| {"event": "hub", "repo": "AdaptiveChunking/ali_mem-mixed_s4", "keep_steps": [0, 1, 2, 4, 6, 8, 16, 32, 64, 128, 256, 512, 1000, 1024, 2048, 3000, 4096, 6000], "t": 4.0} |
| {"event": "start", "run": "ali_mem-mixed_s4", "arch": "ali", "objective": "ratio", "condition": "A", "params": 22785088, "langs": ["en", "tr", "fi"], "premium": {"en": 1.0, "tr": 1.12, "fi": 1.104}, "schedule": {"tau1": 0.4, "tau2": 0.7, "p1_objective": "global", "p2_objective": "global", "p3_objective": "global", "freeze_in_p3": false, "freeze_scope": "router"}, "corpus": "/local/scratch/sas245/beetle-hnet/scratch/data/tier1_corpus", "premium_kind": "flores_per_sent", "doc_filtered": {}, "weights": {"en": 0.74, "tr": 0.1536, "fi": 0.1064}, "device": "cuda", "ali_params": 268096, "ali_config": {"enabled": true, "budget_B": 65536, "staging_capacity": 262144, "policy": "frequency", "value_mode": "learned", "candidate_types": ["surface", "split", "merge", "skel"], "gate": "vector", "retrieval": true, "memory_scope": "persistent", "read_dropout": 0.1, "write_rule": "mean", "eta_min": 0.0, "consolidate_every": 50, "probe_seqs": 2, "probe_reps": 8, "horizon": 32, "predictor": "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": 4.0} |
| {"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": 4.0} |
| {"event": "probe", "step": 0, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 5462.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.313854, "cps": 42.671875, "mean_p": 0.489807, "boundary_entropy": 0.68614, "bpb": 8.077753}, "tr": {"chunks": 13273.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.733193, "cps": 103.695312, "mean_p": 0.515287, "boundary_entropy": 0.686872, "bpb": 8.018145}, "fi": {"chunks": 9462.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.503004, "cps": 73.921875, "mean_p": 0.501981, "boundary_entropy": 0.6873, "bpb": 7.990695}}, "bpb": {"en": 8.07775, "tr": 8.01815, "fi": 7.9907}, "gini_cpc": 0.180355, "gini_cps": 0.184677, "bpic": 2.011987, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 2.3360975145821765, "fi": 1.6026677014027704}, "inventory": {"size": 0, "admitted": 0, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 0.0, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 6.3} |
| {"event": "train", "step": 0, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 5.57735, "bpb": 8.04642, "chunk_loss": 0.037766, "F": 0.27518, "G": 0.48384, "chunks": 281.78, "bytes_per_chunk": 3.634, "boundary_entropy": 0.69036, "router_noise": 0.0, "grad_norm": 13.5991, "lr": 1e-06, "router_trainable": true, "sec_per_step": 0.014, "probe_sec": 0.0063, "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": 34037, "read_suppressed_frac": 0.0, "dropout_reads_logged": 0, "inventory": {"size": 15702, "admitted": 15702, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 3526, "staged_surface": 0, "admitted_split": 3036, "staged_split": 0, "admitted_merge": 6845, "staged_merge": 0, "admitted_skel": 2295, "staged_skel": 0, "mean_count": 2.1676855087280273, "n_util_obs": 0, "util_count_spearman": NaN}, "probe_overhead_sec_total": 0.006, "probe_overhead_seq_frac": 0.0, "util_loss": 0.0, "alpha": 0.03, "t": 7.6} |
| {"event": "probe", "step": 1, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 5494.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.315693, "cps": 42.921875, "mean_p": 0.48996, "boundary_entropy": 0.68615, "bpb": 8.047628}, "tr": {"chunks": 13305.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.734961, "cps": 103.945312, "mean_p": 0.515387, "boundary_entropy": 0.686869, "bpb": 8.059994}, "fi": {"chunks": 9453.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.502525, "cps": 73.851562, "mean_p": 0.50196, "boundary_entropy": 0.687303, "bpb": 7.977565}}, "bpb": {"en": 8.04763, "tr": 8.05999, "fi": 7.97756}, "gini_cpc": 0.179961, "gini_cps": 0.184317, "bpic": 2.00807, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 2.3280901457362626, "fi": 1.5918173704876495}, "inventory": {"size": 15702, "admitted": 15702, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 3526, "staged_surface": 0, "admitted_split": 3036, "staged_split": 0, "admitted_merge": 6845, "staged_merge": 0, "admitted_skel": 2295, "staged_skel": 0, "mean_count": 2.1676855087280273, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 9.4} |
| {"event": "probe", "step": 2, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 5542.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.318451, "cps": 43.296875, "mean_p": 0.490204, "boundary_entropy": 0.686165, "bpb": 7.983023}, "tr": {"chunks": 13354.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.737668, "cps": 104.328125, "mean_p": 0.515562, "boundary_entropy": 0.686865, "bpb": 8.014054}, "fi": {"chunks": 9421.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.500824, "cps": 73.601562, "mean_p": 0.50188, "boundary_entropy": 0.687306, "bpb": 7.91768}}, "bpb": {"en": 7.98302, "tr": 8.01405, "fi": 7.91768}, "gini_cpc": 0.179504, "gini_cps": 0.183918, "bpic": 2.003461, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 2.3164259377646412, "fi": 1.5726885290076422}, "inventory": {"size": 28259, "admitted": 28259, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 6, "admitted_surface": 6247, "staged_surface": 0, "admitted_split": 5209, "staged_split": 0, "admitted_merge": 12654, "staged_merge": 0, "admitted_skel": 4149, "staged_skel": 0, "mean_count": 2.3898580074310303, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 12.1} |
| {"event": "probe", "step": 4, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 5674.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.326036, "cps": 44.328125, "mean_p": 0.490937, "boundary_entropy": 0.686211, "bpb": 7.767068}, "tr": {"chunks": 13476.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.744407, "cps": 105.28125, "mean_p": 0.515981, "boundary_entropy": 0.68686, "bpb": 7.833855}, "fi": {"chunks": 9238.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.491096, "cps": 72.171875, "mean_p": 0.501438, "boundary_entropy": 0.687318, "bpb": 7.716055}}, "bpb": {"en": 7.76707, "tr": 7.83385, "fi": 7.71605}, "gini_cpc": 0.178615, "gini_cps": 0.183223, "bpic": 1.99845, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 2.2832067495497457, "fi": 1.506263190900666}, "inventory": {"size": 49789, "admitted": 49789, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 11, "admitted_surface": 10681, "staged_surface": 0, "admitted_split": 8767, "staged_split": 0, "admitted_merge": 23219, "staged_merge": 0, "admitted_skel": 7122, "staged_skel": 0, "mean_count": 2.8343608379364014, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 15.5} |
| {"event": "probe", "step": 6, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 5926.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.340516, "cps": 46.296875, "mean_p": 0.492051, "boundary_entropy": 0.686279, "bpb": 7.473546}, "tr": {"chunks": 13642.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.753577, "cps": 106.578125, "mean_p": 0.516423, "boundary_entropy": 0.686862, "bpb": 7.58705}, "fi": {"chunks": 8934.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.474935, "cps": 69.796875, "mean_p": 0.500628, "boundary_entropy": 0.68733, "bpb": 7.457559}}, "bpb": {"en": 7.47355, "tr": 7.58705, "fi": 7.45756}, "gini_cpc": 0.175506, "gini_cps": 0.180479, "bpic": 1.990457, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 2.2130435827432064, "fi": 1.3947505384798926}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 17, "admitted_surface": 14004, "staged_surface": 0, "admitted_split": 11323, "staged_split": 0, "admitted_merge": 30779, "staged_merge": 0, "admitted_skel": 9430, "staged_skel": 0, "mean_count": 3.0955657958984375, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 20.1} |
| {"event": "probe", "step": 8, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 6229.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.357927, "cps": 48.664062, "mean_p": 0.493378, "boundary_entropy": 0.68636, "bpb": 7.197037}, "tr": {"chunks": 13754.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.759764, "cps": 107.453125, "mean_p": 0.516498, "boundary_entropy": 0.686892, "bpb": 7.355595}, "fi": {"chunks": 8419.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.447557, "cps": 65.773438, "mean_p": 0.499199, "boundary_entropy": 0.687333, "bpb": 7.225146}}, "bpb": {"en": 7.19704, "tr": 7.3556, "fi": 7.22515}, "gini_cpc": 0.171149, "gini_cps": 0.176631, "bpic": 1.997465, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 2.1226786799536628, "fi": 1.2504156926178984}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 24, "admitted_surface": 14076, "staged_surface": 0, "admitted_split": 11525, "staged_split": 0, "admitted_merge": 30425, "staged_merge": 0, "admitted_skel": 9510, "staged_skel": 0, "mean_count": 3.792724609375, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 24.6} |
| {"event": "probe", "step": 16, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 5306.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.30489, "cps": 41.453125, "mean_p": 0.490503, "boundary_entropy": 0.686434, "bpb": 6.582006}, "tr": {"chunks": 10707.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.591449, "cps": 83.648438, "mean_p": 0.505556, "boundary_entropy": 0.687472, "bpb": 6.85988}, "fi": {"chunks": 4872.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.258997, "cps": 38.0625, "mean_p": 0.488076, "boundary_entropy": 0.687035, "bpb": 6.718783}}, "bpb": {"en": 6.58201, "tr": 6.85988, "fi": 6.71878}, "gini_cpc": 0.191835, "gini_cps": 0.186258, "bpic": 2.716399, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.9398766958572422, "fi": 0.8494782637846351}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 46, "admitted_surface": 14056, "staged_surface": 0, "admitted_split": 11457, "staged_split": 0, "admitted_merge": 30537, "staged_merge": 0, "admitted_skel": 9486, "staged_skel": 0, "mean_count": 7.04937744140625, "n_util_obs": 1, "util_count_spearman": NaN}, "t": 32.7} |
| {"event": "probe", "step": 32, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 4311.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.247716, "cps": 33.679688, "mean_p": 0.492737, "boundary_entropy": 0.687292, "bpb": 6.229797}, "tr": {"chunks": 4955.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.273712, "cps": 38.710938, "mean_p": 0.491833, "boundary_entropy": 0.687814, "bpb": 6.585122}, "fi": {"chunks": 2868.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.152464, "cps": 22.40625, "mean_p": 0.488188, "boundary_entropy": 0.687679, "bpb": 6.376573}}, "bpb": {"en": 6.2298, "tr": 6.58512, "fi": 6.37657}, "gini_cpc": 0.119948, "gini_cps": 0.114664, "bpic": 4.675457, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.1049412949045034, "fi": 0.615479172060855}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 82, "admitted_surface": 14486, "staged_surface": 0, "admitted_split": 12137, "staged_split": 0, "admitted_merge": 29442, "staged_merge": 0, "admitted_skel": 9471, "staged_skel": 0, "mean_count": 10.526168823242188, "n_util_obs": 2, "util_count_spearman": NaN}, "t": 45.5} |
| {"event": "consolidate", "step": 50, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 60.5} |
| {"event": "train", "step": 50, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 4.14507, "bpb": 5.98007, "chunk_loss": 0.030365, "F": 0.17136, "G": 0.48665, "chunks": 175.47, "bytes_per_chunk": 5.836, "boundary_entropy": 0.69169, "router_noise": 0.0, "grad_norm": 2.4166, "lr": 5.1e-05, "router_trainable": true, "sec_per_step": 1.074, "probe_sec": 0.0715, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05239, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": 0.00074, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.02051, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -0.08527, "trunc_abs_frac": 3.4714, "mean_delta_full": -0.021, "read_hit_rate": {"surface": 0.64999, "split": 0.83071, "merge": 0.20857, "skel": 0.63692}, "mean_gate": {"surface": 0.11867, "split": 0.11862, "merge": 0.11875, "skel": 0.11866}, "n_reads": 15756, "n_cand": 25311, "read_suppressed_frac": 0.1014, "dropout_reads_logged": 15756, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 123, "admitted_surface": 15157, "staged_surface": 0, "admitted_split": 13845, "staged_split": 0, "admitted_merge": 26407, "staged_merge": 0, "admitted_skel": 10127, "staged_skel": 0, "mean_count": 14.703659057617188, "n_util_obs": 3, "util_count_spearman": -0.5}, "probe_overhead_sec_total": 4.963, "probe_overhead_seq_frac": 0.4902, "util_loss": 0.020867, "alpha": 0.03, "t": 61.3} |
| {"event": "probe", "step": 64, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 4233.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.243234, "cps": 33.070312, "mean_p": 0.481927, "boundary_entropy": 0.685342, "bpb": 5.499438}, "tr": {"chunks": 5918.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.326907, "cps": 46.234375, "mean_p": 0.490964, "boundary_entropy": 0.687203, "bpb": 5.9081}, "fi": {"chunks": 4563.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.242571, "cps": 35.648438, "mean_p": 0.48404, "boundary_entropy": 0.686484, "bpb": 5.675835}}, "bpb": {"en": 5.49944, "tr": 5.9081, "fi": 5.67583}, "gini_cpc": 0.069181, "gini_cps": 0.076345, "bpic": 3.855648, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.3440030711732098, "fi": 0.9972738631259866}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 154, "admitted_surface": 15843, "staged_surface": 0, "admitted_split": 14391, "staged_split": 0, "admitted_merge": 24913, "staged_merge": 0, "admitted_skel": 10389, "staged_skel": 0, "mean_count": 18.226287841796875, "n_util_obs": 4, "util_count_spearman": -0.8}, "t": 71.5} |
| {"event": "consolidate", "step": 100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 102.7} |
| {"event": "train", "step": 100, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 3.27055, "bpb": 4.71841, "chunk_loss": 0.03187, "F": 0.1951, "G": 0.46307, "chunks": 199.78, "bytes_per_chunk": 5.126, "boundary_entropy": 0.67847, "router_noise": 0.0, "grad_norm": 1.9478, "lr": 0.000101, "router_trainable": true, "sec_per_step": 0.856, "probe_sec": 0.1752, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04825, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.01063, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 3, "mean_delta_split": -0.01813, "n_merge": 1, "mean_delta_merge": -0.00073, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 3.13462, "trunc_abs_frac": 3.13462, "mean_delta_full": -0.0127, "read_hit_rate": {"surface": 0.6216, "split": 0.81821, "merge": 0.17775, "skel": 0.60769}, "mean_gate": {"surface": 0.11896, "split": 0.11894, "merge": 0.11902, "skel": 0.1189}, "n_reads": 16651, "n_cand": 27931, "read_suppressed_frac": 0.0996, "dropout_reads_logged": 16651, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 236, "admitted_surface": 16619, "staged_surface": 0, "admitted_split": 15668, "staged_split": 0, "admitted_merge": 22234, "staged_merge": 0, "admitted_skel": 11015, "staged_skel": 0, "mean_count": 28.757095336914062, "n_util_obs": 9, "util_count_spearman": 0.05}, "probe_overhead_sec_total": 11.318, "probe_overhead_seq_frac": 0.495, "util_loss": 0.001383, "alpha": 0.03, "t": 104.2} |
| {"event": "probe", "step": 128, "phase": 1, "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": 0.452307, "boundary_entropy": 0.63289, "bpb": 4.110949}, "tr": {"chunks": 4163.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.229962, "cps": 32.523438, "mean_p": 0.436456, "boundary_entropy": 0.656018, "bpb": 4.416149}, "fi": {"chunks": 3902.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.207432, "cps": 30.484375, "mean_p": 0.44187, "boundary_entropy": 0.657394, "bpb": 4.351398}}, "bpb": {"en": 4.11095, "tr": 4.41615, "fi": 4.3514}, "gini_cpc": 0.044585, "gini_cps": 0.052268, "bpic": 5.003263, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.2223661211547536, "fi": 1.1026072067491153}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 279, "admitted_surface": 16723, "staged_surface": 0, "admitted_split": 15909, "staged_split": 0, "admitted_merge": 21943, "staged_merge": 0, "admitted_skel": 10961, "staged_skel": 0, "mean_count": 36.013671875, "n_util_obs": 14, "util_count_spearman": -0.27472527472527475}, "t": 130.8} |
| {"event": "consolidate", "step": 150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 153.5} |
| {"event": "train", "step": 150, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.75915, "bpb": 3.98061, "chunk_loss": 0.031326, "F": 0.19006, "G": 0.42095, "chunks": 194.62, "bytes_per_chunk": 5.261, "boundary_entropy": 0.61956, "router_noise": 0.0, "grad_norm": 1.6788, "lr": 0.000151, "router_trainable": true, "sec_per_step": 0.998, "probe_sec": 0.0887, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06715, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.00471, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.02722, "n_merge": 3, "mean_delta_merge": -0.01353, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -0.09993, "trunc_abs_frac": 0.65575, "mean_delta_full": 0.0568, "read_hit_rate": {"surface": 0.63767, "split": 0.81128, "merge": 0.20279, "skel": 0.56427}, "mean_gate": {"surface": 0.1191, "split": 0.11906, "merge": 0.11911, "skel": 0.11898}, "n_reads": 15557, "n_cand": 26352, "read_suppressed_frac": 0.0967, "dropout_reads_logged": 15557, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 316, "admitted_surface": 16803, "staged_surface": 0, "admitted_split": 16233, "staged_split": 0, "admitted_merge": 21237, "staged_merge": 0, "admitted_skel": 11263, "staged_skel": 0, "mean_count": 41.699554443359375, "n_util_obs": 16, "util_count_spearman": -0.23529411764705882}, "probe_overhead_sec_total": 16.787, "probe_overhead_seq_frac": 0.4967, "util_loss": 0.001002, "alpha": 0.03, "t": 154.1} |
| {"event": "consolidate", "step": 200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 190.2} |
| {"event": "probe", "step": 200, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3163.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.18175, "cps": 24.710938, "mean_p": 0.357413, "boundary_entropy": 0.525531, "bpb": 3.555505}, "tr": {"chunks": 4177.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.230735, "cps": 32.632812, "mean_p": 0.388865, "boundary_entropy": 0.601268, "bpb": 3.795033}, "fi": {"chunks": 3512.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.186699, "cps": 27.4375, "mean_p": 0.372107, "boundary_entropy": 0.591519, "bpb": 3.853166}}, "bpb": {"en": 3.55551, "tr": 3.79503, "fi": 3.85317}, "gini_cpc": 0.054502, "gini_cps": 0.062293, "bpic": 5.227792, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.2695179683599602, "fi": 1.0272296634342586}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 390, "admitted_surface": 16792, "staged_surface": 0, "admitted_split": 16614, "staged_split": 0, "admitted_merge": 20738, "staged_merge": 0, "admitted_skel": 11392, "staged_skel": 0, "mean_count": 53.47364807128906, "n_util_obs": 21, "util_count_spearman": -0.17142857142857143}, "t": 191.9} |
| {"event": "eval", "step": 200, "phase": 1, "bpb": {"en": 3.56505, "tr": 3.80402, "fi": 3.8522}, "bytes_per_chunk": {"en": 5.241, "tr": 4.654, "fi": 5.476}, "bpb_mean": 3.74042, "frozen_inventory": true, "t": 194.7} |
| {"event": "train", "step": 200, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.55711, "bpb": 3.68913, "chunk_loss": 0.031158, "F": 0.19107, "G": 0.36834, "chunks": 195.66, "bytes_per_chunk": 5.234, "boundary_entropy": 0.5506, "router_noise": 0.0, "grad_norm": 1.1971, "lr": 0.000201, "router_trainable": true, "sec_per_step": 0.848, "probe_sec": 0.0798, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05024, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.00739, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": 0.07208, "n_merge": 1, "mean_delta_merge": 0.03693, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -0.92742, "trunc_abs_frac": 0.68949, "mean_delta_full": -0.04541, "read_hit_rate": {"surface": 0.65596, "split": 0.82691, "merge": 0.19977, "skel": 0.647}, "mean_gate": {"surface": 0.12011, "split": 0.11996, "merge": 0.12003, "skel": 0.11979}, "n_reads": 17338, "n_cand": 27965, "read_suppressed_frac": 0.097, "dropout_reads_logged": 17338, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 393, "admitted_surface": 16822, "staged_surface": 0, "admitted_split": 16762, "staged_split": 0, "admitted_merge": 20541, "staged_merge": 0, "admitted_skel": 11411, "staged_skel": 0, "mean_count": 53.72261047363281, "n_util_obs": 21, "util_count_spearman": -0.09090909090909091}, "probe_overhead_sec_total": 21.742, "probe_overhead_seq_frac": 0.4975, "util_loss": 0.000435, "alpha": 0.03, "t": 196.5} |
| {"event": "consolidate", "step": 250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 231.2} |
| {"event": "train", "step": 250, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.34829, "bpb": 3.38787, "chunk_loss": 0.030561, "F": 0.18216, "G": 0.32687, "chunks": 186.53, "bytes_per_chunk": 5.49, "boundary_entropy": 0.46436, "router_noise": 0.0, "grad_norm": 1.4816, "lr": 0.000251, "router_trainable": true, "sec_per_step": 0.706, "probe_sec": 0.0877, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06913, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": 0.00453, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.12671, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 0.2974, "trunc_abs_frac": 0.48141, "mean_delta_full": 0.00977, "read_hit_rate": {"surface": 0.65083, "split": 0.83233, "merge": 0.18222, "skel": 0.65439}, "mean_gate": {"surface": 0.12231, "split": 0.12203, "merge": 0.12242, "skel": 0.122}, "n_reads": 16699, "n_cand": 26910, "read_suppressed_frac": 0.1022, "dropout_reads_logged": 16699, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 461, "admitted_surface": 16939, "staged_surface": 0, "admitted_split": 17275, "staged_split": 0, "admitted_merge": 19731, "staged_merge": 0, "admitted_skel": 11591, "staged_skel": 0, "mean_count": 67.09294128417969, "n_util_obs": 32, "util_count_spearman": -0.039589442815249266}, "probe_overhead_sec_total": 26.638, "probe_overhead_seq_frac": 0.498, "util_loss": 0.000609, "alpha": 0.03, "t": 231.8} |
| {"event": "probe", "step": 256, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3115.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.178992, "cps": 24.335938, "mean_p": 0.305144, "boundary_entropy": 0.411697, "bpb": 3.280158}, "tr": {"chunks": 3990.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.220405, "cps": 31.171875, "mean_p": 0.347416, "boundary_entropy": 0.51269, "bpb": 3.499224}, "fi": {"chunks": 2989.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.158896, "cps": 23.351562, "mean_p": 0.32597, "boundary_entropy": 0.514176, "bpb": 3.656771}}, "bpb": {"en": 3.28016, "tr": 3.49922, "fi": 3.65677}, "gini_cpc": 0.073449, "gini_cps": 0.066112, "bpic": 5.620369, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.2313695600766401, "fi": 0.8877283731309495}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 469, "admitted_surface": 16943, "staged_surface": 0, "admitted_split": 17222, "staged_split": 0, "admitted_merge": 19742, "staged_merge": 0, "admitted_skel": 11629, "staged_skel": 0, "mean_count": 68.45793151855469, "n_util_obs": 33, "util_count_spearman": -0.06684491978609626}, "t": 238.4} |
| {"event": "consolidate", "step": 300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 279.9} |
| {"event": "train", "step": 300, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.24315, "bpb": 3.23619, "chunk_loss": 0.030299, "F": 0.17758, "G": 0.2834, "chunks": 181.84, "bytes_per_chunk": 5.631, "boundary_entropy": 0.36227, "router_noise": 0.0, "grad_norm": 1.5783, "lr": 0.0003, "router_trainable": true, "sec_per_step": 0.982, "probe_sec": 0.1268, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09667, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.00394, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.00635, "n_merge": 1, "mean_delta_merge": 0.14221, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.66492, "trunc_abs_frac": 0.6304, "mean_delta_full": -0.12842, "read_hit_rate": {"surface": 0.69224, "split": 0.85881, "merge": 0.18262, "skel": 0.70513}, "mean_gate": {"surface": 0.13211, "split": 0.13063, "merge": 0.13276, "skel": 0.13186}, "n_reads": 17639, "n_cand": 27030, "read_suppressed_frac": 0.099, "dropout_reads_logged": 17639, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 538, "admitted_surface": 16986, "staged_surface": 0, "admitted_split": 17520, "staged_split": 0, "admitted_merge": 19243, "staged_merge": 0, "admitted_skel": 11787, "staged_skel": 0, "mean_count": 80.4422607421875, "n_util_obs": 41, "util_count_spearman": -0.15801393728222995}, "probe_overhead_sec_total": 31.828, "probe_overhead_seq_frac": 0.4983, "util_loss": 0.00076, "alpha": 0.03, "t": 280.9} |
| {"event": "consolidate", "step": 350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 314.8} |
| {"event": "train", "step": 350, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.13479, "bpb": 3.07985, "chunk_loss": 0.030647, "F": 0.19189, "G": 0.28071, "chunks": 196.5, "bytes_per_chunk": 5.211, "boundary_entropy": 0.34951, "router_noise": 0.0, "grad_norm": 1.3129, "lr": 0.0002999, "router_trainable": true, "sec_per_step": 0.688, "probe_sec": 0.0816, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05934, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.0934, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": -0.16306, "n_merge": 1, "mean_delta_merge": 0.34546, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -0.99079, "trunc_abs_frac": 0.44475, "mean_delta_full": -0.03146, "read_hit_rate": {"surface": 0.7038, "split": 0.86036, "merge": 0.20951, "skel": 0.70949}, "mean_gate": {"surface": 0.14737, "split": 0.14199, "merge": 0.1486, "skel": 0.14691}, "n_reads": 18662, "n_cand": 28358, "read_suppressed_frac": 0.1012, "dropout_reads_logged": 18662, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 604, "admitted_surface": 16974, "staged_surface": 0, "admitted_split": 17687, "staged_split": 0, "admitted_merge": 19113, "staged_merge": 0, "admitted_skel": 11762, "staged_skel": 0, "mean_count": 94.0546875, "n_util_obs": 54, "util_count_spearman": 0.03472460453592529}, "probe_overhead_sec_total": 36.335, "probe_overhead_seq_frac": 0.4986, "util_loss": 0.001122, "alpha": 0.03, "t": 315.3} |
| {"event": "consolidate", "step": 400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 351.5} |
| {"event": "probe", "step": 400, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 2987.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.171637, "cps": 23.335938, "mean_p": 0.248727, "boundary_entropy": 0.304613, "bpb": 2.855586}, "tr": {"chunks": 3760.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.2077, "cps": 29.375, "mean_p": 0.302245, "boundary_entropy": 0.416522, "bpb": 3.141745}, "fi": {"chunks": 2874.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.152783, "cps": 22.453125, "mean_p": 0.261834, "boundary_entropy": 0.399296, "bpb": 3.358081}}, "bpb": {"en": 2.85559, "tr": 3.14174, "fi": 3.35808}, "gini_cpc": 0.068803, "gini_cps": 0.061393, "bpic": 5.896684, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.2101137372592545, "fi": 0.8901511924415365}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 663, "admitted_surface": 17027, "staged_surface": 0, "admitted_split": 17899, "staged_split": 0, "admitted_merge": 18778, "staged_merge": 0, "admitted_skel": 11832, "staged_skel": 0, "mean_count": 107.42364501953125, "n_util_obs": 58, "util_count_spearman": 0.16146297948260482}, "t": 352.7} |
| {"event": "eval", "step": 400, "phase": 1, "bpb": {"en": 2.86953, "tr": 3.15328, "fi": 3.35283}, "bytes_per_chunk": {"en": 5.272, "tr": 5.059, "fi": 6.26}, "bpb_mean": 3.12521, "frozen_inventory": true, "t": 355.2} |
| {"event": "train", "step": 400, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.03899, "bpb": 2.94164, "chunk_loss": 0.030483, "F": 0.18872, "G": 0.26305, "chunks": 193.25, "bytes_per_chunk": 5.299, "boundary_entropy": 0.31412, "router_noise": 0.0, "grad_norm": 1.2431, "lr": 0.0002998, "router_trainable": true, "sec_per_step": 0.836, "probe_sec": 0.0828, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06438, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": 0.03251, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.22406, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 5.28541, "trunc_abs_frac": 0.43721, "mean_delta_full": 0.01422, "read_hit_rate": {"surface": 0.71099, "split": 0.86808, "merge": 0.21062, "skel": 0.70975}, "mean_gate": {"surface": 0.15705, "split": 0.14954, "merge": 0.15874, "skel": 0.15617}, "n_reads": 18759, "n_cand": 28243, "read_suppressed_frac": 0.0977, "dropout_reads_logged": 18759, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 664, "admitted_surface": 17013, "staged_surface": 0, "admitted_split": 17899, "staged_split": 0, "admitted_merge": 18711, "staged_merge": 0, "admitted_skel": 11913, "staged_skel": 0, "mean_count": 107.71531677246094, "n_util_obs": 58, "util_count_spearman": 0.16589252207081115}, "probe_overhead_sec_total": 41.354, "probe_overhead_seq_frac": 0.4988, "util_loss": 0.001317, "alpha": 0.03, "t": 357.0} |
| {"event": "consolidate", "step": 450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 397.8} |
| {"event": "train", "step": 450, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.04878, "bpb": 2.95576, "chunk_loss": 0.030759, "F": 0.19455, "G": 0.27719, "chunks": 199.22, "bytes_per_chunk": 5.14, "boundary_entropy": 0.33149, "router_noise": 0.0, "grad_norm": 1.125, "lr": 0.0002995, "router_trainable": true, "sec_per_step": 0.833, "probe_sec": 0.1409, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.12, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.04203, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.07812, "n_merge": 2, "mean_delta_merge": -0.0377, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.34061, "trunc_abs_frac": 0.14875, "mean_delta_full": 0.10559, "read_hit_rate": {"surface": 0.69021, "split": 0.85116, "merge": 0.18777, "skel": 0.69157}, "mean_gate": {"surface": 0.163, "split": 0.1557, "merge": 0.16328, "skel": 0.16553}, "n_reads": 18329, "n_cand": 28498, "read_suppressed_frac": 0.0984, "dropout_reads_logged": 18329, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 734, "admitted_surface": 17011, "staged_surface": 0, "admitted_split": 18204, "staged_split": 0, "admitted_merge": 18451, "staged_merge": 0, "admitted_skel": 11870, "staged_skel": 0, "mean_count": 121.42910766601562, "n_util_obs": 64, "util_count_spearman": 0.12742673992673992}, "probe_overhead_sec_total": 46.54, "probe_overhead_seq_frac": 0.4989, "util_loss": 0.001881, "alpha": 0.03, "t": 398.7} |
| {"event": "consolidate", "step": 500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 442.3} |
| {"event": "train", "step": 500, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.89109, "bpb": 2.72827, "chunk_loss": 0.030308, "F": 0.18277, "G": 0.24032, "chunks": 187.16, "bytes_per_chunk": 5.471, "boundary_entropy": 0.25862, "router_noise": 0.0, "grad_norm": 1.1656, "lr": 0.0002992, "router_trainable": true, "sec_per_step": 0.889, "probe_sec": 0.1094, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0867, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.05483, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": -0.01453, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.91305, "trunc_abs_frac": 0.9256, "mean_delta_full": 0.34009, "read_hit_rate": {"surface": 0.7158, "split": 0.86937, "merge": 0.18633, "skel": 0.7254}, "mean_gate": {"surface": 0.17917, "split": 0.16794, "merge": 0.18232, "skel": 0.18005}, "n_reads": 18326, "n_cand": 27568, "read_suppressed_frac": 0.0999, "dropout_reads_logged": 18326, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 801, "admitted_surface": 16915, "staged_surface": 0, "admitted_split": 18344, "staged_split": 0, "admitted_merge": 18301, "staged_merge": 0, "admitted_skel": 11976, "staged_skel": 0, "mean_count": 135.12545776367188, "n_util_obs": 77, "util_count_spearman": 0.11312371838687628}, "probe_overhead_sec_total": 51.824, "probe_overhead_seq_frac": 0.499, "util_loss": 0.002151, "alpha": 0.03, "t": 443.2} |
| {"event": "probe", "step": 512, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 2904.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.166868, "cps": 22.6875, "mean_p": 0.224449, "boundary_entropy": 0.245839, "bpb": 2.67085}, "tr": {"chunks": 3403.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.18798, "cps": 26.585938, "mean_p": 0.265899, "boundary_entropy": 0.362386, "bpb": 2.991329}, "fi": {"chunks": 2988.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.158843, "cps": 23.34375, "mean_p": 0.262624, "boundary_entropy": 0.380842, "bpb": 3.203582}}, "bpb": {"en": 2.67085, "tr": 2.99133, "fi": 3.20358}, "gini_cpc": 0.037813, "gini_cps": 0.03579, "bpic": 6.103497, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.126519998283468, "fi": 0.9519107204286572}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 815, "admitted_surface": 17001, "staged_surface": 0, "admitted_split": 18413, "staged_split": 0, "admitted_merge": 18116, "staged_merge": 0, "admitted_skel": 12006, "staged_skel": 0, "mean_count": 138.13121032714844, "n_util_obs": 80, "util_count_spearman": 0.050117205813408346}, "t": 451.8} |
| {"event": "consolidate", "step": 550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 481.9} |
| {"event": "train", "step": 550, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.9207, "bpb": 2.77098, "chunk_loss": 0.030438, "F": 0.18924, "G": 0.24541, "chunks": 193.78, "bytes_per_chunk": 5.284, "boundary_entropy": 0.26122, "router_noise": 0.0, "grad_norm": 1.1867, "lr": 0.0002987, "router_trainable": true, "sec_per_step": 0.786, "probe_sec": 0.0632, "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": 8, "mean_delta_loo": -0.03105, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": -0.065, "n_merge": 1, "mean_delta_merge": -0.08179, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -2.28625, "trunc_abs_frac": 1.6403, "mean_delta_full": 0.0162, "read_hit_rate": {"surface": 0.69606, "split": 0.8605, "merge": 0.18331, "skel": 0.70322}, "mean_gate": {"surface": 0.18482, "split": 0.17277, "merge": 0.18717, "skel": 0.18713}, "n_reads": 18292, "n_cand": 28106, "read_suppressed_frac": 0.102, "dropout_reads_logged": 18292, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 864, "admitted_surface": 16955, "staged_surface": 0, "admitted_split": 18521, "staged_split": 0, "admitted_merge": 18120, "staged_merge": 0, "admitted_skel": 11940, "staged_skel": 0, "mean_count": 148.80136108398438, "n_util_obs": 89, "util_count_spearman": 0.007865168539325843}, "probe_overhead_sec_total": 57.037, "probe_overhead_seq_frac": 0.4991, "util_loss": 0.002578, "alpha": 0.03, "t": 482.5} |
| {"event": "consolidate", "step": 600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 524.5} |
| {"event": "probe", "step": 600, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 2911.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.16727, "cps": 22.742188, "mean_p": 0.215189, "boundary_entropy": 0.217428, "bpb": 2.562902}, "tr": {"chunks": 3748.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.207038, "cps": 29.28125, "mean_p": 0.268347, "boundary_entropy": 0.349998, "bpb": 2.888271}, "fi": {"chunks": 3070.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.163202, "cps": 23.984375, "mean_p": 0.260669, "boundary_entropy": 0.364882, "bpb": 3.119547}}, "bpb": {"en": 2.5629, "tr": 2.88827, "fi": 3.11955}, "gini_cpc": 0.054368, "gini_cps": 0.057354, "bpic": 5.831226, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.2377443300182762, "fi": 0.9756822558323526}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 920, "admitted_surface": 17005, "staged_surface": 0, "admitted_split": 18749, "staged_split": 0, "admitted_merge": 17686, "staged_merge": 0, "admitted_skel": 12096, "staged_skel": 0, "mean_count": 162.2275848388672, "n_util_obs": 96, "util_count_spearman": 0.08838849701573522}, "t": 525.7} |
| {"event": "eval", "step": 600, "phase": 1, "bpb": {"en": 2.57944, "tr": 2.89879, "fi": 3.11794}, "bytes_per_chunk": {"en": 5.408, "tr": 5.058, "fi": 5.945}, "bpb_mean": 2.86539, "frozen_inventory": true, "t": 528.3} |
| {"event": "train", "step": 600, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.87714, "bpb": 2.70815, "chunk_loss": 0.03026, "F": 0.18292, "G": 0.2345, "chunks": 187.31, "bytes_per_chunk": 5.467, "boundary_entropy": 0.23764, "router_noise": 0.0, "grad_norm": 1.089, "lr": 0.0002982, "router_trainable": true, "sec_per_step": 0.965, "probe_sec": 0.1743, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10913, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.13785, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": 0.07361, "n_merge": 2, "mean_delta_merge": 0.02185, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 1.0341, "trunc_abs_frac": 1.0341, "mean_delta_full": 0.20764, "read_hit_rate": {"surface": 0.69255, "split": 0.86513, "merge": 0.17841, "skel": 0.70757}, "mean_gate": {"surface": 0.19726, "split": 0.18051, "merge": 0.20184, "skel": 0.19559}, "n_reads": 18059, "n_cand": 27642, "read_suppressed_frac": 0.0992, "dropout_reads_logged": 18059, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 924, "admitted_surface": 16978, "staged_surface": 0, "admitted_split": 18672, "staged_split": 0, "admitted_merge": 17849, "staged_merge": 0, "admitted_skel": 12037, "staged_skel": 0, "mean_count": 162.5076141357422, "n_util_obs": 96, "util_count_spearman": 0.08570265870862724}, "probe_overhead_sec_total": 62.201, "probe_overhead_seq_frac": 0.4992, "util_loss": 0.00259, "alpha": 0.03, "t": 530.7} |
| {"event": "consolidate", "step": 650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 571.5} |
| {"event": "train", "step": 650, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.78619, "bpb": 2.57693, "chunk_loss": 0.030201, "F": 0.18124, "G": 0.22405, "chunks": 185.59, "bytes_per_chunk": 5.517, "boundary_entropy": 0.21116, "router_noise": 0.0, "grad_norm": 1.1938, "lr": 0.0002975, "router_trainable": true, "sec_per_step": 0.835, "probe_sec": 0.1246, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08014, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": 0.07598, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.00183, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.09169, "trunc_abs_frac": 0.44578, "mean_delta_full": 0.07764, "read_hit_rate": {"surface": 0.71407, "split": 0.87541, "merge": 0.19813, "skel": 0.73772}, "mean_gate": {"surface": 0.20684, "split": 0.18657, "merge": 0.21077, "skel": 0.20475}, "n_reads": 18510, "n_cand": 27531, "read_suppressed_frac": 0.1017, "dropout_reads_logged": 18510, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 986, "admitted_surface": 16977, "staged_surface": 0, "admitted_split": 18810, "staged_split": 0, "admitted_merge": 17691, "staged_merge": 0, "admitted_skel": 12058, "staged_skel": 0, "mean_count": 176.3673095703125, "n_util_obs": 106, "util_count_spearman": 0.12679547572864447}, "probe_overhead_sec_total": 67.23, "probe_overhead_seq_frac": 0.4992, "util_loss": 0.00313, "alpha": 0.03, "t": 572.5} |
| {"event": "consolidate", "step": 700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 607.8} |
| {"event": "train", "step": 700, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.84089, "bpb": 2.65584, "chunk_loss": 0.030504, "F": 0.18893, "G": 0.23652, "chunks": 193.47, "bytes_per_chunk": 5.293, "boundary_entropy": 0.23178, "router_noise": 0.0, "grad_norm": 1.1074, "lr": 0.0002967, "router_trainable": true, "sec_per_step": 0.73, "probe_sec": 0.1738, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.1512, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.02908, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 3, "mean_delta_merge": -0.04384, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.69282, "trunc_abs_frac": 0.54831, "mean_delta_full": 0.0918, "read_hit_rate": {"surface": 0.68307, "split": 0.85723, "merge": 0.19341, "skel": 0.67023}, "mean_gate": {"surface": 0.21007, "split": 0.19173, "merge": 0.2098, "skel": 0.21277}, "n_reads": 17930, "n_cand": 27872, "read_suppressed_frac": 0.102, "dropout_reads_logged": 17930, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1043, "admitted_surface": 17008, "staged_surface": 0, "admitted_split": 18878, "staged_split": 0, "admitted_merge": 17568, "staged_merge": 0, "admitted_skel": 12082, "staged_skel": 0, "mean_count": 190.11167907714844, "n_util_obs": 114, "util_count_spearman": 0.1878171766468885}, "probe_overhead_sec_total": 72.854, "probe_overhead_seq_frac": 0.4993, "util_loss": 0.003354, "alpha": 0.03, "t": 608.9} |
| {"event": "consolidate", "step": 750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 645.8} |
| {"event": "train", "step": 750, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.76908, "bpb": 2.55225, "chunk_loss": 0.030384, "F": 0.18979, "G": 0.23381, "chunks": 194.34, "bytes_per_chunk": 5.269, "boundary_entropy": 0.21181, "router_noise": 0.0, "grad_norm": 0.9881, "lr": 0.0002959, "router_trainable": true, "sec_per_step": 0.762, "probe_sec": 0.0884, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0627, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.05447, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": 0.10339, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.44942, "trunc_abs_frac": 1.38572, "mean_delta_full": -0.15849, "read_hit_rate": {"surface": 0.7123, "split": 0.87665, "merge": 0.20601, "skel": 0.72466}, "mean_gate": {"surface": 0.21948, "split": 0.19775, "merge": 0.22007, "skel": 0.21689}, "n_reads": 19035, "n_cand": 28425, "read_suppressed_frac": 0.1012, "dropout_reads_logged": 19035, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1124, "admitted_surface": 16978, "staged_surface": 0, "admitted_split": 18928, "staged_split": 0, "admitted_merge": 17548, "staged_merge": 0, "admitted_skel": 12082, "staged_skel": 0, "mean_count": 203.91293334960938, "n_util_obs": 122, "util_count_spearman": 0.12242045330628079}, "probe_overhead_sec_total": 77.633, "probe_overhead_seq_frac": 0.4993, "util_loss": 0.00366, "alpha": 0.03, "t": 647.0} |
| {"event": "consolidate", "step": 800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 691.4} |
| {"event": "probe", "step": 800, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 2910.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.167213, "cps": 22.734375, "mean_p": 0.199801, "boundary_entropy": 0.166829, "bpb": 2.407962}, "tr": {"chunks": 3688.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.203723, "cps": 28.8125, "mean_p": 0.238046, "boundary_entropy": 0.284022, "bpb": 2.723193}, "fi": {"chunks": 2979.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.158365, "cps": 23.273438, "mean_p": 0.23151, "boundary_entropy": 0.306768, "bpb": 2.964563}}, "bpb": {"en": 2.40796, "tr": 2.72319, "fi": 2.96456}, "gini_cpc": 0.05713, "gini_cps": 0.054158, "bpic": 5.923776, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.2183483856124548, "fi": 0.9470867287017303}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1190, "admitted_surface": 16943, "staged_surface": 0, "admitted_split": 18997, "staged_split": 0, "admitted_merge": 17573, "staged_merge": 0, "admitted_skel": 12023, "staged_skel": 0, "mean_count": 217.4747314453125, "n_util_obs": 129, "util_count_spearman": 0.03538685152057245}, "t": 692.8} |
| {"event": "eval", "step": 800, "phase": 1, "bpb": {"en": 2.42655, "tr": 2.73167, "fi": 2.97027}, "bytes_per_chunk": {"en": 5.43, "tr": 5.141, "fi": 6.006}, "bpb_mean": 2.70949, "frozen_inventory": true, "t": 695.2} |
| {"event": "train", "step": 800, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.74733, "bpb": 2.52086, "chunk_loss": 0.030289, "F": 0.1868, "G": 0.22221, "chunks": 191.28, "bytes_per_chunk": 5.353, "boundary_entropy": 0.18508, "router_noise": 0.0, "grad_norm": 1.1291, "lr": 0.0002949, "router_trainable": true, "sec_per_step": 0.997, "probe_sec": 0.0896, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06464, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.12595, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": 0.05814, "n_merge": 1, "mean_delta_merge": -0.17224, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.99151, "trunc_abs_frac": 0.99151, "mean_delta_full": -0.23718, "read_hit_rate": {"surface": 0.70948, "split": 0.87471, "merge": 0.2067, "skel": 0.72459}, "mean_gate": {"surface": 0.22961, "split": 0.20207, "merge": 0.23161, "skel": 0.22277}, "n_reads": 18833, "n_cand": 28139, "read_suppressed_frac": 0.0996, "dropout_reads_logged": 18833, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1192, "admitted_surface": 16952, "staged_surface": 0, "admitted_split": 19036, "staged_split": 0, "admitted_merge": 17469, "staged_merge": 0, "admitted_skel": 12079, "staged_skel": 0, "mean_count": 217.76051330566406, "n_util_obs": 129, "util_count_spearman": 0.03913796958855099}, "probe_overhead_sec_total": 82.977, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.003901, "alpha": 0.03, "t": 696.9} |
| {"event": "consolidate", "step": 850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 731.1} |
| {"event": "train", "step": 850, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.68815, "bpb": 2.43549, "chunk_loss": 0.03023, "F": 0.18243, "G": 0.21818, "chunks": 186.81, "bytes_per_chunk": 5.481, "boundary_entropy": 0.1795, "router_noise": 0.0, "grad_norm": 1.0225, "lr": 0.0002938, "router_trainable": true, "sec_per_step": 0.713, "probe_sec": 0.111, "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": 8, "mean_delta_loo": 0.02337, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": 0.00583, "n_merge": 1, "mean_delta_merge": 0.00781, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.77134, "trunc_abs_frac": 0.90446, "mean_delta_full": 0.10223, "read_hit_rate": {"surface": 0.72351, "split": 0.88062, "merge": 0.20003, "skel": 0.73258}, "mean_gate": {"surface": 0.23614, "split": 0.20941, "merge": 0.23178, "skel": 0.23125}, "n_reads": 18738, "n_cand": 27730, "read_suppressed_frac": 0.1001, "dropout_reads_logged": 18738, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1252, "admitted_surface": 16868, "staged_surface": 0, "admitted_split": 19170, "staged_split": 0, "admitted_merge": 17371, "staged_merge": 0, "admitted_skel": 12127, "staged_skel": 0, "mean_count": 231.69125366210938, "n_util_obs": 141, "util_count_spearman": -0.005766513977481913}, "probe_overhead_sec_total": 87.472, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.004025, "alpha": 0.03, "t": 732.5} |
| {"event": "consolidate", "step": 900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 765.7} |
| {"event": "train", "step": 900, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.68645, "bpb": 2.43303, "chunk_loss": 0.03024, "F": 0.18546, "G": 0.21791, "chunks": 189.91, "bytes_per_chunk": 5.392, "boundary_entropy": 0.17133, "router_noise": 0.0, "grad_norm": 0.9622, "lr": 0.0002927, "router_trainable": true, "sec_per_step": 0.673, "probe_sec": 0.0609, "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": 12, "mean_delta_loo": -0.00369, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.05884, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -2.19035, "trunc_abs_frac": 26.1934, "mean_delta_full": -0.03121, "read_hit_rate": {"surface": 0.71447, "split": 0.87824, "merge": 0.2009, "skel": 0.73661}, "mean_gate": {"surface": 0.24408, "split": 0.21359, "merge": 0.24059, "skel": 0.23694}, "n_reads": 18849, "n_cand": 28007, "read_suppressed_frac": 0.1006, "dropout_reads_logged": 18849, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1310, "admitted_surface": 16940, "staged_surface": 0, "admitted_split": 19247, "staged_split": 0, "admitted_merge": 17236, "staged_merge": 0, "admitted_skel": 12113, "staged_skel": 0, "mean_count": 245.54490661621094, "n_util_obs": 145, "util_count_spearman": -0.03866520888957243}, "probe_overhead_sec_total": 91.806, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.0043, "alpha": 0.03, "t": 766.2} |
| {"event": "consolidate", "step": 950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 804.2} |
| {"event": "train", "step": 950, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.73067, "bpb": 2.49682, "chunk_loss": 0.030374, "F": 0.19199, "G": 0.22494, "chunks": 196.59, "bytes_per_chunk": 5.209, "boundary_entropy": 0.18037, "router_noise": 0.0, "grad_norm": 1.0388, "lr": 0.0002914, "router_trainable": true, "sec_per_step": 0.772, "probe_sec": 0.0811, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05364, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": -0.03244, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.15039, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.41122, "trunc_abs_frac": 1.68297, "mean_delta_full": -0.11963, "read_hit_rate": {"surface": 0.69995, "split": 0.87412, "merge": 0.20186, "skel": 0.7046}, "mean_gate": {"surface": 0.24738, "split": 0.21391, "merge": 0.2434, "skel": 0.23977}, "n_reads": 18891, "n_cand": 28566, "read_suppressed_frac": 0.0986, "dropout_reads_logged": 18891, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1365, "admitted_surface": 17001, "staged_surface": 0, "admitted_split": 19244, "staged_split": 0, "admitted_merge": 17142, "staged_merge": 0, "admitted_skel": 12149, "staged_skel": 0, "mean_count": 259.34844970703125, "n_util_obs": 154, "util_count_spearman": -0.04370740811330001}, "probe_overhead_sec_total": 95.781, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.004241, "alpha": 0.03, "t": 804.8} |
| {"event": "consolidate", "step": 1000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 837.8} |
| {"event": "probe", "step": 1000, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 2916.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.167557, "cps": 22.78125, "mean_p": 0.194161, "boundary_entropy": 0.144029, "bpb": 2.310507}, "tr": {"chunks": 3703.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.204552, "cps": 28.929688, "mean_p": 0.231984, "boundary_entropy": 0.260293, "bpb": 2.618151}, "fi": {"chunks": 3007.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.159853, "cps": 23.492188, "mean_p": 0.210657, "boundary_entropy": 0.255265, "bpb": 2.844156}}, "bpb": {"en": 2.31051, "tr": 2.61815, "fi": 2.84416}, "gini_cpc": 0.056017, "gini_cps": 0.054505, "bpic": 5.893621, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.2207866213202958, "fi": 0.9540214627944029}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1414, "admitted_surface": 16932, "staged_surface": 0, "admitted_split": 19415, "staged_split": 0, "admitted_merge": 17071, "staged_merge": 0, "admitted_skel": 12118, "staged_skel": 0, "mean_count": 272.9180603027344, "n_util_obs": 159, "util_count_spearman": 0.02090398853594459}, "t": 839.1} |
| {"event": "eval", "step": 1000, "phase": 1, "bpb": {"en": 2.32634, "tr": 2.62797, "fi": 2.85352}, "bytes_per_chunk": {"en": 5.431, "tr": 5.123, "fi": 6.043}, "bpb_mean": 2.60261, "frozen_inventory": true, "t": 841.2} |
| {"event": "train", "step": 1000, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.71731, "bpb": 2.47756, "chunk_loss": 0.030218, "F": 0.18228, "G": 0.21469, "chunks": 186.66, "bytes_per_chunk": 5.486, "boundary_entropy": 0.17027, "router_noise": 0.0, "grad_norm": 1.1382, "lr": 0.0002901, "router_trainable": true, "sec_per_step": 0.768, "probe_sec": 0.0562, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03775, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.07846, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 3, "mean_delta_split": 0.02163, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 7, "trunc_ratio_sum": -0.64966, "trunc_abs_frac": 1.58481, "mean_delta_full": -0.03936, "read_hit_rate": {"surface": 0.69803, "split": 0.86525, "merge": 0.19631, "skel": 0.70873}, "mean_gate": {"surface": 0.25367, "split": 0.21689, "merge": 0.24972, "skel": 0.24559}, "n_reads": 18059, "n_cand": 27435, "read_suppressed_frac": 0.1012, "dropout_reads_logged": 18059, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1416, "admitted_surface": 16984, "staged_surface": 0, "admitted_split": 19456, "staged_split": 0, "admitted_merge": 16904, "staged_merge": 0, "admitted_skel": 12192, "staged_skel": 0, "mean_count": 273.19451904296875, "n_util_obs": 159, "util_count_spearman": 0.021566754239312157}, "probe_overhead_sec_total": 100.069, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.004572, "alpha": 0.03, "t": 843.2} |
| {"event": "probe", "step": 1024, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 2917.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.167615, "cps": 22.789062, "mean_p": 0.191958, "boundary_entropy": 0.13628, "bpb": 2.304965}, "tr": {"chunks": 3780.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.208805, "cps": 29.53125, "mean_p": 0.231804, "boundary_entropy": 0.250005, "bpb": 2.605111}, "fi": {"chunks": 3019.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.160491, "cps": 23.585938, "mean_p": 0.213005, "boundary_entropy": 0.254601, "bpb": 2.843847}}, "bpb": {"en": 2.30496, "tr": 2.60511, "fi": 2.84385}, "gini_cpc": 0.05999, "gini_cps": 0.059215, "bpic": 5.839028, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.2457443883134658, "fi": 0.957500304300832}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1442, "admitted_surface": 16929, "staged_surface": 0, "admitted_split": 19399, "staged_split": 0, "admitted_merge": 17028, "staged_merge": 0, "admitted_skel": 12180, "staged_skel": 0, "mean_count": 279.6255187988281, "n_util_obs": 161, "util_count_spearman": 0.006984702093397746}, "t": 858.0} |
| {"event": "consolidate", "step": 1050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 874.5} |
| {"event": "train", "step": 1050, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.60861, "bpb": 2.32073, "chunk_loss": 0.030272, "F": 0.186, "G": 0.21337, "chunks": 190.47, "bytes_per_chunk": 5.376, "boundary_entropy": 0.15097, "router_noise": 0.0, "grad_norm": 0.9545, "lr": 0.0002886, "router_trainable": true, "sec_per_step": 0.647, "probe_sec": 0.0836, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05718, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.09433, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": 0.12968, "n_merge": 2, "mean_delta_merge": 0.02759, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.60816, "trunc_abs_frac": 0.43811, "mean_delta_full": 0.06299, "read_hit_rate": {"surface": 0.73495, "split": 0.89194, "merge": 0.19881, "skel": 0.74878}, "mean_gate": {"surface": 0.26254, "split": 0.22274, "merge": 0.256, "skel": 0.25169}, "n_reads": 19381, "n_cand": 28272, "read_suppressed_frac": 0.1, "dropout_reads_logged": 19381, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1467, "admitted_surface": 16924, "staged_surface": 0, "admitted_split": 19379, "staged_split": 0, "admitted_merge": 17076, "staged_merge": 0, "admitted_skel": 12157, "staged_skel": 0, "mean_count": 287.1092224121094, "n_util_obs": 166, "util_count_spearman": 0.03255703219434704}, "probe_overhead_sec_total": 104.391, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.004885, "alpha": 0.03, "t": 875.5} |
| {"event": "consolidate", "step": 1100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 907.2} |
| {"event": "train", "step": 1100, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.73418, "bpb": 2.5019, "chunk_loss": 0.030225, "F": 0.18311, "G": 0.21339, "chunks": 187.5, "bytes_per_chunk": 5.461, "boundary_entropy": 0.16931, "router_noise": 0.0, "grad_norm": 0.9725, "lr": 0.0002871, "router_trainable": true, "sec_per_step": 0.645, "probe_sec": 0.0658, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03648, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.02534, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.05518, "n_merge": 1, "mean_delta_merge": 0.08311, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 1.53292, "trunc_abs_frac": 2.70145, "mean_delta_full": 0.06952, "read_hit_rate": {"surface": 0.68314, "split": 0.86139, "merge": 0.17419, "skel": 0.7048}, "mean_gate": {"surface": 0.2564, "split": 0.21552, "merge": 0.24781, "skel": 0.24069}, "n_reads": 17761, "n_cand": 27411, "read_suppressed_frac": 0.0994, "dropout_reads_logged": 17761, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1521, "admitted_surface": 16985, "staged_surface": 0, "admitted_split": 19527, "staged_split": 0, "admitted_merge": 16894, "staged_merge": 0, "admitted_skel": 12130, "staged_skel": 0, "mean_count": 301.0565185546875, "n_util_obs": 173, "util_count_spearman": 0.032807060644969896}, "probe_overhead_sec_total": 108.216, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.005091, "alpha": 0.03, "t": 907.8} |
| {"event": "consolidate", "step": 1150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 948.4} |
| {"event": "train", "step": 1150, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.62267, "bpb": 2.34102, "chunk_loss": 0.03026, "F": 0.18716, "G": 0.21664, "chunks": 191.66, "bytes_per_chunk": 5.343, "boundary_entropy": 0.16646, "router_noise": 0.0, "grad_norm": 0.8412, "lr": 0.0002855, "router_trainable": true, "sec_per_step": 0.823, "probe_sec": 0.0702, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05158, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.01261, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.10278, "n_merge": 1, "mean_delta_merge": -0.11948, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.2041, "trunc_abs_frac": 0.8783, "mean_delta_full": 0.11702, "read_hit_rate": {"surface": 0.7071, "split": 0.87055, "merge": 0.20811, "skel": 0.71987}, "mean_gate": {"surface": 0.27588, "split": 0.22894, "merge": 0.25997, "skel": 0.26052}, "n_reads": 18655, "n_cand": 28007, "read_suppressed_frac": 0.1009, "dropout_reads_logged": 18655, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1575, "admitted_surface": 16972, "staged_surface": 0, "admitted_split": 19615, "staged_split": 0, "admitted_merge": 16758, "staged_merge": 0, "admitted_skel": 12191, "staged_skel": 0, "mean_count": 314.9972229003906, "n_util_obs": 187, "util_count_spearman": -0.03867370870479595}, "probe_overhead_sec_total": 113.052, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.005424, "alpha": 0.03, "t": 948.9} |
| {"event": "consolidate", "step": 1200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 980.2} |
| {"event": "probe", "step": 1200, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 2915.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.1675, "cps": 22.773438, "mean_p": 0.188479, "boundary_entropy": 0.123866, "bpb": 2.242019}, "tr": {"chunks": 3729.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.205988, "cps": 29.132812, "mean_p": 0.221866, "boundary_entropy": 0.233489, "bpb": 2.526101}, "fi": {"chunks": 3024.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.160757, "cps": 23.625, "mean_p": 0.202009, "boundary_entropy": 0.231518, "bpb": 2.720152}}, "bpb": {"en": 2.24202, "tr": 2.5261, "fi": 2.72015}, "gini_cpc": 0.056442, "gini_cps": 0.05613, "bpic": 5.868018, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.2297799072185471, "fi": 0.9597441298579633}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1626, "admitted_surface": 16939, "staged_surface": 0, "admitted_split": 19659, "staged_split": 0, "admitted_merge": 16677, "staged_merge": 0, "admitted_skel": 12261, "staged_skel": 0, "mean_count": 328.73602294921875, "n_util_obs": 195, "util_count_spearman": -0.0312474838755645}, "t": 981.4} |
| {"event": "eval", "step": 1200, "phase": 1, "bpb": {"en": 2.25817, "tr": 2.53809, "fi": 2.73069}, "bytes_per_chunk": {"en": 5.426, "tr": 5.056, "fi": 6.017}, "bpb_mean": 2.50898, "frozen_inventory": true, "t": 983.6} |
| {"event": "train", "step": 1200, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.70947, "bpb": 2.46624, "chunk_loss": 0.030329, "F": 0.18945, "G": 0.21621, "chunks": 194.0, "bytes_per_chunk": 5.278, "boundary_entropy": 0.16529, "router_noise": 0.0, "grad_norm": 0.9515, "lr": 0.0002837, "router_trainable": true, "sec_per_step": 0.732, "probe_sec": 0.0579, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03844, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.03972, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": 0.3278, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.47849, "trunc_abs_frac": 1.09534, "mean_delta_full": 0.08333, "read_hit_rate": {"surface": 0.67957, "split": 0.85984, "merge": 0.18962, "skel": 0.68305}, "mean_gate": {"surface": 0.27441, "split": 0.22458, "merge": 0.2571, "skel": 0.25568}, "n_reads": 18090, "n_cand": 28064, "read_suppressed_frac": 0.0957, "dropout_reads_logged": 18090, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1628, "admitted_surface": 16985, "staged_surface": 0, "admitted_split": 19613, "staged_split": 0, "admitted_merge": 16760, "staged_merge": 0, "admitted_skel": 12178, "staged_skel": 0, "mean_count": 328.9923095703125, "n_util_obs": 195, "util_count_spearman": -0.032457233488161326}, "probe_overhead_sec_total": 116.94, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.005448, "alpha": 0.03, "t": 985.5} |
| {"event": "consolidate", "step": 1250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1017.3} |
| {"event": "train", "step": 1250, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.6238, "bpb": 2.34265, "chunk_loss": 0.030194, "F": 0.18396, "G": 0.20531, "chunks": 188.38, "bytes_per_chunk": 5.436, "boundary_entropy": 0.1323, "router_noise": 0.0, "grad_norm": 0.9225, "lr": 0.0002819, "router_trainable": true, "sec_per_step": 0.651, "probe_sec": 0.0804, "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": 4, "mean_delta_loo": 0.00453, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 4, "mean_delta_split": -0.09183, "n_merge": 1, "mean_delta_merge": -0.24066, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.9018, "trunc_abs_frac": 0.785, "mean_delta_full": -0.08142, "read_hit_rate": {"surface": 0.72198, "split": 0.882, "merge": 0.2054, "skel": 0.73741}, "mean_gate": {"surface": 0.27764, "split": 0.2227, "merge": 0.25952, "skel": 0.25852}, "n_reads": 19059, "n_cand": 28089, "read_suppressed_frac": 0.0925, "dropout_reads_logged": 19059, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1675, "admitted_surface": 16929, "staged_surface": 0, "admitted_split": 19668, "staged_split": 0, "admitted_merge": 16712, "staged_merge": 0, "admitted_skel": 12227, "staged_skel": 0, "mean_count": 343.161376953125, "n_util_obs": 199, "util_count_spearman": 0.001131415510442752}, "probe_overhead_sec_total": 120.991, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.005826, "alpha": 0.03, "t": 1018.1} |
| {"event": "consolidate", "step": 1300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1055.1} |
| {"event": "train", "step": 1300, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.67518, "bpb": 2.41678, "chunk_loss": 0.030279, "F": 0.18756, "G": 0.21117, "chunks": 192.06, "bytes_per_chunk": 5.332, "boundary_entropy": 0.14372, "router_noise": 0.0, "grad_norm": 0.8759, "lr": 0.00028, "router_trainable": true, "sec_per_step": 0.749, "probe_sec": 0.058, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03941, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.06133, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": -0.23332, "n_merge": 1, "mean_delta_merge": 0.29919, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -1.13653, "trunc_abs_frac": 0.41991, "mean_delta_full": 0.13497, "read_hit_rate": {"surface": 0.69071, "split": 0.86431, "merge": 0.1978, "skel": 0.70942}, "mean_gate": {"surface": 0.28734, "split": 0.22984, "merge": 0.26451, "skel": 0.26729}, "n_reads": 18294, "n_cand": 27931, "read_suppressed_frac": 0.0988, "dropout_reads_logged": 18294, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1714, "admitted_surface": 16984, "staged_surface": 0, "admitted_split": 19738, "staged_split": 0, "admitted_merge": 16599, "staged_merge": 0, "admitted_skel": 12215, "staged_skel": 0, "mean_count": 357.04193115234375, "n_util_obs": 205, "util_count_spearman": 0.03949521510190065}, "probe_overhead_sec_total": 125.279, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.005824, "alpha": 0.03, "t": 1055.5} |
| {"event": "consolidate", "step": 1350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1088.8} |
| {"event": "train", "step": 1350, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.57937, "bpb": 2.27854, "chunk_loss": 0.03022, "F": 0.18381, "G": 0.20282, "chunks": 188.22, "bytes_per_chunk": 5.44, "boundary_entropy": 0.11951, "router_noise": 0.0, "grad_norm": 0.8637, "lr": 0.000278, "router_trainable": true, "sec_per_step": 0.673, "probe_sec": 0.0685, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04863, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.04246, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.07837, "n_merge": 1, "mean_delta_merge": -0.2132, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -0.17635, "trunc_abs_frac": 5.38191, "mean_delta_full": 0.11655, "read_hit_rate": {"surface": 0.72559, "split": 0.88233, "merge": 0.20574, "skel": 0.7355}, "mean_gate": {"surface": 0.29289, "split": 0.22991, "merge": 0.26956, "skel": 0.27248}, "n_reads": 18819, "n_cand": 27766, "read_suppressed_frac": 0.1012, "dropout_reads_logged": 18819, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1769, "admitted_surface": 16933, "staged_surface": 0, "admitted_split": 19800, "staged_split": 0, "admitted_merge": 16551, "staged_merge": 0, "admitted_skel": 12252, "staged_skel": 0, "mean_count": 370.88470458984375, "n_util_obs": 212, "util_count_spearman": 0.002435791270068672}, "probe_overhead_sec_total": 129.788, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.006163, "alpha": 0.03, "t": 1089.2} |
| {"event": "consolidate", "step": 1400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1118.9} |
| {"event": "probe", "step": 1400, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 2922.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.167902, "cps": 22.828125, "mean_p": 0.18565, "boundary_entropy": 0.106495, "bpb": 2.188558}, "tr": {"chunks": 3667.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.202563, "cps": 28.648438, "mean_p": 0.219953, "boundary_entropy": 0.207606, "bpb": 2.466663}, "fi": {"chunks": 3055.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.162405, "cps": 23.867188, "mean_p": 0.200375, "boundary_entropy": 0.216624, "bpb": 2.665486}}, "bpb": {"en": 2.18856, "tr": 2.46666, "fi": 2.66549}, "gini_cpc": 0.050241, "gini_cps": 0.0515, "bpic": 5.882621, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.2064359419026038, "fi": 0.9672600253445136}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1823, "admitted_surface": 16937, "staged_surface": 0, "admitted_split": 19818, "staged_split": 0, "admitted_merge": 16549, "staged_merge": 0, "admitted_skel": 12232, "staged_skel": 0, "mean_count": 384.5842590332031, "n_util_obs": 218, "util_count_spearman": 0.02338549640536194}, "t": 1120.2} |
| {"event": "eval", "step": 1400, "phase": 1, "bpb": {"en": 2.20314, "tr": 2.47515, "fi": 2.67642}, "bytes_per_chunk": {"en": 5.429, "tr": 5.172, "fi": 5.949}, "bpb_mean": 2.45157, "frozen_inventory": true, "t": 1122.3} |
| {"event": "train", "step": 1400, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.59312, "bpb": 2.29839, "chunk_loss": 0.030185, "F": 0.1824, "G": 0.2019, "chunks": 186.78, "bytes_per_chunk": 5.482, "boundary_entropy": 0.11696, "router_noise": 0.0, "grad_norm": 0.8765, "lr": 0.0002759, "router_trainable": true, "sec_per_step": 0.693, "probe_sec": 0.0637, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04392, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.14031, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": 0.00922, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.44376, "trunc_abs_frac": 0.43896, "mean_delta_full": 0.34943, "read_hit_rate": {"surface": 0.71001, "split": 0.88564, "merge": 0.19127, "skel": 0.73122}, "mean_gate": {"surface": 0.29439, "split": 0.22825, "merge": 0.26778, "skel": 0.26951}, "n_reads": 18702, "n_cand": 27779, "read_suppressed_frac": 0.1033, "dropout_reads_logged": 18702, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1823, "admitted_surface": 16961, "staged_surface": 0, "admitted_split": 19802, "staged_split": 0, "admitted_merge": 16531, "staged_merge": 0, "admitted_skel": 12242, "staged_skel": 0, "mean_count": 384.8619384765625, "n_util_obs": 218, "util_count_spearman": 0.024445340839890094}, "probe_overhead_sec_total": 133.577, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.005835, "alpha": 0.03, "t": 1123.8} |
| {"event": "consolidate", "step": 1450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1152.5} |
| {"event": "train", "step": 1450, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.58199, "bpb": 2.28233, "chunk_loss": 0.030215, "F": 0.18451, "G": 0.20466, "chunks": 188.94, "bytes_per_chunk": 5.42, "boundary_entropy": 0.12297, "router_noise": 0.0, "grad_norm": 0.8971, "lr": 0.0002738, "router_trainable": true, "sec_per_step": 0.582, "probe_sec": 0.1017, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08171, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.03518, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": -0.1072, "n_merge": 1, "mean_delta_merge": -0.36011, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 1.07467, "trunc_abs_frac": 0.59421, "mean_delta_full": -0.16379, "read_hit_rate": {"surface": 0.70269, "split": 0.87548, "merge": 0.19559, "skel": 0.71718}, "mean_gate": {"surface": 0.31068, "split": 0.23995, "merge": 0.27938, "skel": 0.28509}, "n_reads": 18435, "n_cand": 27738, "read_suppressed_frac": 0.1, "dropout_reads_logged": 18435, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1866, "admitted_surface": 16991, "staged_surface": 0, "admitted_split": 19897, "staged_split": 0, "admitted_merge": 16383, "staged_merge": 0, "admitted_skel": 12265, "staged_skel": 0, "mean_count": 398.78302001953125, "n_util_obs": 222, "util_count_spearman": 0.015477624815530382}, "probe_overhead_sec_total": 137.398, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.005976, "alpha": 0.03, "t": 1153.0} |
| {"event": "consolidate", "step": 1500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1183.7} |
| {"event": "train", "step": 1500, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.60729, "bpb": 2.31883, "chunk_loss": 0.030211, "F": 0.18558, "G": 0.20797, "chunks": 190.03, "bytes_per_chunk": 5.389, "boundary_entropy": 0.13308, "router_noise": 0.0, "grad_norm": 0.8059, "lr": 0.0002715, "router_trainable": true, "sec_per_step": 0.633, "probe_sec": 0.0952, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06125, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.25116, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 3, "mean_delta_merge": 0.10352, "n_dropped_unpaired": 0, "n_full": 0, "read_hit_rate": {"surface": 0.68987, "split": 0.86734, "merge": 0.19295, "skel": 0.70129}, "mean_gate": {"surface": 0.29675, "split": 0.22744, "merge": 0.26282, "skel": 0.26861}, "n_reads": 18108, "n_cand": 27686, "read_suppressed_frac": 0.0975, "dropout_reads_logged": 18108, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1910, "admitted_surface": 16994, "staged_surface": 0, "admitted_split": 19903, "staged_split": 0, "admitted_merge": 16362, "staged_merge": 0, "admitted_skel": 12277, "staged_skel": 0, "mean_count": 412.717041015625, "n_util_obs": 231, "util_count_spearman": 0.03871376000311533}, "probe_overhead_sec_total": 141.376, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.00618, "alpha": 0.03, "t": 1184.6} |
| {"event": "consolidate", "step": 1550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1219.4} |
| {"event": "train", "step": 1550, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.54586, "bpb": 2.23021, "chunk_loss": 0.030172, "F": 0.18127, "G": 0.20204, "chunks": 185.62, "bytes_per_chunk": 5.516, "boundary_entropy": 0.12219, "router_noise": 0.0, "grad_norm": 0.8029, "lr": 0.0002692, "router_trainable": true, "sec_per_step": 0.705, "probe_sec": 0.06, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03959, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.17637, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.18115, "n_merge": 1, "mean_delta_merge": 0.216, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 3.01188, "trunc_abs_frac": 1.8222, "mean_delta_full": 0.0976, "read_hit_rate": {"surface": 0.72089, "split": 0.87836, "merge": 0.19997, "skel": 0.7334}, "mean_gate": {"surface": 0.29576, "split": 0.22582, "merge": 0.26183, "skel": 0.26549}, "n_reads": 18555, "n_cand": 27515, "read_suppressed_frac": 0.1022, "dropout_reads_logged": 18555, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1966, "admitted_surface": 16948, "staged_surface": 0, "admitted_split": 19966, "staged_split": 0, "admitted_merge": 16340, "staged_merge": 0, "admitted_skel": 12282, "staged_skel": 0, "mean_count": 426.69439697265625, "n_util_obs": 238, "util_count_spearman": 0.02037441061616571}, "probe_overhead_sec_total": 145.542, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.006032, "alpha": 0.03, "t": 1219.9} |
| {"event": "consolidate", "step": 1600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1253.6} |
| {"event": "probe", "step": 1600, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 2920.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.167787, "cps": 22.8125, "mean_p": 0.183109, "boundary_entropy": 0.094294, "bpb": 2.15013}, "tr": {"chunks": 3582.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.197868, "cps": 27.984375, "mean_p": 0.216983, "boundary_entropy": 0.19854, "bpb": 2.408686}, "fi": {"chunks": 3004.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.159694, "cps": 23.46875, "mean_p": 0.187012, "boundary_entropy": 0.186526, "bpb": 2.596967}}, "bpb": {"en": 2.15013, "tr": 2.40869, "fi": 2.59697}, "gini_cpc": 0.048443, "gini_cps": 0.046427, "bpic": 5.96802, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.1792782774973345, "fi": 0.9517640873199372}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2001, "admitted_surface": 16933, "staged_surface": 0, "admitted_split": 19970, "staged_split": 0, "admitted_merge": 16406, "staged_merge": 0, "admitted_skel": 12227, "staged_skel": 0, "mean_count": 440.2676696777344, "n_util_obs": 243, "util_count_spearman": 0.004404033713522039}, "t": 1254.8} |
| {"event": "eval", "step": 1600, "phase": 1, "bpb": {"en": 2.16235, "tr": 2.41616, "fi": 2.60674}, "bytes_per_chunk": {"en": 5.43, "tr": 5.293, "fi": 6.064}, "bpb_mean": 2.39508, "frozen_inventory": true, "t": 1257.0} |
| {"event": "train", "step": 1600, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.57723, "bpb": 2.27547, "chunk_loss": 0.030131, "F": 0.17877, "G": 0.19885, "chunks": 183.06, "bytes_per_chunk": 5.594, "boundary_entropy": 0.1147, "router_noise": 0.0, "grad_norm": 1.0167, "lr": 0.0002668, "router_trainable": true, "sec_per_step": 0.774, "probe_sec": 0.0659, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04695, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.1045, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": -0.08221, "n_merge": 1, "mean_delta_merge": 0.22021, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.60036, "trunc_abs_frac": 0.58566, "mean_delta_full": 0.25635, "read_hit_rate": {"surface": 0.71456, "split": 0.87523, "merge": 0.20349, "skel": 0.73129}, "mean_gate": {"surface": 0.3107, "split": 0.23461, "merge": 0.27064, "skel": 0.28166}, "n_reads": 18145, "n_cand": 27004, "read_suppressed_frac": 0.0946, "dropout_reads_logged": 18145, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2002, "admitted_surface": 16958, "staged_surface": 0, "admitted_split": 20058, "staged_split": 0, "admitted_merge": 16268, "staged_merge": 0, "admitted_skel": 12252, "staged_skel": 0, "mean_count": 440.54510498046875, "n_util_obs": 244, "util_count_spearman": 0.01086370188634045}, "probe_overhead_sec_total": 149.609, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.006234, "alpha": 0.03, "t": 1258.5} |
| {"event": "consolidate", "step": 1650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1287.9} |
| {"event": "train", "step": 1650, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.50372, "bpb": 2.16941, "chunk_loss": 0.030278, "F": 0.18951, "G": 0.2071, "chunks": 194.06, "bytes_per_chunk": 5.277, "boundary_entropy": 0.11562, "router_noise": 0.0, "grad_norm": 0.8065, "lr": 0.0002643, "router_trainable": true, "sec_per_step": 0.597, "probe_sec": 0.0762, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04829, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": 0.03305, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.63666, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 1.19354, "trunc_abs_frac": 1.27867, "mean_delta_full": -0.08142, "read_hit_rate": {"surface": 0.74183, "split": 0.89835, "merge": 0.21624, "skel": 0.74979}, "mean_gate": {"surface": 0.31922, "split": 0.238, "merge": 0.27424, "skel": 0.28293}, "n_reads": 19567, "n_cand": 28300, "read_suppressed_frac": 0.1049, "dropout_reads_logged": 19567, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2045, "admitted_surface": 16910, "staged_surface": 0, "admitted_split": 19959, "staged_split": 0, "admitted_merge": 16435, "staged_merge": 0, "admitted_skel": 12232, "staged_skel": 0, "mean_count": 454.6053771972656, "n_util_obs": 248, "util_count_spearman": 0.043839718282821766}, "probe_overhead_sec_total": 153.468, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.006804, "alpha": 0.03, "t": 1288.4} |
| {"event": "consolidate", "step": 1700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1318.8} |
| {"event": "train", "step": 1700, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.57845, "bpb": 2.27722, "chunk_loss": 0.030284, "F": 0.18777, "G": 0.20791, "chunks": 192.28, "bytes_per_chunk": 5.326, "boundary_entropy": 0.13155, "router_noise": 0.0, "grad_norm": 0.7703, "lr": 0.0002618, "router_trainable": true, "sec_per_step": 0.627, "probe_sec": 0.1884, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.17112, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.00473, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": -0.11424, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -0.01114, "trunc_abs_frac": 1.25079, "mean_delta_full": -0.21753, "read_hit_rate": {"surface": 0.67767, "split": 0.85331, "merge": 0.18969, "skel": 0.66808}, "mean_gate": {"surface": 0.30259, "split": 0.2266, "merge": 0.25274, "skel": 0.2651}, "n_reads": 17396, "n_cand": 27304, "read_suppressed_frac": 0.0991, "dropout_reads_logged": 17396, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2092, "admitted_surface": 16974, "staged_surface": 0, "admitted_split": 20138, "staged_split": 0, "admitted_merge": 16147, "staged_merge": 0, "admitted_skel": 12277, "staged_skel": 0, "mean_count": 468.5660400390625, "n_util_obs": 251, "util_count_spearman": 0.04723317154869589}, "probe_overhead_sec_total": 158.041, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.006872, "alpha": 0.03, "t": 1319.7} |
| {"event": "consolidate", "step": 1750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1353.6} |
| {"event": "train", "step": 1750, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.5352, "bpb": 2.21483, "chunk_loss": 0.030172, "F": 0.18396, "G": 0.20094, "chunks": 188.38, "bytes_per_chunk": 5.436, "boundary_entropy": 0.11016, "router_noise": 0.0, "grad_norm": 0.7784, "lr": 0.0002591, "router_trainable": true, "sec_per_step": 0.686, "probe_sec": 0.0735, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05228, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.06978, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": -0.00769, "n_merge": 1, "mean_delta_merge": 0.0036, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.62636, "trunc_abs_frac": 0.50272, "mean_delta_full": 0.08429, "read_hit_rate": {"surface": 0.71631, "split": 0.88256, "merge": 0.19517, "skel": 0.73719}, "mean_gate": {"surface": 0.32127, "split": 0.23253, "merge": 0.27244, "skel": 0.28592}, "n_reads": 18995, "n_cand": 28117, "read_suppressed_frac": 0.0994, "dropout_reads_logged": 18995, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2139, "admitted_surface": 16924, "staged_surface": 0, "admitted_split": 20125, "staged_split": 0, "admitted_merge": 16181, "staged_merge": 0, "admitted_skel": 12306, "staged_skel": 0, "mean_count": 482.6299133300781, "n_util_obs": 259, "util_count_spearman": 0.038905657510308674}, "probe_overhead_sec_total": 162.596, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.007153, "alpha": 0.03, "t": 1354.0} |
| {"event": "consolidate", "step": 1800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1386.8} |
| {"event": "probe", "step": 1800, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 2923.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.16796, "cps": 22.835938, "mean_p": 0.181856, "boundary_entropy": 0.086316, "bpb": 2.123624}, "tr": {"chunks": 3573.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.197371, "cps": 27.914062, "mean_p": 0.210388, "boundary_entropy": 0.180978, "bpb": 2.359191}, "fi": {"chunks": 3003.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.159641, "cps": 23.460938, "mean_p": 0.183035, "boundary_entropy": 0.166714, "bpb": 2.557959}}, "bpb": {"en": 2.12362, "tr": 2.35919, "fi": 2.55796}, "gini_cpc": 0.047914, "gini_cps": 0.045619, "bpic": 5.972418, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.1751079640470656, "fi": 0.950470744028782}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2189, "admitted_surface": 16911, "staged_surface": 0, "admitted_split": 20153, "staged_split": 0, "admitted_merge": 16195, "staged_merge": 0, "admitted_skel": 12277, "staged_skel": 0, "mean_count": 496.33697509765625, "n_util_obs": 264, "util_count_spearman": 0.00981484259337764}, "t": 1388.0} |
| {"event": "eval", "step": 1800, "phase": 1, "bpb": {"en": 2.13586, "tr": 2.36427, "fi": 2.56918}, "bytes_per_chunk": {"en": 5.427, "tr": 5.312, "fi": 6.05}, "bpb_mean": 2.35644, "frozen_inventory": true, "t": 1390.1} |
| {"event": "train", "step": 1800, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.50365, "bpb": 2.16931, "chunk_loss": 0.030109, "F": 0.17892, "G": 0.19814, "chunks": 183.22, "bytes_per_chunk": 5.589, "boundary_entropy": 0.11482, "router_noise": 0.0, "grad_norm": 0.834, "lr": 0.0002564, "router_trainable": true, "sec_per_step": 0.751, "probe_sec": 0.0936, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07053, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.20503, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.37878, "n_merge": 2, "mean_delta_merge": 0.02161, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.61672, "trunc_abs_frac": 0.61672, "mean_delta_full": -0.57678, "read_hit_rate": {"surface": 0.70691, "split": 0.87523, "merge": 0.20417, "skel": 0.71952}, "mean_gate": {"surface": 0.32425, "split": 0.23137, "merge": 0.27041, "skel": 0.28453}, "n_reads": 18089, "n_cand": 27070, "read_suppressed_frac": 0.1023, "dropout_reads_logged": 18089, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2190, "admitted_surface": 16932, "staged_surface": 0, "admitted_split": 20239, "staged_split": 0, "admitted_merge": 16047, "staged_merge": 0, "admitted_skel": 12318, "staged_skel": 0, "mean_count": 496.6181640625, "n_util_obs": 264, "util_count_spearman": 0.009949194216358289}, "probe_overhead_sec_total": 167.205, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.00727, "alpha": 0.03, "t": 1391.6} |
| {"event": "consolidate", "step": 1850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1419.4} |
| {"event": "train", "step": 1850, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.6104, "bpb": 2.32331, "chunk_loss": 0.030119, "F": 0.17999, "G": 0.19929, "chunks": 184.31, "bytes_per_chunk": 5.556, "boundary_entropy": 0.1224, "router_noise": 0.0, "grad_norm": 0.7909, "lr": 0.0002537, "router_trainable": true, "sec_per_step": 0.574, "probe_sec": 0.062, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04439, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": -0.08952, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.20929, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 1.38166, "trunc_abs_frac": 1.45974, "mean_delta_full": -0.19702, "read_hit_rate": {"surface": 0.66076, "split": 0.84554, "merge": 0.18838, "skel": 0.66318}, "mean_gate": {"surface": 0.31626, "split": 0.22911, "merge": 0.26124, "skel": 0.2775}, "n_reads": 16886, "n_cand": 26755, "read_suppressed_frac": 0.0965, "dropout_reads_logged": 16886, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2237, "admitted_surface": 16995, "staged_surface": 0, "admitted_split": 20290, "staged_split": 0, "admitted_merge": 15883, "staged_merge": 0, "admitted_skel": 12368, "staged_skel": 0, "mean_count": 510.4895935058594, "n_util_obs": 269, "util_count_spearman": 0.01783061699884091}, "probe_overhead_sec_total": 170.896, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.00779, "alpha": 0.03, "t": 1420.3} |
| {"event": "consolidate", "step": 1900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1457.9} |
| {"event": "train", "step": 1900, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.52232, "bpb": 2.19624, "chunk_loss": 0.030156, "F": 0.17841, "G": 0.19504, "chunks": 182.69, "bytes_per_chunk": 5.605, "boundary_entropy": 0.10246, "router_noise": 0.0, "grad_norm": 0.7869, "lr": 0.0002508, "router_trainable": true, "sec_per_step": 0.761, "probe_sec": 0.0951, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07374, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.0321, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.29651, "n_merge": 2, "mean_delta_merge": -0.5882, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.57096, "trunc_abs_frac": 1.95632, "mean_delta_full": 0.2677, "read_hit_rate": {"surface": 0.72308, "split": 0.87983, "merge": 0.19543, "skel": 0.73449}, "mean_gate": {"surface": 0.32921, "split": 0.22973, "merge": 0.27049, "skel": 0.28933}, "n_reads": 18416, "n_cand": 27258, "read_suppressed_frac": 0.1015, "dropout_reads_logged": 18416, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2269, "admitted_surface": 16889, "staged_surface": 0, "admitted_split": 20302, "staged_split": 0, "admitted_merge": 16029, "staged_merge": 0, "admitted_skel": 12316, "staged_skel": 0, "mean_count": 524.41650390625, "n_util_obs": 278, "util_count_spearman": -0.00514608836403729}, "probe_overhead_sec_total": 175.26, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.007977, "alpha": 0.03, "t": 1458.3} |
| {"event": "consolidate", "step": 1950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1489.5} |
| {"event": "train", "step": 1950, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.45847, "bpb": 2.10413, "chunk_loss": 0.030134, "F": 0.1806, "G": 0.19326, "chunks": 184.94, "bytes_per_chunk": 5.537, "boundary_entropy": 0.08993, "router_noise": 0.0, "grad_norm": 0.7532, "lr": 0.0002479, "router_trainable": true, "sec_per_step": 0.655, "probe_sec": 0.1005, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05873, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.05655, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": -0.05093, "n_merge": 1, "mean_delta_merge": 0.05988, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.14533, "trunc_abs_frac": 0.14533, "mean_delta_full": 0.41199, "read_hit_rate": {"surface": 0.72783, "split": 0.89218, "merge": 0.19747, "skel": 0.74223}, "mean_gate": {"surface": 0.32375, "split": 0.22145, "merge": 0.25702, "skel": 0.28095}, "n_reads": 18793, "n_cand": 27523, "read_suppressed_frac": 0.1016, "dropout_reads_logged": 18793, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2309, "admitted_surface": 16839, "staged_surface": 0, "admitted_split": 20190, "staged_split": 0, "admitted_merge": 16234, "staged_merge": 0, "admitted_skel": 12273, "staged_skel": 0, "mean_count": 538.4931640625, "n_util_obs": 283, "util_count_spearman": 0.001592859373177951}, "probe_overhead_sec_total": 179.14, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.007736, "alpha": 0.03, "t": 1491.1} |
| {"event": "consolidate", "step": 2000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1526.5} |
| {"event": "probe", "step": 2000, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 2910.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.167213, "cps": 22.734375, "mean_p": 0.179342, "boundary_entropy": 0.076842, "bpb": 2.086442}, "tr": {"chunks": 3562.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.196763, "cps": 27.828125, "mean_p": 0.209462, "boundary_entropy": 0.17225, "bpb": 2.305166}, "fi": {"chunks": 3039.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.161554, "cps": 23.742188, "mean_p": 0.185269, "boundary_entropy": 0.169899, "bpb": 2.500618}}, "bpb": {"en": 2.08644, "tr": 2.30517, "fi": 2.50062}, "gini_cpc": 0.044664, "gini_cps": 0.045701, "bpic": 5.964883, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.1767236848024847, "fi": 0.9661619901055919}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2349, "admitted_surface": 16920, "staged_surface": 0, "admitted_split": 20281, "staged_split": 0, "admitted_merge": 16040, "staged_merge": 0, "admitted_skel": 12295, "staged_skel": 0, "mean_count": 552.2633056640625, "n_util_obs": 289, "util_count_spearman": -0.003927783568958534}, "t": 1527.7} |
| {"event": "eval", "step": 2000, "phase": 1, "bpb": {"en": 2.09916, "tr": 2.31209, "fi": 2.51181}, "bytes_per_chunk": {"en": 5.422, "tr": 5.282, "fi": 5.979}, "bpb_mean": 2.30769, "frozen_inventory": true, "t": 1529.7} |
| {"event": "train", "step": 2000, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.4883, "bpb": 2.14717, "chunk_loss": 0.030084, "F": 0.17715, "G": 0.19228, "chunks": 181.41, "bytes_per_chunk": 5.645, "boundary_entropy": 0.09537, "router_noise": 0.0, "grad_norm": 0.7361, "lr": 0.000245, "router_trainable": true, "sec_per_step": 0.807, "probe_sec": 0.0576, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03942, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.01439, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 0, "read_hit_rate": {"surface": 0.70414, "split": 0.88264, "merge": 0.17349, "skel": 0.72444}, "mean_gate": {"surface": 0.33197, "split": 0.22918, "merge": 0.26524, "skel": 0.29004}, "n_reads": 18026, "n_cand": 27053, "read_suppressed_frac": 0.0969, "dropout_reads_logged": 18026, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2349, "admitted_surface": 16912, "staged_surface": 0, "admitted_split": 20308, "staged_split": 0, "admitted_merge": 16017, "staged_merge": 0, "admitted_skel": 12299, "staged_skel": 0, "mean_count": 552.5396118164062, "n_util_obs": 289, "util_count_spearman": -0.004350614485144971}, "probe_overhead_sec_total": 183.356, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.00811, "alpha": 0.03, "t": 1531.4} |
| {"event": "probe", "step": 2048, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 2911.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.16727, "cps": 22.742188, "mean_p": 0.178652, "boundary_entropy": 0.074415, "bpb": 2.074338}, "tr": {"chunks": 3569.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.19715, "cps": 27.882812, "mean_p": 0.207311, "boundary_entropy": 0.172114, "bpb": 2.313407}, "fi": {"chunks": 2989.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.158896, "cps": 23.351562, "mean_p": 0.178363, "boundary_entropy": 0.169963, "bpb": 2.498016}}, "bpb": {"en": 2.07434, "tr": 2.31341, "fi": 2.49802}, "gini_cpc": 0.048732, "gini_cps": 0.046327, "bpic": 5.99134, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.1786311402975527, "fi": 0.9499394992452449}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2393, "admitted_surface": 16875, "staged_surface": 0, "admitted_split": 20338, "staged_split": 0, "admitted_merge": 16036, "staged_merge": 0, "admitted_skel": 12287, "staged_skel": 0, "mean_count": 565.6760864257812, "n_util_obs": 291, "util_count_spearman": 0.0015424961337157993}, "t": 1560.6} |
| {"event": "consolidate", "step": 2050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1566.1} |
| {"event": "train", "step": 2050, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.49596, "bpb": 2.15821, "chunk_loss": 0.030158, "F": 0.18127, "G": 0.19756, "chunks": 185.62, "bytes_per_chunk": 5.516, "boundary_entropy": 0.10666, "router_noise": 0.0, "grad_norm": 0.7482, "lr": 0.0002419, "router_trainable": true, "sec_per_step": 0.712, "probe_sec": 0.1439, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10964, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.19882, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.27881, "n_merge": 2, "mean_delta_merge": 0.09215, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.96998, "trunc_abs_frac": 0.93938, "mean_delta_full": 0.29586, "read_hit_rate": {"surface": 0.70074, "split": 0.87154, "merge": 0.19673, "skel": 0.69912}, "mean_gate": {"surface": 0.33156, "split": 0.22835, "merge": 0.26409, "skel": 0.28696}, "n_reads": 17850, "n_cand": 27105, "read_suppressed_frac": 0.1017, "dropout_reads_logged": 17850, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2396, "admitted_surface": 16898, "staged_surface": 0, "admitted_split": 20362, "staged_split": 0, "admitted_merge": 15932, "staged_merge": 0, "admitted_skel": 12344, "staged_skel": 0, "mean_count": 566.51904296875, "n_util_obs": 291, "util_count_spearman": 0.002814233329599852}, "probe_overhead_sec_total": 187.404, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.008388, "alpha": 0.03, "t": 1567.0} |
| {"event": "consolidate", "step": 2100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1622.8} |
| {"event": "train", "step": 2100, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.54264, "bpb": 2.22556, "chunk_loss": 0.030176, "F": 0.18484, "G": 0.19905, "chunks": 189.28, "bytes_per_chunk": 5.41, "boundary_entropy": 0.10226, "router_noise": 0.0, "grad_norm": 0.7762, "lr": 0.0002388, "router_trainable": true, "sec_per_step": 1.14, "probe_sec": 0.1353, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05079, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.00581, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": 0.0822, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.22774, "trunc_abs_frac": 0.62431, "mean_delta_full": -0.0946, "read_hit_rate": {"surface": 0.70141, "split": 0.87625, "merge": 0.18472, "skel": 0.71491}, "mean_gate": {"surface": 0.3285, "split": 0.22451, "merge": 0.25262, "skel": 0.27957}, "n_reads": 18275, "n_cand": 27630, "read_suppressed_frac": 0.1053, "dropout_reads_logged": 18275, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2438, "admitted_surface": 16905, "staged_surface": 0, "admitted_split": 20367, "staged_split": 0, "admitted_merge": 15997, "staged_merge": 0, "admitted_skel": 12267, "staged_skel": 0, "mean_count": 580.6834716796875, "n_util_obs": 300, "util_count_spearman": 0.02717363526261403}, "probe_overhead_sec_total": 192.679, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.008595, "alpha": 0.03, "t": 1624.0} |
| {"event": "consolidate", "step": 2150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1657.8} |
| {"event": "train", "step": 2150, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.53872, "bpb": 2.21991, "chunk_loss": 0.030217, "F": 0.18805, "G": 0.20325, "chunks": 192.56, "bytes_per_chunk": 5.318, "boundary_entropy": 0.10858, "router_noise": 0.0, "grad_norm": 0.7623, "lr": 0.0002357, "router_trainable": true, "sec_per_step": 0.701, "probe_sec": 0.208, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08766, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.14485, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.06824, "n_merge": 3, "mean_delta_merge": -0.18642, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 1.50948, "trunc_abs_frac": 0.75908, "mean_delta_full": 0.08368, "read_hit_rate": {"surface": 0.70245, "split": 0.8724, "merge": 0.20712, "skel": 0.7033}, "mean_gate": {"surface": 0.34101, "split": 0.23203, "merge": 0.25872, "skel": 0.29593}, "n_reads": 18487, "n_cand": 27935, "read_suppressed_frac": 0.097, "dropout_reads_logged": 18487, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2475, "admitted_surface": 16926, "staged_surface": 0, "admitted_split": 20467, "staged_split": 0, "admitted_merge": 15808, "staged_merge": 0, "admitted_skel": 12335, "staged_skel": 0, "mean_count": 594.62158203125, "n_util_obs": 308, "util_count_spearman": 0.03488332693654811}, "probe_overhead_sec_total": 196.677, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.008555, "alpha": 0.03, "t": 1659.1} |
| {"event": "consolidate", "step": 2200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1692.7} |
| {"event": "probe", "step": 2200, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 2922.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.167902, "cps": 22.828125, "mean_p": 0.178576, "boundary_entropy": 0.070395, "bpb": 2.05287}, "tr": {"chunks": 3582.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.197868, "cps": 27.984375, "mean_p": 0.205715, "boundary_entropy": 0.163601, "bpb": 2.273938}, "fi": {"chunks": 3028.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.16097, "cps": 23.65625, "mean_p": 0.182586, "boundary_entropy": 0.159807, "bpb": 2.449419}}, "bpb": {"en": 2.05287, "tr": 2.27394, "fi": 2.44942}, "gini_cpc": 0.0467, "gini_cps": 0.04616, "bpic": 5.951742, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.1784711055072612, "fi": 0.9587114097359043}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2524, "admitted_surface": 16903, "staged_surface": 0, "admitted_split": 20467, "staged_split": 0, "admitted_merge": 15829, "staged_merge": 0, "admitted_skel": 12337, "staged_skel": 0, "mean_count": 608.33251953125, "n_util_obs": 314, "util_count_spearman": 0.03965402177543761}, "t": 1693.7} |
| {"event": "eval", "step": 2200, "phase": 1, "bpb": {"en": 2.06713, "tr": 2.27951, "fi": 2.46314}, "bytes_per_chunk": {"en": 5.402, "tr": 5.265, "fi": 6.02}, "bpb_mean": 2.26993, "frozen_inventory": true, "t": 1695.9} |
| {"event": "train", "step": 2200, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.47902, "bpb": 2.13377, "chunk_loss": 0.030174, "F": 0.18353, "G": 0.19756, "chunks": 187.94, "bytes_per_chunk": 5.449, "boundary_entropy": 0.09586, "router_noise": 0.0, "grad_norm": 0.7509, "lr": 0.0002325, "router_trainable": true, "sec_per_step": 0.768, "probe_sec": 0.0769, "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": 10, "mean_delta_loo": 0.08266, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.05652, "n_merge": 1, "mean_delta_merge": -0.0461, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.61498, "trunc_abs_frac": 4.92376, "mean_delta_full": -0.15918, "read_hit_rate": {"surface": 0.71615, "split": 0.88302, "merge": 0.19899, "skel": 0.71533}, "mean_gate": {"surface": 0.33714, "split": 0.22551, "merge": 0.26367, "skel": 0.28662}, "n_reads": 18394, "n_cand": 27451, "read_suppressed_frac": 0.0949, "dropout_reads_logged": 18394, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2524, "admitted_surface": 16879, "staged_surface": 0, "admitted_split": 20439, "staged_split": 0, "admitted_merge": 15888, "staged_merge": 0, "admitted_skel": 12330, "staged_skel": 0, "mean_count": 608.6219482421875, "n_util_obs": 314, "util_count_spearman": 0.040389704649691216}, "probe_overhead_sec_total": 200.579, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.007892, "alpha": 0.03, "t": 1697.5} |
| {"event": "consolidate", "step": 2250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1728.2} |
| {"event": "train", "step": 2250, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.49818, "bpb": 2.16142, "chunk_loss": 0.030149, "F": 0.17975, "G": 0.19296, "chunks": 184.06, "bytes_per_chunk": 5.563, "boundary_entropy": 0.09245, "router_noise": 0.0, "grad_norm": 0.7592, "lr": 0.0002293, "router_trainable": true, "sec_per_step": 0.63, "probe_sec": 0.0608, "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": 6, "mean_delta_loo": 0.03839, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 3, "mean_delta_split": 0.13954, "n_merge": 1, "mean_delta_merge": 0.11432, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.22987, "trunc_abs_frac": 0.21995, "mean_delta_full": 0.37329, "read_hit_rate": {"surface": 0.69478, "split": 0.87124, "merge": 0.18314, "skel": 0.70899}, "mean_gate": {"surface": 0.34623, "split": 0.22984, "merge": 0.26423, "skel": 0.29732}, "n_reads": 17767, "n_cand": 27049, "read_suppressed_frac": 0.0996, "dropout_reads_logged": 17767, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2579, "admitted_surface": 16900, "staged_surface": 0, "admitted_split": 20558, "staged_split": 0, "admitted_merge": 15740, "staged_merge": 0, "admitted_skel": 12338, "staged_skel": 0, "mean_count": 622.623291015625, "n_util_obs": 322, "util_count_spearman": 0.04845856304839351}, "probe_overhead_sec_total": 205.255, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.008117, "alpha": 0.03, "t": 1729.0} |
| {"event": "consolidate", "step": 2300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1760.2} |
| {"event": "train", "step": 2300, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.46521, "bpb": 2.11385, "chunk_loss": 0.030117, "F": 0.18005, "G": 0.19338, "chunks": 184.38, "bytes_per_chunk": 5.554, "boundary_entropy": 0.08897, "router_noise": 0.0, "grad_norm": 0.8211, "lr": 0.000226, "router_trainable": true, "sec_per_step": 0.632, "probe_sec": 0.065, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04473, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.0235, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": -0.17563, "n_merge": 1, "mean_delta_merge": -1.14105, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 26.73832, "trunc_abs_frac": 3.12405, "mean_delta_full": -0.00435, "read_hit_rate": {"surface": 0.71268, "split": 0.87838, "merge": 0.19585, "skel": 0.72914}, "mean_gate": {"surface": 0.34199, "split": 0.2214, "merge": 0.2614, "skel": 0.29179}, "n_reads": 18352, "n_cand": 27347, "read_suppressed_frac": 0.0999, "dropout_reads_logged": 18352, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2629, "admitted_surface": 16870, "staged_surface": 0, "admitted_split": 20552, "staged_split": 0, "admitted_merge": 15786, "staged_merge": 0, "admitted_skel": 12328, "staged_skel": 0, "mean_count": 636.5633544921875, "n_util_obs": 332, "util_count_spearman": 0.04845019008015405}, "probe_overhead_sec_total": 209.397, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.008284, "alpha": 0.03, "t": 1760.6} |
| {"event": "consolidate", "step": 2350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1791.4} |
| {"event": "train", "step": 2350, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.49022, "bpb": 2.14994, "chunk_loss": 0.030206, "F": 0.18533, "G": 0.20008, "chunks": 189.78, "bytes_per_chunk": 5.396, "boundary_entropy": 0.09602, "router_noise": 0.0, "grad_norm": 0.6814, "lr": 0.0002226, "router_trainable": true, "sec_per_step": 0.641, "probe_sec": 0.0774, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03994, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.00153, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.44556, "n_merge": 1, "mean_delta_merge": 0.27025, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.39387, "trunc_abs_frac": 0.81321, "mean_delta_full": -0.40487, "read_hit_rate": {"surface": 0.7166, "split": 0.87518, "merge": 0.21817, "skel": 0.7197}, "mean_gate": {"surface": 0.34742, "split": 0.22511, "merge": 0.26486, "skel": 0.29673}, "n_reads": 18678, "n_cand": 27786, "read_suppressed_frac": 0.1028, "dropout_reads_logged": 18678, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2672, "admitted_surface": 16883, "staged_surface": 0, "admitted_split": 20601, "staged_split": 0, "admitted_merge": 15707, "staged_merge": 0, "admitted_skel": 12345, "staged_skel": 0, "mean_count": 650.5780029296875, "n_util_obs": 339, "util_count_spearman": 0.0749990046942242}, "probe_overhead_sec_total": 213.306, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.008232, "alpha": 0.03, "t": 1792.7} |
| {"event": "consolidate", "step": 2400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1830.3} |
| {"event": "probe", "step": 2400, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 2922.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.167902, "cps": 22.828125, "mean_p": 0.178728, "boundary_entropy": 0.069823, "bpb": 2.024259}, "tr": {"chunks": 3577.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.197592, "cps": 27.945312, "mean_p": 0.205302, "boundary_entropy": 0.160539, "bpb": 2.257387}, "fi": {"chunks": 2997.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.159322, "cps": 23.414062, "mean_p": 0.176829, "boundary_entropy": 0.15508, "bpb": 2.426895}}, "bpb": {"en": 2.02426, "tr": 2.25739, "fi": 2.42689}, "gini_cpc": 0.048614, "gini_cps": 0.045984, "bpic": 5.974305, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.1768261151310644, "fi": 0.948896332555649}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2721, "admitted_surface": 16854, "staged_surface": 0, "admitted_split": 20533, "staged_split": 0, "admitted_merge": 15877, "staged_merge": 0, "admitted_skel": 12272, "staged_skel": 0, "mean_count": 664.1997680664062, "n_util_obs": 346, "util_count_spearman": 0.07046259515104066}, "t": 1831.7} |
| {"event": "eval", "step": 2400, "phase": 1, "bpb": {"en": 2.03993, "tr": 2.26423, "fi": 2.43726}, "bytes_per_chunk": {"en": 5.424, "tr": 5.275, "fi": 6.082}, "bpb_mean": 2.24714, "frozen_inventory": true, "t": 1834.0} |
| {"event": "phase_transition", "step": 2400, "phase": 2, "objective": "global", "router_trainable": true, "encoder_trainable": true, "frozen_router_params": 0, "t": 1835.5} |
| {"event": "train", "step": 2400, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.47954, "bpb": 2.13452, "chunk_loss": 0.030144, "F": 0.18198, "G": 0.19594, "chunks": 186.34, "bytes_per_chunk": 5.495, "boundary_entropy": 0.09626, "router_noise": 0.0, "grad_norm": 0.7394, "lr": 0.0002192, "router_trainable": true, "sec_per_step": 0.882, "probe_sec": 0.1261, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06876, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.08602, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.10681, "n_merge": 2, "mean_delta_merge": 0.257, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.13743, "trunc_abs_frac": 1.30189, "mean_delta_full": -0.21484, "read_hit_rate": {"surface": 0.70545, "split": 0.87459, "merge": 0.2002, "skel": 0.7207}, "mean_gate": {"surface": 0.35349, "split": 0.22757, "merge": 0.26778, "skel": 0.2985}, "n_reads": 18275, "n_cand": 27414, "read_suppressed_frac": 0.1012, "dropout_reads_logged": 18275, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2721, "admitted_surface": 16897, "staged_surface": 0, "admitted_split": 20574, "staged_split": 0, "admitted_merge": 15738, "staged_merge": 0, "admitted_skel": 12327, "staged_skel": 0, "mean_count": 664.4826049804688, "n_util_obs": 346, "util_count_spearman": 0.07107851924278394}, "probe_overhead_sec_total": 217.605, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.009103, "alpha": 0.03, "t": 1836.8} |
| {"event": "consolidate", "step": 2450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1867.8} |
| {"event": "train", "step": 2450, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.49151, "bpb": 2.1518, "chunk_loss": 0.030284, "F": 0.19232, "G": 0.20496, "chunks": 196.94, "bytes_per_chunk": 5.2, "boundary_entropy": 0.09077, "router_noise": 0.0, "grad_norm": 0.7293, "lr": 0.0002158, "router_trainable": true, "sec_per_step": 0.638, "probe_sec": 0.0651, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04266, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.09429, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": 0.08548, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.72045, "trunc_abs_frac": 1.00394, "mean_delta_full": 0.21646, "read_hit_rate": {"surface": 0.72775, "split": 0.88813, "merge": 0.20487, "skel": 0.73043}, "mean_gate": {"surface": 0.34646, "split": 0.22305, "merge": 0.25113, "skel": 0.29025}, "n_reads": 19344, "n_cand": 28549, "read_suppressed_frac": 0.1006, "dropout_reads_logged": 19344, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2761, "admitted_surface": 16814, "staged_surface": 0, "admitted_split": 20547, "staged_split": 0, "admitted_merge": 15917, "staged_merge": 0, "admitted_skel": 12258, "staged_skel": 0, "mean_count": 678.6156005859375, "n_util_obs": 349, "util_count_spearman": 0.06523650475482444}, "probe_overhead_sec_total": 221.565, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.008088, "alpha": 0.03, "t": 1868.7} |
| {"event": "consolidate", "step": 2500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1899.5} |
| {"event": "train", "step": 2500, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.42331, "bpb": 2.0534, "chunk_loss": 0.030161, "F": 0.18228, "G": 0.19674, "chunks": 186.66, "bytes_per_chunk": 5.486, "boundary_entropy": 0.09, "router_noise": 0.0, "grad_norm": 0.7134, "lr": 0.0002123, "router_trainable": true, "sec_per_step": 0.633, "probe_sec": 0.0597, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03941, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": 0.12866, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.06934, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 1.42589, "trunc_abs_frac": 1.06746, "mean_delta_full": 0.02924, "read_hit_rate": {"surface": 0.72311, "split": 0.88384, "merge": 0.20003, "skel": 0.73746}, "mean_gate": {"surface": 0.34047, "split": 0.21329, "merge": 0.24497, "skel": 0.28264}, "n_reads": 18668, "n_cand": 27569, "read_suppressed_frac": 0.1011, "dropout_reads_logged": 18668, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2796, "admitted_surface": 16846, "staged_surface": 0, "admitted_split": 20595, "staged_split": 0, "admitted_merge": 15777, "staged_merge": 0, "admitted_skel": 12318, "staged_skel": 0, "mean_count": 692.5703735351562, "n_util_obs": 352, "util_count_spearman": 0.06728431450101287}, "probe_overhead_sec_total": 225.498, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.009685, "alpha": 0.03, "t": 1900.3} |
| {"event": "consolidate", "step": 2550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1928.1} |
| {"event": "train", "step": 2550, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.4139, "bpb": 2.03982, "chunk_loss": 0.030238, "F": 0.19058, "G": 0.2026, "chunks": 195.16, "bytes_per_chunk": 5.247, "boundary_entropy": 0.08306, "router_noise": 0.0, "grad_norm": 0.675, "lr": 0.0002088, "router_trainable": true, "sec_per_step": 0.574, "probe_sec": 0.0639, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03929, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.08738, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.03931, "n_merge": 1, "mean_delta_merge": -0.02994, "n_dropped_unpaired": 0, "n_full": 0, "read_hit_rate": {"surface": 0.75133, "split": 0.89935, "merge": 0.23103, "skel": 0.75976}, "mean_gate": {"surface": 0.36159, "split": 0.22843, "merge": 0.25913, "skel": 0.30566}, "n_reads": 19968, "n_cand": 28566, "read_suppressed_frac": 0.0974, "dropout_reads_logged": 19968, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2834, "admitted_surface": 16841, "staged_surface": 0, "admitted_split": 20555, "staged_split": 0, "admitted_merge": 15820, "staged_merge": 0, "admitted_skel": 12320, "staged_skel": 0, "mean_count": 706.4804077148438, "n_util_obs": 361, "util_count_spearman": 0.07059708629798934}, "probe_overhead_sec_total": 229.262, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.009207, "alpha": 0.03, "t": 1929.1} |
| {"event": "consolidate", "step": 2600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1969.0} |
| {"event": "probe", "step": 2600, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 2923.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.16796, "cps": 22.835938, "mean_p": 0.178144, "boundary_entropy": 0.066163, "bpb": 2.007317}, "tr": {"chunks": 3574.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.197426, "cps": 27.921875, "mean_p": 0.204199, "boundary_entropy": 0.155022, "bpb": 2.214514}, "fi": {"chunks": 2984.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.158631, "cps": 23.3125, "mean_p": 0.176322, "boundary_entropy": 0.148664, "bpb": 2.397237}}, "bpb": {"en": 2.00732, "tr": 2.21451, "fi": 2.39724}, "gini_cpc": 0.049356, "gini_cps": 0.045776, "bpic": 5.983757, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.17543684956737, "fi": 0.9444571096176775}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2870, "admitted_surface": 16899, "staged_surface": 0, "admitted_split": 20626, "staged_split": 0, "admitted_merge": 15632, "staged_merge": 0, "admitted_skel": 12379, "staged_skel": 0, "mean_count": 720.0575561523438, "n_util_obs": 368, "util_count_spearman": 0.07513643693819987}, "t": 1970.3} |
| {"event": "eval", "step": 2600, "phase": 2, "bpb": {"en": 2.02382, "tr": 2.22271, "fi": 2.40826}, "bytes_per_chunk": {"en": 5.413, "tr": 5.265, "fi": 6.105}, "bpb_mean": 2.21826, "frozen_inventory": true, "t": 1972.7} |
| {"event": "train", "step": 2600, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.50357, "bpb": 2.1692, "chunk_loss": 0.030127, "F": 0.18063, "G": 0.1936, "chunks": 184.97, "bytes_per_chunk": 5.536, "boundary_entropy": 0.08633, "router_noise": 0.0, "grad_norm": 0.6885, "lr": 0.0002053, "router_trainable": true, "sec_per_step": 0.906, "probe_sec": 0.0729, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04458, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": 0.05005, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.32568, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 10.06364, "trunc_abs_frac": 0.79109, "mean_delta_full": 0.00895, "read_hit_rate": {"surface": 0.70715, "split": 0.87113, "merge": 0.18309, "skel": 0.71878}, "mean_gate": {"surface": 0.347, "split": 0.21787, "merge": 0.25014, "skel": 0.29384}, "n_reads": 17978, "n_cand": 27181, "read_suppressed_frac": 0.0982, "dropout_reads_logged": 17978, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2870, "admitted_surface": 16875, "staged_surface": 0, "admitted_split": 20653, "staged_split": 0, "admitted_merge": 15666, "staged_merge": 0, "admitted_skel": 12342, "staged_skel": 0, "mean_count": 720.3295288085938, "n_util_obs": 368, "util_count_spearman": 0.07571362310779314}, "probe_overhead_sec_total": 234.034, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.009134, "alpha": 0.03, "t": 1974.3} |
| {"event": "consolidate", "step": 2650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2007.4} |
| {"event": "train", "step": 2650, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.5081, "bpb": 2.17573, "chunk_loss": 0.030127, "F": 0.17764, "G": 0.19055, "chunks": 181.91, "bytes_per_chunk": 5.629, "boundary_entropy": 0.08282, "router_noise": 0.0, "grad_norm": 0.8091, "lr": 0.0002017, "router_trainable": true, "sec_per_step": 0.672, "probe_sec": 0.0771, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05738, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": 0.07011, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.3844, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.57959, "trunc_abs_frac": 1.06401, "mean_delta_full": 0.07208, "read_hit_rate": {"surface": 0.7015, "split": 0.86644, "merge": 0.18749, "skel": 0.71493}, "mean_gate": {"surface": 0.35672, "split": 0.21991, "merge": 0.25277, "skel": 0.30182}, "n_reads": 17695, "n_cand": 26852, "read_suppressed_frac": 0.1006, "dropout_reads_logged": 17695, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2907, "admitted_surface": 16875, "staged_surface": 0, "admitted_split": 20709, "staged_split": 0, "admitted_merge": 15591, "staged_merge": 0, "admitted_skel": 12361, "staged_skel": 0, "mean_count": 734.3468017578125, "n_util_obs": 370, "util_count_spearman": 0.07603353382047888}, "probe_overhead_sec_total": 238.735, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.009306, "alpha": 0.03, "t": 2007.9} |
| {"event": "consolidate", "step": 2700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2037.7} |
| {"event": "train", "step": 2700, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.42743, "bpb": 2.05934, "chunk_loss": 0.030106, "F": 0.17932, "G": 0.19029, "chunks": 183.62, "bytes_per_chunk": 5.577, "boundary_entropy": 0.07907, "router_noise": 0.0, "grad_norm": 0.7536, "lr": 0.0001981, "router_trainable": true, "sec_per_step": 0.605, "probe_sec": 0.0717, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04727, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.02673, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": 0.02475, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.50264, "trunc_abs_frac": 17.2536, "mean_delta_full": -0.28601, "read_hit_rate": {"surface": 0.71526, "split": 0.88231, "merge": 0.18944, "skel": 0.73024}, "mean_gate": {"surface": 0.3592, "split": 0.21975, "merge": 0.25453, "skel": 0.30077}, "n_reads": 18040, "n_cand": 26913, "read_suppressed_frac": 0.0989, "dropout_reads_logged": 18040, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2956, "admitted_surface": 16879, "staged_surface": 0, "admitted_split": 20673, "staged_split": 0, "admitted_merge": 15636, "staged_merge": 0, "admitted_skel": 12348, "staged_skel": 0, "mean_count": 748.360595703125, "n_util_obs": 375, "util_count_spearman": 0.0866039367391057}, "probe_overhead_sec_total": 242.547, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.00871, "alpha": 0.03, "t": 2038.2} |
| {"event": "consolidate", "step": 2750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2069.3} |
| {"event": "train", "step": 2750, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.46404, "bpb": 2.11216, "chunk_loss": 0.030195, "F": 0.18704, "G": 0.19935, "chunks": 191.53, "bytes_per_chunk": 5.346, "boundary_entropy": 0.09074, "router_noise": 0.0, "grad_norm": 0.7529, "lr": 0.0001945, "router_trainable": true, "sec_per_step": 0.64, "probe_sec": 0.0685, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04745, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": 0.38758, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.12521, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.394, "trunc_abs_frac": 0.75436, "mean_delta_full": -0.23348, "read_hit_rate": {"surface": 0.70707, "split": 0.87892, "merge": 0.19901, "skel": 0.71603}, "mean_gate": {"surface": 0.35722, "split": 0.22199, "merge": 0.25409, "skel": 0.3002}, "n_reads": 18538, "n_cand": 27824, "read_suppressed_frac": 0.0992, "dropout_reads_logged": 18538, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2992, "admitted_surface": 16872, "staged_surface": 0, "admitted_split": 20696, "staged_split": 0, "admitted_merge": 15661, "staged_merge": 0, "admitted_skel": 12307, "staged_skel": 0, "mean_count": 762.3258056640625, "n_util_obs": 378, "util_count_spearman": 0.07275295091065749}, "probe_overhead_sec_total": 246.296, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.00966, "alpha": 0.03, "t": 2070.2} |
| {"event": "consolidate", "step": 2800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2110.4} |
| {"event": "probe", "step": 2800, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 2924.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.168017, "cps": 22.84375, "mean_p": 0.176984, "boundary_entropy": 0.059936, "bpb": 1.991304}, "tr": {"chunks": 3574.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.197426, "cps": 27.921875, "mean_p": 0.202249, "boundary_entropy": 0.140467, "bpb": 2.206418}, "fi": {"chunks": 2982.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.158524, "cps": 23.296875, "mean_p": 0.171105, "boundary_entropy": 0.130183, "bpb": 2.375427}}, "bpb": {"en": 1.9913, "tr": 2.20642, "fi": 2.37543}, "gini_cpc": 0.049496, "gini_cps": 0.04571, "bpic": 5.984388, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.1750348533807873, "fi": 0.9435013102107719}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3037, "admitted_surface": 16832, "staged_surface": 0, "admitted_split": 20598, "staged_split": 0, "admitted_merge": 15832, "staged_merge": 0, "admitted_skel": 12274, "staged_skel": 0, "mean_count": 776.2535400390625, "n_util_obs": 383, "util_count_spearman": 0.08171166519741091}, "t": 2111.9} |
| {"event": "eval", "step": 2800, "phase": 2, "bpb": {"en": 2.00801, "tr": 2.20863, "fi": 2.38607}, "bytes_per_chunk": {"en": 5.402, "tr": 5.272, "fi": 6.111}, "bpb_mean": 2.2009, "frozen_inventory": true, "t": 2114.5} |
| {"event": "train", "step": 2800, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.4224, "bpb": 2.05209, "chunk_loss": 0.030151, "F": 0.1828, "G": 0.19481, "chunks": 187.19, "bytes_per_chunk": 5.47, "boundary_entropy": 0.07906, "router_noise": 0.0, "grad_norm": 0.749, "lr": 0.0001909, "router_trainable": true, "sec_per_step": 0.926, "probe_sec": 0.0639, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04101, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.06682, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": 0.00702, "n_merge": 1, "mean_delta_merge": -0.0285, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -1.4837, "trunc_abs_frac": 1.47748, "mean_delta_full": 0.07581, "read_hit_rate": {"surface": 0.72524, "split": 0.8885, "merge": 0.20621, "skel": 0.73674}, "mean_gate": {"surface": 0.35489, "split": 0.21648, "merge": 0.24591, "skel": 0.29563}, "n_reads": 18779, "n_cand": 27583, "read_suppressed_frac": 0.0988, "dropout_reads_logged": 18779, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3038, "admitted_surface": 16845, "staged_surface": 0, "admitted_split": 20743, "staged_split": 0, "admitted_merge": 15606, "staged_merge": 0, "admitted_skel": 12342, "staged_skel": 0, "mean_count": 776.5372924804688, "n_util_obs": 383, "util_count_spearman": 0.0812847987098103}, "probe_overhead_sec_total": 250.809, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.010344, "alpha": 0.03, "t": 2116.5} |
| {"event": "consolidate", "step": 2850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2149.3} |
| {"event": "train", "step": 2850, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.43325, "bpb": 2.06774, "chunk_loss": 0.030227, "F": 0.18655, "G": 0.19819, "chunks": 191.03, "bytes_per_chunk": 5.36, "boundary_entropy": 0.08628, "router_noise": 0.0, "grad_norm": 0.7513, "lr": 0.0001872, "router_trainable": true, "sec_per_step": 0.668, "probe_sec": 0.1283, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.1016, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.03776, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.77466, "n_merge": 1, "mean_delta_merge": 0.13025, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -0.11704, "trunc_abs_frac": 0.97171, "mean_delta_full": -0.18265, "read_hit_rate": {"surface": 0.71386, "split": 0.88453, "merge": 0.21938, "skel": 0.71621}, "mean_gate": {"surface": 0.36015, "split": 0.21996, "merge": 0.25074, "skel": 0.30111}, "n_reads": 18731, "n_cand": 27774, "read_suppressed_frac": 0.0988, "dropout_reads_logged": 18731, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3065, "admitted_surface": 16893, "staged_surface": 0, "admitted_split": 20781, "staged_split": 0, "admitted_merge": 15483, "staged_merge": 0, "admitted_skel": 12379, "staged_skel": 0, "mean_count": 790.51513671875, "n_util_obs": 392, "util_count_spearman": 0.07982256958541797}, "probe_overhead_sec_total": 255.062, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.008949, "alpha": 0.03, "t": 2149.9} |
| {"event": "consolidate", "step": 2900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2181.2} |
| {"event": "train", "step": 2900, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.33908, "bpb": 1.93188, "chunk_loss": 0.030159, "F": 0.1861, "G": 0.19489, "chunks": 190.56, "bytes_per_chunk": 5.374, "boundary_entropy": 0.06596, "router_noise": 0.0, "grad_norm": 0.6567, "lr": 0.0001835, "router_trainable": true, "sec_per_step": 0.636, "probe_sec": 0.0728, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04581, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": 0.09858, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.14996, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -0.02684, "trunc_abs_frac": 0.77423, "mean_delta_full": 0.24028, "read_hit_rate": {"surface": 0.77943, "split": 0.91386, "merge": 0.22108, "skel": 0.79968}, "mean_gate": {"surface": 0.3634, "split": 0.21298, "merge": 0.24609, "skel": 0.3019}, "n_reads": 20460, "n_cand": 28487, "read_suppressed_frac": 0.1, "dropout_reads_logged": 20460, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3108, "admitted_surface": 16685, "staged_surface": 0, "admitted_split": 20689, "staged_split": 0, "admitted_merge": 15911, "staged_merge": 0, "admitted_skel": 12251, "staged_skel": 0, "mean_count": 804.6676025390625, "n_util_obs": 397, "util_count_spearman": 0.0794919026613443}, "probe_overhead_sec_total": 258.793, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.009185, "alpha": 0.03, "t": 2181.7} |
| {"event": "consolidate", "step": 2950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2218.9} |
| {"event": "train", "step": 2950, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.49813, "bpb": 2.16134, "chunk_loss": 0.030171, "F": 0.18561, "G": 0.19609, "chunks": 190.06, "bytes_per_chunk": 5.388, "boundary_entropy": 0.07841, "router_noise": 0.0, "grad_norm": 0.7471, "lr": 0.0001799, "router_trainable": true, "sec_per_step": 0.769, "probe_sec": 0.1435, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08059, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.14004, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": 0.05582, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 1.4186, "trunc_abs_frac": 5.47157, "mean_delta_full": 0.12422, "read_hit_rate": {"surface": 0.69967, "split": 0.88302, "merge": 0.18744, "skel": 0.71531}, "mean_gate": {"surface": 0.35438, "split": 0.21189, "merge": 0.24606, "skel": 0.29364}, "n_reads": 18581, "n_cand": 27940, "read_suppressed_frac": 0.1011, "dropout_reads_logged": 18581, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3145, "admitted_surface": 16861, "staged_surface": 0, "admitted_split": 20814, "staged_split": 0, "admitted_merge": 15539, "staged_merge": 0, "admitted_skel": 12322, "staged_skel": 0, "mean_count": 818.656982421875, "n_util_obs": 407, "util_count_spearman": 0.07032395893544745}, "probe_overhead_sec_total": 263.119, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.008792, "alpha": 0.03, "t": 2220.1} |
| {"event": "consolidate", "step": 3000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2259.8} |
| {"event": "probe", "step": 3000, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 2922.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.167902, "cps": 22.828125, "mean_p": 0.176577, "boundary_entropy": 0.056993, "bpb": 1.969582}, "tr": {"chunks": 3478.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.192123, "cps": 27.171875, "mean_p": 0.195902, "boundary_entropy": 0.131263, "bpb": 2.185174}, "fi": {"chunks": 2987.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.15879, "cps": 23.335938, "mean_p": 0.175677, "boundary_entropy": 0.138464, "bpb": 2.342616}}, "bpb": {"en": 1.96958, "tr": 2.18517, "fi": 2.34262}, "gini_cpc": 0.042832, "gini_cps": 0.039487, "bpic": 6.043677, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.1442553056823712, "fi": 0.9457301786265344}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3185, "admitted_surface": 16899, "staged_surface": 0, "admitted_split": 20873, "staged_split": 0, "admitted_merge": 15438, "staged_merge": 0, "admitted_skel": 12326, "staged_skel": 0, "mean_count": 832.4312133789062, "n_util_obs": 411, "util_count_spearman": 0.09426897035900592}, "t": 2261.1} |
| {"event": "eval", "step": 3000, "phase": 2, "bpb": {"en": 1.98668, "tr": 2.18949, "fi": 2.35394}, "bytes_per_chunk": {"en": 5.424, "tr": 5.371, "fi": 6.096}, "bpb_mean": 2.1767, "frozen_inventory": true, "t": 2263.3} |
| {"event": "train", "step": 3000, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.41368, "bpb": 2.03951, "chunk_loss": 0.030166, "F": 0.18542, "G": 0.19652, "chunks": 189.88, "bytes_per_chunk": 5.393, "boundary_entropy": 0.07076, "router_noise": 0.0, "grad_norm": 0.7224, "lr": 0.0001761, "router_trainable": true, "sec_per_step": 0.907, "probe_sec": 0.0849, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05055, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.03838, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.3396, "n_merge": 1, "mean_delta_merge": -0.10223, "n_dropped_unpaired": 0, "n_full": 0, "read_hit_rate": {"surface": 0.72915, "split": 0.89241, "merge": 0.21257, "skel": 0.75176}, "mean_gate": {"surface": 0.35562, "split": 0.2082, "merge": 0.2352, "skel": 0.29692}, "n_reads": 19357, "n_cand": 28161, "read_suppressed_frac": 0.101, "dropout_reads_logged": 19357, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3185, "admitted_surface": 16827, "staged_surface": 0, "admitted_split": 20769, "staged_split": 0, "admitted_merge": 15611, "staged_merge": 0, "admitted_skel": 12329, "staged_skel": 0, "mean_count": 832.7359619140625, "n_util_obs": 411, "util_count_spearman": 0.09402101517502422}, "probe_overhead_sec_total": 267.815, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.010089, "alpha": 0.03, "t": 2265.5} |
| {"event": "consolidate", "step": 3050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2299.1} |
| {"event": "train", "step": 3050, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.38987, "bpb": 2.00516, "chunk_loss": 0.030104, "F": 0.17847, "G": 0.18876, "chunks": 182.75, "bytes_per_chunk": 5.603, "boundary_entropy": 0.07346, "router_noise": 0.0, "grad_norm": 0.7058, "lr": 0.0001724, "router_trainable": true, "sec_per_step": 0.683, "probe_sec": 0.0898, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06256, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.21681, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.02942, "n_merge": 3, "mean_delta_merge": -0.11959, "n_dropped_unpaired": 0, "n_full": 0, "read_hit_rate": {"surface": 0.71991, "split": 0.87742, "merge": 0.19104, "skel": 0.721}, "mean_gate": {"surface": 0.36765, "split": 0.21541, "merge": 0.25332, "skel": 0.30039}, "n_reads": 17925, "n_cand": 26817, "read_suppressed_frac": 0.0991, "dropout_reads_logged": 17925, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3226, "admitted_surface": 16823, "staged_surface": 0, "admitted_split": 20943, "staged_split": 0, "admitted_merge": 15398, "staged_merge": 0, "admitted_skel": 12372, "staged_skel": 0, "mean_count": 846.7045288085938, "n_util_obs": 413, "util_count_spearman": 0.07895287554991132}, "probe_overhead_sec_total": 271.922, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.009303, "alpha": 0.03, "t": 2299.7} |
| {"event": "consolidate", "step": 3100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2347.8} |
| {"event": "train", "step": 3100, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.39753, "bpb": 2.01621, "chunk_loss": 0.030162, "F": 0.18387, "G": 0.19237, "chunks": 188.28, "bytes_per_chunk": 5.439, "boundary_entropy": 0.06104, "router_noise": 0.0, "grad_norm": 0.7795, "lr": 0.0001687, "router_trainable": true, "sec_per_step": 0.98, "probe_sec": 0.094, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06634, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": 0.01202, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.15637, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.27373, "trunc_abs_frac": 1.23595, "mean_delta_full": 0.29059, "read_hit_rate": {"surface": 0.74487, "split": 0.89743, "merge": 0.21339, "skel": 0.75086}, "mean_gate": {"surface": 0.36727, "split": 0.21014, "merge": 0.24481, "skel": 0.30562}, "n_reads": 19350, "n_cand": 27931, "read_suppressed_frac": 0.0977, "dropout_reads_logged": 19350, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3277, "admitted_surface": 16802, "staged_surface": 0, "admitted_split": 20883, "staged_split": 0, "admitted_merge": 15488, "staged_merge": 0, "admitted_skel": 12363, "staged_skel": 0, "mean_count": 860.840087890625, "n_util_obs": 421, "util_count_spearman": 0.06981885626393527}, "probe_overhead_sec_total": 277.046, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.009774, "alpha": 0.03, "t": 2348.7} |
| {"event": "consolidate", "step": 3150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2386.2} |
| {"event": "train", "step": 3150, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.32812, "bpb": 1.91607, "chunk_loss": 0.030124, "F": 0.18085, "G": 0.18904, "chunks": 185.19, "bytes_per_chunk": 5.53, "boundary_entropy": 0.06051, "router_noise": 0.0, "grad_norm": 0.6803, "lr": 0.000165, "router_trainable": true, "sec_per_step": 0.777, "probe_sec": 0.0909, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04889, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.04459, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 3, "mean_delta_split": 0.11067, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.8381, "trunc_abs_frac": 4.30711, "mean_delta_full": 0.02092, "read_hit_rate": {"surface": 0.75484, "split": 0.9031, "merge": 0.21409, "skel": 0.76176}, "mean_gate": {"surface": 0.36844, "split": 0.21149, "merge": 0.24665, "skel": 0.30296}, "n_reads": 19313, "n_cand": 27584, "read_suppressed_frac": 0.0986, "dropout_reads_logged": 19313, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3314, "admitted_surface": 16768, "staged_surface": 0, "admitted_split": 20834, "staged_split": 0, "admitted_merge": 15618, "staged_merge": 0, "admitted_skel": 12316, "staged_skel": 0, "mean_count": 874.7879638671875, "n_util_obs": 429, "util_count_spearman": 0.06296191789880888}, "probe_overhead_sec_total": 281.416, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.009935, "alpha": 0.03, "t": 2387.5} |
| {"event": "consolidate", "step": 3200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2423.0} |
| {"event": "probe", "step": 3200, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 2922.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.167902, "cps": 22.828125, "mean_p": 0.17516, "boundary_entropy": 0.050192, "bpb": 1.956626}, "tr": {"chunks": 3412.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.188477, "cps": 26.65625, "mean_p": 0.185086, "boundary_entropy": 0.117891, "bpb": 2.168269}, "fi": {"chunks": 2973.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.158046, "cps": 23.226562, "mean_p": 0.165766, "boundary_entropy": 0.122396, "bpb": 2.315246}}, "bpb": {"en": 1.95663, "tr": 2.16827, "fi": 2.31525}, "gini_cpc": 0.039437, "gini_cps": 0.035099, "bpic": 6.095627, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.122541432716576, "fi": 0.9412975631257738}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3362, "admitted_surface": 16865, "staged_surface": 0, "admitted_split": 20949, "staged_split": 0, "admitted_merge": 15388, "staged_merge": 0, "admitted_skel": 12334, "staged_skel": 0, "mean_count": 888.4251708984375, "n_util_obs": 434, "util_count_spearman": 0.057587357689025}, "t": 2424.2} |
| {"event": "eval", "step": 3200, "phase": 2, "bpb": {"en": 1.97345, "tr": 2.1747, "fi": 2.32725}, "bytes_per_chunk": {"en": 5.419, "tr": 5.471, "fi": 6.107}, "bpb_mean": 2.15847, "frozen_inventory": true, "t": 2426.4} |
| {"event": "train", "step": 3200, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.37688, "bpb": 1.98642, "chunk_loss": 0.030122, "F": 0.18265, "G": 0.19107, "chunks": 187.03, "bytes_per_chunk": 5.475, "boundary_entropy": 0.06761, "router_noise": 0.0, "grad_norm": 0.7102, "lr": 0.0001613, "router_trainable": true, "sec_per_step": 0.807, "probe_sec": 0.0628, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04385, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.11627, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.00024, "n_merge": 1, "mean_delta_merge": 0.19281, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -1.85126, "trunc_abs_frac": 0.51429, "mean_delta_full": -0.06073, "read_hit_rate": {"surface": 0.73425, "split": 0.89275, "merge": 0.20284, "skel": 0.74847}, "mean_gate": {"surface": 0.35427, "split": 0.20592, "merge": 0.23861, "skel": 0.28913}, "n_reads": 18891, "n_cand": 27557, "read_suppressed_frac": 0.0961, "dropout_reads_logged": 18891, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3362, "admitted_surface": 16806, "staged_surface": 0, "admitted_split": 20824, "staged_split": 0, "admitted_merge": 15621, "staged_merge": 0, "admitted_skel": 12285, "staged_skel": 0, "mean_count": 888.722900390625, "n_util_obs": 434, "util_count_spearman": 0.05830731679211001}, "probe_overhead_sec_total": 285.76, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.009162, "alpha": 0.03, "t": 2427.9} |
| {"event": "consolidate", "step": 3250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2459.6} |
| {"event": "train", "step": 3250, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.32441, "bpb": 1.91072, "chunk_loss": 0.030102, "F": 0.17828, "G": 0.18796, "chunks": 182.56, "bytes_per_chunk": 5.609, "boundary_entropy": 0.06956, "router_noise": 0.0, "grad_norm": 0.675, "lr": 0.0001576, "router_trainable": true, "sec_per_step": 0.641, "probe_sec": 0.0548, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03589, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.00442, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.09219, "n_merge": 1, "mean_delta_merge": -0.07156, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -0.39027, "trunc_abs_frac": 1.13393, "mean_delta_full": -0.08695, "read_hit_rate": {"surface": 0.72754, "split": 0.88285, "merge": 0.1999, "skel": 0.73009}, "mean_gate": {"surface": 0.35791, "split": 0.20667, "merge": 0.23412, "skel": 0.29432}, "n_reads": 18281, "n_cand": 26998, "read_suppressed_frac": 0.0978, "dropout_reads_logged": 18281, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3403, "admitted_surface": 16798, "staged_surface": 0, "admitted_split": 20995, "staged_split": 0, "admitted_merge": 15372, "staged_merge": 0, "admitted_skel": 12371, "staged_skel": 0, "mean_count": 902.66552734375, "n_util_obs": 443, "util_count_spearman": 0.05656784942404382}, "probe_overhead_sec_total": 290.002, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.009272, "alpha": 0.03, "t": 2459.9} |
| {"event": "consolidate", "step": 3300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2492.1} |
| {"event": "train", "step": 3300, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.40198, "bpb": 2.02264, "chunk_loss": 0.030118, "F": 0.17987, "G": 0.18989, "chunks": 184.19, "bytes_per_chunk": 5.56, "boundary_entropy": 0.07092, "router_noise": 0.0, "grad_norm": 0.7614, "lr": 0.0001539, "router_trainable": true, "sec_per_step": 0.661, "probe_sec": 0.0607, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04022, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.01642, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.0426, "n_merge": 1, "mean_delta_merge": -0.33746, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.59778, "trunc_abs_frac": 1.92581, "mean_delta_full": -0.16214, "read_hit_rate": {"surface": 0.71494, "split": 0.88187, "merge": 0.19811, "skel": 0.72194}, "mean_gate": {"surface": 0.35634, "split": 0.20488, "merge": 0.23663, "skel": 0.29411}, "n_reads": 18232, "n_cand": 27154, "read_suppressed_frac": 0.1013, "dropout_reads_logged": 18232, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3442, "admitted_surface": 16848, "staged_surface": 0, "admitted_split": 20961, "staged_split": 0, "admitted_merge": 15335, "staged_merge": 0, "admitted_skel": 12392, "staged_skel": 0, "mean_count": 916.5467529296875, "n_util_obs": 452, "util_count_spearman": 0.06624478979376386}, "probe_overhead_sec_total": 293.723, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.010075, "alpha": 0.03, "t": 2493.0} |
| {"event": "consolidate", "step": 3350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2533.3} |
| {"event": "train", "step": 3350, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.56966, "bpb": 2.26454, "chunk_loss": 0.030196, "F": 0.18515, "G": 0.19387, "chunks": 189.59, "bytes_per_chunk": 5.401, "boundary_entropy": 0.07882, "router_noise": 0.0, "grad_norm": 0.7338, "lr": 0.0001501, "router_trainable": true, "sec_per_step": 0.817, "probe_sec": 0.0792, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05615, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.02698, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": -0.06613, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.80992, "trunc_abs_frac": 0.82282, "mean_delta_full": -0.23938, "read_hit_rate": {"surface": 0.66247, "split": 0.8524, "merge": 0.17824, "skel": 0.65441}, "mean_gate": {"surface": 0.35011, "split": 0.20763, "merge": 0.23497, "skel": 0.29011}, "n_reads": 17081, "n_cand": 27148, "read_suppressed_frac": 0.1002, "dropout_reads_logged": 17081, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3472, "admitted_surface": 16867, "staged_surface": 0, "admitted_split": 21046, "staged_split": 0, "admitted_merge": 15252, "staged_merge": 0, "admitted_skel": 12371, "staged_skel": 0, "mean_count": 930.3963623046875, "n_util_obs": 456, "util_count_spearman": 0.07568129689321557}, "probe_overhead_sec_total": 298.143, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008522, "alpha": 0.03, "t": 2533.8} |
| {"event": "consolidate", "step": 3400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2571.9} |
| {"event": "probe", "step": 3400, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 2921.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.167845, "cps": 22.820312, "mean_p": 0.17475, "boundary_entropy": 0.047793, "bpb": 1.939527}, "tr": {"chunks": 3413.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.188532, "cps": 26.664062, "mean_p": 0.183642, "boundary_entropy": 0.104198, "bpb": 2.141456}, "fi": {"chunks": 2985.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.158684, "cps": 23.320312, "mean_p": 0.167903, "boundary_entropy": 0.120405, "bpb": 2.292445}}, "bpb": {"en": 1.93953, "tr": 2.14146, "fi": 2.29245}, "gini_cpc": 0.038634, "gini_cps": 0.035197, "bpic": 6.087778, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.1232548438116001, "fi": 0.9454205003733511}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3503, "admitted_surface": 16779, "staged_surface": 0, "admitted_split": 21016, "staged_split": 0, "admitted_merge": 15364, "staged_merge": 0, "admitted_skel": 12377, "staged_skel": 0, "mean_count": 944.21142578125, "n_util_obs": 462, "util_count_spearman": 0.06946263002004158}, "t": 2573.1} |
| {"event": "eval", "step": 3400, "phase": 2, "bpb": {"en": 1.95691, "tr": 2.15216, "fi": 2.30384}, "bytes_per_chunk": {"en": 5.397, "tr": 5.464, "fi": 6.098}, "bpb_mean": 2.13764, "frozen_inventory": true, "t": 2575.3} |
| {"event": "train", "step": 3400, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.41535, "bpb": 2.04192, "chunk_loss": 0.030164, "F": 0.18582, "G": 0.19361, "chunks": 190.28, "bytes_per_chunk": 5.382, "boundary_entropy": 0.06477, "router_noise": 0.0, "grad_norm": 0.6934, "lr": 0.0001465, "router_trainable": true, "sec_per_step": 0.867, "probe_sec": 0.0572, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03804, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.0069, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.04645, "n_merge": 1, "mean_delta_merge": -0.36743, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.07717, "trunc_abs_frac": 0.07717, "mean_delta_full": 0.28552, "read_hit_rate": {"surface": 0.72313, "split": 0.88741, "merge": 0.2083, "skel": 0.74524}, "mean_gate": {"surface": 0.36002, "split": 0.20491, "merge": 0.23434, "skel": 0.29048}, "n_reads": 19095, "n_cand": 28015, "read_suppressed_frac": 0.1015, "dropout_reads_logged": 19095, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3503, "admitted_surface": 16829, "staged_surface": 0, "admitted_split": 21057, "staged_split": 0, "admitted_merge": 15342, "staged_merge": 0, "admitted_skel": 12308, "staged_skel": 0, "mean_count": 944.5006103515625, "n_util_obs": 462, "util_count_spearman": 0.06930534479825982}, "probe_overhead_sec_total": 303.038, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008986, "alpha": 0.03, "t": 2577.2} |
| {"event": "consolidate", "step": 3450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2607.7} |
| {"event": "train", "step": 3450, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.39608, "bpb": 2.01412, "chunk_loss": 0.030176, "F": 0.1868, "G": 0.19343, "chunks": 191.28, "bytes_per_chunk": 5.353, "boundary_entropy": 0.059, "router_noise": 0.0, "grad_norm": 0.7115, "lr": 0.0001428, "router_trainable": true, "sec_per_step": 0.619, "probe_sec": 0.0713, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05149, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.05236, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": -0.00156, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.59342, "trunc_abs_frac": 0.59342, "mean_delta_full": -0.28748, "read_hit_rate": {"surface": 0.74035, "split": 0.89702, "merge": 0.20571, "skel": 0.74601}, "mean_gate": {"surface": 0.36366, "split": 0.20574, "merge": 0.23379, "skel": 0.29788}, "n_reads": 19250, "n_cand": 27992, "read_suppressed_frac": 0.0962, "dropout_reads_logged": 19250, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3542, "admitted_surface": 16773, "staged_surface": 0, "admitted_split": 21020, "staged_split": 0, "admitted_merge": 15405, "staged_merge": 0, "admitted_skel": 12338, "staged_skel": 0, "mean_count": 958.4954223632812, "n_util_obs": 468, "util_count_spearman": 0.06787210545105539}, "probe_overhead_sec_total": 306.999, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009824, "alpha": 0.03, "t": 2608.1} |
| {"event": "consolidate", "step": 3500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2642.8} |
| {"event": "train", "step": 3500, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.43179, "bpb": 2.06564, "chunk_loss": 0.030175, "F": 0.1846, "G": 0.19272, "chunks": 189.03, "bytes_per_chunk": 5.417, "boundary_entropy": 0.0635, "router_noise": 0.0, "grad_norm": 0.7194, "lr": 0.0001391, "router_trainable": true, "sec_per_step": 0.714, "probe_sec": 0.1331, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05738, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.04403, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.21558, "n_merge": 2, "mean_delta_merge": 0.03886, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.89599, "trunc_abs_frac": 0.89599, "mean_delta_full": 0.2406, "read_hit_rate": {"surface": 0.69985, "split": 0.87423, "merge": 0.2, "skel": 0.70212}, "mean_gate": {"surface": 0.36988, "split": 0.21052, "merge": 0.24142, "skel": 0.30651}, "n_reads": 18014, "n_cand": 27321, "read_suppressed_frac": 0.1013, "dropout_reads_logged": 18014, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3574, "admitted_surface": 16845, "staged_surface": 0, "admitted_split": 21119, "staged_split": 0, "admitted_merge": 15206, "staged_merge": 0, "admitted_skel": 12366, "staged_skel": 0, "mean_count": 972.4354248046875, "n_util_obs": 474, "util_count_spearman": 0.07530743354810142}, "probe_overhead_sec_total": 311.231, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009766, "alpha": 0.03, "t": 2643.8} |
| {"event": "consolidate", "step": 3550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2706.8} |
| {"event": "train", "step": 3550, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.4681, "bpb": 2.11802, "chunk_loss": 0.030115, "F": 0.18277, "G": 0.19009, "chunks": 187.16, "bytes_per_chunk": 5.471, "boundary_entropy": 0.0732, "router_noise": 0.0, "grad_norm": 0.7463, "lr": 0.0001355, "router_trainable": true, "sec_per_step": 1.284, "probe_sec": 0.1117, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06598, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.02782, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 3, "mean_delta_merge": -0.11632, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.74138, "trunc_abs_frac": 0.75695, "mean_delta_full": 0.25859, "read_hit_rate": {"surface": 0.68608, "split": 0.86216, "merge": 0.18498, "skel": 0.69905}, "mean_gate": {"surface": 0.35615, "split": 0.2047, "merge": 0.23098, "skel": 0.29082}, "n_reads": 17664, "n_cand": 27192, "read_suppressed_frac": 0.1041, "dropout_reads_logged": 17664, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3607, "admitted_surface": 16836, "staged_surface": 0, "admitted_split": 21216, "staged_split": 0, "admitted_merge": 15200, "staged_merge": 0, "admitted_skel": 12284, "staged_skel": 0, "mean_count": 986.4506225585938, "n_util_obs": 479, "util_count_spearman": 0.06593774474415366}, "probe_overhead_sec_total": 316.711, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.00984, "alpha": 0.03, "t": 2708.0} |
| {"event": "consolidate", "step": 3600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2752.1} |
| {"event": "probe", "step": 3600, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 2920.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.167787, "cps": 22.8125, "mean_p": 0.174703, "boundary_entropy": 0.047347, "bpb": 1.923304}, "tr": {"chunks": 3417.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.188753, "cps": 26.695312, "mean_p": 0.182103, "boundary_entropy": 0.100839, "bpb": 2.118032}, "fi": {"chunks": 2991.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.159003, "cps": 23.367188, "mean_p": 0.165592, "boundary_entropy": 0.111726, "bpb": 2.278538}}, "bpb": {"en": 1.9233, "tr": 2.11803, "fi": 2.27854}, "gini_cpc": 0.038471, "gini_cps": 0.03552, "bpic": 6.081904, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.1249564137935208, "fi": 0.9476452680339322}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3643, "admitted_surface": 16802, "staged_surface": 0, "admitted_split": 21107, "staged_split": 0, "admitted_merge": 15275, "staged_merge": 0, "admitted_skel": 12352, "staged_skel": 0, "mean_count": 1000.1996459960938, "n_util_obs": 491, "util_count_spearman": 0.07110254364524608}, "t": 2753.4} |
| {"event": "eval", "step": 3600, "phase": 2, "bpb": {"en": 1.94225, "tr": 2.12726, "fi": 2.28887}, "bytes_per_chunk": {"en": 5.401, "tr": 5.463, "fi": 6.097}, "bpb_mean": 2.11946, "frozen_inventory": true, "t": 2755.5} |
| {"event": "train", "step": 3600, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.44954, "bpb": 2.09125, "chunk_loss": 0.030157, "F": 0.18283, "G": 0.1901, "chunks": 187.22, "bytes_per_chunk": 5.47, "boundary_entropy": 0.06127, "router_noise": 0.0, "grad_norm": 0.7482, "lr": 0.0001319, "router_trainable": true, "sec_per_step": 1.006, "probe_sec": 0.071, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04674, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.07982, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.12469, "n_merge": 2, "mean_delta_merge": 0.62668, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 4.31898, "trunc_abs_frac": 1.00086, "mean_delta_full": 0.03613, "read_hit_rate": {"surface": 0.69424, "split": 0.87156, "merge": 0.17581, "skel": 0.70796}, "mean_gate": {"surface": 0.35363, "split": 0.19993, "merge": 0.22971, "skel": 0.28548}, "n_reads": 17871, "n_cand": 27309, "read_suppressed_frac": 0.0994, "dropout_reads_logged": 17871, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3643, "admitted_surface": 16810, "staged_surface": 0, "admitted_split": 21141, "staged_split": 0, "admitted_merge": 15265, "staged_merge": 0, "admitted_skel": 12320, "staged_skel": 0, "mean_count": 1000.4556274414062, "n_util_obs": 491, "util_count_spearman": 0.07166919107572928}, "probe_overhead_sec_total": 321.666, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009453, "alpha": 0.03, "t": 2758.3} |
| {"event": "consolidate", "step": 3650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2789.2} |
| {"event": "train", "step": 3650, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.37953, "bpb": 1.99024, "chunk_loss": 0.030087, "F": 0.17914, "G": 0.18655, "chunks": 183.44, "bytes_per_chunk": 5.582, "boundary_entropy": 0.06028, "router_noise": 0.0, "grad_norm": 0.7146, "lr": 0.0001283, "router_trainable": true, "sec_per_step": 0.628, "probe_sec": 0.0619, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03988, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.02085, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.64624, "n_merge": 2, "mean_delta_merge": 0.06172, "n_dropped_unpaired": 0, "n_full": 0, "read_hit_rate": {"surface": 0.71137, "split": 0.88574, "merge": 0.20014, "skel": 0.72497}, "mean_gate": {"surface": 0.35547, "split": 0.19897, "merge": 0.2313, "skel": 0.28665}, "n_reads": 18303, "n_cand": 27174, "read_suppressed_frac": 0.1011, "dropout_reads_logged": 18303, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3683, "admitted_surface": 16829, "staged_surface": 0, "admitted_split": 21153, "staged_split": 0, "admitted_merge": 15169, "staged_merge": 0, "admitted_skel": 12385, "staged_skel": 0, "mean_count": 1014.6754760742188, "n_util_obs": 497, "util_count_spearman": 0.06400211725082906}, "probe_overhead_sec_total": 325.606, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.010154, "alpha": 0.03, "t": 2789.7} |
| {"event": "consolidate", "step": 3700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2823.7} |
| {"event": "train", "step": 3700, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.358, "bpb": 1.95918, "chunk_loss": 0.030143, "F": 0.18536, "G": 0.19172, "chunks": 189.81, "bytes_per_chunk": 5.395, "boundary_entropy": 0.05224, "router_noise": 0.0, "grad_norm": 0.6694, "lr": 0.0001247, "router_trainable": true, "sec_per_step": 0.689, "probe_sec": 0.0763, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05206, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": -0.18629, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": -0.0186, "n_merge": 3, "mean_delta_merge": -0.09497, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.27854, "trunc_abs_frac": 0.49736, "mean_delta_full": -0.23468, "read_hit_rate": {"surface": 0.74909, "split": 0.89883, "merge": 0.21531, "skel": 0.76086}, "mean_gate": {"surface": 0.361, "split": 0.19949, "merge": 0.22978, "skel": 0.29014}, "n_reads": 19431, "n_cand": 27930, "read_suppressed_frac": 0.0988, "dropout_reads_logged": 19431, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3723, "admitted_surface": 16747, "staged_surface": 0, "admitted_split": 21153, "staged_split": 0, "admitted_merge": 15331, "staged_merge": 0, "admitted_skel": 12305, "staged_skel": 0, "mean_count": 1028.7159423828125, "n_util_obs": 507, "util_count_spearman": 0.06249186973987579}, "probe_overhead_sec_total": 329.871, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.010066, "alpha": 0.03, "t": 2824.2} |
| {"event": "consolidate", "step": 3750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2860.8} |
| {"event": "train", "step": 3750, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.37024, "bpb": 1.97683, "chunk_loss": 0.030131, "F": 0.18109, "G": 0.18647, "chunks": 185.44, "bytes_per_chunk": 5.522, "boundary_entropy": 0.05579, "router_noise": 0.0, "grad_norm": 0.7108, "lr": 0.0001212, "router_trainable": true, "sec_per_step": 0.755, "probe_sec": 0.0683, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04306, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.06617, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.08832, "n_merge": 2, "mean_delta_merge": 0.01852, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.30293, "trunc_abs_frac": 0.63929, "mean_delta_full": 0.40299, "read_hit_rate": {"surface": 0.72518, "split": 0.88858, "merge": 0.21465, "skel": 0.73126}, "mean_gate": {"surface": 0.36744, "split": 0.20502, "merge": 0.23524, "skel": 0.296}, "n_reads": 18524, "n_cand": 27199, "read_suppressed_frac": 0.0987, "dropout_reads_logged": 18524, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3754, "admitted_surface": 16793, "staged_surface": 0, "admitted_split": 21190, "staged_split": 0, "admitted_merge": 15206, "staged_merge": 0, "admitted_skel": 12347, "staged_skel": 0, "mean_count": 1042.7039794921875, "n_util_obs": 511, "util_count_spearman": 0.05659012876168524}, "probe_overhead_sec_total": 334.109, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009094, "alpha": 0.03, "t": 2861.9} |
| {"event": "consolidate", "step": 3800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2900.6} |
| {"event": "probe", "step": 3800, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 2919.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.16773, "cps": 22.804688, "mean_p": 0.1739, "boundary_entropy": 0.04364, "bpb": 1.910431}, "tr": {"chunks": 3417.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.188753, "cps": 26.695312, "mean_p": 0.182302, "boundary_entropy": 0.092249, "bpb": 2.102223}, "fi": {"chunks": 2989.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.158896, "cps": 23.351562, "mean_p": 0.164986, "boundary_entropy": 0.109067, "bpb": 2.255318}}, "bpb": {"en": 1.91043, "tr": 2.10222, "fi": 2.25532}, "gini_cpc": 0.038621, "gini_cps": 0.035603, "bpic": 6.083861, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.1253418048225698, "fi": 0.9473360336769125}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3781, "admitted_surface": 16816, "staged_surface": 0, "admitted_split": 21201, "staged_split": 0, "admitted_merge": 15221, "staged_merge": 0, "admitted_skel": 12298, "staged_skel": 0, "mean_count": 1056.50537109375, "n_util_obs": 518, "util_count_spearman": 0.05476007175930748}, "t": 2901.8} |
| {"event": "eval", "step": 3800, "phase": 2, "bpb": {"en": 1.92792, "tr": 2.10904, "fi": 2.26659}, "bytes_per_chunk": {"en": 5.409, "tr": 5.467, "fi": 6.09}, "bpb_mean": 2.10118, "frozen_inventory": true, "t": 2903.9} |
| {"event": "train", "step": 3800, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.33144, "bpb": 1.92086, "chunk_loss": 0.030093, "F": 0.18048, "G": 0.187, "chunks": 184.81, "bytes_per_chunk": 5.541, "boundary_entropy": 0.05664, "router_noise": 0.0, "grad_norm": 0.7372, "lr": 0.0001177, "router_trainable": true, "sec_per_step": 0.87, "probe_sec": 0.064, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04684, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.00024, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": -0.10052, "n_merge": 2, "mean_delta_merge": -0.36198, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -0.45334, "trunc_abs_frac": 2.25159, "mean_delta_full": 0.04709, "read_hit_rate": {"surface": 0.7324, "split": 0.89788, "merge": 0.21231, "skel": 0.74605}, "mean_gate": {"surface": 0.36779, "split": 0.20287, "merge": 0.23656, "skel": 0.29484}, "n_reads": 18896, "n_cand": 27399, "read_suppressed_frac": 0.1035, "dropout_reads_logged": 18896, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3783, "admitted_surface": 16825, "staged_surface": 0, "admitted_split": 21114, "staged_split": 0, "admitted_merge": 15235, "staged_merge": 0, "admitted_skel": 12362, "staged_skel": 0, "mean_count": 1056.8050537109375, "n_util_obs": 518, "util_count_spearman": 0.05487675649926458}, "probe_overhead_sec_total": 338.282, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.010615, "alpha": 0.03, "t": 2905.4} |
| {"event": "consolidate", "step": 3850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2936.0} |
| {"event": "train", "step": 3850, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.35597, "bpb": 1.95626, "chunk_loss": 0.030185, "F": 0.18689, "G": 0.19281, "chunks": 191.38, "bytes_per_chunk": 5.351, "boundary_entropy": 0.04826, "router_noise": 0.0, "grad_norm": 0.6854, "lr": 0.0001142, "router_trainable": true, "sec_per_step": 0.62, "probe_sec": 0.062, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03983, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.03294, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": -0.11211, "n_merge": 1, "mean_delta_merge": 0.11792, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -1.80805, "trunc_abs_frac": 0.76094, "mean_delta_full": 0.03357, "read_hit_rate": {"surface": 0.75886, "split": 0.91206, "merge": 0.20875, "skel": 0.76963}, "mean_gate": {"surface": 0.35828, "split": 0.19347, "merge": 0.21939, "skel": 0.28632}, "n_reads": 20081, "n_cand": 28491, "read_suppressed_frac": 0.101, "dropout_reads_logged": 20081, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3812, "admitted_surface": 16713, "staged_surface": 0, "admitted_split": 21085, "staged_split": 0, "admitted_merge": 15413, "staged_merge": 0, "admitted_skel": 12325, "staged_skel": 0, "mean_count": 1070.8675537109375, "n_util_obs": 524, "util_count_spearman": 0.058938556350663696}, "probe_overhead_sec_total": 342.011, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009443, "alpha": 0.03, "t": 2936.4} |
| {"event": "consolidate", "step": 3900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2968.9} |
| {"event": "train", "step": 3900, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.38454, "bpb": 1.99746, "chunk_loss": 0.030141, "F": 0.1842, "G": 0.19094, "chunks": 188.62, "bytes_per_chunk": 5.429, "boundary_entropy": 0.05621, "router_noise": 0.0, "grad_norm": 0.6654, "lr": 0.0001108, "router_trainable": true, "sec_per_step": 0.659, "probe_sec": 0.068, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.045, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": 0.00403, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.0741, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -0.59668, "trunc_abs_frac": 1.26789, "mean_delta_full": 0.11301, "read_hit_rate": {"surface": 0.73518, "split": 0.89912, "merge": 0.20328, "skel": 0.75083}, "mean_gate": {"surface": 0.36545, "split": 0.19888, "merge": 0.22806, "skel": 0.29122}, "n_reads": 19250, "n_cand": 27920, "read_suppressed_frac": 0.1004, "dropout_reads_logged": 19250, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3846, "admitted_surface": 16762, "staged_surface": 0, "admitted_split": 21171, "staged_split": 0, "admitted_merge": 15259, "staged_merge": 0, "admitted_skel": 12344, "staged_skel": 0, "mean_count": 1084.9637451171875, "n_util_obs": 535, "util_count_spearman": 0.04869213236446407}, "probe_overhead_sec_total": 346.214, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.010204, "alpha": 0.03, "t": 2969.4} |
| {"event": "consolidate", "step": 3950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3005.4} |
| {"event": "train", "step": 3950, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.37546, "bpb": 1.98436, "chunk_loss": 0.030135, "F": 0.18527, "G": 0.19138, "chunks": 189.72, "bytes_per_chunk": 5.397, "boundary_entropy": 0.05258, "router_noise": 0.0, "grad_norm": 0.714, "lr": 0.0001074, "router_trainable": true, "sec_per_step": 0.73, "probe_sec": 0.141, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.11984, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.02951, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.21912, "n_merge": 1, "mean_delta_merge": -0.32794, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -0.22163, "trunc_abs_frac": 0.64735, "mean_delta_full": -0.10002, "read_hit_rate": {"surface": 0.73257, "split": 0.89647, "merge": 0.2016, "skel": 0.74566}, "mean_gate": {"surface": 0.3634, "split": 0.19826, "merge": 0.22773, "skel": 0.29204}, "n_reads": 19211, "n_cand": 28004, "read_suppressed_frac": 0.0977, "dropout_reads_logged": 19211, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3880, "admitted_surface": 16730, "staged_surface": 0, "admitted_split": 21175, "staged_split": 0, "admitted_merge": 15314, "staged_merge": 0, "admitted_skel": 12317, "staged_skel": 0, "mean_count": 1099.0318603515625, "n_util_obs": 542, "util_count_spearman": 0.04653849559158703}, "probe_overhead_sec_total": 350.479, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008705, "alpha": 0.03, "t": 3005.9} |
| {"event": "consolidate", "step": 4000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3042.0} |
| {"event": "probe", "step": 4000, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 2920.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.167787, "cps": 22.8125, "mean_p": 0.17399, "boundary_entropy": 0.043282, "bpb": 1.902292}, "tr": {"chunks": 3414.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.188588, "cps": 26.671875, "mean_p": 0.181565, "boundary_entropy": 0.087995, "bpb": 2.084089}, "fi": {"chunks": 2998.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.159375, "cps": 23.421875, "mean_p": 0.164993, "boundary_entropy": 0.108219, "bpb": 2.238812}}, "bpb": {"en": 1.90229, "tr": 2.08409, "fi": 2.23881}, "gini_cpc": 0.037761, "gini_cps": 0.035291, "bpic": 6.079297, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.1239687435443606, "fi": 0.9498630938033197}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3908, "admitted_surface": 16754, "staged_surface": 0, "admitted_split": 21239, "staged_split": 0, "admitted_merge": 15191, "staged_merge": 0, "admitted_skel": 12352, "staged_skel": 0, "mean_count": 1112.791015625, "n_util_obs": 549, "util_count_spearman": 0.049995143791367706}, "t": 3043.3} |
| {"event": "eval", "step": 4000, "phase": 2, "bpb": {"en": 1.91959, "tr": 2.09069, "fi": 2.24973}, "bytes_per_chunk": {"en": 5.404, "tr": 5.466, "fi": 6.078}, "bpb_mean": 2.08667, "frozen_inventory": true, "t": 3045.3} |
| {"event": "train", "step": 4000, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.40885, "bpb": 2.03254, "chunk_loss": 0.030102, "F": 0.17862, "G": 0.18568, "chunks": 182.91, "bytes_per_chunk": 5.598, "boundary_entropy": 0.05837, "router_noise": 0.0, "grad_norm": 0.7513, "lr": 0.000104, "router_trainable": true, "sec_per_step": 0.826, "probe_sec": 0.093, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07477, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.22173, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": 0.08063, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 2.32584, "trunc_abs_frac": 0.36955, "mean_delta_full": -0.02307, "read_hit_rate": {"surface": 0.69301, "split": 0.87744, "merge": 0.18051, "skel": 0.71435}, "mean_gate": {"surface": 0.35905, "split": 0.19692, "merge": 0.22242, "skel": 0.28801}, "n_reads": 18017, "n_cand": 27244, "read_suppressed_frac": 0.0984, "dropout_reads_logged": 18017, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3910, "admitted_surface": 16795, "staged_surface": 0, "admitted_split": 21248, "staged_split": 0, "admitted_merge": 15089, "staged_merge": 0, "admitted_skel": 12404, "staged_skel": 0, "mean_count": 1113.049072265625, "n_util_obs": 550, "util_count_spearman": 0.04528402292272772}, "probe_overhead_sec_total": 354.699, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.011003, "alpha": 0.03, "t": 3047.2} |
| {"event": "consolidate", "step": 4050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3081.1} |
| {"event": "train", "step": 4050, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.41349, "bpb": 2.03924, "chunk_loss": 0.030158, "F": 0.18436, "G": 0.19088, "chunks": 188.78, "bytes_per_chunk": 5.424, "boundary_entropy": 0.05419, "router_noise": 0.0, "grad_norm": 0.7441, "lr": 0.0001007, "router_trainable": true, "sec_per_step": 0.688, "probe_sec": 0.0657, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04147, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": 0.03878, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.00134, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.6029, "trunc_abs_frac": 0.33388, "mean_delta_full": 0.04834, "read_hit_rate": {"surface": 0.72608, "split": 0.89412, "merge": 0.20231, "skel": 0.73873}, "mean_gate": {"surface": 0.36376, "split": 0.19964, "merge": 0.22708, "skel": 0.29355}, "n_reads": 18902, "n_cand": 27703, "read_suppressed_frac": 0.0961, "dropout_reads_logged": 18902, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3936, "admitted_surface": 16752, "staged_surface": 0, "admitted_split": 21230, "staged_split": 0, "admitted_merge": 15233, "staged_merge": 0, "admitted_skel": 12321, "staged_skel": 0, "mean_count": 1127.181396484375, "n_util_obs": 558, "util_count_spearman": 0.04813180315054947}, "probe_overhead_sec_total": 358.876, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.010187, "alpha": 0.03, "t": 3081.6} |
| {"event": "probe", "step": 4096, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 2923.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.16796, "cps": 22.835938, "mean_p": 0.173678, "boundary_entropy": 0.040996, "bpb": 1.89338}, "tr": {"chunks": 3413.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.188532, "cps": 26.664062, "mean_p": 0.17967, "boundary_entropy": 0.088326, "bpb": 2.082173}, "fi": {"chunks": 2993.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.159109, "cps": 23.382812, "mean_p": 0.163589, "boundary_entropy": 0.104914, "bpb": 2.235428}}, "bpb": {"en": 1.89338, "tr": 2.08217, "fi": 2.23543}, "gini_cpc": 0.038044, "gini_cps": 0.035016, "bpic": 6.081252, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.1224862807983864, "fi": 0.9473056732860955}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3964, "admitted_surface": 16714, "staged_surface": 0, "admitted_split": 21232, "staged_split": 0, "admitted_merge": 15237, "staged_merge": 0, "admitted_skel": 12353, "staged_skel": 0, "mean_count": 1139.8853759765625, "n_util_obs": 563, "util_count_spearman": 0.045844278373432384}, "t": 3110.6} |
| {"event": "consolidate", "step": 4100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3114.3} |
| {"event": "train", "step": 4100, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.45516, "bpb": 2.09935, "chunk_loss": 0.030142, "F": 0.18155, "G": 0.1885, "chunks": 185.91, "bytes_per_chunk": 5.508, "boundary_entropy": 0.05961, "router_noise": 0.0, "grad_norm": 0.7243, "lr": 9.75e-05, "router_trainable": true, "sec_per_step": 0.662, "probe_sec": 0.0759, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04361, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.00223, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 7, "trunc_ratio_sum": -2.57238, "trunc_abs_frac": 0.69358, "mean_delta_full": -0.01632, "read_hit_rate": {"surface": 0.69731, "split": 0.8707, "merge": 0.2034, "skel": 0.69487}, "mean_gate": {"surface": 0.35175, "split": 0.19519, "merge": 0.22277, "skel": 0.28665}, "n_reads": 17842, "n_cand": 27094, "read_suppressed_frac": 0.0993, "dropout_reads_logged": 17842, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3966, "admitted_surface": 16800, "staged_surface": 0, "admitted_split": 21324, "staged_split": 0, "admitted_merge": 14982, "staged_merge": 0, "admitted_skel": 12430, "staged_skel": 0, "mean_count": 1141.26513671875, "n_util_obs": 565, "util_count_spearman": 0.04185552619970685}, "probe_overhead_sec_total": 362.843, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.010058, "alpha": 0.03, "t": 3114.7} |
| {"event": "consolidate", "step": 4150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3157.1} |
| {"event": "train", "step": 4150, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.37261, "bpb": 1.98026, "chunk_loss": 0.030139, "F": 0.18195, "G": 0.18826, "chunks": 186.31, "bytes_per_chunk": 5.496, "boundary_entropy": 0.0547, "router_noise": 0.0, "grad_norm": 0.7107, "lr": 9.43e-05, "router_trainable": true, "sec_per_step": 0.863, "probe_sec": 0.1101, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07837, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.11821, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 3, "mean_delta_split": 0.00596, "n_merge": 1, "mean_delta_merge": -1.17999, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -1.35724, "trunc_abs_frac": 1.17249, "mean_delta_full": -0.03064, "read_hit_rate": {"surface": 0.71956, "split": 0.88738, "merge": 0.19566, "skel": 0.72449}, "mean_gate": {"surface": 0.36233, "split": 0.19724, "merge": 0.22176, "skel": 0.29201}, "n_reads": 18450, "n_cand": 27374, "read_suppressed_frac": 0.099, "dropout_reads_logged": 18450, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4003, "admitted_surface": 16769, "staged_surface": 0, "admitted_split": 21287, "staged_split": 0, "admitted_merge": 15071, "staged_merge": 0, "admitted_skel": 12409, "staged_skel": 0, "mean_count": 1155.318603515625, "n_util_obs": 568, "util_count_spearman": 0.034451354918080955}, "probe_overhead_sec_total": 368.17, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.010338, "alpha": 0.03, "t": 3157.8} |
| {"event": "consolidate", "step": 4200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3204.6} |
| {"event": "probe", "step": 4200, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 2921.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.167845, "cps": 22.820312, "mean_p": 0.173728, "boundary_entropy": 0.041503, "bpb": 1.888693}, "tr": {"chunks": 3417.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.188753, "cps": 26.695312, "mean_p": 0.180344, "boundary_entropy": 0.086218, "bpb": 2.07373}, "fi": {"chunks": 2993.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.159109, "cps": 23.382812, "mean_p": 0.162992, "boundary_entropy": 0.10382, "bpb": 2.222517}}, "bpb": {"en": 1.88869, "tr": 2.07373, "fi": 2.22252}, "gini_cpc": 0.038322, "gini_cps": 0.035437, "bpic": 6.079949, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.1245712866405617, "fi": 0.947954290659109}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4026, "admitted_surface": 16771, "staged_surface": 0, "admitted_split": 21332, "staged_split": 0, "admitted_merge": 15023, "staged_merge": 0, "admitted_skel": 12410, "staged_skel": 0, "mean_count": 1169.068603515625, "n_util_obs": 571, "util_count_spearman": 0.033625672284943495}, "t": 3206.2} |
| {"event": "eval", "step": 4200, "phase": 2, "bpb": {"en": 1.90735, "tr": 2.08119, "fi": 2.23404}, "bytes_per_chunk": {"en": 5.408, "tr": 5.468, "fi": 6.081}, "bpb_mean": 2.07419, "frozen_inventory": true, "t": 3209.3} |
| {"event": "phase_transition", "step": 4200, "phase": 3, "objective": "global", "router_trainable": true, "encoder_trainable": true, "frozen_router_params": 0, "t": 3210.4} |
| {"event": "train", "step": 4200, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.44391, "bpb": 2.08312, "chunk_loss": 0.030197, "F": 0.18765, "G": 0.1939, "chunks": 192.16, "bytes_per_chunk": 5.329, "boundary_entropy": 0.05848, "router_noise": 0.0, "grad_norm": 0.747, "lr": 9.12e-05, "router_trainable": true, "sec_per_step": 1.07, "probe_sec": 0.2022, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.1749, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.2555, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.13818, "n_merge": 1, "mean_delta_merge": 0.11829, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.299, "trunc_abs_frac": 0.299, "mean_delta_full": -0.8457, "read_hit_rate": {"surface": 0.68955, "split": 0.87103, "merge": 0.20526, "skel": 0.69617}, "mean_gate": {"surface": 0.36323, "split": 0.2001, "merge": 0.23101, "skel": 0.28853}, "n_reads": 18303, "n_cand": 27857, "read_suppressed_frac": 0.0992, "dropout_reads_logged": 18303, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4026, "admitted_surface": 16786, "staged_surface": 0, "admitted_split": 21313, "staged_split": 0, "admitted_merge": 15062, "staged_merge": 0, "admitted_skel": 12375, "staged_skel": 0, "mean_count": 1169.2877197265625, "n_util_obs": 571, "util_count_spearman": 0.03348396058358575}, "probe_overhead_sec_total": 375.987, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.01032, "alpha": 0.03, "t": 3211.4} |
| {"event": "consolidate", "step": 4250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3253.6} |
| {"event": "train", "step": 4250, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35236, "bpb": 1.95105, "chunk_loss": 0.030147, "F": 0.18433, "G": 0.18956, "chunks": 188.75, "bytes_per_chunk": 5.425, "boundary_entropy": 0.05269, "router_noise": 0.0, "grad_norm": 0.6768, "lr": 8.81e-05, "router_trainable": true, "sec_per_step": 0.87, "probe_sec": 0.1478, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.12346, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.00089, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.07344, "n_merge": 2, "mean_delta_merge": 0.30426, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.49842, "trunc_abs_frac": 0.49842, "mean_delta_full": 0.4646, "read_hit_rate": {"surface": 0.72087, "split": 0.88577, "merge": 0.20465, "skel": 0.72829}, "mean_gate": {"surface": 0.3601, "split": 0.19407, "merge": 0.22109, "skel": 0.28645}, "n_reads": 18588, "n_cand": 27526, "read_suppressed_frac": 0.1018, "dropout_reads_logged": 18588, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4042, "admitted_surface": 16797, "staged_surface": 0, "admitted_split": 21353, "staged_split": 0, "admitted_merge": 15000, "staged_merge": 0, "admitted_skel": 12386, "staged_skel": 0, "mean_count": 1183.37109375, "n_util_obs": 574, "util_count_spearman": 0.029417628438579556}, "probe_overhead_sec_total": 384.196, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.011651, "alpha": 0.03, "t": 3254.9} |
| {"event": "consolidate", "step": 4300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3297.5} |
| {"event": "train", "step": 4300, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.41004, "bpb": 2.03426, "chunk_loss": 0.030137, "F": 0.18384, "G": 0.18967, "chunks": 188.25, "bytes_per_chunk": 5.44, "boundary_entropy": 0.05221, "router_noise": 0.0, "grad_norm": 0.6706, "lr": 8.5e-05, "router_trainable": true, "sec_per_step": 0.873, "probe_sec": 0.1497, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.12231, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.07614, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.51944, "n_merge": 1, "mean_delta_merge": -0.13126, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.23377, "trunc_abs_frac": 1.40691, "mean_delta_full": -0.24768, "read_hit_rate": {"surface": 0.71312, "split": 0.88206, "merge": 0.18775, "skel": 0.7245}, "mean_gate": {"surface": 0.3555, "split": 0.19163, "merge": 0.21627, "skel": 0.28371}, "n_reads": 18414, "n_cand": 27547, "read_suppressed_frac": 0.096, "dropout_reads_logged": 18414, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4068, "admitted_surface": 16754, "staged_surface": 0, "admitted_split": 21343, "staged_split": 0, "admitted_merge": 15080, "staged_merge": 0, "admitted_skel": 12359, "staged_skel": 0, "mean_count": 1197.38671875, "n_util_obs": 582, "util_count_spearman": 0.0372468084761229}, "probe_overhead_sec_total": 392.223, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.011425, "alpha": 0.03, "t": 3298.5} |
| {"event": "consolidate", "step": 4350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3345.8} |
| {"event": "train", "step": 4350, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.43096, "bpb": 2.06444, "chunk_loss": 0.030127, "F": 0.18091, "G": 0.18703, "chunks": 185.25, "bytes_per_chunk": 5.528, "boundary_entropy": 0.0559, "router_noise": 0.0, "grad_norm": 0.6852, "lr": 8.21e-05, "router_trainable": true, "sec_per_step": 0.972, "probe_sec": 0.1431, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.12138, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.01404, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.27319, "n_merge": 3, "mean_delta_merge": -0.22487, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -2.9885, "trunc_abs_frac": 0.53961, "mean_delta_full": 0.03271, "read_hit_rate": {"surface": 0.6959, "split": 0.8816, "merge": 0.19969, "skel": 0.70725}, "mean_gate": {"surface": 0.35188, "split": 0.19054, "merge": 0.21684, "skel": 0.28048}, "n_reads": 18096, "n_cand": 27220, "read_suppressed_frac": 0.0995, "dropout_reads_logged": 18096, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4103, "admitted_surface": 16810, "staged_surface": 0, "admitted_split": 21365, "staged_split": 0, "admitted_merge": 14954, "staged_merge": 0, "admitted_skel": 12407, "staged_skel": 0, "mean_count": 1211.478759765625, "n_util_obs": 589, "util_count_spearman": 0.04052375002163717}, "probe_overhead_sec_total": 400.253, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.012889, "alpha": 0.03, "t": 3347.1} |
| {"event": "consolidate", "step": 4400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3389.2} |
| {"event": "probe", "step": 4400, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 2919.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.16773, "cps": 22.804688, "mean_p": 0.173467, "boundary_entropy": 0.040195, "bpb": 1.878619}, "tr": {"chunks": 3412.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.188477, "cps": 26.65625, "mean_p": 0.178093, "boundary_entropy": 0.084905, "bpb": 2.05407}, "fi": {"chunks": 2995.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.159215, "cps": 23.398438, "mean_p": 0.162848, "boundary_entropy": 0.102449, "bpb": 2.208812}}, "bpb": {"en": 1.87862, "tr": 2.05407, "fi": 2.20881}, "gini_cpc": 0.037848, "gini_cps": 0.035242, "bpic": 6.083208, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.123695123808782, "fi": 0.9492376784417375}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4133, "admitted_surface": 16783, "staged_surface": 0, "admitted_split": 21307, "staged_split": 0, "admitted_merge": 15107, "staged_merge": 0, "admitted_skel": 12339, "staged_skel": 0, "mean_count": 1225.29248046875, "n_util_obs": 595, "util_count_spearman": 0.03591848881683821}, "t": 3390.8} |
| {"event": "eval", "step": 4400, "phase": 3, "bpb": {"en": 1.8964, "tr": 2.06163, "fi": 2.21938}, "bytes_per_chunk": {"en": 5.412, "tr": 5.47, "fi": 6.081}, "bpb_mean": 2.05914, "frozen_inventory": true, "t": 3393.8} |
| {"event": "train", "step": 4400, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38944, "bpb": 2.00454, "chunk_loss": 0.030209, "F": 0.18842, "G": 0.19351, "chunks": 192.94, "bytes_per_chunk": 5.307, "boundary_entropy": 0.05283, "router_noise": 0.0, "grad_norm": 0.8006, "lr": 7.92e-05, "router_trainable": true, "sec_per_step": 0.973, "probe_sec": 0.1267, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10733, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.05941, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": -0.1654, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -0.19148, "trunc_abs_frac": 0.63716, "mean_delta_full": -0.1342, "read_hit_rate": {"surface": 0.71687, "split": 0.88803, "merge": 0.20327, "skel": 0.72667}, "mean_gate": {"surface": 0.34938, "split": 0.1918, "merge": 0.21282, "skel": 0.28067}, "n_reads": 18884, "n_cand": 27982, "read_suppressed_frac": 0.101, "dropout_reads_logged": 18884, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4133, "admitted_surface": 16759, "staged_surface": 0, "admitted_split": 21333, "staged_split": 0, "admitted_merge": 15127, "staged_merge": 0, "admitted_skel": 12317, "staged_skel": 0, "mean_count": 1225.579345703125, "n_util_obs": 595, "util_count_spearman": 0.03668003707482586}, "probe_overhead_sec_total": 408.461, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.011201, "alpha": 0.03, "t": 3395.8} |
| {"event": "consolidate", "step": 4450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3442.3} |
| {"event": "train", "step": 4450, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.37213, "bpb": 1.97957, "chunk_loss": 0.030174, "F": 0.18655, "G": 0.19327, "chunks": 191.03, "bytes_per_chunk": 5.36, "boundary_entropy": 0.05812, "router_noise": 0.0, "grad_norm": 0.7601, "lr": 7.63e-05, "router_trainable": true, "sec_per_step": 0.945, "probe_sec": 0.1381, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10749, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.081, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.05484, "n_merge": 3, "mean_delta_merge": 0.30647, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.29659, "trunc_abs_frac": 0.34642, "mean_delta_full": 0.27808, "read_hit_rate": {"surface": 0.71896, "split": 0.89106, "merge": 0.21061, "skel": 0.72574}, "mean_gate": {"surface": 0.34967, "split": 0.19158, "merge": 0.21834, "skel": 0.27636}, "n_reads": 18975, "n_cand": 27947, "read_suppressed_frac": 0.0983, "dropout_reads_logged": 18975, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4155, "admitted_surface": 16791, "staged_surface": 0, "admitted_split": 21349, "staged_split": 0, "admitted_merge": 15019, "staged_merge": 0, "admitted_skel": 12377, "staged_skel": 0, "mean_count": 1239.5631103515625, "n_util_obs": 600, "util_count_spearman": 0.03176470155227533}, "probe_overhead_sec_total": 416.077, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009441, "alpha": 0.03, "t": 3443.0} |
| {"event": "consolidate", "step": 4500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3488.7} |
| {"event": "train", "step": 4500, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.45824, "bpb": 2.10379, "chunk_loss": 0.030324, "F": 0.19116, "G": 0.19542, "chunks": 195.75, "bytes_per_chunk": 5.231, "boundary_entropy": 0.05612, "router_noise": 0.0, "grad_norm": 0.7064, "lr": 7.36e-05, "router_trainable": true, "sec_per_step": 0.931, "probe_sec": 0.192, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.16857, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.05418, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.26227, "n_merge": 2, "mean_delta_merge": -0.1582, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.31784, "trunc_abs_frac": 0.66538, "mean_delta_full": 0.14923, "read_hit_rate": {"surface": 0.69689, "split": 0.88309, "merge": 0.19919, "skel": 0.72337}, "mean_gate": {"surface": 0.3549, "split": 0.19077, "merge": 0.21719, "skel": 0.28199}, "n_reads": 18824, "n_cand": 28279, "read_suppressed_frac": 0.0978, "dropout_reads_logged": 18824, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4181, "admitted_surface": 16798, "staged_surface": 0, "admitted_split": 21325, "staged_split": 0, "admitted_merge": 15092, "staged_merge": 0, "admitted_skel": 12321, "staged_skel": 0, "mean_count": 1253.6500244140625, "n_util_obs": 604, "util_count_spearman": 0.028281312885538278}, "probe_overhead_sec_total": 423.69, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.011034, "alpha": 0.03, "t": 3489.6} |
| {"event": "consolidate", "step": 4550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3531.5} |
| {"event": "train", "step": 4550, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.37227, "bpb": 1.97977, "chunk_loss": 0.030206, "F": 0.18948, "G": 0.19429, "chunks": 194.03, "bytes_per_chunk": 5.278, "boundary_entropy": 0.05212, "router_noise": 0.0, "grad_norm": 0.7548, "lr": 7.09e-05, "router_trainable": true, "sec_per_step": 0.863, "probe_sec": 0.1552, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.12659, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.03685, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.21802, "n_merge": 1, "mean_delta_merge": 0.00134, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -0.22791, "trunc_abs_frac": 0.7655, "mean_delta_full": -0.19202, "read_hit_rate": {"surface": 0.73563, "split": 0.89448, "merge": 0.21709, "skel": 0.74246}, "mean_gate": {"surface": 0.35628, "split": 0.19142, "merge": 0.2146, "skel": 0.27953}, "n_reads": 19426, "n_cand": 28268, "read_suppressed_frac": 0.1009, "dropout_reads_logged": 19426, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4206, "admitted_surface": 16724, "staged_surface": 0, "admitted_split": 21351, "staged_split": 0, "admitted_merge": 15130, "staged_merge": 0, "admitted_skel": 12331, "staged_skel": 0, "mean_count": 1267.9150390625, "n_util_obs": 610, "util_count_spearman": 0.024968713374862773}, "probe_overhead_sec_total": 431.253, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.010731, "alpha": 0.03, "t": 3532.7} |
| {"event": "consolidate", "step": 4600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3574.1} |
| {"event": "probe", "step": 4600, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 2921.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.167845, "cps": 22.820312, "mean_p": 0.173241, "boundary_entropy": 0.03936, "bpb": 1.868273}, "tr": {"chunks": 3413.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.188532, "cps": 26.664062, "mean_p": 0.178509, "boundary_entropy": 0.083039, "bpb": 2.047578}, "fi": {"chunks": 2986.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.158737, "cps": 23.328125, "mean_p": 0.161698, "boundary_entropy": 0.102233, "bpb": 2.195264}}, "bpb": {"en": 1.86827, "tr": 2.04758, "fi": 2.19526}, "gini_cpc": 0.038562, "gini_cps": 0.035193, "bpic": 6.087124, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.1232548438116001, "fi": 0.9457372241590709}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4238, "admitted_surface": 16776, "staged_surface": 0, "admitted_split": 21419, "staged_split": 0, "admitted_merge": 14931, "staged_merge": 0, "admitted_skel": 12410, "staged_skel": 0, "mean_count": 1281.6717529296875, "n_util_obs": 618, "util_count_spearman": 0.01958878685611052}, "t": 3575.7} |
| {"event": "eval", "step": 4600, "phase": 3, "bpb": {"en": 1.88604, "tr": 2.05239, "fi": 2.2064}, "bytes_per_chunk": {"en": 5.416, "tr": 5.469, "fi": 6.096}, "bpb_mean": 2.04827, "frozen_inventory": true, "t": 3578.8} |
| {"event": "train", "step": 4600, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.31684, "bpb": 1.8998, "chunk_loss": 0.030086, "F": 0.18069, "G": 0.18583, "chunks": 185.03, "bytes_per_chunk": 5.534, "boundary_entropy": 0.04968, "router_noise": 0.0, "grad_norm": 0.6706, "lr": 6.82e-05, "router_trainable": true, "sec_per_step": 0.962, "probe_sec": 0.0752, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04392, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": 0.00511, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.03891, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 1.56775, "trunc_abs_frac": 0.70029, "mean_delta_full": 0.04504, "read_hit_rate": {"surface": 0.73374, "split": 0.89539, "merge": 0.21564, "skel": 0.74406}, "mean_gate": {"surface": 0.36275, "split": 0.19268, "merge": 0.21966, "skel": 0.28429}, "n_reads": 18853, "n_cand": 27355, "read_suppressed_frac": 0.1023, "dropout_reads_logged": 18853, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4238, "admitted_surface": 16759, "staged_surface": 0, "admitted_split": 21417, "staged_split": 0, "admitted_merge": 14957, "staged_merge": 0, "admitted_skel": 12403, "staged_skel": 0, "mean_count": 1281.9639892578125, "n_util_obs": 619, "util_count_spearman": 0.02079144329723607}, "probe_overhead_sec_total": 438.669, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.012791, "alpha": 0.03, "t": 3580.8} |
| {"event": "consolidate", "step": 4650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3618.4} |
| {"event": "train", "step": 4650, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.31249, "bpb": 1.89352, "chunk_loss": 0.030125, "F": 0.18262, "G": 0.18806, "chunks": 187.0, "bytes_per_chunk": 5.476, "boundary_entropy": 0.04846, "router_noise": 0.0, "grad_norm": 0.7354, "lr": 6.57e-05, "router_trainable": true, "sec_per_step": 0.758, "probe_sec": 0.058, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03849, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.01126, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.05556, "n_merge": 1, "mean_delta_merge": 0.22534, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 9.38364, "trunc_abs_frac": 0.59295, "mean_delta_full": 0.01001, "read_hit_rate": {"surface": 0.7379, "split": 0.9059, "merge": 0.20524, "skel": 0.74528}, "mean_gate": {"surface": 0.36047, "split": 0.19071, "merge": 0.21799, "skel": 0.28464}, "n_reads": 19145, "n_cand": 27662, "read_suppressed_frac": 0.103, "dropout_reads_logged": 19145, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4263, "admitted_surface": 16751, "staged_surface": 0, "admitted_split": 21324, "staged_split": 0, "admitted_merge": 15046, "staged_merge": 0, "admitted_skel": 12415, "staged_skel": 0, "mean_count": 1296.0308837890625, "n_util_obs": 622, "util_count_spearman": 0.016647023107923615}, "probe_overhead_sec_total": 444.378, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.011267, "alpha": 0.03, "t": 3618.7} |
| {"event": "consolidate", "step": 4700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3643.0} |
| {"event": "train", "step": 4700, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.36533, "bpb": 1.96975, "chunk_loss": 0.030137, "F": 0.18484, "G": 0.19003, "chunks": 189.28, "bytes_per_chunk": 5.41, "boundary_entropy": 0.0514, "router_noise": 0.0, "grad_norm": 0.7371, "lr": 6.32e-05, "router_trainable": true, "sec_per_step": 0.5, "probe_sec": 0.0627, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04323, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.01765, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.07655, "n_merge": 1, "mean_delta_merge": 0.08026, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 1.13487, "trunc_abs_frac": 1.23192, "mean_delta_full": 0.08447, "read_hit_rate": {"surface": 0.72946, "split": 0.88805, "merge": 0.21008, "skel": 0.74122}, "mean_gate": {"surface": 0.35794, "split": 0.18992, "merge": 0.21732, "skel": 0.28479}, "n_reads": 18847, "n_cand": 27646, "read_suppressed_frac": 0.1017, "dropout_reads_logged": 18847, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4296, "admitted_surface": 16751, "staged_surface": 0, "admitted_split": 21434, "staged_split": 0, "admitted_merge": 14975, "staged_merge": 0, "admitted_skel": 12376, "staged_skel": 0, "mean_count": 1310.0902099609375, "n_util_obs": 627, "util_count_spearman": 0.014766853497648504}, "probe_overhead_sec_total": 447.497, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.011051, "alpha": 0.03, "t": 3643.7} |
| {"event": "consolidate", "step": 4750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3668.5} |
| {"event": "train", "step": 4750, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38435, "bpb": 1.9972, "chunk_loss": 0.030157, "F": 0.18692, "G": 0.19205, "chunks": 191.41, "bytes_per_chunk": 5.35, "boundary_entropy": 0.04889, "router_noise": 0.0, "grad_norm": 0.7046, "lr": 6.08e-05, "router_trainable": true, "sec_per_step": 0.51, "probe_sec": 0.0831, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05576, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.04904, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": 0.00073, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.01757, "trunc_abs_frac": 0.04738, "mean_delta_full": -0.49249, "read_hit_rate": {"surface": 0.71902, "split": 0.8925, "merge": 0.19518, "skel": 0.7313}, "mean_gate": {"surface": 0.35672, "split": 0.18913, "merge": 0.21469, "skel": 0.28286}, "n_reads": 18906, "n_cand": 27948, "read_suppressed_frac": 0.1009, "dropout_reads_logged": 18906, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4317, "admitted_surface": 16761, "staged_surface": 0, "admitted_split": 21369, "staged_split": 0, "admitted_merge": 15043, "staged_merge": 0, "admitted_skel": 12363, "staged_skel": 0, "mean_count": 1324.186279296875, "n_util_obs": 633, "util_count_spearman": 0.002575239048867631}, "probe_overhead_sec_total": 450.607, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.011353, "alpha": 0.03, "t": 3669.2} |
| {"event": "consolidate", "step": 4800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3693.7} |
| {"event": "probe", "step": 4800, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 2922.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.167902, "cps": 22.828125, "mean_p": 0.173332, "boundary_entropy": 0.038615, "bpb": 1.860532}, "tr": {"chunks": 3413.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.188532, "cps": 26.664062, "mean_p": 0.179173, "boundary_entropy": 0.082444, "bpb": 2.031813}, "fi": {"chunks": 2990.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.15895, "cps": 23.359375, "mean_p": 0.162261, "boundary_entropy": 0.100819, "bpb": 2.181381}}, "bpb": {"en": 1.86053, "tr": 2.03181, "fi": 2.18138}, "gini_cpc": 0.038266, "gini_cps": 0.035103, "bpic": 6.083861, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.1228704307918151, "fi": 0.9466800248052688}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4346, "admitted_surface": 16803, "staged_surface": 0, "admitted_split": 21480, "staged_split": 0, "admitted_merge": 14873, "staged_merge": 0, "admitted_skel": 12380, "staged_skel": 0, "mean_count": 1337.885986328125, "n_util_obs": 646, "util_count_spearman": -0.0006537814921815302}, "t": 3694.8} |
| {"event": "eval", "step": 4800, "phase": 3, "bpb": {"en": 1.87849, "tr": 2.03865, "fi": 2.19379}, "bytes_per_chunk": {"en": 5.411, "tr": 5.465, "fi": 6.096}, "bpb_mean": 2.03698, "frozen_inventory": true, "t": 3696.6} |
| {"event": "train", "step": 4800, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.41496, "bpb": 2.04136, "chunk_loss": 0.030106, "F": 0.18201, "G": 0.18813, "chunks": 186.38, "bytes_per_chunk": 5.494, "boundary_entropy": 0.05534, "router_noise": 0.0, "grad_norm": 0.7381, "lr": 5.85e-05, "router_trainable": true, "sec_per_step": 0.578, "probe_sec": 0.0915, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07584, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.03973, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": -0.0092, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -0.09726, "trunc_abs_frac": 3.19149, "mean_delta_full": 0.15869, "read_hit_rate": {"surface": 0.69842, "split": 0.87647, "merge": 0.18661, "skel": 0.72484}, "mean_gate": {"surface": 0.34914, "split": 0.1852, "merge": 0.20996, "skel": 0.27632}, "n_reads": 18144, "n_cand": 27358, "read_suppressed_frac": 0.1004, "dropout_reads_logged": 18144, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4348, "admitted_surface": 16784, "staged_surface": 0, "admitted_split": 21470, "staged_split": 0, "admitted_merge": 14935, "staged_merge": 0, "admitted_skel": 12347, "staged_skel": 0, "mean_count": 1338.1658935546875, "n_util_obs": 646, "util_count_spearman": -0.0007495952647390536}, "probe_overhead_sec_total": 453.736, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.010283, "alpha": 0.03, "t": 3698.1} |
| {"event": "consolidate", "step": 4850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3722.3} |
| {"event": "train", "step": 4850, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32622, "bpb": 1.91333, "chunk_loss": 0.030146, "F": 0.18283, "G": 0.18868, "chunks": 187.22, "bytes_per_chunk": 5.47, "boundary_entropy": 0.04502, "router_noise": 0.0, "grad_norm": 0.7212, "lr": 5.62e-05, "router_trainable": true, "sec_per_step": 0.497, "probe_sec": 0.0726, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0452, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": 0.06125, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.65698, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": -0.11323, "trunc_abs_frac": 0.11323, "mean_delta_full": 0.14392, "read_hit_rate": {"surface": 0.74811, "split": 0.90062, "merge": 0.22541, "skel": 0.76253}, "mean_gate": {"surface": 0.35639, "split": 0.18286, "merge": 0.2103, "skel": 0.28099}, "n_reads": 19481, "n_cand": 27851, "read_suppressed_frac": 0.0992, "dropout_reads_logged": 19481, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4375, "admitted_surface": 16705, "staged_surface": 0, "admitted_split": 21415, "staged_split": 0, "admitted_merge": 14998, "staged_merge": 0, "admitted_skel": 12418, "staged_skel": 0, "mean_count": 1352.27001953125, "n_util_obs": 657, "util_count_spearman": 0.012343623769795259}, "probe_overhead_sec_total": 456.851, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.010018, "alpha": 0.03, "t": 3723.0} |
| {"event": "consolidate", "step": 4900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3751.9} |
| {"event": "train", "step": 4900, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.26747, "bpb": 1.82857, "chunk_loss": 0.030129, "F": 0.18399, "G": 0.19003, "chunks": 188.41, "bytes_per_chunk": 5.435, "boundary_entropy": 0.05026, "router_noise": 0.0, "grad_norm": 0.7335, "lr": 5.41e-05, "router_trainable": true, "sec_per_step": 0.588, "probe_sec": 0.0609, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04073, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.0912, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.15958, "n_merge": 1, "mean_delta_merge": -0.34656, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -1.0179, "trunc_abs_frac": 0.64852, "mean_delta_full": -0.16846, "read_hit_rate": {"surface": 0.7347, "split": 0.90228, "merge": 0.22733, "skel": 0.75119}, "mean_gate": {"surface": 0.35733, "split": 0.18789, "merge": 0.21592, "skel": 0.28083}, "n_reads": 19378, "n_cand": 27865, "read_suppressed_frac": 0.0977, "dropout_reads_logged": 19378, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4404, "admitted_surface": 16771, "staged_surface": 0, "admitted_split": 21370, "staged_split": 0, "admitted_merge": 14971, "staged_merge": 0, "admitted_skel": 12424, "staged_skel": 0, "mean_count": 1366.2978515625, "n_util_obs": 668, "util_count_spearman": 0.0046230544260989145}, "probe_overhead_sec_total": 460.162, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009782, "alpha": 0.03, "t": 3752.3} |
| {"event": "consolidate", "step": 4950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3780.5} |
| {"event": "train", "step": 4950, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35038, "bpb": 1.94818, "chunk_loss": 0.030153, "F": 0.18307, "G": 0.18927, "chunks": 187.47, "bytes_per_chunk": 5.462, "boundary_entropy": 0.04992, "router_noise": 0.0, "grad_norm": 0.7327, "lr": 5.2e-05, "router_trainable": true, "sec_per_step": 0.58, "probe_sec": 0.0565, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03759, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.00275, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.10754, "n_merge": 1, "mean_delta_merge": 0.01447, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.89539, "trunc_abs_frac": 0.91489, "mean_delta_full": -0.21041, "read_hit_rate": {"surface": 0.72046, "split": 0.8912, "merge": 0.2091, "skel": 0.73875}, "mean_gate": {"surface": 0.35636, "split": 0.18417, "merge": 0.21009, "skel": 0.28244}, "n_reads": 19031, "n_cand": 27884, "read_suppressed_frac": 0.1032, "dropout_reads_logged": 19031, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4431, "admitted_surface": 16734, "staged_surface": 0, "admitted_split": 21403, "staged_split": 0, "admitted_merge": 14966, "staged_merge": 0, "admitted_skel": 12433, "staged_skel": 0, "mean_count": 1380.4757080078125, "n_util_obs": 680, "util_count_spearman": -0.009314320481270394}, "probe_overhead_sec_total": 463.888, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.010544, "alpha": 0.03, "t": 3781.3} |
| {"event": "consolidate", "step": 5000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3806.6} |
| {"event": "probe", "step": 5000, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 2922.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.167902, "cps": 22.828125, "mean_p": 0.173032, "boundary_entropy": 0.036919, "bpb": 1.855645}, "tr": {"chunks": 3412.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.188477, "cps": 26.65625, "mean_p": 0.17856, "boundary_entropy": 0.079673, "bpb": 2.02418}, "fi": {"chunks": 2990.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.15895, "cps": 23.359375, "mean_p": 0.161086, "boundary_entropy": 0.096506, "bpb": 2.175186}}, "bpb": {"en": 1.85565, "tr": 2.02418, "fi": 2.17519}, "gini_cpc": 0.038199, "gini_cps": 0.035035, "bpic": 6.084513, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.122541432716576, "fi": 0.9466800248052688}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4454, "admitted_surface": 16764, "staged_surface": 0, "admitted_split": 21457, "staged_split": 0, "admitted_merge": 14865, "staged_merge": 0, "admitted_skel": 12450, "staged_skel": 0, "mean_count": 1394.291748046875, "n_util_obs": 687, "util_count_spearman": -0.009458268882243942}, "t": 3807.7} |
| {"event": "eval", "step": 5000, "phase": 3, "bpb": {"en": 1.8734, "tr": 2.03084, "fi": 2.18527}, "bytes_per_chunk": {"en": 5.418, "tr": 5.467, "fi": 6.093}, "bpb_mean": 2.02983, "frozen_inventory": true, "t": 3809.6} |
| {"event": "train", "step": 5000, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.29692, "bpb": 1.87106, "chunk_loss": 0.030124, "F": 0.18112, "G": 0.18683, "chunks": 185.47, "bytes_per_chunk": 5.521, "boundary_entropy": 0.05078, "router_noise": 0.0, "grad_norm": 0.7797, "lr": 5e-05, "router_trainable": true, "sec_per_step": 0.595, "probe_sec": 0.0599, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04189, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.04987, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.12482, "n_merge": 1, "mean_delta_merge": -0.15823, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 1.34568, "trunc_abs_frac": 1.37725, "mean_delta_full": -0.17853, "read_hit_rate": {"surface": 0.72878, "split": 0.8879, "merge": 0.22296, "skel": 0.75092}, "mean_gate": {"surface": 0.3558, "split": 0.18509, "merge": 0.20836, "skel": 0.28196}, "n_reads": 18958, "n_cand": 27558, "read_suppressed_frac": 0.1033, "dropout_reads_logged": 18958, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4454, "admitted_surface": 16735, "staged_surface": 0, "admitted_split": 21457, "staged_split": 0, "admitted_merge": 14936, "staged_merge": 0, "admitted_skel": 12408, "staged_skel": 0, "mean_count": 1394.5928955078125, "n_util_obs": 687, "util_count_spearman": -0.009548590044349188}, "probe_overhead_sec_total": 467.193, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.010876, "alpha": 0.03, "t": 3811.1} |
| {"event": "consolidate", "step": 5050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3837.0} |
| {"event": "train", "step": 5050, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38026, "bpb": 1.99129, "chunk_loss": 0.030111, "F": 0.17877, "G": 0.18453, "chunks": 183.06, "bytes_per_chunk": 5.594, "boundary_entropy": 0.05277, "router_noise": 0.0, "grad_norm": 0.7066, "lr": 4.81e-05, "router_trainable": true, "sec_per_step": 0.526, "probe_sec": 0.0583, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03858, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.04448, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": -0.2126, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 119.72727, "trunc_abs_frac": 119.72727, "mean_delta_full": -0.00134, "read_hit_rate": {"surface": 0.70254, "split": 0.87744, "merge": 0.19658, "skel": 0.71006}, "mean_gate": {"surface": 0.35124, "split": 0.1863, "merge": 0.20959, "skel": 0.27678}, "n_reads": 17957, "n_cand": 27006, "read_suppressed_frac": 0.0992, "dropout_reads_logged": 17957, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4476, "admitted_surface": 16768, "staged_surface": 0, "admitted_split": 21491, "staged_split": 0, "admitted_merge": 14825, "staged_merge": 0, "admitted_skel": 12452, "staged_skel": 0, "mean_count": 1408.62890625, "n_util_obs": 697, "util_count_spearman": -0.015816724091447645}, "probe_overhead_sec_total": 470.271, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.010993, "alpha": 0.03, "t": 3837.4} |
| {"event": "consolidate", "step": 5100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3862.7} |
| {"event": "train", "step": 5100, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32381, "bpb": 1.90985, "chunk_loss": 0.030107, "F": 0.18228, "G": 0.18748, "chunks": 186.66, "bytes_per_chunk": 5.486, "boundary_entropy": 0.05057, "router_noise": 0.0, "grad_norm": 0.7329, "lr": 4.63e-05, "router_trainable": true, "sec_per_step": 0.514, "probe_sec": 0.0598, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0404, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.29293, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": -0.24918, "n_merge": 1, "mean_delta_merge": 0.13843, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.64357, "trunc_abs_frac": 0.64357, "mean_delta_full": -3.64526, "read_hit_rate": {"surface": 0.72766, "split": 0.89291, "merge": 0.20308, "skel": 0.73583}, "mean_gate": {"surface": 0.34984, "split": 0.18428, "merge": 0.20564, "skel": 0.27541}, "n_reads": 18674, "n_cand": 27404, "read_suppressed_frac": 0.0989, "dropout_reads_logged": 18674, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4503, "admitted_surface": 16761, "staged_surface": 0, "admitted_split": 21454, "staged_split": 0, "admitted_merge": 14891, "staged_merge": 0, "admitted_skel": 12430, "staged_skel": 0, "mean_count": 1422.6671142578125, "n_util_obs": 699, "util_count_spearman": -0.012216458310855532}, "probe_overhead_sec_total": 473.41, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.012675, "alpha": 0.03, "t": 3863.1} |
| {"event": "consolidate", "step": 5150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3888.2} |
| {"event": "train", "step": 5150, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.36411, "bpb": 1.968, "chunk_loss": 0.0301, "F": 0.18173, "G": 0.18666, "chunks": 186.09, "bytes_per_chunk": 5.503, "boundary_entropy": 0.05149, "router_noise": 0.0, "grad_norm": 0.7846, "lr": 4.45e-05, "router_trainable": true, "sec_per_step": 0.518, "probe_sec": 0.061, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04168, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.12098, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": -0.00626, "n_merge": 1, "mean_delta_merge": -0.05444, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.43604, "trunc_abs_frac": 0.81027, "mean_delta_full": 0.3575, "read_hit_rate": {"surface": 0.70707, "split": 0.88096, "merge": 0.202, "skel": 0.72243}, "mean_gate": {"surface": 0.35586, "split": 0.18862, "merge": 0.21223, "skel": 0.27913}, "n_reads": 18205, "n_cand": 27195, "read_suppressed_frac": 0.0994, "dropout_reads_logged": 18205, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4526, "admitted_surface": 16768, "staged_surface": 0, "admitted_split": 21557, "staged_split": 0, "admitted_merge": 14809, "staged_merge": 0, "admitted_skel": 12402, "staged_skel": 0, "mean_count": 1436.683837890625, "n_util_obs": 705, "util_count_spearman": -0.006134932294932604}, "probe_overhead_sec_total": 476.554, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.010994, "alpha": 0.03, "t": 3889.0} |
| {"event": "consolidate", "step": 5200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3915.8} |
| {"event": "probe", "step": 5200, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 2920.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.167787, "cps": 22.8125, "mean_p": 0.17297, "boundary_entropy": 0.037016, "bpb": 1.846413}, "tr": {"chunks": 3417.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.188753, "cps": 26.695312, "mean_p": 0.179426, "boundary_entropy": 0.078597, "bpb": 2.015927}, "fi": {"chunks": 2991.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.159003, "cps": 23.367188, "mean_p": 0.162042, "boundary_entropy": 0.096672, "bpb": 2.167941}}, "bpb": {"en": 1.84641, "tr": 2.01593, "fi": 2.16794}, "gini_cpc": 0.038471, "gini_cps": 0.03552, "bpic": 6.081904, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.1249564137935208, "fi": 0.9476452680339322}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4550, "admitted_surface": 16761, "staged_surface": 0, "admitted_split": 21461, "staged_split": 0, "admitted_merge": 14842, "staged_merge": 0, "admitted_skel": 12472, "staged_skel": 0, "mean_count": 1450.4935302734375, "n_util_obs": 708, "util_count_spearman": -0.009816931826126897}, "t": 3917.0} |
| {"event": "eval", "step": 5200, "phase": 3, "bpb": {"en": 1.86528, "tr": 2.02244, "fi": 2.17849}, "bytes_per_chunk": {"en": 5.402, "tr": 5.467, "fi": 6.095}, "bpb_mean": 2.02207, "frozen_inventory": true, "t": 3918.8} |
| {"event": "train", "step": 5200, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32967, "bpb": 1.91831, "chunk_loss": 0.030097, "F": 0.17764, "G": 0.18393, "chunks": 181.91, "bytes_per_chunk": 5.629, "boundary_entropy": 0.05375, "router_noise": 0.0, "grad_norm": 0.7095, "lr": 4.29e-05, "router_trainable": true, "sec_per_step": 0.63, "probe_sec": 0.0615, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04232, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": 0.19644, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.02588, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.4601, "trunc_abs_frac": 0.50176, "mean_delta_full": 0.25381, "read_hit_rate": {"surface": 0.71722, "split": 0.88415, "merge": 0.18638, "skel": 0.73702}, "mean_gate": {"surface": 0.34886, "split": 0.18522, "merge": 0.21263, "skel": 0.27416}, "n_reads": 18081, "n_cand": 26856, "read_suppressed_frac": 0.1029, "dropout_reads_logged": 18081, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4551, "admitted_surface": 16744, "staged_surface": 0, "admitted_split": 21527, "staged_split": 0, "admitted_merge": 14842, "staged_merge": 0, "admitted_skel": 12423, "staged_skel": 0, "mean_count": 1450.771728515625, "n_util_obs": 708, "util_count_spearman": -0.009670251157531612}, "probe_overhead_sec_total": 479.73, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.010488, "alpha": 0.03, "t": 3920.5} |
| {"event": "consolidate", "step": 5250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3946.6} |
| {"event": "train", "step": 5250, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38163, "bpb": 1.99327, "chunk_loss": 0.030203, "F": 0.18518, "G": 0.19108, "chunks": 189.62, "bytes_per_chunk": 5.4, "boundary_entropy": 0.05479, "router_noise": 0.0, "grad_norm": 0.8119, "lr": 4.14e-05, "router_trainable": true, "sec_per_step": 0.537, "probe_sec": 0.0589, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04043, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.07117, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.01175, "n_merge": 1, "mean_delta_merge": -0.12213, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -36.3544, "trunc_abs_frac": 0.68655, "mean_delta_full": 0.0037, "read_hit_rate": {"surface": 0.72416, "split": 0.89289, "merge": 0.1982, "skel": 0.75046}, "mean_gate": {"surface": 0.35409, "split": 0.18302, "merge": 0.20974, "skel": 0.27997}, "n_reads": 19017, "n_cand": 27883, "read_suppressed_frac": 0.1025, "dropout_reads_logged": 19017, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4575, "admitted_surface": 16726, "staged_surface": 0, "admitted_split": 21467, "staged_split": 0, "admitted_merge": 14949, "staged_merge": 0, "admitted_skel": 12394, "staged_skel": 0, "mean_count": 1464.8292236328125, "n_util_obs": 716, "util_count_spearman": -0.004998771920596214}, "probe_overhead_sec_total": 482.856, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.011593, "alpha": 0.03, "t": 3947.3} |
| {"event": "consolidate", "step": 5300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3971.9} |
| {"event": "train", "step": 5300, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.31964, "bpb": 1.90383, "chunk_loss": 0.030102, "F": 0.18008, "G": 0.18543, "chunks": 184.41, "bytes_per_chunk": 5.553, "boundary_entropy": 0.04808, "router_noise": 0.0, "grad_norm": 0.7462, "lr": 3.99e-05, "router_trainable": true, "sec_per_step": 0.499, "probe_sec": 0.0751, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05215, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.04225, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 4, "mean_delta_split": 0.19652, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 1.38281, "trunc_abs_frac": 0.43497, "mean_delta_full": 0.15729, "read_hit_rate": {"surface": 0.72261, "split": 0.88593, "merge": 0.19942, "skel": 0.73034}, "mean_gate": {"surface": 0.35301, "split": 0.18486, "merge": 0.20884, "skel": 0.27772}, "n_reads": 18319, "n_cand": 27089, "read_suppressed_frac": 0.0979, "dropout_reads_logged": 18319, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4605, "admitted_surface": 16718, "staged_surface": 0, "admitted_split": 21563, "staged_split": 0, "admitted_merge": 14828, "staged_merge": 0, "admitted_skel": 12427, "staged_skel": 0, "mean_count": 1478.9208984375, "n_util_obs": 722, "util_count_spearman": -0.0047890948371788426}, "probe_overhead_sec_total": 486.376, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.012578, "alpha": 0.03, "t": 3972.3} |
| {"event": "consolidate", "step": 5350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3996.9} |
| {"event": "train", "step": 5350, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.41575, "bpb": 2.04249, "chunk_loss": 0.030146, "F": 0.18387, "G": 0.1893, "chunks": 188.28, "bytes_per_chunk": 5.439, "boundary_entropy": 0.04996, "router_noise": 0.0, "grad_norm": 0.7888, "lr": 3.86e-05, "router_trainable": true, "sec_per_step": 0.499, "probe_sec": 0.053, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03562, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.08778, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.11328, "n_merge": 2, "mean_delta_merge": -0.4431, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.31806, "trunc_abs_frac": 0.68774, "mean_delta_full": 0.03912, "read_hit_rate": {"surface": 0.69381, "split": 0.87715, "merge": 0.19275, "skel": 0.71284}, "mean_gate": {"surface": 0.34476, "split": 0.18107, "merge": 0.20716, "skel": 0.26863}, "n_reads": 18194, "n_cand": 27518, "read_suppressed_frac": 0.1016, "dropout_reads_logged": 18194, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4631, "admitted_surface": 16802, "staged_surface": 0, "admitted_split": 21546, "staged_split": 0, "admitted_merge": 14775, "staged_merge": 0, "admitted_skel": 12413, "staged_skel": 0, "mean_count": 1493.09228515625, "n_util_obs": 734, "util_count_spearman": -0.012617677463201114}, "probe_overhead_sec_total": 489.934, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.011323, "alpha": 0.03, "t": 3997.2} |
| {"event": "consolidate", "step": 5400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4022.3} |
| {"event": "probe", "step": 5400, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 2921.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.167845, "cps": 22.820312, "mean_p": 0.172784, "boundary_entropy": 0.036263, "bpb": 1.842523}, "tr": {"chunks": 3412.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.188477, "cps": 26.65625, "mean_p": 0.17798, "boundary_entropy": 0.077828, "bpb": 2.009869}, "fi": {"chunks": 2990.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.15895, "cps": 23.359375, "mean_p": 0.160438, "boundary_entropy": 0.093683, "bpb": 2.160795}}, "bpb": {"en": 1.84252, "tr": 2.00987, "fi": 2.1608}, "gini_cpc": 0.038203, "gini_cps": 0.03511, "bpic": 6.085166, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.1229257331043596, "fi": 0.9470041193019497}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4655, "admitted_surface": 16777, "staged_surface": 0, "admitted_split": 21557, "staged_split": 0, "admitted_merge": 14783, "staged_merge": 0, "admitted_skel": 12419, "staged_skel": 0, "mean_count": 1506.9385986328125, "n_util_obs": 744, "util_count_spearman": -0.00319591071683583}, "t": 4023.4} |
| {"event": "eval", "step": 5400, "phase": 3, "bpb": {"en": 1.86023, "tr": 2.01744, "fi": 2.17004}, "bytes_per_chunk": {"en": 5.42, "tr": 5.473, "fi": 6.091}, "bpb_mean": 2.0159, "frozen_inventory": true, "t": 4025.3} |
| {"event": "train", "step": 5400, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.3015, "bpb": 1.87767, "chunk_loss": 0.030155, "F": 0.1861, "G": 0.191, "chunks": 190.56, "bytes_per_chunk": 5.374, "boundary_entropy": 0.04409, "router_noise": 0.0, "grad_norm": 0.6957, "lr": 3.73e-05, "router_trainable": true, "sec_per_step": 0.592, "probe_sec": 0.0637, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04459, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": 0.07341, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.02655, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 1.61188, "trunc_abs_frac": 1.61188, "mean_delta_full": 0.15002, "read_hit_rate": {"surface": 0.75354, "split": 0.9035, "merge": 0.2181, "skel": 0.76236}, "mean_gate": {"surface": 0.35689, "split": 0.18239, "merge": 0.21089, "skel": 0.27813}, "n_reads": 19685, "n_cand": 28127, "read_suppressed_frac": 0.1011, "dropout_reads_logged": 19685, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4655, "admitted_surface": 16690, "staged_surface": 0, "admitted_split": 21497, "staged_split": 0, "admitted_merge": 14928, "staged_merge": 0, "admitted_skel": 12421, "staged_skel": 0, "mean_count": 1507.244140625, "n_util_obs": 744, "util_count_spearman": -0.003202277491092753}, "probe_overhead_sec_total": 493.1, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.010087, "alpha": 0.03, "t": 4026.8} |
| {"event": "consolidate", "step": 5450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4055.0} |
| {"event": "train", "step": 5450, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.41128, "bpb": 2.03605, "chunk_loss": 0.030153, "F": 0.18683, "G": 0.19225, "chunks": 191.31, "bytes_per_chunk": 5.352, "boundary_entropy": 0.05278, "router_noise": 0.0, "grad_norm": 0.775, "lr": 3.62e-05, "router_trainable": true, "sec_per_step": 0.571, "probe_sec": 0.0571, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03893, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.08543, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 1.24237, "trunc_abs_frac": 0.53582, "mean_delta_full": 0.0533, "read_hit_rate": {"surface": 0.70246, "split": 0.88976, "merge": 0.19495, "skel": 0.71226}, "mean_gate": {"surface": 0.35101, "split": 0.18561, "merge": 0.20897, "skel": 0.27736}, "n_reads": 18690, "n_cand": 27946, "read_suppressed_frac": 0.1003, "dropout_reads_logged": 18690, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4683, "admitted_surface": 16742, "staged_surface": 0, "admitted_split": 21537, "staged_split": 0, "admitted_merge": 14833, "staged_merge": 0, "admitted_skel": 12424, "staged_skel": 0, "mean_count": 1521.2626953125, "n_util_obs": 751, "util_count_spearman": -0.003366433455625598}, "probe_overhead_sec_total": 496.649, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.010837, "alpha": 0.03, "t": 4055.4} |
| {"event": "consolidate", "step": 5500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4083.1} |
| {"event": "train", "step": 5500, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.3408, "bpb": 1.93436, "chunk_loss": 0.030199, "F": 0.18802, "G": 0.19266, "chunks": 192.53, "bytes_per_chunk": 5.319, "boundary_entropy": 0.04789, "router_noise": 0.0, "grad_norm": 0.7593, "lr": 3.51e-05, "router_trainable": true, "sec_per_step": 0.562, "probe_sec": 0.0651, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04469, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": -0.04237, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.23474, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -0.11738, "trunc_abs_frac": 0.90921, "mean_delta_full": -0.10132, "read_hit_rate": {"surface": 0.72328, "split": 0.89029, "merge": 0.22208, "skel": 0.73894}, "mean_gate": {"surface": 0.35187, "split": 0.18486, "merge": 0.20799, "skel": 0.27583}, "n_reads": 19232, "n_cand": 28123, "read_suppressed_frac": 0.0956, "dropout_reads_logged": 19232, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4710, "admitted_surface": 16717, "staged_surface": 0, "admitted_split": 21556, "staged_split": 0, "admitted_merge": 14885, "staged_merge": 0, "admitted_skel": 12378, "staged_skel": 0, "mean_count": 1535.3128662109375, "n_util_obs": 757, "util_count_spearman": -0.0024948148366171112}, "probe_overhead_sec_total": 500.089, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.011919, "alpha": 0.03, "t": 4083.5} |
| {"event": "consolidate", "step": 5550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4113.6} |
| {"event": "train", "step": 5550, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.31196, "bpb": 1.89276, "chunk_loss": 0.030116, "F": 0.18213, "G": 0.18743, "chunks": 186.5, "bytes_per_chunk": 5.491, "boundary_entropy": 0.04957, "router_noise": 0.0, "grad_norm": 0.7509, "lr": 3.41e-05, "router_trainable": true, "sec_per_step": 0.612, "probe_sec": 0.0792, "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": 10, "mean_delta_loo": -0.07453, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.22998, "n_merge": 1, "mean_delta_merge": -0.13928, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.44863, "trunc_abs_frac": 0.58078, "mean_delta_full": 0.26556, "read_hit_rate": {"surface": 0.72423, "split": 0.89286, "merge": 0.20054, "skel": 0.73217}, "mean_gate": {"surface": 0.35273, "split": 0.18453, "merge": 0.20767, "skel": 0.27327}, "n_reads": 18704, "n_cand": 27480, "read_suppressed_frac": 0.0998, "dropout_reads_logged": 18704, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4741, "admitted_surface": 16713, "staged_surface": 0, "admitted_split": 21583, "staged_split": 0, "admitted_merge": 14848, "staged_merge": 0, "admitted_skel": 12392, "staged_skel": 0, "mean_count": 1549.539306640625, "n_util_obs": 762, "util_count_spearman": -0.005992772809588668}, "probe_overhead_sec_total": 503.868, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.01009, "alpha": 0.03, "t": 4114.1} |
| {"event": "consolidate", "step": 5600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4143.5} |
| {"event": "probe", "step": 5600, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 2921.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.167845, "cps": 22.820312, "mean_p": 0.172758, "boundary_entropy": 0.035875, "bpb": 1.839195}, "tr": {"chunks": 3412.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.188477, "cps": 26.65625, "mean_p": 0.178958, "boundary_entropy": 0.077993, "bpb": 2.000983}, "fi": {"chunks": 2996.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.159269, "cps": 23.40625, "mean_p": 0.160954, "boundary_entropy": 0.093818, "bpb": 2.154726}}, "bpb": {"en": 1.8392, "tr": 2.00098, "fi": 2.15473}, "gini_cpc": 0.037767, "gini_cps": 0.035088, "bpic": 6.081252, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.1229257331043596, "fi": 0.948904462016268}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4764, "admitted_surface": 16674, "staged_surface": 0, "admitted_split": 21467, "staged_split": 0, "admitted_merge": 15026, "staged_merge": 0, "admitted_skel": 12369, "staged_skel": 0, "mean_count": 1563.4429931640625, "n_util_obs": 770, "util_count_spearman": -0.016598022878587467}, "t": 4144.6} |
| {"event": "eval", "step": 5600, "phase": 3, "bpb": {"en": 1.85726, "tr": 2.00852, "fi": 2.16495}, "bytes_per_chunk": {"en": 5.417, "tr": 5.468, "fi": 6.088}, "bpb_mean": 2.01024, "frozen_inventory": true, "t": 4146.5} |
| {"event": "train", "step": 5600, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35965, "bpb": 1.96157, "chunk_loss": 0.030085, "F": 0.17932, "G": 0.18443, "chunks": 183.62, "bytes_per_chunk": 5.577, "boundary_entropy": 0.04355, "router_noise": 0.0, "grad_norm": 0.7389, "lr": 3.33e-05, "router_trainable": true, "sec_per_step": 0.677, "probe_sec": 0.0566, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03849, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.00622, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -1.05466, "n_merge": 1, "mean_delta_merge": -0.10291, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.58181, "trunc_abs_frac": 1.83234, "mean_delta_full": -0.10818, "read_hit_rate": {"surface": 0.70927, "split": 0.89195, "merge": 0.18909, "skel": 0.73741}, "mean_gate": {"surface": 0.35687, "split": 0.18233, "merge": 0.20838, "skel": 0.27762}, "n_reads": 18441, "n_cand": 27281, "read_suppressed_frac": 0.0982, "dropout_reads_logged": 18441, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4764, "admitted_surface": 16741, "staged_surface": 0, "admitted_split": 21573, "staged_split": 0, "admitted_merge": 14811, "staged_merge": 0, "admitted_skel": 12411, "staged_skel": 0, "mean_count": 1563.7181396484375, "n_util_obs": 771, "util_count_spearman": -0.013419097735490193}, "probe_overhead_sec_total": 507.312, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.011391, "alpha": 0.03, "t": 4147.9} |
| {"event": "consolidate", "step": 5650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4173.1} |
| {"event": "train", "step": 5650, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.27171, "bpb": 1.83468, "chunk_loss": 0.030112, "F": 0.17905, "G": 0.18463, "chunks": 183.34, "bytes_per_chunk": 5.585, "boundary_entropy": 0.04632, "router_noise": 0.0, "grad_norm": 0.7126, "lr": 3.25e-05, "router_trainable": true, "sec_per_step": 0.522, "probe_sec": 0.5033, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04176, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": -0.01667, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.05817, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.64879, "trunc_abs_frac": 3.46818, "mean_delta_full": -0.1286, "read_hit_rate": {"surface": 0.74464, "split": 0.90296, "merge": 0.2192, "skel": 0.76155}, "mean_gate": {"surface": 0.35077, "split": 0.17883, "merge": 0.20471, "skel": 0.27389}, "n_reads": 19149, "n_cand": 27390, "read_suppressed_frac": 0.0992, "dropout_reads_logged": 19149, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4784, "admitted_surface": 16677, "staged_surface": 0, "admitted_split": 21544, "staged_split": 0, "admitted_merge": 14874, "staged_merge": 0, "admitted_skel": 12441, "staged_skel": 0, "mean_count": 1577.77490234375, "n_util_obs": 781, "util_count_spearman": -0.005334998898880316}, "probe_overhead_sec_total": 510.977, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.010738, "alpha": 0.03, "t": 4174.0} |
| {"event": "consolidate", "step": 5700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4200.2} |
| {"event": "train", "step": 5700, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32706, "bpb": 1.91455, "chunk_loss": 0.030119, "F": 0.18018, "G": 0.18556, "chunks": 184.5, "bytes_per_chunk": 5.55, "boundary_entropy": 0.04399, "router_noise": 0.0, "grad_norm": 0.7317, "lr": 3.18e-05, "router_trainable": true, "sec_per_step": 0.537, "probe_sec": 0.0765, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04837, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.01923, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.19104, "n_merge": 1, "mean_delta_merge": 0.00708, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": -0.12942, "trunc_abs_frac": 0.12942, "mean_delta_full": 0.34143, "read_hit_rate": {"surface": 0.72241, "split": 0.89404, "merge": 0.19795, "skel": 0.7416}, "mean_gate": {"surface": 0.35315, "split": 0.18058, "merge": 0.20751, "skel": 0.27294}, "n_reads": 18557, "n_cand": 27227, "read_suppressed_frac": 0.0995, "dropout_reads_logged": 18557, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4812, "admitted_surface": 16738, "staged_surface": 0, "admitted_split": 21594, "staged_split": 0, "admitted_merge": 14787, "staged_merge": 0, "admitted_skel": 12417, "staged_skel": 0, "mean_count": 1591.9219970703125, "n_util_obs": 789, "util_count_spearman": -0.011925965868788576}, "probe_overhead_sec_total": 514.432, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.011327, "alpha": 0.03, "t": 4200.8} |
| {"event": "consolidate", "step": 5750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4226.8} |
| {"event": "train", "step": 5750, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.36092, "bpb": 1.96339, "chunk_loss": 0.030159, "F": 0.18454, "G": 0.18987, "chunks": 188.97, "bytes_per_chunk": 5.419, "boundary_entropy": 0.04863, "router_noise": 0.0, "grad_norm": 0.7574, "lr": 3.13e-05, "router_trainable": true, "sec_per_step": 0.527, "probe_sec": 0.0563, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03799, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.02379, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": 0.07336, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.06764, "trunc_abs_frac": 0.7984, "mean_delta_full": -0.15204, "read_hit_rate": {"surface": 0.71072, "split": 0.88598, "merge": 0.20692, "skel": 0.72072}, "mean_gate": {"surface": 0.35489, "split": 0.1844, "merge": 0.20656, "skel": 0.27435}, "n_reads": 18576, "n_cand": 27591, "read_suppressed_frac": 0.1009, "dropout_reads_logged": 18576, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4836, "admitted_surface": 16747, "staged_surface": 0, "admitted_split": 21635, "staged_split": 0, "admitted_merge": 14709, "staged_merge": 0, "admitted_skel": 12445, "staged_skel": 0, "mean_count": 1606.034423828125, "n_util_obs": 792, "util_count_spearman": -0.0018150644223773815}, "probe_overhead_sec_total": 517.741, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.01151, "alpha": 0.03, "t": 4227.2} |
| {"event": "consolidate", "step": 5800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4252.8} |
| {"event": "probe", "step": 5800, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 2921.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.167845, "cps": 22.820312, "mean_p": 0.172679, "boundary_entropy": 0.035317, "bpb": 1.833639}, "tr": {"chunks": 3412.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.188477, "cps": 26.65625, "mean_p": 0.178748, "boundary_entropy": 0.076616, "bpb": 1.995506}, "fi": {"chunks": 2997.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.159322, "cps": 23.414062, "mean_p": 0.160398, "boundary_entropy": 0.090981, "bpb": 2.148365}}, "bpb": {"en": 1.83364, "tr": 1.99551, "fi": 2.14837}, "gini_cpc": 0.037694, "gini_cps": 0.035084, "bpic": 6.0806, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.1229257331043596, "fi": 0.9492211858019878}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4869, "admitted_surface": 16709, "staged_surface": 0, "admitted_split": 21722, "staged_split": 0, "admitted_merge": 14638, "staged_merge": 0, "admitted_skel": 12467, "staged_skel": 0, "mean_count": 1619.831787109375, "n_util_obs": 799, "util_count_spearman": 0.0031064089117460426}, "t": 4254.0} |
| {"event": "eval", "step": 5800, "phase": 3, "bpb": {"en": 1.85256, "tr": 2.00389, "fi": 2.15958}, "bytes_per_chunk": {"en": 5.423, "tr": 5.473, "fi": 6.074}, "bpb_mean": 2.00534, "frozen_inventory": true, "t": 4255.9} |
| {"event": "train", "step": 5800, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32601, "bpb": 1.91302, "chunk_loss": 0.030285, "F": 0.1893, "G": 0.1945, "chunks": 193.84, "bytes_per_chunk": 5.283, "boundary_entropy": 0.04275, "router_noise": 0.0, "grad_norm": 0.7746, "lr": 3.08e-05, "router_trainable": true, "sec_per_step": 0.604, "probe_sec": 0.0867, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04202, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": -0.03622, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.06659, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -0.30311, "trunc_abs_frac": 4.61183, "mean_delta_full": 0.09424, "read_hit_rate": {"surface": 0.74688, "split": 0.90134, "merge": 0.24304, "skel": 0.75106}, "mean_gate": {"surface": 0.35107, "split": 0.17976, "merge": 0.20463, "skel": 0.2754}, "n_reads": 19607, "n_cand": 28082, "read_suppressed_frac": 0.0952, "dropout_reads_logged": 19607, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4869, "admitted_surface": 16680, "staged_surface": 0, "admitted_split": 21638, "staged_split": 0, "admitted_merge": 14793, "staged_merge": 0, "admitted_skel": 12425, "staged_skel": 0, "mean_count": 1620.1414794921875, "n_util_obs": 801, "util_count_spearman": 0.000968680837334109}, "probe_overhead_sec_total": 520.932, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.011911, "alpha": 0.03, "t": 4257.4} |
| {"event": "consolidate", "step": 5850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4283.1} |
| {"event": "train", "step": 5850, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.3072, "bpb": 1.88589, "chunk_loss": 0.030069, "F": 0.17618, "G": 0.18178, "chunks": 180.41, "bytes_per_chunk": 5.676, "boundary_entropy": 0.04678, "router_noise": 0.0, "grad_norm": 0.7209, "lr": 3.05e-05, "router_trainable": true, "sec_per_step": 0.521, "probe_sec": 0.0552, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0373, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": 0.09063, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": -0.07501, "n_merge": 3, "mean_delta_merge": 0.13128, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.49646, "trunc_abs_frac": 0.47129, "mean_delta_full": 0.32306, "read_hit_rate": {"surface": 0.7227, "split": 0.88757, "merge": 0.20582, "skel": 0.7332}, "mean_gate": {"surface": 0.35093, "split": 0.17854, "merge": 0.2047, "skel": 0.27003}, "n_reads": 18242, "n_cand": 26812, "read_suppressed_frac": 0.1005, "dropout_reads_logged": 18242, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4889, "admitted_surface": 16705, "staged_surface": 0, "admitted_split": 21709, "staged_split": 0, "admitted_merge": 14645, "staged_merge": 0, "admitted_skel": 12477, "staged_skel": 0, "mean_count": 1634.2188720703125, "n_util_obs": 811, "util_count_spearman": -0.002948218723967244}, "probe_overhead_sec_total": 524.587, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.010647, "alpha": 0.03, "t": 4283.4} |
| {"event": "consolidate", "step": 5900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4309.3} |
| {"event": "train", "step": 5900, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.30734, "bpb": 1.8861, "chunk_loss": 0.030074, "F": 0.17587, "G": 0.1821, "chunks": 180.09, "bytes_per_chunk": 5.686, "boundary_entropy": 0.05137, "router_noise": 0.0, "grad_norm": 0.7367, "lr": 3.02e-05, "router_trainable": true, "sec_per_step": 0.535, "probe_sec": 0.058, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03904, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.11817, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": 0.16707, "n_merge": 2, "mean_delta_merge": 0.11536, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 1.11591, "trunc_abs_frac": 1.01809, "mean_delta_full": 0.0932, "read_hit_rate": {"surface": 0.71244, "split": 0.8891, "merge": 0.19003, "skel": 0.73184}, "mean_gate": {"surface": 0.34701, "split": 0.18033, "merge": 0.20462, "skel": 0.26891}, "n_reads": 17984, "n_cand": 26670, "read_suppressed_frac": 0.1039, "dropout_reads_logged": 17984, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4913, "admitted_surface": 16719, "staged_surface": 0, "admitted_split": 21669, "staged_split": 0, "admitted_merge": 14692, "staged_merge": 0, "admitted_skel": 12456, "staged_skel": 0, "mean_count": 1648.375732421875, "n_util_obs": 822, "util_count_spearman": -0.008809966635705295}, "probe_overhead_sec_total": 527.746, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.011527, "alpha": 0.03, "t": 4310.2} |
| {"event": "consolidate", "step": 5950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4336.1} |
| {"event": "train", "step": 5950, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.43986, "bpb": 2.07728, "chunk_loss": 0.030167, "F": 0.1853, "G": 0.19142, "chunks": 189.75, "bytes_per_chunk": 5.397, "boundary_entropy": 0.05476, "router_noise": 0.0, "grad_norm": 0.8146, "lr": 3.01e-05, "router_trainable": true, "sec_per_step": 0.532, "probe_sec": 0.0565, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03841, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.01666, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 7, "trunc_ratio_sum": 0.62439, "trunc_abs_frac": 4.1375, "mean_delta_full": 0.06431, "read_hit_rate": {"surface": 0.67384, "split": 0.87021, "merge": 0.18259, "skel": 0.68235}, "mean_gate": {"surface": 0.34213, "split": 0.18428, "merge": 0.20294, "skel": 0.26755}, "n_reads": 17540, "n_cand": 27207, "read_suppressed_frac": 0.0976, "dropout_reads_logged": 17540, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4932, "admitted_surface": 16766, "staged_surface": 0, "admitted_split": 21684, "staged_split": 0, "admitted_merge": 14651, "staged_merge": 0, "admitted_skel": 12435, "staged_skel": 0, "mean_count": 1662.3575439453125, "n_util_obs": 829, "util_count_spearman": -0.012289774287363938}, "probe_overhead_sec_total": 530.979, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.013359, "alpha": 0.03, "t": 4336.8} |
| {"event": "consolidate", "step": 6000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4362.3} |
| {"event": "train", "step": 5999, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.30314, "bpb": 1.88003, "chunk_loss": 0.030083, "F": 0.17938, "G": 0.18484, "chunks": 183.69, "bytes_per_chunk": 5.575, "boundary_entropy": 0.04416, "router_noise": 0.0, "grad_norm": 0.6959, "lr": 3e-05, "router_trainable": true, "sec_per_step": 0.51, "probe_sec": 0.0564, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03734, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.01568, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -0.18423, "trunc_abs_frac": 0.35975, "mean_delta_full": -0.1027, "read_hit_rate": {"surface": 0.73418, "split": 0.90063, "merge": 0.20519, "skel": 0.75737}, "mean_gate": {"surface": 0.3547, "split": 0.17957, "merge": 0.20269, "skel": 0.27341}, "n_reads": 18879, "n_cand": 27293, "read_suppressed_frac": 0.0975, "dropout_reads_logged": 18879, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4960, "admitted_surface": 16674, "staged_surface": 0, "admitted_split": 21609, "staged_split": 0, "admitted_merge": 14811, "staged_merge": 0, "admitted_skel": 12442, "staged_skel": 0, "mean_count": 1676.1943359375, "n_util_obs": 834, "util_count_spearman": -0.011859470202995654}, "probe_overhead_sec_total": 534.16, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.012896, "alpha": 0.03, "t": 4362.3} |
| {"event": "probe", "step": 6000, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 2921.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.167845, "cps": 22.820312, "mean_p": 0.172694, "boundary_entropy": 0.035198, "bpb": 1.830511}, "tr": {"chunks": 3412.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.188477, "cps": 26.65625, "mean_p": 0.178959, "boundary_entropy": 0.075538, "bpb": 1.991336}, "fi": {"chunks": 2995.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.159215, "cps": 23.398438, "mean_p": 0.160912, "boundary_entropy": 0.091789, "bpb": 2.143726}}, "bpb": {"en": 1.83051, "tr": 1.99134, "fi": 2.14373}, "gini_cpc": 0.03784, "gini_cps": 0.035091, "bpic": 6.081904, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.1229257331043596, "fi": 0.9485877382305483}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4960, "admitted_surface": 16674, "staged_surface": 0, "admitted_split": 21609, "staged_split": 0, "admitted_merge": 14811, "staged_merge": 0, "admitted_skel": 12442, "staged_skel": 0, "mean_count": 1676.1943359375, "n_util_obs": 834, "util_count_spearman": -0.011859470202995654}, "t": 4363.4} |
| {"event": "eval", "step": 6000, "phase": 3, "bpb": {"en": 1.84908, "tr": 1.99998, "fi": 2.15445}, "bytes_per_chunk": {"en": 5.422, "tr": 5.473, "fi": 6.08}, "bpb_mean": 2.00117, "frozen_inventory": true, "t": 4365.2} |
| {"event": "reps_dumped", "dirs": ["/local/scratch/sas245/beetle-hnet/runs/ali_mem-mixed_s4/probe/reps/th0.30", "/local/scratch/sas245/beetle-hnet/runs/ali_mem-mixed_s4/probe/reps/th0.40", "/local/scratch/sas245/beetle-hnet/runs/ali_mem-mixed_s4/probe/reps/th0.50", "/local/scratch/sas245/beetle-hnet/runs/ali_mem-mixed_s4/probe/reps/th0.60", "/local/scratch/sas245/beetle-hnet/runs/ali_mem-mixed_s4/probe/reps/th0.70"], "t": 4367.7} |
| {"event": "final", "run": "ali_mem-mixed_s4", "arch": "ali", "objective": "ratio", "condition": "A", "step": 6000, "params": 22785088, "seed": 4, "per_lang": {"en": {"cpc": 0.16784462449003046, "cps": 22.8203125, "bpb": 1.8551285492300558, "bytes_per_chunk": 5.420790338923468, "reads_per_chunk": 3.3457784486858704}, "tr": {"cpc": 0.18847704800309342, "cps": 26.65625, "bpb": 1.9915023104697922, "bytes_per_chunk": 5.475878635960102, "reads_per_chunk": 2.4824377252075824}, "fi": {"cpc": 0.15921535271915369, "cps": 23.3984375, "bpb": 2.175575309441509, "bytes_per_chunk": 6.055602397810093, "reads_per_chunk": 2.421176035759249}}, "bpb_mean": 2.0074020563804527, "gini_cpc": 0.03783975912870141, "gini_cps": 0.03509148084619795, "bpic": 6.081903945111493, "mean_boundary_entropy": 0.04415855556726456, "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": 534.1603045463562, "probe_overhead_seq_frac": 0.4999166666666667, "wall_sec": 4371.956134080887, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4960, "admitted_surface": 16674, "staged_surface": 0, "admitted_split": 21609, "staged_split": 0, "admitted_merge": 14811, "staged_merge": 0, "admitted_skel": 12442, "staged_skel": 0, "mean_count": 1676.1943359375, "n_util_obs": 834, "util_count_spearman": -0.011859470202995654}, "ali_params": 268096, "ali_config": {"enabled": true, "budget_B": 65536, "staging_capacity": 262144, "policy": "frequency", "value_mode": "learned", "candidate_types": ["surface", "split", "merge", "skel"], "gate": "vector", "retrieval": true, "memory_scope": "persistent", "read_dropout": 0.1, "write_rule": "mean", "eta_min": 0.0, "consolidate_every": 50, "probe_seqs": 2, "probe_reps": 8, "horizon": 32, "predictor": "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": -0.011859470202995654, "paired_vs_dropout": {"n": 2058, "pearson": -0.003318013488018906, "spearman": -0.011604745352086821, "mean_abs_discrepancy": 0.47076502442359924, "mean_paired": 0.022626103833317757, "mean_dropout": 0.05205170437693596}, "predictor_calibration": {"n": 5497, "all": {"n": 5497, "pearson": 0.10150596513511552, "spearman": 0.10327474362629753, "mae": 0.09897452592849731, "reliability": [{"pred": -0.012804780155420303, "obs": 0.0036487141624093056, "n": 550}, {"pred": -0.004629538394510746, "obs": -0.002714421832934022, "n": 550}, {"pred": -0.0007536947377957404, "obs": 0.017434049397706985, "n": 549}, {"pred": 0.001957725500687957, "obs": 0.009500621818006039, "n": 550}, {"pred": 0.004199643153697252, "obs": 0.020444650202989578, "n": 549}, {"pred": 0.006369161885231733, "obs": 0.014687125571072102, "n": 550}, {"pred": 0.008665241301059723, "obs": 0.019541088491678238, "n": 550}, {"pred": 0.011964822188019753, "obs": 0.022163202986121178, "n": 549}, {"pred": 0.01715991459786892, "obs": 0.03169619292020798, "n": 550}, {"pred": 0.028791850432753563, "obs": 0.051099613308906555, "n": 550}]}, "admitted": {"n": 5497, "pearson": 0.10150596513511552, "spearman": 0.10327474362629753, "mae": 0.09897452592849731, "reliability": [{"pred": -0.012804780155420303, "obs": 0.0036487141624093056, "n": 550}, {"pred": -0.004629538394510746, "obs": -0.002714421832934022, "n": 550}, {"pred": -0.0007536947377957404, "obs": 0.017434049397706985, "n": 549}, {"pred": 0.001957725500687957, "obs": 0.009500621818006039, "n": 550}, {"pred": 0.004199643153697252, "obs": 0.020444650202989578, "n": 549}, {"pred": 0.006369161885231733, "obs": 0.014687125571072102, "n": 550}, {"pred": 0.008665241301059723, "obs": 0.019541088491678238, "n": 550}, {"pred": 0.011964822188019753, "obs": 0.022163202986121178, "n": 549}, {"pred": 0.01715991459786892, "obs": 0.03169619292020798, "n": 550}, {"pred": 0.028791850432753563, "obs": 0.051099613308906555, "n": 550}]}, "staged": {"n": 0, "pearson": NaN, "spearman": NaN, "mae": NaN}}, "t": 4372.0} |
| {"event": "hub_done", "repo": "AdaptiveChunking/ali_mem-mixed_s4", "pushed": [0, 1, 2, 4, 6, 8, 16, 32, 64, 128, 256, 512, 1000, 1024, 2048, 3000, 4096, 6000], "failed": [], "ok": true, "t": 4389.7} |
|
|
