| {"event": "hub", "repo": "AdaptiveChunking/ali_mem-mixed_s0", "keep_steps": [0, 1, 2, 4, 6, 8, 16, 32, 64, 128, 256, 512, 1000, 1024, 2048, 3000, 4096, 6000], "t": 3.4} |
| {"event": "start", "run": "ali_mem-mixed_s0", "arch": "ali", "objective": "ratio", "condition": "A", "params": 22785088, "langs": ["en", "tr", "fi"], "premium": {"en": 1.0, "tr": 1.12, "fi": 1.104}, "schedule": {"tau1": 0.4, "tau2": 0.7, "p1_objective": "global", "p2_objective": "global", "p3_objective": "global", "freeze_in_p3": false, "freeze_scope": "router"}, "corpus": "/local/scratch/sas245/beetle-hnet/scratch/data/tier1_corpus", "premium_kind": "flores_per_sent", "doc_filtered": {}, "weights": {"en": 0.74, "tr": 0.1536, "fi": 0.1064}, "device": "cuda", "ali_params": 268096, "ali_config": {"enabled": true, "budget_B": 65536, "staging_capacity": 262144, "policy": "frequency", "value_mode": "learned", "candidate_types": ["surface", "split", "merge", "skel"], "gate": "vector", "retrieval": true, "memory_scope": "persistent", "read_dropout": 0.1, "write_rule": "mean", "eta_min": 0.0, "consolidate_every": 50, "probe_seqs": 2, "probe_reps": 8, "horizon": 32, "predictor": "mlp", "lambda_util": 0.1, "max_span": 64, "byte_classes_path": "/local/scratch/sas245/beetle-hnet/scratch/data/byte_classes.json", "gate_hidden": 0, "type_dim": 16, "gate_bias_init": -2.0}, "t": 3.4} |
| {"event": "probe_schedule", "n": 45, "steps": [0, 1, 2, 4, 6, 8, 16, 32, 64, 128, 200, 256, 400, 512, 600, 800, 1000, 1024, 1200, 1400, 1600, 1800, 2000, 2048, 2200, 2400, 2600, 2800, 3000, 3200, 3400, 3600, 3800, 4000, 4096, 4200, 4400, 4600, 4800, 5000, 5200, 5400, 5600, 5800, 6000], "t": 3.4} |
| {"event": "probe", "step": 0, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 6384.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.366833, "cps": 49.875, "mean_p": 0.496725, "boundary_entropy": 0.686995, "bpb": 8.160238}, "tr": {"chunks": 5290.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.292217, "cps": 41.328125, "mean_p": 0.489567, "boundary_entropy": 0.687288, "bpb": 8.171434}, "fi": {"chunks": 8895.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.472862, "cps": 69.492188, "mean_p": 0.500688, "boundary_entropy": 0.687512, "bpb": 8.214758}}, "bpb": {"en": 8.16024, "tr": 8.17143, "fi": 8.21476}, "gini_cpc": 0.106395, "gini_cps": 0.116842, "bpic": 2.758131, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.7965927738475616, "fi": 1.2890367847879758}, "inventory": {"size": 0, "admitted": 0, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 0.0, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 5.9} |
| {"event": "train", "step": 0, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 5.66332, "bpb": 8.17044, "chunk_loss": 0.038537, "F": 0.28732, "G": 0.4892, "chunks": 294.22, "bytes_per_chunk": 3.48, "boundary_entropy": 0.69137, "router_noise": 0.0, "grad_norm": 11.0314, "lr": 1e-06, "router_trainable": true, "sec_per_step": 0.022, "probe_sec": 0.0127, "probe_extra_seqs": 0, "n_rows": 0, "extra_fwd_seqs": 0, "wall_s": 0.0, "overhead_frac_est": 0.0, "chunk_mismatch": 0, "n_loo": 0, "mean_delta_loo": NaN, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "read_hit_rate": {"surface": 0.0, "split": 0.0, "merge": 0.0, "skel": 0.0}, "mean_gate": {"surface": null, "split": null, "merge": null, "skel": null}, "n_reads": 0, "n_cand": 34933, "read_suppressed_frac": 0.0, "dropout_reads_logged": 0, "inventory": {"size": 16427, "admitted": 16427, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 3696, "staged_surface": 0, "admitted_split": 3051, "staged_split": 0, "admitted_merge": 7179, "staged_merge": 0, "admitted_skel": 2501, "staged_skel": 0, "mean_count": 2.1265599727630615, "n_util_obs": 0, "util_count_spearman": NaN}, "probe_overhead_sec_total": 0.013, "probe_overhead_seq_frac": 0.0, "util_loss": 0.0, "alpha": 0.03, "t": 7.4} |
| {"event": "probe", "step": 1, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 6302.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.362121, "cps": 49.234375, "mean_p": 0.496454, "boundary_entropy": 0.686986, "bpb": 8.14393}, "tr": {"chunks": 5229.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.288847, "cps": 40.851562, "mean_p": 0.489372, "boundary_entropy": 0.687277, "bpb": 8.161818}, "fi": {"chunks": 8819.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.468821, "cps": 68.898438, "mean_p": 0.500418, "boundary_entropy": 0.687508, "bpb": 8.219275}}, "bpb": {"en": 8.14393, "tr": 8.16182, "fi": 8.21928}, "gini_cpc": 0.107148, "gini_cps": 0.117609, "bpic": 2.787813, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.7976526488917842, "fi": 1.2946523993258972}, "inventory": {"size": 16427, "admitted": 16427, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 3696, "staged_surface": 0, "admitted_split": 3051, "staged_split": 0, "admitted_merge": 7179, "staged_merge": 0, "admitted_skel": 2501, "staged_skel": 0, "mean_count": 2.1265599727630615, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 9.0} |
| {"event": "probe", "step": 2, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 6202.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.356375, "cps": 48.453125, "mean_p": 0.496024, "boundary_entropy": 0.686969, "bpb": 8.081262}, "tr": {"chunks": 5154.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.284704, "cps": 40.265625, "mean_p": 0.489068, "boundary_entropy": 0.68726, "bpb": 8.119786}, "fi": {"chunks": 8704.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.462708, "cps": 68.0, "mean_p": 0.499983, "boundary_entropy": 0.6875, "bpb": 8.166307}}, "bpb": {"en": 8.08126, "tr": 8.11979, "fi": 8.16631}, "gini_cpc": 0.107511, "gini_cps": 0.117979, "bpic": 2.828116, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.7988885948286564, "fi": 1.2983726459322422}, "inventory": {"size": 30107, "admitted": 30107, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3, "admitted_surface": 6589, "staged_surface": 0, "admitted_split": 5121, "staged_split": 0, "admitted_merge": 14027, "staged_merge": 0, "admitted_skel": 4370, "staged_skel": 0, "mean_count": 2.4425880908966064, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 11.8} |
| {"event": "probe", "step": 4, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 5818.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.33431, "cps": 45.453125, "mean_p": 0.494569, "boundary_entropy": 0.686905, "bpb": 7.865679}, "tr": {"chunks": 4872.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.269127, "cps": 38.0625, "mean_p": 0.48803, "boundary_entropy": 0.687198, "bpb": 7.966815}, "fi": {"chunks": 8208.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.43634, "cps": 64.125, "mean_p": 0.498414, "boundary_entropy": 0.687465, "bpb": 7.97658}}, "bpb": {"en": 7.86568, "tr": 7.96682, "fi": 7.97658}, "gini_cpc": 0.107211, "gini_cps": 0.117684, "bpic": 3.002011, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.805020855128536, "fi": 1.3051964011241786}, "inventory": {"size": 55336, "admitted": 55336, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 8, "admitted_surface": 11946, "staged_surface": 0, "admitted_split": 9178, "staged_split": 0, "admitted_merge": 26148, "staged_merge": 0, "admitted_skel": 8064, "staged_skel": 0, "mean_count": 2.5957062244415283, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 15.1} |
| {"event": "probe", "step": 6, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 5276.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.303166, "cps": 41.21875, "mean_p": 0.492423, "boundary_entropy": 0.686795, "bpb": 7.570366}, "tr": {"chunks": 4435.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.244987, "cps": 34.648438, "mean_p": 0.48652, "boundary_entropy": 0.687099, "bpb": 7.746204}, "fi": {"chunks": 7524.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.399979, "cps": 58.78125, "mean_p": 0.49604, "boundary_entropy": 0.687393, "bpb": 7.705515}}, "bpb": {"en": 7.57037, "tr": 7.7462, "fi": 7.70551}, "gini_cpc": 0.10898, "gini_cps": 0.119486, "bpic": 3.291674, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.8080949747709771, "fi": 1.3193385026400237}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 12, "admitted_surface": 14235, "staged_surface": 0, "admitted_split": 11435, "staged_split": 0, "admitted_merge": 29989, "staged_merge": 0, "admitted_skel": 9877, "staged_skel": 0, "mean_count": 2.8911285400390625, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 19.8} |
| {"event": "probe", "step": 8, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 4749.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.272884, "cps": 37.101562, "mean_p": 0.490163, "boundary_entropy": 0.686659, "bpb": 7.289205}, "tr": {"chunks": 4068.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.224714, "cps": 31.78125, "mean_p": 0.484947, "boundary_entropy": 0.686986, "bpb": 7.524973}, "fi": {"chunks": 6691.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.355696, "cps": 52.273438, "mean_p": 0.49328, "boundary_entropy": 0.687284, "bpb": 7.441564}}, "bpb": {"en": 7.28921, "tr": 7.52497, "fi": 7.44156}, "gini_cpc": 0.102334, "gini_cps": 0.112759, "bpic": 3.658241, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.823478649179823, "fi": 1.3034701708953575}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 17, "admitted_surface": 14624, "staged_surface": 0, "admitted_split": 12667, "staged_split": 0, "admitted_merge": 27889, "staged_merge": 0, "admitted_skel": 10356, "staged_skel": 0, "mean_count": 3.3327789306640625, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 23.5} |
| {"event": "probe", "step": 16, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 5034.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.28926, "cps": 39.328125, "mean_p": 0.492816, "boundary_entropy": 0.686969, "bpb": 6.66331}, "tr": {"chunks": 5136.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.28371, "cps": 40.125, "mean_p": 0.490318, "boundary_entropy": 0.687456, "bpb": 6.978497}, "fi": {"chunks": 6003.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.319122, "cps": 46.898438, "mean_p": 0.492125, "boundary_entropy": 0.68742, "bpb": 6.844286}}, "bpb": {"en": 6.66331, "tr": 6.9785, "fi": 6.84429}, "gini_cpc": 0.026464, "gini_cps": 0.039943, "bpic": 3.507822, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9808111009857051, "fi": 1.1032333172542708}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 38, "admitted_surface": 15016, "staged_surface": 0, "admitted_split": 15462, "staged_split": 0, "admitted_merge": 23652, "staged_merge": 0, "admitted_skel": 11406, "staged_skel": 0, "mean_count": 4.4781494140625, "n_util_obs": 1, "util_count_spearman": NaN}, "t": 30.5} |
| {"event": "probe", "step": 32, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3904.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.224329, "cps": 30.5, "mean_p": 0.494068, "boundary_entropy": 0.68751, "bpb": 6.306938}, "tr": {"chunks": 3872.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.213887, "cps": 30.25, "mean_p": 0.490997, "boundary_entropy": 0.687895, "bpb": 6.636636}, "fi": {"chunks": 3956.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.210302, "cps": 30.90625, "mean_p": 0.491661, "boundary_entropy": 0.687922, "bpb": 6.430914}}, "bpb": {"en": 6.30694, "tr": 6.63664, "fi": 6.43091}, "gini_cpc": 0.014419, "gini_cps": 0.004773, "bpic": 4.835663, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9534526022767714, "fi": 0.9374728751314847}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 69, "admitted_surface": 15212, "staged_surface": 0, "admitted_split": 15998, "staged_split": 0, "admitted_merge": 23223, "staged_merge": 0, "admitted_skel": 11103, "staged_skel": 0, "mean_count": 7.1539306640625, "n_util_obs": 5, "util_count_spearman": 0.3}, "t": 43.9} |
| {"event": "consolidate", "step": 50, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 59.9} |
| {"event": "train", "step": 50, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 4.22565, "bpb": 6.09632, "chunk_loss": 0.034951, "F": 0.23645, "G": 0.49291, "chunks": 242.12, "bytes_per_chunk": 4.229, "boundary_entropy": 0.69213, "router_noise": 0.0, "grad_norm": 2.4404, "lr": 5.1e-05, "router_trainable": true, "sec_per_step": 1.064, "probe_sec": 0.0787, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04685, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 4e-05, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.00488, "n_merge": 2, "mean_delta_merge": 0.00522, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -1.20833, "trunc_abs_frac": 0.33708, "mean_delta_full": -0.01465, "read_hit_rate": {"surface": 0.67976, "split": 0.81555, "merge": 0.31854, "skel": 0.58149}, "mean_gate": {"surface": 0.1206, "split": 0.12065, "merge": 0.12046, "skel": 0.12056}, "n_reads": 18744, "n_cand": 30103, "read_suppressed_frac": 0.1003, "dropout_reads_logged": 18744, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 110, "admitted_surface": 15441, "staged_surface": 0, "admitted_split": 15422, "staged_split": 0, "admitted_merge": 23856, "staged_merge": 0, "admitted_skel": 10817, "staged_skel": 0, "mean_count": 10.98236083984375, "n_util_obs": 9, "util_count_spearman": -0.16666666666666666}, "probe_overhead_sec_total": 5.028, "probe_overhead_seq_frac": 0.4902, "util_loss": 0.085076, "alpha": 0.03, "t": 60.6} |
| {"event": "probe", "step": 64, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 4110.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.236166, "cps": 32.109375, "mean_p": 0.487671, "boundary_entropy": 0.686496, "bpb": 5.619576}, "tr": {"chunks": 3914.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.216207, "cps": 30.578125, "mean_p": 0.48756, "boundary_entropy": 0.687437, "bpb": 6.008027}, "fi": {"chunks": 5466.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.290575, "cps": 42.703125, "mean_p": 0.492837, "boundary_entropy": 0.687649, "bpb": 5.81772}}, "bpb": {"en": 5.61958, "tr": 6.00803, "fi": 5.81772}, "gini_cpc": 0.066732, "gini_cps": 0.076699, "bpic": 4.205486, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9154878148598995, "fi": 1.2303822076460156}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 135, "admitted_surface": 15807, "staged_surface": 0, "admitted_split": 16630, "staged_split": 0, "admitted_merge": 21751, "staged_merge": 0, "admitted_skel": 11348, "staged_skel": 0, "mean_count": 13.567550659179688, "n_util_obs": 10, "util_count_spearman": -0.7575757575757576}, "t": 74.2} |
| {"event": "consolidate", "step": 100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 113.8} |
| {"event": "train", "step": 100, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 3.21933, "bpb": 4.64451, "chunk_loss": 0.032288, "F": 0.20813, "G": 0.42258, "chunks": 213.12, "bytes_per_chunk": 4.805, "boundary_entropy": 0.66621, "router_noise": 0.0, "grad_norm": 4.1802, "lr": 0.000101, "router_trainable": true, "sec_per_step": 1.093, "probe_sec": 0.1044, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0505, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": -0.0027, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.00732, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.0669, "trunc_abs_frac": 0.14931, "mean_delta_full": 0.09823, "read_hit_rate": {"surface": 0.68002, "split": 0.86761, "merge": 0.27383, "skel": 0.66633}, "mean_gate": {"surface": 0.12038, "split": 0.12041, "merge": 0.12032, "skel": 0.12037}, "n_reads": 20148, "n_cand": 30452, "read_suppressed_frac": 0.0982, "dropout_reads_logged": 20148, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 217, "admitted_surface": 16362, "staged_surface": 0, "admitted_split": 16358, "staged_split": 0, "admitted_merge": 21125, "staged_merge": 0, "admitted_skel": 11691, "staged_skel": 0, "mean_count": 24.75445556640625, "n_util_obs": 14, "util_count_spearman": 0.14725274725274726}, "probe_overhead_sec_total": 12.21, "probe_overhead_seq_frac": 0.495, "util_loss": 0.001905, "alpha": 0.03, "t": 115.2} |
| {"event": "probe", "step": 128, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3578.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.205597, "cps": 27.953125, "mean_p": 0.37894, "boundary_entropy": 0.611458, "bpb": 4.17272}, "tr": {"chunks": 3090.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.17069, "cps": 24.140625, "mean_p": 0.411217, "boundary_entropy": 0.650553, "bpb": 4.493324}, "fi": {"chunks": 3789.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.201425, "cps": 29.601562, "mean_p": 0.395825, "boundary_entropy": 0.63626, "bpb": 4.420068}}, "bpb": {"en": 4.17272, "tr": 4.49332, "fi": 4.42007}, "gini_cpc": 0.040282, "gini_cps": 0.044563, "bpic": 5.425265, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.8302171719883616, "fi": 0.9797076648330741}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 268, "admitted_surface": 16620, "staged_surface": 0, "admitted_split": 16844, "staged_split": 0, "admitted_merge": 20200, "staged_merge": 0, "admitted_skel": 11872, "staged_skel": 0, "mean_count": 32.480499267578125, "n_util_obs": 17, "util_count_spearman": -0.014705882352941176}, "t": 149.4} |
| {"event": "consolidate", "step": 150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 177.2} |
| {"event": "train", "step": 150, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.80573, "bpb": 4.04781, "chunk_loss": 0.031734, "F": 0.20575, "G": 0.37016, "chunks": 210.69, "bytes_per_chunk": 4.86, "boundary_entropy": 0.59609, "router_noise": 0.0, "grad_norm": 2.4181, "lr": 0.000151, "router_trainable": true, "sec_per_step": 1.264, "probe_sec": 0.0829, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05538, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.00741, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.03735, "n_merge": 1, "mean_delta_merge": 0.06393, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": -0.11531, "trunc_abs_frac": 13.706, "mean_delta_full": -0.07613, "read_hit_rate": {"surface": 0.65142, "split": 0.83218, "merge": 0.25678, "skel": 0.59688}, "mean_gate": {"surface": 0.12099, "split": 0.12098, "merge": 0.12101, "skel": 0.12103}, "n_reads": 17708, "n_cand": 28639, "read_suppressed_frac": 0.1001, "dropout_reads_logged": 17708, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 314, "admitted_surface": 16708, "staged_surface": 0, "admitted_split": 16848, "staged_split": 0, "admitted_merge": 20123, "staged_merge": 0, "admitted_skel": 11857, "staged_skel": 0, "mean_count": 38.795562744140625, "n_util_obs": 19, "util_count_spearman": -0.043859649122807015}, "probe_overhead_sec_total": 18.341, "probe_overhead_seq_frac": 0.4967, "util_loss": 0.000617, "alpha": 0.03, "t": 178.4} |
| {"event": "consolidate", "step": 200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 228.2} |
| {"event": "probe", "step": 200, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3004.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.172614, "cps": 23.46875, "mean_p": 0.352407, "boundary_entropy": 0.551427, "bpb": 3.630013}, "tr": {"chunks": 3197.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.176601, "cps": 24.976562, "mean_p": 0.36254, "boundary_entropy": 0.579528, "bpb": 3.781043}, "fi": {"chunks": 3420.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.181809, "cps": 26.71875, "mean_p": 0.359751, "boundary_entropy": 0.5894, "bpb": 3.862472}}, "bpb": {"en": 3.63001, "tr": 3.78104, "fi": 3.86247}, "gini_cpc": 0.011543, "gini_cps": 0.028826, "bpic": 5.896684, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0230957408755772, "fi": 1.0532668472232123}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 401, "admitted_surface": 16962, "staged_surface": 0, "admitted_split": 17190, "staged_split": 0, "admitted_merge": 19334, "staged_merge": 0, "admitted_skel": 12050, "staged_skel": 0, "mean_count": 49.97688293457031, "n_util_obs": 32, "util_count_spearman": 0.12866568914956011}, "t": 229.8} |
| {"event": "eval", "step": 200, "phase": 1, "bpb": {"en": 3.63683, "tr": 3.79122, "fi": 3.86115}, "bytes_per_chunk": {"en": 5.65, "tr": 6.238, "fi": 5.857}, "bpb_mean": 3.76307, "frozen_inventory": true, "t": 232.7} |
| {"event": "train", "step": 200, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.54728, "bpb": 3.67495, "chunk_loss": 0.030431, "F": 0.17563, "G": 0.35783, "chunks": 179.84, "bytes_per_chunk": 5.694, "boundary_entropy": 0.56023, "router_noise": 0.0, "grad_norm": 1.6633, "lr": 0.000201, "router_trainable": true, "sec_per_step": 1.119, "probe_sec": 0.0752, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05809, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.00893, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.05437, "n_merge": 1, "mean_delta_merge": -0.00903, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": -0.7037, "trunc_abs_frac": 0.7037, "mean_delta_full": -0.00659, "read_hit_rate": {"surface": 0.56229, "split": 0.78969, "merge": 0.16412, "skel": 0.53136}, "mean_gate": {"surface": 0.12461, "split": 0.12489, "merge": 0.12481, "skel": 0.12526}, "n_reads": 13859, "n_cand": 25052, "read_suppressed_frac": 0.0985, "dropout_reads_logged": 13859, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 402, "admitted_surface": 17057, "staged_surface": 0, "admitted_split": 17171, "staged_split": 0, "admitted_merge": 19229, "staged_merge": 0, "admitted_skel": 12079, "staged_skel": 0, "mean_count": 50.19068908691406, "n_util_obs": 32, "util_count_spearman": 0.1315982404692082}, "probe_overhead_sec_total": 23.456, "probe_overhead_seq_frac": 0.4975, "util_loss": 0.000502, "alpha": 0.03, "t": 234.4} |
| {"event": "consolidate", "step": 250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 272.4} |
| {"event": "train", "step": 250, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.38223, "bpb": 3.43684, "chunk_loss": 0.031609, "F": 0.20926, "G": 0.34128, "chunks": 214.28, "bytes_per_chunk": 4.779, "boundary_entropy": 0.51424, "router_noise": 0.0, "grad_norm": 1.4261, "lr": 0.000251, "router_trainable": true, "sec_per_step": 0.783, "probe_sec": 0.1132, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07679, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.01977, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.06873, "n_merge": 2, "mean_delta_merge": 0.007, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.68246, "trunc_abs_frac": 0.74537, "mean_delta_full": 0.0863, "read_hit_rate": {"surface": 0.58608, "split": 0.81975, "merge": 0.16517, "skel": 0.58401}, "mean_gate": {"surface": 0.12982, "split": 0.12973, "merge": 0.12946, "skel": 0.12931}, "n_reads": 17628, "n_cand": 30327, "read_suppressed_frac": 0.0961, "dropout_reads_logged": 17628, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 488, "admitted_surface": 17032, "staged_surface": 0, "admitted_split": 17008, "staged_split": 0, "admitted_merge": 19473, "staged_merge": 0, "admitted_skel": 12023, "staged_skel": 0, "mean_count": 61.582733154296875, "n_util_obs": 38, "util_count_spearman": 0.05700842542947806}, "probe_overhead_sec_total": 28.233, "probe_overhead_seq_frac": 0.498, "util_loss": 0.000581, "alpha": 0.03, "t": 273.5} |
| {"event": "probe", "step": 256, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3179.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.18267, "cps": 24.835938, "mean_p": 0.31944, "boundary_entropy": 0.493633, "bpb": 3.335236}, "tr": {"chunks": 3358.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.185494, "cps": 26.234375, "mean_p": 0.339364, "boundary_entropy": 0.520041, "bpb": 3.523966}, "fi": {"chunks": 3851.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.204721, "cps": 30.085938, "mean_p": 0.349652, "boundary_entropy": 0.545321, "bpb": 3.64649}}, "bpb": {"en": 3.33524, "tr": 3.52397, "fi": 3.64649}, "gini_cpc": 0.025661, "gini_cps": 0.043127, "bpic": 5.461302, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0154621321490949, "fi": 1.120715110353618}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 494, "admitted_surface": 17026, "staged_surface": 0, "admitted_split": 16993, "staged_split": 0, "admitted_merge": 19438, "staged_merge": 0, "admitted_skel": 12079, "staged_skel": 0, "mean_count": 62.866119384765625, "n_util_obs": 40, "util_count_spearman": 0.03470919324577861}, "t": 279.2} |
| {"event": "consolidate", "step": 300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 317.3} |
| {"event": "train", "step": 300, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.28221, "bpb": 3.29253, "chunk_loss": 0.031256, "F": 0.21069, "G": 0.29531, "chunks": 215.75, "bytes_per_chunk": 4.746, "boundary_entropy": 0.4105, "router_noise": 0.0, "grad_norm": 1.4603, "lr": 0.0003, "router_trainable": true, "sec_per_step": 0.895, "probe_sec": 0.1093, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08742, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.02757, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": -0.02159, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -1.7655, "trunc_abs_frac": 1.49035, "mean_delta_full": 0.03101, "read_hit_rate": {"surface": 0.59866, "split": 0.84344, "merge": 0.17251, "skel": 0.59197}, "mean_gate": {"surface": 0.13494, "split": 0.13501, "merge": 0.13376, "skel": 0.13502}, "n_reads": 18262, "n_cand": 30588, "read_suppressed_frac": 0.1011, "dropout_reads_logged": 18262, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 566, "admitted_surface": 17061, "staged_surface": 0, "admitted_split": 16894, "staged_split": 0, "admitted_merge": 19545, "staged_merge": 0, "admitted_skel": 12036, "staged_skel": 0, "mean_count": 74.87109375, "n_util_obs": 45, "util_count_spearman": 0.08089591567852437}, "probe_overhead_sec_total": 33.189, "probe_overhead_seq_frac": 0.4983, "util_loss": 0.000751, "alpha": 0.03, "t": 318.3} |
| {"event": "consolidate", "step": 350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 359.9} |
| {"event": "train", "step": 350, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.19788, "bpb": 3.17087, "chunk_loss": 0.031013, "F": 0.20547, "G": 0.28125, "chunks": 210.41, "bytes_per_chunk": 4.867, "boundary_entropy": 0.37441, "router_noise": 0.0, "grad_norm": 1.6466, "lr": 0.0002999, "router_trainable": true, "sec_per_step": 0.843, "probe_sec": 0.0745, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05343, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.08706, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 3, "mean_delta_split": -0.00412, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.25811, "trunc_abs_frac": 0.29471, "mean_delta_full": 0.18823, "read_hit_rate": {"surface": 0.59663, "split": 0.83293, "merge": 0.18009, "skel": 0.60724}, "mean_gate": {"surface": 0.14675, "split": 0.1471, "merge": 0.14476, "skel": 0.15054}, "n_reads": 17994, "n_cand": 30116, "read_suppressed_frac": 0.1022, "dropout_reads_logged": 17994, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 648, "admitted_surface": 17107, "staged_surface": 0, "admitted_split": 17032, "staged_split": 0, "admitted_merge": 19360, "staged_merge": 0, "admitted_skel": 12037, "staged_skel": 0, "mean_count": 89.21432495117188, "n_util_obs": 52, "util_count_spearman": 0.14565013233159738}, "probe_overhead_sec_total": 38.055, "probe_overhead_seq_frac": 0.4986, "util_loss": 0.000987, "alpha": 0.03, "t": 360.4} |
| {"event": "consolidate", "step": 400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 412.6} |
| {"event": "probe", "step": 400, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3657.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.210136, "cps": 28.570312, "mean_p": 0.256534, "boundary_entropy": 0.302452, "bpb": 2.891537}, "tr": {"chunks": 4061.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.224327, "cps": 31.726562, "mean_p": 0.28036, "boundary_entropy": 0.383242, "bpb": 3.177981}, "fi": {"chunks": 4409.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.234384, "cps": 34.445312, "mean_p": 0.291037, "boundary_entropy": 0.385578, "bpb": 3.378356}}, "bpb": {"en": 2.89154, "tr": 3.17798, "fi": 3.37836}, "gini_cpc": 0.024169, "gini_cps": 0.04134, "bpic": 4.678156, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0675337102335343, "fi": 1.1153916147626084}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 723, "admitted_surface": 17094, "staged_surface": 0, "admitted_split": 16818, "staged_split": 0, "admitted_merge": 19629, "staged_merge": 0, "admitted_skel": 11995, "staged_skel": 0, "mean_count": 103.78437805175781, "n_util_obs": 61, "util_count_spearman": 0.08461131676361713}, "t": 414.5} |
| {"event": "eval", "step": 400, "phase": 1, "bpb": {"en": 2.90402, "tr": 3.18404, "fi": 3.3712}, "bytes_per_chunk": {"en": 4.443, "tr": 4.971, "fi": 4.593}, "bpb_mean": 3.15309, "frozen_inventory": true, "t": 418.2} |
| {"event": "train", "step": 400, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.03883, "bpb": 2.94141, "chunk_loss": 0.031374, "F": 0.22519, "G": 0.26993, "chunks": 230.59, "bytes_per_chunk": 4.441, "boundary_entropy": 0.30833, "router_noise": 0.0, "grad_norm": 1.3498, "lr": 0.0002998, "router_trainable": true, "sec_per_step": 1.207, "probe_sec": 0.1045, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07802, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.01544, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": -0.05069, "n_merge": 2, "mean_delta_merge": 0.19315, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 1.02205, "trunc_abs_frac": 1.56086, "mean_delta_full": 0.16333, "read_hit_rate": {"surface": 0.65238, "split": 0.86518, "merge": 0.23308, "skel": 0.6456}, "mean_gate": {"surface": 0.16302, "split": 0.16559, "merge": 0.15815, "skel": 0.17106}, "n_reads": 20621, "n_cand": 32287, "read_suppressed_frac": 0.1007, "dropout_reads_logged": 20621, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 726, "admitted_surface": 17073, "staged_surface": 0, "admitted_split": 16837, "staged_split": 0, "admitted_merge": 19646, "staged_merge": 0, "admitted_skel": 11980, "staged_skel": 0, "mean_count": 104.105224609375, "n_util_obs": 62, "util_count_spearman": 0.06957769887436731}, "probe_overhead_sec_total": 44.233, "probe_overhead_seq_frac": 0.4988, "util_loss": 0.001475, "alpha": 0.03, "t": 420.8} |
| {"event": "consolidate", "step": 450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 481.3} |
| {"event": "train", "step": 450, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.00839, "bpb": 2.8975, "chunk_loss": 0.031217, "F": 0.22, "G": 0.26405, "chunks": 225.28, "bytes_per_chunk": 4.545, "boundary_entropy": 0.29447, "router_noise": 0.0, "grad_norm": 1.2173, "lr": 0.0002995, "router_trainable": true, "sec_per_step": 1.226, "probe_sec": 0.1079, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08067, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.02261, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": -0.06708, "n_merge": 1, "mean_delta_merge": 0.04407, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.19995, "trunc_abs_frac": 0.55004, "mean_delta_full": 0.08557, "read_hit_rate": {"surface": 0.6486, "split": 0.86972, "merge": 0.21669, "skel": 0.64239}, "mean_gate": {"surface": 0.1716, "split": 0.17127, "merge": 0.16912, "skel": 0.17791}, "n_reads": 20449, "n_cand": 32108, "read_suppressed_frac": 0.0957, "dropout_reads_logged": 20449, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 802, "admitted_surface": 17043, "staged_surface": 0, "admitted_split": 16796, "staged_split": 0, "admitted_merge": 19687, "staged_merge": 0, "admitted_skel": 12010, "staged_skel": 0, "mean_count": 119.1416015625, "n_util_obs": 65, "util_count_spearman": 0.18688811188811189}, "probe_overhead_sec_total": 50.737, "probe_overhead_seq_frac": 0.4989, "util_loss": 0.001605, "alpha": 0.03, "t": 482.1} |
| {"event": "consolidate", "step": 500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 526.1} |
| {"event": "train", "step": 500, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.98018, "bpb": 2.85679, "chunk_loss": 0.031426, "F": 0.22974, "G": 0.26517, "chunks": 235.25, "bytes_per_chunk": 4.353, "boundary_entropy": 0.28241, "router_noise": 0.0, "grad_norm": 1.0926, "lr": 0.0002992, "router_trainable": true, "sec_per_step": 0.903, "probe_sec": 0.0834, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05767, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": 0.00225, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.03137, "n_dropped_unpaired": 0, "n_full": 0, "read_hit_rate": {"surface": 0.64514, "split": 0.87468, "merge": 0.21008, "skel": 0.63707}, "mean_gate": {"surface": 0.17647, "split": 0.17445, "merge": 0.17536, "skel": 0.17989}, "n_reads": 20918, "n_cand": 33016, "read_suppressed_frac": 0.1014, "dropout_reads_logged": 20918, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 869, "admitted_surface": 17126, "staged_surface": 0, "admitted_split": 16731, "staged_split": 0, "admitted_merge": 19706, "staged_merge": 0, "admitted_skel": 11973, "staged_skel": 0, "mean_count": 134.74330139160156, "n_util_obs": 76, "util_count_spearman": 0.17200273410799727}, "probe_overhead_sec_total": 56.051, "probe_overhead_seq_frac": 0.499, "util_loss": 0.001786, "alpha": 0.03, "t": 527.3} |
| {"event": "probe", "step": 512, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3813.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.2191, "cps": 29.789062, "mean_p": 0.252937, "boundary_entropy": 0.273361, "bpb": 2.675958}, "tr": {"chunks": 3887.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.214716, "cps": 30.367188, "mean_p": 0.263335, "boundary_entropy": 0.33504, "bpb": 3.011838}, "fi": {"chunks": 4639.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.246611, "cps": 36.242188, "mean_p": 0.284924, "boundary_entropy": 0.337735, "bpb": 3.22627}}, "bpb": {"en": 2.67596, "tr": 3.01184, "fi": 3.22627}, "gini_cpc": 0.03125, "gini_cps": 0.044628, "bpic": 4.597779, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9799892328681499, "fi": 1.1255629887613259}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 888, "admitted_surface": 17111, "staged_surface": 0, "admitted_split": 16765, "staged_split": 0, "admitted_merge": 19777, "staged_merge": 0, "admitted_skel": 11883, "staged_skel": 0, "mean_count": 138.33193969726562, "n_util_obs": 79, "util_count_spearman": 0.17585199610516067}, "t": 537.7} |
| {"event": "consolidate", "step": 550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 576.1} |
| {"event": "train", "step": 550, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.96211, "bpb": 2.83072, "chunk_loss": 0.031757, "F": 0.242, "G": 0.27045, "chunks": 247.81, "bytes_per_chunk": 4.132, "boundary_entropy": 0.26636, "router_noise": 0.0, "grad_norm": 1.3991, "lr": 0.0002987, "router_trainable": true, "sec_per_step": 1.01, "probe_sec": 0.136, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08784, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": 0.09759, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.00623, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.72814, "trunc_abs_frac": 0.73373, "mean_delta_full": 0.25134, "read_hit_rate": {"surface": 0.67422, "split": 0.88542, "merge": 0.22871, "skel": 0.66796}, "mean_gate": {"surface": 0.17701, "split": 0.17226, "merge": 0.1748, "skel": 0.17532}, "n_reads": 22673, "n_cand": 34715, "read_suppressed_frac": 0.0987, "dropout_reads_logged": 22673, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 931, "admitted_surface": 17089, "staged_surface": 0, "admitted_split": 16637, "staged_split": 0, "admitted_merge": 19863, "staged_merge": 0, "admitted_skel": 11947, "staged_skel": 0, "mean_count": 151.12533569335938, "n_util_obs": 87, "util_count_spearman": 0.10539841073120945}, "probe_overhead_sec_total": 63.01, "probe_overhead_seq_frac": 0.4991, "util_loss": 0.00219, "alpha": 0.03, "t": 577.8} |
| {"event": "consolidate", "step": 600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 629.4} |
| {"event": "probe", "step": 600, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3913.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.224846, "cps": 30.570312, "mean_p": 0.251618, "boundary_entropy": 0.240884, "bpb": 2.567005}, "tr": {"chunks": 4160.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.229796, "cps": 32.5, "mean_p": 0.255743, "boundary_entropy": 0.297411, "bpb": 2.902989}, "fi": {"chunks": 4873.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.259051, "cps": 38.070312, "mean_p": 0.276717, "boundary_entropy": 0.292938, "bpb": 3.131561}}, "bpb": {"en": 2.56701, "tr": 2.90299, "fi": 3.13156}, "gini_cpc": 0.03195, "gini_cps": 0.049436, "bpic": 4.382203, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0220144859527514, "fi": 1.152122877030248}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1001, "admitted_surface": 17112, "staged_surface": 0, "admitted_split": 16748, "staged_split": 0, "admitted_merge": 19733, "staged_merge": 0, "admitted_skel": 11943, "staged_skel": 0, "mean_count": 167.6651153564453, "n_util_obs": 92, "util_count_spearman": 0.1504400228102893}, "t": 631.4} |
| {"event": "eval", "step": 600, "phase": 1, "bpb": {"en": 2.57984, "tr": 2.91271, "fi": 3.13372}, "bytes_per_chunk": {"en": 4.172, "tr": 4.987, "fi": 4.032}, "bpb_mean": 2.87542, "frozen_inventory": true, "t": 634.8} |
| {"event": "train", "step": 600, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.85565, "bpb": 2.67714, "chunk_loss": 0.031457, "F": 0.23544, "G": 0.26134, "chunks": 241.09, "bytes_per_chunk": 4.247, "boundary_entropy": 0.24996, "router_noise": 0.0, "grad_norm": 1.368, "lr": 0.0002982, "router_trainable": true, "sec_per_step": 1.216, "probe_sec": 0.1302, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06504, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.00455, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": -0.17801, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.48494, "trunc_abs_frac": 0.65152, "mean_delta_full": -0.06284, "read_hit_rate": {"surface": 0.67552, "split": 0.88846, "merge": 0.23631, "skel": 0.66631}, "mean_gate": {"surface": 0.18275, "split": 0.17603, "merge": 0.18016, "skel": 0.17793}, "n_reads": 22447, "n_cand": 34135, "read_suppressed_frac": 0.0959, "dropout_reads_logged": 22447, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1003, "admitted_surface": 17088, "staged_surface": 0, "admitted_split": 16642, "staged_split": 0, "admitted_merge": 19823, "staged_merge": 0, "admitted_skel": 11983, "staged_skel": 0, "mean_count": 168.00225830078125, "n_util_obs": 92, "util_count_spearman": 0.15544903903950189}, "probe_overhead_sec_total": 70.604, "probe_overhead_seq_frac": 0.4992, "util_loss": 0.00265, "alpha": 0.03, "t": 638.6} |
| {"event": "consolidate", "step": 650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 699.4} |
| {"event": "train", "step": 650, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.88715, "bpb": 2.72259, "chunk_loss": 0.032061, "F": 0.25644, "G": 0.26805, "chunks": 262.59, "bytes_per_chunk": 3.9, "boundary_entropy": 0.23667, "router_noise": 0.0, "grad_norm": 1.1056, "lr": 0.0002975, "router_trainable": true, "sec_per_step": 1.249, "probe_sec": 0.2637, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.14058, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": -0.09331, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 3, "mean_delta_split": -0.09367, "n_merge": 2, "mean_delta_merge": 0.06213, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 1.89419, "trunc_abs_frac": 2.09415, "mean_delta_full": -0.08748, "read_hit_rate": {"surface": 0.70195, "split": 0.8934, "merge": 0.24715, "skel": 0.70996}, "mean_gate": {"surface": 0.18497, "split": 0.17712, "merge": 0.17954, "skel": 0.1788}, "n_reads": 24659, "n_cand": 36640, "read_suppressed_frac": 0.1017, "dropout_reads_logged": 24659, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1066, "admitted_surface": 16990, "staged_surface": 0, "admitted_split": 16599, "staged_split": 0, "admitted_merge": 20182, "staged_merge": 0, "admitted_skel": 11765, "staged_skel": 0, "mean_count": 185.2593231201172, "n_util_obs": 103, "util_count_spearman": 0.09369371348240566}, "probe_overhead_sec_total": 78.326, "probe_overhead_seq_frac": 0.4992, "util_loss": 0.002737, "alpha": 0.03, "t": 701.0} |
| {"event": "consolidate", "step": 700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 778.1} |
| {"event": "train", "step": 700, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.83423, "bpb": 2.64624, "chunk_loss": 0.031594, "F": 0.2438, "G": 0.25821, "chunks": 249.66, "bytes_per_chunk": 4.102, "boundary_entropy": 0.22097, "router_noise": 0.0, "grad_norm": 1.2222, "lr": 0.0002967, "router_trainable": true, "sec_per_step": 1.572, "probe_sec": 0.1179, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07265, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.08755, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.01941, "n_merge": 1, "mean_delta_merge": -0.63391, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.56089, "trunc_abs_frac": 0.56089, "mean_delta_full": -0.59094, "read_hit_rate": {"surface": 0.69034, "split": 0.88744, "merge": 0.25174, "skel": 0.68134}, "mean_gate": {"surface": 0.19221, "split": 0.17953, "merge": 0.1874, "skel": 0.18196}, "n_reads": 23121, "n_cand": 34790, "read_suppressed_frac": 0.1002, "dropout_reads_logged": 23121, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1138, "admitted_surface": 17089, "staged_surface": 0, "admitted_split": 16602, "staged_split": 0, "admitted_merge": 19994, "staged_merge": 0, "admitted_skel": 11851, "staged_skel": 0, "mean_count": 202.85910034179688, "n_util_obs": 108, "util_count_spearman": 0.12840225975782865}, "probe_overhead_sec_total": 87.105, "probe_overhead_seq_frac": 0.4993, "util_loss": 0.00292, "alpha": 0.03, "t": 779.7} |
| {"event": "consolidate", "step": 750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 833.6} |
| {"event": "train", "step": 750, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.78966, "bpb": 2.58194, "chunk_loss": 0.031537, "F": 0.24374, "G": 0.25431, "chunks": 249.59, "bytes_per_chunk": 4.103, "boundary_entropy": 0.22093, "router_noise": 0.0, "grad_norm": 1.0213, "lr": 0.0002959, "router_trainable": true, "sec_per_step": 1.112, "probe_sec": 0.1865, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.14816, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.04417, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": -0.00945, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.15084, "trunc_abs_frac": 0.67205, "mean_delta_full": 0.15039, "read_hit_rate": {"surface": 0.70057, "split": 0.89085, "merge": 0.25016, "skel": 0.70031}, "mean_gate": {"surface": 0.19388, "split": 0.18, "merge": 0.18608, "skel": 0.18405}, "n_reads": 23704, "n_cand": 35256, "read_suppressed_frac": 0.1022, "dropout_reads_logged": 23704, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1226, "admitted_surface": 17045, "staged_surface": 0, "admitted_split": 16553, "staged_split": 0, "admitted_merge": 20090, "staged_merge": 0, "admitted_skel": 11848, "staged_skel": 0, "mean_count": 220.60189819335938, "n_util_obs": 115, "util_count_spearman": 0.12583445119545492}, "probe_overhead_sec_total": 95.09, "probe_overhead_seq_frac": 0.4993, "util_loss": 0.0032, "alpha": 0.03, "t": 835.3} |
| {"event": "consolidate", "step": 800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 884.6} |
| {"event": "probe", "step": 800, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3951.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.22703, "cps": 30.867188, "mean_p": 0.242985, "boundary_entropy": 0.200899, "bpb": 2.402988}, "tr": {"chunks": 4192.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.231564, "cps": 32.75, "mean_p": 0.24486, "boundary_entropy": 0.243128, "bpb": 2.754501}, "fi": {"chunks": 5246.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.278879, "cps": 40.984375, "mean_p": 0.282698, "boundary_entropy": 0.239882, "bpb": 2.919702}}, "bpb": {"en": 2.40299, "tr": 2.7545, "fi": 2.9197}, "gini_cpc": 0.046871, "gini_cps": 0.064481, "bpic": 4.23721, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0199709743255616, "fi": 1.2283821397737078}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1296, "admitted_surface": 17086, "staged_surface": 0, "admitted_split": 16559, "staged_split": 0, "admitted_merge": 20052, "staged_merge": 0, "admitted_skel": 11839, "staged_skel": 0, "mean_count": 238.297119140625, "n_util_obs": 125, "util_count_spearman": 0.142568356374808}, "t": 886.9} |
| {"event": "eval", "step": 800, "phase": 1, "bpb": {"en": 2.41661, "tr": 2.76677, "fi": 2.92655}, "bytes_per_chunk": {"en": 4.1, "tr": 4.864, "fi": 3.771}, "bpb_mean": 2.70331, "frozen_inventory": true, "t": 890.6} |
| {"event": "train", "step": 800, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.78661, "bpb": 2.57753, "chunk_loss": 0.031587, "F": 0.24304, "G": 0.25711, "chunks": 248.88, "bytes_per_chunk": 4.115, "boundary_entropy": 0.20844, "router_noise": 0.0, "grad_norm": 0.9666, "lr": 0.0002949, "router_trainable": true, "sec_per_step": 1.176, "probe_sec": 0.1274, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07865, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.00407, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.03943, "n_merge": 2, "mean_delta_merge": 0.01376, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 1.19751, "trunc_abs_frac": 2.18697, "mean_delta_full": -0.04895, "read_hit_rate": {"surface": 0.68974, "split": 0.8872, "merge": 0.25405, "skel": 0.68967}, "mean_gate": {"surface": 0.20277, "split": 0.18258, "merge": 0.19416, "skel": 0.18714}, "n_reads": 23268, "n_cand": 34888, "read_suppressed_frac": 0.0995, "dropout_reads_logged": 23268, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1298, "admitted_surface": 17049, "staged_surface": 0, "admitted_split": 16631, "staged_split": 0, "admitted_merge": 20011, "staged_merge": 0, "admitted_skel": 11845, "staged_skel": 0, "mean_count": 238.6424102783203, "n_util_obs": 125, "util_count_spearman": 0.14504454685099846}, "probe_overhead_sec_total": 102.56, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.003662, "alpha": 0.03, "t": 894.0} |
| {"event": "consolidate", "step": 850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 944.9} |
| {"event": "train", "step": 850, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.75672, "bpb": 2.53441, "chunk_loss": 0.031529, "F": 0.24539, "G": 0.2526, "chunks": 251.28, "bytes_per_chunk": 4.075, "boundary_entropy": 0.20227, "router_noise": 0.0, "grad_norm": 1.0684, "lr": 0.0002938, "router_trainable": true, "sec_per_step": 1.045, "probe_sec": 0.1793, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.14267, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.13044, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": -0.02985, "n_merge": 2, "mean_delta_merge": -0.0415, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.9198, "trunc_abs_frac": 1.25533, "mean_delta_full": -0.08957, "read_hit_rate": {"surface": 0.69959, "split": 0.88468, "merge": 0.26464, "skel": 0.68506}, "mean_gate": {"surface": 0.20553, "split": 0.18086, "merge": 0.1936, "skel": 0.18723}, "n_reads": 23534, "n_cand": 35152, "read_suppressed_frac": 0.0993, "dropout_reads_logged": 23534, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1360, "admitted_surface": 17035, "staged_surface": 0, "admitted_split": 16619, "staged_split": 0, "admitted_merge": 20011, "staged_merge": 0, "admitted_skel": 11871, "staged_skel": 0, "mean_count": 256.70159912109375, "n_util_obs": 133, "util_count_spearman": 0.11842360310545699}, "probe_overhead_sec_total": 109.745, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.003664, "alpha": 0.03, "t": 946.3} |
| {"event": "consolidate", "step": 900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1005.6} |
| {"event": "train", "step": 900, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.59785, "bpb": 2.30521, "chunk_loss": 0.031301, "F": 0.23737, "G": 0.24696, "chunks": 243.06, "bytes_per_chunk": 4.213, "boundary_entropy": 0.18849, "router_noise": 0.0, "grad_norm": 0.9462, "lr": 0.0002927, "router_trainable": true, "sec_per_step": 1.225, "probe_sec": 0.1372, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0708, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.06893, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.42725, "n_merge": 1, "mean_delta_merge": 0.0954, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 4.24863, "trunc_abs_frac": 0.69369, "mean_delta_full": -0.01114, "read_hit_rate": {"surface": 0.70759, "split": 0.90131, "merge": 0.28299, "skel": 0.70353}, "mean_gate": {"surface": 0.20881, "split": 0.1798, "merge": 0.19757, "skel": 0.18791}, "n_reads": 23662, "n_cand": 34454, "read_suppressed_frac": 0.0975, "dropout_reads_logged": 23662, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1427, "admitted_surface": 17066, "staged_surface": 0, "admitted_split": 16517, "staged_split": 0, "admitted_merge": 20033, "staged_merge": 0, "admitted_skel": 11920, "staged_skel": 0, "mean_count": 274.78277587890625, "n_util_obs": 140, "util_count_spearman": 0.09801666265771578}, "probe_overhead_sec_total": 116.505, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.004475, "alpha": 0.03, "t": 1007.5} |
| {"event": "consolidate", "step": 950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1072.7} |
| {"event": "train", "step": 950, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.7557, "bpb": 2.53294, "chunk_loss": 0.031562, "F": 0.2439, "G": 0.25529, "chunks": 249.75, "bytes_per_chunk": 4.1, "boundary_entropy": 0.18777, "router_noise": 0.0, "grad_norm": 0.9363, "lr": 0.0002914, "router_trainable": true, "sec_per_step": 1.339, "probe_sec": 0.1186, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05601, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": -0.05741, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.28369, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.00578, "trunc_abs_frac": 1.48166, "mean_delta_full": 0.20139, "read_hit_rate": {"surface": 0.69309, "split": 0.88934, "merge": 0.25618, "skel": 0.69989}, "mean_gate": {"surface": 0.20551, "split": 0.17337, "merge": 0.19355, "skel": 0.18336}, "n_reads": 23473, "n_cand": 35002, "read_suppressed_frac": 0.099, "dropout_reads_logged": 23473, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1493, "admitted_surface": 17057, "staged_surface": 0, "admitted_split": 16551, "staged_split": 0, "admitted_merge": 20101, "staged_merge": 0, "admitted_skel": 11827, "staged_skel": 0, "mean_count": 292.69610595703125, "n_util_obs": 150, "util_count_spearman": 0.13490199564425084}, "probe_overhead_sec_total": 123.353, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.004687, "alpha": 0.03, "t": 1074.5} |
| {"event": "consolidate", "step": 1000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1124.4} |
| {"event": "probe", "step": 1000, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3947.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.2268, "cps": 30.835938, "mean_p": 0.236536, "boundary_entropy": 0.182185, "bpb": 2.326328}, "tr": {"chunks": 4430.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.244711, "cps": 34.609375, "mean_p": 0.238183, "boundary_entropy": 0.209009, "bpb": 2.626888}, "fi": {"chunks": 5164.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.27452, "cps": 40.34375, "mean_p": 0.282614, "boundary_entropy": 0.212809, "bpb": 2.791635}}, "bpb": {"en": 2.32633, "tr": 2.62689, "fi": 2.79163}, "gini_cpc": 0.042644, "gini_cps": 0.059917, "bpic": 4.189646, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0789719850550872, "fi": 1.2104067696079965}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1554, "admitted_surface": 17032, "staged_surface": 0, "admitted_split": 16549, "staged_split": 0, "admitted_merge": 20145, "staged_merge": 0, "admitted_skel": 11810, "staged_skel": 0, "mean_count": 310.4193115234375, "n_util_obs": 160, "util_count_spearman": 0.12952224417137262}, "t": 1126.4} |
| {"event": "eval", "step": 1000, "phase": 1, "bpb": {"en": 2.33609, "tr": 2.63727, "fi": 2.80014}, "bytes_per_chunk": {"en": 4.12, "tr": 4.509, "fi": 3.825}, "bpb_mean": 2.59117, "frozen_inventory": true, "t": 1129.8} |
| {"event": "train", "step": 1000, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.70487, "bpb": 2.45961, "chunk_loss": 0.031959, "F": 0.25504, "G": 0.2621, "chunks": 261.16, "bytes_per_chunk": 3.921, "boundary_entropy": 0.18607, "router_noise": 0.0, "grad_norm": 1.0239, "lr": 0.0002901, "router_trainable": true, "sec_per_step": 1.161, "probe_sec": 0.1782, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.1457, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.02526, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": 0.02097, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -5.8625, "trunc_abs_frac": 1.35038, "mean_delta_full": 0.01221, "read_hit_rate": {"surface": 0.70523, "split": 0.89811, "merge": 0.26975, "skel": 0.70426}, "mean_gate": {"surface": 0.20934, "split": 0.16875, "merge": 0.20133, "skel": 0.17987}, "n_reads": 24334, "n_cand": 35927, "read_suppressed_frac": 0.0978, "dropout_reads_logged": 24334, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1555, "admitted_surface": 17025, "staged_surface": 0, "admitted_split": 16443, "staged_split": 0, "admitted_merge": 20298, "staged_merge": 0, "admitted_skel": 11770, "staged_skel": 0, "mean_count": 310.7894287109375, "n_util_obs": 160, "util_count_spearman": 0.12918678160291583}, "probe_overhead_sec_total": 130.437, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.004778, "alpha": 0.03, "t": 1132.5} |
| {"event": "probe", "step": 1024, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 4007.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.230248, "cps": 31.304688, "mean_p": 0.238127, "boundary_entropy": 0.171598, "bpb": 2.303798}, "tr": {"chunks": 4448.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.245705, "cps": 34.75, "mean_p": 0.238936, "boundary_entropy": 0.200177, "bpb": 2.612099}, "fi": {"chunks": 5568.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.295997, "cps": 43.5, "mean_p": 0.285364, "boundary_entropy": 0.202523, "bpb": 2.786727}}, "bpb": {"en": 2.3038, "tr": 2.6121, "fi": 2.78673}, "gini_cpc": 0.056782, "gini_cps": 0.074211, "bpic": 4.045639, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0671341172392603, "fi": 1.2855593190891008}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1586, "admitted_surface": 17009, "staged_surface": 0, "admitted_split": 16545, "staged_split": 0, "admitted_merge": 20139, "staged_merge": 0, "admitted_skel": 11843, "staged_skel": 0, "mean_count": 319.04779052734375, "n_util_obs": 164, "util_count_spearman": 0.1352005767687348}, "t": 1154.5} |
| {"event": "consolidate", "step": 1050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1179.4} |
| {"event": "train", "step": 1050, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.72066, "bpb": 2.48239, "chunk_loss": 0.031495, "F": 0.24466, "G": 0.25261, "chunks": 250.53, "bytes_per_chunk": 4.087, "boundary_entropy": 0.18007, "router_noise": 0.0, "grad_norm": 0.8837, "lr": 0.0002886, "router_trainable": true, "sec_per_step": 0.968, "probe_sec": 0.1275, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08248, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": 0.11796, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.04504, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.94401, "trunc_abs_frac": 1.67636, "mean_delta_full": 0.25943, "read_hit_rate": {"surface": 0.6943, "split": 0.89395, "merge": 0.24582, "skel": 0.68513}, "mean_gate": {"surface": 0.21768, "split": 0.17452, "merge": 0.20348, "skel": 0.18998}, "n_reads": 23470, "n_cand": 35136, "read_suppressed_frac": 0.1005, "dropout_reads_logged": 23470, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1624, "admitted_surface": 17026, "staged_surface": 0, "admitted_split": 16464, "staged_split": 0, "admitted_merge": 20198, "staged_merge": 0, "admitted_skel": 11848, "staged_skel": 0, "mean_count": 328.90960693359375, "n_util_obs": 165, "util_count_spearman": 0.1401063232976251}, "probe_overhead_sec_total": 137.166, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.004951, "alpha": 0.03, "t": 1180.9} |
| {"event": "consolidate", "step": 1100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1228.0} |
| {"event": "train", "step": 1100, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.70676, "bpb": 2.46234, "chunk_loss": 0.03164, "F": 0.24689, "G": 0.25466, "chunks": 252.81, "bytes_per_chunk": 4.05, "boundary_entropy": 0.16932, "router_noise": 0.0, "grad_norm": 0.971, "lr": 0.0002871, "router_trainable": true, "sec_per_step": 0.977, "probe_sec": 0.0959, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05414, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": 0.17117, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.21313, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.44281, "trunc_abs_frac": 0.49116, "mean_delta_full": 0.13127, "read_hit_rate": {"surface": 0.6972, "split": 0.89668, "merge": 0.25941, "skel": 0.69405}, "mean_gate": {"surface": 0.21663, "split": 0.17157, "merge": 0.1967, "skel": 0.18647}, "n_reads": 23931, "n_cand": 35500, "read_suppressed_frac": 0.1027, "dropout_reads_logged": 23931, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1682, "admitted_surface": 17043, "staged_surface": 0, "admitted_split": 16472, "staged_split": 0, "admitted_merge": 20171, "staged_merge": 0, "admitted_skel": 11850, "staged_skel": 0, "mean_count": 346.88616943359375, "n_util_obs": 173, "util_count_spearman": 0.10594212236535315}, "probe_overhead_sec_total": 143.654, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.004766, "alpha": 0.03, "t": 1229.8} |
| {"event": "consolidate", "step": 1150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1286.5} |
| {"event": "train", "step": 1150, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.56525, "bpb": 2.25818, "chunk_loss": 0.03162, "F": 0.24936, "G": 0.25395, "chunks": 255.34, "bytes_per_chunk": 4.01, "boundary_entropy": 0.16484, "router_noise": 0.0, "grad_norm": 0.8487, "lr": 0.0002855, "router_trainable": true, "sec_per_step": 1.154, "probe_sec": 0.1341, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06973, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": 0.02404, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.01099, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -5.54301, "trunc_abs_frac": 0.14494, "mean_delta_full": 0.00568, "read_hit_rate": {"surface": 0.71667, "split": 0.90365, "merge": 0.28593, "skel": 0.71557}, "mean_gate": {"surface": 0.22118, "split": 0.16929, "merge": 0.20218, "skel": 0.18671}, "n_reads": 24682, "n_cand": 35747, "read_suppressed_frac": 0.0996, "dropout_reads_logged": 24682, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1745, "admitted_surface": 17010, "staged_surface": 0, "admitted_split": 16492, "staged_split": 0, "admitted_merge": 20238, "staged_merge": 0, "admitted_skel": 11796, "staged_skel": 0, "mean_count": 364.9102478027344, "n_util_obs": 183, "util_count_spearman": 0.12283333202791506}, "probe_overhead_sec_total": 150.624, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.005102, "alpha": 0.03, "t": 1287.5} |
| {"event": "consolidate", "step": 1200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1356.2} |
| {"event": "probe", "step": 1200, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3961.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.227604, "cps": 30.945312, "mean_p": 0.235492, "boundary_entropy": 0.15761, "bpb": 2.240596}, "tr": {"chunks": 4232.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.233773, "cps": 33.0625, "mean_p": 0.234831, "boundary_entropy": 0.18232, "bpb": 2.523259}, "fi": {"chunks": 5203.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.276593, "cps": 40.648438, "mean_p": 0.28259, "boundary_entropy": 0.186899, "bpb": 2.695209}}, "bpb": {"en": 2.2406, "tr": 2.52326, "fi": 2.69521}, "gini_cpc": 0.044256, "gini_cps": 0.061809, "bpic": 4.234996, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0271039168377345, "fi": 1.2152376613451568}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1800, "admitted_surface": 17029, "staged_surface": 0, "admitted_split": 16602, "staged_split": 0, "admitted_merge": 20049, "staged_merge": 0, "admitted_skel": 11856, "staged_skel": 0, "mean_count": 382.2795715332031, "n_util_obs": 186, "util_count_spearman": 0.10792255080680128}, "t": 1358.4} |
| {"event": "eval", "step": 1200, "phase": 1, "bpb": {"en": 2.25253, "tr": 2.53353, "fi": 2.70476}, "bytes_per_chunk": {"en": 4.133, "tr": 4.755, "fi": 3.841}, "bpb_mean": 2.49694, "frozen_inventory": true, "t": 1362.2} |
| {"event": "train", "step": 1200, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.60668, "bpb": 2.31795, "chunk_loss": 0.0315, "F": 0.24457, "G": 0.25018, "chunks": 250.44, "bytes_per_chunk": 4.089, "boundary_entropy": 0.15767, "router_noise": 0.0, "grad_norm": 0.7446, "lr": 0.0002837, "router_trainable": true, "sec_per_step": 1.549, "probe_sec": 0.1722, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09436, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.07739, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": -0.02563, "n_merge": 1, "mean_delta_merge": -0.06775, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.61575, "trunc_abs_frac": 0.60903, "mean_delta_full": 0.26813, "read_hit_rate": {"surface": 0.70985, "split": 0.89356, "merge": 0.28, "skel": 0.69582}, "mean_gate": {"surface": 0.22744, "split": 0.16993, "merge": 0.20652, "skel": 0.19273}, "n_reads": 23860, "n_cand": 35057, "read_suppressed_frac": 0.0956, "dropout_reads_logged": 23860, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1801, "admitted_surface": 16985, "staged_surface": 0, "admitted_split": 16588, "staged_split": 0, "admitted_merge": 20115, "staged_merge": 0, "admitted_skel": 11848, "staged_skel": 0, "mean_count": 382.6463623046875, "n_util_obs": 186, "util_count_spearman": 0.10801952547683098}, "probe_overhead_sec_total": 158.157, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.005145, "alpha": 0.03, "t": 1364.9} |
| {"event": "consolidate", "step": 1250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1409.3} |
| {"event": "train", "step": 1250, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.66436, "bpb": 2.40117, "chunk_loss": 0.031401, "F": 0.24103, "G": 0.24942, "chunks": 246.81, "bytes_per_chunk": 4.149, "boundary_entropy": 0.1633, "router_noise": 0.0, "grad_norm": 0.8784, "lr": 0.0002819, "router_trainable": true, "sec_per_step": 0.903, "probe_sec": 0.153, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.12837, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.02787, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": 0.10895, "n_merge": 2, "mean_delta_merge": -0.0835, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 2.19505, "trunc_abs_frac": 0.65566, "mean_delta_full": -0.03943, "read_hit_rate": {"surface": 0.67468, "split": 0.88777, "merge": 0.24279, "skel": 0.6788}, "mean_gate": {"surface": 0.22588, "split": 0.167, "merge": 0.20585, "skel": 0.19085}, "n_reads": 23014, "n_cand": 34861, "read_suppressed_frac": 0.0997, "dropout_reads_logged": 23014, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1856, "admitted_surface": 17048, "staged_surface": 0, "admitted_split": 16583, "staged_split": 0, "admitted_merge": 20057, "staged_merge": 0, "admitted_skel": 11848, "staged_skel": 0, "mean_count": 400.2563781738281, "n_util_obs": 194, "util_count_spearman": 0.12195025742378982}, "probe_overhead_sec_total": 163.653, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.005371, "alpha": 0.03, "t": 1410.1} |
| {"event": "consolidate", "step": 1300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1451.5} |
| {"event": "train", "step": 1300, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.60636, "bpb": 2.31749, "chunk_loss": 0.031279, "F": 0.23727, "G": 0.2451, "chunks": 242.97, "bytes_per_chunk": 4.215, "boundary_entropy": 0.15913, "router_noise": 0.0, "grad_norm": 0.8363, "lr": 0.00028, "router_trainable": true, "sec_per_step": 0.841, "probe_sec": 0.1578, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.12333, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.03696, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": 0.02259, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": -0.20719, "trunc_abs_frac": 0.20719, "mean_delta_full": 0.56567, "read_hit_rate": {"surface": 0.68565, "split": 0.88764, "merge": 0.25924, "skel": 0.68122}, "mean_gate": {"surface": 0.22526, "split": 0.16488, "merge": 0.20106, "skel": 0.18664}, "n_reads": 22981, "n_cand": 34427, "read_suppressed_frac": 0.099, "dropout_reads_logged": 22981, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1919, "admitted_surface": 17024, "staged_surface": 0, "admitted_split": 16589, "staged_split": 0, "admitted_merge": 20037, "staged_merge": 0, "admitted_skel": 11886, "staged_skel": 0, "mean_count": 417.7298583984375, "n_util_obs": 200, "util_count_spearman": 0.09900247506187655}, "probe_overhead_sec_total": 169.881, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.005846, "alpha": 0.03, "t": 1452.1} |
| {"event": "consolidate", "step": 1350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1492.5} |
| {"event": "train", "step": 1350, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.58975, "bpb": 2.29352, "chunk_loss": 0.031492, "F": 0.24298, "G": 0.24911, "chunks": 248.81, "bytes_per_chunk": 4.116, "boundary_entropy": 0.15409, "router_noise": 0.0, "grad_norm": 0.9134, "lr": 0.000278, "router_trainable": true, "sec_per_step": 0.822, "probe_sec": 0.116, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08677, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.02109, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.08136, "n_merge": 1, "mean_delta_merge": -0.09674, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.01504, "trunc_abs_frac": 0.76844, "mean_delta_full": -0.38806, "read_hit_rate": {"surface": 0.69887, "split": 0.89332, "merge": 0.2664, "skel": 0.70344}, "mean_gate": {"surface": 0.22436, "split": 0.15962, "merge": 0.20599, "skel": 0.182}, "n_reads": 23803, "n_cand": 35138, "read_suppressed_frac": 0.0992, "dropout_reads_logged": 23803, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1980, "admitted_surface": 16990, "staged_surface": 0, "admitted_split": 16581, "staged_split": 0, "admitted_merge": 20165, "staged_merge": 0, "admitted_skel": 11800, "staged_skel": 0, "mean_count": 435.10552978515625, "n_util_obs": 210, "util_count_spearman": 0.14489999608025436}, "probe_overhead_sec_total": 175.533, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.005774, "alpha": 0.03, "t": 1493.2} |
| {"event": "consolidate", "step": 1400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1547.1} |
| {"event": "probe", "step": 1400, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3943.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.22657, "cps": 30.804688, "mean_p": 0.227917, "boundary_entropy": 0.142186, "bpb": 2.204364}, "tr": {"chunks": 4308.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.237972, "cps": 33.65625, "mean_p": 0.232153, "boundary_entropy": 0.153868, "bpb": 2.469141}, "fi": {"chunks": 5115.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.271915, "cps": 39.960938, "mean_p": 0.274455, "boundary_entropy": 0.165862, "bpb": 2.610115}}, "bpb": {"en": 2.20436, "tr": 2.46914, "fi": 2.61011}, "gini_cpc": 0.041048, "gini_cps": 0.058457, "bpic": 4.244501, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0503220581178832, "fi": 1.200137753212398}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2047, "admitted_surface": 16956, "staged_surface": 0, "admitted_split": 16537, "staged_split": 0, "admitted_merge": 20198, "staged_merge": 0, "admitted_skel": 11845, "staged_skel": 0, "mean_count": 452.08251953125, "n_util_obs": 212, "util_count_spearman": 0.14642194578541626}, "t": 1549.0} |
| {"event": "eval", "step": 1400, "phase": 1, "bpb": {"en": 2.21529, "tr": 2.48067, "fi": 2.61927}, "bytes_per_chunk": {"en": 4.183, "tr": 4.71, "fi": 3.934}, "bpb_mean": 2.43841, "frozen_inventory": true, "t": 1552.5} |
| {"event": "train", "step": 1400, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.68237, "bpb": 2.42715, "chunk_loss": 0.0313, "F": 0.24066, "G": 0.24329, "chunks": 246.44, "bytes_per_chunk": 4.155, "boundary_entropy": 0.141, "router_noise": 0.0, "grad_norm": 0.8789, "lr": 0.0002759, "router_trainable": true, "sec_per_step": 1.227, "probe_sec": 0.109, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07351, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.01128, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.02637, "n_merge": 1, "mean_delta_merge": -0.20071, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.03886, "trunc_abs_frac": 1.74412, "mean_delta_full": -0.1665, "read_hit_rate": {"surface": 0.69175, "split": 0.88245, "merge": 0.2539, "skel": 0.68747}, "mean_gate": {"surface": 0.23282, "split": 0.16356, "merge": 0.20522, "skel": 0.18886}, "n_reads": 22886, "n_cand": 34425, "read_suppressed_frac": 0.1015, "dropout_reads_logged": 22886, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2047, "admitted_surface": 16973, "staged_surface": 0, "admitted_split": 16635, "staged_split": 0, "admitted_merge": 20117, "staged_merge": 0, "admitted_skel": 11811, "staged_skel": 0, "mean_count": 452.4359436035156, "n_util_obs": 212, "util_count_spearman": 0.1453438506384817}, "probe_overhead_sec_total": 183.074, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.006646, "alpha": 0.03, "t": 1554.6} |
| {"event": "consolidate", "step": 1450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1619.8} |
| {"event": "train", "step": 1450, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.63152, "bpb": 2.35378, "chunk_loss": 0.03124, "F": 0.23773, "G": 0.24312, "chunks": 243.44, "bytes_per_chunk": 4.206, "boundary_entropy": 0.14135, "router_noise": 0.0, "grad_norm": 0.7872, "lr": 0.0002738, "router_trainable": true, "sec_per_step": 1.329, "probe_sec": 0.1114, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06125, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.10613, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.13, "n_merge": 2, "mean_delta_merge": 0.29236, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 5.98025, "trunc_abs_frac": 1.54756, "mean_delta_full": 0.017, "read_hit_rate": {"surface": 0.68355, "split": 0.87898, "merge": 0.24929, "skel": 0.68664}, "mean_gate": {"surface": 0.228, "split": 0.15662, "merge": 0.20146, "skel": 0.18195}, "n_reads": 22642, "n_cand": 34239, "read_suppressed_frac": 0.0995, "dropout_reads_logged": 22642, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2104, "admitted_surface": 16973, "staged_surface": 0, "admitted_split": 16675, "staged_split": 0, "admitted_merge": 20094, "staged_merge": 0, "admitted_skel": 11794, "staged_skel": 0, "mean_count": 469.8567810058594, "n_util_obs": 220, "util_count_spearman": 0.10043671075097345}, "probe_overhead_sec_total": 190.618, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.006522, "alpha": 0.03, "t": 1621.0} |
| {"event": "consolidate", "step": 1500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1679.7} |
| {"event": "train", "step": 1500, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.58441, "bpb": 2.28582, "chunk_loss": 0.031106, "F": 0.23276, "G": 0.23564, "chunks": 238.34, "bytes_per_chunk": 4.296, "boundary_entropy": 0.1345, "router_noise": 0.0, "grad_norm": 0.8304, "lr": 0.0002715, "router_trainable": true, "sec_per_step": 1.187, "probe_sec": 0.1365, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10928, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": 0.06173, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.39122, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 1.36264, "trunc_abs_frac": 4.87431, "mean_delta_full": 0.14861, "read_hit_rate": {"surface": 0.69045, "split": 0.88037, "merge": 0.2577, "skel": 0.67603}, "mean_gate": {"surface": 0.23057, "split": 0.15721, "merge": 0.20244, "skel": 0.18771}, "n_reads": 22369, "n_cand": 33693, "read_suppressed_frac": 0.1008, "dropout_reads_logged": 22369, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2163, "admitted_surface": 16960, "staged_surface": 0, "admitted_split": 16648, "staged_split": 0, "admitted_merge": 20035, "staged_merge": 0, "admitted_skel": 11893, "staged_skel": 0, "mean_count": 487.3040771484375, "n_util_obs": 228, "util_count_spearman": 0.08693076960655675}, "probe_overhead_sec_total": 197.136, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.00635, "alpha": 0.03, "t": 1680.4} |
| {"event": "consolidate", "step": 1550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1719.8} |
| {"event": "train", "step": 1550, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.55597, "bpb": 2.24478, "chunk_loss": 0.031226, "F": 0.23785, "G": 0.23914, "chunks": 243.56, "bytes_per_chunk": 4.204, "boundary_entropy": 0.13539, "router_noise": 0.0, "grad_norm": 0.7179, "lr": 0.0002692, "router_trainable": true, "sec_per_step": 0.813, "probe_sec": 0.1062, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06526, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.09779, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.11426, "n_merge": 1, "mean_delta_merge": 0.06348, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.68915, "trunc_abs_frac": 0.99879, "mean_delta_full": 0.19409, "read_hit_rate": {"surface": 0.69518, "split": 0.88667, "merge": 0.26662, "skel": 0.69334}, "mean_gate": {"surface": 0.22684, "split": 0.14982, "merge": 0.19776, "skel": 0.18372}, "n_reads": 22997, "n_cand": 34239, "read_suppressed_frac": 0.0972, "dropout_reads_logged": 22997, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2213, "admitted_surface": 16985, "staged_surface": 0, "admitted_split": 16596, "staged_split": 0, "admitted_merge": 20114, "staged_merge": 0, "admitted_skel": 11841, "staged_skel": 0, "mean_count": 504.616943359375, "n_util_obs": 235, "util_count_spearman": 0.07485274672436591}, "probe_overhead_sec_total": 202.439, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.006381, "alpha": 0.03, "t": 1721.0} |
| {"event": "consolidate", "step": 1600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1760.6} |
| {"event": "probe", "step": 1600, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3947.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.2268, "cps": 30.835938, "mean_p": 0.226231, "boundary_entropy": 0.124399, "bpb": 2.1604}, "tr": {"chunks": 4222.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.233221, "cps": 32.984375, "mean_p": 0.224478, "boundary_entropy": 0.134847, "bpb": 2.400296}, "fi": {"chunks": 5214.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.277178, "cps": 40.734375, "mean_p": 0.273853, "boundary_entropy": 0.14243, "bpb": 2.551887}}, "bpb": {"en": 2.1604, "tr": 2.4003, "fi": 2.55189}, "gini_cpc": 0.045558, "gini_cps": 0.063115, "bpic": 4.239109, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0283114494136745, "fi": 1.2221264323656262}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2255, "admitted_surface": 16959, "staged_surface": 0, "admitted_split": 16668, "staged_split": 0, "admitted_merge": 20025, "staged_merge": 0, "admitted_skel": 11884, "staged_skel": 0, "mean_count": 521.5272216796875, "n_util_obs": 244, "util_count_spearman": 0.04243791019747304}, "t": 1762.6} |
| {"event": "eval", "step": 1600, "phase": 1, "bpb": {"en": 2.17066, "tr": 2.41135, "fi": 2.56251}, "bytes_per_chunk": {"en": 4.182, "tr": 4.792, "fi": 3.822}, "bpb_mean": 2.38151, "frozen_inventory": true, "t": 1765.7} |
| {"event": "train", "step": 1600, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.55584, "bpb": 2.2446, "chunk_loss": 0.031258, "F": 0.2388, "G": 0.24014, "chunks": 244.53, "bytes_per_chunk": 4.188, "boundary_entropy": 0.12373, "router_noise": 0.0, "grad_norm": 0.7803, "lr": 0.0002668, "router_trainable": true, "sec_per_step": 0.932, "probe_sec": 0.1137, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08319, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.2174, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": -0.0535, "n_merge": 1, "mean_delta_merge": 0.57129, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": -0.11553, "trunc_abs_frac": 0.11553, "mean_delta_full": 0.56689, "read_hit_rate": {"surface": 0.69922, "split": 0.88891, "merge": 0.27716, "skel": 0.69003}, "mean_gate": {"surface": 0.23396, "split": 0.15409, "merge": 0.20133, "skel": 0.18917}, "n_reads": 23276, "n_cand": 34453, "read_suppressed_frac": 0.0991, "dropout_reads_logged": 23276, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2257, "admitted_surface": 16942, "staged_surface": 0, "admitted_split": 16637, "staged_split": 0, "admitted_merge": 20111, "staged_merge": 0, "admitted_skel": 11846, "staged_skel": 0, "mean_count": 521.8923950195312, "n_util_obs": 244, "util_count_spearman": 0.04452870401348153}, "probe_overhead_sec_total": 207.856, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.0069, "alpha": 0.03, "t": 1767.6} |
| {"event": "consolidate", "step": 1650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1808.3} |
| {"event": "train", "step": 1650, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.581, "bpb": 2.28091, "chunk_loss": 0.031137, "F": 0.23541, "G": 0.23544, "chunks": 241.06, "bytes_per_chunk": 4.248, "boundary_entropy": 0.12869, "router_noise": 0.0, "grad_norm": 0.7674, "lr": 0.0002643, "router_trainable": true, "sec_per_step": 0.827, "probe_sec": 0.0688, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0447, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.01366, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 3, "mean_delta_split": -0.00857, "n_merge": 1, "mean_delta_merge": 0.15472, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -0.09066, "trunc_abs_frac": 8.39698, "mean_delta_full": -0.2336, "read_hit_rate": {"surface": 0.68719, "split": 0.88511, "merge": 0.25569, "skel": 0.68376}, "mean_gate": {"surface": 0.23286, "split": 0.15013, "merge": 0.20407, "skel": 0.18551}, "n_reads": 22623, "n_cand": 33977, "read_suppressed_frac": 0.0971, "dropout_reads_logged": 22623, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2320, "admitted_surface": 16977, "staged_surface": 0, "admitted_split": 16635, "staged_split": 0, "admitted_merge": 20100, "staged_merge": 0, "admitted_skel": 11824, "staged_skel": 0, "mean_count": 539.1767578125, "n_util_obs": 249, "util_count_spearman": 0.018408083948698018}, "probe_overhead_sec_total": 213.632, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.007507, "alpha": 0.03, "t": 1809.0} |
| {"event": "consolidate", "step": 1700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1854.8} |
| {"event": "train", "step": 1700, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.5979, "bpb": 2.30528, "chunk_loss": 0.03123, "F": 0.2384, "G": 0.23944, "chunks": 244.12, "bytes_per_chunk": 4.195, "boundary_entropy": 0.12091, "router_noise": 0.0, "grad_norm": 0.719, "lr": 0.0002618, "router_trainable": true, "sec_per_step": 0.952, "probe_sec": 0.1315, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07707, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.04539, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 3, "mean_delta_split": 0.08965, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.41612, "trunc_abs_frac": 0.6488, "mean_delta_full": 0.18555, "read_hit_rate": {"surface": 0.68792, "split": 0.88068, "merge": 0.26471, "skel": 0.67379}, "mean_gate": {"surface": 0.23228, "split": 0.14741, "merge": 0.19953, "skel": 0.18605}, "n_reads": 22801, "n_cand": 34343, "read_suppressed_frac": 0.0997, "dropout_reads_logged": 22801, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2367, "admitted_surface": 16961, "staged_surface": 0, "admitted_split": 16671, "staged_split": 0, "admitted_merge": 19998, "staged_merge": 0, "admitted_skel": 11906, "staged_skel": 0, "mean_count": 556.4568481445312, "n_util_obs": 256, "util_count_spearman": 0.054044136975046723}, "probe_overhead_sec_total": 219.955, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.00665, "alpha": 0.03, "t": 1856.6} |
| {"event": "consolidate", "step": 1750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1922.8} |
| {"event": "train", "step": 1750, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.55207, "bpb": 2.23916, "chunk_loss": 0.030905, "F": 0.22562, "G": 0.23161, "chunks": 231.03, "bytes_per_chunk": 4.432, "boundary_entropy": 0.12134, "router_noise": 0.0, "grad_norm": 0.7362, "lr": 0.0002591, "router_trainable": true, "sec_per_step": 1.349, "probe_sec": 0.139, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.11136, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.06947, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": -0.09792, "n_merge": 1, "mean_delta_merge": -0.05774, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.36565, "trunc_abs_frac": 0.36565, "mean_delta_full": -0.58765, "read_hit_rate": {"surface": 0.67138, "split": 0.87565, "merge": 0.24737, "skel": 0.65716}, "mean_gate": {"surface": 0.23222, "split": 0.14468, "merge": 0.20418, "skel": 0.17927}, "n_reads": 21324, "n_cand": 32667, "read_suppressed_frac": 0.1012, "dropout_reads_logged": 21324, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2417, "admitted_surface": 16979, "staged_surface": 0, "admitted_split": 16733, "staged_split": 0, "admitted_merge": 19891, "staged_merge": 0, "admitted_skel": 11933, "staged_skel": 0, "mean_count": 573.3153686523438, "n_util_obs": 264, "util_count_spearman": 0.07320403726692981}, "probe_overhead_sec_total": 227.434, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.00665, "alpha": 0.03, "t": 1924.0} |
| {"event": "consolidate", "step": 1800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1983.2} |
| {"event": "probe", "step": 1800, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3902.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.224214, "cps": 30.484375, "mean_p": 0.222882, "boundary_entropy": 0.128227, "bpb": 2.125188}, "tr": {"chunks": 4222.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.233221, "cps": 32.984375, "mean_p": 0.224849, "boundary_entropy": 0.13749, "bpb": 2.369119}, "fi": {"chunks": 5074.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.269736, "cps": 39.640625, "mean_p": 0.268415, "boundary_entropy": 0.130817, "bpb": 2.498968}}, "bpb": {"en": 2.12519, "tr": 2.36912, "fi": 2.49897}, "gini_cpc": 0.041734, "gini_cps": 0.059201, "bpic": 4.29853, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0401704999579122, "fi": 1.203027166482549}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2462, "admitted_surface": 16960, "staged_surface": 0, "admitted_split": 16734, "staged_split": 0, "admitted_merge": 19947, "staged_merge": 0, "admitted_skel": 11895, "staged_skel": 0, "mean_count": 590.2750244140625, "n_util_obs": 270, "util_count_spearman": 0.07855252983999716}, "t": 1985.1} |
| {"event": "eval", "step": 1800, "phase": 1, "bpb": {"en": 2.13673, "tr": 2.38172, "fi": 2.50792}, "bytes_per_chunk": {"en": 4.197, "tr": 4.891, "fi": 3.904}, "bpb_mean": 2.34213, "frozen_inventory": true, "t": 1988.4} |
| {"event": "train", "step": 1800, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.51444, "bpb": 2.18487, "chunk_loss": 0.031045, "F": 0.23386, "G": 0.23201, "chunks": 239.47, "bytes_per_chunk": 4.276, "boundary_entropy": 0.12409, "router_noise": 0.0, "grad_norm": 0.7626, "lr": 0.0002564, "router_trainable": true, "sec_per_step": 1.327, "probe_sec": 0.1523, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.12282, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.06507, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.04108, "n_merge": 2, "mean_delta_merge": 0.11325, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.37203, "trunc_abs_frac": 0.9116, "mean_delta_full": 0.15438, "read_hit_rate": {"surface": 0.70069, "split": 0.88686, "merge": 0.26925, "skel": 0.69147}, "mean_gate": {"surface": 0.23456, "split": 0.14627, "merge": 0.20412, "skel": 0.18564}, "n_reads": 22767, "n_cand": 33800, "read_suppressed_frac": 0.1026, "dropout_reads_logged": 22767, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2463, "admitted_surface": 16911, "staged_surface": 0, "admitted_split": 16704, "staged_split": 0, "admitted_merge": 20028, "staged_merge": 0, "admitted_skel": 11893, "staged_skel": 0, "mean_count": 590.6304321289062, "n_util_obs": 270, "util_count_spearman": 0.07759929643906105}, "probe_overhead_sec_total": 234.416, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.007038, "alpha": 0.03, "t": 1990.4} |
| {"event": "consolidate", "step": 1850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2029.2} |
| {"event": "train", "step": 1850, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.55118, "bpb": 2.23787, "chunk_loss": 0.031146, "F": 0.23544, "G": 0.23811, "chunks": 241.09, "bytes_per_chunk": 4.247, "boundary_entropy": 0.11873, "router_noise": 0.0, "grad_norm": 0.7889, "lr": 0.0002537, "router_trainable": true, "sec_per_step": 0.797, "probe_sec": 0.1123, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07881, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.15733, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": -0.02921, "n_merge": 1, "mean_delta_merge": -0.125, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.02898, "trunc_abs_frac": 2.30009, "mean_delta_full": -0.04529, "read_hit_rate": {"surface": 0.67708, "split": 0.87927, "merge": 0.24716, "skel": 0.67948}, "mean_gate": {"surface": 0.22177, "split": 0.1347, "merge": 0.19267, "skel": 0.17186}, "n_reads": 22379, "n_cand": 33990, "read_suppressed_frac": 0.1015, "dropout_reads_logged": 22379, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2510, "admitted_surface": 16914, "staged_surface": 0, "admitted_split": 16692, "staged_split": 0, "admitted_merge": 20070, "staged_merge": 0, "admitted_skel": 11860, "staged_skel": 0, "mean_count": 607.589599609375, "n_util_obs": 282, "util_count_spearman": 0.088270729115971}, "probe_overhead_sec_total": 239.845, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.007497, "alpha": 0.03, "t": 2030.3} |
| {"event": "consolidate", "step": 1900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2074.1} |
| {"event": "train", "step": 1900, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.48541, "bpb": 2.143, "chunk_loss": 0.030921, "F": 0.22601, "G": 0.23119, "chunks": 231.44, "bytes_per_chunk": 4.425, "boundary_entropy": 0.11444, "router_noise": 0.0, "grad_norm": 0.7036, "lr": 0.0002508, "router_trainable": true, "sec_per_step": 0.898, "probe_sec": 0.1149, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08144, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.03291, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": -0.05968, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.70886, "trunc_abs_frac": 0.872, "mean_delta_full": -0.10236, "read_hit_rate": {"surface": 0.66816, "split": 0.87295, "merge": 0.2374, "skel": 0.66013}, "mean_gate": {"surface": 0.22785, "split": 0.14045, "merge": 0.1951, "skel": 0.17846}, "n_reads": 21223, "n_cand": 32717, "read_suppressed_frac": 0.0979, "dropout_reads_logged": 21223, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2571, "admitted_surface": 16938, "staged_surface": 0, "admitted_split": 16747, "staged_split": 0, "admitted_merge": 19895, "staged_merge": 0, "admitted_skel": 11956, "staged_skel": 0, "mean_count": 623.6678466796875, "n_util_obs": 286, "util_count_spearman": 0.026722699636691934}, "probe_overhead_sec_total": 245.391, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.007379, "alpha": 0.03, "t": 2075.2} |
| {"event": "consolidate", "step": 1950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2130.3} |
| {"event": "train", "step": 1950, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.56247, "bpb": 2.25417, "chunk_loss": 0.031119, "F": 0.23227, "G": 0.23675, "chunks": 237.84, "bytes_per_chunk": 4.305, "boundary_entropy": 0.11112, "router_noise": 0.0, "grad_norm": 0.7657, "lr": 0.0002479, "router_trainable": true, "sec_per_step": 1.137, "probe_sec": 0.1276, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06885, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": 0.05311, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.04242, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 7, "trunc_ratio_sum": -0.05342, "trunc_abs_frac": 1.53566, "mean_delta_full": 0.17615, "read_hit_rate": {"surface": 0.665, "split": 0.87545, "merge": 0.23851, "skel": 0.66815}, "mean_gate": {"surface": 0.22363, "split": 0.13763, "merge": 0.19726, "skel": 0.17492}, "n_reads": 21818, "n_cand": 33565, "read_suppressed_frac": 0.099, "dropout_reads_logged": 21818, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2618, "admitted_surface": 16956, "staged_surface": 0, "admitted_split": 16741, "staged_split": 0, "admitted_merge": 19946, "staged_merge": 0, "admitted_skel": 11893, "staged_skel": 0, "mean_count": 639.6273193359375, "n_util_obs": 293, "util_count_spearman": 0.03565489456898223}, "probe_overhead_sec_total": 252.873, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.007798, "alpha": 0.03, "t": 2132.1} |
| {"event": "consolidate", "step": 2000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2169.5} |
| {"event": "probe", "step": 2000, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3670.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.210883, "cps": 28.671875, "mean_p": 0.215648, "boundary_entropy": 0.105954, "bpb": 2.106747}, "tr": {"chunks": 4119.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.227531, "cps": 32.179688, "mean_p": 0.219449, "boundary_entropy": 0.118756, "bpb": 2.324177}, "fi": {"chunks": 5132.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.272819, "cps": 40.09375, "mean_p": 0.270046, "boundary_entropy": 0.119823, "bpb": 2.45039}}, "bpb": {"en": 2.10675, "tr": 2.32418, "fi": 2.45039}, "gini_cpc": 0.058055, "gini_cps": 0.075433, "bpic": 4.390682, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0789449744205162, "fi": 1.2936977422190652}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2663, "admitted_surface": 16922, "staged_surface": 0, "admitted_split": 16837, "staged_split": 0, "admitted_merge": 19832, "staged_merge": 0, "admitted_skel": 11945, "staged_skel": 0, "mean_count": 655.31591796875, "n_util_obs": 303, "util_count_spearman": 0.042249606825782896}, "t": 2171.2} |
| {"event": "eval", "step": 2000, "phase": 1, "bpb": {"en": 2.11884, "tr": 2.33549, "fi": 2.46166}, "bytes_per_chunk": {"en": 4.441, "tr": 4.941, "fi": 3.901}, "bpb_mean": 2.30533, "frozen_inventory": true, "t": 2175.0} |
| {"event": "train", "step": 2000, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.52992, "bpb": 2.20721, "chunk_loss": 0.031098, "F": 0.23383, "G": 0.23761, "chunks": 239.44, "bytes_per_chunk": 4.277, "boundary_entropy": 0.10484, "router_noise": 0.0, "grad_norm": 0.7675, "lr": 0.000245, "router_trainable": true, "sec_per_step": 0.918, "probe_sec": 0.1536, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10999, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.02541, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.03186, "n_merge": 1, "mean_delta_merge": 0.14433, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.36946, "trunc_abs_frac": 0.82504, "mean_delta_full": -0.25593, "read_hit_rate": {"surface": 0.6844, "split": 0.87714, "merge": 0.25849, "skel": 0.66618}, "mean_gate": {"surface": 0.22047, "split": 0.13269, "merge": 0.19366, "skel": 0.17001}, "n_reads": 21994, "n_cand": 33425, "read_suppressed_frac": 0.098, "dropout_reads_logged": 21994, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2663, "admitted_surface": 16886, "staged_surface": 0, "admitted_split": 16768, "staged_split": 0, "admitted_merge": 19969, "staged_merge": 0, "admitted_skel": 11913, "staged_skel": 0, "mean_count": 655.6541137695312, "n_util_obs": 303, "util_count_spearman": 0.04205936452829668}, "probe_overhead_sec_total": 257.21, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.007538, "alpha": 0.03, "t": 2177.9} |
| {"event": "probe", "step": 2048, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3670.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.210883, "cps": 28.671875, "mean_p": 0.215421, "boundary_entropy": 0.108704, "bpb": 2.095436}, "tr": {"chunks": 4109.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.226979, "cps": 32.101562, "mean_p": 0.217242, "boundary_entropy": 0.113908, "bpb": 2.32316}, "fi": {"chunks": 5069.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.26947, "cps": 39.601562, "mean_p": 0.265714, "boundary_entropy": 0.114782, "bpb": 2.43454}}, "bpb": {"en": 2.09544, "tr": 2.32316, "fi": 2.43454}, "gini_cpc": 0.055219, "gini_cps": 0.072592, "bpic": 4.415629, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.076325540153897, "fi": 1.2778164176360953}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2698, "admitted_surface": 16965, "staged_surface": 0, "admitted_split": 16786, "staged_split": 0, "admitted_merge": 19855, "staged_merge": 0, "admitted_skel": 11930, "staged_skel": 0, "mean_count": 670.77734375, "n_util_obs": 313, "util_count_spearman": 0.0271436353475791}, "t": 2241.6} |
| {"event": "consolidate", "step": 2050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2245.8} |
| {"event": "train", "step": 2050, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.54234, "bpb": 2.22512, "chunk_loss": 0.030697, "F": 0.21716, "G": 0.22177, "chunks": 222.38, "bytes_per_chunk": 4.605, "boundary_entropy": 0.10615, "router_noise": 0.0, "grad_norm": 0.7537, "lr": 0.0002419, "router_trainable": true, "sec_per_step": 1.381, "probe_sec": 0.0899, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05587, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": 0.03391, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 5, "mean_delta_merge": -0.32641, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 1.01157, "trunc_abs_frac": 2.03628, "mean_delta_full": -0.07471, "read_hit_rate": {"surface": 0.65471, "split": 0.85502, "merge": 0.24135, "skel": 0.63565}, "mean_gate": {"surface": 0.23393, "split": 0.14294, "merge": 0.20774, "skel": 0.18409}, "n_reads": 20102, "n_cand": 31580, "read_suppressed_frac": 0.0994, "dropout_reads_logged": 20102, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2701, "admitted_surface": 16935, "staged_surface": 0, "admitted_split": 16941, "staged_split": 0, "admitted_merge": 19690, "staged_merge": 0, "admitted_skel": 11970, "staged_skel": 0, "mean_count": 671.740234375, "n_util_obs": 313, "util_count_spearman": 0.027434791252173886}, "probe_overhead_sec_total": 263.655, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.008036, "alpha": 0.03, "t": 2247.0} |
| {"event": "consolidate", "step": 2100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2299.2} |
| {"event": "train", "step": 2100, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.46625, "bpb": 2.11535, "chunk_loss": 0.030773, "F": 0.22162, "G": 0.22446, "chunks": 226.94, "bytes_per_chunk": 4.512, "boundary_entropy": 0.09795, "router_noise": 0.0, "grad_norm": 0.69, "lr": 0.0002388, "router_trainable": true, "sec_per_step": 1.06, "probe_sec": 0.083, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04537, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.08034, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": -0.13956, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.00582, "trunc_abs_frac": 1.16651, "mean_delta_full": 0.0918, "read_hit_rate": {"surface": 0.67068, "split": 0.86963, "merge": 0.25035, "skel": 0.65158}, "mean_gate": {"surface": 0.21878, "split": 0.12962, "merge": 0.19144, "skel": 0.17045}, "n_reads": 20814, "n_cand": 32055, "read_suppressed_frac": 0.1014, "dropout_reads_logged": 20814, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2743, "admitted_surface": 16903, "staged_surface": 0, "admitted_split": 16821, "staged_split": 0, "admitted_merge": 19852, "staged_merge": 0, "admitted_skel": 11960, "staged_skel": 0, "mean_count": 687.89013671875, "n_util_obs": 318, "util_count_spearman": 0.04033589957754618}, "probe_overhead_sec_total": 269.671, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.008025, "alpha": 0.03, "t": 2300.0} |
| {"event": "consolidate", "step": 2150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2346.8} |
| {"event": "train", "step": 2150, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.47596, "bpb": 2.12935, "chunk_loss": 0.031035, "F": 0.23178, "G": 0.23432, "chunks": 237.34, "bytes_per_chunk": 4.314, "boundary_entropy": 0.09589, "router_noise": 0.0, "grad_norm": 0.74, "lr": 0.0002357, "router_trainable": true, "sec_per_step": 0.964, "probe_sec": 0.1383, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.1058, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.0705, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.08093, "n_merge": 1, "mean_delta_merge": 0.17047, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 2.17152, "trunc_abs_frac": 15.93133, "mean_delta_full": 0.10767, "read_hit_rate": {"surface": 0.67605, "split": 0.87451, "merge": 0.26238, "skel": 0.67028}, "mean_gate": {"surface": 0.21809, "split": 0.12643, "merge": 0.19087, "skel": 0.16693}, "n_reads": 21980, "n_cand": 33413, "read_suppressed_frac": 0.097, "dropout_reads_logged": 21980, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2781, "admitted_surface": 16929, "staged_surface": 0, "admitted_split": 16773, "staged_split": 0, "admitted_merge": 19937, "staged_merge": 0, "admitted_skel": 11897, "staged_skel": 0, "mean_count": 704.06591796875, "n_util_obs": 326, "util_count_spearman": 0.017830290382775327}, "probe_overhead_sec_total": 275.595, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.008466, "alpha": 0.03, "t": 2348.1} |
| {"event": "consolidate", "step": 2200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2394.7} |
| {"event": "probe", "step": 2200, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3674.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.211113, "cps": 28.703125, "mean_p": 0.214269, "boundary_entropy": 0.101765, "bpb": 2.073457}, "tr": {"chunks": 4048.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.223609, "cps": 31.625, "mean_p": 0.215164, "boundary_entropy": 0.122319, "bpb": 2.282949}, "fi": {"chunks": 5072.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.269629, "cps": 39.625, "mean_p": 0.266019, "boundary_entropy": 0.107625, "bpb": 2.424825}}, "bpb": {"en": 2.07346, "tr": 2.28295, "fi": 2.42483}, "gini_cpc": 0.055386, "gini_cps": 0.072847, "bpic": 4.434266, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0591925578219907, "fi": 1.2771806486822896}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2819, "admitted_surface": 16910, "staged_surface": 0, "admitted_split": 16854, "staged_split": 0, "admitted_merge": 19846, "staged_merge": 0, "admitted_skel": 11926, "staged_skel": 0, "mean_count": 719.8218994140625, "n_util_obs": 332, "util_count_spearman": 0.01481979194642537}, "t": 2396.9} |
| {"event": "eval", "step": 2200, "phase": 1, "bpb": {"en": 2.08655, "tr": 2.29289, "fi": 2.43611}, "bytes_per_chunk": {"en": 4.43, "tr": 5.005, "fi": 3.945}, "bpb_mean": 2.27185, "frozen_inventory": true, "t": 2400.6} |
| {"event": "train", "step": 2200, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.51399, "bpb": 2.18423, "chunk_loss": 0.031025, "F": 0.22842, "G": 0.23126, "chunks": 233.91, "bytes_per_chunk": 4.378, "boundary_entropy": 0.10222, "router_noise": 0.0, "grad_norm": 0.7108, "lr": 0.0002325, "router_trainable": true, "sec_per_step": 1.095, "probe_sec": 0.1455, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10209, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.07587, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 4, "mean_delta_merge": 0.02203, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": -0.06696, "trunc_abs_frac": 0.06696, "mean_delta_full": -0.21057, "read_hit_rate": {"surface": 0.66738, "split": 0.87326, "merge": 0.2439, "skel": 0.66322}, "mean_gate": {"surface": 0.22093, "split": 0.12503, "merge": 0.19229, "skel": 0.16614}, "n_reads": 21460, "n_cand": 33007, "read_suppressed_frac": 0.0987, "dropout_reads_logged": 21460, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2821, "admitted_surface": 16912, "staged_surface": 0, "admitted_split": 16798, "staged_split": 0, "admitted_merge": 19941, "staged_merge": 0, "admitted_skel": 11885, "staged_skel": 0, "mean_count": 720.1441650390625, "n_util_obs": 332, "util_count_spearman": 0.015152963475190009}, "probe_overhead_sec_total": 281.669, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.007931, "alpha": 0.03, "t": 2402.9} |
| {"event": "consolidate", "step": 2250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2452.3} |
| {"event": "train", "step": 2250, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.49551, "bpb": 2.15757, "chunk_loss": 0.030858, "F": 0.22498, "G": 0.22843, "chunks": 230.38, "bytes_per_chunk": 4.445, "boundary_entropy": 0.09485, "router_noise": 0.0, "grad_norm": 0.7556, "lr": 0.0002293, "router_trainable": true, "sec_per_step": 1.012, "probe_sec": 0.1394, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.113, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.0094, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": -0.2439, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.69805, "trunc_abs_frac": 1.84637, "mean_delta_full": -0.08731, "read_hit_rate": {"surface": 0.68324, "split": 0.8781, "merge": 0.25848, "skel": 0.66906}, "mean_gate": {"surface": 0.21631, "split": 0.12305, "merge": 0.18667, "skel": 0.16543}, "n_reads": 21352, "n_cand": 32362, "read_suppressed_frac": 0.1028, "dropout_reads_logged": 21352, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2859, "admitted_surface": 16906, "staged_surface": 0, "admitted_split": 16809, "staged_split": 0, "admitted_merge": 19932, "staged_merge": 0, "admitted_skel": 11889, "staged_skel": 0, "mean_count": 736.1901245117188, "n_util_obs": 337, "util_count_spearman": 0.0263945730309715}, "probe_overhead_sec_total": 288.056, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.009284, "alpha": 0.03, "t": 2453.5} |
| {"event": "consolidate", "step": 2300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2517.0} |
| {"event": "train", "step": 2300, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.52808, "bpb": 2.20456, "chunk_loss": 0.030882, "F": 0.22507, "G": 0.22762, "chunks": 230.47, "bytes_per_chunk": 4.443, "boundary_entropy": 0.09779, "router_noise": 0.0, "grad_norm": 0.6758, "lr": 0.000226, "router_trainable": true, "sec_per_step": 1.308, "probe_sec": 0.1083, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07746, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.02359, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.10428, "n_merge": 2, "mean_delta_merge": 0.03973, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.32724, "trunc_abs_frac": 0.40038, "mean_delta_full": 0.19946, "read_hit_rate": {"surface": 0.66458, "split": 0.861, "merge": 0.25236, "skel": 0.64372}, "mean_gate": {"surface": 0.22384, "split": 0.12762, "merge": 0.19457, "skel": 0.17064}, "n_reads": 20693, "n_cand": 32193, "read_suppressed_frac": 0.0996, "dropout_reads_logged": 20693, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2895, "admitted_surface": 16911, "staged_surface": 0, "admitted_split": 16902, "staged_split": 0, "admitted_merge": 19781, "staged_merge": 0, "admitted_skel": 11942, "staged_skel": 0, "mean_count": 752.2305908203125, "n_util_obs": 349, "util_count_spearman": -0.00287775064575776}, "probe_overhead_sec_total": 294.562, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.007948, "alpha": 0.03, "t": 2518.9} |
| {"event": "consolidate", "step": 2350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2573.9} |
| {"event": "train", "step": 2350, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.4535, "bpb": 2.09696, "chunk_loss": 0.030897, "F": 0.22565, "G": 0.22866, "chunks": 231.06, "bytes_per_chunk": 4.432, "boundary_entropy": 0.08947, "router_noise": 0.0, "grad_norm": 0.6932, "lr": 0.0002226, "router_trainable": true, "sec_per_step": 1.115, "probe_sec": 0.1334, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10608, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": 0.05775, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": 0.01166, "n_merge": 3, "mean_delta_merge": 0.1462, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 1.14603, "trunc_abs_frac": 0.79701, "mean_delta_full": 0.09672, "read_hit_rate": {"surface": 0.67522, "split": 0.87976, "merge": 0.25853, "skel": 0.66929}, "mean_gate": {"surface": 0.21479, "split": 0.11995, "merge": 0.1847, "skel": 0.16568}, "n_reads": 21568, "n_cand": 32712, "read_suppressed_frac": 0.1002, "dropout_reads_logged": 21568, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2943, "admitted_surface": 16919, "staged_surface": 0, "admitted_split": 16807, "staged_split": 0, "admitted_merge": 19873, "staged_merge": 0, "admitted_skel": 11937, "staged_skel": 0, "mean_count": 768.2377319335938, "n_util_obs": 356, "util_count_spearman": -6.649281013244038e-06}, "probe_overhead_sec_total": 300.878, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.008873, "alpha": 0.03, "t": 2574.6} |
| {"event": "consolidate", "step": 2400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2615.6} |
| {"event": "probe", "step": 2400, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3670.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.210883, "cps": 28.671875, "mean_p": 0.21315, "boundary_entropy": 0.095086, "bpb": 2.052416}, "tr": {"chunks": 4061.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.224327, "cps": 31.726562, "mean_p": 0.210684, "boundary_entropy": 0.106198, "bpb": 2.243329}, "fi": {"chunks": 5054.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.268673, "cps": 39.484375, "mean_p": 0.264302, "boundary_entropy": 0.104329, "bpb": 2.372486}}, "bpb": {"en": 2.05242, "tr": 2.24333, "fi": 2.37249}, "gini_cpc": 0.054734, "gini_cps": 0.072168, "bpic": 4.437388, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.063752255674124, "fi": 1.2740351498782454}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2987, "admitted_surface": 16855, "staged_surface": 0, "admitted_split": 16914, "staged_split": 0, "admitted_merge": 19767, "staged_merge": 0, "admitted_skel": 12000, "staged_skel": 0, "mean_count": 783.927734375, "n_util_obs": 361, "util_count_spearman": 0.011865955262007242}, "t": 2617.5} |
| {"event": "eval", "step": 2400, "phase": 1, "bpb": {"en": 2.06529, "tr": 2.25204, "fi": 2.38607}, "bytes_per_chunk": {"en": 4.442, "tr": 5.015, "fi": 3.972}, "bpb_mean": 2.23447, "frozen_inventory": true, "t": 2620.6} |
| {"event": "phase_transition", "step": 2400, "phase": 2, "objective": "global", "router_trainable": true, "encoder_trainable": true, "frozen_router_params": 0, "t": 2621.9} |
| {"event": "train", "step": 2400, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.45069, "bpb": 2.0929, "chunk_loss": 0.030724, "F": 0.21985, "G": 0.2224, "chunks": 225.12, "bytes_per_chunk": 4.549, "boundary_entropy": 0.09587, "router_noise": 0.0, "grad_norm": 0.6705, "lr": 0.0002192, "router_trainable": true, "sec_per_step": 0.969, "probe_sec": 0.128, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09976, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.00278, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": 0.04468, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 1.13854, "trunc_abs_frac": 1.92022, "mean_delta_full": 0.0462, "read_hit_rate": {"surface": 0.67108, "split": 0.86494, "merge": 0.24734, "skel": 0.6598}, "mean_gate": {"surface": 0.21569, "split": 0.12005, "merge": 0.18308, "skel": 0.16796}, "n_reads": 20695, "n_cand": 31899, "read_suppressed_frac": 0.0978, "dropout_reads_logged": 20695, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2988, "admitted_surface": 16872, "staged_surface": 0, "admitted_split": 16937, "staged_split": 0, "admitted_merge": 19784, "staged_merge": 0, "admitted_skel": 11943, "staged_skel": 0, "mean_count": 784.2330322265625, "n_util_obs": 361, "util_count_spearman": 0.010668519170487615}, "probe_overhead_sec_total": 306.346, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.008151, "alpha": 0.03, "t": 2623.1} |
| {"event": "consolidate", "step": 2450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2664.3} |
| {"event": "train", "step": 2450, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.45633, "bpb": 2.10104, "chunk_loss": 0.031101, "F": 0.23303, "G": 0.23611, "chunks": 238.62, "bytes_per_chunk": 4.291, "boundary_entropy": 0.09754, "router_noise": 0.0, "grad_norm": 0.7143, "lr": 0.0002158, "router_trainable": true, "sec_per_step": 0.848, "probe_sec": 0.1867, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.14241, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.04025, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": 0.03719, "n_merge": 1, "mean_delta_merge": 0.10522, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.52414, "trunc_abs_frac": 1.89566, "mean_delta_full": -0.05701, "read_hit_rate": {"surface": 0.67754, "split": 0.88197, "merge": 0.25079, "skel": 0.6954}, "mean_gate": {"surface": 0.21682, "split": 0.11981, "merge": 0.1884, "skel": 0.16571}, "n_reads": 22290, "n_cand": 33635, "read_suppressed_frac": 0.0984, "dropout_reads_logged": 22290, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3026, "admitted_surface": 16883, "staged_surface": 0, "admitted_split": 16812, "staged_split": 0, "admitted_merge": 20019, "staged_merge": 0, "admitted_skel": 11822, "staged_skel": 0, "mean_count": 800.2171630859375, "n_util_obs": 367, "util_count_spearman": 0.015400535175759043}, "probe_overhead_sec_total": 312.328, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.007992, "alpha": 0.03, "t": 2665.5} |
| {"event": "consolidate", "step": 2500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2706.5} |
| {"event": "train", "step": 2500, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.5592, "bpb": 2.24945, "chunk_loss": 0.030993, "F": 0.23114, "G": 0.23135, "chunks": 236.69, "bytes_per_chunk": 4.326, "boundary_entropy": 0.09154, "router_noise": 0.0, "grad_norm": 0.7329, "lr": 0.0002123, "router_trainable": true, "sec_per_step": 0.834, "probe_sec": 0.0931, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06424, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.05692, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.25957, "n_merge": 1, "mean_delta_merge": -0.12842, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 9.1872, "trunc_abs_frac": 0.76947, "mean_delta_full": 0.01913, "read_hit_rate": {"surface": 0.66468, "split": 0.86746, "merge": 0.24115, "skel": 0.6431}, "mean_gate": {"surface": 0.21224, "split": 0.1169, "merge": 0.18427, "skel": 0.1599}, "n_reads": 21104, "n_cand": 32882, "read_suppressed_frac": 0.0997, "dropout_reads_logged": 21104, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3068, "admitted_surface": 16867, "staged_surface": 0, "admitted_split": 16839, "staged_split": 0, "admitted_merge": 19919, "staged_merge": 0, "admitted_skel": 11911, "staged_skel": 0, "mean_count": 816.3248901367188, "n_util_obs": 377, "util_count_spearman": 0.024075358081227488}, "probe_overhead_sec_total": 317.778, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.008854, "alpha": 0.03, "t": 2707.2} |
| {"event": "consolidate", "step": 2550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2755.6} |
| {"event": "train", "step": 2550, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.48384, "bpb": 2.14073, "chunk_loss": 0.030851, "F": 0.22641, "G": 0.22877, "chunks": 231.84, "bytes_per_chunk": 4.417, "boundary_entropy": 0.08997, "router_noise": 0.0, "grad_norm": 0.6581, "lr": 0.0002088, "router_trainable": true, "sec_per_step": 0.982, "probe_sec": 0.1204, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09168, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.03129, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": -0.00572, "n_merge": 1, "mean_delta_merge": 0.00891, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -1.185, "trunc_abs_frac": 0.45977, "mean_delta_full": -0.10541, "read_hit_rate": {"surface": 0.67172, "split": 0.86951, "merge": 0.24786, "skel": 0.66364}, "mean_gate": {"surface": 0.21008, "split": 0.11468, "merge": 0.18343, "skel": 0.16048}, "n_reads": 21418, "n_cand": 32876, "read_suppressed_frac": 0.0986, "dropout_reads_logged": 21418, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3116, "admitted_surface": 16847, "staged_surface": 0, "admitted_split": 16864, "staged_split": 0, "admitted_merge": 19907, "staged_merge": 0, "admitted_skel": 11918, "staged_skel": 0, "mean_count": 832.412841796875, "n_util_obs": 385, "util_count_spearman": 0.035316011708498754}, "probe_overhead_sec_total": 323.748, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.009834, "alpha": 0.03, "t": 2756.3} |
| {"event": "consolidate", "step": 2600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2808.7} |
| {"event": "probe", "step": 2600, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3688.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.211917, "cps": 28.8125, "mean_p": 0.213295, "boundary_entropy": 0.085672, "bpb": 2.033318}, "tr": {"chunks": 4094.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.22615, "cps": 31.984375, "mean_p": 0.211842, "boundary_entropy": 0.100182, "bpb": 2.228221}, "fi": {"chunks": 5076.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.269842, "cps": 39.65625, "mean_p": 0.264953, "boundary_entropy": 0.091718, "bpb": 2.335181}}, "bpb": {"en": 2.03332, "tr": 2.22822, "fi": 2.33518}, "gini_cpc": 0.05455, "gini_cps": 0.071966, "bpic": 4.412195, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0671623499802227, "fi": 1.2733357657188396}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3146, "admitted_surface": 16862, "staged_surface": 0, "admitted_split": 16804, "staged_split": 0, "admitted_merge": 19949, "staged_merge": 0, "admitted_skel": 11921, "staged_skel": 0, "mean_count": 848.15087890625, "n_util_obs": 392, "util_count_spearman": 0.055246595248451026}, "t": 2810.9} |
| {"event": "eval", "step": 2600, "phase": 2, "bpb": {"en": 2.04502, "tr": 2.23758, "fi": 2.3472}, "bytes_per_chunk": {"en": 4.428, "tr": 4.966, "fi": 3.934}, "bpb_mean": 2.20993, "frozen_inventory": true, "t": 2814.6} |
| {"event": "train", "step": 2600, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.49395, "bpb": 2.15531, "chunk_loss": 0.031006, "F": 0.22928, "G": 0.23149, "chunks": 234.78, "bytes_per_chunk": 4.362, "boundary_entropy": 0.08616, "router_noise": 0.0, "grad_norm": 0.6538, "lr": 0.0002053, "router_trainable": true, "sec_per_step": 1.227, "probe_sec": 0.1151, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05404, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.10028, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": 0.09253, "n_merge": 2, "mean_delta_merge": 0.00702, "n_dropped_unpaired": 0, "n_full": 0, "read_hit_rate": {"surface": 0.66595, "split": 0.87383, "merge": 0.24768, "skel": 0.66852}, "mean_gate": {"surface": 0.20643, "split": 0.11023, "merge": 0.17765, "skel": 0.15728}, "n_reads": 21600, "n_cand": 33148, "read_suppressed_frac": 0.1005, "dropout_reads_logged": 21600, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3147, "admitted_surface": 16894, "staged_surface": 0, "admitted_split": 16800, "staged_split": 0, "admitted_merge": 19946, "staged_merge": 0, "admitted_skel": 11896, "staged_skel": 0, "mean_count": 848.4680786132812, "n_util_obs": 392, "util_count_spearman": 0.05553556193824009}, "probe_overhead_sec_total": 329.625, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.008893, "alpha": 0.03, "t": 2817.6} |
| {"event": "consolidate", "step": 2650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2876.3} |
| {"event": "train", "step": 2650, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.42351, "bpb": 2.0537, "chunk_loss": 0.030758, "F": 0.22214, "G": 0.22288, "chunks": 227.47, "bytes_per_chunk": 4.502, "boundary_entropy": 0.08933, "router_noise": 0.0, "grad_norm": 0.6423, "lr": 0.0002017, "router_trainable": true, "sec_per_step": 1.195, "probe_sec": 0.1194, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05664, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.1484, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": 0.10287, "n_merge": 1, "mean_delta_merge": -0.23566, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -2.89683, "trunc_abs_frac": 1.97627, "mean_delta_full": -0.05591, "read_hit_rate": {"surface": 0.67642, "split": 0.8739, "merge": 0.26265, "skel": 0.65958}, "mean_gate": {"surface": 0.20507, "split": 0.10946, "merge": 0.17663, "skel": 0.15873}, "n_reads": 21032, "n_cand": 32055, "read_suppressed_frac": 0.1, "dropout_reads_logged": 21032, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3191, "admitted_surface": 16877, "staged_surface": 0, "admitted_split": 16859, "staged_split": 0, "admitted_merge": 19841, "staged_merge": 0, "admitted_skel": 11959, "staged_skel": 0, "mean_count": 864.5048217773438, "n_util_obs": 402, "util_count_spearman": 0.07098019510777318}, "probe_overhead_sec_total": 335.655, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.009116, "alpha": 0.03, "t": 2877.4} |
| {"event": "consolidate", "step": 2700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2918.7} |
| {"event": "train", "step": 2700, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.51167, "bpb": 2.18088, "chunk_loss": 0.030872, "F": 0.224, "G": 0.22685, "chunks": 229.38, "bytes_per_chunk": 4.464, "boundary_entropy": 0.09264, "router_noise": 0.0, "grad_norm": 0.7235, "lr": 0.0001981, "router_trainable": true, "sec_per_step": 0.85, "probe_sec": 0.1728, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09266, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.07965, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.10986, "n_merge": 2, "mean_delta_merge": -0.44629, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.09233, "trunc_abs_frac": 0.57638, "mean_delta_full": 0.13525, "read_hit_rate": {"surface": 0.64929, "split": 0.86802, "merge": 0.23502, "skel": 0.64514}, "mean_gate": {"surface": 0.21034, "split": 0.11171, "merge": 0.18309, "skel": 0.16056}, "n_reads": 20764, "n_cand": 32468, "read_suppressed_frac": 0.0993, "dropout_reads_logged": 20764, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3239, "admitted_surface": 16931, "staged_surface": 0, "admitted_split": 16839, "staged_split": 0, "admitted_merge": 19806, "staged_merge": 0, "admitted_skel": 11960, "staged_skel": 0, "mean_count": 880.5034790039062, "n_util_obs": 406, "util_count_spearman": 0.08530062398274924}, "probe_overhead_sec_total": 341.396, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.008799, "alpha": 0.03, "t": 2919.8} |
| {"event": "consolidate", "step": 2750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2960.1} |
| {"event": "train", "step": 2750, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.44646, "bpb": 2.08681, "chunk_loss": 0.030857, "F": 0.22684, "G": 0.22834, "chunks": 232.28, "bytes_per_chunk": 4.408, "boundary_entropy": 0.08818, "router_noise": 0.0, "grad_norm": 0.6901, "lr": 0.0001945, "router_trainable": true, "sec_per_step": 0.818, "probe_sec": 0.1551, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.12955, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.07168, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": -0.1765, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -2.82605, "trunc_abs_frac": 14.3368, "mean_delta_full": -0.07684, "read_hit_rate": {"surface": 0.66559, "split": 0.87057, "merge": 0.24997, "skel": 0.64695}, "mean_gate": {"surface": 0.20754, "split": 0.10941, "merge": 0.18227, "skel": 0.15916}, "n_reads": 21164, "n_cand": 32696, "read_suppressed_frac": 0.1006, "dropout_reads_logged": 21164, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3278, "admitted_surface": 16864, "staged_surface": 0, "admitted_split": 16823, "staged_split": 0, "admitted_merge": 19914, "staged_merge": 0, "admitted_skel": 11935, "staged_skel": 0, "mean_count": 896.6510009765625, "n_util_obs": 412, "util_count_spearman": 0.08111534924307977}, "probe_overhead_sec_total": 347.331, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.009657, "alpha": 0.03, "t": 2960.8} |
| {"event": "consolidate", "step": 2800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3006.6} |
| {"event": "probe", "step": 2800, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3649.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.209676, "cps": 28.507812, "mean_p": 0.210779, "boundary_entropy": 0.089461, "bpb": 2.013224}, "tr": {"chunks": 4089.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.225874, "cps": 31.945312, "mean_p": 0.21095, "boundary_entropy": 0.110004, "bpb": 2.20796}, "fi": {"chunks": 5059.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.268938, "cps": 39.523438, "mean_p": 0.26317, "boundary_entropy": 0.092066, "bpb": 2.30996}}, "bpb": {"en": 2.01322, "tr": 2.20796, "fi": 2.30996}, "gini_cpc": 0.05608, "gini_cps": 0.073455, "bpic": 4.433227, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0772507768834791, "fi": 1.2826349002311859}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3319, "admitted_surface": 16852, "staged_surface": 0, "admitted_split": 16846, "staged_split": 0, "admitted_merge": 19939, "staged_merge": 0, "admitted_skel": 11899, "staged_skel": 0, "mean_count": 912.399169921875, "n_util_obs": 416, "util_count_spearman": 0.076042465378948}, "t": 3008.8} |
| {"event": "eval", "step": 2800, "phase": 2, "bpb": {"en": 2.02635, "tr": 2.21725, "fi": 2.3221}, "bytes_per_chunk": {"en": 4.441, "tr": 4.963, "fi": 3.982}, "bpb_mean": 2.18857, "frozen_inventory": true, "t": 3012.6} |
| {"event": "train", "step": 2800, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.43975, "bpb": 2.07712, "chunk_loss": 0.030894, "F": 0.22891, "G": 0.22844, "chunks": 234.41, "bytes_per_chunk": 4.368, "boundary_entropy": 0.08803, "router_noise": 0.0, "grad_norm": 0.7122, "lr": 0.0001909, "router_trainable": true, "sec_per_step": 1.085, "probe_sec": 0.1054, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05685, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.00701, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.05963, "n_merge": 1, "mean_delta_merge": 0.49806, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 1.80191, "trunc_abs_frac": 0.44199, "mean_delta_full": 0.19385, "read_hit_rate": {"surface": 0.68215, "split": 0.87704, "merge": 0.26435, "skel": 0.66542}, "mean_gate": {"surface": 0.20409, "split": 0.10596, "merge": 0.17789, "skel": 0.15542}, "n_reads": 21682, "n_cand": 32867, "read_suppressed_frac": 0.0993, "dropout_reads_logged": 21682, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3320, "admitted_surface": 16874, "staged_surface": 0, "admitted_split": 16825, "staged_split": 0, "admitted_merge": 19942, "staged_merge": 0, "admitted_skel": 11895, "staged_skel": 0, "mean_count": 912.73193359375, "n_util_obs": 416, "util_count_spearman": 0.07543188803915163}, "probe_overhead_sec_total": 353.483, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.009149, "alpha": 0.03, "t": 3015.0} |
| {"event": "consolidate", "step": 2850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3065.2} |
| {"event": "train", "step": 2850, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.53677, "bpb": 2.21709, "chunk_loss": 0.031049, "F": 0.22986, "G": 0.23176, "chunks": 235.38, "bytes_per_chunk": 4.351, "boundary_entropy": 0.0847, "router_noise": 0.0, "grad_norm": 0.6888, "lr": 0.0001872, "router_trainable": true, "sec_per_step": 1.027, "probe_sec": 0.0839, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0554, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": 0.00287, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.31439, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.07368, "trunc_abs_frac": 2.57982, "mean_delta_full": 0.22449, "read_hit_rate": {"surface": 0.66067, "split": 0.86794, "merge": 0.23005, "skel": 0.65907}, "mean_gate": {"surface": 0.20652, "split": 0.10644, "merge": 0.18068, "skel": 0.15561}, "n_reads": 21380, "n_cand": 33244, "read_suppressed_frac": 0.1023, "dropout_reads_logged": 21380, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3364, "admitted_surface": 16895, "staged_surface": 0, "admitted_split": 16856, "staged_split": 0, "admitted_merge": 19896, "staged_merge": 0, "admitted_skel": 11889, "staged_skel": 0, "mean_count": 928.78662109375, "n_util_obs": 427, "util_count_spearman": 0.08054763389591417}, "probe_overhead_sec_total": 359.273, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.009194, "alpha": 0.03, "t": 3066.3} |
| {"event": "consolidate", "step": 2900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3123.6} |
| {"event": "train", "step": 2900, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.48102, "bpb": 2.13666, "chunk_loss": 0.031084, "F": 0.23361, "G": 0.23385, "chunks": 239.22, "bytes_per_chunk": 4.281, "boundary_entropy": 0.09071, "router_noise": 0.0, "grad_norm": 0.6419, "lr": 0.0001835, "router_trainable": true, "sec_per_step": 1.17, "probe_sec": 0.2072, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07806, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": 0.1567, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.80676, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -0.31714, "trunc_abs_frac": 7.73083, "mean_delta_full": -0.1467, "read_hit_rate": {"surface": 0.67887, "split": 0.87079, "merge": 0.25122, "skel": 0.66427}, "mean_gate": {"surface": 0.20565, "split": 0.10856, "merge": 0.1728, "skel": 0.15825}, "n_reads": 21791, "n_cand": 33392, "read_suppressed_frac": 0.1004, "dropout_reads_logged": 21791, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3414, "admitted_surface": 16853, "staged_surface": 0, "admitted_split": 16881, "staged_split": 0, "admitted_merge": 19959, "staged_merge": 0, "admitted_skel": 11843, "staged_skel": 0, "mean_count": 944.891357421875, "n_util_obs": 437, "util_count_spearman": 0.05804540574227084}, "probe_overhead_sec_total": 366.385, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.010411, "alpha": 0.03, "t": 3124.8} |
| {"event": "consolidate", "step": 2950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3187.2} |
| {"event": "train", "step": 2950, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.42264, "bpb": 2.05243, "chunk_loss": 0.03079, "F": 0.22198, "G": 0.22338, "chunks": 227.31, "bytes_per_chunk": 4.505, "boundary_entropy": 0.08274, "router_noise": 0.0, "grad_norm": 0.6345, "lr": 0.0001799, "router_trainable": true, "sec_per_step": 1.291, "probe_sec": 0.2162, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.15729, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.02477, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": 0.09431, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 3.30887, "trunc_abs_frac": 0.46059, "mean_delta_full": 0.01788, "read_hit_rate": {"surface": 0.66404, "split": 0.86334, "merge": 0.24355, "skel": 0.64915}, "mean_gate": {"surface": 0.21126, "split": 0.10805, "merge": 0.18099, "skel": 0.1627}, "n_reads": 20782, "n_cand": 32251, "read_suppressed_frac": 0.0976, "dropout_reads_logged": 20782, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3458, "admitted_surface": 16895, "staged_surface": 0, "admitted_split": 16948, "staged_split": 0, "admitted_merge": 19737, "staged_merge": 0, "admitted_skel": 11956, "staged_skel": 0, "mean_count": 960.9425659179688, "n_util_obs": 449, "util_count_spearman": 0.07018798620572128}, "probe_overhead_sec_total": 372.654, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.008606, "alpha": 0.03, "t": 3189.4} |
| {"event": "consolidate", "step": 3000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3231.4} |
| {"event": "probe", "step": 3000, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3673.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.211056, "cps": 28.695312, "mean_p": 0.212838, "boundary_entropy": 0.082337, "bpb": 1.991042}, "tr": {"chunks": 4079.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.225322, "cps": 31.867188, "mean_p": 0.206932, "boundary_entropy": 0.101219, "bpb": 2.167652}, "fi": {"chunks": 5058.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.268885, "cps": 39.515625, "mean_p": 0.263005, "boundary_entropy": 0.082822, "bpb": 2.286879}}, "bpb": {"en": 1.99104, "tr": 2.16765, "fi": 2.28688}, "gini_cpc": 0.054665, "gini_cps": 0.072079, "bpic": 4.428728, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0675945442660828, "fi": 1.274002069361696}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3496, "admitted_surface": 16888, "staged_surface": 0, "admitted_split": 16892, "staged_split": 0, "admitted_merge": 19853, "staged_merge": 0, "admitted_skel": 11903, "staged_skel": 0, "mean_count": 976.6557006835938, "n_util_obs": 452, "util_count_spearman": 0.050700852572824875}, "t": 3233.2} |
| {"event": "eval", "step": 3000, "phase": 2, "bpb": {"en": 2.00454, "tr": 2.17749, "fi": 2.29889}, "bytes_per_chunk": {"en": 4.428, "tr": 4.973, "fi": 3.947}, "bpb_mean": 2.16031, "frozen_inventory": true, "t": 3236.2} |
| {"event": "train", "step": 3000, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.44901, "bpb": 2.09048, "chunk_loss": 0.030792, "F": 0.22461, "G": 0.22555, "chunks": 230.0, "bytes_per_chunk": 4.452, "boundary_entropy": 0.08175, "router_noise": 0.0, "grad_norm": 0.6596, "lr": 0.0001761, "router_trainable": true, "sec_per_step": 0.986, "probe_sec": 0.094, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06524, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.07736, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.08203, "n_merge": 1, "mean_delta_merge": -0.66913, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.34375, "trunc_abs_frac": 1.17185, "mean_delta_full": -0.14876, "read_hit_rate": {"surface": 0.66526, "split": 0.87207, "merge": 0.24931, "skel": 0.65949}, "mean_gate": {"surface": 0.20058, "split": 0.1022, "merge": 0.17296, "skel": 0.15227}, "n_reads": 21157, "n_cand": 32533, "read_suppressed_frac": 0.1004, "dropout_reads_logged": 21157, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3496, "admitted_surface": 16913, "staged_surface": 0, "admitted_split": 16896, "staged_split": 0, "admitted_merge": 19785, "staged_merge": 0, "admitted_skel": 11942, "staged_skel": 0, "mean_count": 976.9801025390625, "n_util_obs": 452, "util_count_spearman": 0.0494733713105049}, "probe_overhead_sec_total": 378.373, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.010113, "alpha": 0.03, "t": 3238.7} |
| {"event": "consolidate", "step": 3050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3278.8} |
| {"event": "train", "step": 3050, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.43606, "bpb": 2.0718, "chunk_loss": 0.030888, "F": 0.22839, "G": 0.22883, "chunks": 233.88, "bytes_per_chunk": 4.378, "boundary_entropy": 0.08513, "router_noise": 0.0, "grad_norm": 0.6816, "lr": 0.0001724, "router_trainable": true, "sec_per_step": 0.816, "probe_sec": 0.1298, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10661, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": 0.00037, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.25684, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": -0.08245, "trunc_abs_frac": 0.08245, "mean_delta_full": 0.56189, "read_hit_rate": {"surface": 0.6821, "split": 0.87387, "merge": 0.25849, "skel": 0.66575}, "mean_gate": {"surface": 0.20772, "split": 0.10468, "merge": 0.17934, "skel": 0.16017}, "n_reads": 21613, "n_cand": 32892, "read_suppressed_frac": 0.1005, "dropout_reads_logged": 21613, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3527, "admitted_surface": 16858, "staged_surface": 0, "admitted_split": 16905, "staged_split": 0, "admitted_merge": 19859, "staged_merge": 0, "admitted_skel": 11914, "staged_skel": 0, "mean_count": 993.0816650390625, "n_util_obs": 459, "util_count_spearman": 0.0515902475513618}, "probe_overhead_sec_total": 383.888, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.007881, "alpha": 0.03, "t": 3279.5} |
| {"event": "consolidate", "step": 3100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3319.6} |
| {"event": "train", "step": 3100, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.52178, "bpb": 2.19546, "chunk_loss": 0.030833, "F": 0.22305, "G": 0.22268, "chunks": 228.41, "bytes_per_chunk": 4.483, "boundary_entropy": 0.08952, "router_noise": 0.0, "grad_norm": 0.73, "lr": 0.0001687, "router_trainable": true, "sec_per_step": 0.825, "probe_sec": 0.0879, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05862, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.26129, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": -0.05814, "n_merge": 2, "mean_delta_merge": 0.03613, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.08126, "trunc_abs_frac": 0.08126, "mean_delta_full": -0.14722, "read_hit_rate": {"surface": 0.65425, "split": 0.86161, "merge": 0.23257, "skel": 0.64644}, "mean_gate": {"surface": 0.20492, "split": 0.10379, "merge": 0.17952, "skel": 0.15574}, "n_reads": 20460, "n_cand": 32112, "read_suppressed_frac": 0.1002, "dropout_reads_logged": 20460, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3557, "admitted_surface": 16894, "staged_surface": 0, "admitted_split": 16947, "staged_split": 0, "admitted_merge": 19780, "staged_merge": 0, "admitted_skel": 11915, "staged_skel": 0, "mean_count": 1009.0626831054688, "n_util_obs": 468, "util_count_spearman": 0.030061356103373658}, "probe_overhead_sec_total": 389.406, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.009691, "alpha": 0.03, "t": 3320.7} |
| {"event": "consolidate", "step": 3150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3369.9} |
| {"event": "train", "step": 3150, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.43938, "bpb": 2.07659, "chunk_loss": 0.030836, "F": 0.22394, "G": 0.22495, "chunks": 229.31, "bytes_per_chunk": 4.466, "boundary_entropy": 0.08099, "router_noise": 0.0, "grad_norm": 0.6257, "lr": 0.000165, "router_trainable": true, "sec_per_step": 1.011, "probe_sec": 0.1147, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08149, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.03726, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": -0.21395, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -0.52378, "trunc_abs_frac": 0.33765, "mean_delta_full": 0.05241, "read_hit_rate": {"surface": 0.66083, "split": 0.86715, "merge": 0.24649, "skel": 0.65092}, "mean_gate": {"surface": 0.20179, "split": 0.10218, "merge": 0.17684, "skel": 0.15605}, "n_reads": 20904, "n_cand": 32397, "read_suppressed_frac": 0.0985, "dropout_reads_logged": 20904, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3604, "admitted_surface": 16873, "staged_surface": 0, "admitted_split": 16937, "staged_split": 0, "admitted_merge": 19784, "staged_merge": 0, "admitted_skel": 11942, "staged_skel": 0, "mean_count": 1025.1658935546875, "n_util_obs": 472, "util_count_spearman": 0.034441396458071954}, "probe_overhead_sec_total": 395.289, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.00888, "alpha": 0.03, "t": 3371.2} |
| {"event": "consolidate", "step": 3200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3434.8} |
| {"event": "probe", "step": 3200, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3672.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.210998, "cps": 28.6875, "mean_p": 0.211188, "boundary_entropy": 0.077722, "bpb": 1.971663}, "tr": {"chunks": 4096.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.226261, "cps": 32.0, "mean_p": 0.205794, "boundary_entropy": 0.104633, "bpb": 2.149867}, "fi": {"chunks": 5047.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.2683, "cps": 39.429688, "mean_p": 0.26239, "boundary_entropy": 0.084736, "bpb": 2.260693}}, "bpb": {"en": 1.97166, "tr": 2.14987, "fi": 2.26069}, "gini_cpc": 0.054144, "gini_cps": 0.071531, "bpic": 4.427, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0723358963722114, "fi": 1.2715776004375134}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3649, "admitted_surface": 16851, "staged_surface": 0, "admitted_split": 16953, "staged_split": 0, "admitted_merge": 19828, "staged_merge": 0, "admitted_skel": 11904, "staged_skel": 0, "mean_count": 1040.88330078125, "n_util_obs": 478, "util_count_spearman": 0.02125847803477757}, "t": 3436.9} |
| {"event": "eval", "step": 3200, "phase": 2, "bpb": {"en": 1.98633, "tr": 2.15966, "fi": 2.27172}, "bytes_per_chunk": {"en": 4.435, "tr": 4.982, "fi": 3.964}, "bpb_mean": 2.13924, "frozen_inventory": true, "t": 3440.2} |
| {"event": "train", "step": 3200, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.41399, "bpb": 2.03996, "chunk_loss": 0.031019, "F": 0.23126, "G": 0.23044, "chunks": 236.81, "bytes_per_chunk": 4.324, "boundary_entropy": 0.07993, "router_noise": 0.0, "grad_norm": 0.6397, "lr": 0.0001613, "router_trainable": true, "sec_per_step": 1.428, "probe_sec": 0.1106, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06543, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.03433, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.14746, "n_merge": 1, "mean_delta_merge": 0.46225, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": -0.214, "trunc_abs_frac": 0.955, "mean_delta_full": 0.1427, "read_hit_rate": {"surface": 0.68195, "split": 0.87888, "merge": 0.25233, "skel": 0.6808}, "mean_gate": {"surface": 0.19863, "split": 0.09896, "merge": 0.17516, "skel": 0.15034}, "n_reads": 22125, "n_cand": 33465, "read_suppressed_frac": 0.0979, "dropout_reads_logged": 22125, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3650, "admitted_surface": 16852, "staged_surface": 0, "admitted_split": 16869, "staged_split": 0, "admitted_merge": 19938, "staged_merge": 0, "admitted_skel": 11877, "staged_skel": 0, "mean_count": 1041.212646484375, "n_util_obs": 478, "util_count_spearman": 0.02100274421420139}, "probe_overhead_sec_total": 401.548, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.00976, "alpha": 0.03, "t": 3442.6} |
| {"event": "consolidate", "step": 3250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3490.3} |
| {"event": "train", "step": 3250, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.39257, "bpb": 2.00906, "chunk_loss": 0.030704, "F": 0.22089, "G": 0.22112, "chunks": 226.19, "bytes_per_chunk": 4.527, "boundary_entropy": 0.08102, "router_noise": 0.0, "grad_norm": 0.6434, "lr": 0.0001576, "router_trainable": true, "sec_per_step": 0.968, "probe_sec": 0.0945, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06452, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.07073, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": 0.04379, "n_merge": 1, "mean_delta_merge": -0.00903, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 2.08528, "trunc_abs_frac": 0.58445, "mean_delta_full": 0.03142, "read_hit_rate": {"surface": 0.67569, "split": 0.87151, "merge": 0.2477, "skel": 0.66099}, "mean_gate": {"surface": 0.19761, "split": 0.09831, "merge": 0.1681, "skel": 0.15121}, "n_reads": 20875, "n_cand": 31983, "read_suppressed_frac": 0.1007, "dropout_reads_logged": 20875, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3687, "admitted_surface": 16847, "staged_surface": 0, "admitted_split": 16945, "staged_split": 0, "admitted_merge": 19811, "staged_merge": 0, "admitted_skel": 11933, "staged_skel": 0, "mean_count": 1057.3201904296875, "n_util_obs": 484, "util_count_spearman": 0.030414302701495564}, "probe_overhead_sec_total": 407.56, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.009671, "alpha": 0.03, "t": 3491.0} |
| {"event": "consolidate", "step": 3300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3530.0} |
| {"event": "train", "step": 3300, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.43078, "bpb": 2.06418, "chunk_loss": 0.030922, "F": 0.22882, "G": 0.22811, "chunks": 234.31, "bytes_per_chunk": 4.37, "boundary_entropy": 0.08172, "router_noise": 0.0, "grad_norm": 0.6517, "lr": 0.0001539, "router_trainable": true, "sec_per_step": 0.793, "probe_sec": 0.1326, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10642, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.08955, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.08337, "n_merge": 1, "mean_delta_merge": 0.02032, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -0.16757, "trunc_abs_frac": 0.59537, "mean_delta_full": -0.17179, "read_hit_rate": {"surface": 0.67305, "split": 0.87633, "merge": 0.24697, "skel": 0.67516}, "mean_gate": {"surface": 0.193, "split": 0.09582, "merge": 0.16813, "skel": 0.1479}, "n_reads": 21786, "n_cand": 33211, "read_suppressed_frac": 0.097, "dropout_reads_logged": 21786, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3728, "admitted_surface": 16859, "staged_surface": 0, "admitted_split": 16861, "staged_split": 0, "admitted_merge": 19903, "staged_merge": 0, "admitted_skel": 11913, "staged_skel": 0, "mean_count": 1073.39697265625, "n_util_obs": 493, "util_count_spearman": 0.03826650119212426}, "probe_overhead_sec_total": 413.225, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.008572, "alpha": 0.03, "t": 3530.7} |
| {"event": "consolidate", "step": 3350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3569.0} |
| {"event": "train", "step": 3350, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.4599, "bpb": 2.10619, "chunk_loss": 0.031098, "F": 0.23541, "G": 0.23403, "chunks": 241.06, "bytes_per_chunk": 4.248, "boundary_entropy": 0.08326, "router_noise": 0.0, "grad_norm": 0.7202, "lr": 0.0001501, "router_trainable": true, "sec_per_step": 0.788, "probe_sec": 0.0876, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05912, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.2814, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.10748, "n_merge": 1, "mean_delta_merge": -0.25549, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.34598, "trunc_abs_frac": 0.19416, "mean_delta_full": -0.24015, "read_hit_rate": {"surface": 0.68459, "split": 0.86836, "merge": 0.26876, "skel": 0.66879}, "mean_gate": {"surface": 0.19886, "split": 0.09998, "merge": 0.17311, "skel": 0.15217}, "n_reads": 22116, "n_cand": 33641, "read_suppressed_frac": 0.097, "dropout_reads_logged": 22116, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3781, "admitted_surface": 16825, "staged_surface": 0, "admitted_split": 16974, "staged_split": 0, "admitted_merge": 19822, "staged_merge": 0, "admitted_skel": 11915, "staged_skel": 0, "mean_count": 1089.4571533203125, "n_util_obs": 500, "util_count_spearman": 0.05115126183267763}, "probe_overhead_sec_total": 418.179, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008474, "alpha": 0.03, "t": 3570.1} |
| {"event": "consolidate", "step": 3400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3616.6} |
| {"event": "probe", "step": 3400, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3665.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.210596, "cps": 28.632812, "mean_p": 0.211018, "boundary_entropy": 0.076139, "bpb": 1.955303}, "tr": {"chunks": 4087.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.225764, "cps": 31.929688, "mean_p": 0.205006, "boundary_entropy": 0.100684, "bpb": 2.121958}, "fi": {"chunks": 5052.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.268566, "cps": 39.46875, "mean_p": 0.260833, "boundary_entropy": 0.080759, "bpb": 2.24119}}, "bpb": {"en": 1.9553, "tr": 2.12196, "fi": 2.24119}, "gini_cpc": 0.054824, "gini_cps": 0.072217, "bpic": 4.430803, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0720233069839336, "fi": 1.2752684037372404}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3812, "admitted_surface": 16827, "staged_surface": 0, "admitted_split": 16921, "staged_split": 0, "admitted_merge": 19864, "staged_merge": 0, "admitted_skel": 11924, "staged_skel": 0, "mean_count": 1105.1749267578125, "n_util_obs": 506, "util_count_spearman": 0.059382553985892525}, "t": 3618.6} |
| {"event": "eval", "step": 3400, "phase": 2, "bpb": {"en": 1.96777, "tr": 2.13295, "fi": 2.25184}, "bytes_per_chunk": {"en": 4.437, "tr": 4.979, "fi": 3.987}, "bpb_mean": 2.11752, "frozen_inventory": true, "t": 3622.0} |
| {"event": "train", "step": 3400, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.37677, "bpb": 1.98626, "chunk_loss": 0.03075, "F": 0.22269, "G": 0.22163, "chunks": 228.03, "bytes_per_chunk": 4.491, "boundary_entropy": 0.07731, "router_noise": 0.0, "grad_norm": 0.6561, "lr": 0.0001465, "router_trainable": true, "sec_per_step": 1.082, "probe_sec": 0.0674, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04394, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": 0.17528, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.01193, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.73016, "trunc_abs_frac": 0.73016, "mean_delta_full": 1.35657, "read_hit_rate": {"surface": 0.68396, "split": 0.87012, "merge": 0.2584, "skel": 0.67184}, "mean_gate": {"surface": 0.19671, "split": 0.09821, "merge": 0.17104, "skel": 0.154}, "n_reads": 21175, "n_cand": 32193, "read_suppressed_frac": 0.098, "dropout_reads_logged": 21175, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3812, "admitted_surface": 16843, "staged_surface": 0, "admitted_split": 16981, "staged_split": 0, "admitted_merge": 19802, "staged_merge": 0, "admitted_skel": 11910, "staged_skel": 0, "mean_count": 1105.5009765625, "n_util_obs": 506, "util_count_spearman": 0.0590686455617978}, "probe_overhead_sec_total": 424.019, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009374, "alpha": 0.03, "t": 3624.2} |
| {"event": "consolidate", "step": 3450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3678.4} |
| {"event": "train", "step": 3450, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.37274, "bpb": 1.98045, "chunk_loss": 0.03065, "F": 0.21927, "G": 0.21851, "chunks": 224.53, "bytes_per_chunk": 4.561, "boundary_entropy": 0.08004, "router_noise": 0.0, "grad_norm": 0.6438, "lr": 0.0001428, "router_trainable": true, "sec_per_step": 1.111, "probe_sec": 0.1196, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07963, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.11898, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 4, "mean_delta_merge": 0.01967, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.73409, "trunc_abs_frac": 0.74335, "mean_delta_full": 0.2323, "read_hit_rate": {"surface": 0.67622, "split": 0.86559, "merge": 0.25839, "skel": 0.66183}, "mean_gate": {"surface": 0.18572, "split": 0.09048, "merge": 0.16199, "skel": 0.14153}, "n_reads": 20721, "n_cand": 31749, "read_suppressed_frac": 0.0978, "dropout_reads_logged": 20721, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3848, "admitted_surface": 16850, "staged_surface": 0, "admitted_split": 17033, "staged_split": 0, "admitted_merge": 19722, "staged_merge": 0, "admitted_skel": 11931, "staged_skel": 0, "mean_count": 1121.614501953125, "n_util_obs": 512, "util_count_spearman": 0.050998646149774736}, "probe_overhead_sec_total": 430.012, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008178, "alpha": 0.03, "t": 3679.7} |
| {"event": "consolidate", "step": 3500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3742.8} |
| {"event": "train", "step": 3500, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.4569, "bpb": 2.10187, "chunk_loss": 0.030945, "F": 0.22696, "G": 0.22597, "chunks": 232.41, "bytes_per_chunk": 4.406, "boundary_entropy": 0.08039, "router_noise": 0.0, "grad_norm": 0.6503, "lr": 0.0001391, "router_trainable": true, "sec_per_step": 1.291, "probe_sec": 0.1021, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07043, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.06231, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 3, "mean_delta_merge": -0.19129, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.00832, "trunc_abs_frac": 0.48191, "mean_delta_full": -0.18372, "read_hit_rate": {"surface": 0.65982, "split": 0.86762, "merge": 0.23925, "skel": 0.65647}, "mean_gate": {"surface": 0.18924, "split": 0.09192, "merge": 0.16243, "skel": 0.14279}, "n_reads": 21127, "n_cand": 32800, "read_suppressed_frac": 0.102, "dropout_reads_logged": 21127, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3878, "admitted_surface": 16876, "staged_surface": 0, "admitted_split": 16903, "staged_split": 0, "admitted_merge": 19826, "staged_merge": 0, "admitted_skel": 11931, "staged_skel": 0, "mean_count": 1137.697509765625, "n_util_obs": 519, "util_count_spearman": 0.04850478480579067}, "probe_overhead_sec_total": 436.443, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.010069, "alpha": 0.03, "t": 3744.3} |
| {"event": "consolidate", "step": 3550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3787.4} |
| {"event": "train", "step": 3550, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.46021, "bpb": 2.10664, "chunk_loss": 0.031111, "F": 0.23422, "G": 0.23326, "chunks": 239.84, "bytes_per_chunk": 4.269, "boundary_entropy": 0.07888, "router_noise": 0.0, "grad_norm": 0.6713, "lr": 0.0001355, "router_trainable": true, "sec_per_step": 0.875, "probe_sec": 0.1145, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09122, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.02936, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": -0.15335, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -0.32367, "trunc_abs_frac": 0.16467, "mean_delta_full": 0.19666, "read_hit_rate": {"surface": 0.66662, "split": 0.87041, "merge": 0.23624, "skel": 0.67381}, "mean_gate": {"surface": 0.18922, "split": 0.08996, "merge": 0.16704, "skel": 0.14081}, "n_reads": 22062, "n_cand": 33968, "read_suppressed_frac": 0.1001, "dropout_reads_logged": 22062, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3918, "admitted_surface": 16850, "staged_surface": 0, "admitted_split": 16867, "staged_split": 0, "admitted_merge": 19968, "staged_merge": 0, "admitted_skel": 11851, "staged_skel": 0, "mean_count": 1153.8125, "n_util_obs": 526, "util_count_spearman": 0.046127312877656534}, "probe_overhead_sec_total": 442.402, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009659, "alpha": 0.03, "t": 3788.1} |
| {"event": "consolidate", "step": 3600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3826.1} |
| {"event": "probe", "step": 3600, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3656.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.210079, "cps": 28.5625, "mean_p": 0.209773, "boundary_entropy": 0.076087, "bpb": 1.935559}, "tr": {"chunks": 4107.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.226868, "cps": 32.085938, "mean_p": 0.205709, "boundary_entropy": 0.099013, "bpb": 2.098875}, "fi": {"chunks": 5067.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.269364, "cps": 39.585938, "mean_p": 0.261904, "boundary_entropy": 0.079334, "bpb": 2.216311}}, "bpb": {"en": 1.93556, "tr": 2.09888, "fi": 2.21631}, "gini_cpc": 0.055957, "gini_cps": 0.073318, "bpic": 4.421824, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0799212438766692, "fi": 1.2822034880657263}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3948, "admitted_surface": 16891, "staged_surface": 0, "admitted_split": 16921, "staged_split": 0, "admitted_merge": 19769, "staged_merge": 0, "admitted_skel": 11955, "staged_skel": 0, "mean_count": 1169.49267578125, "n_util_obs": 529, "util_count_spearman": 0.047997820088213024}, "t": 3828.1} |
| {"event": "eval", "step": 3600, "phase": 2, "bpb": {"en": 1.95089, "tr": 2.10839, "fi": 2.22789}, "bytes_per_chunk": {"en": 4.451, "tr": 4.972, "fi": 3.984}, "bpb_mean": 2.09573, "frozen_inventory": true, "t": 3831.3} |
| {"event": "train", "step": 3600, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.43538, "bpb": 2.07082, "chunk_loss": 0.030814, "F": 0.22592, "G": 0.22454, "chunks": 231.34, "bytes_per_chunk": 4.426, "boundary_entropy": 0.07699, "router_noise": 0.0, "grad_norm": 0.7028, "lr": 0.0001319, "router_trainable": true, "sec_per_step": 0.902, "probe_sec": 0.0848, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06003, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.00299, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.15038, "n_merge": 1, "mean_delta_merge": -0.09979, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -0.20333, "trunc_abs_frac": 9.1729, "mean_delta_full": -0.06323, "read_hit_rate": {"surface": 0.65744, "split": 0.86635, "merge": 0.23682, "skel": 0.65013}, "mean_gate": {"surface": 0.1914, "split": 0.09226, "merge": 0.16812, "skel": 0.14267}, "n_reads": 20662, "n_cand": 32258, "read_suppressed_frac": 0.1006, "dropout_reads_logged": 20662, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3949, "admitted_surface": 16893, "staged_surface": 0, "admitted_split": 16899, "staged_split": 0, "admitted_merge": 19878, "staged_merge": 0, "admitted_skel": 11866, "staged_skel": 0, "mean_count": 1169.8077392578125, "n_util_obs": 529, "util_count_spearman": 0.047882874568579425}, "probe_overhead_sec_total": 447.868, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008694, "alpha": 0.03, "t": 3833.1} |
| {"event": "consolidate", "step": 3650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3871.8} |
| {"event": "train", "step": 3650, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.45485, "bpb": 2.0989, "chunk_loss": 0.030954, "F": 0.22842, "G": 0.22773, "chunks": 233.91, "bytes_per_chunk": 4.378, "boundary_entropy": 0.07841, "router_noise": 0.0, "grad_norm": 0.6902, "lr": 0.0001283, "router_trainable": true, "sec_per_step": 0.792, "probe_sec": 0.0887, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06078, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.04337, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": -0.10568, "n_merge": 1, "mean_delta_merge": -0.05911, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 2.58405, "trunc_abs_frac": 0.14777, "mean_delta_full": -0.02124, "read_hit_rate": {"surface": 0.66456, "split": 0.867, "merge": 0.24471, "skel": 0.65998}, "mean_gate": {"surface": 0.19664, "split": 0.09574, "merge": 0.16876, "skel": 0.15154}, "n_reads": 21358, "n_cand": 32998, "read_suppressed_frac": 0.1004, "dropout_reads_logged": 21358, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3978, "admitted_surface": 16860, "staged_surface": 0, "admitted_split": 16952, "staged_split": 0, "admitted_merge": 19842, "staged_merge": 0, "admitted_skel": 11882, "staged_skel": 0, "mean_count": 1185.8212890625, "n_util_obs": 538, "util_count_spearman": 0.030163807515226914}, "probe_overhead_sec_total": 453.334, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009657, "alpha": 0.03, "t": 3872.8} |
| {"event": "consolidate", "step": 3700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3920.9} |
| {"event": "train", "step": 3700, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.35573, "bpb": 1.9559, "chunk_loss": 0.030691, "F": 0.2207, "G": 0.21872, "chunks": 226.0, "bytes_per_chunk": 4.531, "boundary_entropy": 0.07894, "router_noise": 0.0, "grad_norm": 0.6474, "lr": 0.0001247, "router_trainable": true, "sec_per_step": 0.984, "probe_sec": 0.0788, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05422, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.05996, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 4, "mean_delta_split": 0.04127, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.71077, "trunc_abs_frac": 1.32548, "mean_delta_full": -0.11163, "read_hit_rate": {"surface": 0.66903, "split": 0.86947, "merge": 0.25014, "skel": 0.65797}, "mean_gate": {"surface": 0.19261, "split": 0.0935, "merge": 0.16862, "skel": 0.14786}, "n_reads": 20776, "n_cand": 31948, "read_suppressed_frac": 0.0958, "dropout_reads_logged": 20776, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4020, "admitted_surface": 16851, "staged_surface": 0, "admitted_split": 16964, "staged_split": 0, "admitted_merge": 19811, "staged_merge": 0, "admitted_skel": 11910, "staged_skel": 0, "mean_count": 1201.9163818359375, "n_util_obs": 547, "util_count_spearman": 0.020247641355315586}, "probe_overhead_sec_total": 459.323, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009951, "alpha": 0.03, "t": 3922.0} |
| {"event": "consolidate", "step": 3750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3980.3} |
| {"event": "train", "step": 3750, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.38098, "bpb": 1.99233, "chunk_loss": 0.031044, "F": 0.23373, "G": 0.23249, "chunks": 239.34, "bytes_per_chunk": 4.278, "boundary_entropy": 0.07254, "router_noise": 0.0, "grad_norm": 0.6352, "lr": 0.0001212, "router_trainable": true, "sec_per_step": 1.188, "probe_sec": 0.1332, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06734, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.03289, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": -0.03371, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.39568, "trunc_abs_frac": 1.01104, "mean_delta_full": -0.13131, "read_hit_rate": {"surface": 0.69621, "split": 0.88039, "merge": 0.25635, "skel": 0.69633}, "mean_gate": {"surface": 0.18452, "split": 0.08685, "merge": 0.15842, "skel": 0.13895}, "n_reads": 22480, "n_cand": 33674, "read_suppressed_frac": 0.0998, "dropout_reads_logged": 22480, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4068, "admitted_surface": 16810, "staged_surface": 0, "admitted_split": 16900, "staged_split": 0, "admitted_merge": 20047, "staged_merge": 0, "admitted_skel": 11779, "staged_skel": 0, "mean_count": 1218.0848388671875, "n_util_obs": 555, "util_count_spearman": 0.003733054342539175}, "probe_overhead_sec_total": 465.743, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009407, "alpha": 0.03, "t": 3981.4} |
| {"event": "consolidate", "step": 3800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4043.3} |
| {"event": "probe", "step": 3800, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3660.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.210309, "cps": 28.59375, "mean_p": 0.209531, "boundary_entropy": 0.073309, "bpb": 1.918684}, "tr": {"chunks": 4034.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.222836, "cps": 31.515625, "mean_p": 0.201498, "boundary_entropy": 0.101079, "bpb": 2.074499}, "fi": {"chunks": 5054.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.268673, "cps": 39.484375, "mean_p": 0.261173, "boundary_entropy": 0.077638, "bpb": 2.187599}}, "bpb": {"en": 1.91868, "tr": 2.0745, "fi": 2.1876}, "gini_cpc": 0.055441, "gini_cps": 0.0729, "bpic": 4.450267, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0595668863869134, "fi": 1.2775161202331038}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4114, "admitted_surface": 16816, "staged_surface": 0, "admitted_split": 17015, "staged_split": 0, "admitted_merge": 19837, "staged_merge": 0, "admitted_skel": 11868, "staged_skel": 0, "mean_count": 1233.8939208984375, "n_util_obs": 559, "util_count_spearman": 0.0047114116120760136}, "t": 4044.9} |
| {"event": "eval", "step": 3800, "phase": 2, "bpb": {"en": 1.93391, "tr": 2.08385, "fi": 2.19867}, "bytes_per_chunk": {"en": 4.448, "tr": 5.046, "fi": 3.977}, "bpb_mean": 2.07214, "frozen_inventory": true, "t": 4048.2} |
| {"event": "train", "step": 3800, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.3605, "bpb": 1.96279, "chunk_loss": 0.030748, "F": 0.22092, "G": 0.21848, "chunks": 226.22, "bytes_per_chunk": 4.527, "boundary_entropy": 0.07708, "router_noise": 0.0, "grad_norm": 0.6048, "lr": 0.0001177, "router_trainable": true, "sec_per_step": 1.371, "probe_sec": 0.1329, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10382, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.00126, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": 0.04822, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.34383, "trunc_abs_frac": 0.74792, "mean_delta_full": 0.27266, "read_hit_rate": {"surface": 0.68378, "split": 0.87033, "merge": 0.26258, "skel": 0.65892}, "mean_gate": {"surface": 0.1902, "split": 0.08939, "merge": 0.16508, "skel": 0.14476}, "n_reads": 20959, "n_cand": 31923, "read_suppressed_frac": 0.1025, "dropout_reads_logged": 20959, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4114, "admitted_surface": 16823, "staged_surface": 0, "admitted_split": 17005, "staged_split": 0, "admitted_merge": 19770, "staged_merge": 0, "admitted_skel": 11938, "staged_skel": 0, "mean_count": 1234.220703125, "n_util_obs": 559, "util_count_spearman": 0.004787632512098925}, "probe_overhead_sec_total": 472.0, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009583, "alpha": 0.03, "t": 4050.0} |
| {"event": "consolidate", "step": 3850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4089.2} |
| {"event": "train", "step": 3850, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.34439, "bpb": 1.93955, "chunk_loss": 0.030606, "F": 0.21698, "G": 0.21404, "chunks": 222.19, "bytes_per_chunk": 4.609, "boundary_entropy": 0.08019, "router_noise": 0.0, "grad_norm": 0.7034, "lr": 0.0001142, "router_trainable": true, "sec_per_step": 0.806, "probe_sec": 0.0861, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06182, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.07804, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": 0.03829, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -89.23923, "trunc_abs_frac": 0.51013, "mean_delta_full": 0.00118, "read_hit_rate": {"surface": 0.67025, "split": 0.85844, "merge": 0.2539, "skel": 0.64691}, "mean_gate": {"surface": 0.19723, "split": 0.09333, "merge": 0.16889, "skel": 0.15359}, "n_reads": 20321, "n_cand": 31483, "read_suppressed_frac": 0.1008, "dropout_reads_logged": 20321, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4139, "admitted_surface": 16815, "staged_surface": 0, "admitted_split": 17039, "staged_split": 0, "admitted_merge": 19761, "staged_merge": 0, "admitted_skel": 11921, "staged_skel": 0, "mean_count": 1250.34716796875, "n_util_obs": 565, "util_count_spearman": 0.0032897588023070435}, "probe_overhead_sec_total": 477.846, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.010148, "alpha": 0.03, "t": 4090.2} |
| {"event": "consolidate", "step": 3900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4128.4} |
| {"event": "train", "step": 3900, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.38149, "bpb": 1.99307, "chunk_loss": 0.030739, "F": 0.22171, "G": 0.21902, "chunks": 227.03, "bytes_per_chunk": 4.51, "boundary_entropy": 0.08264, "router_noise": 0.0, "grad_norm": 0.6422, "lr": 0.0001108, "router_trainable": true, "sec_per_step": 0.776, "probe_sec": 0.1066, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07531, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.09293, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.10565, "n_merge": 1, "mean_delta_merge": -0.09937, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 1.46377, "trunc_abs_frac": 0.60153, "mean_delta_full": -0.07159, "read_hit_rate": {"surface": 0.6668, "split": 0.86512, "merge": 0.25774, "skel": 0.64161}, "mean_gate": {"surface": 0.19607, "split": 0.09294, "merge": 0.17, "skel": 0.15215}, "n_reads": 20718, "n_cand": 32033, "read_suppressed_frac": 0.0985, "dropout_reads_logged": 20718, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4180, "admitted_surface": 16871, "staged_surface": 0, "admitted_split": 16969, "staged_split": 0, "admitted_merge": 19725, "staged_merge": 0, "admitted_skel": 11971, "staged_skel": 0, "mean_count": 1266.469482421875, "n_util_obs": 572, "util_count_spearman": 0.014589866374894376}, "probe_overhead_sec_total": 483.009, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009667, "alpha": 0.03, "t": 4129.0} |
| {"event": "consolidate", "step": 3950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4165.4} |
| {"event": "train", "step": 3950, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.40874, "bpb": 2.03238, "chunk_loss": 0.030875, "F": 0.22736, "G": 0.22583, "chunks": 232.81, "bytes_per_chunk": 4.398, "boundary_entropy": 0.07493, "router_noise": 0.0, "grad_norm": 0.6614, "lr": 0.0001074, "router_trainable": true, "sec_per_step": 0.751, "probe_sec": 0.0969, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06121, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": -0.00092, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": -0.10178, "n_merge": 3, "mean_delta_merge": -0.03485, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.42307, "trunc_abs_frac": 2.03587, "mean_delta_full": -0.29407, "read_hit_rate": {"surface": 0.68226, "split": 0.8732, "merge": 0.25995, "skel": 0.65824}, "mean_gate": {"surface": 0.18492, "split": 0.08613, "merge": 0.16223, "skel": 0.1418}, "n_reads": 21304, "n_cand": 32492, "read_suppressed_frac": 0.0988, "dropout_reads_logged": 21304, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4215, "admitted_surface": 16836, "staged_surface": 0, "admitted_split": 16948, "staged_split": 0, "admitted_merge": 19877, "staged_merge": 0, "admitted_skel": 11875, "staged_skel": 0, "mean_count": 1282.5262451171875, "n_util_obs": 578, "util_count_spearman": 0.024586797717538793}, "probe_overhead_sec_total": 488.056, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.00905, "alpha": 0.03, "t": 4166.6} |
| {"event": "consolidate", "step": 4000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4210.5} |
| {"event": "probe", "step": 4000, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3667.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.210711, "cps": 28.648438, "mean_p": 0.210409, "boundary_entropy": 0.072065, "bpb": 1.902544}, "tr": {"chunks": 4068.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.224714, "cps": 31.78125, "mean_p": 0.202189, "boundary_entropy": 0.097422, "bpb": 2.054922}, "fi": {"chunks": 5079.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.270002, "cps": 39.679688, "mean_p": 0.262772, "boundary_entropy": 0.074361, "bpb": 2.169777}}, "bpb": {"en": 1.90254, "tr": 2.05492, "fi": 2.16978}, "gini_cpc": 0.056033, "gini_cps": 0.073461, "bpic": 4.427345, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0664576233855958, "fi": 1.281384716254525}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4243, "admitted_surface": 16851, "staged_surface": 0, "admitted_split": 16853, "staged_split": 0, "admitted_merge": 20025, "staged_merge": 0, "admitted_skel": 11807, "staged_skel": 0, "mean_count": 1298.326416015625, "n_util_obs": 588, "util_count_spearman": 0.024767867189301383}, "t": 4212.5} |
| {"event": "eval", "step": 4000, "phase": 2, "bpb": {"en": 1.91965, "tr": 2.06413, "fi": 2.17987}, "bytes_per_chunk": {"en": 4.441, "tr": 5.002, "fi": 3.983}, "bpb_mean": 2.05455, "frozen_inventory": true, "t": 4215.6} |
| {"event": "train", "step": 4000, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.42746, "bpb": 2.05938, "chunk_loss": 0.030933, "F": 0.23083, "G": 0.22861, "chunks": 236.38, "bytes_per_chunk": 4.332, "boundary_entropy": 0.0737, "router_noise": 0.0, "grad_norm": 0.6537, "lr": 0.000104, "router_trainable": true, "sec_per_step": 1.015, "probe_sec": 0.1105, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08515, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.0154, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": -0.03613, "n_merge": 2, "mean_delta_merge": 0.10228, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -0.29879, "trunc_abs_frac": 0.77058, "mean_delta_full": -0.06047, "read_hit_rate": {"surface": 0.67764, "split": 0.87303, "merge": 0.25373, "skel": 0.65668}, "mean_gate": {"surface": 0.18995, "split": 0.08845, "merge": 0.16395, "skel": 0.14632}, "n_reads": 21354, "n_cand": 32772, "read_suppressed_frac": 0.0991, "dropout_reads_logged": 21354, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4245, "admitted_surface": 16847, "staged_surface": 0, "admitted_split": 16917, "staged_split": 0, "admitted_merge": 19918, "staged_merge": 0, "admitted_skel": 11854, "staged_skel": 0, "mean_count": 1298.640625, "n_util_obs": 588, "util_count_spearman": 0.02473855956501369}, "probe_overhead_sec_total": 493.405, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008784, "alpha": 0.03, "t": 4217.3} |
| {"event": "consolidate", "step": 4050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4278.0} |
| {"event": "train", "step": 4050, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.34185, "bpb": 1.93589, "chunk_loss": 0.030829, "F": 0.22455, "G": 0.22264, "chunks": 229.94, "bytes_per_chunk": 4.453, "boundary_entropy": 0.07405, "router_noise": 0.0, "grad_norm": 0.6556, "lr": 0.0001007, "router_trainable": true, "sec_per_step": 1.249, "probe_sec": 0.1136, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0776, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.04577, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.09369, "n_merge": 2, "mean_delta_merge": -0.08587, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": -22.38663, "trunc_abs_frac": 22.38663, "mean_delta_full": -0.00525, "read_hit_rate": {"surface": 0.68414, "split": 0.87227, "merge": 0.2705, "skel": 0.66181}, "mean_gate": {"surface": 0.17752, "split": 0.08236, "merge": 0.15291, "skel": 0.13595}, "n_reads": 21337, "n_cand": 32350, "read_suppressed_frac": 0.0997, "dropout_reads_logged": 21337, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4285, "admitted_surface": 16836, "staged_surface": 0, "admitted_split": 16990, "staged_split": 0, "admitted_merge": 19807, "staged_merge": 0, "admitted_skel": 11903, "staged_skel": 0, "mean_count": 1314.731689453125, "n_util_obs": 593, "util_count_spearman": 0.02740018765316524}, "probe_overhead_sec_total": 499.877, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009378, "alpha": 0.03, "t": 4279.8} |
| {"event": "probe", "step": 4096, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3662.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.210423, "cps": 28.609375, "mean_p": 0.209578, "boundary_entropy": 0.072815, "bpb": 1.894949}, "tr": {"chunks": 3939.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.217588, "cps": 30.773438, "mean_p": 0.200918, "boundary_entropy": 0.102192, "bpb": 2.048926}, "fi": {"chunks": 5076.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.269842, "cps": 39.65625, "mean_p": 0.261555, "boundary_entropy": 0.077764, "bpb": 2.157149}}, "bpb": {"en": 1.89495, "tr": 2.04893, "fi": 2.15715}, "gini_cpc": 0.056763, "gini_cps": 0.07436, "bpic": 4.475191, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0340492158575694, "fi": 1.2823763801122556}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4318, "admitted_surface": 16872, "staged_surface": 0, "admitted_split": 16942, "staged_split": 0, "admitted_merge": 19819, "staged_merge": 0, "admitted_skel": 11903, "staged_skel": 0, "mean_count": 1329.28369140625, "n_util_obs": 601, "util_count_spearman": 0.026821771030360732}, "t": 4331.4} |
| {"event": "consolidate", "step": 4100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4336.1} |
| {"event": "train", "step": 4100, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.421, "bpb": 2.05007, "chunk_loss": 0.030971, "F": 0.2309, "G": 0.22944, "chunks": 236.44, "bytes_per_chunk": 4.331, "boundary_entropy": 0.07632, "router_noise": 0.0, "grad_norm": 0.6485, "lr": 9.75e-05, "router_trainable": true, "sec_per_step": 1.14, "probe_sec": 0.178, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.14376, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": 0.00602, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.0697, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.01632, "trunc_abs_frac": 0.66575, "mean_delta_full": -0.05859, "read_hit_rate": {"surface": 0.66857, "split": 0.87341, "merge": 0.2366, "skel": 0.68399}, "mean_gate": {"surface": 0.18056, "split": 0.0812, "merge": 0.1597, "skel": 0.13473}, "n_reads": 21860, "n_cand": 33475, "read_suppressed_frac": 0.0973, "dropout_reads_logged": 21860, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4321, "admitted_surface": 16878, "staged_surface": 0, "admitted_split": 16908, "staged_split": 0, "admitted_merge": 19953, "staged_merge": 0, "admitted_skel": 11797, "staged_skel": 0, "mean_count": 1330.876708984375, "n_util_obs": 601, "util_count_spearman": 0.026630920410360383}, "probe_overhead_sec_total": 506.35, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008851, "alpha": 0.03, "t": 4336.8} |
| {"event": "consolidate", "step": 4150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4375.7} |
| {"event": "train", "step": 4150, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.36795, "bpb": 1.97353, "chunk_loss": 0.031111, "F": 0.23557, "G": 0.23427, "chunks": 241.22, "bytes_per_chunk": 4.245, "boundary_entropy": 0.0707, "router_noise": 0.0, "grad_norm": 0.6464, "lr": 9.43e-05, "router_trainable": true, "sec_per_step": 0.79, "probe_sec": 0.0757, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05407, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.04914, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": -0.03095, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.35071, "trunc_abs_frac": 0.38266, "mean_delta_full": 0.44454, "read_hit_rate": {"surface": 0.6831, "split": 0.87797, "merge": 0.26244, "skel": 0.68588}, "mean_gate": {"surface": 0.17786, "split": 0.08019, "merge": 0.15627, "skel": 0.13447}, "n_reads": 22571, "n_cand": 34030, "read_suppressed_frac": 0.1011, "dropout_reads_logged": 22571, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4353, "admitted_surface": 16840, "staged_surface": 0, "admitted_split": 16906, "staged_split": 0, "admitted_merge": 19966, "staged_merge": 0, "admitted_skel": 11824, "staged_skel": 0, "mean_count": 1346.93408203125, "n_util_obs": 610, "util_count_spearman": 0.02006684419037727}, "probe_overhead_sec_total": 511.501, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009378, "alpha": 0.03, "t": 4376.3} |
| {"event": "consolidate", "step": 4200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4421.6} |
| {"event": "probe", "step": 4200, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3666.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.210653, "cps": 28.640625, "mean_p": 0.209253, "boundary_entropy": 0.07261, "bpb": 1.891835}, "tr": {"chunks": 3931.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.217146, "cps": 30.710938, "mean_p": 0.199801, "boundary_entropy": 0.100331, "bpb": 2.039255}, "fi": {"chunks": 5090.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.270586, "cps": 39.765625, "mean_p": 0.26185, "boundary_entropy": 0.074107, "bpb": 2.145259}}, "bpb": {"en": 1.89183, "tr": 2.03926, "fi": 2.14526}, "gini_cpc": 0.057211, "gini_cps": 0.074827, "bpic": 4.471664, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0308231231488338, "fi": 1.284510203618195}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4382, "admitted_surface": 16831, "staged_surface": 0, "admitted_split": 16858, "staged_split": 0, "admitted_merge": 19982, "staged_merge": 0, "admitted_skel": 11865, "staged_skel": 0, "mean_count": 1362.690185546875, "n_util_obs": 624, "util_count_spearman": 0.013984442696437125}, "t": 4423.3} |
| {"event": "eval", "step": 4200, "phase": 2, "bpb": {"en": 1.90675, "tr": 2.05028, "fi": 2.15518}, "bytes_per_chunk": {"en": 4.447, "tr": 5.185, "fi": 3.973}, "bpb_mean": 2.0374, "frozen_inventory": true, "t": 4426.4} |
| {"event": "phase_transition", "step": 4200, "phase": 3, "objective": "global", "router_trainable": true, "encoder_trainable": true, "frozen_router_params": 0, "t": 4428.0} |
| {"event": "train", "step": 4200, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35883, "bpb": 1.96037, "chunk_loss": 0.03064, "F": 0.21933, "G": 0.2177, "chunks": 224.59, "bytes_per_chunk": 4.559, "boundary_entropy": 0.07536, "router_noise": 0.0, "grad_norm": 0.6528, "lr": 9.12e-05, "router_trainable": true, "sec_per_step": 1.06, "probe_sec": 0.1075, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06264, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.00429, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": -0.11142, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.80656, "trunc_abs_frac": 1.16156, "mean_delta_full": -0.11722, "read_hit_rate": {"surface": 0.67072, "split": 0.86478, "merge": 0.24624, "skel": 0.65992}, "mean_gate": {"surface": 0.17893, "split": 0.08179, "merge": 0.15636, "skel": 0.13705}, "n_reads": 20596, "n_cand": 31770, "read_suppressed_frac": 0.0988, "dropout_reads_logged": 20596, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4384, "admitted_surface": 16827, "staged_surface": 0, "admitted_split": 16949, "staged_split": 0, "admitted_merge": 19857, "staged_merge": 0, "admitted_skel": 11903, "staged_skel": 0, "mean_count": 1363.0020751953125, "n_util_obs": 624, "util_count_spearman": 0.014151697915195677}, "probe_overhead_sec_total": 517.237, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009268, "alpha": 0.03, "t": 4429.3} |
| {"event": "consolidate", "step": 4250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4483.8} |
| {"event": "train", "step": 4250, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35912, "bpb": 1.9608, "chunk_loss": 0.030683, "F": 0.21753, "G": 0.21538, "chunks": 222.75, "bytes_per_chunk": 4.597, "boundary_entropy": 0.0783, "router_noise": 0.0, "grad_norm": 0.6883, "lr": 8.81e-05, "router_trainable": true, "sec_per_step": 1.104, "probe_sec": 0.0926, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06127, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.07074, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": -0.05109, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.15675, "trunc_abs_frac": 0.87635, "mean_delta_full": -0.23041, "read_hit_rate": {"surface": 0.66911, "split": 0.86449, "merge": 0.25184, "skel": 0.6563}, "mean_gate": {"surface": 0.18688, "split": 0.08541, "merge": 0.16405, "skel": 0.14568}, "n_reads": 20574, "n_cand": 31694, "read_suppressed_frac": 0.0974, "dropout_reads_logged": 20574, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4411, "admitted_surface": 16849, "staged_surface": 0, "admitted_split": 16973, "staged_split": 0, "admitted_merge": 19772, "staged_merge": 0, "admitted_skel": 11942, "staged_skel": 0, "mean_count": 1379.171875, "n_util_obs": 635, "util_count_spearman": 0.01563791955242821}, "probe_overhead_sec_total": 523.426, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008751, "alpha": 0.03, "t": 4484.5} |
| {"event": "consolidate", "step": 4300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4545.1} |
| {"event": "train", "step": 4300, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.44446, "bpb": 2.08392, "chunk_loss": 0.030843, "F": 0.22412, "G": 0.22174, "chunks": 229.5, "bytes_per_chunk": 4.462, "boundary_entropy": 0.07799, "router_noise": 0.0, "grad_norm": 0.752, "lr": 8.5e-05, "router_trainable": true, "sec_per_step": 1.233, "probe_sec": 0.1018, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06516, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.12618, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.1543, "n_merge": 1, "mean_delta_merge": -0.18842, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.85276, "trunc_abs_frac": 1.2764, "mean_delta_full": 0.18829, "read_hit_rate": {"surface": 0.65796, "split": 0.86082, "merge": 0.23805, "skel": 0.6541}, "mean_gate": {"surface": 0.18929, "split": 0.08649, "merge": 0.16524, "skel": 0.14563}, "n_reads": 20624, "n_cand": 32235, "read_suppressed_frac": 0.0975, "dropout_reads_logged": 20624, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4449, "admitted_surface": 16829, "staged_surface": 0, "admitted_split": 16982, "staged_split": 0, "admitted_merge": 19830, "staged_merge": 0, "admitted_skel": 11895, "staged_skel": 0, "mean_count": 1395.2593994140625, "n_util_obs": 642, "util_count_spearman": 0.031714112908131156}, "probe_overhead_sec_total": 529.843, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009192, "alpha": 0.03, "t": 4546.2} |
| {"event": "consolidate", "step": 4350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4589.3} |
| {"event": "train", "step": 4350, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.30498, "bpb": 1.88269, "chunk_loss": 0.030775, "F": 0.22253, "G": 0.22096, "chunks": 227.88, "bytes_per_chunk": 4.494, "boundary_entropy": 0.07552, "router_noise": 0.0, "grad_norm": 0.7142, "lr": 8.21e-05, "router_trainable": true, "sec_per_step": 0.883, "probe_sec": 0.1142, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08281, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.06683, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": 0.0434, "n_merge": 1, "mean_delta_merge": 0.1952, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.62588, "trunc_abs_frac": 0.80009, "mean_delta_full": 0.06934, "read_hit_rate": {"surface": 0.6759, "split": 0.8728, "merge": 0.25595, "skel": 0.67618}, "mean_gate": {"surface": 0.17488, "split": 0.07849, "merge": 0.15106, "skel": 0.1358}, "n_reads": 21249, "n_cand": 32315, "read_suppressed_frac": 0.099, "dropout_reads_logged": 21249, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4483, "admitted_surface": 16836, "staged_surface": 0, "admitted_split": 16957, "staged_split": 0, "admitted_merge": 19885, "staged_merge": 0, "admitted_skel": 11858, "staged_skel": 0, "mean_count": 1411.386962890625, "n_util_obs": 650, "util_count_spearman": 0.03804547557417978}, "probe_overhead_sec_total": 535.333, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008995, "alpha": 0.03, "t": 4590.4} |
| {"event": "consolidate", "step": 4400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4628.8} |
| {"event": "probe", "step": 4400, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3671.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.210941, "cps": 28.679688, "mean_p": 0.209217, "boundary_entropy": 0.069608, "bpb": 1.876849}, "tr": {"chunks": 3831.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.211622, "cps": 29.929688, "mean_p": 0.197285, "boundary_entropy": 0.100179, "bpb": 2.027394}, "fi": {"chunks": 5051.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.268513, "cps": 39.460938, "mean_p": 0.260135, "boundary_entropy": 0.073124, "bpb": 2.131719}}, "bpb": {"en": 1.87685, "tr": 2.02739, "fi": 2.13172}, "gini_cpc": 0.055539, "gini_cps": 0.073289, "bpic": 4.519398, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0032319073491403, "fi": 1.2729320483570958}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4514, "admitted_surface": 16839, "staged_surface": 0, "admitted_split": 16887, "staged_split": 0, "admitted_merge": 19939, "staged_merge": 0, "admitted_skel": 11871, "staged_skel": 0, "mean_count": 1427.09619140625, "n_util_obs": 657, "util_count_spearman": 0.034493903306207566}, "t": 4630.6} |
| {"event": "eval", "step": 4400, "phase": 3, "bpb": {"en": 1.89232, "tr": 2.03825, "fi": 2.14017}, "bytes_per_chunk": {"en": 4.441, "tr": 5.296, "fi": 4.001}, "bpb_mean": 2.02358, "frozen_inventory": true, "t": 4633.6} |
| {"event": "train", "step": 4400, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.41197, "bpb": 2.03704, "chunk_loss": 0.031074, "F": 0.23001, "G": 0.22873, "chunks": 235.53, "bytes_per_chunk": 4.348, "boundary_entropy": 0.07357, "router_noise": 0.0, "grad_norm": 0.7013, "lr": 7.92e-05, "router_trainable": true, "sec_per_step": 0.907, "probe_sec": 0.1263, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09251, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.00048, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.17017, "n_merge": 1, "mean_delta_merge": -0.02545, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.48574, "trunc_abs_frac": 0.48574, "mean_delta_full": 0.1626, "read_hit_rate": {"surface": 0.67209, "split": 0.87486, "merge": 0.24113, "skel": 0.66508}, "mean_gate": {"surface": 0.17804, "split": 0.08097, "merge": 0.15693, "skel": 0.13435}, "n_reads": 21535, "n_cand": 33074, "read_suppressed_frac": 0.0998, "dropout_reads_logged": 21535, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4515, "admitted_surface": 16820, "staged_surface": 0, "admitted_split": 16898, "staged_split": 0, "admitted_merge": 19967, "staged_merge": 0, "admitted_skel": 11851, "staged_skel": 0, "mean_count": 1427.419189453125, "n_util_obs": 657, "util_count_spearman": 0.03433772085765387}, "probe_overhead_sec_total": 540.703, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008588, "alpha": 0.03, "t": 4635.7} |
| {"event": "consolidate", "step": 4450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4681.1} |
| {"event": "train", "step": 4450, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35139, "bpb": 1.94964, "chunk_loss": 0.030818, "F": 0.22458, "G": 0.22172, "chunks": 229.97, "bytes_per_chunk": 4.453, "boundary_entropy": 0.07413, "router_noise": 0.0, "grad_norm": 0.6373, "lr": 7.63e-05, "router_trainable": true, "sec_per_step": 0.921, "probe_sec": 0.1237, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07793, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.07719, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": 0.00174, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 0.41727, "trunc_abs_frac": 4.03488, "mean_delta_full": -0.27519, "read_hit_rate": {"surface": 0.67272, "split": 0.86968, "merge": 0.25195, "skel": 0.66547}, "mean_gate": {"surface": 0.18132, "split": 0.08197, "merge": 0.15901, "skel": 0.14113}, "n_reads": 21040, "n_cand": 32279, "read_suppressed_frac": 0.099, "dropout_reads_logged": 21040, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4553, "admitted_surface": 16830, "staged_surface": 0, "admitted_split": 16975, "staged_split": 0, "admitted_merge": 19875, "staged_merge": 0, "admitted_skel": 11856, "staged_skel": 0, "mean_count": 1443.3905029296875, "n_util_obs": 666, "util_count_spearman": 0.02990498308872354}, "probe_overhead_sec_total": 546.209, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008621, "alpha": 0.03, "t": 4681.8} |
| {"event": "consolidate", "step": 4500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4736.6} |
| {"event": "train", "step": 4500, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.36362, "bpb": 1.96728, "chunk_loss": 0.030919, "F": 0.22662, "G": 0.22529, "chunks": 232.06, "bytes_per_chunk": 4.413, "boundary_entropy": 0.07709, "router_noise": 0.0, "grad_norm": 0.7132, "lr": 7.36e-05, "router_trainable": true, "sec_per_step": 1.11, "probe_sec": 0.1015, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07676, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.06984, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": -0.19067, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 3.27328, "trunc_abs_frac": 2.22213, "mean_delta_full": -0.0533, "read_hit_rate": {"surface": 0.66378, "split": 0.87557, "merge": 0.23478, "skel": 0.66661}, "mean_gate": {"surface": 0.17552, "split": 0.0786, "merge": 0.15555, "skel": 0.13287}, "n_reads": 21385, "n_cand": 32924, "read_suppressed_frac": 0.0981, "dropout_reads_logged": 21385, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4581, "admitted_surface": 16843, "staged_surface": 0, "admitted_split": 16898, "staged_split": 0, "admitted_merge": 19918, "staged_merge": 0, "admitted_skel": 11877, "staged_skel": 0, "mean_count": 1459.4052734375, "n_util_obs": 675, "util_count_spearman": 0.025325912997497486}, "probe_overhead_sec_total": 552.2, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008692, "alpha": 0.03, "t": 4737.3} |
| {"event": "consolidate", "step": 4550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4794.7} |
| {"event": "train", "step": 4550, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.3403, "bpb": 1.93364, "chunk_loss": 0.030935, "F": 0.22754, "G": 0.22598, "chunks": 233.0, "bytes_per_chunk": 4.395, "boundary_entropy": 0.07426, "router_noise": 0.0, "grad_norm": 0.6457, "lr": 7.09e-05, "router_trainable": true, "sec_per_step": 1.167, "probe_sec": 0.1201, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06093, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.03448, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": 0.08514, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.19398, "trunc_abs_frac": 0.30514, "mean_delta_full": 0.24684, "read_hit_rate": {"surface": 0.68534, "split": 0.8774, "merge": 0.26204, "skel": 0.67683}, "mean_gate": {"surface": 0.17666, "split": 0.07938, "merge": 0.15715, "skel": 0.13438}, "n_reads": 21696, "n_cand": 32789, "read_suppressed_frac": 0.0999, "dropout_reads_logged": 21696, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4615, "admitted_surface": 16839, "staged_surface": 0, "admitted_split": 16894, "staged_split": 0, "admitted_merge": 19941, "staged_merge": 0, "admitted_skel": 11862, "staged_skel": 0, "mean_count": 1475.244384765625, "n_util_obs": 679, "util_count_spearman": 0.014705997351956018}, "probe_overhead_sec_total": 558.123, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009973, "alpha": 0.03, "t": 4795.6} |
| {"event": "consolidate", "step": 4600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4838.4} |
| {"event": "probe", "step": 4600, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3666.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.210653, "cps": 28.640625, "mean_p": 0.208919, "boundary_entropy": 0.069251, "bpb": 1.867327}, "tr": {"chunks": 3878.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.214219, "cps": 30.296875, "mean_p": 0.198281, "boundary_entropy": 0.09667, "bpb": 2.010154}, "fi": {"chunks": 5056.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.268779, "cps": 39.5, "mean_p": 0.260045, "boundary_entropy": 0.071843, "bpb": 2.118195}}, "bpb": {"en": 1.86733, "tr": 2.01015, "fi": 2.11819}, "gini_cpc": 0.055864, "gini_cps": 0.073545, "bpic": 4.50254, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0169249736889283, "fi": 1.2759299782895077}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4660, "admitted_surface": 16830, "staged_surface": 0, "admitted_split": 16925, "staged_split": 0, "admitted_merge": 19876, "staged_merge": 0, "admitted_skel": 11905, "staged_skel": 0, "mean_count": 1490.91748046875, "n_util_obs": 686, "util_count_spearman": 0.017108079667657997}, "t": 4840.2} |
| {"event": "eval", "step": 4600, "phase": 3, "bpb": {"en": 1.88304, "tr": 2.02071, "fi": 2.12843}, "bytes_per_chunk": {"en": 4.449, "tr": 5.239, "fi": 4.003}, "bpb_mean": 2.01073, "frozen_inventory": true, "t": 4843.4} |
| {"event": "train", "step": 4600, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.39436, "bpb": 2.01163, "chunk_loss": 0.030725, "F": 0.21924, "G": 0.21807, "chunks": 224.5, "bytes_per_chunk": 4.561, "boundary_entropy": 0.07387, "router_noise": 0.0, "grad_norm": 0.6892, "lr": 6.82e-05, "router_trainable": true, "sec_per_step": 0.99, "probe_sec": 0.0811, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05462, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.02461, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 3, "mean_delta_split": 0.08091, "n_merge": 1, "mean_delta_merge": -0.12897, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": -0.66022, "trunc_abs_frac": 0.27624, "mean_delta_full": 0.05473, "read_hit_rate": {"surface": 0.65366, "split": 0.86454, "merge": 0.23034, "skel": 0.65421}, "mean_gate": {"surface": 0.17751, "split": 0.08088, "merge": 0.15449, "skel": 0.13825}, "n_reads": 20454, "n_cand": 31935, "read_suppressed_frac": 0.0987, "dropout_reads_logged": 20454, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4663, "admitted_surface": 16864, "staged_surface": 0, "admitted_split": 16969, "staged_split": 0, "admitted_merge": 19813, "staged_merge": 0, "admitted_skel": 11890, "staged_skel": 0, "mean_count": 1491.229736328125, "n_util_obs": 687, "util_count_spearman": 0.01780214690004888}, "probe_overhead_sec_total": 564.007, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009701, "alpha": 0.03, "t": 4845.1} |
| {"event": "consolidate", "step": 4650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4883.0} |
| {"event": "train", "step": 4650, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32447, "bpb": 1.9108, "chunk_loss": 0.030709, "F": 0.2197, "G": 0.2189, "chunks": 224.97, "bytes_per_chunk": 4.552, "boundary_entropy": 0.07617, "router_noise": 0.0, "grad_norm": 0.6424, "lr": 6.57e-05, "router_trainable": true, "sec_per_step": 0.778, "probe_sec": 0.0677, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04492, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.0297, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.2991, "n_merge": 1, "mean_delta_merge": 0.12225, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.24062, "trunc_abs_frac": 1.94495, "mean_delta_full": -0.13883, "read_hit_rate": {"surface": 0.66946, "split": 0.86755, "merge": 0.25018, "skel": 0.66263}, "mean_gate": {"surface": 0.17787, "split": 0.07967, "merge": 0.15603, "skel": 0.13651}, "n_reads": 20714, "n_cand": 31859, "read_suppressed_frac": 0.1025, "dropout_reads_logged": 20714, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4690, "admitted_surface": 16840, "staged_surface": 0, "admitted_split": 16958, "staged_split": 0, "admitted_merge": 19833, "staged_merge": 0, "admitted_skel": 11905, "staged_skel": 0, "mean_count": 1507.2381591796875, "n_util_obs": 694, "util_count_spearman": 0.017770647007597357}, "probe_overhead_sec_total": 569.564, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009731, "alpha": 0.03, "t": 4884.0} |
| {"event": "consolidate", "step": 4700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4922.1} |
| {"event": "train", "step": 4700, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.33653, "bpb": 1.92821, "chunk_loss": 0.030825, "F": 0.22424, "G": 0.22116, "chunks": 229.62, "bytes_per_chunk": 4.459, "boundary_entropy": 0.07406, "router_noise": 0.0, "grad_norm": 0.6339, "lr": 6.32e-05, "router_trainable": true, "sec_per_step": 0.783, "probe_sec": 0.0782, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0528, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.02131, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 3, "mean_delta_split": 0.07632, "n_merge": 1, "mean_delta_merge": 0.03827, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": -0.0231, "trunc_abs_frac": 0.29501, "mean_delta_full": 0.10325, "read_hit_rate": {"surface": 0.67236, "split": 0.86863, "merge": 0.24849, "skel": 0.65684}, "mean_gate": {"surface": 0.17216, "split": 0.07734, "merge": 0.15045, "skel": 0.13226}, "n_reads": 20996, "n_cand": 32332, "read_suppressed_frac": 0.1011, "dropout_reads_logged": 20996, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4725, "admitted_surface": 16821, "staged_surface": 0, "admitted_split": 16954, "staged_split": 0, "admitted_merge": 19893, "staged_merge": 0, "admitted_skel": 11868, "staged_skel": 0, "mean_count": 1523.2784423828125, "n_util_obs": 702, "util_count_spearman": 0.01400200678312364}, "probe_overhead_sec_total": 574.934, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009372, "alpha": 0.03, "t": 4923.2} |
| {"event": "consolidate", "step": 4750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 4969.4} |
| {"event": "train", "step": 4750, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.29679, "bpb": 1.87088, "chunk_loss": 0.03083, "F": 0.2261, "G": 0.22304, "chunks": 231.53, "bytes_per_chunk": 4.423, "boundary_entropy": 0.07503, "router_noise": 0.0, "grad_norm": 0.6519, "lr": 6.08e-05, "router_trainable": true, "sec_per_step": 0.939, "probe_sec": 0.0944, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06615, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.04856, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.06991, "n_merge": 1, "mean_delta_merge": 0.07608, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.03125, "trunc_abs_frac": 0.91518, "mean_delta_full": 0.03979, "read_hit_rate": {"surface": 0.69215, "split": 0.87801, "merge": 0.27311, "skel": 0.67633}, "mean_gate": {"surface": 0.17314, "split": 0.07752, "merge": 0.14791, "skel": 0.13408}, "n_reads": 21706, "n_cand": 32587, "read_suppressed_frac": 0.0989, "dropout_reads_logged": 21706, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4759, "admitted_surface": 16821, "staged_surface": 0, "admitted_split": 16932, "staged_split": 0, "admitted_merge": 19908, "staged_merge": 0, "admitted_skel": 11875, "staged_skel": 0, "mean_count": 1539.3809814453125, "n_util_obs": 711, "util_count_spearman": 0.017595701851415298}, "probe_overhead_sec_total": 580.752, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008483, "alpha": 0.03, "t": 4970.1} |
| {"event": "consolidate", "step": 4800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 5026.5} |
| {"event": "probe", "step": 4800, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3672.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.210998, "cps": 28.6875, "mean_p": 0.209169, "boundary_entropy": 0.06494, "bpb": 1.858733}, "tr": {"chunks": 3738.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.206485, "cps": 29.203125, "mean_p": 0.198167, "boundary_entropy": 0.096906, "bpb": 2.005123}, "fi": {"chunks": 5075.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.269789, "cps": 39.648438, "mean_p": 0.261036, "boundary_entropy": 0.069565, "bpb": 2.105256}}, "bpb": {"en": 1.85873, "tr": 2.00512, "fi": 2.10526}, "gini_cpc": 0.061406, "gini_cps": 0.074917, "bpic": 4.544013, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9786112257420231, "fi": 1.2786321224926454}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4789, "admitted_surface": 16841, "staged_surface": 0, "admitted_split": 16981, "staged_split": 0, "admitted_merge": 19851, "staged_merge": 0, "admitted_skel": 11863, "staged_skel": 0, "mean_count": 1554.9425048828125, "n_util_obs": 718, "util_count_spearman": 0.015557638231973817}, "t": 5028.6} |
| {"event": "eval", "step": 4800, "phase": 3, "bpb": {"en": 1.87402, "tr": 2.01448, "fi": 2.11497}, "bytes_per_chunk": {"en": 4.442, "tr": 5.432, "fi": 3.988}, "bpb_mean": 2.00116, "frozen_inventory": true, "t": 5031.6} |
| {"event": "train", "step": 4800, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.36042, "bpb": 1.96267, "chunk_loss": 0.031092, "F": 0.23267, "G": 0.23217, "chunks": 238.25, "bytes_per_chunk": 4.298, "boundary_entropy": 0.07021, "router_noise": 0.0, "grad_norm": 0.6902, "lr": 5.85e-05, "router_trainable": true, "sec_per_step": 1.282, "probe_sec": 0.0921, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06161, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.0415, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.10448, "n_merge": 2, "mean_delta_merge": 0.03221, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.28046, "trunc_abs_frac": 68.45152, "mean_delta_full": -0.11422, "read_hit_rate": {"surface": 0.674, "split": 0.87209, "merge": 0.23757, "skel": 0.68252}, "mean_gate": {"surface": 0.17002, "split": 0.07661, "merge": 0.14701, "skel": 0.12918}, "n_reads": 21992, "n_cand": 33645, "read_suppressed_frac": 0.0999, "dropout_reads_logged": 21992, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4789, "admitted_surface": 16814, "staged_surface": 0, "admitted_split": 16948, "staged_split": 0, "admitted_merge": 20014, "staged_merge": 0, "admitted_skel": 11760, "staged_skel": 0, "mean_count": 1555.2847900390625, "n_util_obs": 718, "util_count_spearman": 0.015597741379083686}, "probe_overhead_sec_total": 586.822, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009312, "alpha": 0.03, "t": 5034.2} |
| {"event": "consolidate", "step": 4850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 5089.9} |
| {"event": "train", "step": 4850, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32402, "bpb": 1.91015, "chunk_loss": 0.030964, "F": 0.22894, "G": 0.22769, "chunks": 234.44, "bytes_per_chunk": 4.368, "boundary_entropy": 0.07331, "router_noise": 0.0, "grad_norm": 0.6711, "lr": 5.62e-05, "router_trainable": true, "sec_per_step": 1.126, "probe_sec": 0.1394, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.11347, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.03973, "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": 0.22001, "trunc_abs_frac": 2.41258, "mean_delta_full": -0.21627, "read_hit_rate": {"surface": 0.68527, "split": 0.87608, "merge": 0.25746, "skel": 0.6808}, "mean_gate": {"surface": 0.16927, "split": 0.07618, "merge": 0.14397, "skel": 0.12994}, "n_reads": 21622, "n_cand": 32748, "read_suppressed_frac": 0.0984, "dropout_reads_logged": 21622, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4825, "admitted_surface": 16836, "staged_surface": 0, "admitted_split": 16957, "staged_split": 0, "admitted_merge": 19927, "staged_merge": 0, "admitted_skel": 11816, "staged_skel": 0, "mean_count": 1571.2100830078125, "n_util_obs": 722, "util_count_spearman": 0.022318075645025838}, "probe_overhead_sec_total": 592.741, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008207, "alpha": 0.03, "t": 5090.5} |
| {"event": "consolidate", "step": 4900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 5132.8} |
| {"event": "train", "step": 4900, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32692, "bpb": 1.91434, "chunk_loss": 0.030617, "F": 0.21674, "G": 0.21493, "chunks": 221.94, "bytes_per_chunk": 4.614, "boundary_entropy": 0.0702, "router_noise": 0.0, "grad_norm": 0.6994, "lr": 5.41e-05, "router_trainable": true, "sec_per_step": 0.861, "probe_sec": 0.1571, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.1095, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.09758, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.00183, "n_merge": 1, "mean_delta_merge": -2.33493, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 1.1327, "trunc_abs_frac": 1.1327, "mean_delta_full": 0.24976, "read_hit_rate": {"surface": 0.66393, "split": 0.86652, "merge": 0.24339, "skel": 0.64697}, "mean_gate": {"surface": 0.16939, "split": 0.07654, "merge": 0.1435, "skel": 0.1319}, "n_reads": 20287, "n_cand": 31451, "read_suppressed_frac": 0.1012, "dropout_reads_logged": 20287, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4864, "admitted_surface": 16852, "staged_surface": 0, "admitted_split": 16974, "staged_split": 0, "admitted_merge": 19774, "staged_merge": 0, "admitted_skel": 11936, "staged_skel": 0, "mean_count": 1587.0791015625, "n_util_obs": 727, "util_count_spearman": 0.023261517694701774}, "probe_overhead_sec_total": 598.829, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009289, "alpha": 0.03, "t": 5133.6} |
| {"event": "consolidate", "step": 4950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 5175.2} |
| {"event": "train", "step": 4950, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.30965, "bpb": 1.88943, "chunk_loss": 0.030939, "F": 0.22971, "G": 0.22697, "chunks": 235.22, "bytes_per_chunk": 4.353, "boundary_entropy": 0.06918, "router_noise": 0.0, "grad_norm": 0.6759, "lr": 5.2e-05, "router_trainable": true, "sec_per_step": 0.844, "probe_sec": 0.0917, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06508, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.03326, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": 0.02017, "n_merge": 1, "mean_delta_merge": 0.28912, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.56217, "trunc_abs_frac": 0.53192, "mean_delta_full": 0.22186, "read_hit_rate": {"surface": 0.68659, "split": 0.87837, "merge": 0.26343, "skel": 0.66716}, "mean_gate": {"surface": 0.16534, "split": 0.07295, "merge": 0.14569, "skel": 0.12607}, "n_reads": 21701, "n_cand": 32856, "read_suppressed_frac": 0.1021, "dropout_reads_logged": 21701, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4890, "admitted_surface": 16824, "staged_surface": 0, "admitted_split": 16925, "staged_split": 0, "admitted_merge": 19930, "staged_merge": 0, "admitted_skel": 11857, "staged_skel": 0, "mean_count": 1603.01806640625, "n_util_obs": 731, "util_count_spearman": 0.014478242917705376}, "probe_overhead_sec_total": 605.23, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009238, "alpha": 0.03, "t": 5175.8} |
| {"event": "consolidate", "step": 5000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 5222.8} |
| {"event": "probe", "step": 5000, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3660.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.210309, "cps": 28.59375, "mean_p": 0.208284, "boundary_entropy": 0.067963, "bpb": 1.849697}, "tr": {"chunks": 3775.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.208529, "cps": 29.492188, "mean_p": 0.19792, "boundary_entropy": 0.097469, "bpb": 1.994326}, "fi": {"chunks": 5044.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.268141, "cps": 39.40625, "mean_p": 0.259793, "boundary_entropy": 0.072158, "bpb": 2.093}}, "bpb": {"en": 1.8497, "tr": 1.99433, "fi": 2.093}, "gini_cpc": 0.057849, "gini_cps": 0.073938, "bpic": 4.546198, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9915381745440254, "fi": 1.2749883875060894}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4922, "admitted_surface": 16886, "staged_surface": 0, "admitted_split": 16957, "staged_split": 0, "admitted_merge": 19814, "staged_merge": 0, "admitted_skel": 11879, "staged_skel": 0, "mean_count": 1618.586181640625, "n_util_obs": 739, "util_count_spearman": 0.015659497810850027}, "t": 5225.0} |
| {"event": "eval", "step": 5000, "phase": 3, "bpb": {"en": 1.86522, "tr": 2.00363, "fi": 2.10205}, "bytes_per_chunk": {"en": 4.446, "tr": 5.405, "fi": 4.009}, "bpb_mean": 1.9903, "frozen_inventory": true, "t": 5228.3} |
| {"event": "train", "step": 5000, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.3164, "bpb": 1.89916, "chunk_loss": 0.030832, "F": 0.22479, "G": 0.223, "chunks": 230.19, "bytes_per_chunk": 4.449, "boundary_entropy": 0.07363, "router_noise": 0.0, "grad_norm": 0.6528, "lr": 5e-05, "router_trainable": true, "sec_per_step": 1.089, "probe_sec": 0.1023, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06959, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": -0.00238, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.16156, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.28947, "trunc_abs_frac": 0.68686, "mean_delta_full": 0.04523, "read_hit_rate": {"surface": 0.6823, "split": 0.87362, "merge": 0.25472, "skel": 0.65928}, "mean_gate": {"surface": 0.16809, "split": 0.0748, "merge": 0.14892, "skel": 0.1291}, "n_reads": 21167, "n_cand": 32315, "read_suppressed_frac": 0.1037, "dropout_reads_logged": 21167, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4923, "admitted_surface": 16837, "staged_surface": 0, "admitted_split": 16951, "staged_split": 0, "admitted_merge": 19834, "staged_merge": 0, "admitted_skel": 11914, "staged_skel": 0, "mean_count": 1618.916015625, "n_util_obs": 739, "util_count_spearman": 0.015142711277897257}, "probe_overhead_sec_total": 611.547, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008758, "alpha": 0.03, "t": 5230.3} |
| {"event": "consolidate", "step": 5050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 5287.0} |
| {"event": "train", "step": 5050, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.33066, "bpb": 1.91974, "chunk_loss": 0.030817, "F": 0.22531, "G": 0.22252, "chunks": 230.72, "bytes_per_chunk": 4.438, "boundary_entropy": 0.07351, "router_noise": 0.0, "grad_norm": 0.647, "lr": 4.81e-05, "router_trainable": true, "sec_per_step": 1.162, "probe_sec": 0.101, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06435, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.00329, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.02252, "n_merge": 2, "mean_delta_merge": -0.1319, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.12776, "trunc_abs_frac": 1.3638, "mean_delta_full": 0.13253, "read_hit_rate": {"surface": 0.67746, "split": 0.86769, "merge": 0.25905, "skel": 0.66204}, "mean_gate": {"surface": 0.16687, "split": 0.07566, "merge": 0.14284, "skel": 0.13029}, "n_reads": 21157, "n_cand": 32393, "read_suppressed_frac": 0.1032, "dropout_reads_logged": 21157, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4951, "admitted_surface": 16828, "staged_surface": 0, "admitted_split": 17000, "staged_split": 0, "admitted_merge": 19815, "staged_merge": 0, "admitted_skel": 11893, "staged_skel": 0, "mean_count": 1634.858154296875, "n_util_obs": 748, "util_count_spearman": 0.020697599744739077}, "probe_overhead_sec_total": 617.991, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008594, "alpha": 0.03, "t": 5288.4} |
| {"event": "consolidate", "step": 5100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 5349.7} |
| {"event": "train", "step": 5100, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.29801, "bpb": 1.87264, "chunk_loss": 0.030582, "F": 0.2157, "G": 0.21321, "chunks": 220.88, "bytes_per_chunk": 4.636, "boundary_entropy": 0.07071, "router_noise": 0.0, "grad_norm": 0.6723, "lr": 4.63e-05, "router_trainable": true, "sec_per_step": 1.264, "probe_sec": 0.285, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.16784, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.00459, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": 0.03995, "n_merge": 1, "mean_delta_merge": -0.14737, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.21359, "trunc_abs_frac": 0.21359, "mean_delta_full": -0.13831, "read_hit_rate": {"surface": 0.66359, "split": 0.86384, "merge": 0.24186, "skel": 0.64965}, "mean_gate": {"surface": 0.16433, "split": 0.07325, "merge": 0.14189, "skel": 0.12527}, "n_reads": 19994, "n_cand": 31102, "read_suppressed_frac": 0.1015, "dropout_reads_logged": 19994, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4981, "admitted_surface": 16833, "staged_surface": 0, "admitted_split": 16995, "staged_split": 0, "admitted_merge": 19787, "staged_merge": 0, "admitted_skel": 11921, "staged_skel": 0, "mean_count": 1650.712158203125, "n_util_obs": 753, "util_count_spearman": 0.019528495294978323}, "probe_overhead_sec_total": 624.743, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009146, "alpha": 0.03, "t": 5351.6} |
| {"event": "consolidate", "step": 5150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 5399.6} |
| {"event": "train", "step": 5150, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.28917, "bpb": 1.85988, "chunk_loss": 0.030895, "F": 0.22961, "G": 0.22567, "chunks": 235.12, "bytes_per_chunk": 4.355, "boundary_entropy": 0.07226, "router_noise": 0.0, "grad_norm": 0.6621, "lr": 4.45e-05, "router_trainable": true, "sec_per_step": 0.984, "probe_sec": 0.0948, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06414, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.14457, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": -0.01909, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.84171, "trunc_abs_frac": 0.84171, "mean_delta_full": 0.17737, "read_hit_rate": {"surface": 0.67952, "split": 0.88256, "merge": 0.25992, "skel": 0.68359}, "mean_gate": {"surface": 0.1629, "split": 0.07116, "merge": 0.14214, "skel": 0.12426}, "n_reads": 22042, "n_cand": 33205, "read_suppressed_frac": 0.0978, "dropout_reads_logged": 22042, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5012, "admitted_surface": 16860, "staged_surface": 0, "admitted_split": 16893, "staged_split": 0, "admitted_merge": 19972, "staged_merge": 0, "admitted_skel": 11811, "staged_skel": 0, "mean_count": 1666.6572265625, "n_util_obs": 761, "util_count_spearman": 0.02593057115423768}, "probe_overhead_sec_total": 630.741, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009596, "alpha": 0.03, "t": 5400.8} |
| {"event": "consolidate", "step": 5200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 5441.8} |
| {"event": "probe", "step": 5200, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3661.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.210366, "cps": 28.601562, "mean_p": 0.207961, "boundary_entropy": 0.067885, "bpb": 1.842759}, "tr": {"chunks": 3747.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.206982, "cps": 29.273438, "mean_p": 0.197957, "boundary_entropy": 0.09568, "bpb": 1.985295}, "fi": {"chunks": 5048.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.268354, "cps": 39.4375, "mean_p": 0.259368, "boundary_entropy": 0.071083, "bpb": 2.083349}}, "bpb": {"en": 1.84276, "tr": 1.98529, "fi": 2.08335}, "gini_cpc": 0.059668, "gini_cps": 0.074235, "bpic": 4.554592, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9839148900047399, "fi": 1.275650942088128}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5039, "admitted_surface": 16834, "staged_surface": 0, "admitted_split": 16930, "staged_split": 0, "admitted_merge": 19932, "staged_merge": 0, "admitted_skel": 11840, "staged_skel": 0, "mean_count": 1682.248779296875, "n_util_obs": 770, "util_count_spearman": 0.02169533658229577}, "t": 5443.9} |
| {"event": "eval", "step": 5200, "phase": 3, "bpb": {"en": 1.85908, "tr": 1.99523, "fi": 2.09291}, "bytes_per_chunk": {"en": 4.446, "tr": 5.449, "fi": 4.001}, "bpb_mean": 1.98241, "frozen_inventory": true, "t": 5447.1} |
| {"event": "train", "step": 5200, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32377, "bpb": 1.9098, "chunk_loss": 0.031101, "F": 0.23297, "G": 0.23152, "chunks": 238.56, "bytes_per_chunk": 4.292, "boundary_entropy": 0.07279, "router_noise": 0.0, "grad_norm": 0.668, "lr": 4.29e-05, "router_trainable": true, "sec_per_step": 0.975, "probe_sec": 0.1016, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06606, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.02994, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.06895, "n_merge": 1, "mean_delta_merge": -0.03702, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.55686, "trunc_abs_frac": 0.58442, "mean_delta_full": -0.10681, "read_hit_rate": {"surface": 0.6839, "split": 0.8759, "merge": 0.26328, "skel": 0.6732}, "mean_gate": {"surface": 0.16465, "split": 0.07269, "merge": 0.1413, "skel": 0.12586}, "n_reads": 22108, "n_cand": 33476, "read_suppressed_frac": 0.1001, "dropout_reads_logged": 22108, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5040, "admitted_surface": 16798, "staged_surface": 0, "admitted_split": 16940, "staged_split": 0, "admitted_merge": 19973, "staged_merge": 0, "admitted_skel": 11825, "staged_skel": 0, "mean_count": 1682.581787109375, "n_util_obs": 770, "util_count_spearman": 0.021378784846541456}, "probe_overhead_sec_total": 636.878, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.00925, "alpha": 0.03, "t": 5449.5} |
| {"event": "consolidate", "step": 5250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 5491.8} |
| {"event": "train", "step": 5250, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35755, "bpb": 1.95853, "chunk_loss": 0.030926, "F": 0.22995, "G": 0.22753, "chunks": 235.47, "bytes_per_chunk": 4.349, "boundary_entropy": 0.07368, "router_noise": 0.0, "grad_norm": 0.6623, "lr": 4.14e-05, "router_trainable": true, "sec_per_step": 0.866, "probe_sec": 0.0933, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06128, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.07274, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": -0.1225, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -0.26097, "trunc_abs_frac": 15.28132, "mean_delta_full": 0.1385, "read_hit_rate": {"surface": 0.67586, "split": 0.87205, "merge": 0.26556, "skel": 0.65961}, "mean_gate": {"surface": 0.16629, "split": 0.07341, "merge": 0.14365, "skel": 0.12735}, "n_reads": 21602, "n_cand": 32968, "read_suppressed_frac": 0.1017, "dropout_reads_logged": 21602, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5073, "admitted_surface": 16818, "staged_surface": 0, "admitted_split": 16937, "staged_split": 0, "admitted_merge": 19898, "staged_merge": 0, "admitted_skel": 11883, "staged_skel": 0, "mean_count": 1698.430419921875, "n_util_obs": 773, "util_count_spearman": 0.017747960694021382}, "probe_overhead_sec_total": 643.197, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008862, "alpha": 0.03, "t": 5492.8} |
| {"event": "consolidate", "step": 5300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 5539.2} |
| {"event": "train", "step": 5300, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.34846, "bpb": 1.94542, "chunk_loss": 0.030766, "F": 0.22183, "G": 0.21977, "chunks": 227.16, "bytes_per_chunk": 4.508, "boundary_entropy": 0.07214, "router_noise": 0.0, "grad_norm": 0.6556, "lr": 3.99e-05, "router_trainable": true, "sec_per_step": 0.949, "probe_sec": 0.1486, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.12089, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.00801, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 2, "mean_delta_split": 0.01324, "n_merge": 1, "mean_delta_merge": -0.13635, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.21831, "trunc_abs_frac": 0.25662, "mean_delta_full": 0.40042, "read_hit_rate": {"surface": 0.66381, "split": 0.86966, "merge": 0.24303, "skel": 0.66136}, "mean_gate": {"surface": 0.16368, "split": 0.07143, "merge": 0.13912, "skel": 0.12529}, "n_reads": 20797, "n_cand": 32098, "read_suppressed_frac": 0.1049, "dropout_reads_logged": 20797, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5101, "admitted_surface": 16860, "staged_surface": 0, "admitted_split": 16950, "staged_split": 0, "admitted_merge": 19861, "staged_merge": 0, "admitted_skel": 11865, "staged_skel": 0, "mean_count": 1714.46630859375, "n_util_obs": 783, "util_count_spearman": 0.035454586853823154}, "probe_overhead_sec_total": 649.68, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008866, "alpha": 0.03, "t": 5540.3} |
| {"event": "consolidate", "step": 5350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 5590.5} |
| {"event": "train", "step": 5350, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.31641, "bpb": 1.89918, "chunk_loss": 0.030966, "F": 0.22244, "G": 0.22278, "chunks": 227.78, "bytes_per_chunk": 4.496, "boundary_entropy": 0.07259, "router_noise": 0.0, "grad_norm": 0.7172, "lr": 3.86e-05, "router_trainable": true, "sec_per_step": 1.031, "probe_sec": 0.1109, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07048, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.00651, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.13397, "n_merge": 1, "mean_delta_merge": 0.21167, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": -0.76598, "trunc_abs_frac": 0.76598, "mean_delta_full": 0.06494, "read_hit_rate": {"surface": 0.66171, "split": 0.86453, "merge": 0.25744, "skel": 0.64113}, "mean_gate": {"surface": 0.16481, "split": 0.07018, "merge": 0.14514, "skel": 0.12387}, "n_reads": 20460, "n_cand": 31808, "read_suppressed_frac": 0.099, "dropout_reads_logged": 20460, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5134, "admitted_surface": 16875, "staged_surface": 0, "admitted_split": 16925, "staged_split": 0, "admitted_merge": 19829, "staged_merge": 0, "admitted_skel": 11907, "staged_skel": 0, "mean_count": 1730.39892578125, "n_util_obs": 789, "util_count_spearman": 0.03379470435483522}, "probe_overhead_sec_total": 656.015, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008217, "alpha": 0.03, "t": 5591.9} |
| {"event": "consolidate", "step": 5400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 5653.3} |
| {"event": "probe", "step": 5400, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3663.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.210481, "cps": 28.617188, "mean_p": 0.20813, "boundary_entropy": 0.065854, "bpb": 1.839508}, "tr": {"chunks": 3807.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.210297, "cps": 29.742188, "mean_p": 0.197252, "boundary_entropy": 0.093729, "bpb": 1.97569}, "fi": {"chunks": 5057.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.268832, "cps": 39.507812, "mean_p": 0.259388, "boundary_entropy": 0.069212, "bpb": 2.076547}}, "bpb": {"en": 1.83951, "tr": 1.97569, "fi": 2.07655}, "gini_cpc": 0.056588, "gini_cps": 0.074186, "bpic": 4.528778, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9991243120006309, "fi": 1.2772275322378985}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5169, "admitted_surface": 16850, "staged_surface": 0, "admitted_split": 16883, "staged_split": 0, "admitted_merge": 20037, "staged_merge": 0, "admitted_skel": 11766, "staged_skel": 0, "mean_count": 1745.9146728515625, "n_util_obs": 796, "util_count_spearman": 0.029700300220362737}, "t": 5655.2} |
| {"event": "eval", "step": 5400, "phase": 3, "bpb": {"en": 1.85469, "tr": 1.98569, "fi": 2.08497}, "bytes_per_chunk": {"en": 4.445, "tr": 5.386, "fi": 4.003}, "bpb_mean": 1.97512, "frozen_inventory": true, "t": 5658.8} |
| {"event": "train", "step": 5400, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.2363, "bpb": 1.78361, "chunk_loss": 0.030719, "F": 0.21786, "G": 0.21682, "chunks": 223.09, "bytes_per_chunk": 4.59, "boundary_entropy": 0.07161, "router_noise": 0.0, "grad_norm": 0.6423, "lr": 3.73e-05, "router_trainable": true, "sec_per_step": 1.392, "probe_sec": 0.1149, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08414, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.00406, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.16541, "n_merge": 1, "mean_delta_merge": -0.49756, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.84805, "trunc_abs_frac": 0.32176, "mean_delta_full": -0.20308, "read_hit_rate": {"surface": 0.68425, "split": 0.87375, "merge": 0.26473, "skel": 0.67301}, "mean_gate": {"surface": 0.16262, "split": 0.0713, "merge": 0.14028, "skel": 0.12552}, "n_reads": 20799, "n_cand": 31484, "read_suppressed_frac": 0.1008, "dropout_reads_logged": 20799, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5171, "admitted_surface": 16811, "staged_surface": 0, "admitted_split": 16951, "staged_split": 0, "admitted_merge": 19920, "staged_merge": 0, "admitted_skel": 11854, "staged_skel": 0, "mean_count": 1746.237548828125, "n_util_obs": 796, "util_count_spearman": 0.029697064243896464}, "probe_overhead_sec_total": 662.473, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008881, "alpha": 0.03, "t": 5661.5} |
| {"event": "consolidate", "step": 5450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 5711.6} |
| {"event": "train", "step": 5450, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.28895, "bpb": 1.85956, "chunk_loss": 0.030799, "F": 0.22562, "G": 0.22394, "chunks": 231.03, "bytes_per_chunk": 4.432, "boundary_entropy": 0.06874, "router_noise": 0.0, "grad_norm": 0.6508, "lr": 3.62e-05, "router_trainable": true, "sec_per_step": 1.016, "probe_sec": 0.0928, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06449, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": -0.00215, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.11646, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -1.17602, "trunc_abs_frac": 0.19432, "mean_delta_full": -0.0198, "read_hit_rate": {"surface": 0.67708, "split": 0.88116, "merge": 0.23932, "skel": 0.68401}, "mean_gate": {"surface": 0.16074, "split": 0.07, "merge": 0.14055, "skel": 0.12244}, "n_reads": 21663, "n_cand": 32862, "read_suppressed_frac": 0.0991, "dropout_reads_logged": 21663, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5207, "admitted_surface": 16846, "staged_surface": 0, "admitted_split": 16872, "staged_split": 0, "admitted_merge": 20041, "staged_merge": 0, "admitted_skel": 11777, "staged_skel": 0, "mean_count": 1762.2611083984375, "n_util_obs": 801, "util_count_spearman": 0.02836664224698836}, "probe_overhead_sec_total": 668.88, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009724, "alpha": 0.03, "t": 5712.3} |
| {"event": "consolidate", "step": 5500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 5753.9} |
| {"event": "train", "step": 5500, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.30646, "bpb": 1.88482, "chunk_loss": 0.030776, "F": 0.2233, "G": 0.22138, "chunks": 228.66, "bytes_per_chunk": 4.478, "boundary_entropy": 0.07269, "router_noise": 0.0, "grad_norm": 0.6721, "lr": 3.51e-05, "router_trainable": true, "sec_per_step": 0.846, "probe_sec": 0.1308, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10003, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": 0.02243, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.04492, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -0.93578, "trunc_abs_frac": 1.04408, "mean_delta_full": -0.18051, "read_hit_rate": {"surface": 0.68373, "split": 0.87221, "merge": 0.26403, "skel": 0.65939}, "mean_gate": {"surface": 0.16365, "split": 0.07149, "merge": 0.13955, "skel": 0.12546}, "n_reads": 21097, "n_cand": 32117, "read_suppressed_frac": 0.096, "dropout_reads_logged": 21097, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5229, "admitted_surface": 16795, "staged_surface": 0, "admitted_split": 16953, "staged_split": 0, "admitted_merge": 19904, "staged_merge": 0, "admitted_skel": 11884, "staged_skel": 0, "mean_count": 1778.129150390625, "n_util_obs": 805, "util_count_spearman": 0.02815932069320253}, "probe_overhead_sec_total": 674.954, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009747, "alpha": 0.03, "t": 5754.6} |
| {"event": "consolidate", "step": 5550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 5795.8} |
| {"event": "train", "step": 5550, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.27695, "bpb": 1.84225, "chunk_loss": 0.030906, "F": 0.23099, "G": 0.22787, "chunks": 236.53, "bytes_per_chunk": 4.329, "boundary_entropy": 0.06656, "router_noise": 0.0, "grad_norm": 0.6409, "lr": 3.41e-05, "router_trainable": true, "sec_per_step": 0.838, "probe_sec": 0.1063, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07924, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.10801, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.17822, "n_merge": 1, "mean_delta_merge": 0.15961, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.96767, "trunc_abs_frac": 0.11532, "mean_delta_full": -0.00354, "read_hit_rate": {"surface": 0.68542, "split": 0.88617, "merge": 0.27129, "skel": 0.68938}, "mean_gate": {"surface": 0.15864, "split": 0.06859, "merge": 0.13768, "skel": 0.12123}, "n_reads": 22349, "n_cand": 33360, "read_suppressed_frac": 0.1025, "dropout_reads_logged": 22349, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5262, "admitted_surface": 16859, "staged_surface": 0, "admitted_split": 16850, "staged_split": 0, "admitted_merge": 20026, "staged_merge": 0, "admitted_skel": 11801, "staged_skel": 0, "mean_count": 1793.929931640625, "n_util_obs": 812, "util_count_spearman": 0.02213273559395101}, "probe_overhead_sec_total": 680.588, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.00998, "alpha": 0.03, "t": 5796.5} |
| {"event": "consolidate", "step": 5600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 5844.4} |
| {"event": "probe", "step": 5600, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3655.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.210021, "cps": 28.554688, "mean_p": 0.207269, "boundary_entropy": 0.067255, "bpb": 1.834542}, "tr": {"chunks": 3780.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.208805, "cps": 29.53125, "mean_p": 0.197198, "boundary_entropy": 0.092691, "bpb": 1.970335}, "fi": {"chunks": 5050.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.26846, "cps": 39.453125, "mean_p": 0.259116, "boundary_entropy": 0.068366, "bpb": 2.065995}}, "bpb": {"en": 1.83454, "tr": 1.97033, "fi": 2.066}, "gini_cpc": 0.057865, "gini_cps": 0.074489, "bpic": 4.544013, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9942096800849192, "fi": 1.2782512720494696}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5286, "admitted_surface": 16850, "staged_surface": 0, "admitted_split": 16846, "staged_split": 0, "admitted_merge": 20090, "staged_merge": 0, "admitted_skel": 11750, "staged_skel": 0, "mean_count": 1809.593505859375, "n_util_obs": 817, "util_count_spearman": 0.023044503065859086}, "t": 5846.2} |
| {"event": "eval", "step": 5600, "phase": 3, "bpb": {"en": 1.85116, "tr": 1.98027, "fi": 2.07592}, "bytes_per_chunk": {"en": 4.447, "tr": 5.42, "fi": 4.001}, "bpb_mean": 1.96912, "frozen_inventory": true, "t": 5849.3} |
| {"event": "train", "step": 5600, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.30946, "bpb": 1.88915, "chunk_loss": 0.030892, "F": 0.22717, "G": 0.22566, "chunks": 232.62, "bytes_per_chunk": 4.402, "boundary_entropy": 0.07257, "router_noise": 0.0, "grad_norm": 0.6753, "lr": 3.33e-05, "router_trainable": true, "sec_per_step": 1.101, "probe_sec": 0.0897, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05904, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.04747, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.16661, "n_merge": 1, "mean_delta_merge": -0.20459, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 6.94944, "trunc_abs_frac": 3.70318, "mean_delta_full": 0.01201, "read_hit_rate": {"surface": 0.68295, "split": 0.86989, "merge": 0.25542, "skel": 0.6745}, "mean_gate": {"surface": 0.16404, "split": 0.07149, "merge": 0.13946, "skel": 0.12411}, "n_reads": 21416, "n_cand": 32630, "read_suppressed_frac": 0.0949, "dropout_reads_logged": 21416, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5287, "admitted_surface": 16821, "staged_surface": 0, "admitted_split": 16988, "staged_split": 0, "admitted_merge": 19931, "staged_merge": 0, "admitted_skel": 11796, "staged_skel": 0, "mean_count": 1809.9267578125, "n_util_obs": 817, "util_count_spearman": 0.023120848294934465}, "probe_overhead_sec_total": 686.623, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009553, "alpha": 0.03, "t": 5851.5} |
| {"event": "consolidate", "step": 5650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 5902.4} |
| {"event": "train", "step": 5650, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.29372, "bpb": 1.86644, "chunk_loss": 0.030798, "F": 0.22394, "G": 0.22232, "chunks": 229.31, "bytes_per_chunk": 4.466, "boundary_entropy": 0.07064, "router_noise": 0.0, "grad_norm": 0.6202, "lr": 3.25e-05, "router_trainable": true, "sec_per_step": 1.061, "probe_sec": 0.1401, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06463, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.19038, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.17184, "n_merge": 2, "mean_delta_merge": -0.08026, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.86112, "trunc_abs_frac": 1.25488, "mean_delta_full": 0.1522, "read_hit_rate": {"surface": 0.68177, "split": 0.87348, "merge": 0.25186, "skel": 0.67678}, "mean_gate": {"surface": 0.16077, "split": 0.06923, "merge": 0.13883, "skel": 0.12282}, "n_reads": 21209, "n_cand": 32263, "read_suppressed_frac": 0.0978, "dropout_reads_logged": 21209, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5311, "admitted_surface": 16825, "staged_surface": 0, "admitted_split": 16942, "staged_split": 0, "admitted_merge": 19978, "staged_merge": 0, "admitted_skel": 11791, "staged_skel": 0, "mean_count": 1825.79345703125, "n_util_obs": 824, "util_count_spearman": 0.03963869810122908}, "probe_overhead_sec_total": 692.487, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009853, "alpha": 0.03, "t": 5904.6} |
| {"event": "consolidate", "step": 5700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 5966.9} |
| {"event": "train", "step": 5700, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.37366, "bpb": 1.98178, "chunk_loss": 0.030793, "F": 0.22357, "G": 0.22221, "chunks": 228.94, "bytes_per_chunk": 4.473, "boundary_entropy": 0.07232, "router_noise": 0.0, "grad_norm": 0.6739, "lr": 3.18e-05, "router_trainable": true, "sec_per_step": 1.281, "probe_sec": 0.1333, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08778, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.02521, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.02087, "n_merge": 1, "mean_delta_merge": -0.41406, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.339, "trunc_abs_frac": 0.34666, "mean_delta_full": -0.25504, "read_hit_rate": {"surface": 0.66726, "split": 0.86805, "merge": 0.24497, "skel": 0.66434}, "mean_gate": {"surface": 0.16136, "split": 0.06934, "merge": 0.14125, "skel": 0.12277}, "n_reads": 21031, "n_cand": 32405, "read_suppressed_frac": 0.0978, "dropout_reads_logged": 21031, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5348, "admitted_surface": 16853, "staged_surface": 0, "admitted_split": 16948, "staged_split": 0, "admitted_merge": 19890, "staged_merge": 0, "admitted_skel": 11845, "staged_skel": 0, "mean_count": 1841.73095703125, "n_util_obs": 831, "util_count_spearman": 0.03424529922257674}, "probe_overhead_sec_total": 699.19, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008912, "alpha": 0.03, "t": 5968.7} |
| {"event": "consolidate", "step": 5750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 6014.6} |
| {"event": "train", "step": 5750, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35486, "bpb": 1.95465, "chunk_loss": 0.030831, "F": 0.22375, "G": 0.2234, "chunks": 229.12, "bytes_per_chunk": 4.469, "boundary_entropy": 0.07124, "router_noise": 0.0, "grad_norm": 0.7069, "lr": 3.13e-05, "router_trainable": true, "sec_per_step": 0.941, "probe_sec": 0.0958, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06228, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": 0.03224, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.11487, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 0.36046, "trunc_abs_frac": 0.63401, "mean_delta_full": 0.25334, "read_hit_rate": {"surface": 0.65877, "split": 0.86799, "merge": 0.23817, "skel": 0.64544}, "mean_gate": {"surface": 0.16108, "split": 0.0698, "merge": 0.1409, "skel": 0.12341}, "n_reads": 20607, "n_cand": 32131, "read_suppressed_frac": 0.0991, "dropout_reads_logged": 20607, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5375, "admitted_surface": 16864, "staged_surface": 0, "admitted_split": 16932, "staged_split": 0, "admitted_merge": 19883, "staged_merge": 0, "admitted_skel": 11857, "staged_skel": 0, "mean_count": 1857.686767578125, "n_util_obs": 836, "util_count_spearman": 0.038695209427680934}, "probe_overhead_sec_total": 704.702, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.01005, "alpha": 0.03, "t": 6015.7} |
| {"event": "consolidate", "step": 5800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 6055.4} |
| {"event": "probe", "step": 5800, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3653.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.209906, "cps": 28.539062, "mean_p": 0.207653, "boundary_entropy": 0.066765, "bpb": 1.831013}, "tr": {"chunks": 3731.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.206098, "cps": 29.148438, "mean_p": 0.196995, "boundary_entropy": 0.093662, "bpb": 1.967228}, "fi": {"chunks": 5045.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.268194, "cps": 39.414062, "mean_p": 0.258754, "boundary_entropy": 0.068813, "bpb": 2.061012}}, "bpb": {"en": 1.83101, "tr": 1.96723, "fi": 2.06101}, "gini_cpc": 0.060505, "gini_cps": 0.074664, "bpic": 4.564486, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.981859045796267, "fi": 1.2776848202711126}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5400, "admitted_surface": 16861, "staged_surface": 0, "admitted_split": 16923, "staged_split": 0, "admitted_merge": 19953, "staged_merge": 0, "admitted_skel": 11799, "staged_skel": 0, "mean_count": 1873.3035888671875, "n_util_obs": 845, "util_count_spearman": 0.04467347382045297}, "t": 6057.2} |
| {"event": "eval", "step": 5800, "phase": 3, "bpb": {"en": 1.84684, "tr": 1.97616, "fi": 2.07124}, "bytes_per_chunk": {"en": 4.446, "tr": 5.471, "fi": 4.012}, "bpb_mean": 1.96475, "frozen_inventory": true, "t": 6060.2} |
| {"event": "train", "step": 5800, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.27149, "bpb": 1.83437, "chunk_loss": 0.03072, "F": 0.22226, "G": 0.21979, "chunks": 227.59, "bytes_per_chunk": 4.499, "boundary_entropy": 0.07117, "router_noise": 0.0, "grad_norm": 0.7042, "lr": 3.08e-05, "router_trainable": true, "sec_per_step": 0.924, "probe_sec": 0.0813, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05579, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.0435, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.04333, "n_merge": 1, "mean_delta_merge": 0.27081, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 1.05177, "trunc_abs_frac": 1.05177, "mean_delta_full": 0.39026, "read_hit_rate": {"surface": 0.68211, "split": 0.87869, "merge": 0.25156, "skel": 0.67945}, "mean_gate": {"surface": 0.15702, "split": 0.06817, "merge": 0.13229, "skel": 0.12015}, "n_reads": 21349, "n_cand": 32301, "read_suppressed_frac": 0.1012, "dropout_reads_logged": 21349, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5400, "admitted_surface": 16826, "staged_surface": 0, "admitted_split": 16926, "staged_split": 0, "admitted_merge": 19987, "staged_merge": 0, "admitted_skel": 11797, "staged_skel": 0, "mean_count": 1873.6376953125, "n_util_obs": 845, "util_count_spearman": 0.04484430924293549}, "probe_overhead_sec_total": 709.976, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009877, "alpha": 0.03, "t": 6061.9} |
| {"event": "consolidate", "step": 5850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 6102.0} |
| {"event": "train", "step": 5850, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32767, "bpb": 1.91542, "chunk_loss": 0.030634, "F": 0.21664, "G": 0.21573, "chunks": 221.84, "bytes_per_chunk": 4.616, "boundary_entropy": 0.0689, "router_noise": 0.0, "grad_norm": 0.6693, "lr": 3.05e-05, "router_trainable": true, "sec_per_step": 0.823, "probe_sec": 0.0966, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06935, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.04372, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.36755, "n_merge": 2, "mean_delta_merge": -0.13794, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.05202, "trunc_abs_frac": 1.32165, "mean_delta_full": -0.21863, "read_hit_rate": {"surface": 0.66209, "split": 0.86005, "merge": 0.24549, "skel": 0.65314}, "mean_gate": {"surface": 0.15496, "split": 0.06656, "merge": 0.13448, "skel": 0.11735}, "n_reads": 20135, "n_cand": 31328, "read_suppressed_frac": 0.1016, "dropout_reads_logged": 20135, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5434, "admitted_surface": 16850, "staged_surface": 0, "admitted_split": 17041, "staged_split": 0, "admitted_merge": 19767, "staged_merge": 0, "admitted_skel": 11878, "staged_skel": 0, "mean_count": 1889.6173095703125, "n_util_obs": 848, "util_count_spearman": 0.04325293314765783}, "probe_overhead_sec_total": 715.678, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009205, "alpha": 0.03, "t": 6103.1} |
| {"event": "consolidate", "step": 5900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 6144.7} |
| {"event": "train", "step": 5900, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35568, "bpb": 1.95584, "chunk_loss": 0.030849, "F": 0.22516, "G": 0.22497, "chunks": 230.56, "bytes_per_chunk": 4.441, "boundary_entropy": 0.07039, "router_noise": 0.0, "grad_norm": 0.6948, "lr": 3.02e-05, "router_trainable": true, "sec_per_step": 0.86, "probe_sec": 0.1827, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.14657, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.13772, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": 0.15283, "n_merge": 3, "mean_delta_merge": 0.4466, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -4.99345, "trunc_abs_frac": 6.72635, "mean_delta_full": -0.00854, "read_hit_rate": {"surface": 0.6654, "split": 0.87191, "merge": 0.23681, "skel": 0.65846}, "mean_gate": {"surface": 0.15947, "split": 0.06755, "merge": 0.14004, "skel": 0.12121}, "n_reads": 21149, "n_cand": 32666, "read_suppressed_frac": 0.0992, "dropout_reads_logged": 21149, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5475, "admitted_surface": 16821, "staged_surface": 0, "admitted_split": 16899, "staged_split": 0, "admitted_merge": 19955, "staged_merge": 0, "admitted_skel": 11861, "staged_skel": 0, "mean_count": 1905.5506591796875, "n_util_obs": 852, "util_count_spearman": 0.04220179424123567}, "probe_overhead_sec_total": 721.567, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008147, "alpha": 0.03, "t": 6146.0} |
| {"event": "consolidate", "step": 5950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 6198.2} |
| {"event": "train", "step": 5950, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.28545, "bpb": 1.85451, "chunk_loss": 0.030702, "F": 0.2189, "G": 0.21802, "chunks": 224.16, "bytes_per_chunk": 4.568, "boundary_entropy": 0.07124, "router_noise": 0.0, "grad_norm": 0.7276, "lr": 3.01e-05, "router_trainable": true, "sec_per_step": 1.07, "probe_sec": 0.1073, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05319, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 12, "mean_delta_loo": -0.06639, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.18069, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 1.7681, "trunc_abs_frac": 1.86992, "mean_delta_full": -0.03961, "read_hit_rate": {"surface": 0.67133, "split": 0.87254, "merge": 0.24082, "skel": 0.66205}, "mean_gate": {"surface": 0.15998, "split": 0.06935, "merge": 0.13596, "skel": 0.1223}, "n_reads": 20792, "n_cand": 31912, "read_suppressed_frac": 0.0981, "dropout_reads_logged": 20792, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5509, "admitted_surface": 16836, "staged_surface": 0, "admitted_split": 16935, "staged_split": 0, "admitted_merge": 19887, "staged_merge": 0, "admitted_skel": 11878, "staged_skel": 0, "mean_count": 1921.489013671875, "n_util_obs": 858, "util_count_spearman": 0.046145529108984645}, "probe_overhead_sec_total": 728.201, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008146, "alpha": 0.03, "t": 6199.5} |
| {"event": "consolidate", "step": 6000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 6263.6} |
| {"event": "train", "step": 5999, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.31063, "bpb": 1.89083, "chunk_loss": 0.030935, "F": 0.22745, "G": 0.22619, "chunks": 232.91, "bytes_per_chunk": 4.397, "boundary_entropy": 0.07188, "router_noise": 0.0, "grad_norm": 0.6909, "lr": 3e-05, "router_trainable": true, "sec_per_step": 1.282, "probe_sec": 0.1117, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08256, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.05671, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 1, "mean_delta_split": -0.11548, "n_merge": 2, "mean_delta_merge": 0.08923, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.41149, "trunc_abs_frac": 0.43751, "mean_delta_full": 0.09719, "read_hit_rate": {"surface": 0.68158, "split": 0.87643, "merge": 0.24716, "skel": 0.6848}, "mean_gate": {"surface": 0.15634, "split": 0.06793, "merge": 0.13328, "skel": 0.12078}, "n_reads": 21930, "n_cand": 33198, "read_suppressed_frac": 0.0994, "dropout_reads_logged": 21930, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5540, "admitted_surface": 16810, "staged_surface": 0, "admitted_split": 16925, "staged_split": 0, "admitted_merge": 19981, "staged_merge": 0, "admitted_skel": 11820, "staged_skel": 0, "mean_count": 1937.1162109375, "n_util_obs": 864, "util_count_spearman": 0.04498700863864923}, "probe_overhead_sec_total": 734.98, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.00864, "alpha": 0.03, "t": 6263.7} |
| {"event": "probe", "step": 6000, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3659.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.210251, "cps": 28.585938, "mean_p": 0.207688, "boundary_entropy": 0.065938, "bpb": 1.829139}, "tr": {"chunks": 3660.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.202176, "cps": 28.59375, "mean_p": 0.196182, "boundary_entropy": 0.093753, "bpb": 1.967329}, "fi": {"chunks": 5043.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.268088, "cps": 39.398438, "mean_p": 0.258827, "boundary_entropy": 0.067734, "bpb": 2.056518}}, "bpb": {"en": 1.82914, "tr": 1.96733, "fi": 2.05652}, "gini_cpc": 0.06457, "gini_cps": 0.074637, "bpic": 4.589225, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.961595106752791, "fi": 1.275083997839355}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5540, "admitted_surface": 16810, "staged_surface": 0, "admitted_split": 16925, "staged_split": 0, "admitted_merge": 19981, "staged_merge": 0, "admitted_skel": 11820, "staged_skel": 0, "mean_count": 1937.1162109375, "n_util_obs": 864, "util_count_spearman": 0.04498700863864923}, "t": 6265.9} |
| {"event": "eval", "step": 6000, "phase": 3, "bpb": {"en": 1.84524, "tr": 1.97627, "fi": 2.06556}, "bytes_per_chunk": {"en": 4.449, "tr": 5.603, "fi": 4.015}, "bpb_mean": 1.96236, "frozen_inventory": true, "t": 6269.8} |
| {"event": "reps_dumped", "dirs": ["/local/scratch/sas245/beetle-hnet/runs/ali_mem-mixed_s0/probe/reps/th0.30", "/local/scratch/sas245/beetle-hnet/runs/ali_mem-mixed_s0/probe/reps/th0.40", "/local/scratch/sas245/beetle-hnet/runs/ali_mem-mixed_s0/probe/reps/th0.50", "/local/scratch/sas245/beetle-hnet/runs/ali_mem-mixed_s0/probe/reps/th0.60", "/local/scratch/sas245/beetle-hnet/runs/ali_mem-mixed_s0/probe/reps/th0.70"], "t": 6273.0} |
| {"event": "final", "run": "ali_mem-mixed_s0", "arch": "ali", "objective": "ratio", "condition": "A", "step": 6000, "params": 22785088, "seed": 0, "per_lang": {"en": {"cpc": 0.21025110613112682, "cps": 28.5859375, "bpb": 1.8493486207933525, "bytes_per_chunk": 4.409449878470324, "reads_per_chunk": 2.9068889243992904}, "tr": {"cpc": 0.2021764348450533, "cps": 28.59375, "bpb": 1.969015514375986, "bytes_per_chunk": 5.601128156916371, "reads_per_chunk": 2.5759689756847997}, "fi": {"cpc": 0.26808782095582373, "cps": 39.3984375, "bpb": 2.083401035728936, "bytes_per_chunk": 3.9990846821558788, "reads_per_chunk": 2.9634559350734544}}, "bpb_mean": 1.9672550569660914, "gini_cpc": 0.06457006929145193, "gini_cps": 0.07463732945046653, "bpic": 4.589225044491183, "mean_boundary_entropy": 0.0718768909573555, "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": 734.9796347618103, "probe_overhead_seq_frac": 0.4999166666666667, "wall_sec": 6280.56084394455, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 5540, "admitted_surface": 16810, "staged_surface": 0, "admitted_split": 16925, "staged_split": 0, "admitted_merge": 19981, "staged_merge": 0, "admitted_skel": 11820, "staged_skel": 0, "mean_count": 1937.1162109375, "n_util_obs": 864, "util_count_spearman": 0.04498700863864923}, "ali_params": 268096, "ali_config": {"enabled": true, "budget_B": 65536, "staging_capacity": 262144, "policy": "frequency", "value_mode": "learned", "candidate_types": ["surface", "split", "merge", "skel"], "gate": "vector", "retrieval": true, "memory_scope": "persistent", "read_dropout": 0.1, "write_rule": "mean", "eta_min": 0.0, "consolidate_every": 50, "probe_seqs": 2, "probe_reps": 8, "horizon": 32, "predictor": "mlp", "lambda_util": 0.1, "max_span": 64, "byte_classes_path": "/local/scratch/sas245/beetle-hnet/scratch/data/byte_classes.json", "gate_hidden": 0, "type_dim": 16, "gate_bias_init": -2.0}, "util_vs_frequency_spearman": 0.04498700863864923, "paired_vs_dropout": {"n": 1895, "pearson": -0.01470751455807266, "spearman": 0.012313629004189811, "mean_abs_discrepancy": 0.3884356617927551, "mean_paired": 0.017228469252586365, "mean_dropout": 0.0065397159196436405}, "predictor_calibration": {"n": 4750, "all": {"n": 4750, "pearson": 0.07517200182486088, "spearman": 0.07652358225993569, "mae": 0.08385143429040909, "reliability": [{"pred": -0.0005088937468826771, "obs": 0.0036706055980175734, "n": 475}, {"pred": 0.005684825591742992, "obs": 0.005425000563263893, "n": 475}, {"pred": 0.007983285933732986, "obs": 0.009757690131664276, "n": 475}, {"pred": 0.009891621768474579, "obs": 0.011957990005612373, "n": 475}, {"pred": 0.011289387941360474, "obs": 0.008411113172769547, "n": 475}, {"pred": 0.013064890168607235, "obs": 0.029578158631920815, "n": 475}, {"pred": 0.014604332856833935, "obs": 0.018993722274899483, "n": 475}, {"pred": 0.016015520319342613, "obs": 0.009690395556390285, "n": 475}, {"pred": 0.018993698060512543, "obs": 0.030660323798656464, "n": 475}, {"pred": 0.03003532998263836, "obs": 0.03165699169039726, "n": 475}]}, "admitted": {"n": 4750, "pearson": 0.07517200182486088, "spearman": 0.07652358225993569, "mae": 0.08385143429040909, "reliability": [{"pred": -0.0005088937468826771, "obs": 0.0036706055980175734, "n": 475}, {"pred": 0.005684825591742992, "obs": 0.005425000563263893, "n": 475}, {"pred": 0.007983285933732986, "obs": 0.009757690131664276, "n": 475}, {"pred": 0.009891621768474579, "obs": 0.011957990005612373, "n": 475}, {"pred": 0.011289387941360474, "obs": 0.008411113172769547, "n": 475}, {"pred": 0.013064890168607235, "obs": 0.029578158631920815, "n": 475}, {"pred": 0.014604332856833935, "obs": 0.018993722274899483, "n": 475}, {"pred": 0.016015520319342613, "obs": 0.009690395556390285, "n": 475}, {"pred": 0.018993698060512543, "obs": 0.030660323798656464, "n": 475}, {"pred": 0.03003532998263836, "obs": 0.03165699169039726, "n": 475}]}, "staged": {"n": 0, "pearson": NaN, "spearman": NaN, "mae": NaN}}, "t": 6280.6} |
|
|
