split": 0.43289, "merge": 0.09324, "skel": 0.58189}, "mean_gate": {"surface": 0.15272, "split": 0.07332, "merge": 0.12983, "skel": 0.12274}, "n_reads": 10258, "n_cand": 27985, "read_suppressed_frac": 0.1018, "dropout_reads_logged": 10258, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 965, "admitted_surface": 1352, "staged_surface": 62765, "admitted_split": 1394, "staged_split": 46271, "admitted_merge": 4451, "staged_merge": 111035, "admitted_skel": 58339, "staged_skel": 42073, "mean_count": 363.67230224609375, "n_util_obs": 378, "util_count_spearman": -0.05931371854700802}, "probe_overhead_sec_total": 437.018, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.006081, "alpha": 0.03, "t": 4787.9} |
| {"event": "consolidate", "step": 2950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 5, "uncertain_staged": 262113, "t": 4828.2} |
| {"event": "train", "step": 2950, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.54748, "bpb": 2.23254, "chunk_loss": 0.0306, "F": 0.20114, "G": 0.2271, "chunks": 205.97, "bytes_per_chunk": 4.972, "boundary_entropy": 0.17135, "router_noise": 0.0, "grad_norm": 0.7627, "lr": 0.0001799, "router_trainable": true, "sec_per_step": 0.819, "probe_sec": 0.0767, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04269, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": 0.01238, "n_shadow": 5, "mean_delta_shadow": -0.01695, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": 0.06372, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 3.01451, "trunc_abs_frac": 0.4291, "mean_delta_full": 0.02814, "read_hit_rate": {"surface": 0.30706, "split": 0.40707, "merge": 0.06757, "skel": 0.57467}, "mean_gate": {"surface": 0.14644, "split": 0.07166, "merge": 0.12639, "skel": 0.12055}, "n_reads": 9982, "n_cand": 29542, "read_suppressed_frac": 0.1035, "dropout_reads_logged": 9982, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 971, "admitted_surface": 1353, "staged_surface": 62929, "admitted_split": 1396, "staged_split": 46619, "admitted_merge": 4451, "staged_merge": 110614, "admitted_skel": 58336, "staged_skel": 41982, "mean_count": 371.76116943359375, "n_util_obs": 383, "util_count_spearman": -0.06983779606588672}, "probe_overhead_sec_total": 440.948, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.006187, "alpha": 0.03, "t": 4828.8} |
| {"event": "consolidate", "step": 3000, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 14, "uncertain_staged": 262113, "t": 4867.0} |
| {"event": "probe", "step": 3000, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3227.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.185428, "cps": 25.210938, "mean_p": 0.210231, "boundary_entropy": 0.155779, "bpb": 2.008196}, "tr": {"chunks": 3323.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.183561, "cps": 25.960938, "mean_p": 0.208482, "boundary_entropy": 0.198683, "bpb": 2.185187}, "fi": {"chunks": 4107.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.21833, "cps": 32.085938, "mean_p": 0.233308, "boundary_entropy": 0.165332, "bpb": 2.319524}}, "bpb": {"en": 2.0082, "tr": 2.18519, "fi": 2.31952}, "gini_cpc": 0.039466, "gini_cps": 0.05505, "bpic": 5.323449, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.989931045846683, "fi": 1.1774378002493018}, "inventory": {"size": 327664, "admitted": 65536, "staged": 262128, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 986, "admitted_surface": 1349, "staged_surface": 63045, "admitted_split": 1409, "staged_split": 46096, "admitted_merge": 4451, "staged_merge": 111056, "admitted_skel": 58327, "staged_skel": 41931, "mean_count": 380.9015197753906, "n_util_obs": 387, "util_count_spearman": -0.06563206033932625}, "t": 4868.4} |
| {"event": "eval", "step": 3000, "phase": 2, "bpb": {"en": 2.02244, "tr": 2.19161, "fi": 2.32464}, "bytes_per_chunk": {"en": 5.221, "tr": 5.948, "fi": 4.828}, "bpb_mean": 2.17957, "frozen_inventory": true, "t": 4870.9} |
| {"event": "train", "step": 3000, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.49304, "bpb": 2.15399, "chunk_loss": 0.030351, "F": 0.18994, "G": 0.21483, "chunks": 194.5, "bytes_per_chunk": 5.265, "boundary_entropy": 0.16125, "router_noise": 0.0, "grad_norm": 0.718, "lr": 0.0001761, "router_trainable": true, "sec_per_step": 0.948, "probe_sec": 0.0874, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04871, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.00595, "n_shadow": 4, "mean_delta_shadow": -0.02752, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 1.58055, "trunc_abs_frac": 0.69472, "mean_delta_full": 0.03536, "read_hit_rate": {"surface": 0.32978, "split": 0.42949, "merge": 0.07338, "skel": 0.58377}, "mean_gate": {"surface": 0.15135, "split": 0.07124, "merge": 0.13427, "skel": 0.12376}, "n_reads": 10160, "n_cand": 28430, "read_suppressed_frac": 0.1014, "dropout_reads_logged": 10160, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 985, "admitted_surface": 1349, "staged_surface": 63074, "admitted_split": 1409, "staged_split": 46203, "admitted_merge": 4451, "staged_merge": 110866, "admitted_skel": 58327, "staged_skel": 42001, "mean_count": 381.05657958984375, "n_util_obs": 388, "util_count_spearman": -0.061116901147742084}, "probe_overhead_sec_total": 444.543, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.005981, "alpha": 0.03, "t": 4876.3} |
| {"event": "consolidate", "step": 3050, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 1, "uncertain_staged": 262114, "t": 4915.2} |
| {"event": "train", "step": 3050, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.44547, "bpb": 2.08537, "chunk_loss": 0.030415, "F": 0.19403, "G": 0.21816, "chunks": 198.69, "bytes_per_chunk": 5.154, "boundary_entropy": 0.1579, "router_noise": 0.0, "grad_norm": 0.6942, "lr": 0.0001724, "router_trainable": true, "sec_per_step": 0.803, "probe_sec": 0.0832, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0548, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.05863, "n_shadow": 4, "mean_delta_shadow": -0.05285, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.26001, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -1.37382, "trunc_abs_frac": 2.21389, "mean_delta_full": 0.04861, "read_hit_rate": {"surface": 0.34303, "split": 0.4286, "merge": 0.07753, "skel": 0.59132}, "mean_gate": {"surface": 0.148, "split": 0.06977, "merge": 0.12387, "skel": 0.12069}, "n_reads": 10439, "n_cand": 28885, "read_suppressed_frac": 0.1018, "dropout_reads_logged": 10439, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 999, "admitted_surface": 1347, "staged_surface": 62721, "admitted_split": 1410, "staged_split": 46467, "admitted_merge": 4452, "staged_merge": 110895, "admitted_skel": 58327, "staged_skel": 42061, "mean_count": 388.76953125, "n_util_obs": 392, "util_count_spearman": -0.06326427649695254}, "probe_overhead_sec_total": 448.413, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.005637, "alpha": 0.03, "t": 4916.4} |
| {"event": "consolidate", "step": 3100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 6, "uncertain_staged": 262114, "t": 4950.4} |
| {"event": "train", "step": 3100, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.49143, "bpb": 2.15168, "chunk_loss": 0.030319, "F": 0.18942, "G": 0.21638, "chunks": 193.97, "bytes_per_chunk": 5.279, "boundary_entropy": 0.16572, "router_noise": 0.0, "grad_norm": 0.6907, "lr": 0.0001687, "router_trainable": true, "sec_per_step": 0.689, "probe_sec": 0.0697, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04438, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.02056, "n_shadow": 4, "mean_delta_shadow": -0.01772, "n_split": 1, "mean_delta_split": 0.09357, "n_merge": 1, "mean_delta_merge": 0.01135, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.18933, "trunc_abs_frac": 0.18933, "mean_delta_full": 0.39685, "read_hit_rate": {"surface": 0.32146, "split": 0.42978, "merge": 0.08367, "skel": 0.5841}, "mean_gate": {"surface": 0.14807, "split": 0.0696, "merge": 0.13291, "skel": 0.12097}, "n_reads": 10076, "n_cand": 28191, "read_suppressed_frac": 0.0994, "dropout_reads_logged": 10076, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1012, "admitted_surface": 1349, "staged_surface": 62593, "admitted_split": 1411, "staged_split": 46596, "admitted_merge": 4453, "staged_merge": 110722, "admitted_skel": 58323, "staged_skel": 42233, "mean_count": 396.7842712402344, "n_util_obs": 394, "util_count_spearman": -0.06847421619727545}, "probe_overhead_sec_total": 451.996, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.005996, "alpha": 0.03, "t": 4950.9} |
| {"event": "consolidate", "step": 3150, "retired": 0, "split_retired": 4, "merge_admitted": 0, "admitted": 1, "swapped": 0, "uncertain_staged": 262114, "t": 4997.7} |
| {"event": "train", "step": 3150, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.45643, "bpb": 2.10118, "chunk_loss": 0.030306, "F": 0.18927, "G": 0.21596, "chunks": 193.81, "bytes_per_chunk": 5.283, "boundary_entropy": 0.16169, "router_noise": 0.0, "grad_norm": 0.7313, "lr": 0.000165, "router_trainable": true, "sec_per_step": 0.947, "probe_sec": 0.0592, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03987, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": -0.05037, "n_shadow": 4, "mean_delta_shadow": -0.02576, "n_split": 1, "mean_delta_split": -0.2832, "n_merge": 2, "mean_delta_merge": 0.32366, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": -1.59412, "trunc_abs_frac": 1.59412, "mean_delta_full": -0.0415, "read_hit_rate": {"surface": 0.31977, "split": 0.43934, "merge": 0.07804, "skel": 0.58247}, "mean_gate": {"surface": 0.15159, "split": 0.0697, "merge": 0.13488, "skel": 0.12241}, "n_reads": 10212, "n_cand": 28361, "read_suppressed_frac": 0.0993, "dropout_reads_logged": 10212, "inventory": {"size": 327677, "admitted": 65533, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1010, "admitted_surface": 1345, "staged_surface": 62771, "admitted_split": 1412, "staged_split": 46657, "admitted_merge": 4453, "staged_merge": 110749, "admitted_skel": 58323, "staged_skel": 41967, "mean_count": 402.8968200683594, "n_util_obs": 395, "util_count_spearman": -0.05272489980727898}, "probe_overhead_sec_total": 456.326, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.005935, "alpha": 0.03, "t": 4998.2} |
| {"event": "consolidate", "step": 3200, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 4, "swapped": 10, "uncertain_staged": 262114, "t": 5034.8} |
| {"event": "probe", "step": 3200, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3253.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.186922, "cps": 25.414062, "mean_p": 0.210295, "boundary_entropy": 0.150285, "bpb": 1.990292}, "tr": {"chunks": 3323.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.183561, "cps": 25.960938, "mean_p": 0.205561, "boundary_entropy": 0.191078, "bpb": 2.162771}, "fi": {"chunks": 4155.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.220881, "cps": 32.460938, "mean_p": 0.231965, "boundary_entropy": 0.156017, "bpb": 2.299271}}, "bpb": {"en": 1.99029, "tr": 2.16277, "fi": 2.29927}, "gini_cpc": 0.042073, "gini_cps": 0.056037, "bpic": 5.286739, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9820189009982312, "fi": 1.1816781401637526}, "inventory": {"size": 327666, "admitted": 65536, "staged": 262130, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1023, "admitted_surface": 1342, "staged_surface": 62972, "admitted_split": 1419, "staged_split": 46459, "admitted_merge": 4453, "staged_merge": 110608, "admitted_skel": 58322, "staged_skel": 42091, "mean_count": 410.50384521484375, "n_util_obs": 400, "util_count_spearman": -0.04593320194498833}, "t": 5036.2} |
| {"event": "eval", "step": 3200, "phase": 2, "bpb": {"en": 2.00406, "tr": 2.16909, "fi": 2.30567}, "bytes_per_chunk": {"en": 5.167, "tr": 5.984, "fi": 4.801}, "bpb_mean": 2.15961, "frozen_inventory": true, "t": 5038.6} |
| {"event": "train", "step": 3200, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.43514, "bpb": 2.07047, "chunk_loss": 0.030366, "F": 0.19446, "G": 0.21795, "chunks": 199.12, "bytes_per_chunk": 5.142, "boundary_entropy": 0.15418, "router_noise": 0.0, "grad_norm": 0.646, "lr": 0.0001613, "router_trainable": true, "sec_per_step": 0.894, "probe_sec": 0.069, "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": 7, "mean_delta_loo": -0.05245, "n_shadow": 2, "mean_delta_shadow": -0.01698, "n_split": 1, "mean_delta_split": 0.06702, "n_merge": 2, "mean_delta_merge": -0.02603, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 1.42413, "trunc_abs_frac": 0.34468, "mean_delta_full": -0.01636, "read_hit_rate": {"surface": 0.33943, "split": 0.4447, "merge": 0.08386, "skel": 0.58682}, "mean_gate": {"surface": 0.1476, "split": 0.0673, "merge": 0.12771, "skel": 0.11803}, "n_reads": 10560, "n_cand": 28790, "read_suppressed_frac": 0.0997, "dropout_reads_logged": 10560, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1024, "admitted_surface": 1342, "staged_surface": 62972, "admitted_split": 1419, "staged_split": 46457, "admitted_merge": 4453, "staged_merge": 110623, "admitted_skel": 58322, "staged_skel": 42092, "mean_count": 410.66497802734375, "n_util_obs": 400, "util_count_spearman": -0.04643260705685122}, "probe_overhead_sec_total": 460.018, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.005987, "alpha": 0.03, "t": 5043.0} |
| {"event": "consolidate", "step": 3250, "retired": 1, "split_retired": 4, "merge_admitted": 0, "admitted": 5, "swapped": 5, "uncertain_staged": 262114, "t": 5078.3} |
| {"event": "train", "step": 3250, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.43941, "bpb": 2.07663, "chunk_loss": 0.030398, "F": 0.19183, "G": 0.21722, "chunks": 196.44, "bytes_per_chunk": 5.213, "boundary_entropy": 0.15882, "router_noise": 0.0, "grad_norm": 0.72, "lr": 0.0001576, "router_trainable": true, "sec_per_step": 0.731, "probe_sec": 0.0698, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04348, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.1111, "n_shadow": 5, "mean_delta_shadow": 0.05706, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.31725, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -2.97004, "trunc_abs_frac": 1.00351, "mean_delta_full": -0.02802, "read_hit_rate": {"surface": 0.33467, "split": 0.43384, "merge": 0.08968, "skel": 0.60012}, "mean_gate": {"surface": 0.1532, "split": 0.06917, "merge": 0.13121, "skel": 0.12409}, "n_reads": 10372, "n_cand": 28399, "read_suppressed_frac": 0.0999, "dropout_reads_logged": 10372, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1033, "admitted_surface": 1343, "staged_surface": 62966, "admitted_split": 1421, "staged_split": 46581, "admitted_merge": 4454, "staged_merge": 110559, "admitted_skel": 58318, "staged_skel": 42038, "mean_count": 417.7373046875, "n_util_obs": 400, "util_count_spearman": -0.06255451889798626}, "probe_overhead_sec_total": 463.606, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.006182, "alpha": 0.03, "t": 5079.5} |
| {"event": "consolidate", "step": 3300, "retired": 1, "split_retired": 2, "merge_admitted": 0, "admitted": 3, "swapped": 4, "uncertain_staged": 262114, "t": 5114.7} |
| {"event": "train", "step": 3300, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.41372, "bpb": 2.03957, "chunk_loss": 0.030493, "F": 0.19934, "G": 0.22129, "chunks": 204.12, "bytes_per_chunk": 5.017, "boundary_entropy": 0.15019, "router_noise": 0.0, "grad_norm": 0.6678, "lr": 0.0001539, "router_trainable": true, "sec_per_step": 0.712, "probe_sec": 0.0652, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04142, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.0201, "n_shadow": 5, "mean_delta_shadow": -0.0675, "n_split": 1, "mean_delta_split": 0.03931, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.38389, "trunc_abs_frac": 0.7163, "mean_delta_full": -0.25452, "read_hit_rate": {"surface": 0.34846, "split": 0.44509, "merge": 0.08581, "skel": 0.59714}, "mean_gate": {"surface": 0.14766, "split": 0.06523, "merge": 0.12541, "skel": 0.11725}, "n_reads": 10922, "n_cand": 29453, "read_suppressed_frac": 0.0986, "dropout_reads_logged": 10922, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1043, "admitted_surface": 1340, "staged_surface": 62591, "admitted_split": 1421, "staged_split": 46354, "admitted_merge": 4454, "staged_merge": 111257, "admitted_skel": 58321, "staged_skel": 41942, "mean_count": 425.048095703125, "n_util_obs": 405, "util_count_spearman": -0.0517456343335483}, "probe_overhead_sec_total": 467.27, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.005427, "alpha": 0.03, "t": 5115.1} |
| {"event": "consolidate", "step": 3350, "retired": 1, "split_retired": 0, "merge_admitted": 0, "admitted": 1, "swapped": 10, "uncertain_staged": 262114, "t": 5149.1} |
| {"event": "train", "step": 3350, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.36129, "bpb": 1.96393, "chunk_loss": 0.030366, "F": 0.19455, "G": 0.21662, "chunks": 199.22, "bytes_per_chunk": 5.14, "boundary_entropy": 0.14622, "router_noise": 0.0, "grad_norm": 0.7388, "lr": 0.0001501, "router_trainable": true, "sec_per_step": 0.688, "probe_sec": 0.0659, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04579, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": 0.04453, "n_shadow": 3, "mean_delta_shadow": -0.06738, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.12689, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.52713, "trunc_abs_frac": 0.49609, "mean_delta_full": 0.10676, "read_hit_rate": {"surface": 0.36702, "split": 0.45308, "merge": 0.09572, "skel": 0.60857}, "mean_gate": {"surface": 0.14187, "split": 0.06442, "merge": 0.12193, "skel": 0.1157}, "n_reads": 11088, "n_cand": 28941, "read_suppressed_frac": 0.1033, "dropout_reads_logged": 11088, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1058, "admitted_surface": 1341, "staged_surface": 62833, "admitted_split": 1425, "staged_split": 46201, "admitted_merge": 4454, "staged_merge": 111346, "admitted_skel": 58316, "staged_skel": 41764, "mean_count": 434.5997619628906, "n_util_obs": 414, "util_count_spearman": -0.04550316993894919}, "probe_overhead_sec_total": 470.897, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006089, "alpha": 0.03, "t": 5149.5} |
| {"event": "consolidate", "step": 3400, "retired": 1, "split_retired": 1, "merge_admitted": 0, "admitted": 2, "swapped": 0, "uncertain_staged": 262115, "t": 5185.7} |
| {"event": "probe", "step": 3400, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3284.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.188703, "cps": 25.65625, "mean_p": 0.210153, "boundary_entropy": 0.141784, "bpb": 1.979906}, "tr": {"chunks": 3322.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.183505, "cps": 25.953125, "mean_p": 0.206972, "boundary_entropy": 0.187155, "bpb": 2.154533}, "fi": {"chunks": 4195.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.223008, "cps": 32.773438, "mean_p": 0.233373, "boundary_entropy": 0.152858, "bpb": 2.26251}}, "bpb": {"en": 1.97991, "tr": 2.15453, "fi": 2.26251}, "gini_cpc": 0.044244, "gini_cps": 0.056229, "bpic": 5.252477, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9724561974943353, "fi": 1.1817920210341593}, "inventory": {"size": 327678, "admitted": 65536, "staged": 262142, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1064, "admitted_surface": 1341, "staged_surface": 62744, "admitted_split": 1424, "staged_split": 46502, "admitted_merge": 4454, "staged_merge": 110742, "admitted_skel": 58317, "staged_skel": 42154, "mean_count": 442.18719482421875, "n_util_obs": 418, "util_count_spearman": -0.05241469429656626}, "t": 5187.4} |
| {"event": "eval", "step": 3400, "phase": 2, "bpb": {"en": 1.99335, "tr": 2.1616, "fi": 2.26908}, "bytes_per_chunk": {"en": 5.156, "tr": 5.919, "fi": 4.76}, "bpb_mean": 2.14135, "frozen_inventory": true, "t": 5189.8} |
| {"event": "train", "step": 3400, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.42741, "bpb": 2.05931, "chunk_loss": 0.030327, "F": 0.19, "G": 0.2145, "chunks": 194.56, "bytes_per_chunk": 5.263, "boundary_entropy": 0.1503, "router_noise": 0.0, "grad_norm": 0.75, "lr": 0.0001465, "router_trainable": true, "sec_per_step": 0.907, "probe_sec": 0.0729, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04488, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": -0.01621, "n_shadow": 3, "mean_delta_shadow": 0.10491, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": -0.00073, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.05803, "trunc_abs_frac": 2.67491, "mean_delta_full": -0.12639, "read_hit_rate": {"surface": 0.33177, "split": 0.43595, "merge": 0.08536, "skel": 0.58465}, "mean_gate": {"surface": 0.14022, "split": 0.06373, "merge": 0.12368, "skel": 0.11348}, "n_reads": 10253, "n_cand": 28270, "read_suppressed_frac": 0.1018, "dropout_reads_logged": 10253, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1066, "admitted_surface": 1341, "staged_surface": 62713, "admitted_split": 1424, "staged_split": 46506, "admitted_merge": 4454, "staged_merge": 110733, "admitted_skel": 58317, "staged_skel": 42192, "mean_count": 442.3436584472656, "n_util_obs": 418, "util_count_spearman": -0.054824172256376265}, "probe_overhead_sec_total": 475.17, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.005661, "alpha": 0.03, "t": 5194.9} |
| {"event": "consolidate", "step": 3450, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 0, "uncertain_staged": 262115, "t": 5229.2} |
| {"event": "train", "step": 3450, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.38752, "bpb": 2.00176, "chunk_loss": 0.030365, "F": 0.19449, "G": 0.21718, "chunks": 199.16, "bytes_per_chunk": 5.142, "boundary_entropy": 0.15088, "router_noise": 0.0, "grad_norm": 0.7969, "lr": 0.0001428, "router_trainable": true, "sec_per_step": 0.706, "probe_sec": 0.0987, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05535, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.00101, "n_shadow": 5, "mean_delta_shadow": 0.04626, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.00388, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -7.83143, "trunc_abs_frac": 1.73905, "mean_delta_full": -0.01068, "read_hit_rate": {"surface": 0.34301, "split": 0.43121, "merge": 0.08638, "skel": 0.60036}, "mean_gate": {"surface": 0.14392, "split": 0.06537, "merge": 0.12485, "skel": 0.11782}, "n_reads": 10608, "n_cand": 28963, "read_suppressed_frac": 0.1009, "dropout_reads_logged": 10608, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1085, "admitted_surface": 1341, "staged_surface": 62730, "admitted_split": 1424, "staged_split": 46258, "admitted_merge": 4454, "staged_merge": 111239, "admitted_skel": 58317, "staged_skel": 41917, "mean_count": 450.17254638671875, "n_util_obs": 420, "util_count_spearman": -0.05174139337385255}, "probe_overhead_sec_total": 478.834, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.005586, "alpha": 0.03, "t": 5230.2} |
| {"event": "consolidate", "step": 3500, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 5, "uncertain_staged": 262115, "t": 5265.4} |
| {"event": "train", "step": 3500, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.36344, "bpb": 1.96703, "chunk_loss": 0.030257, "F": 0.18625, "G": 0.20983, "chunks": 190.72, "bytes_per_chunk": 5.369, "boundary_entropy": 0.14545, "router_noise": 0.0, "grad_norm": 0.6522, "lr": 0.0001391, "router_trainable": true, "sec_per_step": 0.728, "probe_sec": 0.0712, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04278, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": -0.08531, "n_shadow": 5, "mean_delta_shadow": -0.01746, "n_split": 1, "mean_delta_split": 0.19482, "n_merge": 1, "mean_delta_merge": 0.13495, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 13.97959, "trunc_abs_frac": 1.07922, "mean_delta_full": 0.00299, "read_hit_rate": {"surface": 0.36666, "split": 0.43762, "merge": 0.09687, "skel": 0.60517}, "mean_gate": {"surface": 0.14656, "split": 0.06551, "merge": 0.12286, "skel": 0.11845}, "n_reads": 10489, "n_cand": 27802, "read_suppressed_frac": 0.0961, "dropout_reads_logged": 10489, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1095, "admitted_surface": 1342, "staged_surface": 62854, "admitted_split": 1424, "staged_split": 46505, "admitted_merge": 4454, "staged_merge": 110656, "admitted_skel": 58316, "staged_skel": 42129, "mean_count": 458.01715087890625, "n_util_obs": 425, "util_count_spearman": -0.04048311111609604}, "probe_overhead_sec_total": 482.476, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006085, "alpha": 0.03, "t": 5266.6} |
| {"event": "consolidate", "step": 3550, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 0, "uncertain_staged": 262115, "t": 5300.8} |
| {"event": "train", "step": 3550, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.40826, "bpb": 2.03168, "chunk_loss": 0.030326, "F": 0.19159, "G": 0.21494, "chunks": 196.19, "bytes_per_chunk": 5.219, "boundary_entropy": 0.1477, "router_noise": 0.0, "grad_norm": 0.6754, "lr": 0.0001355, "router_trainable": true, "sec_per_step": 0.694, "probe_sec": 0.068, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04112, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": 0.66395, "n_shadow": 7, "mean_delta_shadow": 0.03234, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.10254, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.90113, "trunc_abs_frac": 0.75092, "mean_delta_full": 0.25007, "read_hit_rate": {"surface": 0.32997, "split": 0.43639, "merge": 0.08159, "skel": 0.58664}, "mean_gate": {"surface": 0.14958, "split": 0.06727, "merge": 0.13761, "skel": 0.12421}, "n_reads": 10292, "n_cand": 28449, "read_suppressed_frac": 0.0946, "dropout_reads_logged": 10292, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1090, "admitted_surface": 1341, "staged_surface": 62811, "admitted_split": 1424, "staged_split": 46230, "admitted_merge": 4454, "staged_merge": 110997, "admitted_skel": 58317, "staged_skel": 42106, "mean_count": 465.9288635253906, "n_util_obs": 433, "util_count_spearman": -0.039643443957076335}, "probe_overhead_sec_total": 486.118, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006069, "alpha": 0.03, "t": 5301.3} |
| {"event": "consolidate", "step": 3600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 6, "uncertain_staged": 262115, "t": 5340.5} |
| {"event": "probe", "step": 3600, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3335.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.191634, "cps": 26.054688, "mean_p": 0.210431, "boundary_entropy": 0.137499, "bpb": 1.963334}, "tr": {"chunks": 3431.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.189527, "cps": 26.804688, "mean_p": 0.206301, "boundary_entropy": 0.178883, "bpb": 2.120442}, "fi": {"chunks": 4129.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.219499, "cps": 32.257812, "mean_p": 0.231647, "boundary_entropy": 0.149969, "bpb": 2.243752}}, "bpb": {"en": 1.96333, "tr": 2.12044, "fi": 2.24375}, "gini_cpc": 0.033266, "gini_cps": 0.048585, "bpic": 5.207159, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9890049120056886, "fi": 1.145410820186632}, "inventory": {"size": 327674, "admitted": 65536, "staged": 262138, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1122, "admitted_surface": 1345, "staged_surface": 62760, "admitted_split": 1424, "staged_split": 46316, "admitted_merge": 4454, "staged_merge": 110978, "admitted_skel": 58313, "staged_skel": 42084, "mean_count": 473.88812255859375, "n_util_obs": 439, "util_count_spearman": -0.03790724986312717}, "t": 5342.0} |
| {"event": "eval", "step": 3600, "phase": 2, "bpb": {"en": 1.97854, "tr": 2.12592, "fi": 2.24927}, "bytes_per_chunk": {"en": 5.089, "tr": 5.649, "fi": 4.827}, "bpb_mean": 2.11791, "frozen_inventory": true, "t": 5344.4} |
| {"event": "train", "step": 3600, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.40784, "bpb": 2.03109, "chunk_loss": 0.030388, "F": 0.19559, "G": 0.21598, "chunks": 200.28, "bytes_per_chunk": 5.113, "boundary_entropy": 0.14164, "router_noise": 0.0, "grad_norm": 0.6431, "lr": 0.0001319, "router_trainable": true, "sec_per_step": 1.008, "probe_sec": 0.0775, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04667, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.18254, "n_shadow": 4, "mean_delta_shadow": -0.03522, "n_split": 2, "mean_delta_split": -0.00485, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -0.41525, "trunc_abs_frac": 1.92219, "mean_delta_full": 0.0623, "read_hit_rate": {"surface": 0.33966, "split": 0.43362, "merge": 0.08826, "skel": 0.59225}, "mean_gate": {"surface": 0.14194, "split": 0.0625, "merge": 0.12375, "skel": 0.11351}, "n_reads": 10766, "n_cand": 29384, "read_suppressed_frac": 0.1017, "dropout_reads_logged": 10766, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1122, "admitted_surface": 1345, "staged_surface": 62785, "admitted_split": 1424, "staged_split": 46283, "admitted_merge": 4454, "staged_merge": 110999, "admitted_skel": 58313, "staged_skel": 42077, "mean_count": 474.0523986816406, "n_util_obs": 439, "util_count_spearman": -0.03776555590813906}, "probe_overhead_sec_total": 490.467, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.005888, "alpha": 0.03, "t": 5351.7} |
| {"event": "consolidate", "step": 3650, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 1, "uncertain_staged": 262115, "t": 5385.6} |
| {"event": "train", "step": 3650, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.44258, "bpb": 2.0812, "chunk_loss": 0.030441, "F": 0.1965, "G": 0.21792, "chunks": 201.22, "bytes_per_chunk": 5.089, "boundary_entropy": 0.14365, "router_noise": 0.0, "grad_norm": 0.7254, "lr": 0.0001283, "router_trainable": true, "sec_per_step": 0.699, "probe_sec": 0.0767, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05089, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.01757, "n_shadow": 4, "mean_delta_shadow": -0.00887, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.02911, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 1.09345, "trunc_abs_frac": 1.6929, "mean_delta_full": 0.07942, "read_hit_rate": {"surface": 0.33713, "split": 0.43597, "merge": 0.08047, "skel": 0.59582}, "mean_gate": {"surface": 0.13694, "split": 0.0607, "merge": 0.12104, "skel": 0.11128}, "n_reads": 10703, "n_cand": 29322, "read_suppressed_frac": 0.0989, "dropout_reads_logged": 10703, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1131, "admitted_surface": 1343, "staged_surface": 62831, "admitted_split": 1424, "staged_split": 45934, "admitted_merge": 4455, "staged_merge": 111238, "admitted_skel": 58314, "staged_skel": 42141, "mean_count": 481.22705078125, "n_util_obs": 439, "util_count_spearman": -0.043391524380384495}, "probe_overhead_sec_total": 494.168, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006161, "alpha": 0.03, "t": 5386.6} |
| {"event": "consolidate", "step": 3700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1, "uncertain_staged": 262115, "t": 5421.2} |
| {"event": "train", "step": 3700, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.39567, "bpb": 2.01352, "chunk_loss": 0.030349, "F": 0.19354, "G": 0.21657, "chunks": 198.19, "bytes_per_chunk": 5.167, "boundary_entropy": 0.14511, "router_noise": 0.0, "grad_norm": 0.6575, "lr": 0.0001247, "router_trainable": true, "sec_per_step": 0.699, "probe_sec": 0.0614, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04247, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.0019, "n_shadow": 2, "mean_delta_shadow": -0.03143, "n_split": 2, "mean_delta_split": 0.10241, "n_merge": 1, "mean_delta_merge": 0.0708, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -2.12849, "trunc_abs_frac": 0.50878, "mean_delta_full": -0.03064, "read_hit_rate": {"surface": 0.34485, "split": 0.43728, "merge": 0.08251, "skel": 0.58608}, "mean_gate": {"surface": 0.14823, "split": 0.06322, "merge": 0.12673, "skel": 0.11999}, "n_reads": 10468, "n_cand": 28694, "read_suppressed_frac": 0.0943, "dropout_reads_logged": 10468, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1131, "admitted_surface": 1344, "staged_surface": 62795, "admitted_split": 1424, "staged_split": 46036, "admitted_merge": 4455, "staged_merge": 111042, "admitted_skel": 58313, "staged_skel": 42271, "mean_count": 489.3497314453125, "n_util_obs": 445, "util_count_spearman": -0.02697541598634963}, "probe_overhead_sec_total": 497.759, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.005565, "alpha": 0.03, "t": 5421.6} |
| {"event": "consolidate", "step": 3750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 3, "uncertain_staged": 262115, "t": 5457.0} |
| {"event": "train", "step": 3750, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.37952, "bpb": 1.99023, "chunk_loss": 0.030299, "F": 0.19086, "G": 0.21286, "chunks": 195.44, "bytes_per_chunk": 5.24, "boundary_entropy": 0.14506, "router_noise": 0.0, "grad_norm": 0.6773, "lr": 0.0001212, "router_trainable": true, "sec_per_step": 0.727, "probe_sec": 0.0624, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.039, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": 0.23634, "n_shadow": 8, "mean_delta_shadow": 0.02539, "n_split": 1, "mean_delta_split": -0.16376, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.20391, "trunc_abs_frac": 0.38902, "mean_delta_full": -0.08112, "read_hit_rate": {"surface": 0.34153, "split": 0.42884, "merge": 0.08741, "skel": 0.59447}, "mean_gate": {"surface": 0.14504, "split": 0.06344, "merge": 0.12437, "skel": 0.11865}, "n_reads": 10424, "n_cand": 28574, "read_suppressed_frac": 0.0935, "dropout_reads_logged": 10424, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1150, "admitted_surface": 1344, "staged_surface": 62707, "admitted_split": 1424, "staged_split": 46035, "admitted_merge": 4456, "staged_merge": 111609, "admitted_skel": 58312, "staged_skel": 41793, "mean_count": 497.42449951171875, "n_util_obs": 456, "util_count_spearman": -0.026127471515010384}, "probe_overhead_sec_total": 501.477, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.005659, "alpha": 0.03, "t": 5458.0} |
| {"event": "consolidate", "step": 3800, "retired": 1, "split_retired": 1, "merge_admitted": 0, "admitted": 2, "swapped": 1, "uncertain_staged": 262115, "t": 5491.4} |
| {"event": "probe", "step": 3800, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3295.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.189335, "cps": 25.742188, "mean_p": 0.208578, "boundary_entropy": 0.130123, "bpb": 1.950955}, "tr": {"chunks": 3340.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.1845, "cps": 26.09375, "mean_p": 0.203899, "boundary_entropy": 0.176108, "bpb": 2.113883}, "fi": {"chunks": 4145.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.22035, "cps": 32.382812, "mean_p": 0.230844, "boundary_entropy": 0.142283, "bpb": 2.219769}}, "bpb": {"en": 1.95096, "tr": 2.11388, "fi": 2.21977}, "gini_cpc": 0.040223, "gini_cps": 0.052566, "bpic": 5.262709, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.974461346080936, "fi": 1.1638080388988814}, "inventory": {"size": 327677, "admitted": 65536, "staged": 262141, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1162, "admitted_surface": 1344, "staged_surface": 63009, "admitted_split": 1424, "staged_split": 46184, "admitted_merge": 4457, "staged_merge": 110806, "admitted_skel": 58311, "staged_skel": 42142, "mean_count": 504.64849853515625, "n_util_obs": 461, "util_count_spearman": -0.01913223588803696}, "t": 5492.8} |
| {"event": "eval", "step": 3800, "phase": 2, "bpb": {"en": 1.9653, "tr": 2.11862, "fi": 2.22667}, "bytes_per_chunk": {"en": 5.116, "tr": 5.908, "fi": 4.81}, "bpb_mean": 2.10353, "frozen_inventory": true, "t": 5495.3} |
| {"event": "train", "step": 3800, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.45231, "bpb": 2.09525, "chunk_loss": 0.030389, "F": 0.19672, "G": 0.21615, "chunks": 201.44, "bytes_per_chunk": 5.083, "boundary_entropy": 0.13911, "router_noise": 0.0, "grad_norm": 0.6704, "lr": 0.0001177, "router_trainable": true, "sec_per_step": 0.844, "probe_sec": 0.0651, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0431, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.00798, "n_shadow": 6, "mean_delta_shadow": -0.16173, "n_split": 1, "mean_delta_split": 0.08142, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.27284, "trunc_abs_frac": 0.26311, "mean_delta_full": -0.36432, "read_hit_rate": {"surface": 0.31915, "split": 0.43102, "merge": 0.07255, "skel": 0.60441}, "mean_gate": {"surface": 0.13663, "split": 0.05969, "merge": 0.12499, "skel": 0.11115}, "n_reads": 10632, "n_cand": 29561, "read_suppressed_frac": 0.0973, "dropout_reads_logged": 10632, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1161, "admitted_surface": 1344, "staged_surface": 62999, "admitted_split": 1424, "staged_split": 46223, "admitted_merge": 4457, "staged_merge": 110857, "admitted_skel": 58311, "staged_skel": 42065, "mean_count": 504.81072998046875, "n_util_obs": 461, "util_count_spearman": -0.018701642802737974}, "probe_overhead_sec_total": 505.452, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006144, "alpha": 0.03, "t": 5500.2} |
| {"event": "consolidate", "step": 3850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 22, "uncertain_staged": 262115, "t": 5534.9} |
| {"event": "train", "step": 3850, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.54755, "bpb": 2.23264, "chunk_loss": 0.030596, "F": 0.19412, "G": 0.21819, "chunks": 198.78, "bytes_per_chunk": 5.151, "boundary_entropy": 0.14703, "router_noise": 0.0, "grad_norm": 0.7297, "lr": 0.0001142, "router_trainable": true, "sec_per_step": 0.709, "probe_sec": 0.0672, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04105, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": 0.03298, "n_shadow": 5, "mean_delta_shadow": -0.00476, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": 0.1245, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 8.88439, "trunc_abs_frac": 0.56166, "mean_delta_full": -0.01056, "read_hit_rate": {"surface": 0.31, "split": 0.40703, "merge": 0.0775, "skel": 0.56222}, "mean_gate": {"surface": 0.13931, "split": 0.06128, "merge": 0.12906, "skel": 0.1163}, "n_reads": 9649, "n_cand": 28468, "read_suppressed_frac": 0.1008, "dropout_reads_logged": 9649, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1183, "admitted_surface": 1353, "staged_surface": 62752, "admitted_split": 1425, "staged_split": 46142, "admitted_merge": 4458, "staged_merge": 111361, "admitted_skel": 58300, "staged_skel": 41889, "mean_count": 513.0, "n_util_obs": 467, "util_count_spearman": -0.013949323441641803}, "probe_overhead_sec_total": 509.07, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006151, "alpha": 0.03, "t": 5535.6} |
| {"event": "consolidate", "step": 3900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 7, "uncertain_staged": 262115, "t": 5569.0} |
| {"event": "train", "step": 3900, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.43621, "bpb": 2.07202, "chunk_loss": 0.030481, "F": 0.19727, "G": 0.21808, "chunks": 202.0, "bytes_per_chunk": 5.069, "boundary_entropy": 0.1422, "router_noise": 0.0, "grad_norm": 0.7257, "lr": 0.0001108, "router_trainable": true, "sec_per_step": 0.693, "probe_sec": 0.0758, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05015, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.039, "n_shadow": 6, "mean_delta_shadow": -0.01994, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.11505, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.35367, "trunc_abs_frac": 0.35367, "mean_delta_full": -0.31909, "read_hit_rate": {"surface": 0.32603, "split": 0.42199, "merge": 0.07203, "skel": 0.58601}, "mean_gate": {"surface": 0.13579, "split": 0.06048, "merge": 0.12099, "skel": 0.11312}, "n_reads": 10336, "n_cand": 29280, "read_suppressed_frac": 0.1008, "dropout_reads_logged": 10336, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1192, "admitted_surface": 1353, "staged_surface": 62780, "admitted_split": 1425, "staged_split": 46127, "admitted_merge": 4460, "staged_merge": 111051, "admitted_skel": 58298, "staged_skel": 42186, "mean_count": 521.039306640625, "n_util_obs": 473, "util_count_spearman": -0.015151081323856962}, "probe_overhead_sec_total": 512.652, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006385, "alpha": 0.03, "t": 5570.3} |
| {"event": "consolidate", "step": 3950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 10, "uncertain_staged": 262115, "t": 5604.7} |
| {"event": "train", "step": 3950, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.49131, "bpb": 2.15151, "chunk_loss": 0.030361, "F": 0.1907, "G": 0.21206, "chunks": 195.28, "bytes_per_chunk": 5.244, "boundary_entropy": 0.14315, "router_noise": 0.0, "grad_norm": 0.7715, "lr": 0.0001074, "router_trainable": true, "sec_per_step": 0.697, "probe_sec": 0.0683, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04667, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.00114, "n_shadow": 5, "mean_delta_shadow": 0.03077, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.10278, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -0.3112, "trunc_abs_frac": 0.71176, "mean_delta_full": 0.34665, "read_hit_rate": {"surface": 0.33489, "split": 0.42981, "merge": 0.06614, "skel": 0.56777}, "mean_gate": {"surface": 0.1394, "split": 0.06069, "merge": 0.12692, "skel": 0.11314}, "n_reads": 10027, "n_cand": 28357, "read_suppressed_frac": 0.1031, "dropout_reads_logged": 10027, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1207, "admitted_surface": 1359, "staged_surface": 62831, "admitted_split": 1426, "staged_split": 46432, "admitted_merge": 4460, "staged_merge": 110543, "admitted_skel": 58291, "staged_skel": 42338, "mean_count": 529.1815185546875, "n_util_obs": 481, "util_count_spearman": -0.03431453687366152}, "probe_overhead_sec_total": 516.225, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006477, "alpha": 0.03, "t": 5605.1} |
| {"event": "consolidate", "step": 4000, "retired": 0, "split_retired": 5, "merge_admitted": 0, "admitted": 4, "swapped": 0, "uncertain_staged": 262116, "t": 5640.9} |
| {"event": "probe", "step": 4000, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3285.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.188761, "cps": 25.664062, "mean_p": 0.2095, "boundary_entropy": 0.136594, "bpb": 1.941705}, "tr": {"chunks": 3459.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.191073, "cps": 27.023438, "mean_p": 0.205331, "boundary_entropy": 0.171232, "bpb": 2.095635}, "fi": {"chunks": 4143.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.220243, "cps": 32.367188, "mean_p": 0.230161, "boundary_entropy": 0.139634, "bpb": 2.202481}}, "bpb": {"en": 1.94171, "tr": 2.09564, "fi": 2.20248}, "gini_cpc": 0.034977, "gini_cps": 0.05254, "bpic": 5.210986, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0122522642504572, "fi": 1.1667875761995858}, "inventory": {"size": 327675, "admitted": 65535, "staged": 262140, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1205, "admitted_surface": 1354, "staged_surface": 62409, "admitted_split": 1427, "staged_split": 46216, "admitted_merge": 4460, "staged_merge": 111705, "admitted_skel": 58294, "staged_skel": 41810, "mean_count": 535.7836303710938, "n_util_obs": 484, "util_count_spearman": -0.03832902457020315}, "t": 5642.3} |
| {"event": "eval", "step": 4000, "phase": 2, "bpb": {"en": 1.9563, "tr": 2.1001, "fi": 2.20913}, "bytes_per_chunk": {"en": 5.162, "tr": 5.653, "fi": 4.811}, "bpb_mean": 2.08851, "frozen_inventory": true, "t": 5644.7} |
| {"event": "train", "step": 4000, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.41426, "bpb": 2.04035, "chunk_loss": 0.030452, "F": 0.19489, "G": 0.2177, "chunks": 199.56, "bytes_per_chunk": 5.131, "boundary_entropy": 0.14471, "router_noise": 0.0, "grad_norm": 0.685, "lr": 0.000104, "router_trainable": true, "sec_per_step": 0.88, "probe_sec": 0.0637, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04094, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.1118, "n_shadow": 5, "mean_delta_shadow": -0.013, "n_split": 1, "mean_delta_split": -0.14636, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -0.00145, "trunc_abs_frac": 0.6464, "mean_delta_full": 0.23926, "read_hit_rate": {"surface": 0.33837, "split": 0.43311, "merge": 0.08336, "skel": 0.5929}, "mean_gate": {"surface": 0.14157, "split": 0.06186, "merge": 0.12655, "skel": 0.11684}, "n_reads": 10508, "n_cand": 28918, "read_suppressed_frac": 0.097, "dropout_reads_logged": 10508, "inventory": {"size": 327679, "admitted": 65535, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1205, "admitted_surface": 1354, "staged_surface": 62453, "admitted_split": 1427, "staged_split": 46220, "admitted_merge": 4460, "staged_merge": 111645, "admitted_skel": 58294, "staged_skel": 41826, "mean_count": 535.9439697265625, "n_util_obs": 484, "util_count_spearman": -0.037550521408159385}, "probe_overhead_sec_total": 519.769, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.0061, "alpha": 0.03, "t": 5649.2} |
| {"event": "consolidate", "step": 4050, "retired": 2, "split_retired": 3, "merge_admitted": 0, "admitted": 6, "swapped": 0, "uncertain_staged": 262116, "t": 5685.6} |
| {"event": "train", "step": 4050, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.42889, "bpb": 2.06146, "chunk_loss": 0.030397, "F": 0.19492, "G": 0.21661, "chunks": 199.59, "bytes_per_chunk": 5.13, "boundary_entropy": 0.14151, "router_noise": 0.0, "grad_norm": 0.7199, "lr": 0.0001007, "router_trainable": true, "sec_per_step": 0.742, "probe_sec": 0.0597, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03974, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.046, "n_shadow": 3, "mean_delta_shadow": -0.0239, "n_split": 2, "mean_delta_split": -0.03133, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.12695, "trunc_abs_frac": 0.64178, "mean_delta_full": -0.17212, "read_hit_rate": {"surface": 0.3443, "split": 0.42648, "merge": 0.09204, "skel": 0.57226}, "mean_gate": {"surface": 0.14187, "split": 0.06118, "merge": 0.1293, "skel": 0.11633}, "n_reads": 10407, "n_cand": 28882, "read_suppressed_frac": 0.0958, "dropout_reads_logged": 10407, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1207, "admitted_surface": 1353, "staged_surface": 62860, "admitted_split": 1426, "staged_split": 46277, "admitted_merge": 4461, "staged_merge": 110974, "admitted_skel": 58296, "staged_skel": 42033, "mean_count": 542.4915771484375, "n_util_obs": 484, "util_count_spearman": -0.023177700226366504}, "probe_overhead_sec_total": 523.352, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006741, "alpha": 0.03, "t": 5686.2} |
| {"event": "probe", "step": 4096, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3271.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.187956, "cps": 25.554688, "mean_p": 0.20739, "boundary_entropy": 0.129655, "bpb": 1.934978}, "tr": {"chunks": 3403.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.18798, "cps": 26.585938, "mean_p": 0.203926, "boundary_entropy": 0.172273, "bpb": 2.090654}, "fi": {"chunks": 4176.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.221998, "cps": 32.625, "mean_p": 0.230483, "boundary_entropy": 0.137486, "bpb": 2.195858}}, "bpb": {"en": 1.93498, "tr": 2.09065, "fi": 2.19586}, "gini_cpc": 0.037955, "gini_cps": 0.055607, "bpic": 5.228756, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.000126589732556, "fi": 1.1811149934859433}, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1208, "admitted_surface": 1353, "staged_surface": 62959, "admitted_split": 1426, "staged_split": 46187, "admitted_merge": 4461, "staged_merge": 110871, "admitted_skel": 58296, "staged_skel": 42127, "mean_count": 549.6715087890625, "n_util_obs": 489, "util_count_spearman": -0.018729771814626234}, "t": 5716.6} |
| {"event": "consolidate", "step": 4100, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 6, "uncertain_staged": 262116, "t": 5726.8} |
| {"event": "train", "step": 4100, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.36478, "bpb": 1.96896, "chunk_loss": 0.03039, "F": 0.19519, "G": 0.216, "chunks": 199.88, "bytes_per_chunk": 5.123, "boundary_entropy": 0.13574, "router_noise": 0.0, "grad_norm": 0.7107, "lr": 9.75e-05, "router_trainable": true, "sec_per_step": 0.834, "probe_sec": 0.0712, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0458, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 2, "mean_delta_loo": -0.01556, "n_shadow": 4, "mean_delta_shadow": 0.00113, "n_split": 3, "mean_delta_split": 0.00247, "n_merge": 1, "mean_delta_merge": 0.00439, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.24603, "trunc_abs_frac": 0.42291, "mean_delta_full": -0.0743, "read_hit_rate": {"surface": 0.36188, "split": 0.44136, "merge": 0.09697, "skel": 0.61424}, "mean_gate": {"surface": 0.14205, "split": 0.06047, "merge": 0.12562, "skel": 0.11634}, "n_reads": 11042, "n_cand": 29130, "read_suppressed_frac": 0.0997, "dropout_reads_logged": 11042, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1209, "admitted_surface": 1355, "staged_surface": 62936, "admitted_split": 1427, "staged_split": 46198, "admitted_merge": 4462, "staged_merge": 110859, "admitted_skel": 58292, "staged_skel": 42151, "mean_count": 550.320556640625, "n_util_obs": 488, "util_count_spearman": -0.019680505877353743}, "probe_overhead_sec_total": 527.484, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006231, "alpha": 0.03, "t": 5727.9} |
| {"event": "consolidate", "step": 4150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 13, "uncertain_staged": 262116, "t": 5760.9} |
| {"event": "train", "step": 4150, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.42615, "bpb": 2.05749, "chunk_loss": 0.030564, "F": 0.20279, "G": 0.22107, "chunks": 207.66, "bytes_per_chunk": 4.931, "boundary_entropy": 0.13444, "router_noise": 0.0, "grad_norm": 0.7599, "lr": 9.43e-05, "router_trainable": true, "sec_per_step": 0.674, "probe_sec": 0.0906, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05399, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.08721, "n_shadow": 7, "mean_delta_shadow": 0.02614, "n_split": 1, "mean_delta_split": 0.04572, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -0.12224, "trunc_abs_frac": 0.49576, "mean_delta_full": 0.06641, "read_hit_rate": {"surface": 0.34568, "split": 0.41504, "merge": 0.07978, "skel": 0.5954}, "mean_gate": {"surface": 0.13256, "split": 0.05757, "merge": 0.11752, "skel": 0.10821}, "n_reads": 10735, "n_cand": 30036, "read_suppressed_frac": 0.0961, "dropout_reads_logged": 10735, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1219, "admitted_surface": 1357, "staged_surface": 62807, "admitted_split": 1429, "staged_split": 46302, "admitted_merge": 4462, "staged_merge": 110988, "admitted_skel": 58288, "staged_skel": 42047, "mean_count": 558.4100341796875, "n_util_obs": 491, "util_count_spearman": -0.022315112660802015}, "probe_overhead_sec_total": 531.574, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006012, "alpha": 0.03, "t": 5761.6} |
| {"event": "consolidate", "step": 4200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 20, "uncertain_staged": 262116, "t": 5796.7} |
| {"event": "probe", "step": 4200, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3260.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.187324, "cps": 25.46875, "mean_p": 0.206471, "boundary_entropy": 0.130427, "bpb": 1.933771}, "tr": {"chunks": 3416.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.188698, "cps": 26.6875, "mean_p": 0.203391, "boundary_entropy": 0.171237, "bpb": 2.079278}, "fi": {"chunks": 4130.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.219552, "cps": 32.265625, "mean_p": 0.228126, "boundary_entropy": 0.139992, "bpb": 2.190852}}, "bpb": {"en": 1.93377, "tr": 2.07928, "fi": 2.19085}, "gini_cpc": 0.036075, "gini_cps": 0.053674, "bpic": 5.250046, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0073347840187827, "fi": 1.1720460845093572}, "inventory": {"size": 327660, "admitted": 65536, "staged": 262124, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1244, "admitted_surface": 1361, "staged_surface": 63049, "admitted_split": 1431, "staged_split": 46060, "admitted_merge": 4463, "staged_merge": 111211, "admitted_skel": 58281, "staged_skel": 41804, "mean_count": 569.1753540039062, "n_util_obs": 498, "util_count_spearman": -0.019069824191055608}, "t": 5798.0} |
| {"event": "eval", "step": 4200, "phase": 2, "bpb": {"en": 1.94698, "tr": 2.08553, "fi": 2.19885}, "bytes_per_chunk": {"en": 5.201, "tr": 5.754, "fi": 4.822}, "bpb_mean": 2.07712, "frozen_inventory": true, "t": 5800.4} |
| {"event": "phase_transition", "step": 4200, "phase": 3, "objective": "global", "router_trainable": true, "encoder_trainable": true, "frozen_router_params": 0, "t": 5804.2} |
| {"event": "train", "step": 4200, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.3537, "bpb": 1.95297, "chunk_loss": 0.030231, "F": 0.18613, "G": 0.20781, "chunks": 190.59, "bytes_per_chunk": 5.373, "boundary_entropy": 0.13705, "router_noise": 0.0, "grad_norm": 0.6998, "lr": 9.12e-05, "router_trainable": true, "sec_per_step": 0.859, "probe_sec": 0.0641, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03936, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.07644, "n_shadow": 3, "mean_delta_shadow": -0.05984, "n_split": 2, "mean_delta_split": 0.07855, "n_merge": 1, "mean_delta_merge": -0.25354, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.88075, "trunc_abs_frac": 2.15529, "mean_delta_full": 0.08087, "read_hit_rate": {"surface": 0.34811, "split": 0.43084, "merge": 0.08766, "skel": 0.58408}, "mean_gate": {"surface": 0.14056, "split": 0.06017, "merge": 0.12981, "skel": 0.11534}, "n_reads": 10219, "n_cand": 27948, "read_suppressed_frac": 0.0927, "dropout_reads_logged": 10219, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1244, "admitted_surface": 1361, "staged_surface": 63020, "admitted_split": 1431, "staged_split": 46091, "admitted_merge": 4463, "staged_merge": 111222, "admitted_skel": 58281, "staged_skel": 41811, "mean_count": 569.331298828125, "n_util_obs": 498, "util_count_spearman": -0.0190551055863254}, "probe_overhead_sec_total": 535.066, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.005631, "alpha": 0.03, "t": 5804.6} |
| {"event": "consolidate", "step": 4250, "retired": 1, "split_retired": 0, "merge_admitted": 0, "admitted": 1, "swapped": 45, "uncertain_staged": 262117, "t": 5837.7} |
| {"event": "train", "step": 4250, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38856, "bpb": 2.00327, "chunk_loss": 0.030406, "F": 0.19641, "G": 0.2175, "chunks": 201.12, "bytes_per_chunk": 5.091, "boundary_entropy": 0.13691, "router_noise": 0.0, "grad_norm": 0.6841, "lr": 8.81e-05, "router_trainable": true, "sec_per_step": 0.681, "probe_sec": 0.0636, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0417, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": -0.00837, "n_shadow": 5, "mean_delta_shadow": 0.01604, "n_split": 1, "mean_delta_split": -0.0564, "n_merge": 2, "mean_delta_merge": -0.16046, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.3524, "trunc_abs_frac": 0.6681, "mean_delta_full": -0.02667, "read_hit_rate": {"surface": 0.35353, "split": 0.44219, "merge": 0.086, "skel": 0.59415}, "mean_gate": {"surface": 0.13784, "split": 0.05847, "merge": 0.12755, "skel": 0.11432}, "n_reads": 10925, "n_cand": 29373, "read_suppressed_frac": 0.0989, "dropout_reads_logged": 10925, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1258, "admitted_surface": 1368, "staged_surface": 62834, "admitted_split": 1437, "staged_split": 45834, "admitted_merge": 4463, "staged_merge": 111592, "admitted_skel": 58268, "staged_skel": 41884, "mean_count": 577.2846069335938, "n_util_obs": 506, "util_count_spearman": -0.011645364495393282}, "probe_overhead_sec_total": 538.578, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006213, "alpha": 0.03, "t": 5838.6} |
| {"event": "consolidate", "step": 4300, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 21, "uncertain_staged": 262117, "t": 5869.6} |
| {"event": "train", "step": 4300, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.39193, "bpb": 2.00813, "chunk_loss": 0.030253, "F": 0.18884, "G": 0.20904, "chunks": 193.38, "bytes_per_chunk": 5.295, "boundary_entropy": 0.13686, "router_noise": 0.0, "grad_norm": 0.7284, "lr": 8.5e-05, "router_trainable": true, "sec_per_step": 0.629, "probe_sec": 0.0756, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04736, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.01978, "n_shadow": 7, "mean_delta_shadow": -0.02795, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.68215, "trunc_abs_frac": 0.83348, "mean_delta_full": 0.10974, "read_hit_rate": {"surface": 0.34081, "split": 0.43946, "merge": 0.08377, "skel": 0.58272}, "mean_gate": {"surface": 0.13785, "split": 0.05786, "merge": 0.12391, "skel": 0.11304}, "n_reads": 10270, "n_cand": 28120, "read_suppressed_frac": 0.0951, "dropout_reads_logged": 10270, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1262, "admitted_surface": 1367, "staged_surface": 62718, "admitted_split": 1438, "staged_split": 46054, "admitted_merge": 4465, "staged_merge": 111364, "admitted_skel": 58266, "staged_skel": 42008, "mean_count": 585.3675537109375, "n_util_obs": 515, "util_count_spearman": -0.015397983374433615}, "probe_overhead_sec_total": 542.008, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006528, "alpha": 0.03, "t": 5870.1} |
| {"event": "consolidate", "step": 4350, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 46, "uncertain_staged": 262117, "t": 5899.6} |
| {"event": "train", "step": 4350, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.41149, "bpb": 2.03634, "chunk_loss": 0.030351, "F": 0.19458, "G": 0.2134, "chunks": 199.25, "bytes_per_chunk": 5.139, "boundary_entropy": 0.13091, "router_noise": 0.0, "grad_norm": 0.6743, "lr": 8.21e-05, "router_trainable": true, "sec_per_step": 0.599, "probe_sec": 0.0574, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03811, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.04167, "n_shadow": 5, "mean_delta_shadow": 0.0281, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.10199, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -5.49405, "trunc_abs_frac": 1.65578, "mean_delta_full": 0.00342, "read_hit_rate": {"surface": 0.34726, "split": 0.43647, "merge": 0.07811, "skel": 0.60429}, "mean_gate": {"surface": 0.13708, "split": 0.05646, "merge": 0.12692, "skel": 0.11095}, "n_reads": 10681, "n_cand": 29013, "read_suppressed_frac": 0.0944, "dropout_reads_logged": 10681, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1272, "admitted_surface": 1368, "staged_surface": 62804, "admitted_split": 1439, "staged_split": 45944, "admitted_merge": 4468, "staged_merge": 111787, "admitted_skel": 58261, "staged_skel": 41609, "mean_count": 593.151123046875, "n_util_obs": 517, "util_count_spearman": -0.01891919762288868}, "probe_overhead_sec_total": 545.251, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.005989, "alpha": 0.03, "t": 5900.0} |
| {"event": "consolidate", "step": 4400, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 25, "uncertain_staged": 262117, "t": 5930.2} |
| {"event": "probe", "step": 4400, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3275.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.188186, "cps": 25.585938, "mean_p": 0.205846, "boundary_entropy": 0.122784, "bpb": 1.922647}, "tr": {"chunks": 3334.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.184168, "cps": 26.046875, "mean_p": 0.199951, "boundary_entropy": 0.166303, "bpb": 2.067311}, "fi": {"chunks": 4122.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.219127, "cps": 32.203125, "mean_p": 0.226484, "boundary_entropy": 0.134713, "bpb": 2.175208}}, "bpb": {"en": 1.92265, "tr": 2.06731, "fi": 2.17521}, "gini_cpc": 0.039402, "gini_cps": 0.05262, "bpic": 5.286739, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.97865103544476, "fi": 1.164418025672002}, "inventory": {"size": 327653, "admitted": 65536, "staged": 262117, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1273, "admitted_surface": 1366, "staged_surface": 62623, "admitted_split": 1439, "staged_split": 46363, "admitted_merge": 4468, "staged_merge": 111211, "admitted_skel": 58263, "staged_skel": 41920, "mean_count": 600.5020141601562, "n_util_obs": 525, "util_count_spearman": -0.015127565715926295}, "t": 5931.5} |
| {"event": "eval", "step": 4400, "phase": 3, "bpb": {"en": 1.93631, "tr": 2.07402, "fi": 2.18246}, "bytes_per_chunk": {"en": 5.16, "tr": 5.813, "fi": 4.845}, "bpb_mean": 2.06426, "frozen_inventory": true, "t": 5933.7} |
| {"event": "train", "step": 4400, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.36539, "bpb": 1.96985, "chunk_loss": 0.030247, "F": 0.19025, "G": 0.20848, "chunks": 194.81, "bytes_per_chunk": 5.256, "boundary_entropy": 0.12642, "router_noise": 0.0, "grad_norm": 0.6447, "lr": 7.92e-05, "router_trainable": true, "sec_per_step": 0.761, "probe_sec": 0.0648, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04328, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.01248, "n_shadow": 6, "mean_delta_shadow": 0.02156, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -0.4906, "trunc_abs_frac": 0.80533, "mean_delta_full": 0.00325, "read_hit_rate": {"surface": 0.35424, "split": 0.42115, "merge": 0.0872, "skel": 0.59457}, "mean_gate": {"surface": 0.13839, "split": 0.05745, "merge": 0.12409, "skel": 0.11351}, "n_reads": 10390, "n_cand": 28502, "read_suppressed_frac": 0.0964, "dropout_reads_logged": 10390, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1272, "admitted_surface": 1366, "staged_surface": 62640, "admitted_split": 1439, "staged_split": 46300, "admitted_merge": 4468, "staged_merge": 111296, "admitted_skel": 58263, "staged_skel": 41908, "mean_count": 600.6605224609375, "n_util_obs": 525, "util_count_spearman": -0.015217877924723044}, "probe_overhead_sec_total": 548.445, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006188, "alpha": 0.03, "t": 5938.1} |
| {"event": "consolidate", "step": 4450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 32, "uncertain_staged": 262117, "t": 5968.5} |
| {"event": "train", "step": 4450, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.36208, "bpb": 1.96506, "chunk_loss": 0.030225, "F": 0.18607, "G": 0.20545, "chunks": 190.53, "bytes_per_chunk": 5.374, "boundary_entropy": 0.13107, "router_noise": 0.0, "grad_norm": 0.6945, "lr": 7.63e-05, "router_trainable": true, "sec_per_step": 0.616, "probe_sec": 0.0637, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0395, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.04769, "n_shadow": 5, "mean_delta_shadow": -0.0011, "n_split": 1, "mean_delta_split": -0.12476, "n_merge": 1, "mean_delta_merge": -0.08435, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.9031, "trunc_abs_frac": 3.14988, "mean_delta_full": -0.08191, "read_hit_rate": {"surface": 0.3496, "split": 0.42582, "merge": 0.08887, "skel": 0.58199}, "mean_gate": {"surface": 0.13984, "split": 0.05845, "merge": 0.12737, "skel": 0.11639}, "n_reads": 10017, "n_cand": 27626, "read_suppressed_frac": 0.0957, "dropout_reads_logged": 10017, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1283, "admitted_surface": 1368, "staged_surface": 62707, "admitted_split": 1442, "staged_split": 45941, "admitted_merge": 4468, "staged_merge": 111433, "admitted_skel": 58258, "staged_skel": 42063, "mean_count": 608.8529663085938, "n_util_obs": 530, "util_count_spearman": -0.008536742076234253}, "probe_overhead_sec_total": 551.712, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006328, "alpha": 0.03, "t": 5968.9} |
| {"event": "consolidate", "step": 4500, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 33, "uncertain_staged": 262117, "t": 5999.0} |
| {"event": "train", "step": 4500, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.33555, "bpb": 1.9268, "chunk_loss": 0.030327, "F": 0.19601, "G": 0.21294, "chunks": 200.72, "bytes_per_chunk": 5.102, "boundary_entropy": 0.12597, "router_noise": 0.0, "grad_norm": 0.6572, "lr": 7.36e-05, "router_trainable": true, "sec_per_step": 0.61, "probe_sec": 0.0648, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04224, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.0406, "n_shadow": 6, "mean_delta_shadow": -0.00257, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.05127, "n_dropped_unpaired": 0, "n_full": 7, "trunc_ratio_sum": -0.03083, "trunc_abs_frac": 0.8437, "mean_delta_full": -0.24976, "read_hit_rate": {"surface": 0.35128, "split": 0.45696, "merge": 0.08508, "skel": 0.60491}, "mean_gate": {"surface": 0.13344, "split": 0.05482, "merge": 0.12043, "skel": 0.10752}, "n_reads": 11168, "n_cand": 29431, "read_suppressed_frac": 0.1079, "dropout_reads_logged": 11168, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1295, "admitted_surface": 1367, "staged_surface": 62759, "admitted_split": 1446, "staged_split": 46001, "admitted_merge": 4469, "staged_merge": 111402, "admitted_skel": 58254, "staged_skel": 41982, "mean_count": 616.7590942382812, "n_util_obs": 534, "util_count_spearman": -0.004786121924984346}, "probe_overhead_sec_total": 555.544, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006443, "alpha": 0.03, "t": 5999.4} |
| {"event": "consolidate", "step": 4550, "retired": 2, "split_retired": 1, "merge_admitted": 0, "admitted": 3, "swapped": 37, "uncertain_staged": 262117, "t": 6030.8} |
| {"event": "train", "step": 4550, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.36794, "bpb": 1.97353, "chunk_loss": 0.030237, "F": 0.18826, "G": 0.20899, "chunks": 192.78, "bytes_per_chunk": 5.312, "boundary_entropy": 0.1313, "router_noise": 0.0, "grad_norm": 0.6721, "lr": 7.09e-05, "router_trainable": true, "sec_per_step": 0.64, "probe_sec": 0.0674, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04274, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.02483, "n_shadow": 5, "mean_delta_shadow": 0.04427, "n_split": 1, "mean_delta_split": -0.1167, "n_merge": 1, "mean_delta_merge": -1.26831, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 1.36752, "trunc_abs_frac": 2.30363, "mean_delta_full": 0.14996, "read_hit_rate": {"surface": 0.33795, "split": 0.42642, "merge": 0.08337, "skel": 0.56798}, "mean_gate": {"surface": 0.1367, "split": 0.0561, "merge": 0.12315, "skel": 0.11168}, "n_reads": 10138, "n_cand": 28311, "read_suppressed_frac": 0.1013, "dropout_reads_logged": 10138, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1308, "admitted_surface": 1368, "staged_surface": 62726, "admitted_split": 1463, "staged_split": 45738, "admitted_merge": 4471, "staged_merge": 111682, "admitted_skel": 58234, "staged_skel": 41998, "mean_count": 624.7330322265625, "n_util_obs": 542, "util_count_spearman": -0.018888560683784906}, "probe_overhead_sec_total": 558.99, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006256, "alpha": 0.03, "t": 6031.4} |
| {"event": "consolidate", "step": 4600, "retired": 1, "split_retired": 0, "merge_admitted": 0, "admitted": 1, "swapped": 88, "uncertain_staged": 262117, "t": 6062.7} |
| {"event": "probe", "step": 4600, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3317.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.190599, "cps": 25.914062, "mean_p": 0.207546, "boundary_entropy": 0.121978, "bpb": 1.913443}, "tr": {"chunks": 3379.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.186654, "cps": 26.398438, "mean_p": 0.199984, "boundary_entropy": 0.159057, "bpb": 2.057733}, "fi": {"chunks": 4227.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.224709, "cps": 33.023438, "mean_p": 0.230682, "boundary_entropy": 0.131345, "bpb": 2.16155}}, "bpb": {"en": 1.91344, "tr": 2.05773, "fi": 2.16155}, "gini_cpc": 0.042145, "gini_cps": 0.05554, "bpic": 5.193811, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9793012053044339, "fi": 1.1789598440684268}, "inventory": {"size": 327591, "admitted": 65536, "staged": 262055, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1332, "admitted_surface": 1373, "staged_surface": 62586, "admitted_split": 1499, "staged_split": 45801, "admitted_merge": 4472, "staged_merge": 111836, "admitted_skel": 58192, "staged_skel": 41832, "mean_count": 632.7537841796875, "n_util_obs": 545, "util_count_spearman": -0.02195948991907039}, "t": 6064.0} |
| {"event": "eval", "step": 4600, "phase": 3, "bpb": {"en": 1.92683, "tr": 2.06351, "fi": 2.16762}, "bytes_per_chunk": {"en": 5.116, "tr": 5.748, "fi": 4.731}, "bpb_mean": 2.05266, "frozen_inventory": true, "t": 6066.2} |
| {"event": "train", "step": 4600, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.3658, "bpb": 1.97044, "chunk_loss": 0.030472, "F": 0.20206, "G": 0.21886, "chunks": 206.91, "bytes_per_chunk": 4.949, "boundary_entropy": 0.12802, "router_noise": 0.0, "grad_norm": 0.6855, "lr": 6.82e-05, "router_trainable": true, "sec_per_step": 0.783, "probe_sec": 0.0676, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04391, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": -0.05826, "n_shadow": 5, "mean_delta_shadow": 0.0837, "n_split": 2, "mean_delta_split": 0.03784, "n_merge": 1, "mean_delta_merge": -0.1626, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.29645, "trunc_abs_frac": 4.18335, "mean_delta_full": 0.12417, "read_hit_rate": {"surface": 0.38033, "split": 0.45121, "merge": 0.09654, "skel": 0.61066}, "mean_gate": {"surface": 0.13364, "split": 0.05597, "merge": 0.1194, "skel": 0.11148}, "n_reads": 11497, "n_cand": 29846, "read_suppressed_frac": 0.1007, "dropout_reads_logged": 11497, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1333, "admitted_surface": 1373, "staged_surface": 62633, "admitted_split": 1499, "staged_split": 45790, "admitted_merge": 4472, "staged_merge": 111838, "admitted_skel": 58192, "staged_skel": 41883, "mean_count": 632.92919921875, "n_util_obs": 545, "util_count_spearman": -0.0214777203242485}, "probe_overhead_sec_total": 562.336, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006293, "alpha": 0.03, "t": 6070.6} |
| {"event": "consolidate", "step": 4650, "retired": 1, "split_retired": 2, "merge_admitted": 0, "admitted": 3, "swapped": 189, "uncertain_staged": 262116, "t": 6101.1} |
| {"event": "train", "step": 4650, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.39625, "bpb": 2.01436, "chunk_loss": 0.030365, "F": 0.194, "G": 0.21394, "chunks": 198.66, "bytes_per_chunk": 5.155, "boundary_entropy": 0.13024, "router_noise": 0.0, "grad_norm": 0.6839, "lr": 6.57e-05, "router_trainable": true, "sec_per_step": 0.62, "probe_sec": 0.0621, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04071, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.05545, "n_shadow": 4, "mean_delta_shadow": 0.00093, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -0.19173, "trunc_abs_frac": 1.65289, "mean_delta_full": -0.06379, "read_hit_rate": {"surface": 0.33312, "split": 0.44075, "merge": 0.08215, "skel": 0.57864}, "mean_gate": {"surface": 0.13263, "split": 0.05526, "merge": 0.12221, "skel": 0.11094}, "n_reads": 10381, "n_cand": 28674, "read_suppressed_frac": 0.0992, "dropout_reads_logged": 10381, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1324, "admitted_surface": 1381, "staged_surface": 62959, "admitted_split": 1586, "staged_split": 45784, "admitted_merge": 4473, "staged_merge": 111398, "admitted_skel": 58096, "staged_skel": 42003, "mean_count": 640.7261962890625, "n_util_obs": 545, "util_count_spearman": -0.02297488459693843}, "probe_overhead_sec_total": 565.725, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006111, "alpha": 0.03, "t": 6101.6} |
| {"event": "consolidate", "step": 4700, "retired": 0, "split_retired": 4, "merge_admitted": 0, "admitted": 4, "swapped": 290, "uncertain_staged": 262117, "t": 6133.4} |
| {"event": "train", "step": 4700, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.42248, "bpb": 2.05221, "chunk_loss": 0.030615, "F": 0.20428, "G": 0.22079, "chunks": 209.19, "bytes_per_chunk": 4.895, "boundary_entropy": 0.12426, "router_noise": 0.0, "grad_norm": 0.6833, "lr": 6.32e-05, "router_trainable": true, "sec_per_step": 0.655, "probe_sec": 0.0634, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04214, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 2, "mean_delta_loo": 0.14285, "n_shadow": 4, "mean_delta_shadow": 0.0234, "n_split": 0, "mean_delta_split": NaN, "n_merge": 4, "mean_delta_merge": -0.09467, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 2.21067, "trunc_abs_frac": 1.7585, "mean_delta_full": -0.01785, "read_hit_rate": {"surface": 0.37721, "split": 0.45664, "merge": 0.09502, "skel": 0.606}, "mean_gate": {"surface": 0.12945, "split": 0.05398, "merge": 0.11457, "skel": 0.10599}, "n_reads": 11575, "n_cand": 30077, "read_suppressed_frac": 0.1006, "dropout_reads_logged": 11575, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1328, "admitted_surface": 1416, "staged_surface": 62881, "admitted_split": 1787, "staged_split": 45590, "admitted_merge": 4477, "staged_merge": 111914, "admitted_skel": 57856, "staged_skel": 41759, "mean_count": 648.1238403320312, "n_util_obs": 555, "util_count_spearman": -0.020715234933355607}, "probe_overhead_sec_total": 569.145, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006326, "alpha": 0.03, "t": 6134.3} |
| {"event": "consolidate", "step": 4750, "retired": 1, "split_retired": 0, "merge_admitted": 0, "admitted": 1, "swapped": 155, "uncertain_staged": 262117, "t": 6168.3} |
| {"event": "train", "step": 4750, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.413, "bpb": 2.03853, "chunk_loss": 0.030391, "F": 0.19794, "G": 0.21545, "chunks": 202.69, "bytes_per_chunk": 5.052, "boundary_entropy": 0.12722, "router_noise": 0.0, "grad_norm": 0.7002, "lr": 6.08e-05, "router_trainable": true, "sec_per_step": 0.69, "probe_sec": 0.066, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04337, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": 0.01208, "n_shadow": 9, "mean_delta_shadow": -0.07587, "n_split": 1, "mean_delta_split": 0.00125, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 1.13101, "trunc_abs_frac": 1.13101, "mean_delta_full": -0.27161, "read_hit_rate": {"surface": 0.36241, "split": 0.46227, "merge": 0.08798, "skel": 0.59103}, "mean_gate": {"surface": 0.13378, "split": 0.0549, "merge": 0.11848, "skel": 0.10934}, "n_reads": 11177, "n_cand": 29362, "read_suppressed_frac": 0.1022, "dropout_reads_logged": 11177, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1337, "admitted_surface": 1432, "staged_surface": 62828, "admitted_split": 1885, "staged_split": 45681, "admitted_merge": 4476, "staged_merge": 111501, "admitted_skel": 57743, "staged_skel": 42134, "mean_count": 656.5406494140625, "n_util_obs": 561, "util_count_spearman": -0.021957691444062014}, "probe_overhead_sec_total": 572.674, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006266, "alpha": 0.03, "t": 6168.8} |
| {"event": "consolidate", "step": 4800, "retired": 1, "split_retired": 2, "merge_admitted": 0, "admitted": 3, "swapped": 133, "uncertain_staged": 262117, "t": 6206.3} |
| {"event": "probe", "step": 4800, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3274.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.188128, "cps": 25.578125, "mean_p": 0.205173, "boundary_entropy": 0.119935, "bpb": 1.905484}, "tr": {"chunks": 3396.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.187593, "cps": 26.53125, "mean_p": 0.199207, "boundary_entropy": 0.158945, "bpb": 2.042992}, "fi": {"chunks": 4136.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.219871, "cps": 32.3125, "mean_p": 0.226428, "boundary_entropy": 0.127559, "bpb": 2.150009}}, "bpb": {"en": 1.90548, "tr": 2.04299, "fi": 2.15001}, "gini_cpc": 0.03613, "gini_cps": 0.05318, "bpic": 5.250046, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9971547795920115, "fi": 1.1687297301676935}, "inventory": {"size": 327544, "admitted": 65536, "staged": 262008, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1337, "admitted_surface": 1452, "staged_surface": 62901, "admitted_split": 1976, "staged_split": 45604, "admitted_merge": 4477, "staged_merge": 111814, "admitted_skel": 57631, "staged_skel": 41689, "mean_count": 664.6360473632812, "n_util_obs": 569, "util_count_spearman": -0.032331768520118855}, "t": 6207.6} |
| {"event": "eval", "step": 4800, "phase": 3, "bpb": {"en": 1.92044, "tr": 2.04967, "fi": 2.15573}, "bytes_per_chunk": {"en": 5.163, "tr": 5.757, "fi": 4.833}, "bpb_mean": 2.04195, "frozen_inventory": true, "t": 6209.7} |
| {"event": "train", "step": 4800, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.3812, "bpb": 1.99265, "chunk_loss": 0.030391, "F": 0.19687, "G": 0.21392, "chunks": 201.59, "bytes_per_chunk": 5.08, "boundary_entropy": 0.1236, "router_noise": 0.0, "grad_norm": 0.669, "lr": 5.85e-05, "router_trainable": true, "sec_per_step": 0.914, "probe_sec": 0.0733, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05059, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": 0.00236, "n_shadow": 3, "mean_delta_shadow": -0.03503, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.20135, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -0.0291, "trunc_abs_frac": 3.04571, "mean_delta_full": -0.14116, "read_hit_rate": {"surface": 0.37778, "split": 0.48161, "merge": 0.08909, "skel": 0.60218}, "mean_gate": {"surface": 0.12976, "split": 0.05241, "merge": 0.11916, "skel": 0.10526}, "n_reads": 11482, "n_cand": 29166, "read_suppressed_frac": 0.1011, "dropout_reads_logged": 11482, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1337, "admitted_surface": 1452, "staged_surface": 62875, "admitted_split": 1976, "staged_split": 45673, "admitted_merge": 4477, "staged_merge": 111865, "admitted_skel": 57631, "staged_skel": 41731, "mean_count": 664.811279296875, "n_util_obs": 569, "util_count_spearman": -0.03242081344543465}, "probe_overhead_sec_total": 576.397, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006048, "alpha": 0.03, "t": 6214.5} |
| {"event": "consolidate", "step": 4850, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 166, "uncertain_staged": 262117, "t": 6251.5} |
| {"event": "train", "step": 4850, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35682, "bpb": 1.95748, "chunk_loss": 0.03034, "F": 0.19385, "G": 0.21126, "chunks": 198.5, "bytes_per_chunk": 5.159, "boundary_entropy": 0.1265, "router_noise": 0.0, "grad_norm": 0.7184, "lr": 5.62e-05, "router_trainable": true, "sec_per_step": 0.747, "probe_sec": 0.0655, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0426, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": -0.03517, "n_shadow": 6, "mean_delta_shadow": 0.03098, "n_split": 1, "mean_delta_split": 0.15491, "n_merge": 1, "mean_delta_merge": -0.08569, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 3.40519, "trunc_abs_frac": 0.77695, "mean_delta_full": -0.0144, "read_hit_rate": {"surface": 0.35918, "split": 0.48221, "merge": 0.08238, "skel": 0.59061}, "mean_gate": {"surface": 0.13004, "split": 0.05352, "merge": 0.1187, "skel": 0.10626}, "n_reads": 11077, "n_cand": 28696, "read_suppressed_frac": 0.1035, "dropout_reads_logged": 11077, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1359, "admitted_surface": 1457, "staged_surface": 62863, "admitted_split": 2118, "staged_split": 45416, "admitted_merge": 4481, "staged_merge": 112046, "admitted_skel": 57480, "staged_skel": 41819, "mean_count": 672.7343139648438, "n_util_obs": 574, "util_count_spearman": -0.03564308607737937}, "probe_overhead_sec_total": 579.9, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006763, "alpha": 0.03, "t": 6251.9} |
| {"event": "consolidate", "step": 4900, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 127, "uncertain_staged": 262117, "t": 6287.6} |
| {"event": "train", "step": 4900, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38542, "bpb": 1.99873, "chunk_loss": 0.030332, "F": 0.18951, "G": 0.20657, "chunks": 194.06, "bytes_per_chunk": 5.277, "boundary_entropy": 0.12256, "router_noise": 0.0, "grad_norm": 0.682, "lr": 5.41e-05, "router_trainable": true, "sec_per_step": 0.733, "probe_sec": 0.0681, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04032, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": 0.00787, "n_shadow": 5, "mean_delta_shadow": 0.03661, "n_split": 2, "mean_delta_split": 0.17912, "n_merge": 1, "mean_delta_merge": 0.22156, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.64108, "trunc_abs_frac": 2.73996, "mean_delta_full": -0.08447, "read_hit_rate": {"surface": 0.33538, "split": 0.47652, "merge": 0.07078, "skel": 0.5865}, "mean_gate": {"surface": 0.12614, "split": 0.05132, "merge": 0.12147, "skel": 0.10276}, "n_reads": 10591, "n_cand": 28160, "read_suppressed_frac": 0.1037, "dropout_reads_logged": 10591, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1371, "admitted_surface": 1460, "staged_surface": 63047, "admitted_split": 2229, "staged_split": 45472, "admitted_merge": 4481, "staged_merge": 112052, "admitted_skel": 57366, "staged_skel": 41573, "mean_count": 681.27392578125, "n_util_obs": 578, "util_count_spearman": -0.030732930560616194}, "probe_overhead_sec_total": 583.468, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.005875, "alpha": 0.03, "t": 6288.5} |
| {"event": "consolidate", "step": 4950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 185, "uncertain_staged": 262117, "t": 6323.4} |
| {"event": "train", "step": 4950, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.34663, "bpb": 1.94278, "chunk_loss": 0.030337, "F": 0.19235, "G": 0.2105, "chunks": 196.97, "bytes_per_chunk": 5.199, "boundary_entropy": 0.12831, "router_noise": 0.0, "grad_norm": 0.6821, "lr": 5.2e-05, "router_trainable": true, "sec_per_step": 0.707, "probe_sec": 0.0674, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04004, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.01553, "n_shadow": 4, "mean_delta_shadow": -0.15487, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 1.61104, "trunc_abs_frac": 1.39218, "mean_delta_full": 0.01257, "read_hit_rate": {"surface": 0.37809, "split": 0.4766, "merge": 0.0928, "skel": 0.58535}, "mean_gate": {"surface": 0.13181, "split": 0.0532, "merge": 0.11591, "skel": 0.10856}, "n_reads": 11094, "n_cand": 28495, "read_suppressed_frac": 0.1007, "dropout_reads_logged": 11094, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1371, "admitted_surface": 1464, "staged_surface": 62924, "admitted_split": 2364, "staged_split": 45178, "admitted_merge": 4483, "staged_merge": 112198, "admitted_skel": 57225, "staged_skel": 41844, "mean_count": 690.126708984375, "n_util_obs": 584, "util_count_spearman": -0.03930435034079387}, "probe_overhead_sec_total": 586.938, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.005811, "alpha": 0.03, "t": 6323.9} |
| {"event": "consolidate", "step": 5000, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 50, "uncertain_staged": 262117, "t": 6362.3} |
| {"event": "probe", "step": 5000, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3305.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.18991, "cps": 25.820312, "mean_p": 0.206208, "boundary_entropy": 0.117847, "bpb": 1.897016}, "tr": {"chunks": 3364.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.185826, "cps": 26.28125, "mean_p": 0.199124, "boundary_entropy": 0.156823, "bpb": 2.037619}, "fi": {"chunks": 4195.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.223008, "cps": 32.773438, "mean_p": 0.228671, "boundary_entropy": 0.12651, "bpb": 2.139097}}, "bpb": {"en": 1.89702, "tr": 2.03762, "fi": 2.1391}, "gini_cpc": 0.0414, "gini_cps": 0.054615, "bpic": 5.222018, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9784938312729404, "fi": 1.1742829038051978}, "inventory": {"size": 327629, "admitted": 65536, "staged": 262093, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1388, "admitted_surface": 1465, "staged_surface": 62658, "admitted_split": 2398, "staged_split": 44981, "admitted_merge": 4486, "staged_merge": 112864, "admitted_skel": 57187, "staged_skel": 41590, "mean_count": 698.4751586914062, "n_util_obs": 589, "util_count_spearman": -0.041544071329502434}, "t": 6363.7} |
| {"event": "eval", "step": 5000, "phase": 3, "bpb": {"en": 1.91194, "tr": 2.04486, "fi": 2.145}, "bytes_per_chunk": {"en": 5.125, "tr": 5.778, "fi": 4.771}, "bpb_mean": 2.03393, "frozen_inventory": true, "t": 6365.9} |
| {"event": "train", "step": 5000, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.33094, "bpb": 1.92014, "chunk_loss": 0.030341, "F": 0.19571, "G": 0.21202, "chunks": 200.41, "bytes_per_chunk": 5.11, "boundary_entropy": 0.12179, "router_noise": 0.0, "grad_norm": 0.6214, "lr": 5e-05, "router_trainable": true, "sec_per_step": 0.936, "probe_sec": 0.0781, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04821, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": -0.04066, "n_shadow": 7, "mean_delta_shadow": -0.07877, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.08014, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.04658, "trunc_abs_frac": 0.04658, "mean_delta_full": -0.33154, "read_hit_rate": {"surface": 0.38709, "split": 0.49653, "merge": 0.09214, "skel": 0.58941}, "mean_gate": {"surface": 0.12274, "split": 0.05004, "merge": 0.11339, "skel": 0.10005}, "n_reads": 11568, "n_cand": 28946, "read_suppressed_frac": 0.1002, "dropout_reads_logged": 11568, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1387, "admitted_surface": 1465, "staged_surface": 62671, "admitted_split": 2398, "staged_split": 44973, "admitted_merge": 4486, "staged_merge": 112888, "admitted_skel": 57187, "staged_skel": 41612, "mean_count": 698.6516723632812, "n_util_obs": 589, "util_count_spearman": -0.04169053484425547}, "probe_overhead_sec_total": 590.653, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.00625, "alpha": 0.03, "t": 6370.7} |
| {"event": "consolidate", "step": 5050, "retired": 1, "split_retired": 1, "merge_admitted": 0, "admitted": 2, "swapped": 71, "uncertain_staged": 262117, "t": 6407.0} |
| {"event": "train", "step": 5050, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35078, "bpb": 1.94876, "chunk_loss": 0.030423, "F": 0.19867, "G": 0.21513, "chunks": 203.44, "bytes_per_chunk": 5.033, "boundary_entropy": 0.12264, "router_noise": 0.0, "grad_norm": 0.6724, "lr": 4.81e-05, "router_trainable": true, "sec_per_step": 0.742, "probe_sec": 0.0656, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04119, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": 0.0763, "n_shadow": 7, "mean_delta_shadow": 0.12337, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": 0.60709, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.51366, "trunc_abs_frac": 0.44027, "mean_delta_full": 0.25638, "read_hit_rate": {"surface": 0.38901, "split": 0.5082, "merge": 0.0951, "skel": 0.60598}, "mean_gate": {"surface": 0.12841, "split": 0.05151, "merge": 0.11589, "skel": 0.10428}, "n_reads": 12120, "n_cand": 29624, "read_suppressed_frac": 0.1029, "dropout_reads_logged": 12120, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1411, "admitted_surface": 1465, "staged_surface": 62920, "admitted_split": 2453, "staged_split": 44880, "admitted_merge": 4489, "staged_merge": 112452, "admitted_skel": 57129, "staged_skel": 41892, "mean_count": 713.685546875, "n_util_obs": 591, "util_count_spearman": -0.03765524321533958}, "probe_overhead_sec_total": 594.047, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006065, "alpha": 0.03, "t": 6407.7} |
| {"event": "consolidate", "step": 5100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 72, "uncertain_staged": 262118, "t": 6442.6} |
| {"event": "train", "step": 5100, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38588, "bpb": 1.9994, "chunk_loss": 0.030417, "F": 0.19653, "G": 0.21434, "chunks": 201.25, "bytes_per_chunk": 5.088, "boundary_entropy": 0.12607, "router_noise": 0.0, "grad_norm": 0.7118, "lr": 4.63e-05, "router_trainable": true, "sec_per_step": 0.712, "probe_sec": 0.0717, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04259, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.05052, "n_shadow": 3, "mean_delta_shadow": -0.15381, "n_split": 1, "mean_delta_split": 0.13623, "n_merge": 1, "mean_delta_merge": 0.03308, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.01541, "trunc_abs_frac": 1.1844, "mean_delta_full": 0.20303, "read_hit_rate": {"surface": 0.37516, "split": 0.48952, "merge": 0.08642, "skel": 0.57958}, "mean_gate": {"surface": 0.12844, "split": 0.05169, "merge": 0.11571, "skel": 0.10696}, "n_reads": 11309, "n_cand": 28947, "read_suppressed_frac": 0.0964, "dropout_reads_logged": 11309, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1427, "admitted_surface": 1467, "staged_surface": 62879, "admitted_split": 2509, "staged_split": 44847, "admitted_merge": 4491, "staged_merge": 112712, "admitted_skel": 57069, "staged_skel": 41706, "mean_count": 722.607177734375, "n_util_obs": 597, "util_count_spearman": -0.028815206964059013}, "probe_overhead_sec_total": 597.459, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006563, "alpha": 0.03, "t": 6443.3} |
| {"event": "consolidate", "step": 5150, "retired": 0, "split_retired": 3, "merge_admitted": 0, "admitted": 3, "swapped": 39, "uncertain_staged": 262118, "t": 6479.8} |
| {"event": "train", "step": 5150, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.3722, "bpb": 1.97966, "chunk_loss": 0.030541, "F": 0.20218, "G": 0.21827, "chunks": 207.03, "bytes_per_chunk": 4.946, "boundary_entropy": 0.12236, "router_noise": 0.0, "grad_norm": 0.7084, "lr": 4.45e-05, "router_trainable": true, "sec_per_step": 0.738, "probe_sec": 0.0649, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04267, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.02924, "n_shadow": 6, "mean_delta_shadow": -0.32413, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.27209, "trunc_abs_frac": 0.61933, "mean_delta_full": 0.13171, "read_hit_rate": {"surface": 0.38981, "split": 0.49606, "merge": 0.0884, "skel": 0.60052}, "mean_gate": {"surface": 0.12661, "split": 0.0508, "merge": 0.11591, "skel": 0.103}, "n_reads": 11792, "n_cand": 29532, "read_suppressed_frac": 0.0968, "dropout_reads_logged": 11792, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1423, "admitted_surface": 1466, "staged_surface": 62947, "admitted_split": 2543, "staged_split": 44858, "admitted_merge": 4491, "staged_merge": 112526, "admitted_skel": 57036, "staged_skel": 41813, "mean_count": 730.0638427734375, "n_util_obs": 600, "util_count_spearman": -0.029663027253728862}, "probe_overhead_sec_total": 600.933, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006541, "alpha": 0.03, "t": 6480.2} |
| {"event": "consolidate", "step": 5200, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 123, "uncertain_staged": 262118, "t": 6510.7} |
| {"event": "probe", "step": 5200, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3303.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.189795, "cps": 25.804688, "mean_p": 0.20523, "boundary_entropy": 0.114327, "bpb": 1.892084}, "tr": {"chunks": 3421.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.188974, "cps": 26.726562, "mean_p": 0.198181, "boundary_entropy": 0.153253, "bpb": 2.029644}, "fi": {"chunks": 4162.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.221254, "cps": 32.515625, "mean_p": 0.226978, "boundary_entropy": 0.121147, "bpb": 2.131424}}, "bpb": {"en": 1.89208, "tr": 2.02964, "fi": 2.13142}, "gini_cpc": 0.035865, "gini_cps": 0.052606, "bpic": 5.211464, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.995676069567205, "fi": 1.165750845049742}, "inventory": {"size": 327555, "admitted": 65536, "staged": 262019, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1435, "admitted_surface": 1465, "staged_surface": 62834, "admitted_split": 2653, "staged_split": 44788, "admitted_merge": 4493, "staged_merge": 112515, "admitted_skel": 56925, "staged_skel": 41882, "mean_count": 738.3915405273438, "n_util_obs": 606, "util_count_spearman": -0.031623424217264426}, "t": 6512.0} |
| {"event": "eval", "step": 5200, "phase": 3, "bpb": {"en": 1.90733, "tr": 2.03635, "fi": 2.13646}, "bytes_per_chunk": {"en": 5.124, "tr": 5.736, "fi": 4.797}, "bpb_mean": 2.02671, "frozen_inventory": true, "t": 6514.1} |
| {"event": "train", "step": 5200, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32834, "bpb": 1.91638, "chunk_loss": 0.030293, "F": 0.19342, "G": 0.20824, "chunks": 198.06, "bytes_per_chunk": 5.17, "boundary_entropy": 0.11633, "router_noise": 0.0, "grad_norm": 0.6914, "lr": 4.29e-05, "router_trainable": true, "sec_per_step": 0.755, "probe_sec": 0.0612, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04029, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": -0.08882, "n_shadow": 4, "mean_delta_shadow": 0.13162, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.01782, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -0.30191, "trunc_abs_frac": 0.4376, "mean_delta_full": -0.03005, "read_hit_rate": {"surface": 0.3687, "split": 0.50389, "merge": 0.09117, "skel": 0.61727}, "mean_gate": {"surface": 0.12235, "split": 0.04949, "merge": 0.11002, "skel": 0.09911}, "n_reads": 11760, "n_cand": 29015, "read_suppressed_frac": 0.1, "dropout_reads_logged": 11760, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1435, "admitted_surface": 1465, "staged_surface": 62934, "admitted_split": 2653, "staged_split": 44821, "admitted_merge": 4493, "staged_merge": 112504, "admitted_skel": 56925, "staged_skel": 41885, "mean_count": 738.5709838867188, "n_util_obs": 606, "util_count_spearman": -0.03225851589398359}, "probe_overhead_sec_total": 604.139, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.005889, "alpha": 0.03, "t": 6518.0} |
| {"event": "consolidate", "step": 5250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 51, "uncertain_staged": 262118, "t": 6547.3} |
| {"event": "train", "step": 5250, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32713, "bpb": 1.91465, "chunk_loss": 0.030323, "F": 0.19275, "G": 0.21039, "chunks": 197.38, "bytes_per_chunk": 5.188, "boundary_entropy": 0.12028, "router_noise": 0.0, "grad_norm": 0.6655, "lr": 4.14e-05, "router_trainable": true, "sec_per_step": 0.595, "probe_sec": 0.0673, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04371, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.02004, "n_shadow": 4, "mean_delta_shadow": -0.1012, "n_split": 2, "mean_delta_split": 0.03796, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.76441, "trunc_abs_frac": 1.83815, "mean_delta_full": -0.03918, "read_hit_rate": {"surface": 0.3816, "split": 0.51892, "merge": 0.09181, "skel": 0.59433}, "mean_gate": {"surface": 0.12489, "split": 0.04985, "merge": 0.11385, "skel": 0.10054}, "n_reads": 11765, "n_cand": 28753, "read_suppressed_frac": 0.102, "dropout_reads_logged": 11765, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1452, "admitted_surface": 1470, "staged_surface": 63065, "admitted_split": 2693, "staged_split": 44974, "admitted_merge": 4495, "staged_merge": 111983, "admitted_skel": 56878, "staged_skel": 42122, "mean_count": 747.5083618164062, "n_util_obs": 607, "util_count_spearman": -0.03419532747373059}, "probe_overhead_sec_total": 607.427, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006255, "alpha": 0.03, "t": 6547.7} |
| {"event": "consolidate", "step": 5300, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 105, "uncertain_staged": 262118, "t": 6578.2} |
| {"event": "train", "step": 5300, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38026, "bpb": 1.9913, "chunk_loss": 0.030357, "F": 0.19592, "G": 0.21197, "chunks": 200.62, "bytes_per_chunk": 5.104, "boundary_entropy": 0.11911, "router_noise": 0.0, "grad_norm": 0.6944, "lr": 3.99e-05, "router_trainable": true, "sec_per_step": 0.616, "probe_sec": 0.0525, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03428, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.03631, "n_shadow": 4, "mean_delta_shadow": -0.05997, "n_split": 1, "mean_delta_split": -0.19604, "n_merge": 1, "mean_delta_merge": -0.2254, "n_dropped_unpaired": 0, "n_full": 0, "read_hit_rate": {"surface": 0.36709, "split": 0.49956, "merge": 0.08244, "skel": 0.5989}, "mean_gate": {"surface": 0.12427, "split": 0.04982, "merge": 0.11415, "skel": 0.10197}, "n_reads": 11573, "n_cand": 29159, "read_suppressed_frac": 0.0969, "dropout_reads_logged": 11573, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1466, "admitted_surface": 1471, "staged_surface": 62976, "admitted_split": 2781, "staged_split": 44762, "admitted_merge": 4495, "staged_merge": 112422, "admitted_skel": 56789, "staged_skel": 41984, "mean_count": 756.0698852539062, "n_util_obs": 616, "util_count_spearman": -0.029448003027205766}, "probe_overhead_sec_total": 610.714, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.0058, "alpha": 0.03, "t": 6578.6} |
| {"event": "consolidate", "step": 5350, "retired": 1, "split_retired": 1, "merge_admitted": 0, "admitted": 2, "swapped": 160, "uncertain_staged": 262118, "t": 6606.6} |
| {"event": "train", "step": 5350, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.31408, "bpb": 1.89582, "chunk_loss": 0.030213, "F": 0.18802, "G": 0.205, "chunks": 192.53, "bytes_per_chunk": 5.319, "boundary_entropy": 0.11796, "router_noise": 0.0, "grad_norm": 0.7524, "lr": 3.86e-05, "router_trainable": true, "sec_per_step": 0.568, "probe_sec": 0.0649, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04302, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.0819, "n_shadow": 3, "mean_delta_shadow": -0.06729, "n_split": 1, "mean_delta_split": 0.05298, "n_merge": 1, "mean_delta_merge": -0.2076, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.24742, "trunc_abs_frac": 0.79675, "mean_delta_full": 0.24359, "read_hit_rate": {"surface": 0.36205, "split": 0.51861, "merge": 0.08906, "skel": 0.608}, "mean_gate": {"surface": 0.12183, "split": 0.0484, "merge": 0.11387, "skel": 0.09973}, "n_reads": 11544, "n_cand": 28298, "read_suppressed_frac": 0.0978, "dropout_reads_logged": 11544, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1462, "admitted_surface": 1472, "staged_surface": 63008, "admitted_split": 2910, "staged_split": 44634, "admitted_merge": 4495, "staged_merge": 112667, "admitted_skel": 56659, "staged_skel": 41835, "mean_count": 764.842041015625, "n_util_obs": 622, "util_count_spearman": -0.02984989927785192}, "probe_overhead_sec_total": 613.945, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006581, "alpha": 0.03, "t": 6606.9} |
| {"event": "consolidate", "step": 5400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 105, "uncertain_staged": 262118, "t": 6638.0} |
| {"event": "probe", "step": 5400, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3314.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.190427, "cps": 25.890625, "mean_p": 0.20624, "boundary_entropy": 0.11447, "bpb": 1.886501}, "tr": {"chunks": 3332.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.184058, "cps": 26.03125, "mean_p": 0.199668, "boundary_entropy": 0.156423, "bpb": 2.024688}, "fi": {"chunks": 4200.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.223274, "cps": 32.8125, "mean_p": 0.228516, "boundary_entropy": 0.123038, "bpb": 2.124653}}, "bpb": {"en": 1.8865, "tr": 2.02469, "fi": 2.12465}, "gini_cpc": 0.043736, "gini_cps": 0.054459, "bpic": 5.230684, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9665538552594719, "fi": 1.1724896644437584}, "inventory": {"size": 327575, "admitted": 65536, "staged": 262039, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1474, "admitted_surface": 1476, "staged_surface": 63020, "admitted_split": 2990, "staged_split": 44611, "admitted_merge": 4498, "staged_merge": 112170, "admitted_skel": 56572, "staged_skel": 42238, "mean_count": 773.630126953125, "n_util_obs": 626, "util_count_spearman": -0.023285323671648945}, "t": 6639.3} |
| {"event": "eval", "step": 5400, "phase": 3, "bpb": {"en": 1.90108, "tr": 2.03161, "fi": 2.13047}, "bytes_per_chunk": {"en": 5.099, "tr": 5.886, "fi": 4.765}, "bpb_mean": 2.02106, "frozen_inventory": true, "t": 6641.5} |
| {"event": "train", "step": 5400, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38229, "bpb": 1.99423, "chunk_loss": 0.03059, "F": 0.20236, "G": 0.21806, "chunks": 207.22, "bytes_per_chunk": 4.942, "boundary_entropy": 0.12198, "router_noise": 0.0, "grad_norm": 0.7334, "lr": 3.73e-05, "router_trainable": true, "sec_per_step": 0.774, "probe_sec": 0.0647, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04382, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": 0.17919, "n_shadow": 6, "mean_delta_shadow": 0.05708, "n_split": 1, "mean_delta_split": -0.04932, "n_merge": 1, "mean_delta_merge": -0.02429, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.14371, "trunc_abs_frac": 0.16781, "mean_delta_full": 0.31213, "read_hit_rate": {"surface": 0.35975, "split": 0.50502, "merge": 0.07187, "skel": 0.60843}, "mean_gate": {"surface": 0.12021, "split": 0.04818, "merge": 0.10855, "skel": 0.09738}, "n_reads": 11811, "n_cand": 29863, "read_suppressed_frac": 0.0975, "dropout_reads_logged": 11811, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1475, "admitted_surface": 1476, "staged_surface": 63013, "admitted_split": 2990, "staged_split": 44554, "admitted_merge": 4498, "staged_merge": 112345, "admitted_skel": 56572, "staged_skel": 42232, "mean_count": 773.8103637695312, "n_util_obs": 626, "util_count_spearman": -0.022834016181335532}, "probe_overhead_sec_total": 617.29, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006131, "alpha": 0.03, "t": 6645.6} |
| {"event": "consolidate", "step": 5450, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 57, "uncertain_staged": 262118, "t": 6673.8} |
| {"event": "train", "step": 5450, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32172, "bpb": 1.90684, "chunk_loss": 0.030532, "F": 0.20218, "G": 0.21808, "chunks": 207.03, "bytes_per_chunk": 4.946, "boundary_entropy": 0.12101, "router_noise": 0.0, "grad_norm": 0.7003, "lr": 3.62e-05, "router_trainable": true, "sec_per_step": 0.571, "probe_sec": 0.064, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04166, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": -0.00121, "n_shadow": 5, "mean_delta_shadow": 0.05885, "n_split": 1, "mean_delta_split": -0.1356, "n_merge": 1, "mean_delta_merge": 0.02557, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -7.03548, "trunc_abs_frac": 10.58975, "mean_delta_full": 0.03741, "read_hit_rate": {"surface": 0.39906, "split": 0.51689, "merge": 0.09968, "skel": 0.62248}, "mean_gate": {"surface": 0.12474, "split": 0.04972, "merge": 0.1108, "skel": 0.10312}, "n_reads": 12494, "n_cand": 29927, "read_suppressed_frac": 0.0997, "dropout_reads_logged": 12494, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1475, "admitted_surface": 1478, "staged_surface": 63001, "admitted_split": 3035, "staged_split": 44301, "admitted_merge": 4501, "staged_merge": 112847, "admitted_skel": 56522, "staged_skel": 41995, "mean_count": 782.6924438476562, "n_util_obs": 629, "util_count_spearman": -0.018927574700896063}, "probe_overhead_sec_total": 620.959, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006581, "alpha": 0.03, "t": 6674.2} |
| {"event": "consolidate", "step": 5500, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 27, "uncertain_staged": 262118, "t": 6706.2} |
| {"event": "train", "step": 5500, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.30585, "bpb": 1.88394, "chunk_loss": 0.030404, "F": 0.19836, "G": 0.21354, "chunks": 203.12, "bytes_per_chunk": 5.041, "boundary_entropy": 0.11837, "router_noise": 0.0, "grad_norm": 0.6857, "lr": 3.51e-05, "router_trainable": true, "sec_per_step": 0.648, "probe_sec": 0.0643, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03968, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": -0.07638, "n_shadow": 3, "mean_delta_shadow": -0.04915, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.2052, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.29517, "trunc_abs_frac": 0.13117, "mean_delta_full": -0.14119, "read_hit_rate": {"surface": 0.37941, "split": 0.51877, "merge": 0.09913, "skel": 0.60753}, "mean_gate": {"surface": 0.12337, "split": 0.04881, "merge": 0.11154, "skel": 0.09985}, "n_reads": 12117, "n_cand": 29433, "read_suppressed_frac": 0.1014, "dropout_reads_logged": 12117, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1501, "admitted_surface": 1480, "staged_surface": 62959, "admitted_split": 3050, "staged_split": 44145, "admitted_merge": 4501, "staged_merge": 113227, "admitted_skel": 56505, "staged_skel": 41813, "mean_count": 788.978515625, "n_util_obs": 632, "util_count_spearman": -0.017558953040977755}, "probe_overhead_sec_total": 624.36, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006322, "alpha": 0.03, "t": 6706.6} |
| {"event": "consolidate", "step": 5550, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 7, "uncertain_staged": 262118, "t": 6734.4} |
| {"event": "train", "step": 5550, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.36438, "bpb": 1.96838, "chunk_loss": 0.030331, "F": 0.19217, "G": 0.2099, "chunks": 196.78, "bytes_per_chunk": 5.204, "boundary_entropy": 0.12216, "router_noise": 0.0, "grad_norm": 0.7132, "lr": 3.41e-05, "router_trainable": true, "sec_per_step": 0.566, "probe_sec": 0.0622, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04177, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.01866, "n_shadow": 4, "mean_delta_shadow": 0.04964, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.80388, "trunc_abs_frac": 0.5782, "mean_delta_full": 0.05762, "read_hit_rate": {"surface": 0.39377, "split": 0.51173, "merge": 0.09369, "skel": 0.59273}, "mean_gate": {"surface": 0.12573, "split": 0.04945, "merge": 0.11518, "skel": 0.10373}, "n_reads": 11531, "n_cand": 28307, "read_suppressed_frac": 0.1008, "dropout_reads_logged": 11531, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1505, "admitted_surface": 1482, "staged_surface": 63156, "admitted_split": 3054, "staged_split": 44113, "admitted_merge": 4502, "staged_merge": 112946, "admitted_skel": 56498, "staged_skel": 41929, "mean_count": 800.962646484375, "n_util_obs": 635, "util_count_spearman": -0.01427436660239885}, "probe_overhead_sec_total": 627.961, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.005802, "alpha": 0.03, "t": 6734.8} |
| {"event": "consolidate", "step": 5600, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 62, "uncertain_staged": 262119, "t": 6765.8} |
| {"event": "probe", "step": 5600, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3288.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.188933, "cps": 25.6875, "mean_p": 0.205093, "boundary_entropy": 0.113935, "bpb": 1.881622}, "tr": {"chunks": 3400.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.187814, "cps": 26.5625, "mean_p": 0.198221, "boundary_entropy": 0.152163, "bpb": 2.015822}, "fi": {"chunks": 4143.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.220243, "cps": 32.367188, "mean_p": 0.226801, "boundary_entropy": 0.121847, "bpb": 2.114822}}, "bpb": {"en": 1.88162, "tr": 2.01582, "fi": 2.11482}, "gini_cpc": 0.036214, "gini_cps": 0.052627, "bpic": 5.237928, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9940784908417405, "fi": 1.1657229889950242}, "inventory": {"size": 327616, "admitted": 65536, "staged": 262080, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1511, "admitted_surface": 1486, "staged_surface": 63109, "admitted_split": 3103, "staged_split": 44276, "admitted_merge": 4504, "staged_merge": 113033, "admitted_skel": 56443, "staged_skel": 41662, "mean_count": 808.6175537109375, "n_util_obs": 638, "util_count_spearman": -0.014270033983340164}, "t": 6767.1} |
| {"event": "eval", "step": 5600, "phase": 3, "bpb": {"en": 1.89696, "tr": 2.02268, "fi": 2.12148}, "bytes_per_chunk": {"en": 5.135, "tr": 5.721, "fi": 4.808}, "bpb_mean": 2.01371, "frozen_inventory": true, "t": 6769.2} |
| {"event": "train", "step": 5600, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35142, "bpb": 1.94969, "chunk_loss": 0.030369, "F": 0.19537, "G": 0.21125, "chunks": 200.06, "bytes_per_chunk": 5.118, "boundary_entropy": 0.12161, "router_noise": 0.0, "grad_norm": 0.747, "lr": 3.33e-05, "router_trainable": true, "sec_per_step": 0.776, "probe_sec": 0.0713, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04251, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 11, "mean_delta_loo": 0.09167, "n_shadow": 3, "mean_delta_shadow": 0.0287, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": -5.62032, "trunc_abs_frac": 2.29112, "mean_delta_full": -0.01737, "read_hit_rate": {"surface": 0.35542, "split": 0.50826, "merge": 0.07337, "skel": 0.60426}, "mean_gate": {"surface": 0.12301, "split": 0.0482, "merge": 0.1113, "skel": 0.09965}, "n_reads": 11608, "n_cand": 29211, "read_suppressed_frac": 0.1004, "dropout_reads_logged": 11608, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1511, "admitted_surface": 1486, "staged_surface": 63112, "admitted_split": 3103, "staged_split": 44282, "admitted_merge": 4504, "staged_merge": 113066, "admitted_skel": 56443, "staged_skel": 41684, "mean_count": 808.794677734375, "n_util_obs": 638, "util_count_spearman": -0.013722973857878169}, "probe_overhead_sec_total": 631.441, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006101, "alpha": 0.03, "t": 6773.6} |
| {"event": "consolidate", "step": 5650, "retired": 1, "split_retired": 4, "merge_admitted": 0, "admitted": 5, "swapped": 17, "uncertain_staged": 262119, "t": 6805.0} |
| {"event": "train", "step": 5650, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.33738, "bpb": 1.92944, "chunk_loss": 0.030409, "F": 0.19803, "G": 0.21323, "chunks": 202.78, "bytes_per_chunk": 5.05, "boundary_entropy": 0.11693, "router_noise": 0.0, "grad_norm": 0.7137, "lr": 3.25e-05, "router_trainable": true, "sec_per_step": 0.644, "probe_sec": 0.0629, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04169, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": -0.02146, "n_shadow": 5, "mean_delta_shadow": -0.01643, "n_split": 1, "mean_delta_split": 0.04297, "n_merge": 2, "mean_delta_merge": 0.15736, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -0.39952, "trunc_abs_frac": 0.98303, "mean_delta_full": -0.04605, "read_hit_rate": {"surface": 0.39027, "split": 0.50793, "merge": 0.09105, "skel": 0.61101}, "mean_gate": {"surface": 0.11906, "split": 0.04684, "merge": 0.10631, "skel": 0.09674}, "n_reads": 12040, "n_cand": 29441, "read_suppressed_frac": 0.0958, "dropout_reads_logged": 12040, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1521, "admitted_surface": 1482, "staged_surface": 62946, "admitted_split": 3120, "staged_split": 44337, "admitted_merge": 4506, "staged_merge": 113038, "admitted_skel": 56428, "staged_skel": 41823, "mean_count": 816.5992431640625, "n_util_obs": 640, "util_count_spearman": -0.012805673024363758}, "probe_overhead_sec_total": 634.858, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006282, "alpha": 0.03, "t": 6805.8} |
| {"event": "consolidate", "step": 5700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 14, "uncertain_staged": 262119, "t": 6835.6} |
| {"event": "train", "step": 5700, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.34935, "bpb": 1.94671, "chunk_loss": 0.030283, "F": 0.19162, "G": 0.20801, "chunks": 196.22, "bytes_per_chunk": 5.219, "boundary_entropy": 0.12056, "router_noise": 0.0, "grad_norm": 0.7523, "lr": 3.18e-05, "router_trainable": true, "sec_per_step": 0.604, "probe_sec": 0.058, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03949, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.04614, "n_shadow": 5, "mean_delta_shadow": 0.06072, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.07843, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.81311, "trunc_abs_frac": 4.68986, "mean_delta_full": 0.08099, "read_hit_rate": {"surface": 0.37474, "split": 0.51669, "merge": 0.09751, "skel": 0.60304}, "mean_gate": {"surface": 0.12183, "split": 0.04725, "merge": 0.11069, "skel": 0.09958}, "n_reads": 11751, "n_cand": 28666, "read_suppressed_frac": 0.101, "dropout_reads_logged": 11751, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1532, "admitted_surface": 1483, "staged_surface": 63004, "admitted_split": 3133, "staged_split": 44683, "admitted_merge": 4506, "staged_merge": 112505, "admitted_skel": 56414, "staged_skel": 41952, "mean_count": 825.6182250976562, "n_util_obs": 644, "util_count_spearman": -0.004356025786249312}, "probe_overhead_sec_total": 638.141, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006383, "alpha": 0.03, "t": 6836.0} |
| {"event": "consolidate", "step": 5750, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 21, "uncertain_staged": 262119, "t": 6865.6} |
| {"event": "train", "step": 5750, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.39617, "bpb": 2.01424, "chunk_loss": 0.030516, "F": 0.20203, "G": 0.21727, "chunks": 206.88, "bytes_per_chunk": 4.95, "boundary_entropy": 0.1185, "router_noise": 0.0, "grad_norm": 0.7484, "lr": 3.13e-05, "router_trainable": true, "sec_per_step": 0.612, "probe_sec": 0.0695, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04419, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 3, "mean_delta_loo": 0.093, "n_shadow": 3, "mean_delta_shadow": -0.00352, "n_split": 1, "mean_delta_split": -0.15869, "n_merge": 3, "mean_delta_merge": 0.01133, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.51566, "trunc_abs_frac": 0.68694, "mean_delta_full": -0.38385, "read_hit_rate": {"surface": 0.3602, "split": 0.51471, "merge": 0.07322, "skel": 0.60366}, "mean_gate": {"surface": 0.12004, "split": 0.04745, "merge": 0.11181, "skel": 0.09744}, "n_reads": 11834, "n_cand": 29715, "read_suppressed_frac": 0.1009, "dropout_reads_logged": 11834, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1546, "admitted_surface": 1485, "staged_surface": 63074, "admitted_split": 3149, "staged_split": 44622, "admitted_merge": 4506, "staged_merge": 112657, "admitted_skel": 56396, "staged_skel": 41791, "mean_count": 834.5458984375, "n_util_obs": 651, "util_count_spearman": 0.002160418341858426}, "probe_overhead_sec_total": 642.209, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.00616, "alpha": 0.03, "t": 6866.6} |
| {"event": "consolidate", "step": 5800, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 14, "uncertain_staged": 262119, "t": 6895.2} |
| {"event": "probe", "step": 5800, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3307.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.190025, "cps": 25.835938, "mean_p": 0.205322, "boundary_entropy": 0.113003, "bpb": 1.877705}, "tr": {"chunks": 3399.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.187759, "cps": 26.554688, "mean_p": 0.196513, "boundary_entropy": 0.149973, "bpb": 2.014963}, "fi": {"chunks": 4152.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.220722, "cps": 32.4375, "mean_p": 0.226387, "boundary_entropy": 0.118324, "bpb": 2.111255}}, "bpb": {"en": 1.87771, "tr": 2.01496, "fi": 2.11126}, "gini_cpc": 0.036717, "gini_cps": 0.051882, "bpic": 5.224903, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9880764274302369, "fi": 1.161543253528402}, "inventory": {"size": 327665, "admitted": 65536, "staged": 262129, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1556, "admitted_surface": 1484, "staged_surface": 63068, "admitted_split": 3159, "staged_split": 44351, "admitted_merge": 4507, "staged_merge": 112939, "admitted_skel": 56386, "staged_skel": 41771, "mean_count": 843.1324462890625, "n_util_obs": 657, "util_count_spearman": -0.0012019984109535736}, "t": 6896.5} |
| {"event": "eval", "step": 5800, "phase": 3, "bpb": {"en": 1.89355, "tr": 2.02093, "fi": 2.11715}, "bytes_per_chunk": {"en": 5.122, "tr": 5.727, "fi": 4.797}, "bpb_mean": 2.01054, "frozen_inventory": true, "t": 6898.7} |
| {"event": "train", "step": 5800, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32479, "bpb": 1.91127, "chunk_loss": 0.030464, "F": 0.19662, "G": 0.21274, "chunks": 201.34, "bytes_per_chunk": 5.086, "boundary_entropy": 0.12168, "router_noise": 0.0, "grad_norm": 0.7453, "lr": 3.08e-05, "router_trainable": true, "sec_per_step": 0.73, "probe_sec": 0.0686, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04735, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": 0.02777, "n_shadow": 2, "mean_delta_shadow": -0.02054, "n_split": 1, "mean_delta_split": -0.06274, "n_merge": 1, "mean_delta_merge": 0.21545, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.38557, "trunc_abs_frac": 0.97115, "mean_delta_full": 0.16073, "read_hit_rate": {"surface": 0.38481, "split": 0.51138, "merge": 0.08669, "skel": 0.60679}, "mean_gate": {"surface": 0.12281, "split": 0.04848, "merge": 0.11114, "skel": 0.10107}, "n_reads": 11903, "n_cand": 29226, "read_suppressed_frac": 0.1015, "dropout_reads_logged": 11903, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1557, "admitted_surface": 1484, "staged_surface": 63046, "admitted_split": 3159, "staged_split": 44338, "admitted_merge": 4507, "staged_merge": 112999, "admitted_skel": 56386, "staged_skel": 41761, "mean_count": 843.3140869140625, "n_util_obs": 657, "util_count_spearman": -0.0011092668230797066}, "probe_overhead_sec_total": 645.391, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.005782, "alpha": 0.03, "t": 6903.1} |
| {"event": "consolidate", "step": 5850, "retired": 1, "split_retired": 1, "merge_admitted": 0, "admitted": 2, "swapped": 16, "uncertain_staged": 262119, "t": 6930.4} |
| {"event": "train", "step": 5850, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.34553, "bpb": 1.94118, "chunk_loss": 0.030419, "F": 0.19897, "G": 0.21345, "chunks": 203.75, "bytes_per_chunk": 5.026, "boundary_entropy": 0.11151, "router_noise": 0.0, "grad_norm": 0.7415, "lr": 3.05e-05, "router_trainable": true, "sec_per_step": 0.56, "probe_sec": 0.063, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03987, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.02395, "n_shadow": 2, "mean_delta_shadow": -0.08788, "n_split": 2, "mean_delta_split": -0.02044, "n_merge": 1, "mean_delta_merge": 0.11566, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -0.10269, "trunc_abs_frac": 0.5679, "mean_delta_full": -0.28658, "read_hit_rate": {"surface": 0.36621, "split": 0.51522, "merge": 0.07481, "skel": 0.60462}, "mean_gate": {"surface": 0.1171, "split": 0.04606, "merge": 0.10884, "skel": 0.09327}, "n_reads": 12019, "n_cand": 29814, "read_suppressed_frac": 0.0997, "dropout_reads_logged": 12019, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1558, "admitted_surface": 1484, "staged_surface": 63196, "admitted_split": 3172, "staged_split": 44181, "admitted_merge": 4509, "staged_merge": 112994, "admitted_skel": 56371, "staged_skel": 41773, "mean_count": 851.1753540039062, "n_util_obs": 662, "util_count_spearman": -0.01168639423437852}, "probe_overhead_sec_total": 648.572, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.005836, "alpha": 0.03, "t": 6931.1} |
| {"event": "consolidate", "step": 5900, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 9, "uncertain_staged": 262119, "t": 6961.0} |
| {"event": "train", "step": 5900, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.27894, "bpb": 1.84512, "chunk_loss": 0.030269, "F": 0.19171, "G": 0.20692, "chunks": 196.31, "bytes_per_chunk": 5.216, "boundary_entropy": 0.11572, "router_noise": 0.0, "grad_norm": 0.7032, "lr": 3.02e-05, "router_trainable": true, "sec_per_step": 0.612, "probe_sec": 0.0632, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04181, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": -0.04502, "n_shadow": 7, "mean_delta_shadow": 0.01617, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.12048, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -0.22785, "trunc_abs_frac": 0.86287, "mean_delta_full": -0.0959, "read_hit_rate": {"surface": 0.37136, "split": 0.5193, "merge": 0.09119, "skel": 0.60898}, "mean_gate": {"surface": 0.12216, "split": 0.04776, "merge": 0.11056, "skel": 0.09943}, "n_reads": 11751, "n_cand": 28670, "read_suppressed_frac": 0.0939, "dropout_reads_logged": 11751, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1578, "admitted_surface": 1484, "staged_surface": 63134, "admitted_split": 3177, "staged_split": 44010, "admitted_merge": 4511, "staged_merge": 113306, "admitted_skel": 56364, "staged_skel": 41694, "mean_count": 860.1357421875, "n_util_obs": 665, "util_count_spearman": -0.002867576906593123}, "probe_overhead_sec_total": 651.929, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006253, "alpha": 0.03, "t": 6961.7} |
| {"event": "consolidate", "step": 5950, "retired": 0, "split_retired": 3, "merge_admitted": 0, "admitted": 3, "swapped": 33, "uncertain_staged": 262119, "t": 6988.6} |
| {"event": "train", "step": 5950, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.37911, "bpb": 1.98964, "chunk_loss": 0.030406, "F": 0.19904, "G": 0.21327, "chunks": 203.81, "bytes_per_chunk": 5.024, "boundary_entropy": 0.11786, "router_noise": 0.0, "grad_norm": 0.7344, "lr": 3.01e-05, "router_trainable": true, "sec_per_step": 0.545, "probe_sec": 0.0653, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04233, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.0679, "n_shadow": 3, "mean_delta_shadow": -0.08004, "n_split": 1, "mean_delta_split": 0.10327, "n_merge": 1, "mean_delta_merge": 0.01599, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.57131, "trunc_abs_frac": 0.8111, "mean_delta_full": 0.14209, "read_hit_rate": {"surface": 0.34977, "split": 0.52161, "merge": 0.09167, "skel": 0.61817}, "mean_gate": {"surface": 0.11967, "split": 0.04658, "merge": 0.10506, "skel": 0.09566}, "n_reads": 12148, "n_cand": 29791, "read_suppressed_frac": 0.1001, "dropout_reads_logged": 12148, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1583, "admitted_surface": 1482, "staged_surface": 63078, "admitted_split": 3207, "staged_split": 44018, "admitted_merge": 4512, "staged_merge": 113214, "admitted_skel": 56335, "staged_skel": 41834, "mean_count": 864.7684326171875, "n_util_obs": 669, "util_count_spearman": 0.003420640420836163}, "probe_overhead_sec_total": 655.214, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006018, "alpha": 0.03, "t": 6989.0} |
| {"event": "consolidate", "step": 6000, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 15, "uncertain_staged": 262119, "t": 7018.1} |
| {"event": "train", "step": 5999, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38424, "bpb": 1.99704, "chunk_loss": 0.030475, "F": 0.19913, "G": 0.21406, "chunks": 203.91, "bytes_per_chunk": 5.022, "boundary_entropy": 0.11464, "router_noise": 0.0, "grad_norm": 0.7015, "lr": 3e-05, "router_trainable": true, "sec_per_step": 0.582, "probe_sec": 0.0645, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04273, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": 0.01491, "n_shadow": 4, "mean_delta_shadow": -0.04259, "n_split": 3, "mean_delta_split": -0.18958, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -0.0366, "trunc_abs_frac": 1.14252, "mean_delta_full": 0.02351, "read_hit_rate": {"surface": 0.37718, "split": 0.51178, "merge": 0.08915, "skel": 0.60653}, "mean_gate": {"surface": 0.12083, "split": 0.04701, "merge": 0.10899, "skel": 0.09709}, "n_reads": 11810, "n_cand": 29201, "read_suppressed_frac": 0.0999, "dropout_reads_logged": 11810, "inventory": {"size": 327663, "admitted": 65536, "staged": 262127, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1592, "admitted_surface": 1484, "staged_surface": 62975, "admitted_split": 3216, "staged_split": 43766, "admitted_merge": 4513, "staged_merge": 113647, "admitted_skel": 56323, "staged_skel": 41739, "mean_count": 873.39892578125, "n_util_obs": 673, "util_count_spearman": 0.0020731477701752825}, "probe_overhead_sec_total": 658.485, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006017, "alpha": 0.03, "t": 7018.1} |
| {"event": "probe", "step": 6000, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3310.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.190197, "cps": 25.859375, "mean_p": 0.205075, "boundary_entropy": 0.111035, "bpb": 1.876649}, "tr": {"chunks": 3400.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.187814, "cps": 26.5625, "mean_p": 0.196723, "boundary_entropy": 0.148765, "bpb": 2.009317}, "fi": {"chunks": 4182.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.222317, "cps": 32.671875, "mean_p": 0.22731, "boundary_entropy": 0.119273, "bpb": 2.10764}}, "bpb": {"en": 1.87665, "tr": 2.00932, "fi": 2.10764}, "gini_cpc": 0.038315, "gini_cps": 0.053373, "bpic": 5.208593, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9874713226246656, "fi": 1.1688755422238803}, "inventory": {"size": 327663, "admitted": 65536, "staged": 262127, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1592, "admitted_surface": 1484, "staged_surface": 62975, "admitted_split": 3216, "staged_split": 43766, "admitted_merge": 4513, "staged_merge": 113647, "admitted_skel": 56323, "staged_skel": 41739, "mean_count": 873.39892578125, "n_util_obs": 673, "util_count_spearman": 0.0020731477701752825}, "t": 7019.4} |
| {"event": "eval", "step": 6000, "phase": 3, "bpb": {"en": 1.89206, "tr": 2.01579, "fi": 2.11394}, "bytes_per_chunk": {"en": 5.116, "tr": 5.745, "fi": 4.773}, "bpb_mean": 2.00726, "frozen_inventory": true, "t": 7021.6} |
| {"event": "reps_dumped", "dirs": ["/local/scratch/sas245/beetle-hnet/runs/ali_ali-ema_s2/probe/reps/th0.30", "/local/scratch/sas245/beetle-hnet/runs/ali_ali-ema_s2/probe/reps/th0.40", "/local/scratch/sas245/beetle-hnet/runs/ali_ali-ema_s2/probe/reps/th0.50", "/local/scratch/sas245/beetle-hnet/runs/ali_ali-ema_s2/probe/reps/th0.60", "/local/scratch/sas245/beetle-hnet/runs/ali_ali-ema_s2/probe/reps/th0.70"], "t": 7025.6} |
| {"event": "final", "run": "ali_ali-ema_s2", "arch": "ali", "objective": "ratio", "condition": "A", "step": 6000, "params": 22785088, "seed": 2, "per_lang": {"en": {"cpc": 0.1901970924553238, "cps": 25.859375, "bpb": 1.8971362146122137, "bytes_per_chunk": 5.036726773173989, "reads_per_chunk": 1.9068237057246884}, "tr": {"cpc": 0.1878141744462244, "cps": 26.5625, "bpb": 2.0067071783945005, "bytes_per_chunk": 5.765905266746582, "reads_per_chunk": 1.5557742854314904}, "fi": {"cpc": 0.2223167295731221, "cps": 32.671875, "bpb": 2.1323793084353833, "bytes_per_chunk": 4.751481756720016, "reads_per_chunk": 1.729984955864494}}, "bpb_mean": 2.0120742338140327, "gini_cpc": 0.03831522693095413, "gini_cps": 0.05337250581466524, "bpic": 5.2085934630921775, "mean_boundary_entropy": 0.1146434098482132, "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": 658.484516620636, "probe_overhead_seq_frac": 0.4999166666666667, "wall_sec": 7032.369393110275, "inventory": {"size": 327663, "admitted": 65536, "staged": 262127, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1592, "admitted_surface": 1484, "staged_surface": 62975, "admitted_split": 3216, "staged_split": 43766, "admitted_merge": 4513, "staged_merge": 113647, "admitted_skel": 56323, "staged_skel": 41739, "mean_count": 873.39892578125, "n_util_obs": 673, "util_count_spearman": 0.0020731477701752825}, "ali_params": 268096, "ali_config": {"enabled": true, "budget_B": 65536, "staging_capacity": 262144, "policy": "utility", "candidate_types": ["surface", "split", "merge", "skel"], "gate": "vector", "retrieval": true, "memory_scope": "persistent", "read_dropout": 0.1, "write_rule": "ema_floor", "eta_min": 0.02, "consolidate_every": 50, "probe_seqs": 2, "probe_reps": 8, "horizon": 32, "predictor": "mlp", "lambda_util": 0.1, "max_span": 64, "byte_classes_path": "/local/scratch/sas245/beetle-hnet/scratch/data/byte_classes.json", "gate_hidden": 0, "type_dim": 16, "gate_bias_init": -2.0}, "util_vs_frequency_spearman": 0.0020731477701752825, "paired_vs_dropout": {"n": 1337, "pearson": -0.022845278606275488, "spearman": 0.010523693405067655, "mean_abs_discrepancy": 0.3949146270751953, "mean_paired": 0.013803702779114246, "mean_dropout": 0.048447128385305405}, "predictor_calibration": {"n": 3088, "all": {"n": 3088, "pearson": 0.10416681401858108, "spearman": 0.10541703130472703, "mae": 0.06698639690876007, "reliability": [{"pred": -0.009958121925592422, "obs": -0.00588121498003602, "n": 309}, {"pred": -1.1075781912950333e-05, "obs": 0.003863766323775053, "n": 309}, {"pred": 0.0025206152349710464, "obs": 0.007261715829372406, "n": 309}, {"pred": 0.004377542529255152, "obs": -0.0008440149249508977, "n": 308}, {"pred": 0.005750237964093685, "obs": 0.0087474649772048, "n": 309}, {"pred": 0.007714975159615278, "obs": 0.0159766748547554, "n": 309}, {"pred": 0.010088569484651089, "obs": 0.020149748772382736, "n": 308}, {"pred": 0.012256436049938202, "obs": 0.02656889706850052, "n": 309}, {"pred": 0.015597963705658913, "obs": 0.02468051202595234, "n": 309}, {"pred": 0.023524438962340355, "obs": 0.026402730494737625, "n": 309}]}, "admitted": {"n": 3054, "pearson": 0.10747404201838114, "spearman": 0.10335375987408908, "mae": 0.06729769706726074, "reliability": [{"pred": -0.008902370929718018, "obs": -0.004839307628571987, "n": 306}, {"pred": 0.00021819565154146403, "obs": 0.006557357031852007, "n": 305}, {"pred": 0.002646495122462511, "obs": 0.006762208882719278, "n": 305}, {"pred": 0.004467288497835398, "obs": -0.0003836536779999733, "n": 306}, {"pred": 0.005811069626361132, "obs": 0.005709864664822817, "n": 305}, {"pred": 0.0077991425059735775, "obs": 0.016588244587183, "n": 305}, {"pred": 0.010151953436434269, "obs": 0.02165837027132511, "n": 306}, {"pred": 0.01232284028083086, "obs": 0.026581838726997375, "n": 305}, {"pred": 0.015664149075746536, "obs": 0.02482587844133377, "n": 305}, {"pred": 0.023578040301799774, "obs": 0.026963677257299423, "n": 306}]}, "staged": {"n": 34, "pearson": -0.2859285368408656, "spearman": -0.5288006111535523, "mae": 0.03902382403612137, "reliability": [{"pred": -0.05824749916791916, "obs": 0.0076751708984375, "n": 4}, {"pred": -0.025906706228852272, "obs": 0.012908935546875, "n": 3}, {"pred": -0.011162939481437206, "obs": -0.08101648837327957, "n": 3}, {"pred": -0.006388402543962002, "obs": -0.004583537578582764, "n": 4}, {"pred": -0.004260601941496134, "obs": -0.011403027921915054, "n": 3}, {"pred": -0.002395186573266983, "obs": -0.022086994722485542, "n": 3}, {"pred": -0.0005985461175441742, "obs": -0.010456628166139126, "n": 4}, {"pred": 0.00019091616559308022, "obs": -0.01719760149717331, "n": 3}, {"pred": 0.003857092233374715, "obs": -0.04803817346692085, "n": 3}, {"pred": 0.011402986943721771, "obs": -0.022997580468654633, "n": 4}]}}, "t": 7032.4} |
| {"event": "hub_done", "repo": "AdaptiveChunking/ali_ali-ema_s2", "pushed": [0, 1, 2, 4, 6, 8, 16, 32, 64, 128, 256, 512, 1000, 1024, 2048, 3000, 4096, 6000], "failed": [], "ok": true, "t": 7049.1} |
|
|
