| {"event": "hub", "repo": "AdaptiveChunking/ali_mem-merge_s2", "keep_steps": [0, 1, 2, 4, 6, 8, 16, 32, 64, 128, 256, 512, 1000, 1024, 2048, 3000, 4096, 6000], "t": 2.1} |
| {"event": "start", "run": "ali_mem-merge_s2", "arch": "ali", "objective": "ratio", "condition": "A", "params": 22785088, "langs": ["en", "tr", "fi"], "premium": {"en": 1.0, "tr": 1.12, "fi": 1.104}, "schedule": {"tau1": 0.4, "tau2": 0.7, "p1_objective": "global", "p2_objective": "global", "p3_objective": "global", "freeze_in_p3": false, "freeze_scope": "router"}, "corpus": "/local/scratch/sas245/beetle-hnet/scratch/data/tier1_corpus", "premium_kind": "flores_per_sent", "doc_filtered": {}, "weights": {"en": 0.74, "tr": 0.1536, "fi": 0.1064}, "device": "cuda", "ali_params": 268096, "ali_config": {"enabled": true, "budget_B": 65536, "staging_capacity": 262144, "policy": "frequency", "value_mode": "learned", "candidate_types": ["merge"], "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": 2.1} |
| {"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": 2.1} |
| {"event": "probe", "step": 0, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 10933.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.628225, "cps": 85.414062, "mean_p": 0.513249, "boundary_entropy": 0.686295, "bpb": 8.12598}, "tr": {"chunks": 8187.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.452245, "cps": 63.960938, "mean_p": 0.498002, "boundary_entropy": 0.686826, "bpb": 8.097536}, "fi": {"chunks": 13540.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.719792, "cps": 105.78125, "mean_p": 0.517169, "boundary_entropy": 0.686661, "bpb": 8.065117}}, "bpb": {"en": 8.12598, "tr": 8.09754, "fi": 8.06512}, "gini_cpc": 0.099077, "gini_cps": 0.109267, "bpic": 1.737048, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.7198781817505053, "fi": 1.1457544508645294}, "inventory": {"size": 0, "admitted": 0, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 0.0, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 3.8} |
| {"event": "train", "step": 0, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 5.62735, "bpb": 8.11854, "chunk_loss": 0.062086, "F": 0.59833, "G": 0.50836, "chunks": 612.69, "bytes_per_chunk": 1.671, "boundary_entropy": 0.69105, "router_noise": 0.0, "grad_norm": 11.4936, "lr": 1e-06, "router_trainable": true, "sec_per_step": 0.01, "probe_sec": 0.01, "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": {"merge": 0.0}, "mean_gate": {"merge": null}, "n_reads": 0, "n_cand": 19542, "read_suppressed_frac": 0.0, "dropout_reads_logged": 0, "inventory": {"size": 7355, "admitted": 7355, "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": 7355, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 2.656968116760254, "n_util_obs": 0, "util_count_spearman": NaN}, "probe_overhead_sec_total": 0.01, "probe_overhead_seq_frac": 0.0, "util_loss": -0.0, "alpha": 0.03, "t": 4.6} |
| {"event": "probe", "step": 1, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 10908.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.626788, "cps": 85.21875, "mean_p": 0.513143, "boundary_entropy": 0.686304, "bpb": 8.103122}, "tr": {"chunks": 8204.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.453185, "cps": 64.09375, "mean_p": 0.498082, "boundary_entropy": 0.686828, "bpb": 8.078603}, "fi": {"chunks": 13506.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.717984, "cps": 105.515625, "mean_p": 0.517038, "boundary_entropy": 0.686672, "bpb": 8.058114}}, "bpb": {"en": 8.10312, "tr": 8.0786, "fi": 8.05811}, "gini_cpc": 0.098185, "gini_cps": 0.108366, "bpic": 1.739285, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.723026293680575, "fi": 1.145496727653734}, "inventory": {"size": 7355, "admitted": 7355, "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": 7355, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 2.656968116760254, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 5.8} |
| {"event": "probe", "step": 2, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 10882.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.625294, "cps": 85.015625, "mean_p": 0.51299, "boundary_entropy": 0.686317, "bpb": 8.04112}, "tr": {"chunks": 8248.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.455615, "cps": 64.4375, "mean_p": 0.498256, "boundary_entropy": 0.686832, "bpb": 8.035076}, "fi": {"chunks": 13468.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.715964, "cps": 105.21875, "mean_p": 0.516806, "boundary_entropy": 0.686691, "bpb": 8.002541}}, "bpb": {"en": 8.04112, "tr": 8.03508, "fi": 8.00254}, "gini_cpc": 0.096593, "gini_cps": 0.106755, "bpic": 1.740352, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.728640823010943, "fi": 1.14500299563312}, "inventory": {"size": 12617, "admitted": 12617, "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": 12617, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 2.9243876934051514, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 8.3} |
| {"event": "probe", "step": 4, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 10722.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.616101, "cps": 83.765625, "mean_p": 0.512132, "boundary_entropy": 0.686378, "bpb": 7.831652}, "tr": {"chunks": 8392.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.46357, "cps": 65.5625, "mean_p": 0.498701, "boundary_entropy": 0.686846, "bpb": 7.883028}, "fi": {"chunks": 13218.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.702674, "cps": 103.265625, "mean_p": 0.515693, "boundary_entropy": 0.686774, "bpb": 7.803572}}, "bpb": {"en": 7.83165, "tr": 7.88303, "fi": 7.80357}, "gini_cpc": 0.089434, "gini_cps": 0.099509, "bpic": 1.75467, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.7524250417951298, "fi": 1.1405180986716836}, "inventory": {"size": 21902, "admitted": 21902, "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": 21902, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 3.3778648376464844, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 10.7} |
| {"event": "probe", "step": 6, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 10336.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.593921, "cps": 80.75, "mean_p": 0.510414, "boundary_entropy": 0.686481, "bpb": 7.552874}, "tr": {"chunks": 8519.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.470585, "cps": 66.554688, "mean_p": 0.499232, "boundary_entropy": 0.686868, "bpb": 7.677954}, "fi": {"chunks": 12730.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.676732, "cps": 99.453125, "mean_p": 0.513625, "boundary_entropy": 0.686908, "bpb": 7.543857}}, "bpb": {"en": 7.55287, "tr": 7.67795, "fi": 7.54386}, "gini_cpc": 0.078927, "gini_cps": 0.088882, "bpic": 1.796169, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.7923365431366336, "fi": 1.139431285511919}, "inventory": {"size": 29546, "admitted": 29546, "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": 29546, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 3.765145778656006, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 13.4} |
| {"event": "probe", "step": 8, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 9666.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.555421, "cps": 75.515625, "mean_p": 0.507391, "boundary_entropy": 0.686612, "bpb": 7.298639}, "tr": {"chunks": 8645.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.477545, "cps": 67.539062, "mean_p": 0.499639, "boundary_entropy": 0.686901, "bpb": 7.47919}, "fi": {"chunks": 11844.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.629632, "cps": 92.53125, "mean_p": 0.510217, "boundary_entropy": 0.687082, "bpb": 7.310451}}, "bpb": {"en": 7.29864, "tr": 7.47919, "fi": 7.31045}, "gini_cpc": 0.060983, "gini_cps": 0.070723, "bpic": 1.881346, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.8597887843179451, "fi": 1.1336104603006265}, "inventory": {"size": 37529, "admitted": 37529, "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": 37529, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 3.8738577365875244, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 15.5} |
| {"event": "probe", "step": 16, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3795.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.218066, "cps": 29.648438, "mean_p": 0.48468, "boundary_entropy": 0.686187, "bpb": 6.726534}, "tr": {"chunks": 7328.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.404795, "cps": 57.25, "mean_p": 0.494626, "boundary_entropy": 0.687003, "bpb": 7.013899}, "fi": {"chunks": 4531.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.24087, "cps": 35.398438, "mean_p": 0.486361, "boundary_entropy": 0.686891, "bpb": 6.787905}}, "bpb": {"en": 6.72653, "tr": 7.0139, "fi": 6.7879}, "gini_cpc": 0.144126, "gini_cps": 0.150462, "bpic": 3.624122, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.856296087015473, "fi": 1.1045732429284607}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 3.3566741943359375, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 19.3} |
| {"event": "probe", "step": 32, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 2633.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.151296, "cps": 20.570312, "mean_p": 0.48691, "boundary_entropy": 0.686962, "bpb": 6.305624}, "tr": {"chunks": 4634.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.25598, "cps": 36.203125, "mean_p": 0.490089, "boundary_entropy": 0.687554, "bpb": 6.652063}, "fi": {"chunks": 1368.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.072723, "cps": 10.6875, "mean_p": 0.479153, "boundary_entropy": 0.686948, "bpb": 6.422029}}, "bpb": {"en": 6.30562, "tr": 6.65206, "fi": 6.42203}, "gini_cpc": 0.254523, "gini_cps": 0.252152, "bpic": 6.570006, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.6919157727632692, "fi": 0.4806705065033847}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 3.983367919921875, "n_util_obs": 1, "util_count_spearman": NaN}, "t": 25.9} |
| {"event": "consolidate", "step": 50, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 32.0} |
| {"event": "train", "step": 50, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 4.21292, "bpb": 6.07795, "chunk_loss": 0.034205, "F": 0.22577, "G": 0.48856, "chunks": 231.19, "bytes_per_chunk": 4.429, "boundary_entropy": 0.69161, "router_noise": 0.0, "grad_norm": 2.4792, "lr": 5.1e-05, "router_trainable": true, "sec_per_step": 0.555, "probe_sec": 0.0681, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06083, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.00358, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -2.99167, "trunc_abs_frac": 1.12683, "mean_delta_full": 0.00366, "read_hit_rate": {"merge": 0.38921}, "mean_gate": {"merge": 0.11912}, "n_reads": 2849, "n_cand": 7320, "read_suppressed_frac": 0.1046, "dropout_reads_logged": 2849, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 4.51416015625, "n_util_obs": 2, "util_count_spearman": NaN}, "probe_overhead_sec_total": 2.393, "probe_overhead_seq_frac": 0.4773, "util_loss": 0.052657, "alpha": 0.03, "t": 32.4} |
| {"event": "probe", "step": 64, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 4880.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.280411, "cps": 38.125, "mean_p": 0.485727, "boundary_entropy": 0.685564, "bpb": 5.553172}, "tr": {"chunks": 5519.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.304867, "cps": 43.117188, "mean_p": 0.487045, "boundary_entropy": 0.686515, "bpb": 5.955713}, "fi": {"chunks": 4764.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.253256, "cps": 37.21875, "mean_p": 0.484218, "boundary_entropy": 0.686343, "bpb": 5.718636}}, "bpb": {"en": 5.55317, "tr": 5.95571, "fi": 5.71864}, "gini_cpc": 0.041032, "gini_cps": 0.033195, "bpic": 3.741476, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.087211758670558, "fi": 0.9031589033622637}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 4.9490814208984375, "n_util_obs": 5, "util_count_spearman": -0.7}, "t": 37.3} |
| {"event": "consolidate", "step": 100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 49.4} |
| {"event": "train", "step": 100, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 3.24995, "bpb": 4.68868, "chunk_loss": 0.029535, "F": 0.15897, "G": 0.45248, "chunks": 162.78, "bytes_per_chunk": 6.291, "boundary_entropy": 0.68196, "router_noise": 0.0, "grad_norm": 1.6277, "lr": 0.000101, "router_trainable": true, "sec_per_step": 0.348, "probe_sec": 0.0231, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01849, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.00554, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.08534, "trunc_abs_frac": 0.09513, "mean_delta_full": -0.05078, "read_hit_rate": {"merge": 0.2091}, "mean_gate": {"merge": 0.11943}, "n_reads": 1075, "n_cand": 5141, "read_suppressed_frac": 0.0958, "dropout_reads_logged": 1075, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 6.0265960693359375, "n_util_obs": 11, "util_count_spearman": -0.33636363636363636}, "probe_overhead_sec_total": 5.09, "probe_overhead_seq_frac": 0.4873, "util_loss": 0.001235, "alpha": 0.03, "t": 49.7} |
| {"event": "probe", "step": 128, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3838.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.220537, "cps": 29.984375, "mean_p": 0.416231, "boundary_entropy": 0.644935, "bpb": 4.150495}, "tr": {"chunks": 4832.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.266917, "cps": 37.75, "mean_p": 0.437007, "boundary_entropy": 0.659426, "bpb": 4.394721}, "fi": {"chunks": 3831.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.203657, "cps": 29.929688, "mean_p": 0.429086, "boundary_entropy": 0.660151, "bpb": 4.345524}}, "bpb": {"en": 4.1505, "tr": 4.39472, "fi": 4.34552}, "gini_cpc": 0.061022, "gini_cps": 0.053382, "bpic": 4.538197, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.2103069411422225, "fi": 0.9234628275800952}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 6.813079833984375, "n_util_obs": 14, "util_count_spearman": -0.23956043956043957}, "t": 59.1} |
| {"event": "consolidate", "step": 150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 67.1} |
| {"event": "train", "step": 150, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.79368, "bpb": 4.03043, "chunk_loss": 0.03182, "F": 0.20438, "G": 0.38693, "chunks": 209.28, "bytes_per_chunk": 4.893, "boundary_entropy": 0.62196, "router_noise": 0.0, "grad_norm": 1.2182, "lr": 0.000151, "router_trainable": true, "sec_per_step": 0.353, "probe_sec": 0.0256, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.019, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.00593, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.21106, "trunc_abs_frac": 0.4908, "mean_delta_full": -0.04858, "read_hit_rate": {"merge": 0.30981}, "mean_gate": {"merge": 0.121}, "n_reads": 2055, "n_cand": 6633, "read_suppressed_frac": 0.108, "dropout_reads_logged": 2055, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 7.5304412841796875, "n_util_obs": 16, "util_count_spearman": -0.24705882352941178}, "probe_overhead_sec_total": 7.666, "probe_overhead_seq_frac": 0.4915, "util_loss": 0.000394, "alpha": 0.03, "t": 67.4} |
| {"event": "consolidate", "step": 200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 82.6} |
| {"event": "probe", "step": 200, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3577.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.205539, "cps": 27.945312, "mean_p": 0.380689, "boundary_entropy": 0.593481, "bpb": 3.64097}, "tr": {"chunks": 4417.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.243993, "cps": 34.507812, "mean_p": 0.384489, "boundary_entropy": 0.608153, "bpb": 3.835863}, "fi": {"chunks": 3566.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.18957, "cps": 27.859375, "mean_p": 0.398353, "boundary_entropy": 0.635398, "bpb": 3.87154}}, "bpb": {"en": 3.64097, "tr": 3.83586, "fi": 3.87154}, "gini_cpc": 0.05677, "gini_cps": 0.049077, "bpic": 4.907612, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.187085575657363, "fi": 0.922305153776705}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 8.735916137695312, "n_util_obs": 21, "util_count_spearman": 0.14935064935064934}, "t": 83.3} |
| {"event": "eval", "step": 200, "phase": 1, "bpb": {"en": 3.64791, "tr": 3.84669, "fi": 3.86993}, "bytes_per_chunk": {"en": 5.302, "tr": 5.22, "fi": 5.856}, "bpb_mean": 3.78818, "frozen_inventory": true, "t": 84.6} |
| {"event": "train", "step": 200, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.5409, "bpb": 3.66574, "chunk_loss": 0.030651, "F": 0.18088, "G": 0.37049, "chunks": 185.22, "bytes_per_chunk": 5.529, "boundary_entropy": 0.59675, "router_noise": 0.0, "grad_norm": 1.3922, "lr": 0.000201, "router_trainable": true, "sec_per_step": 0.383, "probe_sec": 0.0548, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02154, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.01693, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.14702, "trunc_abs_frac": 2.82207, "mean_delta_full": 0.19102, "read_hit_rate": {"merge": 0.21614}, "mean_gate": {"merge": 0.12341}, "n_reads": 1267, "n_cand": 5862, "read_suppressed_frac": 0.0923, "dropout_reads_logged": 1267, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 8.75665283203125, "n_util_obs": 21, "util_count_spearman": 0.15974025974025974}, "probe_overhead_sec_total": 10.22, "probe_overhead_seq_frac": 0.4936, "util_loss": 0.000421, "alpha": 0.03, "t": 86.5} |
| {"event": "consolidate", "step": 250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 102.9} |
| {"event": "train", "step": 250, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.38112, "bpb": 3.43524, "chunk_loss": 0.031153, "F": 0.1933, "G": 0.35128, "chunks": 197.94, "bytes_per_chunk": 5.173, "boundary_entropy": 0.55197, "router_noise": 0.0, "grad_norm": 1.7205, "lr": 0.000251, "router_trainable": true, "sec_per_step": 0.333, "probe_sec": 0.0299, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01959, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.01391, "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": 0.04148, "trunc_abs_frac": 0.72347, "mean_delta_full": -0.07308, "read_hit_rate": {"merge": 0.24689}, "mean_gate": {"merge": 0.12986}, "n_reads": 1548, "n_cand": 6270, "read_suppressed_frac": 0.0853, "dropout_reads_logged": 1548, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 10.0242919921875, "n_util_obs": 28, "util_count_spearman": -0.04324028461959496}, "probe_overhead_sec_total": 12.938, "probe_overhead_seq_frac": 0.4949, "util_loss": 0.000645, "alpha": 0.03, "t": 103.1} |
| {"event": "probe", "step": 256, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3262.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.187439, "cps": 25.484375, "mean_p": 0.341649, "boundary_entropy": 0.535707, "bpb": 3.348075}, "tr": {"chunks": 4711.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.260233, "cps": 36.804688, "mean_p": 0.335051, "boundary_entropy": 0.52984, "bpb": 3.506208}, "fi": {"chunks": 5069.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.26947, "cps": 39.601562, "mean_p": 0.382407, "boundary_entropy": 0.586245, "bpb": 3.651388}}, "bpb": {"en": 3.34808, "tr": 3.50621, "fi": 3.65139}, "gini_cpc": 0.076257, "gini_cps": 0.092368, "bpic": 4.349946, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.3883619934476037, "fi": 1.4376414018162076}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 10.155548095703125, "n_util_obs": 31, "util_count_spearman": -0.04738858737205862}, "t": 105.6} |
| {"event": "consolidate", "step": 300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 121.9} |
| {"event": "train", "step": 300, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.27735, "bpb": 3.28552, "chunk_loss": 0.030473, "F": 0.17981, "G": 0.31886, "chunks": 184.12, "bytes_per_chunk": 5.561, "boundary_entropy": 0.49948, "router_noise": 0.0, "grad_norm": 1.3192, "lr": 0.0003, "router_trainable": true, "sec_per_step": 0.381, "probe_sec": 0.0383, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03135, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.00748, "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": 0.59074, "trunc_abs_frac": 0.66687, "mean_delta_full": -0.20044, "read_hit_rate": {"merge": 0.18572}, "mean_gate": {"merge": 0.13452}, "n_reads": 1082, "n_cand": 5826, "read_suppressed_frac": 0.097, "dropout_reads_logged": 1082, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 11.077255249023438, "n_util_obs": 34, "util_count_spearman": -0.09488158899923606}, "probe_overhead_sec_total": 15.511, "probe_overhead_seq_frac": 0.4957, "util_loss": 0.001094, "alpha": 0.03, "t": 122.2} |
| {"event": "consolidate", "step": 350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 137.8} |
| {"event": "train", "step": 350, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.16771, "bpb": 3.12734, "chunk_loss": 0.031098, "F": 0.20172, "G": 0.30169, "chunks": 206.56, "bytes_per_chunk": 4.957, "boundary_entropy": 0.45215, "router_noise": 0.0, "grad_norm": 1.0779, "lr": 0.0002999, "router_trainable": true, "sec_per_step": 0.319, "probe_sec": 0.0249, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01979, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.05847, "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": 0.27264, "trunc_abs_frac": 0.50018, "mean_delta_full": -0.23364, "read_hit_rate": {"merge": 0.23881}, "mean_gate": {"merge": 0.14361}, "n_reads": 1563, "n_cand": 6545, "read_suppressed_frac": 0.0947, "dropout_reads_logged": 1563, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 12.16912841796875, "n_util_obs": 42, "util_count_spearman": 0.09715975859844864}, "probe_overhead_sec_total": 17.899, "probe_overhead_seq_frac": 0.4963, "util_loss": 0.001516, "alpha": 0.03, "t": 138.2} |
| {"event": "consolidate", "step": 400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 155.1} |
| {"event": "probe", "step": 400, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 2975.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.170948, "cps": 23.242188, "mean_p": 0.272825, "boundary_entropy": 0.410388, "bpb": 2.910227}, "tr": {"chunks": 3379.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.186654, "cps": 26.398438, "mean_p": 0.287432, "boundary_entropy": 0.463183, "bpb": 3.177092}, "fi": {"chunks": 3833.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.203764, "cps": 29.945312, "mean_p": 0.291174, "boundary_entropy": 0.47242, "bpb": 3.372343}}, "bpb": {"en": 2.91023, "tr": 3.17709, "fi": 3.37234}, "gini_cpc": 0.038972, "gini_cps": 0.05615, "bpic": 5.569059, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0918796968049773, "fi": 1.1919665991961614}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 13.30413818359375, "n_util_obs": 44, "util_count_spearman": 0.24214235377026075}, "t": 155.8} |
| {"event": "eval", "step": 400, "phase": 1, "bpb": {"en": 2.92223, "tr": 3.18554, "fi": 3.36379}, "bytes_per_chunk": {"en": 5.678, "tr": 5.561, "fi": 5.376}, "bpb_mean": 3.15719, "frozen_inventory": true, "t": 157.2} |
| {"event": "train", "step": 400, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.08924, "bpb": 3.01413, "chunk_loss": 0.030245, "F": 0.17563, "G": 0.27826, "chunks": 179.84, "bytes_per_chunk": 5.694, "boundary_entropy": 0.41848, "router_noise": 0.0, "grad_norm": 1.3879, "lr": 0.0002998, "router_trainable": true, "sec_per_step": 0.406, "probe_sec": 0.0287, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02097, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.03079, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.29312, "trunc_abs_frac": 0.34595, "mean_delta_full": 0.31494, "read_hit_rate": {"merge": 0.19156}, "mean_gate": {"merge": 0.16786}, "n_reads": 1090, "n_cand": 5690, "read_suppressed_frac": 0.1083, "dropout_reads_logged": 1090, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 13.322525024414062, "n_util_obs": 44, "util_count_spearman": 0.24214235377026075}, "probe_overhead_sec_total": 20.367, "probe_overhead_seq_frac": 0.4968, "util_loss": 0.002352, "alpha": 0.03, "t": 158.5} |
| {"event": "consolidate", "step": 450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 175.4} |
| {"event": "train", "step": 450, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.99805, "bpb": 2.88258, "chunk_loss": 0.030795, "F": 0.1987, "G": 0.27438, "chunks": 203.47, "bytes_per_chunk": 5.033, "boundary_entropy": 0.37487, "router_noise": 0.0, "grad_norm": 1.3429, "lr": 0.0002995, "router_trainable": true, "sec_per_step": 0.343, "probe_sec": 0.0244, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01939, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.04132, "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": 7.31846, "trunc_abs_frac": 1.25691, "mean_delta_full": -0.01961, "read_hit_rate": {"merge": 0.24554}, "mean_gate": {"merge": 0.17851}, "n_reads": 1583, "n_cand": 6447, "read_suppressed_frac": 0.0954, "dropout_reads_logged": 1583, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 14.476821899414062, "n_util_obs": 49, "util_count_spearman": 0.1620408163265306}, "probe_overhead_sec_total": 22.975, "probe_overhead_seq_frac": 0.4971, "util_loss": 0.003179, "alpha": 0.03, "t": 175.6} |
| {"event": "consolidate", "step": 500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 192.7} |
| {"event": "train", "step": 500, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.90756, "bpb": 2.75203, "chunk_loss": 0.031163, "F": 0.21307, "G": 0.27491, "chunks": 218.19, "bytes_per_chunk": 4.693, "boundary_entropy": 0.34322, "router_noise": 0.0, "grad_norm": 1.6323, "lr": 0.0002992, "router_trainable": true, "sec_per_step": 0.346, "probe_sec": 0.0264, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02123, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.07839, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.49734, "trunc_abs_frac": 0.86685, "mean_delta_full": -0.32141, "read_hit_rate": {"merge": 0.27204}, "mean_gate": {"merge": 0.17958}, "n_reads": 1882, "n_cand": 6918, "read_suppressed_frac": 0.0956, "dropout_reads_logged": 1882, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 15.745681762695312, "n_util_obs": 53, "util_count_spearman": 0.005483207680947271}, "probe_overhead_sec_total": 25.027, "probe_overhead_seq_frac": 0.4974, "util_loss": 0.005137, "alpha": 0.03, "t": 192.9} |
| {"event": "probe", "step": 512, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3397.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.195196, "cps": 26.539062, "mean_p": 0.264391, "boundary_entropy": 0.342062, "bpb": 2.70854}, "tr": {"chunks": 3509.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.193835, "cps": 27.414062, "mean_p": 0.266707, "boundary_entropy": 0.383408, "bpb": 3.031172}, "fi": {"chunks": 4401.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.233959, "cps": 34.382812, "mean_p": 0.267861, "boundary_entropy": 0.389981, "bpb": 3.220461}}, "bpb": {"en": 2.70854, "tr": 3.03117, "fi": 3.22046}, "gini_cpc": 0.042937, "gini_cps": 0.059196, "bpic": 5.017423, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9930277618060692, "fi": 1.1985828477368738}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 16.039306640625, "n_util_obs": 55, "util_count_spearman": 0.01215750645050682}, "t": 196.7} |
| {"event": "consolidate", "step": 550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 209.4} |
| {"event": "train", "step": 550, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.91047, "bpb": 2.75622, "chunk_loss": 0.030854, "F": 0.2038, "G": 0.26676, "chunks": 208.69, "bytes_per_chunk": 4.907, "boundary_entropy": 0.3353, "router_noise": 0.0, "grad_norm": 1.7662, "lr": 0.0002987, "router_trainable": true, "sec_per_step": 0.335, "probe_sec": 0.1073, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09442, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.0197, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.60533, "trunc_abs_frac": 1.09895, "mean_delta_full": -0.27359, "read_hit_rate": {"merge": 0.26565}, "mean_gate": {"merge": 0.19169}, "n_reads": 1757, "n_cand": 6614, "read_suppressed_frac": 0.0973, "dropout_reads_logged": 1757, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 17.058380126953125, "n_util_obs": 56, "util_count_spearman": -0.0031442241968557758}, "probe_overhead_sec_total": 26.974, "probe_overhead_seq_frac": 0.4976, "util_loss": 0.006091, "alpha": 0.03, "t": 209.7} |
| {"event": "consolidate", "step": 600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 227.2} |
| {"event": "probe", "step": 600, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3456.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.198586, "cps": 27.0, "mean_p": 0.255706, "boundary_entropy": 0.314043, "bpb": 2.587089}, "tr": {"chunks": 3703.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.204552, "cps": 28.929688, "mean_p": 0.261443, "boundary_entropy": 0.344846, "bpb": 2.92013}, "fi": {"chunks": 4771.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.253628, "cps": 37.273438, "mean_p": 0.272752, "boundary_entropy": 0.357436, "bpb": 3.136534}}, "bpb": {"en": 2.58709, "tr": 2.92013, "fi": 3.13653}, "gini_cpc": 0.055871, "gini_cps": 0.073484, "bpic": 4.755407, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0300387117389997, "fi": 1.2771676792981648}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 18.429580688476562, "n_util_obs": 63, "util_count_spearman": 0.005424347158218126}, "t": 227.9} |
| {"event": "eval", "step": 600, "phase": 1, "bpb": {"en": 2.59867, "tr": 2.92652, "fi": 3.13844}, "bytes_per_chunk": {"en": 4.883, "tr": 5.419, "fi": 4.245}, "bpb_mean": 2.88788, "frozen_inventory": true, "t": 229.3} |
| {"event": "train", "step": 600, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.88637, "bpb": 2.72146, "chunk_loss": 0.030842, "F": 0.20593, "G": 0.25851, "chunks": 210.88, "bytes_per_chunk": 4.856, "boundary_entropy": 0.31645, "router_noise": 0.0, "grad_norm": 1.0923, "lr": 0.0002982, "router_trainable": true, "sec_per_step": 0.415, "probe_sec": 0.0458, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03357, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.03318, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": -0.30816, "trunc_abs_frac": 1.44451, "mean_delta_full": -0.04075, "read_hit_rate": {"merge": 0.26107}, "mean_gate": {"merge": 0.19374}, "n_reads": 1745, "n_cand": 6684, "read_suppressed_frac": 0.1037, "dropout_reads_logged": 1745, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 18.454757690429688, "n_util_obs": 64, "util_count_spearman": 0.02142857142857143}, "probe_overhead_sec_total": 29.463, "probe_overhead_seq_frac": 0.4978, "util_loss": 0.007284, "alpha": 0.03, "t": 230.4} |
| {"event": "consolidate", "step": 650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 250.5} |
| {"event": "train", "step": 650, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.83619, "bpb": 2.64906, "chunk_loss": 0.030801, "F": 0.20331, "G": 0.26097, "chunks": 208.19, "bytes_per_chunk": 4.919, "boundary_entropy": 0.30362, "router_noise": 0.0, "grad_norm": 1.1583, "lr": 0.0002975, "router_trainable": true, "sec_per_step": 0.407, "probe_sec": 0.0381, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03049, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.12777, "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": -0.13966, "trunc_abs_frac": 2.08661, "mean_delta_full": -0.08187, "read_hit_rate": {"merge": 0.26372}, "mean_gate": {"merge": 0.19619}, "n_reads": 1740, "n_cand": 6598, "read_suppressed_frac": 0.104, "dropout_reads_logged": 1740, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 19.837509155273438, "n_util_obs": 70, "util_count_spearman": 0.04246347651124136}, "probe_overhead_sec_total": 32.29, "probe_overhead_seq_frac": 0.498, "util_loss": 0.008275, "alpha": 0.03, "t": 250.8} |
| {"event": "consolidate", "step": 700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 287.3} |
| {"event": "train", "step": 700, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.9826, "bpb": 2.86029, "chunk_loss": 0.031128, "F": 0.21161, "G": 0.26299, "chunks": 216.69, "bytes_per_chunk": 4.726, "boundary_entropy": 0.30272, "router_noise": 0.0, "grad_norm": 1.2017, "lr": 0.0002967, "router_trainable": true, "sec_per_step": 0.74, "probe_sec": 0.0552, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04607, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.06908, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 1.0, "trunc_abs_frac": 1.0, "mean_delta_full": 0.13794, "read_hit_rate": {"merge": 0.24643}, "mean_gate": {"merge": 0.21132}, "n_reads": 1693, "n_cand": 6870, "read_suppressed_frac": 0.0945, "dropout_reads_logged": 1693, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 21.2314453125, "n_util_obs": 78, "util_count_spearman": 0.015364576748237792}, "probe_overhead_sec_total": 35.585, "probe_overhead_seq_frac": 0.4981, "util_loss": 0.009696, "alpha": 0.03, "t": 287.8} |
| {"event": "consolidate", "step": 750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 306.9} |
| {"event": "train", "step": 750, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.78231, "bpb": 2.57133, "chunk_loss": 0.030996, "F": 0.21317, "G": 0.26196, "chunks": 218.28, "bytes_per_chunk": 4.691, "boundary_entropy": 0.28765, "router_noise": 0.0, "grad_norm": 1.2038, "lr": 0.0002959, "router_trainable": true, "sec_per_step": 0.388, "probe_sec": 0.0349, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02877, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.12233, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 0.58777, "trunc_abs_frac": 0.56689, "mean_delta_full": 0.15299, "read_hit_rate": {"merge": 0.28002}, "mean_gate": {"merge": 0.20517}, "n_reads": 1938, "n_cand": 6921, "read_suppressed_frac": 0.1011, "dropout_reads_logged": 1938, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 22.643600463867188, "n_util_obs": 83, "util_count_spearman": 0.021546227836175894}, "probe_overhead_sec_total": 38.64, "probe_overhead_seq_frac": 0.4983, "util_loss": 0.012254, "alpha": 0.03, "t": 307.2} |
| {"event": "consolidate", "step": 800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 331.8} |
| {"event": "probe", "step": 800, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3435.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.19738, "cps": 26.835938, "mean_p": 0.248027, "boundary_entropy": 0.278259, "bpb": 2.437515}, "tr": {"chunks": 3713.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.205104, "cps": 29.007812, "mean_p": 0.255209, "boundary_entropy": 0.320483, "bpb": 2.761437}, "fi": {"chunks": 4040.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.214768, "cps": 31.5625, "mean_p": 0.26303, "boundary_entropy": 0.319951, "bpb": 2.989438}}, "bpb": {"en": 2.43751, "tr": 2.76144, "fi": 2.98944}, "gini_cpc": 0.01878, "gini_cps": 0.036051, "bpic": 5.07079, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0391345302848547, "fi": 1.088095114839199}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 24.016921997070312, "n_util_obs": 99, "util_count_spearman": 0.023172613397183746}, "t": 332.6} |
| {"event": "eval", "step": 800, "phase": 1, "bpb": {"en": 2.44805, "tr": 2.77053, "fi": 2.99306}, "bytes_per_chunk": {"en": 4.892, "tr": 5.35, "fi": 4.83}, "bpb_mean": 2.73721, "frozen_inventory": true, "t": 333.7} |
| {"event": "train", "step": 800, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.85786, "bpb": 2.68032, "chunk_loss": 0.031025, "F": 0.20908, "G": 0.26152, "chunks": 214.09, "bytes_per_chunk": 4.783, "boundary_entropy": 0.28487, "router_noise": 0.0, "grad_norm": 1.3777, "lr": 0.0002949, "router_trainable": true, "sec_per_step": 0.563, "probe_sec": 0.0521, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03487, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.07195, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.09683, "trunc_abs_frac": 0.6975, "mean_delta_full": -0.25403, "read_hit_rate": {"merge": 0.26433}, "mean_gate": {"merge": 0.19782}, "n_reads": 1794, "n_cand": 6787, "read_suppressed_frac": 0.1026, "dropout_reads_logged": 1794, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 24.043502807617188, "n_util_obs": 99, "util_count_spearman": 0.023723090433913355}, "probe_overhead_sec_total": 41.313, "probe_overhead_seq_frac": 0.4984, "util_loss": 0.010572, "alpha": 0.03, "t": 335.3} |
| {"event": "consolidate", "step": 850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 361.1} |
| {"event": "train", "step": 850, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.81002, "bpb": 2.61131, "chunk_loss": 0.030918, "F": 0.21109, "G": 0.25755, "chunks": 216.16, "bytes_per_chunk": 4.737, "boundary_entropy": 0.28252, "router_noise": 0.0, "grad_norm": 1.1139, "lr": 0.0002938, "router_trainable": true, "sec_per_step": 0.522, "probe_sec": 0.0781, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06813, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.07876, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.02346, "trunc_abs_frac": 0.4813, "mean_delta_full": -0.19366, "read_hit_rate": {"merge": 0.2485}, "mean_gate": {"merge": 0.20143}, "n_reads": 1703, "n_cand": 6853, "read_suppressed_frac": 0.0969, "dropout_reads_logged": 1703, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 25.414764404296875, "n_util_obs": 111, "util_count_spearman": 0.025763425763425763}, "probe_overhead_sec_total": 44.279, "probe_overhead_seq_frac": 0.4985, "util_loss": 0.013153, "alpha": 0.03, "t": 361.4} |
| {"event": "consolidate", "step": 900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 380.3} |
| {"event": "train", "step": 900, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.60579, "bpb": 2.31667, "chunk_loss": 0.030911, "F": 0.21133, "G": 0.25504, "chunks": 216.41, "bytes_per_chunk": 4.732, "boundary_entropy": 0.26924, "router_noise": 0.0, "grad_norm": 1.1601, "lr": 0.0002927, "router_trainable": true, "sec_per_step": 0.387, "probe_sec": 0.0506, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04197, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.03371, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.45663, "trunc_abs_frac": 3.42355, "mean_delta_full": -0.18646, "read_hit_rate": {"merge": 0.28771}, "mean_gate": {"merge": 0.19353}, "n_reads": 1974, "n_cand": 6861, "read_suppressed_frac": 0.1033, "dropout_reads_logged": 1974, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 26.801589965820312, "n_util_obs": 130, "util_count_spearman": 0.025785077921453715}, "probe_overhead_sec_total": 47.102, "probe_overhead_seq_frac": 0.4985, "util_loss": 0.014416, "alpha": 0.03, "t": 380.8} |
| {"event": "consolidate", "step": 950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 398.3} |
| {"event": "train", "step": 950, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.68508, "bpb": 2.43106, "chunk_loss": 0.030794, "F": 0.20563, "G": 0.25182, "chunks": 210.56, "bytes_per_chunk": 4.863, "boundary_entropy": 0.26071, "router_noise": 0.0, "grad_norm": 1.1932, "lr": 0.0002914, "router_trainable": true, "sec_per_step": 0.355, "probe_sec": 0.0771, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06483, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.13073, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 1.20518, "trunc_abs_frac": 1.1542, "mean_delta_full": 0.27063, "read_hit_rate": {"merge": 0.26925}, "mean_gate": {"merge": 0.19021}, "n_reads": 1797, "n_cand": 6674, "read_suppressed_frac": 0.0952, "dropout_reads_logged": 1797, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 28.201766967773438, "n_util_obs": 142, "util_count_spearman": 0.033450023523314866}, "probe_overhead_sec_total": 49.86, "probe_overhead_seq_frac": 0.4986, "util_loss": 0.011644, "alpha": 0.03, "t": 398.5} |
| {"event": "consolidate", "step": 1000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 415.9} |
| {"event": "probe", "step": 1000, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3496.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.200885, "cps": 27.3125, "mean_p": 0.242657, "boundary_entropy": 0.260126, "bpb": 2.337974}, "tr": {"chunks": 3542.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.195658, "cps": 27.671875, "mean_p": 0.243789, "boundary_entropy": 0.292237, "bpb": 2.647798}, "fi": {"chunks": 4460.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.237095, "cps": 34.84375, "mean_p": 0.262235, "boundary_entropy": 0.289156, "bpb": 2.857167}}, "bpb": {"en": 2.33797, "tr": 2.6478, "fi": 2.85717}, "gini_cpc": 0.043597, "gini_cps": 0.055894, "bpic": 4.934075, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9739814860578502, "fi": 1.1802545178115877}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 29.538223266601562, "n_util_obs": 158, "util_count_spearman": 0.056672401754560196}, "t": 416.6} |
| {"event": "eval", "step": 1000, "phase": 1, "bpb": {"en": 2.34768, "tr": 2.65398, "fi": 2.86201}, "bytes_per_chunk": {"en": 4.853, "tr": 5.672, "fi": 4.4}, "bpb_mean": 2.62122, "frozen_inventory": true, "t": 417.7} |
| {"event": "train", "step": 1000, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.74223, "bpb": 2.51351, "chunk_loss": 0.0308, "F": 0.20761, "G": 0.25089, "chunks": 212.59, "bytes_per_chunk": 4.817, "boundary_entropy": 0.264, "router_noise": 0.0, "grad_norm": 1.0247, "lr": 0.0002901, "router_trainable": true, "sec_per_step": 0.422, "probe_sec": 0.0531, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04401, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.06842, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 7.06113, "trunc_abs_frac": 0.43959, "mean_delta_full": 0.01228, "read_hit_rate": {"merge": 0.2585}, "mean_gate": {"merge": 0.21194}, "n_reads": 1742, "n_cand": 6739, "read_suppressed_frac": 0.0941, "dropout_reads_logged": 1742, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 29.56085205078125, "n_util_obs": 159, "util_count_spearman": 0.05711598816311924}, "probe_overhead_sec_total": 52.864, "probe_overhead_seq_frac": 0.4987, "util_loss": 0.013766, "alpha": 0.03, "t": 419.6} |
| {"event": "probe", "step": 1024, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3268.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.187784, "cps": 25.53125, "mean_p": 0.232548, "boundary_entropy": 0.259807, "bpb": 2.333726}, "tr": {"chunks": 3619.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.199912, "cps": 28.273438, "mean_p": 0.249565, "boundary_entropy": 0.297689, "bpb": 2.636992}, "fi": {"chunks": 4634.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.246345, "cps": 36.203125, "mean_p": 0.266125, "boundary_entropy": 0.283602, "bpb": 2.838935}}, "bpb": {"en": 2.33373, "tr": 2.63699, "fi": 2.83893}, "gini_cpc": 0.061575, "gini_cps": 0.079044, "bpic": 4.924225, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0645844149934642, "fi": 1.3118561582792223}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 30.194992065429688, "n_util_obs": 166, "util_count_spearman": 0.0702405188159992}, "t": 428.0} |
| {"event": "consolidate", "step": 1050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 438.8} |
| {"event": "train", "step": 1050, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.64217, "bpb": 2.36914, "chunk_loss": 0.030876, "F": 0.20569, "G": 0.24786, "chunks": 210.62, "bytes_per_chunk": 4.862, "boundary_entropy": 0.25822, "router_noise": 0.0, "grad_norm": 1.1499, "lr": 0.0002886, "router_trainable": true, "sec_per_step": 0.39, "probe_sec": 0.0824, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06786, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.08377, "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": 1.68614, "trunc_abs_frac": 0.44892, "mean_delta_full": 0.03727, "read_hit_rate": {"merge": 0.27112}, "mean_gate": {"merge": 0.19486}, "n_reads": 1810, "n_cand": 6676, "read_suppressed_frac": 0.0939, "dropout_reads_logged": 1810, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 30.925735473632812, "n_util_obs": 172, "util_count_spearman": 0.07322501734907447}, "probe_overhead_sec_total": 55.779, "probe_overhead_seq_frac": 0.4988, "util_loss": 0.013889, "alpha": 0.03, "t": 439.1} |
| {"event": "consolidate", "step": 1100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 456.2} |
| {"event": "train", "step": 1100, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.74403, "bpb": 2.51611, "chunk_loss": 0.030808, "F": 0.20596, "G": 0.24898, "chunks": 210.91, "bytes_per_chunk": 4.855, "boundary_entropy": 0.258, "router_noise": 0.0, "grad_norm": 1.0002, "lr": 0.0002871, "router_trainable": true, "sec_per_step": 0.348, "probe_sec": 0.0272, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0224, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.0154, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.17632, "trunc_abs_frac": 0.17632, "mean_delta_full": -0.37732, "read_hit_rate": {"merge": 0.25221}, "mean_gate": {"merge": 0.19915}, "n_reads": 1686, "n_cand": 6685, "read_suppressed_frac": 0.1044, "dropout_reads_logged": 1686, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 32.281585693359375, "n_util_obs": 196, "util_count_spearman": 0.046789092081418016}, "probe_overhead_sec_total": 58.442, "probe_overhead_seq_frac": 0.4988, "util_loss": 0.013544, "alpha": 0.03, "t": 456.5} |
| {"event": "consolidate", "step": 1150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 474.0} |
| {"event": "train", "step": 1150, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.64427, "bpb": 2.37218, "chunk_loss": 0.030664, "F": 0.20312, "G": 0.24308, "chunks": 208.0, "bytes_per_chunk": 4.923, "boundary_entropy": 0.25053, "router_noise": 0.0, "grad_norm": 0.8774, "lr": 0.0002855, "router_trainable": true, "sec_per_step": 0.361, "probe_sec": 0.0463, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04167, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.60405, "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": 0.76565, "trunc_abs_frac": 9.92389, "mean_delta_full": 0.38033, "read_hit_rate": {"merge": 0.26456}, "mean_gate": {"merge": 0.20234}, "n_reads": 1744, "n_cand": 6592, "read_suppressed_frac": 0.0889, "dropout_reads_logged": 1744, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 33.64100646972656, "n_util_obs": 216, "util_count_spearman": 0.028588158979423836}, "probe_overhead_sec_total": 61.264, "probe_overhead_seq_frac": 0.4989, "util_loss": 0.015847, "alpha": 0.03, "t": 474.6} |
| {"event": "consolidate", "step": 1200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 491.7} |
| {"event": "probe", "step": 1200, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3366.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.193415, "cps": 26.296875, "mean_p": 0.232932, "boundary_entropy": 0.242652, "bpb": 2.271154}, "tr": {"chunks": 3675.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.203005, "cps": 28.710938, "mean_p": 0.243173, "boundary_entropy": 0.280711, "bpb": 2.561717}, "fi": {"chunks": 4081.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.216948, "cps": 31.882812, "mean_p": 0.258785, "boundary_entropy": 0.275008, "bpb": 2.747053}}, "bpb": {"en": 2.27115, "tr": 2.56172, "fi": 2.74705}, "gini_cpc": 0.025577, "gini_cps": 0.042858, "bpic": 5.100881, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0495830306732628, "fi": 1.121669006765962}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 34.91175842285156, "n_util_obs": 229, "util_count_spearman": -0.0008149073586265179}, "t": 492.4} |
| {"event": "eval", "step": 1200, "phase": 1, "bpb": {"en": 2.28144, "tr": 2.56896, "fi": 2.75741}, "bytes_per_chunk": {"en": 4.995, "tr": 5.432, "fi": 4.91}, "bpb_mean": 2.53593, "frozen_inventory": true, "t": 493.6} |
| {"event": "train", "step": 1200, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.62226, "bpb": 2.34043, "chunk_loss": 0.030586, "F": 0.19955, "G": 0.24361, "chunks": 204.34, "bytes_per_chunk": 5.011, "boundary_entropy": 0.24988, "router_noise": 0.0, "grad_norm": 0.9461, "lr": 0.0002837, "router_trainable": true, "sec_per_step": 0.406, "probe_sec": 0.0837, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07432, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.03256, "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": 0.99045, "trunc_abs_frac": 0.98406, "mean_delta_full": -0.28109, "read_hit_rate": {"merge": 0.25838}, "mean_gate": {"merge": 0.21312}, "n_reads": 1673, "n_cand": 6475, "read_suppressed_frac": 0.1052, "dropout_reads_logged": 1673, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 34.93553161621094, "n_util_obs": 229, "util_count_spearman": -0.0011106922490660635}, "probe_overhead_sec_total": 63.857, "probe_overhead_seq_frac": 0.4989, "util_loss": 0.014251, "alpha": 0.03, "t": 494.9} |
| {"event": "consolidate", "step": 1250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 511.2} |
| {"event": "train", "step": 1250, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.63955, "bpb": 2.36536, "chunk_loss": 0.030604, "F": 0.20047, "G": 0.24173, "chunks": 205.28, "bytes_per_chunk": 4.988, "boundary_entropy": 0.24247, "router_noise": 0.0, "grad_norm": 0.9767, "lr": 0.0002819, "router_trainable": true, "sec_per_step": 0.334, "probe_sec": 0.1, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09036, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.09594, "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": 0.33667, "trunc_abs_frac": 1.46054, "mean_delta_full": -0.26298, "read_hit_rate": {"merge": 0.26472}, "mean_gate": {"merge": 0.20569}, "n_reads": 1722, "n_cand": 6505, "read_suppressed_frac": 0.097, "dropout_reads_logged": 1722, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 36.225555419921875, "n_util_obs": 248, "util_count_spearman": 0.002040819136346654}, "probe_overhead_sec_total": 66.396, "probe_overhead_seq_frac": 0.499, "util_loss": 0.01575, "alpha": 0.03, "t": 511.6} |
| {"event": "consolidate", "step": 1300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 528.0} |
| {"event": "train", "step": 1300, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.67587, "bpb": 2.41777, "chunk_loss": 0.030719, "F": 0.20267, "G": 0.2466, "chunks": 207.53, "bytes_per_chunk": 4.934, "boundary_entropy": 0.24136, "router_noise": 0.0, "grad_norm": 0.9136, "lr": 0.00028, "router_trainable": true, "sec_per_step": 0.333, "probe_sec": 0.0264, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02074, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.03927, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 1.33055, "trunc_abs_frac": 0.56774, "mean_delta_full": 0.12691, "read_hit_rate": {"merge": 0.2489}, "mean_gate": {"merge": 0.20546}, "n_reads": 1637, "n_cand": 6577, "read_suppressed_frac": 0.1063, "dropout_reads_logged": 1637, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 37.522918701171875, "n_util_obs": 262, "util_count_spearman": 0.013679503997496439}, "probe_overhead_sec_total": 69.221, "probe_overhead_seq_frac": 0.499, "util_loss": 0.017275, "alpha": 0.03, "t": 528.2} |
| {"event": "consolidate", "step": 1350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 545.3} |
| {"event": "train", "step": 1350, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.60448, "bpb": 2.31477, "chunk_loss": 0.030615, "F": 0.19791, "G": 0.24397, "chunks": 202.66, "bytes_per_chunk": 5.053, "boundary_entropy": 0.24488, "router_noise": 0.0, "grad_norm": 0.981, "lr": 0.000278, "router_trainable": true, "sec_per_step": 0.352, "probe_sec": 0.0249, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02019, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.01199, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 0.50595, "trunc_abs_frac": 4.23682, "mean_delta_full": -0.04704, "read_hit_rate": {"merge": 0.26429}, "mean_gate": {"merge": 0.20617}, "n_reads": 1697, "n_cand": 6421, "read_suppressed_frac": 0.0919, "dropout_reads_logged": 1697, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 38.78569030761719, "n_util_obs": 282, "util_count_spearman": 0.007873026643127778}, "probe_overhead_sec_total": 71.806, "probe_overhead_seq_frac": 0.499, "util_loss": 0.014964, "alpha": 0.03, "t": 545.8} |
| {"event": "consolidate", "step": 1400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 562.0} |
| {"event": "probe", "step": 1400, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3402.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.195484, "cps": 26.578125, "mean_p": 0.231649, "boundary_entropy": 0.23029, "bpb": 2.217944}, "tr": {"chunks": 3594.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.198531, "cps": 28.078125, "mean_p": 0.236469, "boundary_entropy": 0.263849, "bpb": 2.489831}, "fi": {"chunks": 4082.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.217001, "cps": 31.890625, "mean_p": 0.249084, "boundary_entropy": 0.255486, "bpb": 2.656384}}, "bpb": {"en": 2.21794, "tr": 2.48983, "fi": 2.65638}, "gini_cpc": 0.023477, "gini_cps": 0.040922, "bpic": 5.121141, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0155874658443294, "fi": 1.1100714364939197}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 40.008148193359375, "n_util_obs": 296, "util_count_spearman": 0.013650037335186226}, "t": 562.6} |
| {"event": "eval", "step": 1400, "phase": 1, "bpb": {"en": 2.22714, "tr": 2.49676, "fi": 2.66469}, "bytes_per_chunk": {"en": 4.949, "tr": 5.549, "fi": 4.939}, "bpb_mean": 2.46286, "frozen_inventory": true, "t": 563.8} |
| {"event": "train", "step": 1400, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.60291, "bpb": 2.31251, "chunk_loss": 0.030559, "F": 0.19693, "G": 0.23782, "chunks": 201.66, "bytes_per_chunk": 5.078, "boundary_entropy": 0.2368, "router_noise": 0.0, "grad_norm": 0.8389, "lr": 0.0002759, "router_trainable": true, "sec_per_step": 0.387, "probe_sec": 0.0803, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07356, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.15498, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 1.00293, "trunc_abs_frac": 1.00866, "mean_delta_full": 0.78137, "read_hit_rate": {"merge": 0.26984}, "mean_gate": {"merge": 0.20958}, "n_reads": 1724, "n_cand": 6389, "read_suppressed_frac": 0.0969, "dropout_reads_logged": 1724, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 40.03639221191406, "n_util_obs": 296, "util_count_spearman": 0.01372637706240717}, "probe_overhead_sec_total": 74.462, "probe_overhead_seq_frac": 0.4991, "util_loss": 0.015457, "alpha": 0.03, "t": 565.2} |
| {"event": "consolidate", "step": 1450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 582.6} |
| {"event": "train", "step": 1450, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.65624, "bpb": 2.38945, "chunk_loss": 0.030507, "F": 0.19684, "G": 0.23661, "chunks": 201.56, "bytes_per_chunk": 5.08, "boundary_entropy": 0.23311, "router_noise": 0.0, "grad_norm": 0.9239, "lr": 0.0002738, "router_trainable": true, "sec_per_step": 0.353, "probe_sec": 0.0295, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02456, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.1693, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.35296, "trunc_abs_frac": 0.71071, "mean_delta_full": -0.31079, "read_hit_rate": {"merge": 0.24945}, "mean_gate": {"merge": 0.21259}, "n_reads": 1593, "n_cand": 6386, "read_suppressed_frac": 0.1092, "dropout_reads_logged": 1593, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 41.31047058105469, "n_util_obs": 316, "util_count_spearman": 0.0036042854012281844}, "probe_overhead_sec_total": 77.164, "probe_overhead_seq_frac": 0.4991, "util_loss": 0.016207, "alpha": 0.03, "t": 582.9} |
| {"event": "consolidate", "step": 1500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 600.3} |
| {"event": "train", "step": 1500, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.67104, "bpb": 2.41081, "chunk_loss": 0.030756, "F": 0.20654, "G": 0.24366, "chunks": 211.5, "bytes_per_chunk": 4.842, "boundary_entropy": 0.23452, "router_noise": 0.0, "grad_norm": 0.9032, "lr": 0.0002715, "router_trainable": true, "sec_per_step": 0.357, "probe_sec": 0.1369, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.1272, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.17581, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 1.00894, "trunc_abs_frac": 2.50183, "mean_delta_full": -0.40863, "read_hit_rate": {"merge": 0.26715}, "mean_gate": {"merge": 0.20912}, "n_reads": 1791, "n_cand": 6704, "read_suppressed_frac": 0.1061, "dropout_reads_logged": 1791, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 42.59764099121094, "n_util_obs": 333, "util_count_spearman": 0.03574183626096895}, "probe_overhead_sec_total": 79.788, "probe_overhead_seq_frac": 0.4991, "util_loss": 0.017858, "alpha": 0.03, "t": 600.7} |
| {"event": "consolidate", "step": 1550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 618.1} |
| {"event": "train", "step": 1550, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.58541, "bpb": 2.28727, "chunk_loss": 0.030695, "F": 0.20187, "G": 0.23971, "chunks": 206.72, "bytes_per_chunk": 4.954, "boundary_entropy": 0.2322, "router_noise": 0.0, "grad_norm": 0.8415, "lr": 0.0002692, "router_trainable": true, "sec_per_step": 0.352, "probe_sec": 0.0354, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02985, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.19363, "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": 1.14953, "trunc_abs_frac": 0.85813, "mean_delta_full": 0.23511, "read_hit_rate": {"merge": 0.27675}, "mean_gate": {"merge": 0.20711}, "n_reads": 1813, "n_cand": 6551, "read_suppressed_frac": 0.096, "dropout_reads_logged": 1813, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 43.86436462402344, "n_util_obs": 353, "util_count_spearman": 0.03327103247378791}, "probe_overhead_sec_total": 82.425, "probe_overhead_seq_frac": 0.4992, "util_loss": 0.016373, "alpha": 0.03, "t": 618.3} |
| {"event": "consolidate", "step": 1600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 635.4} |
| {"event": "probe", "step": 1600, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3161.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.181635, "cps": 24.695312, "mean_p": 0.22001, "boundary_entropy": 0.223257, "bpb": 2.177401}, "tr": {"chunks": 3622.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.200077, "cps": 28.296875, "mean_p": 0.235211, "boundary_entropy": 0.255875, "bpb": 2.428886}, "fi": {"chunks": 4020.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.213705, "cps": 31.40625, "mean_p": 0.24813, "boundary_entropy": 0.244171, "bpb": 2.59614}}, "bpb": {"en": 2.1774, "tr": 2.42889, "fi": 2.59614}, "gini_cpc": 0.035907, "gini_cps": 0.05301, "bpic": 5.251504, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.101533016553682, "fi": 1.1765592267987672}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 45.04231262207031, "n_util_obs": 376, "util_count_spearman": 0.04649693074818521}, "t": 636.1} |
| {"event": "eval", "step": 1600, "phase": 1, "bpb": {"en": 2.18909, "tr": 2.43749, "fi": 2.60435}, "bytes_per_chunk": {"en": 5.273, "tr": 5.317, "fi": 5.021}, "bpb_mean": 2.41031, "frozen_inventory": true, "t": 637.3} |
| {"event": "train", "step": 1600, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.52129, "bpb": 2.19476, "chunk_loss": 0.030388, "F": 0.19159, "G": 0.23144, "chunks": 196.19, "bytes_per_chunk": 5.219, "boundary_entropy": 0.22933, "router_noise": 0.0, "grad_norm": 0.8543, "lr": 0.0002668, "router_trainable": true, "sec_per_step": 0.41, "probe_sec": 0.0264, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01916, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.06639, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 153.19415, "trunc_abs_frac": 1.99579, "mean_delta_full": 0.00143, "read_hit_rate": {"merge": 0.25024}, "mean_gate": {"merge": 0.20495}, "n_reads": 1555, "n_cand": 6214, "read_suppressed_frac": 0.0984, "dropout_reads_logged": 1555, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 45.067413330078125, "n_util_obs": 379, "util_count_spearman": 0.04687098235068841}, "probe_overhead_sec_total": 85.073, "probe_overhead_seq_frac": 0.4992, "util_loss": 0.016566, "alpha": 0.03, "t": 638.8} |
| {"event": "consolidate", "step": 1650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 655.2} |
| {"event": "train", "step": 1650, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.58575, "bpb": 2.28775, "chunk_loss": 0.030428, "F": 0.1944, "G": 0.23131, "chunks": 199.06, "bytes_per_chunk": 5.144, "boundary_entropy": 0.22486, "router_noise": 0.0, "grad_norm": 0.8709, "lr": 0.0002643, "router_trainable": true, "sec_per_step": 0.335, "probe_sec": 0.033, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02785, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.12895, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.62399, "trunc_abs_frac": 1.1266, "mean_delta_full": 0.28584, "read_hit_rate": {"merge": 0.24992}, "mean_gate": {"merge": 0.21545}, "n_reads": 1576, "n_cand": 6306, "read_suppressed_frac": 0.1009, "dropout_reads_logged": 1576, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 46.28450012207031, "n_util_obs": 393, "util_count_spearman": 0.030960549855826976}, "probe_overhead_sec_total": 88.138, "probe_overhead_seq_frac": 0.4992, "util_loss": 0.016926, "alpha": 0.03, "t": 655.5} |
| {"event": "consolidate", "step": 1700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 672.9} |
| {"event": "train", "step": 1700, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.54337, "bpb": 2.22662, "chunk_loss": 0.030353, "F": 0.18985, "G": 0.22852, "chunks": 194.41, "bytes_per_chunk": 5.267, "boundary_entropy": 0.22118, "router_noise": 0.0, "grad_norm": 0.8018, "lr": 0.0002618, "router_trainable": true, "sec_per_step": 0.352, "probe_sec": 0.029, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02436, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.07378, "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": 0.59152, "trunc_abs_frac": 0.91805, "mean_delta_full": -0.15784, "read_hit_rate": {"merge": 0.2394}, "mean_gate": {"merge": 0.21416}, "n_reads": 1474, "n_cand": 6157, "read_suppressed_frac": 0.1045, "dropout_reads_logged": 1474, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 47.49977111816406, "n_util_obs": 410, "util_count_spearman": 0.03345324251451588}, "probe_overhead_sec_total": 91.381, "probe_overhead_seq_frac": 0.4992, "util_loss": 0.018718, "alpha": 0.03, "t": 673.1} |
| {"event": "consolidate", "step": 1750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 690.4} |
| {"event": "train", "step": 1750, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.57443, "bpb": 2.27142, "chunk_loss": 0.030473, "F": 0.19617, "G": 0.23089, "chunks": 200.88, "bytes_per_chunk": 5.098, "boundary_entropy": 0.22077, "router_noise": 0.0, "grad_norm": 0.7835, "lr": 0.0002591, "router_trainable": true, "sec_per_step": 0.351, "probe_sec": 0.0668, "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": 14, "mean_delta_loo": 0.406, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -20.31044, "trunc_abs_frac": 0.69223, "mean_delta_full": -0.01633, "read_hit_rate": {"merge": 0.24749}, "mean_gate": {"merge": 0.2158}, "n_reads": 1575, "n_cand": 6364, "read_suppressed_frac": 0.099, "dropout_reads_logged": 1575, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 48.681243896484375, "n_util_obs": 433, "util_count_spearman": 0.03495876184619153}, "probe_overhead_sec_total": 94.374, "probe_overhead_seq_frac": 0.4993, "util_loss": 0.017404, "alpha": 0.03, "t": 690.7} |
| {"event": "consolidate", "step": 1800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 708.0} |
| {"event": "probe", "step": 1800, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3245.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.186462, "cps": 25.351562, "mean_p": 0.219676, "boundary_entropy": 0.206119, "bpb": 2.142819}, "tr": {"chunks": 3695.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.20411, "cps": 28.867188, "mean_p": 0.230873, "boundary_entropy": 0.23587, "bpb": 2.364583}, "fi": {"chunks": 4176.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.221998, "cps": 32.625, "mean_p": 0.246283, "boundary_entropy": 0.221266, "bpb": 2.551054}}, "bpb": {"en": 2.14282, "tr": 2.36458, "fi": 2.55105}, "gini_cpc": 0.038674, "gini_cps": 0.055835, "bpic": 5.103634, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0946450319763292, "fi": 1.190578472632518}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 49.8580322265625, "n_util_obs": 448, "util_count_spearman": 0.04183112509890965}, "t": 708.7} |
| {"event": "eval", "step": 1800, "phase": 1, "bpb": {"en": 2.15678, "tr": 2.37449, "fi": 2.55634}, "bytes_per_chunk": {"en": 5.2, "tr": 5.444, "fi": 4.783}, "bpb_mean": 2.36254, "frozen_inventory": true, "t": 709.9} |
| {"event": "train", "step": 1800, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.54813, "bpb": 2.23348, "chunk_loss": 0.030383, "F": 0.18939, "G": 0.22612, "chunks": 193.94, "bytes_per_chunk": 5.28, "boundary_entropy": 0.21511, "router_noise": 0.0, "grad_norm": 0.8306, "lr": 0.0002564, "router_trainable": true, "sec_per_step": 0.411, "probe_sec": 0.0258, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01812, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.09915, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 1.78142, "trunc_abs_frac": 2.87194, "mean_delta_full": 0.09226, "read_hit_rate": {"merge": 0.24169}, "mean_gate": {"merge": 0.21045}, "n_reads": 1484, "n_cand": 6140, "read_suppressed_frac": 0.1004, "dropout_reads_logged": 1484, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 49.88050842285156, "n_util_obs": 448, "util_count_spearman": 0.042760543797475674}, "probe_overhead_sec_total": 97.001, "probe_overhead_seq_frac": 0.4993, "util_loss": 0.018334, "alpha": 0.03, "t": 711.2} |
| {"event": "consolidate", "step": 1850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 733.1} |
| {"event": "train", "step": 1850, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.55779, "bpb": 2.24742, "chunk_loss": 0.030509, "F": 0.19641, "G": 0.23072, "chunks": 201.12, "bytes_per_chunk": 5.091, "boundary_entropy": 0.20683, "router_noise": 0.0, "grad_norm": 0.7792, "lr": 0.0002537, "router_trainable": true, "sec_per_step": 0.443, "probe_sec": 0.0669, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05907, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.13827, "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": 0.3659, "trunc_abs_frac": 0.82562, "mean_delta_full": 0.09497, "read_hit_rate": {"merge": 0.24262}, "mean_gate": {"merge": 0.20421}, "n_reads": 1546, "n_cand": 6372, "read_suppressed_frac": 0.0925, "dropout_reads_logged": 1546, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 51.079681396484375, "n_util_obs": 458, "util_count_spearman": 0.05430008574881225}, "probe_overhead_sec_total": 99.704, "probe_overhead_seq_frac": 0.4993, "util_loss": 0.019735, "alpha": 0.03, "t": 733.4} |
| {"event": "consolidate", "step": 1900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 750.4} |
| {"event": "train", "step": 1900, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.53723, "bpb": 2.21775, "chunk_loss": 0.030525, "F": 0.1933, "G": 0.22735, "chunks": 197.94, "bytes_per_chunk": 5.173, "boundary_entropy": 0.21005, "router_noise": 0.0, "grad_norm": 0.8301, "lr": 0.0002508, "router_trainable": true, "sec_per_step": 0.348, "probe_sec": 0.0543, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0454, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.1067, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": -1.7195, "trunc_abs_frac": 1.90565, "mean_delta_full": -0.07756, "read_hit_rate": {"merge": 0.23158}, "mean_gate": {"merge": 0.20752}, "n_reads": 1452, "n_cand": 6270, "read_suppressed_frac": 0.1054, "dropout_reads_logged": 1452, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 52.26963806152344, "n_util_obs": 479, "util_count_spearman": 0.030548176000962855}, "probe_overhead_sec_total": 102.529, "probe_overhead_seq_frac": 0.4993, "util_loss": 0.018796, "alpha": 0.03, "t": 750.8} |
| {"event": "consolidate", "step": 1950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 767.8} |
| {"event": "train", "step": 1950, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.48684, "bpb": 2.14505, "chunk_loss": 0.030537, "F": 0.19919, "G": 0.22884, "chunks": 203.97, "bytes_per_chunk": 5.02, "boundary_entropy": 0.20582, "router_noise": 0.0, "grad_norm": 0.7677, "lr": 0.0002479, "router_trainable": true, "sec_per_step": 0.351, "probe_sec": 0.0324, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02829, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.00973, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": -0.7384, "trunc_abs_frac": 1.20312, "mean_delta_full": -0.05711, "read_hit_rate": {"merge": 0.27773}, "mean_gate": {"merge": 0.21252}, "n_reads": 1795, "n_cand": 6463, "read_suppressed_frac": 0.0886, "dropout_reads_logged": 1795, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 53.48078918457031, "n_util_obs": 501, "util_count_spearman": 0.017083495678181317}, "probe_overhead_sec_total": 105.24, "probe_overhead_seq_frac": 0.4992, "util_loss": 0.018978, "alpha": 0.03, "t": 768.4} |
| {"event": "consolidate", "step": 2000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 791.6} |
| {"event": "probe", "step": 2000, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3237.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.186002, "cps": 25.289062, "mean_p": 0.213523, "boundary_entropy": 0.192482, "bpb": 2.116372}, "tr": {"chunks": 3746.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.206927, "cps": 29.265625, "mean_p": 0.229862, "boundary_entropy": 0.220459, "bpb": 2.318123}, "fi": {"chunks": 4227.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.224709, "cps": 33.023438, "mean_p": 0.24389, "boundary_entropy": 0.208058, "bpb": 2.49676}}, "bpb": {"en": 2.11637, "tr": 2.31812, "fi": 2.49676}, "gini_cpc": 0.041779, "gini_cps": 0.058876, "bpic": 5.060839, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.1124964720208537, "fi": 1.2080969424698709}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 54.678253173828125, "n_util_obs": 521, "util_count_spearman": 0.019964817159850896}, "t": 792.3} |
| {"event": "eval", "step": 2000, "phase": 1, "bpb": {"en": 2.12742, "tr": 2.32815, "fi": 2.50816}, "bytes_per_chunk": {"en": 5.222, "tr": 5.439, "fi": 4.813}, "bpb_mean": 2.32124, "frozen_inventory": true, "t": 793.6} |
| {"event": "train", "step": 2000, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.64588, "bpb": 2.37451, "chunk_loss": 0.030699, "F": 0.20496, "G": 0.23585, "chunks": 209.88, "bytes_per_chunk": 4.879, "boundary_entropy": 0.20491, "router_noise": 0.0, "grad_norm": 0.7903, "lr": 0.000245, "router_trainable": true, "sec_per_step": 0.536, "probe_sec": 0.0841, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07854, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.17912, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.19762, "trunc_abs_frac": 1.85378, "mean_delta_full": -0.06146, "read_hit_rate": {"merge": 0.22986}, "mean_gate": {"merge": 0.20753}, "n_reads": 1529, "n_cand": 6652, "read_suppressed_frac": 0.102, "dropout_reads_logged": 1529, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 54.69647216796875, "n_util_obs": 521, "util_count_spearman": 0.02172977394305366}, "probe_overhead_sec_total": 108.063, "probe_overhead_seq_frac": 0.4993, "util_loss": 0.017215, "alpha": 0.03, "t": 795.2} |
| {"event": "probe", "step": 2048, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3285.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.188761, "cps": 25.664062, "mean_p": 0.216021, "boundary_entropy": 0.189289, "bpb": 2.104871}, "tr": {"chunks": 3694.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.204055, "cps": 28.859375, "mean_p": 0.226837, "boundary_entropy": 0.221082, "bpb": 2.309412}, "fi": {"chunks": 4217.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.224177, "cps": 32.945312, "mean_p": 0.244066, "boundary_entropy": 0.205499, "bpb": 2.483777}}, "bpb": {"en": 2.10487, "tr": 2.30941, "fi": 2.48378}, "gini_cpc": 0.038268, "gini_cps": 0.055496, "bpic": 5.067167, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0810233778956893, "fi": 1.1876280977150986}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 55.859619140625, "n_util_obs": 534, "util_count_spearman": 0.012848612218122056}, "t": 827.7} |
| {"event": "consolidate", "step": 2050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 830.3} |
| {"event": "train", "step": 2050, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.48774, "bpb": 2.14635, "chunk_loss": 0.030387, "F": 0.18716, "G": 0.21899, "chunks": 191.66, "bytes_per_chunk": 5.343, "boundary_entropy": 0.19662, "router_noise": 0.0, "grad_norm": 0.8326, "lr": 0.0002419, "router_trainable": true, "sec_per_step": 0.71, "probe_sec": 0.0439, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03644, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.13748, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 1.06118, "trunc_abs_frac": 2.03334, "mean_delta_full": 0.18611, "read_hit_rate": {"merge": 0.23925}, "mean_gate": {"merge": 0.22017}, "n_reads": 1452, "n_cand": 6069, "read_suppressed_frac": 0.0833, "dropout_reads_logged": 1452, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 55.93229675292969, "n_util_obs": 534, "util_count_spearman": 0.013512081959382189}, "probe_overhead_sec_total": 111.006, "probe_overhead_seq_frac": 0.4993, "util_loss": 0.021366, "alpha": 0.03, "t": 830.6} |
| {"event": "consolidate", "step": 2100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 850.2} |
| {"event": "train", "step": 2100, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.54739, "bpb": 2.23241, "chunk_loss": 0.030422, "F": 0.19611, "G": 0.22368, "chunks": 200.81, "bytes_per_chunk": 5.099, "boundary_entropy": 0.19714, "router_noise": 0.0, "grad_norm": 0.7825, "lr": 0.0002388, "router_trainable": true, "sec_per_step": 0.402, "probe_sec": 0.0524, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04169, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.20351, "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": 0.96739, "trunc_abs_frac": 0.75997, "mean_delta_full": 0.72778, "read_hit_rate": {"merge": 0.23845}, "mean_gate": {"merge": 0.20805}, "n_reads": 1517, "n_cand": 6362, "read_suppressed_frac": 0.0995, "dropout_reads_logged": 1517, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 57.12213134765625, "n_util_obs": 551, "util_count_spearman": 0.01656178508119342}, "probe_overhead_sec_total": 113.481, "probe_overhead_seq_frac": 0.4993, "util_loss": 0.02331, "alpha": 0.03, "t": 850.7} |
| {"event": "consolidate", "step": 2150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 873.1} |
| {"event": "train", "step": 2150, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.50499, "bpb": 2.17124, "chunk_loss": 0.030447, "F": 0.19479, "G": 0.22512, "chunks": 199.47, "bytes_per_chunk": 5.134, "boundary_entropy": 0.19633, "router_noise": 0.0, "grad_norm": 0.7892, "lr": 0.0002357, "router_trainable": true, "sec_per_step": 0.469, "probe_sec": 0.0724, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06503, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.24361, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.65671, "trunc_abs_frac": 0.52558, "mean_delta_full": -0.15735, "read_hit_rate": {"merge": 0.25589}, "mean_gate": {"merge": 0.20405}, "n_reads": 1617, "n_cand": 6319, "read_suppressed_frac": 0.1039, "dropout_reads_logged": 1617, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 58.3387451171875, "n_util_obs": 568, "util_count_spearman": 0.012599338266593006}, "probe_overhead_sec_total": 116.414, "probe_overhead_seq_frac": 0.4993, "util_loss": 0.021385, "alpha": 0.03, "t": 874.2} |
| {"event": "consolidate", "step": 2200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 898.3} |
| {"event": "probe", "step": 2200, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3258.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.187209, "cps": 25.453125, "mean_p": 0.212782, "boundary_entropy": 0.181883, "bpb": 2.08645}, "tr": {"chunks": 3692.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.203944, "cps": 28.84375, "mean_p": 0.227616, "boundary_entropy": 0.214655, "bpb": 2.289538}, "fi": {"chunks": 4170.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.221679, "cps": 32.578125, "mean_p": 0.2428, "boundary_entropy": 0.203919, "bpb": 2.450298}}, "bpb": {"en": 2.08645, "tr": 2.28954, "fi": 2.4503}, "gini_cpc": 0.037498, "gini_cps": 0.054676, "bpic": 5.101799, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0893919986604177, "fi": 1.1841240769257202}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 59.50450134277344, "n_util_obs": 588, "util_count_spearman": 0.005617278240118956}, "t": 898.9} |
| {"event": "eval", "step": 2200, "phase": 1, "bpb": {"en": 2.09768, "tr": 2.29671, "fi": 2.4602}, "bytes_per_chunk": {"en": 5.168, "tr": 5.357, "fi": 4.902}, "bpb_mean": 2.28486, "frozen_inventory": true, "t": 900.1} |
| {"event": "train", "step": 2200, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.48485, "bpb": 2.14219, "chunk_loss": 0.030366, "F": 0.19327, "G": 0.22144, "chunks": 197.91, "bytes_per_chunk": 5.174, "boundary_entropy": 0.18905, "router_noise": 0.0, "grad_norm": 0.6729, "lr": 0.0002325, "router_trainable": true, "sec_per_step": 0.564, "probe_sec": 0.1182, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03664, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.12083, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.79616, "trunc_abs_frac": 1.80737, "mean_delta_full": -0.18842, "read_hit_rate": {"merge": 0.25411}, "mean_gate": {"merge": 0.2003}, "n_reads": 1593, "n_cand": 6269, "read_suppressed_frac": 0.1067, "dropout_reads_logged": 1593, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 59.529937744140625, "n_util_obs": 588, "util_count_spearman": 0.005190365748075382}, "probe_overhead_sec_total": 119.117, "probe_overhead_seq_frac": 0.4993, "util_loss": 0.018876, "alpha": 0.03, "t": 902.4} |
| {"event": "consolidate", "step": 2250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 924.2} |
| {"event": "train", "step": 2250, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.49944, "bpb": 2.16324, "chunk_loss": 0.030405, "F": 0.19458, "G": 0.22201, "chunks": 199.25, "bytes_per_chunk": 5.139, "boundary_entropy": 0.18521, "router_noise": 0.0, "grad_norm": 0.7847, "lr": 0.0002293, "router_trainable": true, "sec_per_step": 0.443, "probe_sec": 0.0324, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02456, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.02136, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.59302, "trunc_abs_frac": 0.59302, "mean_delta_full": -0.39893, "read_hit_rate": {"merge": 0.24461}, "mean_gate": {"merge": 0.21621}, "n_reads": 1544, "n_cand": 6312, "read_suppressed_frac": 0.101, "dropout_reads_logged": 1544, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 60.733612060546875, "n_util_obs": 605, "util_count_spearman": 0.004481081799829217}, "probe_overhead_sec_total": 121.968, "probe_overhead_seq_frac": 0.4993, "util_loss": 0.024217, "alpha": 0.03, "t": 924.5} |
| {"event": "consolidate", "step": 2300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 943.4} |
| {"event": "train", "step": 2300, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.51944, "bpb": 2.19209, "chunk_loss": 0.030391, "F": 0.19498, "G": 0.22219, "chunks": 199.66, "bytes_per_chunk": 5.129, "boundary_entropy": 0.1911, "router_noise": 0.0, "grad_norm": 0.791, "lr": 0.000226, "router_trainable": true, "sec_per_step": 0.383, "probe_sec": 0.0833, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07313, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.08755, "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": 1.07904, "trunc_abs_frac": 0.80114, "mean_delta_full": 0.04504, "read_hit_rate": {"merge": 0.23858}, "mean_gate": {"merge": 0.20695}, "n_reads": 1509, "n_cand": 6325, "read_suppressed_frac": 0.0848, "dropout_reads_logged": 1509, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 61.91722106933594, "n_util_obs": 626, "util_count_spearman": 0.022897608958909264}, "probe_overhead_sec_total": 124.846, "probe_overhead_seq_frac": 0.4993, "util_loss": 0.029674, "alpha": 0.03, "t": 943.7} |
| {"event": "consolidate", "step": 2350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 960.1} |
| {"event": "train", "step": 2350, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.56184, "bpb": 2.25325, "chunk_loss": 0.030496, "F": 0.20129, "G": 0.22851, "chunks": 206.12, "bytes_per_chunk": 4.968, "boundary_entropy": 0.18931, "router_noise": 0.0, "grad_norm": 0.7506, "lr": 0.0002226, "router_trainable": true, "sec_per_step": 0.341, "probe_sec": 0.055, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04983, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.24195, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 4.22184, "trunc_abs_frac": 1.02967, "mean_delta_full": -0.00659, "read_hit_rate": {"merge": 0.2477}, "mean_gate": {"merge": 0.20863}, "n_reads": 1618, "n_cand": 6532, "read_suppressed_frac": 0.1044, "dropout_reads_logged": 1618, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 63.124542236328125, "n_util_obs": 646, "util_count_spearman": 0.02212665788437123}, "probe_overhead_sec_total": 127.723, "probe_overhead_seq_frac": 0.4993, "util_loss": 0.022429, "alpha": 0.03, "t": 960.7} |
| {"event": "consolidate", "step": 2400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 978.6} |
| {"event": "probe", "step": 2400, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3297.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.18945, "cps": 25.757812, "mean_p": 0.213013, "boundary_entropy": 0.17791, "bpb": 2.064034}, "tr": {"chunks": 3700.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.204386, "cps": 28.90625, "mean_p": 0.225459, "boundary_entropy": 0.208063, "bpb": 2.251924}, "fi": {"chunks": 4205.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.223539, "cps": 32.851562, "mean_p": 0.240664, "boundary_entropy": 0.187308, "bpb": 2.403062}}, "bpb": {"en": 2.06403, "tr": 2.25192, "fi": 2.40306}, "gini_cpc": 0.036811, "gini_cps": 0.054038, "bpic": 5.064453, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0788382743835108, "fi": 1.1799382768750535}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 64.35185241699219, "n_util_obs": 663, "util_count_spearman": 0.03315699387548129}, "t": 979.3} |
| {"event": "eval", "step": 2400, "phase": 1, "bpb": {"en": 2.07757, "tr": 2.26059, "fi": 2.41354}, "bytes_per_chunk": {"en": 5.077, "tr": 5.325, "fi": 4.807}, "bpb_mean": 2.25057, "frozen_inventory": true, "t": 980.4} |
| {"event": "phase_transition", "step": 2400, "phase": 2, "objective": "global", "router_trainable": true, "encoder_trainable": true, "frozen_router_params": 0, "t": 981.4} |
| {"event": "train", "step": 2400, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.50037, "bpb": 2.16458, "chunk_loss": 0.030555, "F": 0.20322, "G": 0.22763, "chunks": 208.09, "bytes_per_chunk": 4.921, "boundary_entropy": 0.18333, "router_noise": 0.0, "grad_norm": 0.7191, "lr": 0.0002192, "router_trainable": true, "sec_per_step": 0.419, "probe_sec": 0.052, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04633, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.06577, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.35634, "trunc_abs_frac": 2.12799, "mean_delta_full": 0.08064, "read_hit_rate": {"merge": 0.24412}, "mean_gate": {"merge": 0.20711}, "n_reads": 1610, "n_cand": 6595, "read_suppressed_frac": 0.1037, "dropout_reads_logged": 1610, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 64.3726806640625, "n_util_obs": 664, "util_count_spearman": 0.032310993897823506}, "probe_overhead_sec_total": 130.724, "probe_overhead_seq_frac": 0.4993, "util_loss": 0.019355, "alpha": 0.03, "t": 981.7} |
| {"event": "consolidate", "step": 2450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 998.7} |
| {"event": "train", "step": 2450, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.46834, "bpb": 2.11836, "chunk_loss": 0.030386, "F": 0.19513, "G": 0.21948, "chunks": 199.81, "bytes_per_chunk": 5.125, "boundary_entropy": 0.18415, "router_noise": 0.0, "grad_norm": 0.7169, "lr": 0.0002158, "router_trainable": true, "sec_per_step": 0.346, "probe_sec": 0.0282, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02059, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.0575, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.67774, "trunc_abs_frac": 0.77503, "mean_delta_full": -0.14581, "read_hit_rate": {"merge": 0.25482}, "mean_gate": {"merge": 0.20536}, "n_reads": 1613, "n_cand": 6330, "read_suppressed_frac": 0.106, "dropout_reads_logged": 1613, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 65.60118103027344, "n_util_obs": 685, "util_count_spearman": 0.04716788498154715}, "probe_overhead_sec_total": 133.565, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.019218, "alpha": 0.03, "t": 999.0} |
| {"event": "consolidate", "step": 2500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1016.3} |
| {"event": "train", "step": 2500, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.44011, "bpb": 2.07764, "chunk_loss": 0.030458, "F": 0.19995, "G": 0.2225, "chunks": 204.75, "bytes_per_chunk": 5.001, "boundary_entropy": 0.17923, "router_noise": 0.0, "grad_norm": 0.8219, "lr": 0.0002123, "router_trainable": true, "sec_per_step": 0.351, "probe_sec": 0.0251, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02024, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.25696, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.8223, "trunc_abs_frac": 0.87394, "mean_delta_full": 0.72214, "read_hit_rate": {"merge": 0.26064}, "mean_gate": {"merge": 0.20004}, "n_reads": 1691, "n_cand": 6488, "read_suppressed_frac": 0.0988, "dropout_reads_logged": 1691, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 66.82986450195312, "n_util_obs": 709, "util_count_spearman": 0.05096527526230172}, "probe_overhead_sec_total": 136.148, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.025166, "alpha": 0.03, "t": 1016.5} |
| {"event": "consolidate", "step": 2550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1033.1} |
| {"event": "train", "step": 2550, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.47446, "bpb": 2.1272, "chunk_loss": 0.030368, "F": 0.19522, "G": 0.21892, "chunks": 199.91, "bytes_per_chunk": 5.122, "boundary_entropy": 0.17694, "router_noise": 0.0, "grad_norm": 0.7566, "lr": 0.0002088, "router_trainable": true, "sec_per_step": 0.337, "probe_sec": 0.0643, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05699, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.02926, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 0.39858, "trunc_abs_frac": 1.28292, "mean_delta_full": -0.02081, "read_hit_rate": {"merge": 0.25691}, "mean_gate": {"merge": 0.20939}, "n_reads": 1627, "n_cand": 6333, "read_suppressed_frac": 0.1094, "dropout_reads_logged": 1627, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 68.06626892089844, "n_util_obs": 738, "util_count_spearman": 0.056947029263401144}, "probe_overhead_sec_total": 138.613, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.018379, "alpha": 0.03, "t": 1033.4} |
| {"event": "consolidate", "step": 2600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1050.8} |
| {"event": "probe", "step": 2600, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3224.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.185255, "cps": 25.1875, "mean_p": 0.206884, "boundary_entropy": 0.169853, "bpb": 2.044967}, "tr": {"chunks": 3686.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.203613, "cps": 28.796875, "mean_p": 0.220943, "boundary_entropy": 0.196357, "bpb": 2.232612}, "fi": {"chunks": 4120.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.219021, "cps": 32.1875, "mean_p": 0.234575, "boundary_entropy": 0.179552, "bpb": 2.38774}}, "bpb": {"en": 2.04497, "tr": 2.23261, "fi": 2.38774}, "gini_cpc": 0.03703, "gini_cps": 0.054155, "bpic": 5.143427, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0990915438525262, "fi": 1.1822638752232109}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 69.29756164550781, "n_util_obs": 753, "util_count_spearman": 0.06075215721230339}, "t": 1051.3} |
| {"event": "eval", "step": 2600, "phase": 2, "bpb": {"en": 2.05768, "tr": 2.24208, "fi": 2.39378}, "bytes_per_chunk": {"en": 5.173, "tr": 5.248, "fi": 4.875}, "bpb_mean": 2.23118, "frozen_inventory": true, "t": 1052.2} |
| {"event": "train", "step": 2600, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.42152, "bpb": 2.05082, "chunk_loss": 0.030321, "F": 0.19278, "G": 0.21497, "chunks": 197.41, "bytes_per_chunk": 5.187, "boundary_entropy": 0.17396, "router_noise": 0.0, "grad_norm": 0.7496, "lr": 0.0002053, "router_trainable": true, "sec_per_step": 0.405, "probe_sec": 0.0399, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02746, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.07957, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 2.53846, "trunc_abs_frac": 0.79123, "mean_delta_full": -0.0123, "read_hit_rate": {"merge": 0.25108}, "mean_gate": {"merge": 0.20149}, "n_reads": 1570, "n_cand": 6253, "read_suppressed_frac": 0.0994, "dropout_reads_logged": 1570, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 69.32243347167969, "n_util_obs": 753, "util_count_spearman": 0.06046543230707246}, "probe_overhead_sec_total": 141.532, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.018705, "alpha": 0.03, "t": 1053.6} |
| {"event": "consolidate", "step": 2650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1071.0} |
| {"event": "train", "step": 2650, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.46678, "bpb": 2.11612, "chunk_loss": 0.030346, "F": 0.19296, "G": 0.21747, "chunks": 197.59, "bytes_per_chunk": 5.182, "boundary_entropy": 0.17779, "router_noise": 0.0, "grad_norm": 0.7175, "lr": 0.0002017, "router_trainable": true, "sec_per_step": 0.352, "probe_sec": 0.0297, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02447, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.0787, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -0.23688, "trunc_abs_frac": 0.89117, "mean_delta_full": 0.28009, "read_hit_rate": {"merge": 0.24445}, "mean_gate": {"merge": 0.21515}, "n_reads": 1530, "n_cand": 6259, "read_suppressed_frac": 0.1065, "dropout_reads_logged": 1530, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 70.55268859863281, "n_util_obs": 778, "util_count_spearman": 0.05436876986568598}, "probe_overhead_sec_total": 144.309, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.01947, "alpha": 0.03, "t": 1071.2} |
| {"event": "consolidate", "step": 2700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1088.3} |
| {"event": "train", "step": 2700, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.44889, "bpb": 2.09031, "chunk_loss": 0.030441, "F": 0.19754, "G": 0.22197, "chunks": 202.28, "bytes_per_chunk": 5.062, "boundary_entropy": 0.17643, "router_noise": 0.0, "grad_norm": 0.6958, "lr": 0.0001981, "router_trainable": true, "sec_per_step": 0.348, "probe_sec": 0.0355, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02981, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.00572, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.03362, "trunc_abs_frac": 0.03362, "mean_delta_full": 0.58826, "read_hit_rate": {"merge": 0.25698}, "mean_gate": {"merge": 0.19846}, "n_reads": 1647, "n_cand": 6409, "read_suppressed_frac": 0.0917, "dropout_reads_logged": 1647, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 71.80476379394531, "n_util_obs": 804, "util_count_spearman": 0.058443057271000166}, "probe_overhead_sec_total": 146.939, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.0201, "alpha": 0.03, "t": 1088.6} |
| {"event": "consolidate", "step": 2750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1105.9} |
| {"event": "train", "step": 2750, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.46066, "bpb": 2.10729, "chunk_loss": 0.030492, "F": 0.19937, "G": 0.21982, "chunks": 204.16, "bytes_per_chunk": 5.016, "boundary_entropy": 0.17003, "router_noise": 0.0, "grad_norm": 0.68, "lr": 0.0001945, "router_trainable": true, "sec_per_step": 0.353, "probe_sec": 0.0363, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03072, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.05879, "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": -0.75042, "trunc_abs_frac": 1.22304, "mean_delta_full": 0.09603, "read_hit_rate": {"merge": 0.25383}, "mean_gate": {"merge": 0.19998}, "n_reads": 1642, "n_cand": 6469, "read_suppressed_frac": 0.0914, "dropout_reads_logged": 1642, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 73.06039428710938, "n_util_obs": 825, "util_count_spearman": 0.05811371972814655}, "probe_overhead_sec_total": 149.566, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.020372, "alpha": 0.03, "t": 1106.3} |
| {"event": "consolidate", "step": 2800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1123.0} |
| {"event": "probe", "step": 2800, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3211.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.184508, "cps": 25.085938, "mean_p": 0.205424, "boundary_entropy": 0.162406, "bpb": 2.031639}, "tr": {"chunks": 3676.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.20306, "cps": 28.71875, "mean_p": 0.218188, "boundary_entropy": 0.194473, "bpb": 2.197843}, "fi": {"chunks": 4158.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.221041, "cps": 32.484375, "mean_p": 0.235823, "boundary_entropy": 0.174471, "bpb": 2.342507}}, "bpb": {"en": 2.03164, "tr": 2.19784, "fi": 2.34251}, "gini_cpc": 0.040017, "gini_cps": 0.05716, "bpic": 5.136442, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.100547434949253, "fi": 1.1979988914627684}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 74.27445983886719, "n_util_obs": 845, "util_count_spearman": 0.06148372743460524}, "t": 1123.6} |
| {"event": "eval", "step": 2800, "phase": 2, "bpb": {"en": 2.04323, "tr": 2.2092, "fi": 2.34883}, "bytes_per_chunk": {"en": 5.214, "tr": 5.374, "fi": 4.878}, "bpb_mean": 2.20042, "frozen_inventory": true, "t": 1124.7} |
| {"event": "train", "step": 2800, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.41641, "bpb": 2.04345, "chunk_loss": 0.030342, "F": 0.19357, "G": 0.2139, "chunks": 198.22, "bytes_per_chunk": 5.166, "boundary_entropy": 0.16592, "router_noise": 0.0, "grad_norm": 0.7333, "lr": 0.0001909, "router_trainable": true, "sec_per_step": 0.398, "probe_sec": 0.1722, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.16333, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.16233, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -3.23755, "trunc_abs_frac": 2.28763, "mean_delta_full": 0.01593, "read_hit_rate": {"merge": 0.2459}, "mean_gate": {"merge": 0.20978}, "n_reads": 1544, "n_cand": 6279, "read_suppressed_frac": 0.1023, "dropout_reads_logged": 1544, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 74.30043029785156, "n_util_obs": 845, "util_count_spearman": 0.06207576109370472}, "probe_overhead_sec_total": 152.438, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.017712, "alpha": 0.03, "t": 1126.2} |
| {"event": "consolidate", "step": 2850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1144.5} |
| {"event": "train", "step": 2850, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.46265, "bpb": 2.11016, "chunk_loss": 0.030543, "F": 0.20435, "G": 0.22575, "chunks": 209.25, "bytes_per_chunk": 4.894, "boundary_entropy": 0.16765, "router_noise": 0.0, "grad_norm": 0.7565, "lr": 0.0001872, "router_trainable": true, "sec_per_step": 0.372, "probe_sec": 0.0545, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04941, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.05967, "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": 0.21702, "trunc_abs_frac": 1.41467, "mean_delta_full": -0.14156, "read_hit_rate": {"merge": 0.26071}, "mean_gate": {"merge": 0.18807}, "n_reads": 1729, "n_cand": 6632, "read_suppressed_frac": 0.0885, "dropout_reads_logged": 1729, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 75.52821350097656, "n_util_obs": 868, "util_count_spearman": 0.051376607629476795}, "probe_overhead_sec_total": 155.612, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.02529, "alpha": 0.03, "t": 1144.8} |
| {"event": "consolidate", "step": 2900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1161.9} |
| {"event": "train", "step": 2900, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.39654, "bpb": 2.01478, "chunk_loss": 0.030337, "F": 0.19537, "G": 0.21418, "chunks": 200.06, "bytes_per_chunk": 5.118, "boundary_entropy": 0.16324, "router_noise": 0.0, "grad_norm": 0.6695, "lr": 0.0001835, "router_trainable": true, "sec_per_step": 0.345, "probe_sec": 0.0224, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01783, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.01152, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.90586, "trunc_abs_frac": 0.28373, "mean_delta_full": 0.02917, "read_hit_rate": {"merge": 0.27375}, "mean_gate": {"merge": 0.19831}, "n_reads": 1735, "n_cand": 6338, "read_suppressed_frac": 0.0974, "dropout_reads_logged": 1735, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 76.80059814453125, "n_util_obs": 883, "util_count_spearman": 0.05358290800614839}, "probe_overhead_sec_total": 158.411, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.016981, "alpha": 0.03, "t": 1162.1} |
| {"event": "consolidate", "step": 2950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1179.3} |
| {"event": "train", "step": 2950, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.54614, "bpb": 2.23062, "chunk_loss": 0.030606, "F": 0.20529, "G": 0.22402, "chunks": 210.22, "bytes_per_chunk": 4.871, "boundary_entropy": 0.16375, "router_noise": 0.0, "grad_norm": 0.7496, "lr": 0.0001799, "router_trainable": true, "sec_per_step": 0.352, "probe_sec": 0.0843, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07857, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.03348, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.63699, "trunc_abs_frac": 0.63699, "mean_delta_full": 0.08911, "read_hit_rate": {"merge": 0.24178}, "mean_gate": {"merge": 0.20222}, "n_reads": 1611, "n_cand": 6663, "read_suppressed_frac": 0.0987, "dropout_reads_logged": 1611, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 78.04356384277344, "n_util_obs": 905, "util_count_spearman": 0.05909658998716808}, "probe_overhead_sec_total": 161.532, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.01884, "alpha": 0.03, "t": 1179.7} |
| {"event": "consolidate", "step": 3000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1199.1} |
| {"event": "probe", "step": 3000, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3217.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.184853, "cps": 25.132812, "mean_p": 0.203755, "boundary_entropy": 0.155181, "bpb": 2.010652}, "tr": {"chunks": 3648.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.201514, "cps": 28.5, "mean_p": 0.21394, "boundary_entropy": 0.181405, "bpb": 2.17576}, "fi": {"chunks": 4141.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.220137, "cps": 32.351562, "mean_p": 0.236083, "boundary_entropy": 0.173533, "bpb": 2.326792}}, "bpb": {"en": 2.01065, "tr": 2.17576, "fi": 2.32679}, "gini_cpc": 0.038784, "gini_cps": 0.055969, "bpic": 5.154643, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0901276055636528, "fi": 1.190875625570241}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 79.27723693847656, "n_util_obs": 914, "util_count_spearman": 0.05670814144966714}, "t": 1199.9} |
| {"event": "eval", "step": 3000, "phase": 2, "bpb": {"en": 2.02456, "tr": 2.186, "fi": 2.3297}, "bytes_per_chunk": {"en": 5.17, "tr": 5.321, "fi": 4.912}, "bpb_mean": 2.18009, "frozen_inventory": true, "t": 1201.3} |
| {"event": "train", "step": 3000, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.49056, "bpb": 2.15042, "chunk_loss": 0.030341, "F": 0.19156, "G": 0.21079, "chunks": 196.16, "bytes_per_chunk": 5.22, "boundary_entropy": 0.16207, "router_noise": 0.0, "grad_norm": 0.7019, "lr": 0.0001761, "router_trainable": true, "sec_per_step": 0.475, "probe_sec": 0.0823, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07394, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.017, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.93256, "trunc_abs_frac": 0.90371, "mean_delta_full": 0.46338, "read_hit_rate": {"merge": 0.23129}, "mean_gate": {"merge": 0.19379}, "n_reads": 1437, "n_cand": 6213, "read_suppressed_frac": 0.0974, "dropout_reads_logged": 1437, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 79.2989501953125, "n_util_obs": 915, "util_count_spearman": 0.05728454619631181}, "probe_overhead_sec_total": 164.412, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.022008, "alpha": 0.03, "t": 1203.5} |
| {"event": "consolidate", "step": 3050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1221.6} |
| {"event": "train", "step": 3050, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.43976, "bpb": 2.07713, "chunk_loss": 0.030391, "F": 0.19565, "G": 0.21291, "chunks": 200.34, "bytes_per_chunk": 5.111, "boundary_entropy": 0.15651, "router_noise": 0.0, "grad_norm": 0.6873, "lr": 0.0001724, "router_trainable": true, "sec_per_step": 0.37, "probe_sec": 0.0704, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06232, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.06697, "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": 1.11903, "trunc_abs_frac": 1.23183, "mean_delta_full": 0.37728, "read_hit_rate": {"merge": 0.24248}, "mean_gate": {"merge": 0.18865}, "n_reads": 1539, "n_cand": 6347, "read_suppressed_frac": 0.0949, "dropout_reads_logged": 1539, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 80.55233764648438, "n_util_obs": 942, "util_count_spearman": 0.06035325732469576}, "probe_overhead_sec_total": 167.956, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.022331, "alpha": 0.03, "t": 1221.9} |
| {"event": "consolidate", "step": 3100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1240.1} |
| {"event": "train", "step": 3100, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.49149, "bpb": 2.15177, "chunk_loss": 0.030384, "F": 0.19699, "G": 0.21593, "chunks": 201.72, "bytes_per_chunk": 5.076, "boundary_entropy": 0.15709, "router_noise": 0.0, "grad_norm": 0.7002, "lr": 0.0001687, "router_trainable": true, "sec_per_step": 0.368, "probe_sec": 0.0252, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02007, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.06329, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -0.05017, "trunc_abs_frac": 14.55746, "mean_delta_full": -0.20105, "read_hit_rate": {"merge": 0.25223}, "mean_gate": {"merge": 0.21392}, "n_reads": 1612, "n_cand": 6391, "read_suppressed_frac": 0.0974, "dropout_reads_logged": 1612, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 81.80574035644531, "n_util_obs": 973, "util_count_spearman": 0.06239275504421023}, "probe_overhead_sec_total": 170.687, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.020489, "alpha": 0.03, "t": 1240.4} |
| {"event": "consolidate", "step": 3150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1257.6} |
| {"event": "train", "step": 3150, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.45457, "bpb": 2.0985, "chunk_loss": 0.030305, "F": 0.19186, "G": 0.20937, "chunks": 196.47, "bytes_per_chunk": 5.212, "boundary_entropy": 0.15638, "router_noise": 0.0, "grad_norm": 0.7094, "lr": 0.000165, "router_trainable": true, "sec_per_step": 0.352, "probe_sec": 0.0348, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03011, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.12824, "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": 0.36355, "trunc_abs_frac": 0.81012, "mean_delta_full": 0.10258, "read_hit_rate": {"merge": 0.23027}, "mean_gate": {"merge": 0.20078}, "n_reads": 1433, "n_cand": 6223, "read_suppressed_frac": 0.0949, "dropout_reads_logged": 1433, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 83.09161376953125, "n_util_obs": 990, "util_count_spearman": 0.0635214023611339}, "probe_overhead_sec_total": 173.797, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.019361, "alpha": 0.03, "t": 1258.0} |
| {"event": "consolidate", "step": 3200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1285.6} |
| {"event": "probe", "step": 3200, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3293.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.18922, "cps": 25.726562, "mean_p": 0.207417, "boundary_entropy": 0.155687, "bpb": 1.99289}, "tr": {"chunks": 3881.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.214384, "cps": 30.320312, "mean_p": 0.221212, "boundary_entropy": 0.178406, "bpb": 2.147219}, "fi": {"chunks": 4197.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.223114, "cps": 32.789062, "mean_p": 0.234749, "boundary_entropy": 0.157339, "bpb": 2.298516}}, "bpb": {"en": 1.99289, "tr": 2.14722, "fi": 2.29852}, "gini_cpc": 0.036054, "gini_cps": 0.053, "bpic": 4.989183, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.1329884453905739, "fi": 1.1791239895026886}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 84.31842041015625, "n_util_obs": 1006, "util_count_spearman": 0.06796198444506697}, "t": 1286.2} |
| {"event": "eval", "step": 3200, "phase": 2, "bpb": {"en": 2.00563, "tr": 2.15651, "fi": 2.30546}, "bytes_per_chunk": {"en": 5.094, "tr": 5.261, "fi": 4.787}, "bpb_mean": 2.15586, "frozen_inventory": true, "t": 1287.3} |
| {"event": "train", "step": 3200, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.43442, "bpb": 2.06943, "chunk_loss": 0.030385, "F": 0.19836, "G": 0.21536, "chunks": 203.12, "bytes_per_chunk": 5.041, "boundary_entropy": 0.15606, "router_noise": 0.0, "grad_norm": 0.6341, "lr": 0.0001613, "router_trainable": true, "sec_per_step": 0.616, "probe_sec": 0.0398, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0264, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.04424, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.72153, "trunc_abs_frac": 0.72153, "mean_delta_full": 0.48132, "read_hit_rate": {"merge": 0.25994}, "mean_gate": {"merge": 0.18922}, "n_reads": 1673, "n_cand": 6436, "read_suppressed_frac": 0.1195, "dropout_reads_logged": 1673, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 84.3421630859375, "n_util_obs": 1007, "util_count_spearman": 0.0675429584838338}, "probe_overhead_sec_total": 176.877, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.018861, "alpha": 0.03, "t": 1288.8} |
| {"event": "consolidate", "step": 3250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1313.9} |
| {"event": "train", "step": 3250, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.43778, "bpb": 2.07427, "chunk_loss": 0.030452, "F": 0.19904, "G": 0.21484, "chunks": 203.81, "bytes_per_chunk": 5.024, "boundary_entropy": 0.15372, "router_noise": 0.0, "grad_norm": 0.7386, "lr": 0.0001576, "router_trainable": true, "sec_per_step": 0.511, "probe_sec": 0.1274, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.12049, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.07527, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.34742, "trunc_abs_frac": 0.70341, "mean_delta_full": 0.17511, "read_hit_rate": {"merge": 0.2687}, "mean_gate": {"merge": 0.2046}, "n_reads": 1735, "n_cand": 6457, "read_suppressed_frac": 0.0968, "dropout_reads_logged": 1735, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 85.60700988769531, "n_util_obs": 1024, "util_count_spearman": 0.06565049415757102}, "probe_overhead_sec_total": 180.235, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.019395, "alpha": 0.03, "t": 1314.3} |
| {"event": "consolidate", "step": 3300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1341.1} |
| {"event": "train", "step": 3300, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.41393, "bpb": 2.03987, "chunk_loss": 0.030559, "F": 0.2041, "G": 0.21921, "chunks": 209.0, "bytes_per_chunk": 4.9, "boundary_entropy": 0.14822, "router_noise": 0.0, "grad_norm": 0.6791, "lr": 0.0001539, "router_trainable": true, "sec_per_step": 0.544, "probe_sec": 0.0508, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0271, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.10408, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 0.35934, "trunc_abs_frac": 0.51376, "mean_delta_full": 0.2772, "read_hit_rate": {"merge": 0.26947}, "mean_gate": {"merge": 0.1882}, "n_reads": 1785, "n_cand": 6624, "read_suppressed_frac": 0.0908, "dropout_reads_logged": 1785, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 86.87800598144531, "n_util_obs": 1052, "util_count_spearman": 0.06789689849381587}, "probe_overhead_sec_total": 183.094, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.016876, "alpha": 0.03, "t": 1341.5} |
| {"event": "consolidate", "step": 3350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1360.2} |
| {"event": "train", "step": 3350, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.35703, "bpb": 1.95778, "chunk_loss": 0.030404, "F": 0.19922, "G": 0.2141, "chunks": 204.0, "bytes_per_chunk": 5.02, "boundary_entropy": 0.14644, "router_noise": 0.0, "grad_norm": 0.7112, "lr": 0.0001501, "router_trainable": true, "sec_per_step": 0.401, "probe_sec": 0.087, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06096, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.07348, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -1.65373, "trunc_abs_frac": 6.26635, "mean_delta_full": -0.05151, "read_hit_rate": {"merge": 0.28512}, "mean_gate": {"merge": 0.18961}, "n_reads": 1843, "n_cand": 6464, "read_suppressed_frac": 0.102, "dropout_reads_logged": 1843, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 88.12701416015625, "n_util_obs": 1074, "util_count_spearman": 0.07487968163651323}, "probe_overhead_sec_total": 186.012, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.023141, "alpha": 0.03, "t": 1361.5} |
| {"event": "consolidate", "step": 3400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1387.8} |
| {"event": "probe", "step": 3400, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3306.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.189967, "cps": 25.828125, "mean_p": 0.203844, "boundary_entropy": 0.139202, "bpb": 1.97998}, "tr": {"chunks": 3714.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.205159, "cps": 29.015625, "mean_p": 0.210633, "boundary_entropy": 0.16481, "bpb": 2.14252}, "fi": {"chunks": 4112.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.218596, "cps": 32.125, "mean_p": 0.232884, "boundary_entropy": 0.158452, "bpb": 2.264101}}, "bpb": {"en": 1.97998, "tr": 2.14252, "fi": 2.2641}, "gini_cpc": 0.031098, "gini_cps": 0.048269, "bpic": 5.096299, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.079972306081163, "fi": 1.1507010079537006}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 89.39927673339844, "n_util_obs": 1105, "util_count_spearman": 0.06601652380550255}, "t": 1388.5} |
| {"event": "eval", "step": 3400, "phase": 2, "bpb": {"en": 1.99247, "tr": 2.15081, "fi": 2.2708}, "bytes_per_chunk": {"en": 5.028, "tr": 5.183, "fi": 4.884}, "bpb_mean": 2.13803, "frozen_inventory": true, "t": 1389.6} |
| {"event": "train", "step": 3400, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.4224, "bpb": 2.0521, "chunk_loss": 0.030351, "F": 0.19669, "G": 0.21165, "chunks": 201.41, "bytes_per_chunk": 5.084, "boundary_entropy": 0.14342, "router_noise": 0.0, "grad_norm": 0.748, "lr": 0.0001465, "router_trainable": true, "sec_per_step": 0.605, "probe_sec": 0.0377, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03166, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.08301, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 1.64753, "trunc_abs_frac": 1.13107, "mean_delta_full": 0.14132, "read_hit_rate": {"merge": 0.25764}, "mean_gate": {"merge": 0.19311}, "n_reads": 1644, "n_cand": 6381, "read_suppressed_frac": 0.104, "dropout_reads_logged": 1644, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 89.42521667480469, "n_util_obs": 1105, "util_count_spearman": 0.06623428283979257}, "probe_overhead_sec_total": 188.925, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.019761, "alpha": 0.03, "t": 1391.8} |
| {"event": "consolidate", "step": 3450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1413.4} |
| {"event": "train", "step": 3450, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.38672, "bpb": 2.00062, "chunk_loss": 0.030366, "F": 0.19821, "G": 0.21256, "chunks": 202.97, "bytes_per_chunk": 5.045, "boundary_entropy": 0.14576, "router_noise": 0.0, "grad_norm": 0.7753, "lr": 0.0001428, "router_trainable": true, "sec_per_step": 0.439, "probe_sec": 0.0331, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02389, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.15403, "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": {"merge": 0.2603}, "mean_gate": {"merge": 0.19253}, "n_reads": 1674, "n_cand": 6431, "read_suppressed_frac": 0.0956, "dropout_reads_logged": 1674, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 90.6904296875, "n_util_obs": 1121, "util_count_spearman": 0.069849575508986}, "probe_overhead_sec_total": 191.595, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.022818, "alpha": 0.03, "t": 1413.7} |
| {"event": "consolidate", "step": 3500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1431.5} |
| {"event": "train", "step": 3500, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.36311, "bpb": 1.96655, "chunk_loss": 0.030306, "F": 0.19345, "G": 0.20642, "chunks": 198.09, "bytes_per_chunk": 5.169, "boundary_entropy": 0.13805, "router_noise": 0.0, "grad_norm": 0.674, "lr": 0.0001391, "router_trainable": true, "sec_per_step": 0.362, "probe_sec": 0.0234, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01765, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.26433, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.15209, "trunc_abs_frac": 0.15209, "mean_delta_full": -0.42297, "read_hit_rate": {"merge": 0.28112}, "mean_gate": {"merge": 0.18978}, "n_reads": 1764, "n_cand": 6275, "read_suppressed_frac": 0.0998, "dropout_reads_logged": 1764, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 91.98207092285156, "n_util_obs": 1144, "util_count_spearman": 0.0814831153844608}, "probe_overhead_sec_total": 194.182, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.020363, "alpha": 0.03, "t": 1431.8} |
| {"event": "consolidate", "step": 3550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1448.9} |
| {"event": "train", "step": 3550, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.40913, "bpb": 2.03294, "chunk_loss": 0.030364, "F": 0.19711, "G": 0.21137, "chunks": 201.84, "bytes_per_chunk": 5.073, "boundary_entropy": 0.14032, "router_noise": 0.0, "grad_norm": 0.6662, "lr": 0.0001355, "router_trainable": true, "sec_per_step": 0.349, "probe_sec": 0.0656, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06075, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.05229, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.3566, "trunc_abs_frac": 4.28383, "mean_delta_full": -0.14443, "read_hit_rate": {"merge": 0.26099}, "mean_gate": {"merge": 0.20011}, "n_reads": 1669, "n_cand": 6395, "read_suppressed_frac": 0.0833, "dropout_reads_logged": 1669, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 93.26115417480469, "n_util_obs": 1166, "util_count_spearman": 0.09525113210786444}, "probe_overhead_sec_total": 197.064, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.022435, "alpha": 0.03, "t": 1449.3} |
| {"event": "consolidate", "step": 3600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1466.2} |
| {"event": "probe", "step": 3600, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3301.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.18968, "cps": 25.789062, "mean_p": 0.204085, "boundary_entropy": 0.139415, "bpb": 1.965945}, "tr": {"chunks": 3759.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.207645, "cps": 29.367188, "mean_p": 0.207889, "boundary_entropy": 0.153801, "bpb": 2.108647}, "fi": {"chunks": 4123.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.21918, "cps": 32.210938, "mean_p": 0.231229, "boundary_entropy": 0.152414, "bpb": 2.248988}}, "bpb": {"en": 1.96594, "tr": 2.10865, "fi": 2.24899}, "gini_cpc": 0.031901, "gini_cps": 0.049003, "bpic": 5.073057, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0947132386602678, "fi": 1.1555268658751574}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 94.49418640136719, "n_util_obs": 1190, "util_count_spearman": 0.09399562820910996}, "t": 1466.9} |
| {"event": "eval", "step": 3600, "phase": 2, "bpb": {"en": 1.98016, "tr": 2.11525, "fi": 2.25525}, "bytes_per_chunk": {"en": 5.072, "tr": 5.162, "fi": 4.851}, "bpb_mean": 2.11689, "frozen_inventory": true, "t": 1468.1} |
| {"event": "train", "step": 3600, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.40955, "bpb": 2.03356, "chunk_loss": 0.030404, "F": 0.19794, "G": 0.21236, "chunks": 202.69, "bytes_per_chunk": 5.052, "boundary_entropy": 0.14319, "router_noise": 0.0, "grad_norm": 0.6671, "lr": 0.0001319, "router_trainable": true, "sec_per_step": 0.403, "probe_sec": 0.0397, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03373, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.18833, "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": 0.73061, "trunc_abs_frac": 8.49271, "mean_delta_full": -0.64693, "read_hit_rate": {"merge": 0.24899}, "mean_gate": {"merge": 0.1861}, "n_reads": 1599, "n_cand": 6422, "read_suppressed_frac": 0.0982, "dropout_reads_logged": 1599, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 94.51860046386719, "n_util_obs": 1192, "util_count_spearman": 0.0954636684731947}, "probe_overhead_sec_total": 199.543, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.019111, "alpha": 0.03, "t": 1469.4} |
| {"event": "consolidate", "step": 3650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1486.9} |
| {"event": "train", "step": 3650, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.44426, "bpb": 2.08362, "chunk_loss": 0.030464, "F": 0.20151, "G": 0.2146, "chunks": 206.34, "bytes_per_chunk": 4.963, "boundary_entropy": 0.14189, "router_noise": 0.0, "grad_norm": 0.7053, "lr": 0.0001283, "router_trainable": true, "sec_per_step": 0.356, "probe_sec": 0.0801, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0674, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.04996, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 1.49005, "trunc_abs_frac": 0.55723, "mean_delta_full": 0.05829, "read_hit_rate": {"merge": 0.25921}, "mean_gate": {"merge": 0.19181}, "n_reads": 1695, "n_cand": 6539, "read_suppressed_frac": 0.1009, "dropout_reads_logged": 1695, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 95.80496215820312, "n_util_obs": 1204, "util_count_spearman": 0.09192803628581893}, "probe_overhead_sec_total": 202.505, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.020457, "alpha": 0.03, "t": 1487.2} |
| {"event": "consolidate", "step": 3700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1505.5} |
| {"event": "train", "step": 3700, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.39475, "bpb": 2.0122, "chunk_loss": 0.0304, "F": 0.19931, "G": 0.21352, "chunks": 204.09, "bytes_per_chunk": 5.017, "boundary_entropy": 0.14308, "router_noise": 0.0, "grad_norm": 0.6685, "lr": 0.0001247, "router_trainable": true, "sec_per_step": 0.372, "probe_sec": 0.0786, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06651, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.10948, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 91.7, "trunc_abs_frac": 1.15625, "mean_delta_full": 0.00046, "read_hit_rate": {"merge": 0.26983}, "mean_gate": {"merge": 0.18641}, "n_reads": 1745, "n_cand": 6467, "read_suppressed_frac": 0.0997, "dropout_reads_logged": 1745, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 97.07667541503906, "n_util_obs": 1232, "util_count_spearman": 0.1029027010572872}, "probe_overhead_sec_total": 205.18, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.023212, "alpha": 0.03, "t": 1505.8} |
| {"event": "consolidate", "step": 3750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1526.0} |
| {"event": "train", "step": 3750, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.37948, "bpb": 1.99017, "chunk_loss": 0.030371, "F": 0.19818, "G": 0.211, "chunks": 202.94, "bytes_per_chunk": 5.046, "boundary_entropy": 0.14161, "router_noise": 0.0, "grad_norm": 0.6663, "lr": 0.0001212, "router_trainable": true, "sec_per_step": 0.409, "probe_sec": 0.0451, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03797, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.06935, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": -0.2914, "trunc_abs_frac": 0.31182, "mean_delta_full": 0.18697, "read_hit_rate": {"merge": 0.26407}, "mean_gate": {"merge": 0.19397}, "n_reads": 1698, "n_cand": 6430, "read_suppressed_frac": 0.1007, "dropout_reads_logged": 1698, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 98.36012268066406, "n_util_obs": 1247, "util_count_spearman": 0.10099101727592402}, "probe_overhead_sec_total": 208.095, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.017964, "alpha": 0.03, "t": 1526.3} |
| {"event": "consolidate", "step": 3800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1544.5} |
| {"event": "probe", "step": 3800, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3276.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.188243, "cps": 25.59375, "mean_p": 0.202952, "boundary_entropy": 0.134424, "bpb": 1.951817}, "tr": {"chunks": 3625.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.200243, "cps": 28.320312, "mean_p": 0.207546, "boundary_entropy": 0.160325, "bpb": 2.101837}, "fi": {"chunks": 4168.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.221572, "cps": 32.5625, "mean_p": 0.231351, "boundary_entropy": 0.144533, "bpb": 2.230959}}, "bpb": {"en": 1.95182, "tr": 2.10184, "fi": 2.23096}, "gini_cpc": 0.036422, "gini_cps": 0.053724, "bpic": 5.125305, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0637453792593825, "fi": 1.1770530959303487}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 99.60874938964844, "n_util_obs": 1273, "util_count_spearman": 0.09640991238296526}, "t": 1545.2} |
| {"event": "eval", "step": 3800, "phase": 2, "bpb": {"en": 1.96649, "tr": 2.10909, "fi": 2.23606}, "bytes_per_chunk": {"en": 5.09, "tr": 5.352, "fi": 4.835}, "bpb_mean": 2.10388, "frozen_inventory": true, "t": 1546.4} |
| {"event": "train", "step": 3800, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.45274, "bpb": 2.09586, "chunk_loss": 0.030372, "F": 0.19769, "G": 0.21232, "chunks": 202.44, "bytes_per_chunk": 5.058, "boundary_entropy": 0.14119, "router_noise": 0.0, "grad_norm": 0.7067, "lr": 0.0001177, "router_trainable": true, "sec_per_step": 0.431, "probe_sec": 0.0287, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02349, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.04624, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": -0.88828, "trunc_abs_frac": 0.88828, "mean_delta_full": 0.06665, "read_hit_rate": {"merge": 0.22046}, "mean_gate": {"merge": 0.1748}, "n_reads": 1414, "n_cand": 6414, "read_suppressed_frac": 0.1089, "dropout_reads_logged": 1414, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 99.617919921875, "n_util_obs": 1273, "util_count_spearman": 0.0961562617888678}, "probe_overhead_sec_total": 210.918, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.021952, "alpha": 0.03, "t": 1547.8} |
| {"event": "consolidate", "step": 3850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1564.5} |
| {"event": "train", "step": 3850, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.55266, "bpb": 2.24001, "chunk_loss": 0.030618, "F": 0.19965, "G": 0.21397, "chunks": 204.44, "bytes_per_chunk": 5.009, "boundary_entropy": 0.14047, "router_noise": 0.0, "grad_norm": 0.7091, "lr": 0.0001142, "router_trainable": true, "sec_per_step": 0.338, "probe_sec": 0.023, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01852, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.09188, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.62966, "trunc_abs_frac": 0.59626, "mean_delta_full": 0.15704, "read_hit_rate": {"merge": 0.23835}, "mean_gate": {"merge": 0.19217}, "n_reads": 1544, "n_cand": 6478, "read_suppressed_frac": 0.1017, "dropout_reads_logged": 1544, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 100.90333557128906, "n_util_obs": 1294, "util_count_spearman": 0.10134528980715929}, "probe_overhead_sec_total": 213.386, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.019104, "alpha": 0.03, "t": 1564.7} |
| {"event": "consolidate", "step": 3900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1582.2} |
| {"event": "train", "step": 3900, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.43179, "bpb": 2.06564, "chunk_loss": 0.030466, "F": 0.19885, "G": 0.21214, "chunks": 203.62, "bytes_per_chunk": 5.029, "boundary_entropy": 0.13875, "router_noise": 0.0, "grad_norm": 0.7551, "lr": 0.0001108, "router_trainable": true, "sec_per_step": 0.357, "probe_sec": 0.0623, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05644, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.01582, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 1.26479, "trunc_abs_frac": 1.26479, "mean_delta_full": 0.61096, "read_hit_rate": {"merge": 0.23667}, "mean_gate": {"merge": 0.18538}, "n_reads": 1527, "n_cand": 6452, "read_suppressed_frac": 0.1015, "dropout_reads_logged": 1527, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 102.14251708984375, "n_util_obs": 1316, "util_count_spearman": 0.10038760089404501}, "probe_overhead_sec_total": 216.266, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.027427, "alpha": 0.03, "t": 1582.6} |
| {"event": "consolidate", "step": 3950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1598.9} |
| {"event": "train", "step": 3950, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.49034, "bpb": 2.1501, "chunk_loss": 0.030375, "F": 0.19583, "G": 0.20963, "chunks": 200.53, "bytes_per_chunk": 5.106, "boundary_entropy": 0.13905, "router_noise": 0.0, "grad_norm": 0.7701, "lr": 0.0001074, "router_trainable": true, "sec_per_step": 0.335, "probe_sec": 0.0265, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01986, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.10667, "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": 1.45885, "trunc_abs_frac": 1.50857, "mean_delta_full": -0.07218, "read_hit_rate": {"merge": 0.22985}, "mean_gate": {"merge": 0.17949}, "n_reads": 1460, "n_cand": 6352, "read_suppressed_frac": 0.0904, "dropout_reads_logged": 1460, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 103.42344665527344, "n_util_obs": 1336, "util_count_spearman": 0.1064481680876982}, "probe_overhead_sec_total": 218.855, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.017658, "alpha": 0.03, "t": 1599.3} |
| {"event": "consolidate", "step": 4000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1615.7} |
| {"event": "probe", "step": 4000, "phase": 2, "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": 0.203564, "boundary_entropy": 0.135302, "bpb": 1.94063}, "tr": {"chunks": 3726.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.205822, "cps": 29.109375, "mean_p": 0.206946, "boundary_entropy": 0.146786, "bpb": 2.085362}, "fi": {"chunks": 4283.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.227686, "cps": 33.460938, "mean_p": 0.235684, "boundary_entropy": 0.148365, "bpb": 2.204594}}, "bpb": {"en": 1.94063, "tr": 2.08536, "fi": 2.20459}, "gini_cpc": 0.042423, "gini_cps": 0.059618, "bpic": 5.028095, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.094051445453426, "fi": 1.2102682384691081}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 104.68721008300781, "n_util_obs": 1350, "util_count_spearman": 0.1059902119894246}, "t": 1616.4} |
| {"event": "eval", "step": 4000, "phase": 2, "bpb": {"en": 1.95523, "tr": 2.09214, "fi": 2.21102}, "bytes_per_chunk": {"en": 5.085, "tr": 5.201, "fi": 4.772}, "bpb_mean": 2.08613, "frozen_inventory": true, "t": 1617.7} |
| {"event": "train", "step": 4000, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.4173, "bpb": 2.04473, "chunk_loss": 0.030453, "F": 0.19717, "G": 0.21323, "chunks": 201.91, "bytes_per_chunk": 5.072, "boundary_entropy": 0.14043, "router_noise": 0.0, "grad_norm": 0.6745, "lr": 0.000104, "router_trainable": true, "sec_per_step": 0.393, "probe_sec": 0.0711, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06001, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.04455, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.53439, "trunc_abs_frac": 0.72874, "mean_delta_full": 0.51733, "read_hit_rate": {"merge": 0.26059}, "mean_gate": {"merge": 0.18212}, "n_reads": 1667, "n_cand": 6397, "read_suppressed_frac": 0.1092, "dropout_reads_logged": 1667, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 104.71353149414062, "n_util_obs": 1351, "util_count_spearman": 0.1040634371308208}, "probe_overhead_sec_total": 221.467, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.023372, "alpha": 0.03, "t": 1619.0} |
| {"event": "consolidate", "step": 4050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1635.8} |
| {"event": "train", "step": 4050, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.42573, "bpb": 2.05689, "chunk_loss": 0.030485, "F": 0.20203, "G": 0.21544, "chunks": 206.88, "bytes_per_chunk": 4.95, "boundary_entropy": 0.1341, "router_noise": 0.0, "grad_norm": 0.7095, "lr": 0.0001007, "router_trainable": true, "sec_per_step": 0.341, "probe_sec": 0.0228, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01818, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.1199, "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": -0.8721, "trunc_abs_frac": 0.8284, "mean_delta_full": 0.0579, "read_hit_rate": {"merge": 0.26514}, "mean_gate": {"merge": 0.19433}, "n_reads": 1738, "n_cand": 6555, "read_suppressed_frac": 0.0995, "dropout_reads_logged": 1738, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 105.96951293945312, "n_util_obs": 1370, "util_count_spearman": 0.10590002445606911}, "probe_overhead_sec_total": 224.022, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.020315, "alpha": 0.03, "t": 1636.0} |
| {"event": "probe", "step": 4096, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3275.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.188186, "cps": 25.585938, "mean_p": 0.20151, "boundary_entropy": 0.125237, "bpb": 1.93621}, "tr": {"chunks": 3668.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.202618, "cps": 28.65625, "mean_p": 0.207137, "boundary_entropy": 0.153843, "bpb": 2.082374}, "fi": {"chunks": 4159.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.221094, "cps": 32.492188, "mean_p": 0.231509, "boundary_entropy": 0.14224, "bpb": 2.200561}}, "bpb": {"en": 1.93621, "tr": 2.08237, "fi": 2.20056}, "gini_cpc": 0.035854, "gini_cps": 0.053084, "bpic": 5.11007, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0766922609512235, "fi": 1.1748701040198584}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 107.1265869140625, "n_util_obs": 1386, "util_count_spearman": 0.1075887057892853}, "t": 1651.9} |
| {"event": "consolidate", "step": 4100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1654.8} |
| {"event": "train", "step": 4100, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.36112, "bpb": 1.96369, "chunk_loss": 0.030387, "F": 0.19739, "G": 0.20971, "chunks": 202.12, "bytes_per_chunk": 5.066, "boundary_entropy": 0.13199, "router_noise": 0.0, "grad_norm": 0.7489, "lr": 9.75e-05, "router_trainable": true, "sec_per_step": 0.381, "probe_sec": 0.0389, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03337, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.00495, "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": -0.26715, "trunc_abs_frac": 1.29318, "mean_delta_full": 0.27583, "read_hit_rate": {"merge": 0.2728}, "mean_gate": {"merge": 0.18198}, "n_reads": 1747, "n_cand": 6404, "read_suppressed_frac": 0.1053, "dropout_reads_logged": 1747, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 107.24932861328125, "n_util_obs": 1389, "util_count_spearman": 0.10755115804733846}, "probe_overhead_sec_total": 226.617, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.022479, "alpha": 0.03, "t": 1655.1} |
| {"event": "consolidate", "step": 4150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1671.6} |
| {"event": "train", "step": 4150, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.42659, "bpb": 2.05814, "chunk_loss": 0.030579, "F": 0.20657, "G": 0.21753, "chunks": 211.53, "bytes_per_chunk": 4.841, "boundary_entropy": 0.13133, "router_noise": 0.0, "grad_norm": 0.7617, "lr": 9.43e-05, "router_trainable": true, "sec_per_step": 0.336, "probe_sec": 0.034, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02587, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.10907, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.7373, "trunc_abs_frac": 0.77692, "mean_delta_full": 0.25977, "read_hit_rate": {"merge": 0.26025}, "mean_gate": {"merge": 0.18637}, "n_reads": 1745, "n_cand": 6705, "read_suppressed_frac": 0.0923, "dropout_reads_logged": 1745, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 108.51741027832031, "n_util_obs": 1421, "util_count_spearman": 0.10878655452903643}, "probe_overhead_sec_total": 229.002, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.021313, "alpha": 0.03, "t": 1671.9} |
| {"event": "consolidate", "step": 4200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1688.5} |
| {"event": "probe", "step": 4200, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3280.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.188473, "cps": 25.625, "mean_p": 0.199985, "boundary_entropy": 0.12375, "bpb": 1.933819}, "tr": {"chunks": 3715.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.205215, "cps": 29.023438, "mean_p": 0.206179, "boundary_entropy": 0.143615, "bpb": 2.071053}, "fi": {"chunks": 4151.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.220669, "cps": 32.429688, "mean_p": 0.229863, "boundary_entropy": 0.140325, "bpb": 2.195667}}, "bpb": {"en": 1.93382, "tr": 2.07105, "fi": 2.19567}, "gini_cpc": 0.034937, "gini_cps": 0.052096, "bpic": 5.089898, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0888261513049313, "fi": 1.1708226796464445}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 109.77699279785156, "n_util_obs": 1448, "util_count_spearman": 0.10987705776556463}, "t": 1689.3} |
| {"event": "eval", "step": 4200, "phase": 2, "bpb": {"en": 1.94711, "tr": 2.07836, "fi": 2.20251}, "bytes_per_chunk": {"en": 5.082, "tr": 5.199, "fi": 4.868}, "bpb_mean": 2.076, "frozen_inventory": true, "t": 1690.4} |
| {"event": "phase_transition", "step": 4200, "phase": 3, "objective": "global", "router_trainable": true, "encoder_trainable": true, "frozen_router_params": 0, "t": 1691.1} |
| {"event": "train", "step": 4200, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35221, "bpb": 1.95083, "chunk_loss": 0.030262, "F": 0.1937, "G": 0.20417, "chunks": 198.34, "bytes_per_chunk": 5.163, "boundary_entropy": 0.12601, "router_noise": 0.0, "grad_norm": 0.718, "lr": 9.12e-05, "router_trainable": true, "sec_per_step": 0.397, "probe_sec": 0.0638, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05828, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.14409, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.31562, "trunc_abs_frac": 0.88796, "mean_delta_full": 0.24966, "read_hit_rate": {"merge": 0.26341}, "mean_gate": {"merge": 0.18986}, "n_reads": 1655, "n_cand": 6283, "read_suppressed_frac": 0.0997, "dropout_reads_logged": 1655, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 109.80287170410156, "n_util_obs": 1448, "util_count_spearman": 0.11008414712267979}, "probe_overhead_sec_total": 231.212, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.022381, "alpha": 0.03, "t": 1691.7} |
| {"event": "consolidate", "step": 4250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1708.1} |
| {"event": "train", "step": 4250, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.3853, "bpb": 1.99857, "chunk_loss": 0.030429, "F": 0.20166, "G": 0.21391, "chunks": 206.5, "bytes_per_chunk": 4.959, "boundary_entropy": 0.13006, "router_noise": 0.0, "grad_norm": 0.7068, "lr": 8.81e-05, "router_trainable": true, "sec_per_step": 0.336, "probe_sec": 0.1066, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10168, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.03139, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.2861, "trunc_abs_frac": 1.04996, "mean_delta_full": 0.16022, "read_hit_rate": {"merge": 0.26543}, "mean_gate": {"merge": 0.18961}, "n_reads": 1737, "n_cand": 6544, "read_suppressed_frac": 0.1077, "dropout_reads_logged": 1737, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 111.09506225585938, "n_util_obs": 1475, "util_count_spearman": 0.11251364704287889}, "probe_overhead_sec_total": 233.776, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.025638, "alpha": 0.03, "t": 1708.5} |
| {"event": "consolidate", "step": 4300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1726.1} |
| {"event": "train", "step": 4300, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38908, "bpb": 2.00402, "chunk_loss": 0.03025, "F": 0.19119, "G": 0.2043, "chunks": 195.78, "bytes_per_chunk": 5.23, "boundary_entropy": 0.13242, "router_noise": 0.0, "grad_norm": 0.7195, "lr": 8.5e-05, "router_trainable": true, "sec_per_step": 0.358, "probe_sec": 0.0694, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06278, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.42568, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 1.27116, "trunc_abs_frac": 1.02678, "mean_delta_full": -0.06807, "read_hit_rate": {"merge": 0.23915}, "mean_gate": {"merge": 0.18613}, "n_reads": 1483, "n_cand": 6201, "read_suppressed_frac": 0.1119, "dropout_reads_logged": 1483, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 112.37286376953125, "n_util_obs": 1503, "util_count_spearman": 0.10581114270177616}, "probe_overhead_sec_total": 236.804, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.02142, "alpha": 0.03, "t": 1726.4} |
| {"event": "consolidate", "step": 4350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1743.9} |
| {"event": "train", "step": 4350, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.41433, "bpb": 2.04045, "chunk_loss": 0.030319, "F": 0.19547, "G": 0.2081, "chunks": 200.16, "bytes_per_chunk": 5.116, "boundary_entropy": 0.12687, "router_noise": 0.0, "grad_norm": 0.678, "lr": 8.21e-05, "router_trainable": true, "sec_per_step": 0.357, "probe_sec": 0.0662, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05886, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.10036, "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": -1.92946, "trunc_abs_frac": 25.91735, "mean_delta_full": 0.0424, "read_hit_rate": {"merge": 0.24933}, "mean_gate": {"merge": 0.18232}, "n_reads": 1581, "n_cand": 6341, "read_suppressed_frac": 0.0904, "dropout_reads_logged": 1581, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 113.62376403808594, "n_util_obs": 1517, "util_count_spearman": 0.10450599671470534}, "probe_overhead_sec_total": 239.546, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.021074, "alpha": 0.03, "t": 1744.3} |
| {"event": "consolidate", "step": 4400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1764.5} |
| {"event": "probe", "step": 4400, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3279.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.188416, "cps": 25.617188, "mean_p": 0.199864, "boundary_entropy": 0.123753, "bpb": 1.923684}, "tr": {"chunks": 3695.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.20411, "cps": 28.867188, "mean_p": 0.204041, "boundary_entropy": 0.140504, "bpb": 2.056981}, "fi": {"chunks": 4131.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.219606, "cps": 32.273438, "mean_p": 0.23082, "boundary_entropy": 0.143188, "bpb": 2.17781}}, "bpb": {"en": 1.92368, "tr": 2.05698, "fi": 2.17781}, "gini_cpc": 0.033968, "gini_cps": 0.051148, "bpic": 5.10869, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0832946412818505, "fi": 1.1655368666296408}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 114.89256286621094, "n_util_obs": 1535, "util_count_spearman": 0.10211906426268302}, "t": 1765.2} |
| {"event": "eval", "step": 4400, "phase": 3, "bpb": {"en": 1.93671, "tr": 2.06365, "fi": 2.18603}, "bytes_per_chunk": {"en": 5.09, "tr": 5.254, "fi": 4.887}, "bpb_mean": 2.06213, "frozen_inventory": true, "t": 1766.4} |
| {"event": "train", "step": 4400, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.36448, "bpb": 1.96852, "chunk_loss": 0.030278, "F": 0.19467, "G": 0.20574, "chunks": 199.34, "bytes_per_chunk": 5.137, "boundary_entropy": 0.12597, "router_noise": 0.0, "grad_norm": 0.6622, "lr": 7.92e-05, "router_trainable": true, "sec_per_step": 0.47, "probe_sec": 0.0249, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01845, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.03776, "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": 0.74426, "trunc_abs_frac": 0.42594, "mean_delta_full": 0.41288, "read_hit_rate": {"merge": 0.24909}, "mean_gate": {"merge": 0.17446}, "n_reads": 1573, "n_cand": 6315, "read_suppressed_frac": 0.0909, "dropout_reads_logged": 1573, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 114.91461181640625, "n_util_obs": 1536, "util_count_spearman": 0.10164750252710568}, "probe_overhead_sec_total": 242.451, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.019803, "alpha": 0.03, "t": 1767.8} |
| {"event": "consolidate", "step": 4450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1785.1} |
| {"event": "train", "step": 4450, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35805, "bpb": 1.95925, "chunk_loss": 0.030248, "F": 0.19003, "G": 0.20161, "chunks": 194.59, "bytes_per_chunk": 5.262, "boundary_entropy": 0.12793, "router_noise": 0.0, "grad_norm": 0.6967, "lr": 7.63e-05, "router_trainable": true, "sec_per_step": 0.351, "probe_sec": 0.0303, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02587, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.00752, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 3.92112, "trunc_abs_frac": 0.92747, "mean_delta_full": 0.01584, "read_hit_rate": {"merge": 0.26441}, "mean_gate": {"merge": 0.18134}, "n_reads": 1629, "n_cand": 6161, "read_suppressed_frac": 0.1001, "dropout_reads_logged": 1629, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 116.19398498535156, "n_util_obs": 1560, "util_count_spearman": 0.10029409951245577}, "probe_overhead_sec_total": 245.147, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.017644, "alpha": 0.03, "t": 1785.3} |
| {"event": "consolidate", "step": 4500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1802.5} |
| {"event": "train", "step": 4500, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.33069, "bpb": 1.91978, "chunk_loss": 0.030349, "F": 0.19836, "G": 0.2095, "chunks": 203.12, "bytes_per_chunk": 5.041, "boundary_entropy": 0.12558, "router_noise": 0.0, "grad_norm": 0.6522, "lr": 7.36e-05, "router_trainable": true, "sec_per_step": 0.35, "probe_sec": 0.0801, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07507, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.18213, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.77296, "trunc_abs_frac": 0.43356, "mean_delta_full": 0.4176, "read_hit_rate": {"merge": 0.24285}, "mean_gate": {"merge": 0.17204}, "n_reads": 1563, "n_cand": 6436, "read_suppressed_frac": 0.0825, "dropout_reads_logged": 1563, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 117.45849609375, "n_util_obs": 1582, "util_count_spearman": 0.10556216717683242}, "probe_overhead_sec_total": 247.537, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.024981, "alpha": 0.03, "t": 1802.8} |
| {"event": "consolidate", "step": 4550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1824.1} |
| {"event": "train", "step": 4550, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.36342, "bpb": 1.967, "chunk_loss": 0.030317, "F": 0.19696, "G": 0.20789, "chunks": 201.69, "bytes_per_chunk": 5.077, "boundary_entropy": 0.12452, "router_noise": 0.0, "grad_norm": 0.6685, "lr": 7.09e-05, "router_trainable": true, "sec_per_step": 0.432, "probe_sec": 0.0432, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03255, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.23284, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.8447, "trunc_abs_frac": 0.77712, "mean_delta_full": 0.59885, "read_hit_rate": {"merge": 0.24773}, "mean_gate": {"merge": 0.18189}, "n_reads": 1583, "n_cand": 6390, "read_suppressed_frac": 0.1118, "dropout_reads_logged": 1583, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 118.77423095703125, "n_util_obs": 1597, "util_count_spearman": 0.10631220518560294}, "probe_overhead_sec_total": 250.065, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.024183, "alpha": 0.03, "t": 1824.4} |
| {"event": "consolidate", "step": 4600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1848.1} |
| {"event": "probe", "step": 4600, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3318.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.190657, "cps": 25.921875, "mean_p": 0.201504, "boundary_entropy": 0.119468, "bpb": 1.914613}, "tr": {"chunks": 3746.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.206927, "cps": 29.265625, "mean_p": 0.206148, "boundary_entropy": 0.138989, "bpb": 2.045006}, "fi": {"chunks": 4160.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.221147, "cps": 32.5, "mean_p": 0.230911, "boundary_entropy": 0.135208, "bpb": 2.168408}}, "bpb": {"en": 1.91461, "tr": 2.04501, "fi": 2.16841}, "gini_cpc": 0.032853, "gini_cps": 0.050012, "bpic": 5.054526, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0853378782192595, "fi": 1.1599230683674273}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 120.0291748046875, "n_util_obs": 1621, "util_count_spearman": 0.1092461686859911}, "t": 1848.9} |
| {"event": "eval", "step": 4600, "phase": 3, "bpb": {"en": 1.92745, "tr": 2.05083, "fi": 2.17446}, "bytes_per_chunk": {"en": 5.043, "tr": 5.192, "fi": 4.844}, "bpb_mean": 2.05091, "frozen_inventory": true, "t": 1850.5} |
| {"event": "train", "step": 4600, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.36492, "bpb": 1.96917, "chunk_loss": 0.030504, "F": 0.20596, "G": 0.21612, "chunks": 210.91, "bytes_per_chunk": 4.855, "boundary_entropy": 0.12694, "router_noise": 0.0, "grad_norm": 0.7246, "lr": 6.82e-05, "router_trainable": true, "sec_per_step": 0.572, "probe_sec": 0.0336, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02112, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.31134, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 1.4167, "trunc_abs_frac": 1.4167, "mean_delta_full": 0.31726, "read_hit_rate": {"merge": 0.28003}, "mean_gate": {"merge": 0.18204}, "n_reads": 1872, "n_cand": 6685, "read_suppressed_frac": 0.0994, "dropout_reads_logged": 1872, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 120.05876159667969, "n_util_obs": 1623, "util_count_spearman": 0.10927570387331804}, "probe_overhead_sec_total": 252.77, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.016045, "alpha": 0.03, "t": 1853.0} |
| {"event": "consolidate", "step": 4650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1873.9} |
| {"event": "train", "step": 4650, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.39243, "bpb": 2.00885, "chunk_loss": 0.030439, "F": 0.20013, "G": 0.2119, "chunks": 204.94, "bytes_per_chunk": 4.997, "boundary_entropy": 0.12406, "router_noise": 0.0, "grad_norm": 0.6763, "lr": 6.57e-05, "router_trainable": true, "sec_per_step": 0.422, "probe_sec": 0.0228, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01784, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.06788, "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": -0.30417, "trunc_abs_frac": 0.51783, "mean_delta_full": 0.09521, "read_hit_rate": {"merge": 0.25839}, "mean_gate": {"merge": 0.1861}, "n_reads": 1678, "n_cand": 6494, "read_suppressed_frac": 0.1055, "dropout_reads_logged": 1678, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 121.31338500976562, "n_util_obs": 1650, "util_count_spearman": 0.11008577977365326}, "probe_overhead_sec_total": 254.633, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.024637, "alpha": 0.03, "t": 1874.1} |
| {"event": "consolidate", "step": 4700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1894.1} |
| {"event": "train", "step": 4700, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.42279, "bpb": 2.05266, "chunk_loss": 0.030577, "F": 0.20554, "G": 0.21664, "chunks": 210.47, "bytes_per_chunk": 4.865, "boundary_entropy": 0.12841, "router_noise": 0.0, "grad_norm": 0.6952, "lr": 6.32e-05, "router_trainable": true, "sec_per_step": 0.403, "probe_sec": 0.0235, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01863, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.05207, "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": 0.87263, "trunc_abs_frac": 0.58, "mean_delta_full": 0.3835, "read_hit_rate": {"merge": 0.27387}, "mean_gate": {"merge": 0.1743}, "n_reads": 1827, "n_cand": 6671, "read_suppressed_frac": 0.087, "dropout_reads_logged": 1827, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 122.60098266601562, "n_util_obs": 1685, "util_count_spearman": 0.11223079124171084}, "probe_overhead_sec_total": 257.344, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.02239, "alpha": 0.03, "t": 1894.3} |
| {"event": "consolidate", "step": 4750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1921.9} |
| {"event": "train", "step": 4750, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.40545, "bpb": 2.02764, "chunk_loss": 0.030409, "F": 0.20148, "G": 0.21159, "chunks": 206.31, "bytes_per_chunk": 4.963, "boundary_entropy": 0.12729, "router_noise": 0.0, "grad_norm": 0.6762, "lr": 6.08e-05, "router_trainable": true, "sec_per_step": 0.567, "probe_sec": 0.0439, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02586, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.05067, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.64129, "trunc_abs_frac": 0.70529, "mean_delta_full": -0.40314, "read_hit_rate": {"merge": 0.26354}, "mean_gate": {"merge": 0.17124}, "n_reads": 1723, "n_cand": 6538, "read_suppressed_frac": 0.0934, "dropout_reads_logged": 1723, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 123.86383056640625, "n_util_obs": 1713, "util_count_spearman": 0.11716596412335595}, "probe_overhead_sec_total": 260.319, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.021102, "alpha": 0.03, "t": 1922.6} |
| {"event": "consolidate", "step": 4800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1950.0} |
| {"event": "probe", "step": 4800, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3275.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.188186, "cps": 25.585938, "mean_p": 0.198774, "boundary_entropy": 0.120966, "bpb": 1.906227}, "tr": {"chunks": 3713.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.205104, "cps": 29.007812, "mean_p": 0.204731, "boundary_entropy": 0.140577, "bpb": 2.034734}, "fi": {"chunks": 4116.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.218808, "cps": 32.15625, "mean_p": 0.229056, "boundary_entropy": 0.134668, "bpb": 2.154861}}, "bpb": {"en": 1.90623, "tr": 2.03473, "fi": 2.15486}, "gini_cpc": 0.033352, "gini_cps": 0.050492, "bpic": 5.10915, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.089901408100298, "fi": 1.1627230940480253}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 125.11738586425781, "n_util_obs": 1738, "util_count_spearman": 0.11754910689019789}, "t": 1950.8} |
| {"event": "eval", "step": 4800, "phase": 3, "bpb": {"en": 1.91996, "tr": 2.0408, "fi": 2.16111}, "bytes_per_chunk": {"en": 5.093, "tr": 5.211, "fi": 4.88}, "bpb_mean": 2.04062, "frozen_inventory": true, "t": 1951.8} |
| {"event": "train", "step": 4800, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38087, "bpb": 1.99217, "chunk_loss": 0.030405, "F": 0.20044, "G": 0.21075, "chunks": 205.25, "bytes_per_chunk": 4.989, "boundary_entropy": 0.12371, "router_noise": 0.0, "grad_norm": 0.6648, "lr": 5.85e-05, "router_trainable": true, "sec_per_step": 0.613, "probe_sec": 0.0578, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05219, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.11251, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 2.47386, "trunc_abs_frac": 0.87528, "mean_delta_full": 0.06521, "read_hit_rate": {"merge": 0.25384}, "mean_gate": {"merge": 0.16924}, "n_reads": 1651, "n_cand": 6504, "read_suppressed_frac": 0.1163, "dropout_reads_logged": 1651, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 125.1400146484375, "n_util_obs": 1738, "util_count_spearman": 0.11749337794480237}, "probe_overhead_sec_total": 263.137, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.024114, "alpha": 0.03, "t": 1953.2} |
| {"event": "consolidate", "step": 4850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1973.4} |
| {"event": "train", "step": 4850, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35198, "bpb": 1.9505, "chunk_loss": 0.03036, "F": 0.19736, "G": 0.2087, "chunks": 202.09, "bytes_per_chunk": 5.067, "boundary_entropy": 0.1237, "router_noise": 0.0, "grad_norm": 0.6812, "lr": 5.62e-05, "router_trainable": true, "sec_per_step": 0.427, "probe_sec": 0.0426, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02369, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.03858, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.57431, "trunc_abs_frac": 0.57736, "mean_delta_full": -1.21051, "read_hit_rate": {"merge": 0.25754}, "mean_gate": {"merge": 0.18688}, "n_reads": 1649, "n_cand": 6403, "read_suppressed_frac": 0.1073, "dropout_reads_logged": 1649, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 126.42037963867188, "n_util_obs": 1756, "util_count_spearman": 0.11402040640296933}, "probe_overhead_sec_total": 265.806, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.019754, "alpha": 0.03, "t": 1974.6} |
| {"event": "consolidate", "step": 4900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1991.5} |
| {"event": "train", "step": 4900, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38597, "bpb": 1.99954, "chunk_loss": 0.030376, "F": 0.19336, "G": 0.20479, "chunks": 198.0, "bytes_per_chunk": 5.172, "boundary_entropy": 0.12353, "router_noise": 0.0, "grad_norm": 0.6994, "lr": 5.41e-05, "router_trainable": true, "sec_per_step": 0.345, "probe_sec": 0.0769, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06873, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.10122, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.58514, "trunc_abs_frac": 0.84009, "mean_delta_full": 0.3952, "read_hit_rate": {"merge": 0.23039}, "mean_gate": {"merge": 0.17787}, "n_reads": 1445, "n_cand": 6272, "read_suppressed_frac": 0.1024, "dropout_reads_logged": 1445, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 127.69465637207031, "n_util_obs": 1787, "util_count_spearman": 0.11567199224285102}, "probe_overhead_sec_total": 268.487, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.020677, "alpha": 0.03, "t": 1991.8} |
| {"event": "consolidate", "step": 4950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2009.8} |
| {"event": "train", "step": 4950, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.34263, "bpb": 1.937, "chunk_loss": 0.030324, "F": 0.19504, "G": 0.20553, "chunks": 199.72, "bytes_per_chunk": 5.127, "boundary_entropy": 0.12365, "router_noise": 0.0, "grad_norm": 0.6788, "lr": 5.2e-05, "router_trainable": true, "sec_per_step": 0.367, "probe_sec": 0.0416, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03666, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.04708, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -0.74702, "trunc_abs_frac": 3.02326, "mean_delta_full": -0.07888, "read_hit_rate": {"merge": 0.2581}, "mean_gate": {"merge": 0.17553}, "n_reads": 1633, "n_cand": 6327, "read_suppressed_frac": 0.0931, "dropout_reads_logged": 1633, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 128.97898864746094, "n_util_obs": 1804, "util_count_spearman": 0.11295044804851931}, "probe_overhead_sec_total": 271.528, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.018485, "alpha": 0.03, "t": 2010.1} |
| {"event": "consolidate", "step": 5000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2027.3} |
| {"event": "probe", "step": 5000, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3296.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.189393, "cps": 25.75, "mean_p": 0.200522, "boundary_entropy": 0.114936, "bpb": 1.897941}, "tr": {"chunks": 3710.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.204938, "cps": 28.984375, "mean_p": 0.206797, "boundary_entropy": 0.138529, "bpb": 2.02743}, "fi": {"chunks": 4144.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.220297, "cps": 32.375, "mean_p": 0.229701, "boundary_entropy": 0.129709, "bpb": 2.141503}}, "bpb": {"en": 1.89794, "tr": 2.02743, "fi": 2.1415}, "gini_cpc": 0.033521, "gini_cps": 0.050703, "bpic": 5.088072, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0820822555828058, "fi": 1.1631742530320774}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 130.23460388183594, "n_util_obs": 1831, "util_count_spearman": 0.11044934721147304}, "t": 2028.0} |
| {"event": "eval", "step": 5000, "phase": 3, "bpb": {"en": 1.91207, "tr": 2.03419, "fi": 2.14863}, "bytes_per_chunk": {"en": 5.058, "tr": 5.243, "fi": 4.824}, "bpb_mean": 2.03163, "frozen_inventory": true, "t": 2029.2} |
| {"event": "train", "step": 5000, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.33455, "bpb": 1.92534, "chunk_loss": 0.030355, "F": 0.19803, "G": 0.20838, "chunks": 202.78, "bytes_per_chunk": 5.05, "boundary_entropy": 0.11876, "router_noise": 0.0, "grad_norm": 0.6467, "lr": 5e-05, "router_trainable": true, "sec_per_step": 0.41, "probe_sec": 0.1487, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.13928, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.4154, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.95898, "trunc_abs_frac": 1.00641, "mean_delta_full": 0.80493, "read_hit_rate": {"merge": 0.26148}, "mean_gate": {"merge": 0.16125}, "n_reads": 1680, "n_cand": 6425, "read_suppressed_frac": 0.0976, "dropout_reads_logged": 1680, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 130.25942993164062, "n_util_obs": 1832, "util_count_spearman": 0.1092477213555317}, "probe_overhead_sec_total": 274.722, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.026085, "alpha": 0.03, "t": 2030.6} |
| {"event": "consolidate", "step": 5050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2047.6} |
| {"event": "train", "step": 5050, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.3527, "bpb": 1.95153, "chunk_loss": 0.030407, "F": 0.20071, "G": 0.21056, "chunks": 205.53, "bytes_per_chunk": 4.982, "boundary_entropy": 0.11961, "router_noise": 0.0, "grad_norm": 0.676, "lr": 4.81e-05, "router_trainable": true, "sec_per_step": 0.346, "probe_sec": 0.033, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02888, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.20617, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 1.45464, "trunc_abs_frac": 1.13681, "mean_delta_full": 0.26489, "read_hit_rate": {"merge": 0.26685}, "mean_gate": {"merge": 0.16761}, "n_reads": 1738, "n_cand": 6513, "read_suppressed_frac": 0.0972, "dropout_reads_logged": 1738, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 131.53981018066406, "n_util_obs": 1855, "util_count_spearman": 0.10466188515506364}, "probe_overhead_sec_total": 277.261, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.021115, "alpha": 0.03, "t": 2047.9} |
| {"event": "consolidate", "step": 5100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2064.4} |
| {"event": "train", "step": 5100, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.37845, "bpb": 1.98868, "chunk_loss": 0.03045, "F": 0.20163, "G": 0.21265, "chunks": 206.47, "bytes_per_chunk": 4.96, "boundary_entropy": 0.12419, "router_noise": 0.0, "grad_norm": 0.7185, "lr": 4.63e-05, "router_trainable": true, "sec_per_step": 0.337, "probe_sec": 0.0301, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02344, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.0137, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.5769, "trunc_abs_frac": 0.67897, "mean_delta_full": 0.20874, "read_hit_rate": {"merge": 0.26578}, "mean_gate": {"merge": 0.17339}, "n_reads": 1739, "n_cand": 6543, "read_suppressed_frac": 0.1001, "dropout_reads_logged": 1739, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 132.83441162109375, "n_util_obs": 1887, "util_count_spearman": 0.10691764431924684}, "probe_overhead_sec_total": 280.216, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.019957, "alpha": 0.03, "t": 2064.7} |
| {"event": "consolidate", "step": 5150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2081.5} |
| {"event": "train", "step": 5150, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.37227, "bpb": 1.97977, "chunk_loss": 0.030524, "F": 0.20499, "G": 0.21526, "chunks": 209.91, "bytes_per_chunk": 4.878, "boundary_entropy": 0.12424, "router_noise": 0.0, "grad_norm": 0.7314, "lr": 4.45e-05, "router_trainable": true, "sec_per_step": 0.347, "probe_sec": 0.0431, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03522, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.26997, "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": 1.19282, "trunc_abs_frac": 1.25938, "mean_delta_full": 0.4517, "read_hit_rate": {"merge": 0.29299}, "mean_gate": {"merge": 0.16652}, "n_reads": 1949, "n_cand": 6652, "read_suppressed_frac": 0.0959, "dropout_reads_logged": 1949, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 134.15728759765625, "n_util_obs": 1909, "util_count_spearman": 0.10471726968167486}, "probe_overhead_sec_total": 283.011, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.026963, "alpha": 0.03, "t": 2082.1} |
| {"event": "consolidate", "step": 5200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2099.7} |
| {"event": "probe", "step": 5200, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3298.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.189508, "cps": 25.765625, "mean_p": 0.199325, "boundary_entropy": 0.117289, "bpb": 1.891727}, "tr": {"chunks": 3699.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.204331, "cps": 28.898438, "mean_p": 0.203033, "boundary_entropy": 0.134033, "bpb": 2.019832}, "fi": {"chunks": 4142.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.22019, "cps": 32.359375, "mean_p": 0.229298, "boundary_entropy": 0.130182, "bpb": 2.13335}}, "bpb": {"en": 1.89173, "tr": 2.01983, "fi": 2.13335}, "gini_cpc": 0.033313, "gini_cps": 0.050513, "bpic": 5.093096, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0782196659389283, "fi": 1.1619078343352192}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 135.4091033935547, "n_util_obs": 1931, "util_count_spearman": 0.09965724888438417}, "t": 2100.5} |
| {"event": "eval", "step": 5200, "phase": 3, "bpb": {"en": 1.90699, "tr": 2.02615, "fi": 2.1399}, "bytes_per_chunk": {"en": 5.078, "tr": 5.226, "fi": 4.849}, "bpb_mean": 2.02435, "frozen_inventory": true, "t": 2101.7} |
| {"event": "train", "step": 5200, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32613, "bpb": 1.91319, "chunk_loss": 0.030275, "F": 0.19339, "G": 0.20441, "chunks": 198.03, "bytes_per_chunk": 5.171, "boundary_entropy": 0.12082, "router_noise": 0.0, "grad_norm": 0.6639, "lr": 4.29e-05, "router_trainable": true, "sec_per_step": 0.42, "probe_sec": 0.0623, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05309, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.28636, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.98451, "trunc_abs_frac": 1.04664, "mean_delta_full": 0.38187, "read_hit_rate": {"merge": 0.25729}, "mean_gate": {"merge": 0.16253}, "n_reads": 1614, "n_cand": 6273, "read_suppressed_frac": 0.1041, "dropout_reads_logged": 1614, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 135.43507385253906, "n_util_obs": 1931, "util_count_spearman": 0.09988883768138565}, "probe_overhead_sec_total": 285.645, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.018038, "alpha": 0.03, "t": 2103.0} |
| {"event": "consolidate", "step": 5250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2120.9} |
| {"event": "train", "step": 5250, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32767, "bpb": 1.91542, "chunk_loss": 0.03033, "F": 0.19565, "G": 0.20519, "chunks": 200.34, "bytes_per_chunk": 5.111, "boundary_entropy": 0.12002, "router_noise": 0.0, "grad_norm": 0.6567, "lr": 4.14e-05, "router_trainable": true, "sec_per_step": 0.363, "probe_sec": 0.0674, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06176, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.09002, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.52051, "trunc_abs_frac": 0.56467, "mean_delta_full": 0.49261, "read_hit_rate": {"merge": 0.2576}, "mean_gate": {"merge": 0.1651}, "n_reads": 1635, "n_cand": 6347, "read_suppressed_frac": 0.1113, "dropout_reads_logged": 1635, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 136.7294464111328, "n_util_obs": 1952, "util_count_spearman": 0.09652103675742811}, "probe_overhead_sec_total": 288.682, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.021593, "alpha": 0.03, "t": 2121.2} |
| {"event": "consolidate", "step": 5300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2139.0} |
| {"event": "train", "step": 5300, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38446, "bpb": 1.99735, "chunk_loss": 0.030366, "F": 0.19968, "G": 0.20978, "chunks": 204.47, "bytes_per_chunk": 5.008, "boundary_entropy": 0.11864, "router_noise": 0.0, "grad_norm": 0.6859, "lr": 3.99e-05, "router_trainable": true, "sec_per_step": 0.363, "probe_sec": 0.0721, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06249, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.15712, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 0.39625, "trunc_abs_frac": 0.66698, "mean_delta_full": 0.23033, "read_hit_rate": {"merge": 0.25421}, "mean_gate": {"merge": 0.17583}, "n_reads": 1647, "n_cand": 6479, "read_suppressed_frac": 0.0953, "dropout_reads_logged": 1647, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 138.01927185058594, "n_util_obs": 1973, "util_count_spearman": 0.09936929964521267}, "probe_overhead_sec_total": 291.532, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.021092, "alpha": 0.03, "t": 2139.4} |
| {"event": "consolidate", "step": 5350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2156.8} |
| {"event": "train", "step": 5350, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.31224, "bpb": 1.89317, "chunk_loss": 0.030245, "F": 0.19226, "G": 0.20252, "chunks": 196.88, "bytes_per_chunk": 5.201, "boundary_entropy": 0.11888, "router_noise": 0.0, "grad_norm": 0.7243, "lr": 3.86e-05, "router_trainable": true, "sec_per_step": 0.359, "probe_sec": 0.0645, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0552, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.31251, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 1.1209, "trunc_abs_frac": 0.62021, "mean_delta_full": 0.31589, "read_hit_rate": {"merge": 0.24936}, "mean_gate": {"merge": 0.16726}, "n_reads": 1555, "n_cand": 6236, "read_suppressed_frac": 0.1035, "dropout_reads_logged": 1555, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 139.33413696289062, "n_util_obs": 1995, "util_count_spearman": 0.0917565650690771}, "probe_overhead_sec_total": 294.499, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.01636, "alpha": 0.03, "t": 2157.3} |
| {"event": "consolidate", "step": 5400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2174.4} |
| {"event": "probe", "step": 5400, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3314.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.190427, "cps": 25.890625, "mean_p": 0.20017, "boundary_entropy": 0.116469, "bpb": 1.887393}, "tr": {"chunks": 3672.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.202839, "cps": 28.6875, "mean_p": 0.205036, "boundary_entropy": 0.137847, "bpb": 2.01364}, "fi": {"chunks": 4136.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.219871, "cps": 32.3125, "mean_p": 0.229393, "boundary_entropy": 0.128755, "bpb": 2.12869}}, "bpb": {"en": 1.88739, "tr": 2.01364, "fi": 2.12869}, "gini_cpc": 0.032015, "gini_cps": 0.049272, "bpic": 5.100881, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0651817996737036, "fi": 1.154623155271282}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 140.6128692626953, "n_util_obs": 2022, "util_count_spearman": 0.08754568148572581}, "t": 2175.0} |
| {"event": "eval", "step": 5400, "phase": 3, "bpb": {"en": 1.9012, "tr": 2.01976, "fi": 2.13585}, "bytes_per_chunk": {"en": 5.068, "tr": 5.247, "fi": 4.844}, "bpb_mean": 2.01893, "frozen_inventory": true, "t": 2176.1} |
| {"event": "train", "step": 5400, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38191, "bpb": 1.99368, "chunk_loss": 0.030617, "F": 0.20401, "G": 0.21567, "chunks": 208.91, "bytes_per_chunk": 4.902, "boundary_entropy": 0.12231, "router_noise": 0.0, "grad_norm": 0.7383, "lr": 3.73e-05, "router_trainable": true, "sec_per_step": 0.402, "probe_sec": 0.08, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07128, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.03316, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.22762, "trunc_abs_frac": 0.67658, "mean_delta_full": -0.10056, "read_hit_rate": {"merge": 0.25721}, "mean_gate": {"merge": 0.16637}, "n_reads": 1703, "n_cand": 6621, "read_suppressed_frac": 0.0963, "dropout_reads_logged": 1703, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 140.63148498535156, "n_util_obs": 2022, "util_count_spearman": 0.08751347660459981}, "probe_overhead_sec_total": 297.221, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.016158, "alpha": 0.03, "t": 2177.4} |
| {"event": "consolidate", "step": 5450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2195.4} |
| {"event": "train", "step": 5450, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32085, "bpb": 1.90558, "chunk_loss": 0.030542, "F": 0.20462, "G": 0.21425, "chunks": 209.53, "bytes_per_chunk": 4.887, "boundary_entropy": 0.12008, "router_noise": 0.0, "grad_norm": 0.7017, "lr": 3.62e-05, "router_trainable": true, "sec_per_step": 0.366, "probe_sec": 0.0377, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03041, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.20285, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.48874, "trunc_abs_frac": 0.58038, "mean_delta_full": 0.15433, "read_hit_rate": {"merge": 0.28189}, "mean_gate": {"merge": 0.17443}, "n_reads": 1872, "n_cand": 6641, "read_suppressed_frac": 0.0994, "dropout_reads_logged": 1872, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 141.93911743164062, "n_util_obs": 2042, "util_count_spearman": 0.08945841260303732}, "probe_overhead_sec_total": 300.262, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.018617, "alpha": 0.03, "t": 2195.7} |
| {"event": "consolidate", "step": 5500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2212.1} |
| {"event": "train", "step": 5500, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.30814, "bpb": 1.88725, "chunk_loss": 0.030396, "F": 0.20053, "G": 0.21042, "chunks": 205.34, "bytes_per_chunk": 4.987, "boundary_entropy": 0.11851, "router_noise": 0.0, "grad_norm": 0.6802, "lr": 3.51e-05, "router_trainable": true, "sec_per_step": 0.344, "probe_sec": 0.1418, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.13659, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.12832, "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": 0.40854, "trunc_abs_frac": 0.45071, "mean_delta_full": -0.3682, "read_hit_rate": {"merge": 0.27524}, "mean_gate": {"merge": 0.17205}, "n_reads": 1791, "n_cand": 6507, "read_suppressed_frac": 0.0905, "dropout_reads_logged": 1791, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 143.22975158691406, "n_util_obs": 2066, "util_count_spearman": 0.09014213579816857}, "probe_overhead_sec_total": 302.811, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.021379, "alpha": 0.03, "t": 2212.9} |
| {"event": "consolidate", "step": 5550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2229.7} |
| {"event": "train", "step": 5550, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35872, "bpb": 1.96022, "chunk_loss": 0.030333, "F": 0.19647, "G": 0.20556, "chunks": 201.19, "bytes_per_chunk": 5.09, "boundary_entropy": 0.11545, "router_noise": 0.0, "grad_norm": 0.7237, "lr": 3.41e-05, "router_trainable": true, "sec_per_step": 0.346, "probe_sec": 0.0238, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01779, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.10958, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 6.48322, "trunc_abs_frac": 0.79921, "mean_delta_full": 0.03577, "read_hit_rate": {"merge": 0.26498}, "mean_gate": {"merge": 0.16849}, "n_reads": 1689, "n_cand": 6374, "read_suppressed_frac": 0.1024, "dropout_reads_logged": 1689, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 144.5160369873047, "n_util_obs": 2083, "util_count_spearman": 0.08947427334360465}, "probe_overhead_sec_total": 305.933, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.018009, "alpha": 0.03, "t": 2230.2} |
| {"event": "consolidate", "step": 5600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2247.0} |
| {"event": "probe", "step": 5600, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3297.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.18945, "cps": 25.757812, "mean_p": 0.199687, "boundary_entropy": 0.114427, "bpb": 1.881645}, "tr": {"chunks": 3718.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.20538, "cps": 29.046875, "mean_p": 0.202861, "boundary_entropy": 0.130432, "bpb": 2.007828}, "fi": {"chunks": 4179.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.222157, "cps": 32.648438, "mean_p": 0.230636, "boundary_entropy": 0.127553, "bpb": 2.119366}}, "bpb": {"en": 1.88165, "tr": 2.00783, "fi": 2.11937}, "gini_cpc": 0.035341, "gini_cps": 0.052528, "bpic": 5.068072, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0840866767994304, "fi": 1.172642582416373}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 145.7813720703125, "n_util_obs": 2102, "util_count_spearman": 0.08596173206454293}, "t": 2247.8} |
| {"event": "eval", "step": 5600, "phase": 3, "bpb": {"en": 1.89681, "tr": 2.01387, "fi": 2.12722}, "bytes_per_chunk": {"en": 5.051, "tr": 5.232, "fi": 4.819}, "bpb_mean": 2.01264, "frozen_inventory": true, "t": 2249.1} |
| {"event": "train", "step": 5600, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35221, "bpb": 1.95082, "chunk_loss": 0.03038, "F": 0.1976, "G": 0.20924, "chunks": 202.34, "bytes_per_chunk": 5.061, "boundary_entropy": 0.12116, "router_noise": 0.0, "grad_norm": 0.764, "lr": 3.33e-05, "router_trainable": true, "sec_per_step": 0.405, "probe_sec": 0.0422, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03724, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.01449, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.06881, "trunc_abs_frac": 1.62976, "mean_delta_full": -0.11554, "read_hit_rate": {"merge": 0.24208}, "mean_gate": {"merge": 0.17559}, "n_reads": 1552, "n_cand": 6411, "read_suppressed_frac": 0.0889, "dropout_reads_logged": 1552, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 145.80572509765625, "n_util_obs": 2102, "util_count_spearman": 0.08647346063348575}, "probe_overhead_sec_total": 308.653, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.022684, "alpha": 0.03, "t": 2250.5} |
| {"event": "consolidate", "step": 5650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2268.1} |
| {"event": "train", "step": 5650, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.33878, "bpb": 1.93145, "chunk_loss": 0.030413, "F": 0.20114, "G": 0.21094, "chunks": 205.97, "bytes_per_chunk": 4.972, "boundary_entropy": 0.11954, "router_noise": 0.0, "grad_norm": 0.7239, "lr": 3.25e-05, "router_trainable": true, "sec_per_step": 0.358, "probe_sec": 0.0649, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06048, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.00648, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.94765, "trunc_abs_frac": 0.55951, "mean_delta_full": -0.32483, "read_hit_rate": {"merge": 0.26444}, "mean_gate": {"merge": 0.15703}, "n_reads": 1726, "n_cand": 6527, "read_suppressed_frac": 0.0979, "dropout_reads_logged": 1726, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 147.0949249267578, "n_util_obs": 2127, "util_count_spearman": 0.08806165292717615}, "probe_overhead_sec_total": 311.836, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.021128, "alpha": 0.03, "t": 2268.4} |
| {"event": "consolidate", "step": 5700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2284.9} |
| {"event": "train", "step": 5700, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.34588, "bpb": 1.9417, "chunk_loss": 0.030283, "F": 0.19415, "G": 0.20382, "chunks": 198.81, "bytes_per_chunk": 5.151, "boundary_entropy": 0.11901, "router_noise": 0.0, "grad_norm": 0.7401, "lr": 3.18e-05, "router_trainable": true, "sec_per_step": 0.338, "probe_sec": 0.0895, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08019, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.12469, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.72042, "trunc_abs_frac": 0.75701, "mean_delta_full": 0.28306, "read_hit_rate": {"merge": 0.25532}, "mean_gate": {"merge": 0.17097}, "n_reads": 1608, "n_cand": 6298, "read_suppressed_frac": 0.0958, "dropout_reads_logged": 1608, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 148.40310668945312, "n_util_obs": 2147, "util_count_spearman": 0.08594907327080405}, "probe_overhead_sec_total": 314.402, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.019252, "alpha": 0.03, "t": 2285.3} |
| {"event": "consolidate", "step": 5750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2306.2} |
| {"event": "train", "step": 5750, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.40213, "bpb": 2.02285, "chunk_loss": 0.030512, "F": 0.20367, "G": 0.21363, "chunks": 208.56, "bytes_per_chunk": 4.91, "boundary_entropy": 0.1179, "router_noise": 0.0, "grad_norm": 0.7226, "lr": 3.13e-05, "router_trainable": true, "sec_per_step": 0.425, "probe_sec": 0.0888, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0757, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.13177, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 3.53045, "trunc_abs_frac": 0.3779, "mean_delta_full": 0.03023, "read_hit_rate": {"merge": 0.25507}, "mean_gate": {"merge": 0.16016}, "n_reads": 1686, "n_cand": 6610, "read_suppressed_frac": 0.102, "dropout_reads_logged": 1686, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 149.70262145996094, "n_util_obs": 2167, "util_count_spearman": 0.08452567725297229}, "probe_overhead_sec_total": 317.029, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.022554, "alpha": 0.03, "t": 2306.6} |
| {"event": "consolidate", "step": 5800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2324.2} |
| {"event": "probe", "step": 5800, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3290.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.189048, "cps": 25.703125, "mean_p": 0.198877, "boundary_entropy": 0.114195, "bpb": 1.879368}, "tr": {"chunks": 3678.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.203171, "cps": 28.734375, "mean_p": 0.202184, "boundary_entropy": 0.133163, "bpb": 2.004862}, "fi": {"chunks": 4161.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.2212, "cps": 32.507812, "mean_p": 0.229762, "boundary_entropy": 0.127797, "bpb": 2.114837}}, "bpb": {"en": 1.87937, "tr": 2.00486, "fi": 2.11484}, "gini_cpc": 0.034943, "gini_cps": 0.052176, "bpic": 5.097673, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.074705312575608, "fi": 1.1700759547176933}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 150.96340942382812, "n_util_obs": 2193, "util_count_spearman": 0.08947142882333287}, "t": 2324.7} |
| {"event": "eval", "step": 5800, "phase": 3, "bpb": {"en": 1.89364, "tr": 2.00965, "fi": 2.122}, "bytes_per_chunk": {"en": 5.078, "tr": 5.252, "fi": 4.833}, "bpb_mean": 2.00843, "frozen_inventory": true, "t": 2325.9} |
| {"event": "train", "step": 5800, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32285, "bpb": 1.90847, "chunk_loss": 0.030505, "F": 0.20212, "G": 0.21095, "chunks": 206.97, "bytes_per_chunk": 4.948, "boundary_entropy": 0.12048, "router_noise": 0.0, "grad_norm": 0.7283, "lr": 3.08e-05, "router_trainable": true, "sec_per_step": 0.418, "probe_sec": 0.071, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.064, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.11875, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 0.79561, "trunc_abs_frac": 0.56397, "mean_delta_full": -0.13328, "read_hit_rate": {"merge": 0.27306}, "mean_gate": {"merge": 0.16618}, "n_reads": 1791, "n_cand": 6559, "read_suppressed_frac": 0.1055, "dropout_reads_logged": 1791, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 150.99420166015625, "n_util_obs": 2193, "util_count_spearman": 0.08919556999095034}, "probe_overhead_sec_total": 319.86, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.024759, "alpha": 0.03, "t": 2327.5} |
| {"event": "consolidate", "step": 5850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2344.4} |
| {"event": "train", "step": 5850, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.34414, "bpb": 1.93919, "chunk_loss": 0.030405, "F": 0.20071, "G": 0.21042, "chunks": 205.53, "bytes_per_chunk": 4.982, "boundary_entropy": 0.11558, "router_noise": 0.0, "grad_norm": 0.7409, "lr": 3.05e-05, "router_trainable": true, "sec_per_step": 0.351, "probe_sec": 0.0298, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02438, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.0449, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 1.0, "trunc_abs_frac": 1.0, "mean_delta_full": 0.16113, "read_hit_rate": {"merge": 0.24919}, "mean_gate": {"merge": 0.16086}, "n_reads": 1623, "n_cand": 6513, "read_suppressed_frac": 0.0961, "dropout_reads_logged": 1623, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 152.27931213378906, "n_util_obs": 2209, "util_count_spearman": 0.08642262694858852}, "probe_overhead_sec_total": 322.768, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.021843, "alpha": 0.03, "t": 2345.0} |
| {"event": "consolidate", "step": 5900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2364.0} |
| {"event": "train", "step": 5900, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.27843, "bpb": 1.84438, "chunk_loss": 0.030271, "F": 0.19403, "G": 0.20385, "chunks": 198.69, "bytes_per_chunk": 5.154, "boundary_entropy": 0.11505, "router_noise": 0.0, "grad_norm": 0.7158, "lr": 3.02e-05, "router_trainable": true, "sec_per_step": 0.387, "probe_sec": 0.0749, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06644, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.09971, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.56486, "trunc_abs_frac": 0.56486, "mean_delta_full": 0.52417, "read_hit_rate": {"merge": 0.25278}, "mean_gate": {"merge": 0.16503}, "n_reads": 1591, "n_cand": 6294, "read_suppressed_frac": 0.1094, "dropout_reads_logged": 1591, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 153.56475830078125, "n_util_obs": 2235, "util_count_spearman": 0.0892492344498247}, "probe_overhead_sec_total": 325.398, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.024726, "alpha": 0.03, "t": 2364.4} |
| {"event": "consolidate", "step": 5950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2391.4} |
| {"event": "train", "step": 5950, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38119, "bpb": 1.99264, "chunk_loss": 0.030408, "F": 0.20135, "G": 0.2102, "chunks": 206.19, "bytes_per_chunk": 4.966, "boundary_entropy": 0.1164, "router_noise": 0.0, "grad_norm": 0.7331, "lr": 3.01e-05, "router_trainable": true, "sec_per_step": 0.566, "probe_sec": 0.0969, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06072, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.06571, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 1.29926, "trunc_abs_frac": 1.64994, "mean_delta_full": 0.08224, "read_hit_rate": {"merge": 0.25069}, "mean_gate": {"merge": 0.16513}, "n_reads": 1638, "n_cand": 6534, "read_suppressed_frac": 0.102, "dropout_reads_logged": 1638, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 154.87002563476562, "n_util_obs": 2248, "util_count_spearman": 0.08837246384308978}, "probe_overhead_sec_total": 328.475, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.023106, "alpha": 0.03, "t": 2392.7} |
| {"event": "consolidate", "step": 6000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2421.2} |
| {"event": "train", "step": 5999, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38513, "bpb": 1.99832, "chunk_loss": 0.030475, "F": 0.2012, "G": 0.20907, "chunks": 206.03, "bytes_per_chunk": 4.97, "boundary_entropy": 0.1132, "router_noise": 0.0, "grad_norm": 0.7083, "lr": 3e-05, "router_trainable": true, "sec_per_step": 0.57, "probe_sec": 0.0872, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07333, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.18265, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 1.16892, "trunc_abs_frac": 0.54632, "mean_delta_full": 0.1084, "read_hit_rate": {"merge": 0.26467}, "mean_gate": {"merge": 0.16004}, "n_reads": 1728, "n_cand": 6529, "read_suppressed_frac": 0.103, "dropout_reads_logged": 1728, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 156.15919494628906, "n_util_obs": 2272, "util_count_spearman": 0.09220505925747007}, "probe_overhead_sec_total": 331.318, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.024544, "alpha": 0.03, "t": 2421.2} |
| {"event": "probe", "step": 6000, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3306.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.189967, "cps": 25.828125, "mean_p": 0.198652, "boundary_entropy": 0.112455, "bpb": 1.876682}, "tr": {"chunks": 3715.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.205215, "cps": 29.023438, "mean_p": 0.202188, "boundary_entropy": 0.127927, "bpb": 1.99823}, "fi": {"chunks": 4171.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.221732, "cps": 32.585938, "mean_p": 0.228969, "boundary_entropy": 0.124598, "bpb": 2.1119}}, "bpb": {"en": 1.87668, "tr": 1.99823, "fi": 2.1119}, "gini_cpc": 0.034326, "gini_cps": 0.051525, "bpic": 5.068978, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0802630902238881, "fi": 1.1672115525717133}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 156.15919494628906, "n_util_obs": 2272, "util_count_spearman": 0.09220505925747007}, "t": 2421.9} |
| {"event": "eval", "step": 6000, "phase": 3, "bpb": {"en": 1.89089, "tr": 2.00393, "fi": 2.11893}, "bytes_per_chunk": {"en": 5.058, "tr": 5.216, "fi": 4.825}, "bpb_mean": 2.00458, "frozen_inventory": true, "t": 2422.9} |
| {"event": "reps_dumped", "dirs": ["/local/scratch/sas245/beetle-hnet/runs/ali_mem-merge_s2/probe/reps/th0.30", "/local/scratch/sas245/beetle-hnet/runs/ali_mem-merge_s2/probe/reps/th0.40", "/local/scratch/sas245/beetle-hnet/runs/ali_mem-merge_s2/probe/reps/th0.50", "/local/scratch/sas245/beetle-hnet/runs/ali_mem-merge_s2/probe/reps/th0.60", "/local/scratch/sas245/beetle-hnet/runs/ali_mem-merge_s2/probe/reps/th0.70"], "t": 2425.2} |
| {"event": "final", "run": "ali_mem-merge_s2", "arch": "ali", "objective": "ratio", "condition": "A", "step": 6000, "params": 22785088, "seed": 2, "per_lang": {"en": {"cpc": 0.18996724702637477, "cps": 25.828125, "bpb": 1.896335646121363, "bytes_per_chunk": 4.990319909385976, "reads_per_chunk": 0.27372669211220146}, "tr": {"cpc": 0.20521460531403635, "cps": 29.0234375, "bpb": 1.9957856996232657, "bytes_per_chunk": 5.2245939212755355, "reads_per_chunk": 0.25641255605381164}, "fi": {"cpc": 0.22173196533942907, "cps": 32.5859375, "bpb": 2.1353870915184916, "bytes_per_chunk": 4.7983599355688975, "reads_per_chunk": 0.1633749450871284}}, "bpb_mean": 2.009169479087707, "gini_cpc": 0.03432647846179293, "gini_cps": 0.05152489873719324, "bpic": 5.068977841315225, "mean_boundary_entropy": 0.1131967306137085, "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": 331.3176290988922, "probe_overhead_seq_frac": 0.49972916666666667, "wall_sec": 2427.437506198883, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 65536, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 156.15919494628906, "n_util_obs": 2272, "util_count_spearman": 0.09220505925747007}, "ali_params": 268096, "ali_config": {"enabled": true, "budget_B": 65536, "staging_capacity": 262144, "policy": "frequency", "value_mode": "learned", "candidate_types": ["merge"], "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": 0.09220505925747007, "paired_vs_dropout": {"n": 5401, "pearson": 0.03920025825735148, "spearman": 0.03775594935826767, "mean_abs_discrepancy": 1.7015271186828613, "mean_paired": 0.07435416430234909, "mean_dropout": 0.06816203892230988}, "predictor_calibration": {"n": 11405, "all": {"n": 11405, "pearson": 0.1021127872669672, "spearman": 0.1405262786708346, "mae": 0.1321927011013031, "reliability": [{"pred": 0.010875605046749115, "obs": 0.02351362816989422, "n": 1141}, {"pred": 0.017546283081173897, "obs": 0.03070252761244774, "n": 1140}, {"pred": 0.021329494193196297, "obs": 0.04522925615310669, "n": 1141}, {"pred": 0.02455827221274376, "obs": 0.041162483394145966, "n": 1140}, {"pred": 0.027789145708084106, "obs": 0.04716715216636658, "n": 1140}, {"pred": 0.031354840844869614, "obs": 0.07182937860488892, "n": 1141}, {"pred": 0.03524461016058922, "obs": 0.054299842566251755, "n": 1140}, {"pred": 0.04003944993019104, "obs": 0.08206656575202942, "n": 1141}, {"pred": 0.04686075076460838, "obs": 0.08314508944749832, "n": 1140}, {"pred": 0.0648217722773552, "obs": 0.10733778774738312, "n": 1141}]}, "admitted": {"n": 11405, "pearson": 0.1021127872669672, "spearman": 0.1405262786708346, "mae": 0.1321927011013031, "reliability": [{"pred": 0.010875605046749115, "obs": 0.02351362816989422, "n": 1141}, {"pred": 0.017546283081173897, "obs": 0.03070252761244774, "n": 1140}, {"pred": 0.021329494193196297, "obs": 0.04522925615310669, "n": 1141}, {"pred": 0.02455827221274376, "obs": 0.041162483394145966, "n": 1140}, {"pred": 0.027789145708084106, "obs": 0.04716715216636658, "n": 1140}, {"pred": 0.031354840844869614, "obs": 0.07182937860488892, "n": 1141}, {"pred": 0.03524461016058922, "obs": 0.054299842566251755, "n": 1140}, {"pred": 0.04003944993019104, "obs": 0.08206656575202942, "n": 1141}, {"pred": 0.04686075076460838, "obs": 0.08314508944749832, "n": 1140}, {"pred": 0.0648217722773552, "obs": 0.10733778774738312, "n": 1141}]}, "staged": {"n": 0, "pearson": NaN, "spearman": NaN, "mae": NaN}}, "t": 2427.7} |
| {"event": "hub_done", "repo": "AdaptiveChunking/ali_mem-merge_s2", "pushed": [0, 1, 2, 4, 6, 8, 16, 32, 64, 128, 256, 512, 1000, 1024, 2048, 3000, 4096, 6000], "failed": [], "ok": true, "t": 2443.0} |
|
|
