| {"event": "hub", "repo": "AdaptiveChunking/ali_lexmem-v1_s2", "keep_steps": [0, 1, 2, 4, 6, 8, 16, 32, 64, 128, 256, 512, 1000, 1024, 2048, 3000, 4096, 6000], "t": 3.1} |
| {"event": "start", "run": "ali_lexmem-v1_s2", "arch": "lexmem_v1", "objective": "ratio", "condition": "A", "params": 22719553, "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": 202561, "ali_config": {"enabled": true, "budget_B": 65536, "staging_capacity": 262144, "policy": "frequency", "candidate_types": ["surface"], "gate": "scalar", "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": 3.1} |
| {"event": "probe_schedule", "n": 45, "steps": [0, 1, 2, 4, 6, 8, 16, 32, 64, 128, 200, 256, 400, 512, 600, 800, 1000, 1024, 1200, 1400, 1600, 1800, 2000, 2048, 2200, 2400, 2600, 2800, 3000, 3200, 3400, 3600, 3800, 4000, 4096, 4200, 4400, 4600, 4800, 5000, 5200, 5400, 5600, 5800, 6000], "t": 3.1} |
| {"event": "probe", "step": 0, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 10933.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.628225, "cps": 85.414062, "mean_p": 0.513249, "boundary_entropy": 0.686295, "bpb": 8.12598}, "tr": {"chunks": 8187.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.452245, "cps": 63.960938, "mean_p": 0.498002, "boundary_entropy": 0.686826, "bpb": 8.097536}, "fi": {"chunks": 13540.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.719792, "cps": 105.78125, "mean_p": 0.517169, "boundary_entropy": 0.686661, "bpb": 8.065117}}, "bpb": {"en": 8.12598, "tr": 8.09754, "fi": 8.06512}, "gini_cpc": 0.099077, "gini_cps": 0.109267, "bpic": 1.737048, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.7198781817505053, "fi": 1.1457544508645294}, "inventory": {"size": 0, "admitted": 0, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 0.0, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 4.9} |
| {"event": "train", "step": 0, "arch": "lexmem_v1", "phase": 1, "objective": "global", "lm_loss": 5.62735, "bpb": 8.11854, "chunk_loss": 0.062086, "F": 0.59833, "G": 0.50836, "chunks": 612.69, "bytes_per_chunk": 1.671, "boundary_entropy": 0.69105, "router_noise": 0.0, "grad_norm": 11.4934, "lr": 1e-06, "router_trainable": true, "sec_per_step": 0.03, "probe_sec": 0.0408, "probe_extra_seqs": 0, "n_rows": 0, "extra_fwd_seqs": 0, "wall_s": 0.0, "overhead_frac_est": 0.0, "chunk_mismatch": 0, "n_loo": 0, "mean_delta_loo": NaN, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "read_hit_rate": {"surface": 0.0}, "mean_gate": {"surface": null}, "n_reads": 0, "n_cand": 19574, "read_suppressed_frac": 0.0, "dropout_reads_logged": 0, "inventory": {"size": 2006, "admitted": 2006, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 2006, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 9.757726669311523, "n_util_obs": 0, "util_count_spearman": NaN}, "probe_overhead_sec_total": 0.041, "probe_overhead_seq_frac": 0.0, "util_loss": 0.0, "alpha": 0.03, "t": 7.0} |
| {"event": "probe", "step": 1, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 10909.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.626846, "cps": 85.226562, "mean_p": 0.513143, "boundary_entropy": 0.686304, "bpb": 8.084484}, "tr": {"chunks": 8205.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.45324, "cps": 64.101562, "mean_p": 0.498082, "boundary_entropy": 0.686828, "bpb": 8.066492}, "fi": {"chunks": 13504.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.717878, "cps": 105.5, "mean_p": 0.517038, "boundary_entropy": 0.686672, "bpb": 8.02165}}, "bpb": {"en": 8.08448, "tr": 8.06649, "fi": 8.02165}, "gini_cpc": 0.098125, "gini_cps": 0.108304, "bpic": 1.739285, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.7230481385868147, "fi": 1.1452221106325746}, "inventory": {"size": 2006, "admitted": 2006, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 2006, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 9.757726669311523, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 8.1} |
| {"event": "probe", "step": 2, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 10879.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.625122, "cps": 84.992188, "mean_p": 0.512981, "boundary_entropy": 0.686318, "bpb": 8.021794}, "tr": {"chunks": 8245.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.455449, "cps": 64.414062, "mean_p": 0.498242, "boundary_entropy": 0.686832, "bpb": 8.020366}, "fi": {"chunks": 13468.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.715964, "cps": 105.21875, "mean_p": 0.516802, "boundary_entropy": 0.686692, "bpb": 7.962109}}, "bpb": {"en": 8.02179, "tr": 8.02037, "fi": 7.96211}, "gini_cpc": 0.096673, "gini_cps": 0.106836, "bpic": 1.740673, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.7285766558394645, "fi": 1.1453187423917284}, "inventory": {"size": 3532, "admitted": 3532, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 3532, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 10.465458869934082, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 10.8} |
| {"event": "probe", "step": 4, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 10716.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.615756, "cps": 83.71875, "mean_p": 0.512097, "boundary_entropy": 0.68638, "bpb": 7.813735}, "tr": {"chunks": 8372.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.462465, "cps": 65.40625, "mean_p": 0.498616, "boundary_entropy": 0.686845, "bpb": 7.866858}, "fi": {"chunks": 13211.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.702302, "cps": 103.210938, "mean_p": 0.515667, "boundary_entropy": 0.686776, "bpb": 7.762904}}, "bpb": {"en": 7.81374, "tr": 7.86686, "fi": 7.7629}, "gini_cpc": 0.0898, "gini_cps": 0.099879, "bpic": 1.756463, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.7510521323824557, "fi": 1.140552351876743}, "inventory": {"size": 5985, "admitted": 5985, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 5985, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 12.37710952758789, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 14.3} |
| {"event": "probe", "step": 6, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 10309.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.592369, "cps": 80.539062, "mean_p": 0.510305, "boundary_entropy": 0.686485, "bpb": 7.540577}, "tr": {"chunks": 8481.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.468486, "cps": 66.257812, "mean_p": 0.499046, "boundary_entropy": 0.686868, "bpb": 7.663901}, "fi": {"chunks": 12708.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.675562, "cps": 99.28125, "mean_p": 0.513577, "boundary_entropy": 0.686911, "bpb": 7.511057}}, "bpb": {"en": 7.54058, "tr": 7.6639, "fi": 7.51106}, "gini_cpc": 0.079503, "gini_cps": 0.089466, "bpic": 1.80113, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.7908681617594306, "fi": 1.1404412128366739}, "inventory": {"size": 7733, "admitted": 7733, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 7733, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 14.395318031311035, "n_util_obs": 2, "util_count_spearman": NaN}, "t": 16.7} |
| {"event": "probe", "step": 8, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 9611.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.552261, "cps": 75.085938, "mean_p": 0.507191, "boundary_entropy": 0.686618, "bpb": 7.290816}, "tr": {"chunks": 8561.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.472905, "cps": 66.882812, "mean_p": 0.499325, "boundary_entropy": 0.686901, "bpb": 7.467938}, "fi": {"chunks": 11825.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.628622, "cps": 92.382812, "mean_p": 0.51013, "boundary_entropy": 0.687087, "bpb": 7.288916}}, "bpb": {"en": 7.29082, "tr": 7.46794, "fi": 7.28892}, "gini_cpc": 0.062772, "gini_cps": 0.072541, "bpic": 1.891256, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.8563069888300816, "fi": 1.1382687395569864}, "inventory": {"size": 9835, "admitted": 9835, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 9835, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 14.775901794433594, "n_util_obs": 2, "util_count_spearman": NaN}, "t": 21.2} |
| {"event": "probe", "step": 16, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3714.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.213411, "cps": 29.015625, "mean_p": 0.484174, "boundary_entropy": 0.686145, "bpb": 6.723808}, "tr": {"chunks": 7124.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.393526, "cps": 55.65625, "mean_p": 0.493765, "boundary_entropy": 0.68698, "bpb": 7.011114}, "fi": {"chunks": 4441.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.236085, "cps": 34.695312, "mean_p": 0.48584, "boundary_entropy": 0.68685, "bpb": 6.78689}}, "bpb": {"en": 6.72381, "tr": 7.01111, "fi": 6.78689}, "gini_cpc": 0.142435, "gini_cps": 0.148788, "bpic": 3.71307, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.8439773412898386, "fi": 1.106244464428016}, "inventory": {"size": 23912, "admitted": 23912, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 23912, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 9.452032089233398, "n_util_obs": 5, "util_count_spearman": -0.1}, "t": 27.8} |
| {"event": "probe", "step": 32, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 2611.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.150032, "cps": 20.398438, "mean_p": 0.48676, "boundary_entropy": 0.686948, "bpb": 6.304671}, "tr": {"chunks": 4495.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.248301, "cps": 35.117188, "mean_p": 0.48969, "boundary_entropy": 0.687534, "bpb": 6.651901}, "fi": {"chunks": 1345.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.071501, "cps": 10.507812, "mean_p": 0.478861, "boundary_entropy": 0.686918, "bpb": 6.422253}}, "bpb": {"en": 6.30467, "tr": 6.6519, "fi": 6.42225}, "gini_cpc": 0.25087, "gini_cps": 0.248491, "bpic": 6.713052, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.654993883356045, "fi": 0.4765710415654554}, "inventory": {"size": 50566, "admitted": 50566, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 50566, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 6.35850191116333, "n_util_obs": 8, "util_count_spearman": -0.8809523809523809}, "t": 43.1} |
| {"event": "consolidate", "step": 50, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 57.0} |
| {"event": "train", "step": 50, "arch": "lexmem_v1", "phase": 1, "objective": "global", "lm_loss": 4.21185, "bpb": 6.07641, "chunk_loss": 0.034047, "F": 0.2236, "G": 0.48837, "chunks": 228.97, "bytes_per_chunk": 4.472, "boundary_entropy": 0.69159, "router_noise": 0.0, "grad_norm": 2.4785, "lr": 5.1e-05, "router_trainable": true, "sec_per_step": 1.018, "probe_sec": 0.0783, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04594, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.00344, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 0.4145, "trunc_abs_frac": 0.39184, "mean_delta_full": 0.02588, "read_hit_rate": {"surface": 0.69661}, "mean_gate": {"surface": 0.13458}, "n_reads": 5079, "n_cand": 7291, "read_suppressed_frac": 0.1032, "dropout_reads_logged": 5079, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 6.1507415771484375, "n_util_obs": 9, "util_count_spearman": -0.6833333333333333}, "probe_overhead_sec_total": 3.408, "probe_overhead_seq_frac": 0.4871, "util_loss": 0.065329, "alpha": 0.03, "t": 57.9} |
| {"event": "probe", "step": 64, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 4930.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.283284, "cps": 38.515625, "mean_p": 0.485749, "boundary_entropy": 0.685552, "bpb": 5.54737}, "tr": {"chunks": 5333.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.294592, "cps": 41.664062, "mean_p": 0.486252, "boundary_entropy": 0.68647, "bpb": 5.953626}, "fi": {"chunks": 4704.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.250066, "cps": 36.75, "mean_p": 0.483748, "boundary_entropy": 0.686302, "bpb": 5.716498}}, "bpb": {"en": 5.54737, "tr": 5.95363, "fi": 5.7165}, "gini_cpc": 0.035852, "gini_cps": 0.028017, "bpic": 3.790472, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0399159351732967, "fi": 0.882739642540083}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 7.0018157958984375, "n_util_obs": 11, "util_count_spearman": -0.6818181818181818}, "t": 69.6} |
| {"event": "consolidate", "step": 100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 101.5} |
| {"event": "train", "step": 100, "arch": "lexmem_v1", "phase": 1, "objective": "global", "lm_loss": 3.24515, "bpb": 4.68176, "chunk_loss": 0.029269, "F": 0.15466, "G": 0.45153, "chunks": 158.38, "bytes_per_chunk": 6.466, "boundary_entropy": 0.68162, "router_noise": 0.0, "grad_norm": 1.5941, "lr": 0.000101, "router_trainable": true, "sec_per_step": 0.877, "probe_sec": 0.0326, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02677, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.00073, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.34337, "trunc_abs_frac": 0.91852, "mean_delta_full": -0.04053, "read_hit_rate": {"surface": 0.58184}, "mean_gate": {"surface": 0.16132}, "n_reads": 2929, "n_cand": 5034, "read_suppressed_frac": 0.1021, "dropout_reads_logged": 2929, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 9.382949829101562, "n_util_obs": 22, "util_count_spearman": 0.04009034443817053}, "probe_overhead_sec_total": 6.82, "probe_overhead_seq_frac": 0.4935, "util_loss": 0.001536, "alpha": 0.03, "t": 101.8} |
| {"event": "probe", "step": 128, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3927.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.225651, "cps": 30.679688, "mean_p": 0.417888, "boundary_entropy": 0.645766, "bpb": 4.148002}, "tr": {"chunks": 4663.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.257582, "cps": 36.429688, "mean_p": 0.43604, "boundary_entropy": 0.659666, "bpb": 4.399325}, "fi": {"chunks": 3673.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.195258, "cps": 28.695312, "mean_p": 0.428559, "boundary_entropy": 0.660762, "bpb": 4.342142}}, "bpb": {"en": 4.148, "tr": 4.39933, "fi": 4.34214}, "gini_cpc": 0.061237, "gini_cps": 0.05382, "bpic": 4.626274, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.1415056961001573, "fi": 0.8653110782059114}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 11.265625, "n_util_obs": 29, "util_count_spearman": 0.2019704433497537}, "t": 125.0} |
| {"event": "consolidate", "step": 150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 133.9} |
| {"event": "train", "step": 150, "arch": "lexmem_v1", "phase": 1, "objective": "global", "lm_loss": 2.78109, "bpb": 4.01227, "chunk_loss": 0.032097, "F": 0.2106, "G": 0.38561, "chunks": 215.66, "bytes_per_chunk": 4.748, "boundary_entropy": 0.61996, "router_noise": 0.0, "grad_norm": 1.7489, "lr": 0.000151, "router_trainable": true, "sec_per_step": 0.647, "probe_sec": 0.0351, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02676, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.00526, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -0.27729, "trunc_abs_frac": 0.59956, "mean_delta_full": -0.03892, "read_hit_rate": {"surface": 0.70782}, "mean_gate": {"surface": 0.20261}, "n_reads": 4862, "n_cand": 6869, "read_suppressed_frac": 0.1057, "dropout_reads_logged": 4862, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 12.85247802734375, "n_util_obs": 34, "util_count_spearman": 0.22566844919786097}, "probe_overhead_sec_total": 9.863, "probe_overhead_seq_frac": 0.4957, "util_loss": 0.001222, "alpha": 0.03, "t": 134.1} |
| {"event": "consolidate", "step": 200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 149.3} |
| {"event": "probe", "step": 200, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3797.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.218181, "cps": 29.664062, "mean_p": 0.389886, "boundary_entropy": 0.597031, "bpb": 3.646479}, "tr": {"chunks": 4667.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.257803, "cps": 36.460938, "mean_p": 0.385631, "boundary_entropy": 0.601809, "bpb": 3.819036}, "fi": {"chunks": 4090.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.217426, "cps": 31.953125, "mean_p": 0.405121, "boundary_entropy": 0.635121, "bpb": 3.8757}}, "bpb": {"en": 3.64648, "tr": 3.81904, "fi": 3.8757}, "gini_cpc": 0.038819, "gini_cps": 0.0462, "bpic": 4.519038, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.181600789594078, "fi": 0.9965404867993188}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 15.730453491210938, "n_util_obs": 42, "util_count_spearman": -0.02665910380034033}, "t": 149.8} |
| {"event": "eval", "step": 200, "phase": 1, "bpb": {"en": 3.65332, "tr": 3.8295, "fi": 3.87415}, "bytes_per_chunk": {"en": 5.187, "tr": 4.872, "fi": 5.285}, "bpb_mean": 3.78565, "frozen_inventory": true, "t": 150.8} |
| {"event": "train", "step": 200, "arch": "lexmem_v1", "phase": 1, "objective": "global", "lm_loss": 2.53958, "bpb": 3.66384, "chunk_loss": 0.031028, "F": 0.18884, "G": 0.37709, "chunks": 193.38, "bytes_per_chunk": 5.295, "boundary_entropy": 0.59788, "router_noise": 0.0, "grad_norm": 1.9666, "lr": 0.000201, "router_trainable": true, "sec_per_step": 0.384, "probe_sec": 0.0582, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04497, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.00022, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -4.64375, "trunc_abs_frac": 0.57933, "mean_delta_full": 0.00391, "read_hit_rate": {"surface": 0.63759}, "mean_gate": {"surface": 0.27486}, "n_reads": 3925, "n_cand": 6156, "read_suppressed_frac": 0.0968, "dropout_reads_logged": 3925, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 15.79156494140625, "n_util_obs": 42, "util_count_spearman": -0.022445506847095047}, "probe_overhead_sec_total": 12.563, "probe_overhead_seq_frac": 0.4967, "util_loss": 0.000648, "alpha": 0.03, "t": 153.3} |
| {"event": "consolidate", "step": 250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 196.5} |
| {"event": "train", "step": 250, "arch": "lexmem_v1", "phase": 1, "objective": "global", "lm_loss": 2.38816, "bpb": 3.44539, "chunk_loss": 0.030715, "F": 0.18628, "G": 0.32726, "chunks": 190.75, "bytes_per_chunk": 5.368, "boundary_entropy": 0.52494, "router_noise": 0.0, "grad_norm": 1.8771, "lr": 0.000251, "router_trainable": true, "sec_per_step": 0.868, "probe_sec": 0.0316, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02299, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.02589, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 5.93662, "trunc_abs_frac": 0.77648, "mean_delta_full": 0.00867, "read_hit_rate": {"surface": 0.63867}, "mean_gate": {"surface": 0.30218}, "n_reads": 3878, "n_cand": 6072, "read_suppressed_frac": 0.091, "dropout_reads_logged": 3878, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 19.076217651367188, "n_util_obs": 49, "util_count_spearman": -0.020918367346938777}, "probe_overhead_sec_total": 16.474, "probe_overhead_seq_frac": 0.4974, "util_loss": 0.000873, "alpha": 0.03, "t": 196.8} |
| {"event": "probe", "step": 256, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3681.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.211515, "cps": 28.757812, "mean_p": 0.327629, "boundary_entropy": 0.502598, "bpb": 3.35089}, "tr": {"chunks": 4902.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.270784, "cps": 38.296875, "mean_p": 0.313787, "boundary_entropy": 0.498656, "bpb": 3.518348}, "fi": {"chunks": 4732.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.251555, "cps": 36.96875, "mean_p": 0.350487, "boundary_entropy": 0.558658, "bpb": 3.670988}}, "bpb": {"en": 3.35089, "tr": 3.51835, "fi": 3.67099}, "gini_cpc": 0.053842, "gini_cps": 0.061134, "bpic": 4.260759, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.2802095371945945, "fi": 1.1892992781062357}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 19.434478759765625, "n_util_obs": 51, "util_count_spearman": -0.002986425339366516}, "t": 200.9} |
| {"event": "consolidate", "step": 300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 229.4} |
| {"event": "train", "step": 300, "arch": "lexmem_v1", "phase": 1, "objective": "global", "lm_loss": 2.28745, "bpb": 3.3001, "chunk_loss": 0.030833, "F": 0.19781, "G": 0.28591, "chunks": 202.56, "bytes_per_chunk": 5.055, "boundary_entropy": 0.43896, "router_noise": 0.0, "grad_norm": 1.2897, "lr": 0.0003, "router_trainable": true, "sec_per_step": 0.658, "probe_sec": 0.0377, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03187, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.07211, "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": -9.7818, "trunc_abs_frac": 0.81918, "mean_delta_full": 0.00928, "read_hit_rate": {"surface": 0.64496}, "mean_gate": {"surface": 0.37581}, "n_reads": 4160, "n_cand": 6450, "read_suppressed_frac": 0.0993, "dropout_reads_logged": 4160, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 22.550216674804688, "n_util_obs": 65, "util_count_spearman": 0.14860302233210448}, "probe_overhead_sec_total": 19.538, "probe_overhead_seq_frac": 0.4978, "util_loss": 0.001119, "alpha": 0.03, "t": 229.7} |
| {"event": "consolidate", "step": 350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 267.7} |
| {"event": "train", "step": 350, "arch": "lexmem_v1", "phase": 1, "objective": "global", "lm_loss": 2.1682, "bpb": 3.12806, "chunk_loss": 0.031136, "F": 0.21661, "G": 0.26635, "chunks": 221.81, "bytes_per_chunk": 4.617, "boundary_entropy": 0.37476, "router_noise": 0.0, "grad_norm": 1.0976, "lr": 0.0002999, "router_trainable": true, "sec_per_step": 0.784, "probe_sec": 0.0663, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03405, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.02747, "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": 0.51536, "trunc_abs_frac": 1.0059, "mean_delta_full": 0.12581, "read_hit_rate": {"surface": 0.69983}, "mean_gate": {"surface": 0.41102}, "n_reads": 4945, "n_cand": 7066, "read_suppressed_frac": 0.1015, "dropout_reads_logged": 4945, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 26.364181518554688, "n_util_obs": 75, "util_count_spearman": 0.12219061166429587}, "probe_overhead_sec_total": 22.805, "probe_overhead_seq_frac": 0.4981, "util_loss": 0.001687, "alpha": 0.03, "t": 268.8} |
| {"event": "consolidate", "step": 400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 307.2} |
| {"event": "probe", "step": 400, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3788.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.217664, "cps": 29.59375, "mean_p": 0.258498, "boundary_entropy": 0.329023, "bpb": 2.890088}, "tr": {"chunks": 3383.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.186875, "cps": 26.429688, "mean_p": 0.25739, "boundary_entropy": 0.375192, "bpb": 3.178569}, "fi": {"chunks": 4337.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.230557, "cps": 33.882812, "mean_p": 0.265977, "boundary_entropy": 0.401303, "bpb": 3.368847}}, "bpb": {"en": 2.89009, "tr": 3.17857, "fi": 3.36885}, "gini_cpc": 0.045853, "gini_cps": 0.055266, "bpic": 4.929788, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.8585500072592938, "fi": 1.0592334536337311}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 30.194732666015625, "n_util_obs": 93, "util_count_spearman": 0.13135985198889916}, "t": 307.7} |
| {"event": "eval", "step": 400, "phase": 1, "bpb": {"en": 2.90246, "tr": 3.18546, "fi": 3.36098}, "bytes_per_chunk": {"en": 4.422, "tr": 5.826, "fi": 4.829}, "bpb_mean": 3.14964, "frozen_inventory": true, "t": 308.6} |
| {"event": "train", "step": 400, "arch": "lexmem_v1", "phase": 1, "objective": "global", "lm_loss": 2.08265, "bpb": 3.00463, "chunk_loss": 0.030943, "F": 0.21094, "G": 0.25793, "chunks": 216.0, "bytes_per_chunk": 4.741, "boundary_entropy": 0.33752, "router_noise": 0.0, "grad_norm": 1.1415, "lr": 0.0002998, "router_trainable": true, "sec_per_step": 0.823, "probe_sec": 0.0362, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02058, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.14562, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.72211, "trunc_abs_frac": 0.56577, "mean_delta_full": 0.35339, "read_hit_rate": {"surface": 0.69637}, "mean_gate": {"surface": 0.47175}, "n_reads": 4791, "n_cand": 6880, "read_suppressed_frac": 0.1039, "dropout_reads_logged": 4791, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 30.268417358398438, "n_util_obs": 94, "util_count_spearman": 0.10959072354874833}, "probe_overhead_sec_total": 25.874, "probe_overhead_seq_frac": 0.4984, "util_loss": 0.002696, "alpha": 0.03, "t": 310.0} |
| {"event": "consolidate", "step": 450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 349.3} |
| {"event": "train", "step": 450, "arch": "lexmem_v1", "phase": 1, "objective": "global", "lm_loss": 1.99496, "bpb": 2.87812, "chunk_loss": 0.031379, "F": 0.233, "G": 0.259, "chunks": 238.59, "bytes_per_chunk": 4.292, "boundary_entropy": 0.30102, "router_noise": 0.0, "grad_norm": 1.6933, "lr": 0.0002995, "router_trainable": true, "sec_per_step": 0.79, "probe_sec": 0.0235, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01781, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.00397, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.22925, "trunc_abs_frac": 1.81427, "mean_delta_full": 0.10942, "read_hit_rate": {"surface": 0.75076}, "mean_gate": {"surface": 0.46407}, "n_reads": 5708, "n_cand": 7603, "read_suppressed_frac": 0.0993, "dropout_reads_logged": 5708, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 34.23805236816406, "n_util_obs": 106, "util_count_spearman": 0.11089502985112225}, "probe_overhead_sec_total": 29.15, "probe_overhead_seq_frac": 0.4985, "util_loss": 0.003876, "alpha": 0.03, "t": 349.5} |
| {"event": "consolidate", "step": 500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 381.2} |
| {"event": "train", "step": 500, "arch": "lexmem_v1", "phase": 1, "objective": "global", "lm_loss": 1.89089, "bpb": 2.72798, "chunk_loss": 0.031413, "F": 0.23236, "G": 0.2622, "chunks": 237.94, "bytes_per_chunk": 4.304, "boundary_entropy": 0.28144, "router_noise": 0.0, "grad_norm": 1.4329, "lr": 0.0002992, "router_trainable": true, "sec_per_step": 0.654, "probe_sec": 0.1158, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07665, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.0543, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.3493, "trunc_abs_frac": 0.5993, "mean_delta_full": -0.23694, "read_hit_rate": {"surface": 0.75125}, "mean_gate": {"surface": 0.44566}, "n_reads": 5696, "n_cand": 7582, "read_suppressed_frac": 0.1011, "dropout_reads_logged": 5696, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 38.31129455566406, "n_util_obs": 123, "util_count_spearman": 0.0849340422997873}, "probe_overhead_sec_total": 32.113, "probe_overhead_seq_frac": 0.4987, "util_loss": 0.00497, "alpha": 0.03, "t": 382.2} |
| {"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": 0.253983, "boundary_entropy": 0.283853, "bpb": 2.68678}, "tr": {"chunks": 3974.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.219522, "cps": 31.046875, "mean_p": 0.252009, "boundary_entropy": 0.320539, "bpb": 3.045035}, "fi": {"chunks": 4136.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.219871, "cps": 32.3125, "mean_p": 0.264158, "boundary_entropy": 0.350567, "bpb": 3.218011}}, "bpb": {"en": 2.68678, "tr": 3.04504, "fi": 3.21801}, "gini_cpc": 0.000781, "gini_cps": 0.01806, "bpic": 4.758198, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0019236458497627, "fi": 1.0035198364985651}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 39.18843078613281, "n_util_obs": 128, "util_count_spearman": 0.09126075810291155}, "t": 392.6} |
| {"event": "consolidate", "step": 550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 420.0} |
| {"event": "train", "step": 550, "arch": "lexmem_v1", "phase": 1, "objective": "global", "lm_loss": 1.89223, "bpb": 2.72992, "chunk_loss": 0.031066, "F": 0.22119, "G": 0.25233, "chunks": 226.5, "bytes_per_chunk": 4.521, "boundary_entropy": 0.2727, "router_noise": 0.0, "grad_norm": 1.3965, "lr": 0.0002987, "router_trainable": true, "sec_per_step": 0.772, "probe_sec": 0.0493, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02567, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.07718, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": -2.1962, "trunc_abs_frac": 3.89335, "mean_delta_full": 0.0225, "read_hit_rate": {"surface": 0.73933}, "mean_gate": {"surface": 0.45308}, "n_reads": 5335, "n_cand": 7216, "read_suppressed_frac": 0.0958, "dropout_reads_logged": 5335, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 42.36277770996094, "n_util_obs": 133, "util_count_spearman": 0.11210877274875311}, "probe_overhead_sec_total": 35.126, "probe_overhead_seq_frac": 0.4988, "util_loss": 0.005816, "alpha": 0.03, "t": 420.8} |
| {"event": "consolidate", "step": 600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 442.1} |
| {"event": "probe", "step": 600, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3898.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.223984, "cps": 30.453125, "mean_p": 0.250723, "boundary_entropy": 0.260663, "bpb": 2.563944}, "tr": {"chunks": 4071.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.22488, "cps": 31.804688, "mean_p": 0.243669, "boundary_entropy": 0.279038, "bpb": 2.90947}, "fi": {"chunks": 4347.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.231088, "cps": 33.960938, "mean_p": 0.267865, "boundary_entropy": 0.309899, "bpb": 3.138103}}, "bpb": {"en": 2.56394, "tr": 2.90947, "fi": 3.1381}, "gini_cpc": 0.006965, "gini_cps": 0.024304, "bpic": 4.606366, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0039979738539067, "fi": 1.0317157065534752}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 46.51246643066406, "n_util_obs": 148, "util_count_spearman": 0.10761206350851604}, "t": 442.7} |
| {"event": "eval", "step": 600, "phase": 1, "bpb": {"en": 2.57652, "tr": 2.91677, "fi": 3.13918}, "bytes_per_chunk": {"en": 4.351, "tr": 4.881, "fi": 4.683}, "bpb_mean": 2.87749, "frozen_inventory": true, "t": 443.7} |
| {"event": "train", "step": 600, "arch": "lexmem_v1", "phase": 1, "objective": "global", "lm_loss": 1.87564, "bpb": 2.70598, "chunk_loss": 0.031116, "F": 0.2254, "G": 0.25041, "chunks": 230.81, "bytes_per_chunk": 4.437, "boundary_entropy": 0.26238, "router_noise": 0.0, "grad_norm": 1.1551, "lr": 0.0002982, "router_trainable": true, "sec_per_step": 0.497, "probe_sec": 0.0783, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04758, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.23884, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.75128, "trunc_abs_frac": 0.75393, "mean_delta_full": 0.67458, "read_hit_rate": {"surface": 0.7347}, "mean_gate": {"surface": 0.44548}, "n_reads": 5403, "n_cand": 7354, "read_suppressed_frac": 0.1027, "dropout_reads_logged": 5403, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 46.59352111816406, "n_util_obs": 148, "util_count_spearman": 0.11279287249888453}, "probe_overhead_sec_total": 38.046, "probe_overhead_seq_frac": 0.4989, "util_loss": 0.006963, "alpha": 0.03, "t": 445.7} |
| {"event": "consolidate", "step": 650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 472.8} |
| {"event": "train", "step": 650, "arch": "lexmem_v1", "phase": 1, "objective": "global", "lm_loss": 1.82919, "bpb": 2.63896, "chunk_loss": 0.031136, "F": 0.22443, "G": 0.25187, "chunks": 229.81, "bytes_per_chunk": 4.456, "boundary_entropy": 0.24919, "router_noise": 0.0, "grad_norm": 1.0952, "lr": 0.0002975, "router_trainable": true, "sec_per_step": 0.556, "probe_sec": 0.05, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03279, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.04712, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.79619, "trunc_abs_frac": 0.74067, "mean_delta_full": 0.12488, "read_hit_rate": {"surface": 0.75307}, "mean_gate": {"surface": 0.45537}, "n_reads": 5514, "n_cand": 7322, "read_suppressed_frac": 0.0947, "dropout_reads_logged": 5514, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 50.66845703125, "n_util_obs": 161, "util_count_spearman": 0.16789476635450748}, "probe_overhead_sec_total": 40.876, "probe_overhead_seq_frac": 0.499, "util_loss": 0.007735, "alpha": 0.03, "t": 473.5} |
| {"event": "consolidate", "step": 700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 513.8} |
| {"event": "train", "step": 700, "arch": "lexmem_v1", "phase": 1, "objective": "global", "lm_loss": 1.97494, "bpb": 2.84923, "chunk_loss": 0.031328, "F": 0.22974, "G": 0.25508, "chunks": 235.25, "bytes_per_chunk": 4.353, "boundary_entropy": 0.24365, "router_noise": 0.0, "grad_norm": 1.0432, "lr": 0.0002967, "router_trainable": true, "sec_per_step": 0.823, "probe_sec": 0.0585, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02816, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.00535, "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": 0.76945, "trunc_abs_frac": 0.92192, "mean_delta_full": 0.45251, "read_hit_rate": {"surface": 0.73372}, "mean_gate": {"surface": 0.41612}, "n_reads": 5500, "n_cand": 7496, "read_suppressed_frac": 0.0945, "dropout_reads_logged": 5500, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 54.748687744140625, "n_util_obs": 173, "util_count_spearman": 0.14753940680657343}, "probe_overhead_sec_total": 44.088, "probe_overhead_seq_frac": 0.4991, "util_loss": 0.008437, "alpha": 0.03, "t": 514.6} |
| {"event": "consolidate", "step": 750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 552.7} |
| {"event": "train", "step": 750, "arch": "lexmem_v1", "phase": 1, "objective": "global", "lm_loss": 1.77369, "bpb": 2.55889, "chunk_loss": 0.031114, "F": 0.22589, "G": 0.24875, "chunks": 231.31, "bytes_per_chunk": 4.427, "boundary_entropy": 0.23277, "router_noise": 0.0, "grad_norm": 1.236, "lr": 0.0002959, "router_trainable": true, "sec_per_step": 0.779, "probe_sec": 0.0837, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04277, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.00563, "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": -0.11623, "trunc_abs_frac": 0.25375, "mean_delta_full": 0.22721, "read_hit_rate": {"surface": 0.75102}, "mean_gate": {"surface": 0.42642}, "n_reads": 5535, "n_cand": 7370, "read_suppressed_frac": 0.1017, "dropout_reads_logged": 5535, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 58.775238037109375, "n_util_obs": 182, "util_count_spearman": 0.16567519067292638}, "probe_overhead_sec_total": 47.461, "probe_overhead_seq_frac": 0.4991, "util_loss": 0.009675, "alpha": 0.03, "t": 553.5} |
| {"event": "consolidate", "step": 800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 590.1} |
| {"event": "probe", "step": 800, "phase": 1, "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": 0.241306, "boundary_entropy": 0.233233, "bpb": 2.4145}, "tr": {"chunks": 4103.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.226648, "cps": 32.054688, "mean_p": 0.233426, "boundary_entropy": 0.245514, "bpb": 2.742881}, "fi": {"chunks": 4295.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.228324, "cps": 33.554688, "mean_p": 0.25667, "boundary_entropy": 0.257276, "bpb": 2.968894}}, "bpb": {"en": 2.4145, "tr": 2.74288, "fi": 2.96889}, "gini_cpc": 0.016401, "gini_cps": 0.033482, "bpic": 4.694026, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.069508334628445, "fi": 1.0774186591336516}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 62.71028137207031, "n_util_obs": 188, "util_count_spearman": 0.1606716187442095}, "t": 590.8} |
| {"event": "eval", "step": 800, "phase": 1, "bpb": {"en": 2.42598, "tr": 2.75199, "fi": 2.97206}, "bytes_per_chunk": {"en": 4.58, "tr": 4.855, "fi": 4.801}, "bpb_mean": 2.71668, "frozen_inventory": true, "t": 591.7} |
| {"event": "train", "step": 800, "arch": "lexmem_v1", "phase": 1, "objective": "global", "lm_loss": 1.85507, "bpb": 2.6763, "chunk_loss": 0.030982, "F": 0.21933, "G": 0.24641, "chunks": 224.59, "bytes_per_chunk": 4.559, "boundary_entropy": 0.23182, "router_noise": 0.0, "grad_norm": 1.1584, "lr": 0.0002949, "router_trainable": true, "sec_per_step": 0.809, "probe_sec": 0.0596, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03838, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.18104, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 1.03736, "trunc_abs_frac": 2.68057, "mean_delta_full": 0.65344, "read_hit_rate": {"surface": 0.71754}, "mean_gate": {"surface": 0.41561}, "n_reads": 5134, "n_cand": 7155, "read_suppressed_frac": 0.0989, "dropout_reads_logged": 5134, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 62.78791809082031, "n_util_obs": 188, "util_count_spearman": 0.1621236502337879}, "probe_overhead_sec_total": 50.907, "probe_overhead_seq_frac": 0.4992, "util_loss": 0.009955, "alpha": 0.03, "t": 594.0} |
| {"event": "consolidate", "step": 850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 632.3} |
| {"event": "train", "step": 850, "arch": "lexmem_v1", "phase": 1, "objective": "global", "lm_loss": 1.804, "bpb": 2.60262, "chunk_loss": 0.030881, "F": 0.21494, "G": 0.24277, "chunks": 220.09, "bytes_per_chunk": 4.653, "boundary_entropy": 0.23589, "router_noise": 0.0, "grad_norm": 1.0528, "lr": 0.0002938, "router_trainable": true, "sec_per_step": 0.784, "probe_sec": 0.0701, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03407, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.04677, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": -0.71673, "trunc_abs_frac": 1.39276, "mean_delta_full": 0.08958, "read_hit_rate": {"surface": 0.70717}, "mean_gate": {"surface": 0.42385}, "n_reads": 4958, "n_cand": 7011, "read_suppressed_frac": 0.0928, "dropout_reads_logged": 4958, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 66.77836608886719, "n_util_obs": 200, "util_count_spearman": 0.18647716192904823}, "probe_overhead_sec_total": 54.061, "probe_overhead_seq_frac": 0.4992, "util_loss": 0.010833, "alpha": 0.03, "t": 633.2} |
| {"event": "consolidate", "step": 900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 665.0} |
| {"event": "train", "step": 900, "arch": "lexmem_v1", "phase": 1, "objective": "global", "lm_loss": 1.59758, "bpb": 2.30481, "chunk_loss": 0.03072, "F": 0.20914, "G": 0.23878, "chunks": 214.16, "bytes_per_chunk": 4.782, "boundary_entropy": 0.22003, "router_noise": 0.0, "grad_norm": 1.158, "lr": 0.0002927, "router_trainable": true, "sec_per_step": 0.653, "probe_sec": 0.0828, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05814, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.05192, "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": 0.989, "trunc_abs_frac": 0.96602, "mean_delta_full": 0.18038, "read_hit_rate": {"surface": 0.73391}, "mean_gate": {"surface": 0.41617}, "n_reads": 5006, "n_cand": 6821, "read_suppressed_frac": 0.0997, "dropout_reads_logged": 5006, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 70.74334716796875, "n_util_obs": 213, "util_count_spearman": 0.09621104663057833}, "probe_overhead_sec_total": 57.232, "probe_overhead_seq_frac": 0.4993, "util_loss": 0.012639, "alpha": 0.03, "t": 665.9} |
| {"event": "consolidate", "step": 950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 706.6} |
| {"event": "train", "step": 950, "arch": "lexmem_v1", "phase": 1, "objective": "global", "lm_loss": 1.67941, "bpb": 2.42288, "chunk_loss": 0.030943, "F": 0.22073, "G": 0.24108, "chunks": 226.03, "bytes_per_chunk": 4.53, "boundary_entropy": 0.20344, "router_noise": 0.0, "grad_norm": 1.2289, "lr": 0.0002914, "router_trainable": true, "sec_per_step": 0.819, "probe_sec": 0.0846, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07332, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.07331, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": -0.72963, "trunc_abs_frac": 0.72963, "mean_delta_full": -0.1438, "read_hit_rate": {"surface": 0.74906}, "mean_gate": {"surface": 0.43368}, "n_reads": 5394, "n_cand": 7201, "read_suppressed_frac": 0.0944, "dropout_reads_logged": 5394, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 74.66146850585938, "n_util_obs": 220, "util_count_spearman": 0.09538388023639131}, "probe_overhead_sec_total": 60.533, "probe_overhead_seq_frac": 0.4993, "util_loss": 0.01252, "alpha": 0.03, "t": 706.8} |
| {"event": "consolidate", "step": 1000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 740.5} |
| {"event": "probe", "step": 1000, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3630.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.208585, "cps": 28.359375, "mean_p": 0.235518, "boundary_entropy": 0.211273, "bpb": 2.326551}, "tr": {"chunks": 4124.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.227808, "cps": 32.21875, "mean_p": 0.229074, "boundary_entropy": 0.213765, "bpb": 2.632658}, "fi": {"chunks": 4536.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.241136, "cps": 35.4375, "mean_p": 0.26107, "boundary_entropy": 0.227451, "bpb": 2.832165}}, "bpb": {"en": 2.32655, "tr": 2.63266, "fi": 2.83216}, "gini_cpc": 0.032029, "gini_cps": 0.049146, "bpic": 4.616111, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0921583245186064, "fi": 1.1560554291470921}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 78.46987915039062, "n_util_obs": 230, "util_count_spearman": 0.11211885622532605}, "t": 741.0} |
| {"event": "eval", "step": 1000, "phase": 1, "bpb": {"en": 2.33644, "tr": 2.64056, "fi": 2.83896}, "bytes_per_chunk": {"en": 4.736, "tr": 4.818, "fi": 4.538}, "bpb_mean": 2.60532, "frozen_inventory": true, "t": 741.9} |
| {"event": "train", "step": 1000, "arch": "lexmem_v1", "phase": 1, "objective": "global", "lm_loss": 1.73666, "bpb": 2.50546, "chunk_loss": 0.030841, "F": 0.21613, "G": 0.24059, "chunks": 221.31, "bytes_per_chunk": 4.627, "boundary_entropy": 0.20689, "router_noise": 0.0, "grad_norm": 1.0105, "lr": 0.0002901, "router_trainable": true, "sec_per_step": 0.728, "probe_sec": 0.0321, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02422, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.0012, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.66263, "trunc_abs_frac": 0.46439, "mean_delta_full": 0.20355, "read_hit_rate": {"surface": 0.7156}, "mean_gate": {"surface": 0.37597}, "n_reads": 5045, "n_cand": 7050, "read_suppressed_frac": 0.094, "dropout_reads_logged": 5045, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 78.54470825195312, "n_util_obs": 231, "util_count_spearman": 0.10504977342086737}, "probe_overhead_sec_total": 63.307, "probe_overhead_seq_frac": 0.4993, "util_loss": 0.012803, "alpha": 0.03, "t": 743.2} |
| {"event": "probe", "step": 1024, "phase": 1, "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": 0.236692, "boundary_entropy": 0.216651, "bpb": 2.315874}, "tr": {"chunks": 3933.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.217257, "cps": 30.726562, "mean_p": 0.222994, "boundary_entropy": 0.220359, "bpb": 2.629791}, "fi": {"chunks": 4532.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.240923, "cps": 35.40625, "mean_p": 0.260115, "boundary_entropy": 0.233191, "bpb": 2.812642}}, "bpb": {"en": 2.31587, "tr": 2.62979, "fi": 2.81264}, "gini_cpc": 0.030351, "gini_cps": 0.047768, "bpic": 4.67777, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0321922561330394, "fi": 1.144630250366256}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 80.33705139160156, "n_util_obs": 237, "util_count_spearman": 0.13878536550300213}, "t": 756.2} |
| {"event": "consolidate", "step": 1050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 776.9} |
| {"event": "train", "step": 1050, "arch": "lexmem_v1", "phase": 1, "objective": "global", "lm_loss": 1.62982, "bpb": 2.35133, "chunk_loss": 0.030981, "F": 0.2146, "G": 0.23884, "chunks": 219.75, "bytes_per_chunk": 4.66, "boundary_entropy": 0.20568, "router_noise": 0.0, "grad_norm": 0.9357, "lr": 0.0002886, "router_trainable": true, "sec_per_step": 0.691, "probe_sec": 0.0927, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06746, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.42613, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 1.02277, "trunc_abs_frac": 1.13997, "mean_delta_full": 0.41142, "read_hit_rate": {"surface": 0.72529}, "mean_gate": {"surface": 0.42055}, "n_reads": 5077, "n_cand": 7000, "read_suppressed_frac": 0.0951, "dropout_reads_logged": 5077, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 82.40510559082031, "n_util_obs": 240, "util_count_spearman": 0.13058088273273455}, "probe_overhead_sec_total": 66.148, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.013691, "alpha": 0.03, "t": 777.8} |
| {"event": "consolidate", "step": 1100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 815.9} |
| {"event": "train", "step": 1100, "arch": "lexmem_v1", "phase": 1, "objective": "global", "lm_loss": 1.73758, "bpb": 2.50679, "chunk_loss": 0.03085, "F": 0.21481, "G": 0.23754, "chunks": 219.97, "bytes_per_chunk": 4.655, "boundary_entropy": 0.20258, "router_noise": 0.0, "grad_norm": 0.9526, "lr": 0.0002871, "router_trainable": true, "sec_per_step": 0.767, "probe_sec": 0.0731, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06545, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.08769, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.73909, "trunc_abs_frac": 0.49229, "mean_delta_full": -0.15598, "read_hit_rate": {"surface": 0.72399}, "mean_gate": {"surface": 0.40984}, "n_reads": 5073, "n_cand": 7007, "read_suppressed_frac": 0.0952, "dropout_reads_logged": 5073, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 86.20928955078125, "n_util_obs": 255, "util_count_spearman": 0.12480862262934418}, "probe_overhead_sec_total": 69.295, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.014128, "alpha": 0.03, "t": 816.1} |
| {"event": "consolidate", "step": 1150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 845.1} |
| {"event": "train", "step": 1150, "arch": "lexmem_v1", "phase": 1, "objective": "global", "lm_loss": 1.63814, "bpb": 2.36333, "chunk_loss": 0.030793, "F": 0.21609, "G": 0.2367, "chunks": 221.28, "bytes_per_chunk": 4.628, "boundary_entropy": 0.1939, "router_noise": 0.0, "grad_norm": 0.8352, "lr": 0.0002855, "router_trainable": true, "sec_per_step": 0.595, "probe_sec": 0.0672, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02694, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.19257, "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": 0.28895, "trunc_abs_frac": 0.22487, "mean_delta_full": 0.04309, "read_hit_rate": {"surface": 0.73713}, "mean_gate": {"surface": 0.38671}, "n_reads": 5196, "n_cand": 7049, "read_suppressed_frac": 0.0937, "dropout_reads_logged": 5196, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 90.00779724121094, "n_util_obs": 277, "util_count_spearman": 0.14254259516026926}, "probe_overhead_sec_total": 72.315, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.015728, "alpha": 0.03, "t": 845.8} |
| {"event": "consolidate", "step": 1200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 886.5} |
| {"event": "probe", "step": 1200, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3632.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.2087, "cps": 28.375, "mean_p": 0.233769, "boundary_entropy": 0.195426, "bpb": 2.256038}, "tr": {"chunks": 3909.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.215931, "cps": 30.539062, "mean_p": 0.21919, "boundary_entropy": 0.199806, "bpb": 2.55064}, "fi": {"chunks": 4622.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.245707, "cps": 36.109375, "mean_p": 0.261206, "boundary_entropy": 0.207364, "bpb": 2.714408}}, "bpb": {"en": 2.25604, "tr": 2.55064, "fi": 2.71441}, "gini_cpc": 0.036805, "gini_cps": 0.054263, "bpic": 4.66431, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0346498505492676, "fi": 1.177324923766369}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 93.709228515625, "n_util_obs": 299, "util_count_spearman": 0.1402268323735901}, "t": 887.1} |
| {"event": "eval", "step": 1200, "phase": 1, "bpb": {"en": 2.26724, "tr": 2.56037, "fi": 2.724}, "bytes_per_chunk": {"en": 4.635, "tr": 5.114, "fi": 4.395}, "bpb_mean": 2.5172, "frozen_inventory": true, "t": 888.2} |
| {"event": "train", "step": 1200, "arch": "lexmem_v1", "phase": 1, "objective": "global", "lm_loss": 1.61322, "bpb": 2.32739, "chunk_loss": 0.030818, "F": 0.21646, "G": 0.23755, "chunks": 221.66, "bytes_per_chunk": 4.62, "boundary_entropy": 0.19513, "router_noise": 0.0, "grad_norm": 0.8368, "lr": 0.0002837, "router_trainable": true, "sec_per_step": 0.874, "probe_sec": 0.0273, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01931, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.05448, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.26807, "trunc_abs_frac": 0.26807, "mean_delta_full": 0.47974, "read_hit_rate": {"surface": 0.74225}, "mean_gate": {"surface": 0.40714}, "n_reads": 5241, "n_cand": 7061, "read_suppressed_frac": 0.1065, "dropout_reads_logged": 5241, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 93.78953552246094, "n_util_obs": 300, "util_count_spearman": 0.13006393295390659}, "probe_overhead_sec_total": 75.599, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.015568, "alpha": 0.03, "t": 889.5} |
| {"event": "consolidate", "step": 1250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 928.4} |
| {"event": "train", "step": 1250, "arch": "lexmem_v1", "phase": 1, "objective": "global", "lm_loss": 1.63449, "bpb": 2.35807, "chunk_loss": 0.030547, "F": 0.203, "G": 0.22676, "chunks": 207.88, "bytes_per_chunk": 4.926, "boundary_entropy": 0.19059, "router_noise": 0.0, "grad_norm": 0.9088, "lr": 0.0002819, "router_trainable": true, "sec_per_step": 0.782, "probe_sec": 0.0949, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08732, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.10248, "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": 0.95251, "trunc_abs_frac": 5.69188, "mean_delta_full": 0.23454, "read_hit_rate": {"surface": 0.71707}, "mean_gate": {"surface": 0.39513}, "n_reads": 4747, "n_cand": 6620, "read_suppressed_frac": 0.0961, "dropout_reads_logged": 4747, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 97.55459594726562, "n_util_obs": 310, "util_count_spearman": 0.13347124220398207}, "probe_overhead_sec_total": 79.108, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.015335, "alpha": 0.03, "t": 928.6} |
| {"event": "consolidate", "step": 1300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 965.2} |
| {"event": "train", "step": 1300, "arch": "lexmem_v1", "phase": 1, "objective": "global", "lm_loss": 1.66826, "bpb": 2.40679, "chunk_loss": 0.030844, "F": 0.21588, "G": 0.23703, "chunks": 221.06, "bytes_per_chunk": 4.632, "boundary_entropy": 0.178, "router_noise": 0.0, "grad_norm": 0.9444, "lr": 0.00028, "router_trainable": true, "sec_per_step": 0.754, "probe_sec": 0.0945, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04971, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.15511, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.51441, "trunc_abs_frac": 0.15793, "mean_delta_full": 0.12708, "read_hit_rate": {"surface": 0.7184}, "mean_gate": {"surface": 0.40232}, "n_reads": 5059, "n_cand": 7042, "read_suppressed_frac": 0.1069, "dropout_reads_logged": 5059, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 101.29327392578125, "n_util_obs": 324, "util_count_spearman": 0.13564731208441588}, "probe_overhead_sec_total": 82.454, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.014354, "alpha": 0.03, "t": 966.4} |
| {"event": "consolidate", "step": 1350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1004.9} |
| {"event": "train", "step": 1350, "arch": "lexmem_v1", "phase": 1, "objective": "global", "lm_loss": 1.59265, "bpb": 2.2977, "chunk_loss": 0.030793, "F": 0.21429, "G": 0.23345, "chunks": 219.44, "bytes_per_chunk": 4.666, "boundary_entropy": 0.17685, "router_noise": 0.0, "grad_norm": 0.9179, "lr": 0.000278, "router_trainable": true, "sec_per_step": 0.777, "probe_sec": 0.0415, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02597, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.06768, "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": 0.29562, "trunc_abs_frac": 0.799, "mean_delta_full": 0.18306, "read_hit_rate": {"surface": 0.73591}, "mean_gate": {"surface": 0.38165}, "n_reads": 5144, "n_cand": 6990, "read_suppressed_frac": 0.099, "dropout_reads_logged": 5144, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 105.06669616699219, "n_util_obs": 337, "util_count_spearman": 0.12462329913376394}, "probe_overhead_sec_total": 85.336, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.015908, "alpha": 0.03, "t": 1005.2} |
| {"event": "consolidate", "step": 1400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1049.9} |
| {"event": "probe", "step": 1400, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3648.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.209619, "cps": 28.5, "mean_p": 0.231083, "boundary_entropy": 0.183531, "bpb": 2.206021}, "tr": {"chunks": 3473.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.191847, "cps": 27.132812, "mean_p": 0.205063, "boundary_entropy": 0.197175, "bpb": 2.487687}, "fi": {"chunks": 4559.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.242358, "cps": 35.617188, "mean_p": 0.25408, "boundary_entropy": 0.189249, "bpb": 2.626872}}, "bpb": {"en": 2.20602, "tr": 2.48769, "fi": 2.62687}, "gini_cpc": 0.052304, "gini_cps": 0.061986, "bpic": 4.857192, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9152158282139919, "fi": 1.1561841178453818}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 108.74443054199219, "n_util_obs": 349, "util_count_spearman": 0.11854224414365065}, "t": 1050.5} |
| {"event": "eval", "step": 1400, "phase": 1, "bpb": {"en": 2.21545, "tr": 2.49719, "fi": 2.63474}, "bytes_per_chunk": {"en": 4.716, "tr": 5.666, "fi": 4.456}, "bpb_mean": 2.44913, "frozen_inventory": true, "t": 1051.5} |
| {"event": "train", "step": 1400, "arch": "lexmem_v1", "phase": 1, "objective": "global", "lm_loss": 1.59766, "bpb": 2.30493, "chunk_loss": 0.030651, "F": 0.2059, "G": 0.22694, "chunks": 210.84, "bytes_per_chunk": 4.857, "boundary_entropy": 0.18069, "router_noise": 0.0, "grad_norm": 0.8819, "lr": 0.0002759, "router_trainable": true, "sec_per_step": 0.946, "probe_sec": 0.023, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01718, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.02013, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.91556, "trunc_abs_frac": 1.15094, "mean_delta_full": 0.16589, "read_hit_rate": {"surface": 0.7166}, "mean_gate": {"surface": 0.38503}, "n_reads": 4812, "n_cand": 6715, "read_suppressed_frac": 0.0993, "dropout_reads_logged": 4812, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 108.81834411621094, "n_util_obs": 349, "util_count_spearman": 0.11757860973194818}, "probe_overhead_sec_total": 88.696, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.016791, "alpha": 0.03, "t": 1052.5} |
| {"event": "consolidate", "step": 1450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1076.2} |
| {"event": "train", "step": 1450, "arch": "lexmem_v1", "phase": 1, "objective": "global", "lm_loss": 1.64516, "bpb": 2.37346, "chunk_loss": 0.030542, "F": 0.20648, "G": 0.22208, "chunks": 211.44, "bytes_per_chunk": 4.843, "boundary_entropy": 0.17056, "router_noise": 0.0, "grad_norm": 0.8547, "lr": 0.0002738, "router_trainable": true, "sec_per_step": 0.477, "probe_sec": 0.0231, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01716, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.27665, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 1.05951, "trunc_abs_frac": 1.23637, "mean_delta_full": 0.11911, "read_hit_rate": {"surface": 0.71547}, "mean_gate": {"surface": 0.38131}, "n_reads": 4818, "n_cand": 6734, "read_suppressed_frac": 0.1002, "dropout_reads_logged": 4818, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 112.57679748535156, "n_util_obs": 360, "util_count_spearman": 0.13910258409356505}, "probe_overhead_sec_total": 91.226, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.015562, "alpha": 0.03, "t": 1076.3} |
| {"event": "consolidate", "step": 1500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1112.1} |
| {"event": "train", "step": 1500, "arch": "lexmem_v1", "phase": 1, "objective": "global", "lm_loss": 1.65765, "bpb": 2.39148, "chunk_loss": 0.030775, "F": 0.21417, "G": 0.23194, "chunks": 219.31, "bytes_per_chunk": 4.669, "boundary_entropy": 0.17153, "router_noise": 0.0, "grad_norm": 0.7584, "lr": 0.0002715, "router_trainable": true, "sec_per_step": 0.738, "probe_sec": 0.0575, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03025, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.13525, "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": -0.80108, "trunc_abs_frac": 1.66994, "mean_delta_full": -0.10136, "read_hit_rate": {"surface": 0.7286}, "mean_gate": {"surface": 0.36153}, "n_reads": 5090, "n_cand": 6986, "read_suppressed_frac": 0.0996, "dropout_reads_logged": 5090, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 116.30645751953125, "n_util_obs": 379, "util_count_spearman": 0.11019251411042966}, "probe_overhead_sec_total": 94.365, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.015984, "alpha": 0.03, "t": 1113.2} |
| {"event": "consolidate", "step": 1550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1153.7} |
| {"event": "train", "step": 1550, "arch": "lexmem_v1", "phase": 1, "objective": "global", "lm_loss": 1.58284, "bpb": 2.28356, "chunk_loss": 0.030626, "F": 0.2067, "G": 0.22458, "chunks": 211.66, "bytes_per_chunk": 4.838, "boundary_entropy": 0.16905, "router_noise": 0.0, "grad_norm": 0.8847, "lr": 0.0002692, "router_trainable": true, "sec_per_step": 0.814, "probe_sec": 0.0332, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02113, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.16889, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.48206, "trunc_abs_frac": 0.48206, "mean_delta_full": 0.27222, "read_hit_rate": {"surface": 0.71236}, "mean_gate": {"surface": 0.37277}, "n_reads": 4802, "n_cand": 6741, "read_suppressed_frac": 0.0973, "dropout_reads_logged": 4802, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 120.062744140625, "n_util_obs": 393, "util_count_spearman": 0.09984487970825534}, "probe_overhead_sec_total": 97.997, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.017191, "alpha": 0.03, "t": 1153.9} |
| {"event": "consolidate", "step": 1600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1194.0} |
| {"event": "probe", "step": 1600, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3568.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.205022, "cps": 27.875, "mean_p": 0.222709, "boundary_entropy": 0.164208, "bpb": 2.163723}, "tr": {"chunks": 3617.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.199801, "cps": 28.257812, "mean_p": 0.207024, "boundary_entropy": 0.177407, "bpb": 2.42092}, "fi": {"chunks": 4522.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.240391, "cps": 35.328125, "mean_p": 0.250555, "boundary_entropy": 0.170634, "bpb": 2.571557}}, "bpb": {"en": 2.16372, "tr": 2.42092, "fi": 2.57156}, "gini_cpc": 0.04194, "gini_cps": 0.054326, "bpic": 4.84599, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9745345301140533, "fi": 1.1725137626991935}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 123.679931640625, "n_util_obs": 406, "util_count_spearman": 0.09342467175628419}, "t": 1194.6} |
| {"event": "eval", "step": 1600, "phase": 1, "bpb": {"en": 2.1745, "tr": 2.43023, "fi": 2.57872}, "bytes_per_chunk": {"en": 4.868, "tr": 5.287, "fi": 4.565}, "bpb_mean": 2.39448, "frozen_inventory": true, "t": 1195.8} |
| {"event": "train", "step": 1600, "arch": "lexmem_v1", "phase": 1, "objective": "global", "lm_loss": 1.51373, "bpb": 2.18385, "chunk_loss": 0.030515, "F": 0.20517, "G": 0.22347, "chunks": 210.09, "bytes_per_chunk": 4.874, "boundary_entropy": 0.16326, "router_noise": 0.0, "grad_norm": 0.8649, "lr": 0.0002668, "router_trainable": true, "sec_per_step": 0.882, "probe_sec": 0.0483, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02488, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.12531, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 1.40458, "trunc_abs_frac": 0.87809, "mean_delta_full": 0.04861, "read_hit_rate": {"surface": 0.73263}, "mean_gate": {"surface": 0.38232}, "n_reads": 4902, "n_cand": 6691, "read_suppressed_frac": 0.1006, "dropout_reads_logged": 4902, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 123.75537109375, "n_util_obs": 406, "util_count_spearman": 0.09308201756706928}, "probe_overhead_sec_total": 101.074, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.017612, "alpha": 0.03, "t": 1198.0} |
| {"event": "consolidate", "step": 1650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1234.2} |
| {"event": "train", "step": 1650, "arch": "lexmem_v1", "phase": 1, "objective": "global", "lm_loss": 1.5788, "bpb": 2.27773, "chunk_loss": 0.030609, "F": 0.20932, "G": 0.22617, "chunks": 214.34, "bytes_per_chunk": 4.777, "boundary_entropy": 0.1589, "router_noise": 0.0, "grad_norm": 0.8487, "lr": 0.0002643, "router_trainable": true, "sec_per_step": 0.738, "probe_sec": 0.0488, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03233, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.05479, "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": 0.67838, "trunc_abs_frac": 3.15857, "mean_delta_full": 0.41756, "read_hit_rate": {"surface": 0.72096}, "mean_gate": {"surface": 0.34922}, "n_reads": 4922, "n_cand": 6827, "read_suppressed_frac": 0.0961, "dropout_reads_logged": 4922, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 127.46766662597656, "n_util_obs": 422, "util_count_spearman": 0.08351720979050047}, "probe_overhead_sec_total": 104.042, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.015884, "alpha": 0.03, "t": 1235.0} |
| {"event": "consolidate", "step": 1700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1271.0} |
| {"event": "train", "step": 1700, "arch": "lexmem_v1", "phase": 1, "objective": "global", "lm_loss": 1.53465, "bpb": 2.21403, "chunk_loss": 0.030489, "F": 0.20285, "G": 0.22051, "chunks": 207.72, "bytes_per_chunk": 4.93, "boundary_entropy": 0.15985, "router_noise": 0.0, "grad_norm": 0.8127, "lr": 0.0002618, "router_trainable": true, "sec_per_step": 0.742, "probe_sec": 0.0431, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02358, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.04112, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -5.23006, "trunc_abs_frac": 2.80822, "mean_delta_full": 0.01326, "read_hit_rate": {"surface": 0.71534}, "mean_gate": {"surface": 0.36587}, "n_reads": 4732, "n_cand": 6615, "read_suppressed_frac": 0.1006, "dropout_reads_logged": 4732, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 131.14532470703125, "n_util_obs": 431, "util_count_spearman": 0.08126010702159985}, "probe_overhead_sec_total": 107.104, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.018975, "alpha": 0.03, "t": 1272.0} |
| {"event": "consolidate", "step": 1750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1302.9} |
| {"event": "train", "step": 1750, "arch": "lexmem_v1", "phase": 1, "objective": "global", "lm_loss": 1.56019, "bpb": 2.25087, "chunk_loss": 0.030663, "F": 0.20898, "G": 0.22628, "chunks": 214.0, "bytes_per_chunk": 4.785, "boundary_entropy": 0.15591, "router_noise": 0.0, "grad_norm": 0.7341, "lr": 0.0002591, "router_trainable": true, "sec_per_step": 0.621, "probe_sec": 0.0239, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01826, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.26036, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 1.05699, "trunc_abs_frac": 0.95017, "mean_delta_full": 0.16404, "read_hit_rate": {"surface": 0.71464}, "mean_gate": {"surface": 0.37412}, "n_reads": 4871, "n_cand": 6816, "read_suppressed_frac": 0.1029, "dropout_reads_logged": 4871, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 134.8401336669922, "n_util_obs": 443, "util_count_spearman": 0.06489623602411251}, "probe_overhead_sec_total": 109.954, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.015544, "alpha": 0.03, "t": 1303.1} |
| {"event": "consolidate", "step": 1800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1336.0} |
| {"event": "probe", "step": 1800, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3632.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.2087, "cps": 28.375, "mean_p": 0.226127, "boundary_entropy": 0.154847, "bpb": 2.124397}, "tr": {"chunks": 3503.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.193504, "cps": 27.367188, "mean_p": 0.202962, "boundary_entropy": 0.173826, "bpb": 2.36491}, "fi": {"chunks": 4697.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.249694, "cps": 36.695312, "mean_p": 0.256351, "boundary_entropy": 0.16221, "bpb": 2.513035}}, "bpb": {"en": 2.1244, "tr": 2.36491, "fi": 2.51304}, "gini_cpc": 0.057463, "gini_cps": 0.067275, "bpic": 4.794794, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9271881367291085, "fi": 1.196429071166299}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 138.45884704589844, "n_util_obs": 451, "util_count_spearman": 0.06238017855443778}, "t": 1336.6} |
| {"event": "eval", "step": 1800, "phase": 1, "bpb": {"en": 2.13724, "tr": 2.37445, "fi": 2.51815}, "bytes_per_chunk": {"en": 4.811, "tr": 5.514, "fi": 4.339}, "bpb_mean": 2.34328, "frozen_inventory": true, "t": 1338.0} |
| {"event": "train", "step": 1800, "arch": "lexmem_v1", "phase": 1, "objective": "global", "lm_loss": 1.53907, "bpb": 2.22041, "chunk_loss": 0.030491, "F": 0.2012, "G": 0.22004, "chunks": 206.03, "bytes_per_chunk": 4.97, "boundary_entropy": 0.15554, "router_noise": 0.0, "grad_norm": 0.7441, "lr": 0.0002564, "router_trainable": true, "sec_per_step": 0.729, "probe_sec": 0.0484, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03471, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.14161, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.72382, "trunc_abs_frac": 1.07864, "mean_delta_full": 0.681, "read_hit_rate": {"surface": 0.70793}, "mean_gate": {"surface": 0.36585}, "n_reads": 4644, "n_cand": 6560, "read_suppressed_frac": 0.0952, "dropout_reads_logged": 4644, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 138.5289306640625, "n_util_obs": 451, "util_count_spearman": 0.06267607140005053}, "probe_overhead_sec_total": 112.787, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.017511, "alpha": 0.03, "t": 1339.5} |
| {"event": "consolidate", "step": 1850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1373.3} |
| {"event": "train", "step": 1850, "arch": "lexmem_v1", "phase": 1, "objective": "global", "lm_loss": 1.54732, "bpb": 2.23231, "chunk_loss": 0.030635, "F": 0.20898, "G": 0.2243, "chunks": 214.0, "bytes_per_chunk": 4.785, "boundary_entropy": 0.14725, "router_noise": 0.0, "grad_norm": 0.7616, "lr": 0.0002537, "router_trainable": true, "sec_per_step": 0.698, "probe_sec": 0.0502, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0237, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.0182, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": -0.23658, "trunc_abs_frac": 0.23658, "mean_delta_full": -0.14551, "read_hit_rate": {"surface": 0.72036}, "mean_gate": {"surface": 0.37003}, "n_reads": 4910, "n_cand": 6816, "read_suppressed_frac": 0.0923, "dropout_reads_logged": 4910, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 142.2401580810547, "n_util_obs": 457, "util_count_spearman": 0.04551056192765014}, "probe_overhead_sec_total": 115.852, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.016536, "alpha": 0.03, "t": 1374.4} |
| {"event": "consolidate", "step": 1900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1410.0} |
| {"event": "train", "step": 1900, "arch": "lexmem_v1", "phase": 1, "objective": "global", "lm_loss": 1.53337, "bpb": 2.21219, "chunk_loss": 0.030789, "F": 0.21182, "G": 0.22511, "chunks": 216.91, "bytes_per_chunk": 4.721, "boundary_entropy": 0.14683, "router_noise": 0.0, "grad_norm": 0.9001, "lr": 0.0002508, "router_trainable": true, "sec_per_step": 0.731, "probe_sec": 0.0676, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02987, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.23625, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.97106, "trunc_abs_frac": 1.27688, "mean_delta_full": -0.23859, "read_hit_rate": {"surface": 0.7318}, "mean_gate": {"surface": 0.36008}, "n_reads": 5056, "n_cand": 6909, "read_suppressed_frac": 0.0975, "dropout_reads_logged": 5056, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 145.9194793701172, "n_util_obs": 464, "util_count_spearman": 0.052916560256912916}, "probe_overhead_sec_total": 118.832, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.017885, "alpha": 0.03, "t": 1411.0} |
| {"event": "consolidate", "step": 1950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1448.1} |
| {"event": "train", "step": 1950, "arch": "lexmem_v1", "phase": 1, "objective": "global", "lm_loss": 1.48043, "bpb": 2.13581, "chunk_loss": 0.030563, "F": 0.20398, "G": 0.21957, "chunks": 208.88, "bytes_per_chunk": 4.902, "boundary_entropy": 0.14931, "router_noise": 0.0, "grad_norm": 0.7262, "lr": 0.0002479, "router_trainable": true, "sec_per_step": 0.762, "probe_sec": 0.0535, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02273, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.04473, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.48729, "trunc_abs_frac": 0.55142, "mean_delta_full": 0.34589, "read_hit_rate": {"surface": 0.7258}, "mean_gate": {"surface": 0.34539}, "n_reads": 4828, "n_cand": 6652, "read_suppressed_frac": 0.1, "dropout_reads_logged": 4828, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 149.61964416503906, "n_util_obs": 475, "util_count_spearman": 0.0527307918993494}, "probe_overhead_sec_total": 122.005, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.019975, "alpha": 0.03, "t": 1449.1} |
| {"event": "consolidate", "step": 2000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1474.4} |
| {"event": "probe", "step": 2000, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3593.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.206459, "cps": 28.070312, "mean_p": 0.221174, "boundary_entropy": 0.146659, "bpb": 2.09884}, "tr": {"chunks": 3385.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.186986, "cps": 26.445312, "mean_p": 0.201883, "boundary_entropy": 0.181297, "bpb": 2.318018}, "fi": {"chunks": 4544.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.241561, "cps": 35.5, "mean_p": 0.246423, "boundary_entropy": 0.150713, "bpb": 2.466672}}, "bpb": {"en": 2.09884, "tr": 2.31802, "fi": 2.46667}, "gini_cpc": 0.057296, "gini_cps": 0.06706, "bpic": 4.923798, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9056805155162548, "fi": 1.1700201528607412}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 153.22540283203125, "n_util_obs": 490, "util_count_spearman": 0.05450069980085293}, "t": 1475.0} |
| {"event": "eval", "step": 2000, "phase": 1, "bpb": {"en": 2.1116, "tr": 2.32892, "fi": 2.47633}, "bytes_per_chunk": {"en": 4.837, "tr": 5.89, "fi": 4.517}, "bpb_mean": 2.30562, "frozen_inventory": true, "t": 1475.9} |
| {"event": "train", "step": 2000, "arch": "lexmem_v1", "phase": 1, "objective": "global", "lm_loss": 1.64153, "bpb": 2.36822, "chunk_loss": 0.030872, "F": 0.21564, "G": 0.23143, "chunks": 220.81, "bytes_per_chunk": 4.637, "boundary_entropy": 0.15285, "router_noise": 0.0, "grad_norm": 0.8158, "lr": 0.000245, "router_trainable": true, "sec_per_step": 0.558, "probe_sec": 0.0253, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01864, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.17513, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.4308, "trunc_abs_frac": 0.3404, "mean_delta_full": 0.34662, "read_hit_rate": {"surface": 0.71083}, "mean_gate": {"surface": 0.37129}, "n_reads": 5000, "n_cand": 7034, "read_suppressed_frac": 0.1002, "dropout_reads_logged": 5000, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 153.3014373779297, "n_util_obs": 491, "util_count_spearman": 0.057907701735736214}, "probe_overhead_sec_total": 124.587, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.017396, "alpha": 0.03, "t": 1477.0} |
| {"event": "probe", "step": 2048, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3559.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.204505, "cps": 27.804688, "mean_p": 0.218503, "boundary_entropy": 0.143131, "bpb": 2.092553}, "tr": {"chunks": 3438.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.189913, "cps": 26.859375, "mean_p": 0.19685, "boundary_entropy": 0.166702, "bpb": 2.313115}, "fi": {"chunks": 4456.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.236883, "cps": 34.8125, "mean_p": 0.243485, "boundary_entropy": 0.151196, "bpb": 2.461545}}, "bpb": {"en": 2.09255, "tr": 2.31312, "fi": 2.46154}, "gini_cpc": 0.049601, "gini_cps": 0.059257, "bpic": 4.953462, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9286486957487825, "fi": 1.1583223359249681}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 156.8063507080078, "n_util_obs": 506, "util_count_spearman": 0.0690530756949015}, "t": 1517.4} |
| {"event": "consolidate", "step": 2050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1521.8} |
| {"event": "train", "step": 2050, "arch": "lexmem_v1", "phase": 1, "objective": "global", "lm_loss": 1.47796, "bpb": 2.13225, "chunk_loss": 0.030553, "F": 0.19965, "G": 0.21322, "chunks": 204.44, "bytes_per_chunk": 5.009, "boundary_entropy": 0.14605, "router_noise": 0.0, "grad_norm": 0.7765, "lr": 0.0002419, "router_trainable": true, "sec_per_step": 0.919, "probe_sec": 0.0729, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04534, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.15202, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 1.58884, "trunc_abs_frac": 5.15764, "mean_delta_full": 0.21075, "read_hit_rate": {"surface": 0.70737}, "mean_gate": {"surface": 0.36799}, "n_reads": 4605, "n_cand": 6510, "read_suppressed_frac": 0.0953, "dropout_reads_logged": 4605, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 157.025634765625, "n_util_obs": 506, "util_count_spearman": 0.06816793643005314}, "probe_overhead_sec_total": 127.507, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.020462, "alpha": 0.03, "t": 1522.9} |
| {"event": "consolidate", "step": 2100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1562.9} |
| {"event": "train", "step": 2100, "arch": "lexmem_v1", "phase": 1, "objective": "global", "lm_loss": 1.53627, "bpb": 2.21636, "chunk_loss": 0.030433, "F": 0.20105, "G": 0.21603, "chunks": 205.88, "bytes_per_chunk": 4.974, "boundary_entropy": 0.14401, "router_noise": 0.0, "grad_norm": 0.7432, "lr": 0.0002388, "router_trainable": true, "sec_per_step": 0.813, "probe_sec": 0.0302, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01823, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.24044, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 1.24328, "trunc_abs_frac": 0.62777, "mean_delta_full": 0.48938, "read_hit_rate": {"surface": 0.70592}, "mean_gate": {"surface": 0.34589}, "n_reads": 4628, "n_cand": 6556, "read_suppressed_frac": 0.0966, "dropout_reads_logged": 4628, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 160.711181640625, "n_util_obs": 513, "util_count_spearman": 0.054010161434508125}, "probe_overhead_sec_total": 130.533, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.01651, "alpha": 0.03, "t": 1563.6} |
| {"event": "consolidate", "step": 2150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1600.1} |
| {"event": "train", "step": 2150, "arch": "lexmem_v1", "phase": 1, "objective": "global", "lm_loss": 1.49679, "bpb": 2.15941, "chunk_loss": 0.030597, "F": 0.20865, "G": 0.22322, "chunks": 213.66, "bytes_per_chunk": 4.793, "boundary_entropy": 0.145, "router_noise": 0.0, "grad_norm": 0.6907, "lr": 0.0002357, "router_trainable": true, "sec_per_step": 0.736, "probe_sec": 0.0363, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02463, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.02973, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.37647, "trunc_abs_frac": 0.68759, "mean_delta_full": 0.13144, "read_hit_rate": {"surface": 0.72946}, "mean_gate": {"surface": 0.33616}, "n_reads": 4964, "n_cand": 6805, "read_suppressed_frac": 0.1086, "dropout_reads_logged": 4964, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 164.43084716796875, "n_util_obs": 525, "util_count_spearman": 0.062016763878500865}, "probe_overhead_sec_total": 133.673, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.021716, "alpha": 0.03, "t": 1600.4} |
| {"event": "consolidate", "step": 2200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1638.8} |
| {"event": "probe", "step": 2200, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3531.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.202896, "cps": 27.585938, "mean_p": 0.218396, "boundary_entropy": 0.138768, "bpb": 2.071851}, "tr": {"chunks": 3490.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.192786, "cps": 27.265625, "mean_p": 0.198101, "boundary_entropy": 0.15764, "bpb": 2.28149}, "fi": {"chunks": 4427.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.235341, "cps": 34.585938, "mean_p": 0.24213, "boundary_entropy": 0.145673, "bpb": 2.422977}}, "bpb": {"en": 2.07185, "tr": 2.28149, "fi": 2.42298}, "gini_cpc": 0.044959, "gini_cps": 0.054566, "bpic": 4.955625, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9501699211877883, "fi": 1.1599093283467659}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 168.04649353027344, "n_util_obs": 537, "util_count_spearman": 0.07855505370301868}, "t": 1639.4} |
| {"event": "eval", "step": 2200, "phase": 1, "bpb": {"en": 2.08236, "tr": 2.28973, "fi": 2.42974}, "bytes_per_chunk": {"en": 4.877, "tr": 5.589, "fi": 4.655}, "bpb_mean": 2.26727, "frozen_inventory": true, "t": 1640.2} |
| {"event": "train", "step": 2200, "arch": "lexmem_v1", "phase": 1, "objective": "global", "lm_loss": 1.47076, "bpb": 2.12185, "chunk_loss": 0.03049, "F": 0.20422, "G": 0.21729, "chunks": 209.12, "bytes_per_chunk": 4.897, "boundary_entropy": 0.13731, "router_noise": 0.0, "grad_norm": 0.6846, "lr": 0.0002325, "router_trainable": true, "sec_per_step": 0.831, "probe_sec": 0.0498, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03179, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.01009, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.43157, "trunc_abs_frac": 0.69991, "mean_delta_full": -0.04246, "read_hit_rate": {"surface": 0.72778}, "mean_gate": {"surface": 0.35286}, "n_reads": 4847, "n_cand": 6660, "read_suppressed_frac": 0.098, "dropout_reads_logged": 4847, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 168.12124633789062, "n_util_obs": 537, "util_count_spearman": 0.07867012709650113}, "probe_overhead_sec_total": 136.853, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.020869, "alpha": 0.03, "t": 1641.9} |
| {"event": "consolidate", "step": 2250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1676.1} |
| {"event": "train", "step": 2250, "arch": "lexmem_v1", "phase": 1, "objective": "global", "lm_loss": 1.49525, "bpb": 2.1572, "chunk_loss": 0.030486, "F": 0.20264, "G": 0.21544, "chunks": 207.5, "bytes_per_chunk": 4.935, "boundary_entropy": 0.13354, "router_noise": 0.0, "grad_norm": 0.7227, "lr": 0.0002293, "router_trainable": true, "sec_per_step": 0.694, "probe_sec": 0.0424, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02714, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.08165, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.73145, "trunc_abs_frac": 0.98746, "mean_delta_full": 0.50205, "read_hit_rate": {"surface": 0.71898}, "mean_gate": {"surface": 0.34124}, "n_reads": 4751, "n_cand": 6608, "read_suppressed_frac": 0.1027, "dropout_reads_logged": 4751, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 171.8083038330078, "n_util_obs": 550, "util_count_spearman": 0.08645202367515921}, "probe_overhead_sec_total": 139.848, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.019009, "alpha": 0.03, "t": 1676.6} |
| {"event": "consolidate", "step": 2300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1705.9} |
| {"event": "train", "step": 2300, "arch": "lexmem_v1", "phase": 1, "objective": "global", "lm_loss": 1.50825, "bpb": 2.17594, "chunk_loss": 0.030473, "F": 0.20456, "G": 0.21583, "chunks": 209.47, "bytes_per_chunk": 4.889, "boundary_entropy": 0.13251, "router_noise": 0.0, "grad_norm": 0.7708, "lr": 0.000226, "router_trainable": true, "sec_per_step": 0.589, "probe_sec": 0.0294, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01965, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.05883, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.7192, "trunc_abs_frac": 0.62366, "mean_delta_full": 0.17511, "read_hit_rate": {"surface": 0.71009}, "mean_gate": {"surface": 0.32883}, "n_reads": 4737, "n_cand": 6671, "read_suppressed_frac": 0.095, "dropout_reads_logged": 4737, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 175.51531982421875, "n_util_obs": 560, "util_count_spearman": 0.0910924109285705}, "probe_overhead_sec_total": 142.446, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.017438, "alpha": 0.03, "t": 1706.1} |
| {"event": "consolidate", "step": 2350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1747.1} |
| {"event": "train", "step": 2350, "arch": "lexmem_v1", "phase": 1, "objective": "global", "lm_loss": 1.55583, "bpb": 2.24459, "chunk_loss": 0.030565, "F": 0.20975, "G": 0.22147, "chunks": 214.78, "bytes_per_chunk": 4.768, "boundary_entropy": 0.13092, "router_noise": 0.0, "grad_norm": 0.7133, "lr": 0.0002226, "router_trainable": true, "sec_per_step": 0.84, "probe_sec": 0.0925, "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": 14, "mean_delta_loo": 0.19306, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.27304, "trunc_abs_frac": 1.07849, "mean_delta_full": -0.01788, "read_hit_rate": {"surface": 0.71817}, "mean_gate": {"surface": 0.34883}, "n_reads": 4913, "n_cand": 6841, "read_suppressed_frac": 0.1034, "dropout_reads_logged": 4913, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 179.1917266845703, "n_util_obs": 573, "util_count_spearman": 0.09645843654881217}, "probe_overhead_sec_total": 145.977, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.016279, "alpha": 0.03, "t": 1748.1} |
| {"event": "consolidate", "step": 2400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1790.8} |
| {"event": "probe", "step": 2400, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3620.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.20801, "cps": 28.28125, "mean_p": 0.2208, "boundary_entropy": 0.132265, "bpb": 2.052656}, "tr": {"chunks": 3397.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.187648, "cps": 26.539062, "mean_p": 0.194768, "boundary_entropy": 0.153757, "bpb": 2.252844}, "fi": {"chunks": 4538.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.241242, "cps": 35.453125, "mean_p": 0.24546, "boundary_entropy": 0.134141, "bpb": 2.384888}}, "bpb": {"en": 2.05266, "tr": 2.25284, "fi": 2.38489}, "gini_cpc": 0.056098, "gini_cps": 0.06583, "bpic": 4.909736, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9021121770055511, "fi": 1.1597600851271048}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 182.8261260986328, "n_util_obs": 589, "util_count_spearman": 0.09206026255304785}, "t": 1791.4} |
| {"event": "eval", "step": 2400, "phase": 1, "bpb": {"en": 2.0642, "tr": 2.26192, "fi": 2.39524}, "bytes_per_chunk": {"en": 4.796, "tr": 5.824, "fi": 4.574}, "bpb_mean": 2.24046, "frozen_inventory": true, "t": 1792.5} |
| {"event": "phase_transition", "step": 2400, "phase": 2, "objective": "global", "router_trainable": true, "encoder_trainable": true, "frozen_router_params": 0, "t": 1794.7} |
| {"event": "train", "step": 2400, "arch": "lexmem_v1", "phase": 2, "objective": "global", "lm_loss": 1.49859, "bpb": 2.162, "chunk_loss": 0.030546, "F": 0.2052, "G": 0.21947, "chunks": 210.12, "bytes_per_chunk": 4.873, "boundary_entropy": 0.12961, "router_noise": 0.0, "grad_norm": 0.7491, "lr": 0.0002192, "router_trainable": true, "sec_per_step": 0.953, "probe_sec": 0.0746, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03157, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.14621, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 8, "trunc_ratio_sum": 0.61478, "trunc_abs_frac": 0.90703, "mean_delta_full": 0.23923, "read_hit_rate": {"surface": 0.71175}, "mean_gate": {"surface": 0.34228}, "n_reads": 4763, "n_cand": 6692, "read_suppressed_frac": 0.0966, "dropout_reads_logged": 4763, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 182.8994598388672, "n_util_obs": 589, "util_count_spearman": 0.09205825404730977}, "probe_overhead_sec_total": 149.21, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.01928, "alpha": 0.03, "t": 1795.7} |
| {"event": "consolidate", "step": 2450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1820.2} |
| {"event": "train", "step": 2450, "arch": "lexmem_v1", "phase": 2, "objective": "global", "lm_loss": 1.46186, "bpb": 2.10902, "chunk_loss": 0.030404, "F": 0.19928, "G": 0.21147, "chunks": 204.06, "bytes_per_chunk": 5.018, "boundary_entropy": 0.13014, "router_noise": 0.0, "grad_norm": 0.6982, "lr": 0.0002158, "router_trainable": true, "sec_per_step": 0.493, "probe_sec": 0.0351, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02922, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.11048, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.33982, "trunc_abs_frac": 0.50252, "mean_delta_full": -0.1004, "read_hit_rate": {"surface": 0.7113}, "mean_gate": {"surface": 0.326}, "n_reads": 4622, "n_cand": 6498, "read_suppressed_frac": 0.106, "dropout_reads_logged": 4622, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 186.5843505859375, "n_util_obs": 599, "util_count_spearman": 0.10285787684058739}, "probe_overhead_sec_total": 151.751, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.020498, "alpha": 0.03, "t": 1820.4} |
| {"event": "consolidate", "step": 2500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1847.3} |
| {"event": "train", "step": 2500, "arch": "lexmem_v1", "phase": 2, "objective": "global", "lm_loss": 1.43489, "bpb": 2.07011, "chunk_loss": 0.03057, "F": 0.20926, "G": 0.22065, "chunks": 214.28, "bytes_per_chunk": 4.779, "boundary_entropy": 0.12928, "router_noise": 0.0, "grad_norm": 0.7483, "lr": 0.0002123, "router_trainable": true, "sec_per_step": 0.562, "probe_sec": 0.0826, "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": 14, "mean_delta_loo": 0.03806, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.68606, "trunc_abs_frac": 0.95607, "mean_delta_full": 0.25381, "read_hit_rate": {"surface": 0.73744}, "mean_gate": {"surface": 0.33419}, "n_reads": 5033, "n_cand": 6825, "read_suppressed_frac": 0.097, "dropout_reads_logged": 5033, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 190.27325439453125, "n_util_obs": 611, "util_count_spearman": 0.114303948515712}, "probe_overhead_sec_total": 154.315, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.016465, "alpha": 0.03, "t": 1848.4} |
| {"event": "consolidate", "step": 2550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1889.8} |
| {"event": "train", "step": 2550, "arch": "lexmem_v1", "phase": 2, "objective": "global", "lm_loss": 1.4619, "bpb": 2.10908, "chunk_loss": 0.030476, "F": 0.20459, "G": 0.21521, "chunks": 209.5, "bytes_per_chunk": 4.888, "boundary_entropy": 0.12277, "router_noise": 0.0, "grad_norm": 0.7, "lr": 0.0002088, "router_trainable": true, "sec_per_step": 0.847, "probe_sec": 0.0823, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03186, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.21846, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.96138, "trunc_abs_frac": 2.84571, "mean_delta_full": -0.17383, "read_hit_rate": {"surface": 0.72662}, "mean_gate": {"surface": 0.3349}, "n_reads": 4848, "n_cand": 6672, "read_suppressed_frac": 0.1068, "dropout_reads_logged": 4848, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 193.94285583496094, "n_util_obs": 620, "util_count_spearman": 0.10492526614990635}, "probe_overhead_sec_total": 157.458, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.016356, "alpha": 0.03, "t": 1890.8} |
| {"event": "consolidate", "step": 2600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1929.4} |
| {"event": "probe", "step": 2600, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3523.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.202436, "cps": 27.523438, "mean_p": 0.212204, "boundary_entropy": 0.122765, "bpb": 2.033071}, "tr": {"chunks": 3683.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.203447, "cps": 28.773438, "mean_p": 0.201228, "boundary_entropy": 0.137113, "bpb": 2.232648}, "fi": {"chunks": 4590.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.244006, "cps": 35.859375, "mean_p": 0.242884, "boundary_entropy": 0.125501, "bpb": 2.361674}}, "bpb": {"en": 2.03307, "tr": 2.23265, "fi": 2.36167}, "gini_cpc": 0.042643, "gini_cps": 0.060303, "bpic": 4.809427, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0049920920388864, "fi": 1.2053475212661533}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 197.59312438964844, "n_util_obs": 641, "util_count_spearman": 0.10526275982171954}, "t": 1930.1} |
| {"event": "eval", "step": 2600, "phase": 2, "bpb": {"en": 2.04463, "tr": 2.24271, "fi": 2.36969}, "bytes_per_chunk": {"en": 4.837, "tr": 5.122, "fi": 4.467}, "bpb_mean": 2.21901, "frozen_inventory": true, "t": 1931.2} |
| {"event": "train", "step": 2600, "arch": "lexmem_v1", "phase": 2, "objective": "global", "lm_loss": 1.41504, "bpb": 2.04147, "chunk_loss": 0.030473, "F": 0.20694, "G": 0.21365, "chunks": 211.91, "bytes_per_chunk": 4.832, "boundary_entropy": 0.12131, "router_noise": 0.0, "grad_norm": 0.7159, "lr": 0.0002053, "router_trainable": true, "sec_per_step": 0.857, "probe_sec": 0.0442, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02413, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.03864, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.25473, "trunc_abs_frac": 0.25473, "mean_delta_full": -0.49695, "read_hit_rate": {"surface": 0.73018}, "mean_gate": {"surface": 0.3312}, "n_reads": 4928, "n_cand": 6749, "read_suppressed_frac": 0.1002, "dropout_reads_logged": 4928, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 197.6685028076172, "n_util_obs": 641, "util_count_spearman": 0.1055937527511973}, "probe_overhead_sec_total": 160.859, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.020015, "alpha": 0.03, "t": 1933.7} |
| {"event": "consolidate", "step": 2650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1967.4} |
| {"event": "train", "step": 2650, "arch": "lexmem_v1", "phase": 2, "objective": "global", "lm_loss": 1.46197, "bpb": 2.10917, "chunk_loss": 0.030375, "F": 0.19843, "G": 0.20912, "chunks": 203.19, "bytes_per_chunk": 5.04, "boundary_entropy": 0.12426, "router_noise": 0.0, "grad_norm": 0.6866, "lr": 0.0002017, "router_trainable": true, "sec_per_step": 0.678, "probe_sec": 0.0321, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01897, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.01933, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 16.87712, "trunc_abs_frac": 16.87712, "mean_delta_full": -0.0072, "read_hit_rate": {"surface": 0.70912}, "mean_gate": {"surface": 0.33948}, "n_reads": 4588, "n_cand": 6470, "read_suppressed_frac": 0.099, "dropout_reads_logged": 4588, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 201.30368041992188, "n_util_obs": 657, "util_count_spearman": 0.09778750318163382}, "probe_overhead_sec_total": 163.685, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.02025, "alpha": 0.03, "t": 1967.6} |
| {"event": "consolidate", "step": 2700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2003.7} |
| {"event": "train", "step": 2700, "arch": "lexmem_v1", "phase": 2, "objective": "global", "lm_loss": 1.44519, "bpb": 2.08496, "chunk_loss": 0.030541, "F": 0.20648, "G": 0.21694, "chunks": 211.44, "bytes_per_chunk": 4.843, "boundary_entropy": 0.12278, "router_noise": 0.0, "grad_norm": 0.7263, "lr": 0.0001981, "router_trainable": true, "sec_per_step": 0.741, "probe_sec": 0.069, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0232, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.09131, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 1.04992, "trunc_abs_frac": 1.14552, "mean_delta_full": -0.34445, "read_hit_rate": {"surface": 0.72008}, "mean_gate": {"surface": 0.31312}, "n_reads": 4849, "n_cand": 6734, "read_suppressed_frac": 0.0942, "dropout_reads_logged": 4849, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 204.97972106933594, "n_util_obs": 662, "util_count_spearman": 0.10618133580307189}, "probe_overhead_sec_total": 166.467, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.020801, "alpha": 0.03, "t": 2004.6} |
| {"event": "consolidate", "step": 2750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2038.8} |
| {"event": "train", "step": 2750, "arch": "lexmem_v1", "phase": 2, "objective": "global", "lm_loss": 1.44932, "bpb": 2.09092, "chunk_loss": 0.030535, "F": 0.20544, "G": 0.21648, "chunks": 210.38, "bytes_per_chunk": 4.867, "boundary_entropy": 0.12657, "router_noise": 0.0, "grad_norm": 0.6843, "lr": 0.0001945, "router_trainable": true, "sec_per_step": 0.69, "probe_sec": 0.046, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03496, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.10821, "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": 0.20688, "trunc_abs_frac": 1.43164, "mean_delta_full": 0.03373, "read_hit_rate": {"surface": 0.72896}, "mean_gate": {"surface": 0.32157}, "n_reads": 4884, "n_cand": 6700, "read_suppressed_frac": 0.0987, "dropout_reads_logged": 4884, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 208.652099609375, "n_util_obs": 675, "util_count_spearman": 0.11876189535008005}, "probe_overhead_sec_total": 169.047, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.019881, "alpha": 0.03, "t": 2039.1} |
| {"event": "consolidate", "step": 2800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2080.2} |
| {"event": "probe", "step": 2800, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3528.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.202724, "cps": 27.5625, "mean_p": 0.213656, "boundary_entropy": 0.115975, "bpb": 2.015051}, "tr": {"chunks": 3415.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.188643, "cps": 26.679688, "mean_p": 0.197218, "boundary_entropy": 0.14854, "bpb": 2.197171}, "fi": {"chunks": 4535.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.241082, "cps": 35.429688, "mean_p": 0.242133, "boundary_entropy": 0.124339, "bpb": 2.316313}}, "bpb": {"en": 2.01505, "tr": 2.19717, "fi": 2.31631}, "gini_cpc": 0.055277, "gini_cps": 0.065052, "bpic": 4.942673, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9305414012260155, "fi": 1.1892165696218415}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 212.22805786132812, "n_util_obs": 681, "util_count_spearman": 0.12904022257225226}, "t": 2080.9} |
| {"event": "eval", "step": 2800, "phase": 2, "bpb": {"en": 2.02728, "tr": 2.20639, "fi": 2.3225}, "bytes_per_chunk": {"en": 4.848, "tr": 5.731, "fi": 4.553}, "bpb_mean": 2.18539, "frozen_inventory": true, "t": 2082.1} |
| {"event": "train", "step": 2800, "arch": "lexmem_v1", "phase": 2, "objective": "global", "lm_loss": 1.40445, "bpb": 2.0262, "chunk_loss": 0.030467, "F": 0.2038, "G": 0.21488, "chunks": 208.69, "bytes_per_chunk": 4.907, "boundary_entropy": 0.11639, "router_noise": 0.0, "grad_norm": 0.6959, "lr": 0.0001909, "router_trainable": true, "sec_per_step": 0.932, "probe_sec": 0.0744, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04449, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.12813, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 1.35442, "trunc_abs_frac": 1.35442, "mean_delta_full": -0.24695, "read_hit_rate": {"surface": 0.73367}, "mean_gate": {"surface": 0.33689}, "n_reads": 4876, "n_cand": 6646, "read_suppressed_frac": 0.0937, "dropout_reads_logged": 4876, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 212.30360412597656, "n_util_obs": 682, "util_count_spearman": 0.12730179553426926}, "probe_overhead_sec_total": 172.277, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.021366, "alpha": 0.03, "t": 2085.7} |
| {"event": "consolidate", "step": 2850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2128.6} |
| {"event": "train", "step": 2850, "arch": "lexmem_v1", "phase": 2, "objective": "global", "lm_loss": 1.4574, "bpb": 2.10258, "chunk_loss": 0.030676, "F": 0.21606, "G": 0.22403, "chunks": 221.25, "bytes_per_chunk": 4.628, "boundary_entropy": 0.11545, "router_noise": 0.0, "grad_norm": 0.7226, "lr": 0.0001872, "router_trainable": true, "sec_per_step": 0.878, "probe_sec": 0.0702, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03787, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.01911, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.28629, "trunc_abs_frac": 0.37841, "mean_delta_full": 0.18604, "read_hit_rate": {"surface": 0.73127}, "mean_gate": {"surface": 0.31497}, "n_reads": 5154, "n_cand": 7048, "read_suppressed_frac": 0.0906, "dropout_reads_logged": 5154, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 215.99908447265625, "n_util_obs": 688, "util_count_spearman": 0.12675753507654616}, "probe_overhead_sec_total": 175.481, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.020217, "alpha": 0.03, "t": 2129.6} |
| {"event": "consolidate", "step": 2900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2159.6} |
| {"event": "train", "step": 2900, "arch": "lexmem_v1", "phase": 2, "objective": "global", "lm_loss": 1.38557, "bpb": 1.99895, "chunk_loss": 0.030351, "F": 0.1994, "G": 0.21041, "chunks": 204.19, "bytes_per_chunk": 5.015, "boundary_entropy": 0.12075, "router_noise": 0.0, "grad_norm": 0.6388, "lr": 0.0001835, "router_trainable": true, "sec_per_step": 0.613, "probe_sec": 0.0481, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02949, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.32168, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 1.02772, "trunc_abs_frac": 1.6454, "mean_delta_full": 0.8125, "read_hit_rate": {"surface": 0.72532}, "mean_gate": {"surface": 0.31991}, "n_reads": 4716, "n_cand": 6502, "read_suppressed_frac": 0.0958, "dropout_reads_logged": 4716, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 219.66380310058594, "n_util_obs": 706, "util_count_spearman": 0.12514435838924}, "probe_overhead_sec_total": 178.717, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.017297, "alpha": 0.03, "t": 2160.3} |
| {"event": "consolidate", "step": 2950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2180.4} |
| {"event": "train", "step": 2950, "arch": "lexmem_v1", "phase": 2, "objective": "global", "lm_loss": 1.53483, "bpb": 2.2143, "chunk_loss": 0.030647, "F": 0.20764, "G": 0.21758, "chunks": 212.62, "bytes_per_chunk": 4.816, "boundary_entropy": 0.12059, "router_noise": 0.0, "grad_norm": 0.7018, "lr": 0.0001799, "router_trainable": true, "sec_per_step": 0.413, "probe_sec": 0.4016, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06295, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.05276, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 1.25888, "trunc_abs_frac": 1.184, "mean_delta_full": 0.25781, "read_hit_rate": {"surface": 0.70585}, "mean_gate": {"surface": 0.31981}, "n_reads": 4780, "n_cand": 6772, "read_suppressed_frac": 0.0964, "dropout_reads_logged": 4780, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 223.315673828125, "n_util_obs": 716, "util_count_spearman": 0.12733424195635054}, "probe_overhead_sec_total": 181.403, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.018973, "alpha": 0.03, "t": 2180.9} |
| {"event": "consolidate", "step": 3000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2213.2} |
| {"event": "probe", "step": 3000, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3564.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.204792, "cps": 27.84375, "mean_p": 0.212931, "boundary_entropy": 0.115102, "bpb": 1.995162}, "tr": {"chunks": 3564.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.196873, "cps": 27.84375, "mean_p": 0.198422, "boundary_entropy": 0.131699, "bpb": 2.174585}, "fi": {"chunks": 4588.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.2439, "cps": 35.84375, "mean_p": 0.241944, "boundary_entropy": 0.117316, "bpb": 2.299406}}, "bpb": {"en": 1.99516, "tr": 2.17458, "fi": 2.29941}, "gini_cpc": 0.048564, "gini_cps": 0.058268, "bpic": 4.842267, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9613323758493069, "fi": 1.1909621260001357}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 226.9815216064453, "n_util_obs": 726, "util_count_spearman": 0.13421434302138588}, "t": 2214.0} |
| {"event": "eval", "step": 3000, "phase": 2, "bpb": {"en": 2.00988, "tr": 2.18598, "fi": 2.30207}, "bytes_per_chunk": {"en": 4.835, "tr": 5.449, "fi": 4.486}, "bpb_mean": 2.16598, "frozen_inventory": true, "t": 2214.9} |
| {"event": "train", "step": 3000, "arch": "lexmem_v1", "phase": 2, "objective": "global", "lm_loss": 1.48225, "bpb": 2.13844, "chunk_loss": 0.030473, "F": 0.20218, "G": 0.21083, "chunks": 207.03, "bytes_per_chunk": 4.946, "boundary_entropy": 0.11268, "router_noise": 0.0, "grad_norm": 0.7209, "lr": 0.0001761, "router_trainable": true, "sec_per_step": 0.755, "probe_sec": 0.0862, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04812, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.13037, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 1.05455, "trunc_abs_frac": 3.31113, "mean_delta_full": -0.11515, "read_hit_rate": {"surface": 0.70499}, "mean_gate": {"surface": 0.30562}, "n_reads": 4648, "n_cand": 6593, "read_suppressed_frac": 0.0957, "dropout_reads_logged": 4648, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 227.04881286621094, "n_util_obs": 726, "util_count_spearman": 0.13515188186552374}, "probe_overhead_sec_total": 184.283, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.017539, "alpha": 0.03, "t": 2218.6} |
| {"event": "consolidate", "step": 3050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2266.5} |
| {"event": "train", "step": 3050, "arch": "lexmem_v1", "phase": 2, "objective": "global", "lm_loss": 1.43041, "bpb": 2.06365, "chunk_loss": 0.030572, "F": 0.20688, "G": 0.21462, "chunks": 211.84, "bytes_per_chunk": 4.834, "boundary_entropy": 0.10834, "router_noise": 0.0, "grad_norm": 0.6582, "lr": 0.0001724, "router_trainable": true, "sec_per_step": 0.974, "probe_sec": 0.0705, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04615, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.26287, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.51603, "trunc_abs_frac": 0.65174, "mean_delta_full": 0.79449, "read_hit_rate": {"surface": 0.72388}, "mean_gate": {"surface": 0.32536}, "n_reads": 4884, "n_cand": 6747, "read_suppressed_frac": 0.0962, "dropout_reads_logged": 4884, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 230.7264404296875, "n_util_obs": 737, "util_count_spearman": 0.14974049193484207}, "probe_overhead_sec_total": 187.689, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.020053, "alpha": 0.03, "t": 2267.3} |
| {"event": "consolidate", "step": 3100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2297.9} |
| {"event": "train", "step": 3100, "arch": "lexmem_v1", "phase": 2, "objective": "global", "lm_loss": 1.48452, "bpb": 2.14172, "chunk_loss": 0.030357, "F": 0.19684, "G": 0.20773, "chunks": 201.56, "bytes_per_chunk": 5.08, "boundary_entropy": 0.11328, "router_noise": 0.0, "grad_norm": 0.6804, "lr": 0.0001687, "router_trainable": true, "sec_per_step": 0.63, "probe_sec": 0.0862, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05294, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.21867, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 1.02988, "trunc_abs_frac": 0.95343, "mean_delta_full": 0.46292, "read_hit_rate": {"surface": 0.7105}, "mean_gate": {"surface": 0.31839}, "n_reads": 4560, "n_cand": 6418, "read_suppressed_frac": 0.0952, "dropout_reads_logged": 4560, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 234.36439514160156, "n_util_obs": 756, "util_count_spearman": 0.16049216894719354}, "probe_overhead_sec_total": 190.329, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.018175, "alpha": 0.03, "t": 2298.8} |
| {"event": "consolidate", "step": 3150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2332.3} |
| {"event": "train", "step": 3150, "arch": "lexmem_v1", "phase": 2, "objective": "global", "lm_loss": 1.44699, "bpb": 2.08757, "chunk_loss": 0.030491, "F": 0.20563, "G": 0.2138, "chunks": 210.56, "bytes_per_chunk": 4.863, "boundary_entropy": 0.11203, "router_noise": 0.0, "grad_norm": 0.7216, "lr": 0.000165, "router_trainable": true, "sec_per_step": 0.684, "probe_sec": 0.044, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02265, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.15761, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 1.21213, "trunc_abs_frac": 1.10923, "mean_delta_full": 0.61768, "read_hit_rate": {"surface": 0.72472}, "mean_gate": {"surface": 0.29523}, "n_reads": 4860, "n_cand": 6706, "read_suppressed_frac": 0.0969, "dropout_reads_logged": 4860, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 238.06204223632812, "n_util_obs": 765, "util_count_spearman": 0.16369105111788174}, "probe_overhead_sec_total": 193.231, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.021761, "alpha": 0.03, "t": 2333.1} |
| {"event": "consolidate", "step": 3200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2370.5} |
| {"event": "probe", "step": 3200, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3593.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.206459, "cps": 28.070312, "mean_p": 0.214633, "boundary_entropy": 0.107709, "bpb": 1.978198}, "tr": {"chunks": 3442.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.190134, "cps": 26.890625, "mean_p": 0.191954, "boundary_entropy": 0.123801, "bpb": 2.153622}, "fi": {"chunks": 4496.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.239009, "cps": 35.125, "mean_p": 0.239797, "boundary_entropy": 0.113058, "bpb": 2.279377}}, "bpb": {"en": 1.9782, "tr": 2.15362, "fi": 2.27938}, "gini_cpc": 0.051264, "gini_cps": 0.060937, "bpic": 4.919955, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9209312657036777, "fi": 1.1576607850488319}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 241.64048767089844, "n_util_obs": 776, "util_count_spearman": 0.16686310453147002}, "t": 2371.1} |
| {"event": "eval", "step": 3200, "phase": 2, "bpb": {"en": 1.99145, "tr": 2.16078, "fi": 2.28545}, "bytes_per_chunk": {"en": 4.871, "tr": 5.774, "fi": 4.601}, "bpb_mean": 2.14589, "frozen_inventory": true, "t": 2371.9} |
| {"event": "train", "step": 3200, "arch": "lexmem_v1", "phase": 2, "objective": "global", "lm_loss": 1.43161, "bpb": 2.06537, "chunk_loss": 0.030434, "F": 0.20297, "G": 0.21217, "chunks": 207.84, "bytes_per_chunk": 4.927, "boundary_entropy": 0.10587, "router_noise": 0.0, "grad_norm": 0.6223, "lr": 0.0001613, "router_trainable": true, "sec_per_step": 0.812, "probe_sec": 0.0465, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03785, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.15995, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -0.60017, "trunc_abs_frac": 1.05513, "mean_delta_full": -0.15967, "read_hit_rate": {"surface": 0.72458}, "mean_gate": {"surface": 0.31458}, "n_reads": 4796, "n_cand": 6619, "read_suppressed_frac": 0.1049, "dropout_reads_logged": 4796, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 241.71441650390625, "n_util_obs": 776, "util_count_spearman": 0.16686830582513346}, "probe_overhead_sec_total": 196.086, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.020604, "alpha": 0.03, "t": 2373.6} |
| {"event": "consolidate", "step": 3250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2412.1} |
| {"event": "train", "step": 3250, "arch": "lexmem_v1", "phase": 2, "objective": "global", "lm_loss": 1.42947, "bpb": 2.06228, "chunk_loss": 0.030562, "F": 0.20865, "G": 0.21646, "chunks": 213.66, "bytes_per_chunk": 4.793, "boundary_entropy": 0.10756, "router_noise": 0.0, "grad_norm": 0.6976, "lr": 0.0001576, "router_trainable": true, "sec_per_step": 0.781, "probe_sec": 0.0284, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0177, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.14783, "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": 0.35294, "trunc_abs_frac": 0.40155, "mean_delta_full": 0.28402, "read_hit_rate": {"surface": 0.73843}, "mean_gate": {"surface": 0.31623}, "n_reads": 5025, "n_cand": 6805, "read_suppressed_frac": 0.0983, "dropout_reads_logged": 5025, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 245.38320922851562, "n_util_obs": 791, "util_count_spearman": 0.1678530675506022}, "probe_overhead_sec_total": 198.883, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.015728, "alpha": 0.03, "t": 2412.7} |
| {"event": "consolidate", "step": 3300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2443.5} |
| {"event": "train", "step": 3300, "arch": "lexmem_v1", "phase": 2, "objective": "global", "lm_loss": 1.40711, "bpb": 2.03003, "chunk_loss": 0.030544, "F": 0.20837, "G": 0.21683, "chunks": 213.38, "bytes_per_chunk": 4.799, "boundary_entropy": 0.10429, "router_noise": 0.0, "grad_norm": 0.6614, "lr": 0.0001539, "router_trainable": true, "sec_per_step": 0.622, "probe_sec": 0.0512, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03992, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.22809, "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": 0.44514, "trunc_abs_frac": 0.90702, "mean_delta_full": 0.58716, "read_hit_rate": {"surface": 0.72543}, "mean_gate": {"surface": 0.29586}, "n_reads": 4930, "n_cand": 6796, "read_suppressed_frac": 0.0992, "dropout_reads_logged": 4930, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 249.04298400878906, "n_util_obs": 806, "util_count_spearman": 0.1706656947903494}, "probe_overhead_sec_total": 202.162, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.019608, "alpha": 0.03, "t": 2443.8} |
| {"event": "consolidate", "step": 3350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2486.8} |
| {"event": "train", "step": 3350, "arch": "lexmem_v1", "phase": 2, "objective": "global", "lm_loss": 1.35191, "bpb": 1.9504, "chunk_loss": 0.030448, "F": 0.20392, "G": 0.21133, "chunks": 208.81, "bytes_per_chunk": 4.904, "boundary_entropy": 0.10612, "router_noise": 0.0, "grad_norm": 0.7391, "lr": 0.0001501, "router_trainable": true, "sec_per_step": 0.878, "probe_sec": 0.0491, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02176, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.15448, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.95777, "trunc_abs_frac": 1.05224, "mean_delta_full": 0.87805, "read_hit_rate": {"surface": 0.73263}, "mean_gate": {"surface": 0.30368}, "n_reads": 4872, "n_cand": 6650, "read_suppressed_frac": 0.1051, "dropout_reads_logged": 4872, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 252.71377563476562, "n_util_obs": 816, "util_count_spearman": 0.17670351037168355}, "probe_overhead_sec_total": 205.589, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.018349, "alpha": 0.03, "t": 2487.7} |
| {"event": "consolidate", "step": 3400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2529.2} |
| {"event": "probe", "step": 3400, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3614.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.207665, "cps": 28.234375, "mean_p": 0.214314, "boundary_entropy": 0.100216, "bpb": 1.96764}, "tr": {"chunks": 3353.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.185218, "cps": 26.195312, "mean_p": 0.189728, "boundary_entropy": 0.134079, "bpb": 2.150827}, "fi": {"chunks": 4587.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.243847, "cps": 35.835938, "mean_p": 0.243223, "boundary_entropy": 0.109884, "bpb": 2.244207}}, "bpb": {"en": 1.96764, "tr": 2.15083, "fi": 2.24421}, "gini_cpc": 0.061385, "gini_cps": 0.071202, "bpic": 4.910161, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.891905770952608, "fi": 1.1742290721877133}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 256.29608154296875, "n_util_obs": 830, "util_count_spearman": 0.17591901410444263}, "t": 2529.9} |
| {"event": "eval", "step": 3400, "phase": 2, "bpb": {"en": 1.98043, "tr": 2.16013, "fi": 2.24909}, "bytes_per_chunk": {"en": 4.818, "tr": 5.954, "fi": 4.474}, "bpb_mean": 2.12988, "frozen_inventory": true, "t": 2530.8} |
| {"event": "train", "step": 3400, "arch": "lexmem_v1", "phase": 2, "objective": "global", "lm_loss": 1.41528, "bpb": 2.04182, "chunk_loss": 0.030449, "F": 0.20148, "G": 0.21062, "chunks": 206.31, "bytes_per_chunk": 4.963, "boundary_entropy": 0.10398, "router_noise": 0.0, "grad_norm": 0.7125, "lr": 0.0001465, "router_trainable": true, "sec_per_step": 0.904, "probe_sec": 0.0527, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02556, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.14199, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 1.03587, "trunc_abs_frac": 0.63404, "mean_delta_full": 0.08609, "read_hit_rate": {"surface": 0.71142}, "mean_gate": {"surface": 0.31068}, "n_reads": 4674, "n_cand": 6570, "read_suppressed_frac": 0.0993, "dropout_reads_logged": 4674, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 256.36798095703125, "n_util_obs": 830, "util_count_spearman": 0.1753536556358558}, "probe_overhead_sec_total": 208.41, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.021962, "alpha": 0.03, "t": 2532.9} |
| {"event": "consolidate", "step": 3450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2555.2} |
| {"event": "train", "step": 3450, "arch": "lexmem_v1", "phase": 2, "objective": "global", "lm_loss": 1.38011, "bpb": 1.99108, "chunk_loss": 0.030449, "F": 0.20465, "G": 0.2131, "chunks": 209.56, "bytes_per_chunk": 4.886, "boundary_entropy": 0.10732, "router_noise": 0.0, "grad_norm": 0.7695, "lr": 0.0001428, "router_trainable": true, "sec_per_step": 0.452, "probe_sec": 0.0361, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02025, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.04266, "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": 0.28416, "trunc_abs_frac": 0.31847, "mean_delta_full": 0.39591, "read_hit_rate": {"surface": 0.72101}, "mean_gate": {"surface": 0.30171}, "n_reads": 4812, "n_cand": 6674, "read_suppressed_frac": 0.1016, "dropout_reads_logged": 4812, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 259.99853515625, "n_util_obs": 843, "util_count_spearman": 0.1718598929888316}, "probe_overhead_sec_total": 211.059, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.020339, "alpha": 0.03, "t": 2555.5} |
| {"event": "consolidate", "step": 3500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2591.6} |
| {"event": "train", "step": 3500, "arch": "lexmem_v1", "phase": 2, "objective": "global", "lm_loss": 1.36279, "bpb": 1.96608, "chunk_loss": 0.030309, "F": 0.19748, "G": 0.20461, "chunks": 202.22, "bytes_per_chunk": 5.064, "boundary_entropy": 0.1001, "router_noise": 0.0, "grad_norm": 0.6814, "lr": 0.0001391, "router_trainable": true, "sec_per_step": 0.742, "probe_sec": 0.0477, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02329, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.01145, "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": -0.17918, "trunc_abs_frac": 6.26854, "mean_delta_full": -0.18343, "read_hit_rate": {"surface": 0.7217}, "mean_gate": {"surface": 0.30139}, "n_reads": 4647, "n_cand": 6439, "read_suppressed_frac": 0.0981, "dropout_reads_logged": 4647, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 263.6689758300781, "n_util_obs": 848, "util_count_spearman": 0.1731501808572313}, "probe_overhead_sec_total": 214.174, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.023212, "alpha": 0.03, "t": 2592.6} |
| {"event": "consolidate", "step": 3550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2630.5} |
| {"event": "train", "step": 3550, "arch": "lexmem_v1", "phase": 2, "objective": "global", "lm_loss": 1.40363, "bpb": 2.02501, "chunk_loss": 0.030349, "F": 0.19522, "G": 0.20506, "chunks": 199.91, "bytes_per_chunk": 5.122, "boundary_entropy": 0.1062, "router_noise": 0.0, "grad_norm": 0.6642, "lr": 0.0001355, "router_trainable": true, "sec_per_step": 0.77, "probe_sec": 0.0398, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02194, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.18457, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.46127, "trunc_abs_frac": 0.66152, "mean_delta_full": -0.25946, "read_hit_rate": {"surface": 0.69631}, "mean_gate": {"surface": 0.30311}, "n_reads": 4432, "n_cand": 6365, "read_suppressed_frac": 0.09, "dropout_reads_logged": 4432, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 267.3263854980469, "n_util_obs": 862, "util_count_spearman": 0.18308330521285662}, "probe_overhead_sec_total": 217.127, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.018314, "alpha": 0.03, "t": 2631.1} |
| {"event": "consolidate", "step": 3600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2672.5} |
| {"event": "probe", "step": 3600, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3549.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.20393, "cps": 27.726562, "mean_p": 0.209795, "boundary_entropy": 0.101848, "bpb": 1.951953}, "tr": {"chunks": 3433.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.189637, "cps": 26.820312, "mean_p": 0.190079, "boundary_entropy": 0.119814, "bpb": 2.1135}, "fi": {"chunks": 4492.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.238796, "cps": 35.09375, "mean_p": 0.237396, "boundary_entropy": 0.106503, "bpb": 2.227213}}, "bpb": {"en": 1.95195, "tr": 2.1135, "fi": 2.22721}, "gini_cpc": 0.051826, "gini_cps": 0.06153, "bpic": 4.944396, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9299109738773373, "fi": 1.1709705832539035}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 270.9185791015625, "n_util_obs": 875, "util_count_spearman": 0.18387915358057103}, "t": 2673.3} |
| {"event": "eval", "step": 3600, "phase": 2, "bpb": {"en": 1.96458, "tr": 2.12189, "fi": 2.23292}, "bytes_per_chunk": {"en": 4.824, "tr": 5.749, "fi": 4.615}, "bpb_mean": 2.10646, "frozen_inventory": true, "t": 2674.2} |
| {"event": "train", "step": 3600, "arch": "lexmem_v1", "phase": 2, "objective": "global", "lm_loss": 1.39832, "bpb": 2.01735, "chunk_loss": 0.030486, "F": 0.20554, "G": 0.21113, "chunks": 210.47, "bytes_per_chunk": 4.865, "boundary_entropy": 0.10147, "router_noise": 0.0, "grad_norm": 0.6184, "lr": 0.0001319, "router_trainable": true, "sec_per_step": 0.909, "probe_sec": 0.0458, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02311, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.13778, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 1.64315, "trunc_abs_frac": 1.46351, "mean_delta_full": -0.15213, "read_hit_rate": {"surface": 0.71401}, "mean_gate": {"surface": 0.28892}, "n_reads": 4786, "n_cand": 6703, "read_suppressed_frac": 0.1055, "dropout_reads_logged": 4786, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 270.9913635253906, "n_util_obs": 875, "util_count_spearman": 0.18374166488746527}, "probe_overhead_sec_total": 220.023, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.019418, "alpha": 0.03, "t": 2676.6} |
| {"event": "consolidate", "step": 3650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2708.8} |
| {"event": "train", "step": 3650, "arch": "lexmem_v1", "phase": 2, "objective": "global", "lm_loss": 1.43719, "bpb": 2.07342, "chunk_loss": 0.030568, "F": 0.20654, "G": 0.21456, "chunks": 211.5, "bytes_per_chunk": 4.842, "boundary_entropy": 0.10561, "router_noise": 0.0, "grad_norm": 0.7263, "lr": 0.0001283, "router_trainable": true, "sec_per_step": 0.668, "probe_sec": 0.077, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02969, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.04494, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.28891, "trunc_abs_frac": 19.68645, "mean_delta_full": 0.15295, "read_hit_rate": {"surface": 0.71882}, "mean_gate": {"surface": 0.28503}, "n_reads": 4842, "n_cand": 6736, "read_suppressed_frac": 0.0995, "dropout_reads_logged": 4842, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 274.6850280761719, "n_util_obs": 883, "util_count_spearman": 0.1857206256001008}, "probe_overhead_sec_total": 223.157, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.018715, "alpha": 0.03, "t": 2710.0} |
| {"event": "consolidate", "step": 3700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2746.3} |
| {"event": "train", "step": 3700, "arch": "lexmem_v1", "phase": 2, "objective": "global", "lm_loss": 1.38535, "bpb": 1.99864, "chunk_loss": 0.030447, "F": 0.20401, "G": 0.21139, "chunks": 208.91, "bytes_per_chunk": 4.902, "boundary_entropy": 0.10457, "router_noise": 0.0, "grad_norm": 0.6358, "lr": 0.0001247, "router_trainable": true, "sec_per_step": 0.732, "probe_sec": 0.0361, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02489, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.06789, "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": 0.798, "trunc_abs_frac": 1.03916, "mean_delta_full": -0.64412, "read_hit_rate": {"surface": 0.73065}, "mean_gate": {"surface": 0.28538}, "n_reads": 4861, "n_cand": 6653, "read_suppressed_frac": 0.1014, "dropout_reads_logged": 4861, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 278.37567138671875, "n_util_obs": 903, "util_count_spearman": 0.19440675408030028}, "probe_overhead_sec_total": 226.287, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.020326, "alpha": 0.03, "t": 2746.6} |
| {"event": "consolidate", "step": 3750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2777.4} |
| {"event": "train", "step": 3750, "arch": "lexmem_v1", "phase": 2, "objective": "global", "lm_loss": 1.37633, "bpb": 1.98563, "chunk_loss": 0.030402, "F": 0.20071, "G": 0.20881, "chunks": 205.53, "bytes_per_chunk": 4.982, "boundary_entropy": 0.10364, "router_noise": 0.0, "grad_norm": 0.6784, "lr": 0.0001212, "router_trainable": true, "sec_per_step": 0.633, "probe_sec": 0.0499, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02484, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.20982, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.94491, "trunc_abs_frac": 1.05036, "mean_delta_full": 0.46002, "read_hit_rate": {"surface": 0.71948}, "mean_gate": {"surface": 0.28688}, "n_reads": 4709, "n_cand": 6545, "read_suppressed_frac": 0.0996, "dropout_reads_logged": 4709, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 282.00958251953125, "n_util_obs": 909, "util_count_spearman": 0.19255342661622787}, "probe_overhead_sec_total": 228.921, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.022986, "alpha": 0.03, "t": 2778.2} |
| {"event": "consolidate", "step": 3800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2821.7} |
| {"event": "probe", "step": 3800, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3532.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.202954, "cps": 27.59375, "mean_p": 0.210283, "boundary_entropy": 0.099826, "bpb": 1.940202}, "tr": {"chunks": 3384.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.18693, "cps": 26.4375, "mean_p": 0.186167, "boundary_entropy": 0.119088, "bpb": 2.103317}, "fi": {"chunks": 4506.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.239541, "cps": 35.203125, "mean_p": 0.237832, "boundary_entropy": 0.101671, "bpb": 2.208308}}, "bpb": {"en": 1.9402, "tr": 2.10332, "fi": 2.20831}, "gini_cpc": 0.055723, "gini_cps": 0.065488, "bpic": 4.966906, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9210500452644548, "fi": 1.1802736983186086}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 285.5693664550781, "n_util_obs": 921, "util_count_spearman": 0.19677970761497995}, "t": 2822.3} |
| {"event": "eval", "step": 3800, "phase": 2, "bpb": {"en": 1.95305, "tr": 2.11318, "fi": 2.21328}, "bytes_per_chunk": {"en": 4.844, "tr": 5.903, "fi": 4.599}, "bpb_mean": 2.09317, "frozen_inventory": true, "t": 2823.6} |
| {"event": "train", "step": 3800, "arch": "lexmem_v1", "phase": 2, "objective": "global", "lm_loss": 1.44575, "bpb": 2.08578, "chunk_loss": 0.030498, "F": 0.20572, "G": 0.21256, "chunks": 210.66, "bytes_per_chunk": 4.861, "boundary_entropy": 0.09811, "router_noise": 0.0, "grad_norm": 0.6707, "lr": 0.0001177, "router_trainable": true, "sec_per_step": 0.955, "probe_sec": 0.0341, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01844, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.13376, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 1.12753, "trunc_abs_frac": 1.46023, "mean_delta_full": 0.09586, "read_hit_rate": {"surface": 0.70741}, "mean_gate": {"surface": 0.26973}, "n_reads": 4746, "n_cand": 6709, "read_suppressed_frac": 0.106, "dropout_reads_logged": 4746, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 285.64178466796875, "n_util_obs": 921, "util_count_spearman": 0.19649485867925603}, "probe_overhead_sec_total": 231.969, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.018172, "alpha": 0.03, "t": 2826.0} |
| {"event": "consolidate", "step": 3850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2866.8} |
| {"event": "train", "step": 3850, "arch": "lexmem_v1", "phase": 2, "objective": "global", "lm_loss": 1.54098, "bpb": 2.22316, "chunk_loss": 0.030753, "F": 0.20691, "G": 0.21562, "chunks": 211.88, "bytes_per_chunk": 4.833, "boundary_entropy": 0.1035, "router_noise": 0.0, "grad_norm": 0.6929, "lr": 0.0001142, "router_trainable": true, "sec_per_step": 0.838, "probe_sec": 0.0405, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03049, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.19636, "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": 0.74037, "trunc_abs_frac": 0.79879, "mean_delta_full": 0.37606, "read_hit_rate": {"surface": 0.70465}, "mean_gate": {"surface": 0.29581}, "n_reads": 4755, "n_cand": 6748, "read_suppressed_frac": 0.1016, "dropout_reads_logged": 4755, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 289.3359375, "n_util_obs": 933, "util_count_spearman": 0.18859349225956992}, "probe_overhead_sec_total": 234.861, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.017075, "alpha": 0.03, "t": 2867.9} |
| {"event": "consolidate", "step": 3900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2902.8} |
| {"event": "train", "step": 3900, "arch": "lexmem_v1", "phase": 2, "objective": "global", "lm_loss": 1.42727, "bpb": 2.05911, "chunk_loss": 0.030573, "F": 0.20688, "G": 0.21302, "chunks": 211.84, "bytes_per_chunk": 4.834, "boundary_entropy": 0.09778, "router_noise": 0.0, "grad_norm": 0.6807, "lr": 0.0001108, "router_trainable": true, "sec_per_step": 0.7, "probe_sec": 0.0292, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01974, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.06605, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -0.38876, "trunc_abs_frac": 0.42226, "mean_delta_full": -0.15073, "read_hit_rate": {"surface": 0.72106}, "mean_gate": {"surface": 0.27078}, "n_reads": 4865, "n_cand": 6747, "read_suppressed_frac": 0.0985, "dropout_reads_logged": 4865, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 292.97784423828125, "n_util_obs": 947, "util_count_spearman": 0.19325957445389938}, "probe_overhead_sec_total": 237.91, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.017689, "alpha": 0.03, "t": 2902.9} |
| {"event": "consolidate", "step": 3950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2920.1} |
| {"event": "train", "step": 3950, "arch": "lexmem_v1", "phase": 2, "objective": "global", "lm_loss": 1.48474, "bpb": 2.14203, "chunk_loss": 0.030467, "F": 0.1994, "G": 0.20711, "chunks": 204.19, "bytes_per_chunk": 5.015, "boundary_entropy": 0.0996, "router_noise": 0.0, "grad_norm": 0.7633, "lr": 0.0001074, "router_trainable": true, "sec_per_step": 0.363, "probe_sec": 0.0497, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01731, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.04652, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.83556, "trunc_abs_frac": 0.83556, "mean_delta_full": -0.15515, "read_hit_rate": {"surface": 0.70132}, "mean_gate": {"surface": 0.27982}, "n_reads": 4560, "n_cand": 6502, "read_suppressed_frac": 0.1059, "dropout_reads_logged": 4560, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 296.6395263671875, "n_util_obs": 956, "util_count_spearman": 0.18394506960251555}, "probe_overhead_sec_total": 240.236, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.017242, "alpha": 0.03, "t": 2921.0} |
| {"event": "consolidate", "step": 4000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2951.8} |
| {"event": "probe", "step": 4000, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3574.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.205367, "cps": 27.921875, "mean_p": 0.210279, "boundary_entropy": 0.097362, "bpb": 1.928454}, "tr": {"chunks": 3457.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.190963, "cps": 27.007812, "mean_p": 0.19223, "boundary_entropy": 0.117133, "bpb": 2.08608}, "fi": {"chunks": 4507.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.239594, "cps": 35.210938, "mean_p": 0.238031, "boundary_entropy": 0.09861, "bpb": 2.186484}}, "bpb": {"en": 1.92845, "tr": 2.08608, "fi": 2.18648}, "gini_cpc": 0.050982, "gini_cps": 0.060669, "bpic": 4.91697, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9298617860411453, "fi": 1.1666625217233948}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 300.2243347167969, "n_util_obs": 972, "util_count_spearman": 0.17781788628059605}, "t": 2952.4} |
| {"event": "eval", "step": 4000, "phase": 2, "bpb": {"en": 1.94182, "tr": 2.09511, "fi": 2.19217}, "bytes_per_chunk": {"en": 4.84, "tr": 5.747, "fi": 4.594}, "bpb_mean": 2.07637, "frozen_inventory": true, "t": 2953.1} |
| {"event": "train", "step": 4000, "arch": "lexmem_v1", "phase": 2, "objective": "global", "lm_loss": 1.40666, "bpb": 2.02938, "chunk_loss": 0.030518, "F": 0.20462, "G": 0.2116, "chunks": 209.53, "bytes_per_chunk": 4.887, "boundary_entropy": 0.09785, "router_noise": 0.0, "grad_norm": 0.6782, "lr": 0.000104, "router_trainable": true, "sec_per_step": 0.666, "probe_sec": 0.0515, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02796, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.08216, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.09002, "trunc_abs_frac": 0.82636, "mean_delta_full": -0.23774, "read_hit_rate": {"surface": 0.72726}, "mean_gate": {"surface": 0.28396}, "n_reads": 4853, "n_cand": 6673, "read_suppressed_frac": 0.1016, "dropout_reads_logged": 4853, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 300.29949951171875, "n_util_obs": 972, "util_count_spearman": 0.17807314354138845}, "probe_overhead_sec_total": 243.298, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.016356, "alpha": 0.03, "t": 2954.3} |
| {"event": "consolidate", "step": 4050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2973.6} |
| {"event": "train", "step": 4050, "arch": "lexmem_v1", "phase": 2, "objective": "global", "lm_loss": 1.4194, "bpb": 2.04777, "chunk_loss": 0.030508, "F": 0.2059, "G": 0.21301, "chunks": 210.84, "bytes_per_chunk": 4.857, "boundary_entropy": 0.09851, "router_noise": 0.0, "grad_norm": 0.6954, "lr": 0.0001007, "router_trainable": true, "sec_per_step": 0.391, "probe_sec": 0.0305, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02333, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.09289, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 1.36186, "trunc_abs_frac": 0.66959, "mean_delta_full": 0.40104, "read_hit_rate": {"surface": 0.71988}, "mean_gate": {"surface": 0.28899}, "n_reads": 4834, "n_cand": 6715, "read_suppressed_frac": 0.1086, "dropout_reads_logged": 4834, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 303.96044921875, "n_util_obs": 979, "util_count_spearman": 0.18138334479076243}, "probe_overhead_sec_total": 245.945, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.021258, "alpha": 0.03, "t": 2973.9} |
| {"event": "probe", "step": 4096, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3517.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.202092, "cps": 27.476562, "mean_p": 0.209018, "boundary_entropy": 0.095554, "bpb": 1.923176}, "tr": {"chunks": 3399.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.187759, "cps": 26.554688, "mean_p": 0.188018, "boundary_entropy": 0.114791, "bpb": 2.081132}, "fi": {"chunks": 4496.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.239009, "cps": 35.125, "mean_p": 0.237089, "boundary_entropy": 0.095175, "bpb": 2.181668}}, "bpb": {"en": 1.92318, "tr": 2.08113, "fi": 2.18167}, "gini_cpc": 0.054331, "gini_cps": 0.064085, "bpic": 4.971258, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9290784036143854, "fi": 1.182677054501124}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 307.25628662109375, "n_util_obs": 987, "util_count_spearman": 0.17671379661344652}, "t": 3000.7} |
| {"event": "consolidate", "step": 4100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3005.0} |
| {"event": "train", "step": 4100, "arch": "lexmem_v1", "phase": 2, "objective": "global", "lm_loss": 1.35278, "bpb": 1.95165, "chunk_loss": 0.030494, "F": 0.20407, "G": 0.21075, "chunks": 208.97, "bytes_per_chunk": 4.9, "boundary_entropy": 0.09516, "router_noise": 0.0, "grad_norm": 0.6952, "lr": 9.75e-05, "router_trainable": true, "sec_per_step": 0.627, "probe_sec": 0.043, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03596, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.12542, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.36404, "trunc_abs_frac": 0.34881, "mean_delta_full": 0.4133, "read_hit_rate": {"surface": 0.73704}, "mean_gate": {"surface": 0.26669}, "n_reads": 4905, "n_cand": 6655, "read_suppressed_frac": 0.1046, "dropout_reads_logged": 4905, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 307.625, "n_util_obs": 987, "util_count_spearman": 0.1765628306978255}, "probe_overhead_sec_total": 248.525, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.020755, "alpha": 0.03, "t": 3005.2} |
| {"event": "consolidate", "step": 4150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3028.7} |
| {"event": "train", "step": 4150, "arch": "lexmem_v1", "phase": 2, "objective": "global", "lm_loss": 1.42246, "bpb": 2.05217, "chunk_loss": 0.030563, "F": 0.20709, "G": 0.21435, "chunks": 212.06, "bytes_per_chunk": 4.829, "boundary_entropy": 0.09451, "router_noise": 0.0, "grad_norm": 0.7642, "lr": 9.43e-05, "router_trainable": true, "sec_per_step": 0.485, "probe_sec": 0.0458, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03174, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.42503, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.62589, "trunc_abs_frac": 0.40196, "mean_delta_full": 0.23248, "read_hit_rate": {"surface": 0.71676}, "mean_gate": {"surface": 0.26042}, "n_reads": 4841, "n_cand": 6754, "read_suppressed_frac": 0.0954, "dropout_reads_logged": 4841, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 311.29327392578125, "n_util_obs": 996, "util_count_spearman": 0.1737992160053516}, "probe_overhead_sec_total": 250.928, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.017818, "alpha": 0.03, "t": 3029.4} |
| {"event": "consolidate", "step": 4200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3052.8} |
| {"event": "probe", "step": 4200, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3551.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.204045, "cps": 27.742188, "mean_p": 0.209641, "boundary_entropy": 0.097245, "bpb": 1.921061}, "tr": {"chunks": 3411.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.188422, "cps": 26.648438, "mean_p": 0.188075, "boundary_entropy": 0.114129, "bpb": 2.070419}, "fi": {"chunks": 4489.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.238637, "cps": 35.070312, "mean_p": 0.236536, "boundary_entropy": 0.09692, "bpb": 2.174893}}, "bpb": {"en": 1.92106, "tr": 2.07042, "fi": 2.17489}, "gini_cpc": 0.053045, "gini_cps": 0.06276, "bpic": 4.954327, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.923431352864541, "fi": 1.1695294704122337}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 314.9080810546875, "n_util_obs": 1007, "util_count_spearman": 0.16259023881286505}, "t": 3053.3} |
| {"event": "eval", "step": 4200, "phase": 2, "bpb": {"en": 1.93384, "tr": 2.07954, "fi": 2.18127}, "bytes_per_chunk": {"en": 4.856, "tr": 5.81, "fi": 4.606}, "bpb_mean": 2.06489, "frozen_inventory": true, "t": 3054.1} |
| {"event": "phase_transition", "step": 4200, "phase": 3, "objective": "global", "router_trainable": true, "encoder_trainable": true, "frozen_router_params": 0, "t": 3055.8} |
| {"event": "train", "step": 4200, "arch": "lexmem_v1", "phase": 3, "objective": "global", "lm_loss": 1.34689, "bpb": 1.94315, "chunk_loss": 0.030295, "F": 0.19495, "G": 0.2026, "chunks": 199.62, "bytes_per_chunk": 5.13, "boundary_entropy": 0.09549, "router_noise": 0.0, "grad_norm": 0.6719, "lr": 9.12e-05, "router_trainable": true, "sec_per_step": 0.541, "probe_sec": 0.0858, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03775, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.13834, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.90813, "trunc_abs_frac": 1.3903, "mean_delta_full": 0.17124, "read_hit_rate": {"surface": 0.71271}, "mean_gate": {"surface": 0.27711}, "n_reads": 4530, "n_cand": 6356, "read_suppressed_frac": 0.1035, "dropout_reads_logged": 4530, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 314.97686767578125, "n_util_obs": 1008, "util_count_spearman": 0.16170108194409588}, "probe_overhead_sec_total": 253.331, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.018853, "alpha": 0.03, "t": 3056.5} |
| {"event": "consolidate", "step": 4250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3083.2} |
| {"event": "train", "step": 4250, "arch": "lexmem_v1", "phase": 3, "objective": "global", "lm_loss": 1.37898, "bpb": 1.98944, "chunk_loss": 0.030525, "F": 0.20749, "G": 0.21427, "chunks": 212.47, "bytes_per_chunk": 4.82, "boundary_entropy": 0.09566, "router_noise": 0.0, "grad_norm": 0.6527, "lr": 8.81e-05, "router_trainable": true, "sec_per_step": 0.547, "probe_sec": 0.0706, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06091, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.39599, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 1.34225, "trunc_abs_frac": 1.03953, "mean_delta_full": 0.44128, "read_hit_rate": {"surface": 0.72499}, "mean_gate": {"surface": 0.26922}, "n_reads": 4906, "n_cand": 6767, "read_suppressed_frac": 0.0991, "dropout_reads_logged": 4906, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 318.6830139160156, "n_util_obs": 1029, "util_count_spearman": 0.16167535979495126}, "probe_overhead_sec_total": 256.093, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.017243, "alpha": 0.03, "t": 3083.8} |
| {"event": "consolidate", "step": 4300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3111.2} |
| {"event": "train", "step": 4300, "arch": "lexmem_v1", "phase": 3, "objective": "global", "lm_loss": 1.38543, "bpb": 1.99875, "chunk_loss": 0.030352, "F": 0.19897, "G": 0.20568, "chunks": 203.75, "bytes_per_chunk": 5.026, "boundary_entropy": 0.09394, "router_noise": 0.0, "grad_norm": 0.7653, "lr": 8.5e-05, "router_trainable": true, "sec_per_step": 0.553, "probe_sec": 0.0319, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02124, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.04972, "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": 0.60054, "trunc_abs_frac": 3.16999, "mean_delta_full": 0.2915, "read_hit_rate": {"surface": 0.71239}, "mean_gate": {"surface": 0.26501}, "n_reads": 4622, "n_cand": 6488, "read_suppressed_frac": 0.1, "dropout_reads_logged": 4622, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 322.3778991699219, "n_util_obs": 1041, "util_count_spearman": 0.16194736136995436}, "probe_overhead_sec_total": 258.703, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.017492, "alpha": 0.03, "t": 3111.5} |
| {"event": "consolidate", "step": 4350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3137.3} |
| {"event": "train", "step": 4350, "arch": "lexmem_v1", "phase": 3, "objective": "global", "lm_loss": 1.4013, "bpb": 2.02164, "chunk_loss": 0.030456, "F": 0.20514, "G": 0.2108, "chunks": 210.06, "bytes_per_chunk": 4.875, "boundary_entropy": 0.09261, "router_noise": 0.0, "grad_norm": 0.659, "lr": 8.21e-05, "router_trainable": true, "sec_per_step": 0.534, "probe_sec": 0.0921, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05182, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.06901, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.07287, "trunc_abs_frac": 0.97859, "mean_delta_full": -0.26196, "read_hit_rate": {"surface": 0.7145}, "mean_gate": {"surface": 0.26343}, "n_reads": 4780, "n_cand": 6690, "read_suppressed_frac": 0.0897, "dropout_reads_logged": 4780, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 326.04290771484375, "n_util_obs": 1052, "util_count_spearman": 0.1663832381848113}, "probe_overhead_sec_total": 261.593, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.019381, "alpha": 0.03, "t": 3138.2} |
| {"event": "consolidate", "step": 4400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3168.5} |
| {"event": "probe", "step": 4400, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3577.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.205539, "cps": 27.945312, "mean_p": 0.209857, "boundary_entropy": 0.094267, "bpb": 1.909894}, "tr": {"chunks": 3447.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.19041, "cps": 26.929688, "mean_p": 0.188953, "boundary_entropy": 0.110891, "bpb": 2.056798}, "fi": {"chunks": 4489.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.238637, "cps": 35.070312, "mean_p": 0.23605, "boundary_entropy": 0.096049, "bpb": 2.158633}}, "bpb": {"en": 1.90989, "tr": 2.0568, "fi": 2.15863}, "gini_cpc": 0.050665, "gini_cps": 0.060338, "bpic": 4.927647, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.926394380640917, "fi": 1.1610285572921}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 329.6482849121094, "n_util_obs": 1069, "util_count_spearman": 0.16618154714463115}, "t": 3169.1} |
| {"event": "eval", "step": 4400, "phase": 3, "bpb": {"en": 1.92274, "tr": 2.06671, "fi": 2.16528}, "bytes_per_chunk": {"en": 4.848, "tr": 5.764, "fi": 4.59}, "bpb_mean": 2.05157, "frozen_inventory": true, "t": 3170.0} |
| {"event": "train", "step": 4400, "arch": "lexmem_v1", "phase": 3, "objective": "global", "lm_loss": 1.35779, "bpb": 1.95888, "chunk_loss": 0.030355, "F": 0.20215, "G": 0.20678, "chunks": 207.0, "bytes_per_chunk": 4.947, "boundary_entropy": 0.09277, "router_noise": 0.0, "grad_norm": 0.6294, "lr": 7.92e-05, "router_trainable": true, "sec_per_step": 0.667, "probe_sec": 0.0463, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03461, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.08995, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": -0.00582, "trunc_abs_frac": 0.00582, "mean_delta_full": -0.43494, "read_hit_rate": {"surface": 0.71951}, "mean_gate": {"surface": 0.26557}, "n_reads": 4743, "n_cand": 6592, "read_suppressed_frac": 0.0995, "dropout_reads_logged": 4743, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 329.71893310546875, "n_util_obs": 1069, "util_count_spearman": 0.166343708104403}, "probe_overhead_sec_total": 264.352, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.01781, "alpha": 0.03, "t": 3171.5} |
| {"event": "consolidate", "step": 4450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3196.9} |
| {"event": "train", "step": 4450, "arch": "lexmem_v1", "phase": 3, "objective": "global", "lm_loss": 1.35091, "bpb": 1.94895, "chunk_loss": 0.030304, "F": 0.19559, "G": 0.20164, "chunks": 200.28, "bytes_per_chunk": 5.113, "boundary_entropy": 0.09302, "router_noise": 0.0, "grad_norm": 0.7136, "lr": 7.63e-05, "router_trainable": true, "sec_per_step": 0.512, "probe_sec": 0.0355, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02539, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.01105, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.30251, "trunc_abs_frac": 1.09819, "mean_delta_full": -0.32605, "read_hit_rate": {"surface": 0.70786}, "mean_gate": {"surface": 0.26082}, "n_reads": 4514, "n_cand": 6377, "read_suppressed_frac": 0.0973, "dropout_reads_logged": 4514, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 333.403564453125, "n_util_obs": 1076, "util_count_spearman": 0.1632755254579667}, "probe_overhead_sec_total": 267.321, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.015491, "alpha": 0.03, "t": 3197.1} |
| {"event": "consolidate", "step": 4500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3209.1} |
| {"event": "train", "step": 4500, "arch": "lexmem_v1", "phase": 3, "objective": "global", "lm_loss": 1.32816, "bpb": 1.91613, "chunk_loss": 0.03041, "F": 0.20374, "G": 0.2094, "chunks": 208.62, "bytes_per_chunk": 4.908, "boundary_entropy": 0.089, "router_noise": 0.0, "grad_norm": 0.6562, "lr": 7.36e-05, "router_trainable": true, "sec_per_step": 0.248, "probe_sec": 0.0443, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03106, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.27194, "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": 0.7316, "trunc_abs_frac": 1.087, "mean_delta_full": 0.30245, "read_hit_rate": {"surface": 0.72486}, "mean_gate": {"surface": 0.26072}, "n_reads": 4816, "n_cand": 6644, "read_suppressed_frac": 0.0914, "dropout_reads_logged": 4816, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 337.0630187988281, "n_util_obs": 1082, "util_count_spearman": 0.157871212845505}, "probe_overhead_sec_total": 269.448, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.016536, "alpha": 0.03, "t": 3209.5} |
| {"event": "consolidate", "step": 4550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3221.7} |
| {"event": "train", "step": 4550, "arch": "lexmem_v1", "phase": 3, "objective": "global", "lm_loss": 1.35871, "bpb": 1.9602, "chunk_loss": 0.030358, "F": 0.20007, "G": 0.20554, "chunks": 204.88, "bytes_per_chunk": 4.998, "boundary_entropy": 0.09104, "router_noise": 0.0, "grad_norm": 0.6602, "lr": 7.09e-05, "router_trainable": true, "sec_per_step": 0.249, "probe_sec": 0.1029, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09518, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.2069, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 1.64059, "trunc_abs_frac": 1.80153, "mean_delta_full": 0.12482, "read_hit_rate": {"surface": 0.70494}, "mean_gate": {"surface": 0.26683}, "n_reads": 4599, "n_cand": 6524, "read_suppressed_frac": 0.1, "dropout_reads_logged": 4599, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 340.7364501953125, "n_util_obs": 1093, "util_count_spearman": 0.15951145524556923}, "probe_overhead_sec_total": 271.809, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.01915, "alpha": 0.03, "t": 3222.0} |
| {"event": "consolidate", "step": 4600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3234.4} |
| {"event": "probe", "step": 4600, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3560.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.204562, "cps": 27.8125, "mean_p": 0.209962, "boundary_entropy": 0.092523, "bpb": 1.898706}, "tr": {"chunks": 3423.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.189085, "cps": 26.742188, "mean_p": 0.187315, "boundary_entropy": 0.108902, "bpb": 2.051619}, "fi": {"chunks": 4508.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.239647, "cps": 35.21875, "mean_p": 0.236487, "boundary_entropy": 0.090443, "bpb": 2.147158}}, "bpb": {"en": 1.89871, "tr": 2.05162, "fi": 2.14716}, "gini_cpc": 0.053227, "gini_cps": 0.062948, "bpic": 4.937081, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.924337281610162, "fi": 1.1715103940498597}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 344.3352966308594, "n_util_obs": 1106, "util_count_spearman": 0.15793786893933975}, "t": 3235.0} |
| {"event": "eval", "step": 4600, "phase": 3, "bpb": {"en": 1.91259, "tr": 2.05919, "fi": 2.1524}, "bytes_per_chunk": {"en": 4.834, "tr": 5.834, "fi": 4.589}, "bpb_mean": 2.04139, "frozen_inventory": true, "t": 3235.8} |
| {"event": "train", "step": 4600, "arch": "lexmem_v1", "phase": 3, "objective": "global", "lm_loss": 1.35772, "bpb": 1.95877, "chunk_loss": 0.030556, "F": 0.21133, "G": 0.21596, "chunks": 216.41, "bytes_per_chunk": 4.732, "boundary_entropy": 0.09317, "router_noise": 0.0, "grad_norm": 0.684, "lr": 6.82e-05, "router_trainable": true, "sec_per_step": 0.298, "probe_sec": 0.0245, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01848, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.15353, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.9354, "trunc_abs_frac": 1.11363, "mean_delta_full": -0.34271, "read_hit_rate": {"surface": 0.73524}, "mean_gate": {"surface": 0.26218}, "n_reads": 5068, "n_cand": 6893, "read_suppressed_frac": 0.101, "dropout_reads_logged": 5068, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 344.41351318359375, "n_util_obs": 1106, "util_count_spearman": 0.15819336111423665}, "probe_overhead_sec_total": 274.284, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.01862, "alpha": 0.03, "t": 3236.9} |
| {"event": "consolidate", "step": 4650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3249.5} |
| {"event": "train", "step": 4650, "arch": "lexmem_v1", "phase": 3, "objective": "global", "lm_loss": 1.38693, "bpb": 2.00092, "chunk_loss": 0.030382, "F": 0.20126, "G": 0.20831, "chunks": 206.09, "bytes_per_chunk": 4.969, "boundary_entropy": 0.08957, "router_noise": 0.0, "grad_norm": 0.6488, "lr": 6.57e-05, "router_trainable": true, "sec_per_step": 0.256, "probe_sec": 0.023, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01786, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.20237, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.79438, "trunc_abs_frac": 1.03488, "mean_delta_full": 0.90302, "read_hit_rate": {"surface": 0.70517}, "mean_gate": {"surface": 0.26616}, "n_reads": 4628, "n_cand": 6563, "read_suppressed_frac": 0.1013, "dropout_reads_logged": 4628, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 348.05853271484375, "n_util_obs": 1116, "util_count_spearman": 0.1578646635995393}, "probe_overhead_sec_total": 276.542, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.018033, "alpha": 0.03, "t": 3249.7} |
| {"event": "consolidate", "step": 4700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3259.3} |
| {"event": "train", "step": 4700, "arch": "lexmem_v1", "phase": 3, "objective": "global", "lm_loss": 1.41427, "bpb": 2.04035, "chunk_loss": 0.030744, "F": 0.2157, "G": 0.21989, "chunks": 220.88, "bytes_per_chunk": 4.636, "boundary_entropy": 0.08941, "router_noise": 0.0, "grad_norm": 0.6597, "lr": 6.32e-05, "router_trainable": true, "sec_per_step": 0.196, "probe_sec": 0.0274, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01862, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.41489, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.36985, "trunc_abs_frac": 1.79992, "mean_delta_full": 0.11115, "read_hit_rate": {"surface": 0.7392}, "mean_gate": {"surface": 0.25288}, "n_reads": 5201, "n_cand": 7036, "read_suppressed_frac": 0.0906, "dropout_reads_logged": 5201, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 351.75262451171875, "n_util_obs": 1127, "util_count_spearman": 0.15828077090594359}, "probe_overhead_sec_total": 278.82, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.01965, "alpha": 0.03, "t": 3259.5} |
| {"event": "consolidate", "step": 4750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3272.2} |
| {"event": "train", "step": 4750, "arch": "lexmem_v1", "phase": 3, "objective": "global", "lm_loss": 1.40117, "bpb": 2.02146, "chunk_loss": 0.030527, "F": 0.20734, "G": 0.21284, "chunks": 212.31, "bytes_per_chunk": 4.823, "boundary_entropy": 0.09069, "router_noise": 0.0, "grad_norm": 0.665, "lr": 6.08e-05, "router_trainable": true, "sec_per_step": 0.258, "probe_sec": 0.023, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0173, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.00427, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 7, "trunc_ratio_sum": -3.2416, "trunc_abs_frac": 2.21129, "mean_delta_full": 0.04543, "read_hit_rate": {"surface": 0.72567}, "mean_gate": {"surface": 0.26098}, "n_reads": 4907, "n_cand": 6762, "read_suppressed_frac": 0.105, "dropout_reads_logged": 4907, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 355.4190673828125, "n_util_obs": 1139, "util_count_spearman": 0.16210114149349972}, "probe_overhead_sec_total": 281.16, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.015816, "alpha": 0.03, "t": 3272.4} |
| {"event": "consolidate", "step": 4800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3283.3} |
| {"event": "probe", "step": 4800, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3517.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.202092, "cps": 27.476562, "mean_p": 0.207507, "boundary_entropy": 0.08901, "bpb": 1.891519}, "tr": {"chunks": 3459.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.191073, "cps": 27.023438, "mean_p": 0.188662, "boundary_entropy": 0.107535, "bpb": 2.034864}, "fi": {"chunks": 4488.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.238584, "cps": 35.0625, "mean_p": 0.23547, "boundary_entropy": 0.089978, "bpb": 2.138057}}, "bpb": {"en": 1.89152, "tr": 2.03486, "fi": 2.13806}, "gini_cpc": 0.050137, "gini_cps": 0.059839, "bpic": 4.948709, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9454787284795997, "fi": 1.1805726469308373}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 359.0134582519531, "n_util_obs": 1146, "util_count_spearman": 0.16099524953922426}, "t": 3283.8} |
| {"event": "eval", "step": 4800, "phase": 3, "bpb": {"en": 1.90535, "tr": 2.04464, "fi": 2.14159}, "bytes_per_chunk": {"en": 4.866, "tr": 5.749, "fi": 4.627}, "bpb_mean": 2.03053, "frozen_inventory": true, "t": 3284.6} |
| {"event": "train", "step": 4800, "arch": "lexmem_v1", "phase": 3, "objective": "global", "lm_loss": 1.37395, "bpb": 1.98219, "chunk_loss": 0.030479, "F": 0.20654, "G": 0.21119, "chunks": 211.5, "bytes_per_chunk": 4.842, "boundary_entropy": 0.08885, "router_noise": 0.0, "grad_norm": 0.6454, "lr": 5.85e-05, "router_trainable": true, "sec_per_step": 0.268, "probe_sec": 0.0383, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02175, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.18273, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.49624, "trunc_abs_frac": 0.74232, "mean_delta_full": 0.32671, "read_hit_rate": {"surface": 0.71882}, "mean_gate": {"surface": 0.25492}, "n_reads": 4842, "n_cand": 6736, "read_suppressed_frac": 0.1051, "dropout_reads_logged": 4842, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 359.08642578125, "n_util_obs": 1146, "util_count_spearman": 0.16092515527015275}, "probe_overhead_sec_total": 283.34, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.023694, "alpha": 0.03, "t": 3285.8} |
| {"event": "consolidate", "step": 4850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3302.1} |
| {"event": "train", "step": 4850, "arch": "lexmem_v1", "phase": 3, "objective": "global", "lm_loss": 1.34574, "bpb": 1.94149, "chunk_loss": 0.030381, "F": 0.20078, "G": 0.20754, "chunks": 205.59, "bytes_per_chunk": 4.981, "boundary_entropy": 0.08813, "router_noise": 0.0, "grad_norm": 0.6689, "lr": 5.62e-05, "router_trainable": true, "sec_per_step": 0.332, "probe_sec": 0.1004, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09105, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.0768, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.38211, "trunc_abs_frac": 0.53145, "mean_delta_full": 0.36139, "read_hit_rate": {"surface": 0.71559}, "mean_gate": {"surface": 0.25824}, "n_reads": 4685, "n_cand": 6547, "read_suppressed_frac": 0.1027, "dropout_reads_logged": 4685, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 362.7866516113281, "n_util_obs": 1155, "util_count_spearman": 0.16731825453229474}, "probe_overhead_sec_total": 285.806, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.017582, "alpha": 0.03, "t": 3302.4} |
| {"event": "consolidate", "step": 4900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3323.1} |
| {"event": "train", "step": 4900, "arch": "lexmem_v1", "phase": 3, "objective": "global", "lm_loss": 1.37925, "bpb": 1.98984, "chunk_loss": 0.030425, "F": 0.19867, "G": 0.20452, "chunks": 203.44, "bytes_per_chunk": 5.033, "boundary_entropy": 0.08514, "router_noise": 0.0, "grad_norm": 0.6764, "lr": 5.41e-05, "router_trainable": true, "sec_per_step": 0.428, "probe_sec": 0.1026, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09235, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.22178, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.65843, "trunc_abs_frac": 0.69668, "mean_delta_full": -0.2012, "read_hit_rate": {"surface": 0.69775}, "mean_gate": {"surface": 0.24385}, "n_reads": 4520, "n_cand": 6478, "read_suppressed_frac": 0.096, "dropout_reads_logged": 4520, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 366.46087646484375, "n_util_obs": 1166, "util_count_spearman": 0.168135116916557}, "probe_overhead_sec_total": 288.473, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.020036, "alpha": 0.03, "t": 3323.8} |
| {"event": "consolidate", "step": 4950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3341.2} |
| {"event": "train", "step": 4950, "arch": "lexmem_v1", "phase": 3, "objective": "global", "lm_loss": 1.33679, "bpb": 1.92858, "chunk_loss": 0.03042, "F": 0.20288, "G": 0.20861, "chunks": 207.75, "bytes_per_chunk": 4.929, "boundary_entropy": 0.0898, "router_noise": 0.0, "grad_norm": 0.6693, "lr": 5.2e-05, "router_trainable": true, "sec_per_step": 0.353, "probe_sec": 0.0282, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02061, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.01967, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 1.6369, "trunc_abs_frac": 0.50371, "mean_delta_full": -0.09073, "read_hit_rate": {"surface": 0.7308}, "mean_gate": {"surface": 0.25982}, "n_reads": 4835, "n_cand": 6616, "read_suppressed_frac": 0.0962, "dropout_reads_logged": 4835, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 370.137939453125, "n_util_obs": 1173, "util_count_spearman": 0.16874678790342085}, "probe_overhead_sec_total": 291.202, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.018462, "alpha": 0.03, "t": 3341.4} |
| {"event": "consolidate", "step": 5000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3357.1} |
| {"event": "probe", "step": 5000, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3527.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.202666, "cps": 27.554688, "mean_p": 0.207756, "boundary_entropy": 0.089233, "bpb": 1.884894}, "tr": {"chunks": 3447.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.19041, "cps": 26.929688, "mean_p": 0.187817, "boundary_entropy": 0.107726, "bpb": 2.026373}, "fi": {"chunks": 4500.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.239222, "cps": 35.15625, "mean_p": 0.236188, "boundary_entropy": 0.088922, "bpb": 2.123939}}, "bpb": {"en": 1.88489, "tr": 2.02637, "fi": 2.12394}, "gini_cpc": 0.051464, "gini_cps": 0.061182, "bpic": 4.944396, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9395272751779303, "fi": 1.180373065322598}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 373.74957275390625, "n_util_obs": 1178, "util_count_spearman": 0.16682103503813742}, "t": 3357.7} |
| {"event": "eval", "step": 5000, "phase": 3, "bpb": {"en": 1.89826, "tr": 2.03598, "fi": 2.12751}, "bytes_per_chunk": {"en": 4.868, "tr": 5.777, "fi": 4.618}, "bpb_mean": 2.02058, "frozen_inventory": true, "t": 3358.5} |
| {"event": "train", "step": 5000, "arch": "lexmem_v1", "phase": 3, "objective": "global", "lm_loss": 1.32681, "bpb": 1.91418, "chunk_loss": 0.030432, "F": 0.20575, "G": 0.21001, "chunks": 210.69, "bytes_per_chunk": 4.86, "boundary_entropy": 0.08644, "router_noise": 0.0, "grad_norm": 0.6251, "lr": 5e-05, "router_trainable": true, "sec_per_step": 0.377, "probe_sec": 0.0393, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0193, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.08319, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -1.0627, "trunc_abs_frac": 1.18307, "mean_delta_full": -0.17108, "read_hit_rate": {"surface": 0.72757}, "mean_gate": {"surface": 0.24284}, "n_reads": 4882, "n_cand": 6710, "read_suppressed_frac": 0.1028, "dropout_reads_logged": 4882, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 373.82342529296875, "n_util_obs": 1178, "util_count_spearman": 0.1667477162621277}, "probe_overhead_sec_total": 293.516, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.017258, "alpha": 0.03, "t": 3360.3} |
| {"event": "consolidate", "step": 5050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3371.8} |
| {"event": "train", "step": 5050, "arch": "lexmem_v1", "phase": 3, "objective": "global", "lm_loss": 1.34426, "bpb": 1.93936, "chunk_loss": 0.030479, "F": 0.20639, "G": 0.21153, "chunks": 211.34, "bytes_per_chunk": 4.845, "boundary_entropy": 0.08656, "router_noise": 0.0, "grad_norm": 0.6661, "lr": 4.81e-05, "router_trainable": true, "sec_per_step": 0.235, "probe_sec": 0.0359, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02979, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.23289, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 2.97255, "trunc_abs_frac": 0.5255, "mean_delta_full": 0.02712, "read_hit_rate": {"surface": 0.72055}, "mean_gate": {"surface": 0.2496}, "n_reads": 4850, "n_cand": 6731, "read_suppressed_frac": 0.0971, "dropout_reads_logged": 4850, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 377.47088623046875, "n_util_obs": 1193, "util_count_spearman": 0.17327662446122183}, "probe_overhead_sec_total": 296.014, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.018078, "alpha": 0.03, "t": 3372.0} |
| {"event": "consolidate", "step": 5100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3387.6} |
| {"event": "train", "step": 5100, "arch": "lexmem_v1", "phase": 3, "objective": "global", "lm_loss": 1.3687, "bpb": 1.97462, "chunk_loss": 0.030554, "F": 0.20844, "G": 0.21335, "chunks": 213.44, "bytes_per_chunk": 4.798, "boundary_entropy": 0.08708, "router_noise": 0.0, "grad_norm": 0.6986, "lr": 4.63e-05, "router_trainable": true, "sec_per_step": 0.319, "probe_sec": 0.1045, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09496, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.3648, "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": 0.50863, "trunc_abs_frac": 0.57529, "mean_delta_full": 0.27336, "read_hit_rate": {"surface": 0.73904}, "mean_gate": {"surface": 0.24825}, "n_reads": 5024, "n_cand": 6798, "read_suppressed_frac": 0.1027, "dropout_reads_logged": 5024, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 381.16448974609375, "n_util_obs": 1203, "util_count_spearman": 0.16448657878312564}, "probe_overhead_sec_total": 298.289, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.01766, "alpha": 0.03, "t": 3388.0} |
| {"event": "consolidate", "step": 5150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3405.7} |
| {"event": "train", "step": 5150, "arch": "lexmem_v1", "phase": 3, "objective": "global", "lm_loss": 1.36773, "bpb": 1.97322, "chunk_loss": 0.030567, "F": 0.20856, "G": 0.21324, "chunks": 213.56, "bytes_per_chunk": 4.795, "boundary_entropy": 0.08589, "router_noise": 0.0, "grad_norm": 0.7098, "lr": 4.45e-05, "router_trainable": true, "sec_per_step": 0.368, "probe_sec": 0.0424, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02359, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.12067, "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": -0.01597, "trunc_abs_frac": 0.42157, "mean_delta_full": 0.23625, "read_hit_rate": {"surface": 0.72846}, "mean_gate": {"surface": 0.24902}, "n_reads": 4955, "n_cand": 6802, "read_suppressed_frac": 0.0955, "dropout_reads_logged": 4955, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 384.8824462890625, "n_util_obs": 1211, "util_count_spearman": 0.1625985457095501}, "probe_overhead_sec_total": 300.681, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.017388, "alpha": 0.03, "t": 3406.4} |
| {"event": "consolidate", "step": 5200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3422.4} |
| {"event": "probe", "step": 5200, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3539.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.203356, "cps": 27.648438, "mean_p": 0.208097, "boundary_entropy": 0.088, "bpb": 1.876995}, "tr": {"chunks": 3547.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.195934, "cps": 27.710938, "mean_p": 0.193374, "boundary_entropy": 0.105707, "bpb": 2.016641}, "fi": {"chunks": 4521.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.240338, "cps": 35.320312, "mean_p": 0.236885, "boundary_entropy": 0.088624, "bpb": 2.115125}}, "bpb": {"en": 1.87699, "tr": 2.01664, "fi": 2.11512}, "gini_cpc": 0.046281, "gini_cps": 0.056403, "bpic": 4.88774, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9635054922682935, "fi": 1.1818603998679387}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 388.4793395996094, "n_util_obs": 1219, "util_count_spearman": 0.16254218633696133}, "t": 3422.9} |
| {"event": "eval", "step": 5200, "phase": 3, "bpb": {"en": 1.89224, "tr": 2.02657, "fi": 2.11845}, "bytes_per_chunk": {"en": 4.841, "tr": 5.555, "fi": 4.607}, "bpb_mean": 2.01242, "frozen_inventory": true, "t": 3423.7} |
| {"event": "train", "step": 5200, "arch": "lexmem_v1", "phase": 3, "objective": "global", "lm_loss": 1.32121, "bpb": 1.9061, "chunk_loss": 0.030333, "F": 0.20016, "G": 0.20506, "chunks": 204.97, "bytes_per_chunk": 4.996, "boundary_entropy": 0.0858, "router_noise": 0.0, "grad_norm": 0.6752, "lr": 4.29e-05, "router_trainable": true, "sec_per_step": 0.375, "probe_sec": 0.03, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01764, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.04862, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -2.34683, "trunc_abs_frac": 3.29829, "mean_delta_full": -0.02502, "read_hit_rate": {"surface": 0.71717}, "mean_gate": {"surface": 0.24231}, "n_reads": 4681, "n_cand": 6527, "read_suppressed_frac": 0.1036, "dropout_reads_logged": 4681, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 388.5507507324219, "n_util_obs": 1219, "util_count_spearman": 0.16257558339211955}, "probe_overhead_sec_total": 303.008, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.017823, "alpha": 0.03, "t": 3425.1} |
| {"event": "consolidate", "step": 5250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3440.4} |
| {"event": "train", "step": 5250, "arch": "lexmem_v1", "phase": 3, "objective": "global", "lm_loss": 1.31725, "bpb": 1.90038, "chunk_loss": 0.030404, "F": 0.20151, "G": 0.20713, "chunks": 206.34, "bytes_per_chunk": 4.963, "boundary_entropy": 0.08569, "router_noise": 0.0, "grad_norm": 0.6558, "lr": 4.14e-05, "router_trainable": true, "sec_per_step": 0.308, "probe_sec": 0.0302, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01999, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.02885, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 8, "trunc_ratio_sum": -0.31797, "trunc_abs_frac": 0.48281, "mean_delta_full": 0.14412, "read_hit_rate": {"surface": 0.72348}, "mean_gate": {"surface": 0.23967}, "n_reads": 4754, "n_cand": 6571, "read_suppressed_frac": 0.1062, "dropout_reads_logged": 4754, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 392.21893310546875, "n_util_obs": 1223, "util_count_spearman": 0.16767048287560504}, "probe_overhead_sec_total": 305.767, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.019484, "alpha": 0.03, "t": 3440.6} |
| {"event": "consolidate", "step": 5300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3453.2} |
| {"event": "train", "step": 5300, "arch": "lexmem_v1", "phase": 3, "objective": "global", "lm_loss": 1.36905, "bpb": 1.97512, "chunk_loss": 0.030436, "F": 0.20386, "G": 0.20877, "chunks": 208.75, "bytes_per_chunk": 4.905, "boundary_entropy": 0.08246, "router_noise": 0.0, "grad_norm": 0.6775, "lr": 3.99e-05, "router_trainable": true, "sec_per_step": 0.258, "probe_sec": 0.034, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0268, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.16703, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.114, "trunc_abs_frac": 0.80396, "mean_delta_full": 0.37567, "read_hit_rate": {"surface": 0.7148}, "mean_gate": {"surface": 0.24218}, "n_reads": 4752, "n_cand": 6648, "read_suppressed_frac": 0.0968, "dropout_reads_logged": 4752, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 395.9039306640625, "n_util_obs": 1230, "util_count_spearman": 0.17218528425945098}, "probe_overhead_sec_total": 308.243, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.019748, "alpha": 0.03, "t": 3453.4} |
| {"event": "consolidate", "step": 5350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3468.0} |
| {"event": "train", "step": 5350, "arch": "lexmem_v1", "phase": 3, "objective": "global", "lm_loss": 1.3063, "bpb": 1.88459, "chunk_loss": 0.030318, "F": 0.19971, "G": 0.20451, "chunks": 204.5, "bytes_per_chunk": 5.007, "boundary_entropy": 0.08526, "router_noise": 0.0, "grad_norm": 0.6829, "lr": 3.86e-05, "router_trainable": true, "sec_per_step": 0.298, "probe_sec": 0.0232, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0182, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.06729, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 1.20309, "trunc_abs_frac": 0.4931, "mean_delta_full": -0.10669, "read_hit_rate": {"surface": 0.72512}, "mean_gate": {"surface": 0.23823}, "n_reads": 4722, "n_cand": 6512, "read_suppressed_frac": 0.1048, "dropout_reads_logged": 4722, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 399.57293701171875, "n_util_obs": 1241, "util_count_spearman": 0.1746632670082461}, "probe_overhead_sec_total": 310.731, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.021506, "alpha": 0.03, "t": 3468.3} |
| {"event": "consolidate", "step": 5400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3491.9} |
| {"event": "probe", "step": 5400, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3543.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.203586, "cps": 27.679688, "mean_p": 0.208674, "boundary_entropy": 0.087446, "bpb": 1.87268}, "tr": {"chunks": 3442.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.190134, "cps": 26.890625, "mean_p": 0.186847, "boundary_entropy": 0.105177, "bpb": 2.012645}, "fi": {"chunks": 4505.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.239488, "cps": 35.195312, "mean_p": 0.236043, "boundary_entropy": 0.084389, "bpb": 2.108488}}, "bpb": {"en": 1.87268, "tr": 2.01265, "fi": 2.10849}, "gini_cpc": 0.051961, "gini_cps": 0.061677, "bpic": 4.937511, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.933927755482166, "fi": 1.1763481660409931}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 403.152099609375, "n_util_obs": 1254, "util_count_spearman": 0.1740650835640866}, "t": 3492.5} |
| {"event": "eval", "step": 5400, "phase": 3, "bpb": {"en": 1.88659, "tr": 2.02291, "fi": 2.11217}, "bytes_per_chunk": {"en": 4.841, "tr": 5.798, "fi": 4.604}, "bpb_mean": 2.00722, "frozen_inventory": true, "t": 3493.5} |
| {"event": "train", "step": 5400, "arch": "lexmem_v1", "phase": 3, "objective": "global", "lm_loss": 1.37406, "bpb": 1.98236, "chunk_loss": 0.030652, "F": 0.20978, "G": 0.21487, "chunks": 214.81, "bytes_per_chunk": 4.767, "boundary_entropy": 0.0853, "router_noise": 0.0, "grad_norm": 0.7024, "lr": 3.73e-05, "router_trainable": true, "sec_per_step": 0.544, "probe_sec": 0.0713, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03712, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.27092, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -1.80107, "trunc_abs_frac": 5.31062, "mean_delta_full": -0.05725, "read_hit_rate": {"surface": 0.72201}, "mean_gate": {"surface": 0.23953}, "n_reads": 4940, "n_cand": 6842, "read_suppressed_frac": 0.0931, "dropout_reads_logged": 4940, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 403.22528076171875, "n_util_obs": 1254, "util_count_spearman": 0.17335537462225675}, "probe_overhead_sec_total": 313.616, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.019646, "alpha": 0.03, "t": 3495.5} |
| {"event": "consolidate", "step": 5450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3510.0} |
| {"event": "train", "step": 5450, "arch": "lexmem_v1", "phase": 3, "objective": "global", "lm_loss": 1.31177, "bpb": 1.89249, "chunk_loss": 0.030528, "F": 0.20792, "G": 0.21244, "chunks": 212.91, "bytes_per_chunk": 4.81, "boundary_entropy": 0.08563, "router_noise": 0.0, "grad_norm": 0.6869, "lr": 3.62e-05, "router_trainable": true, "sec_per_step": 0.308, "probe_sec": 0.0524, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02455, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.09407, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 8.91127, "trunc_abs_frac": 2.13392, "mean_delta_full": 0.01083, "read_hit_rate": {"surface": 0.73499}, "mean_gate": {"surface": 0.24909}, "n_reads": 4984, "n_cand": 6781, "read_suppressed_frac": 0.0975, "dropout_reads_logged": 4984, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 406.9345703125, "n_util_obs": 1262, "util_count_spearman": 0.16454257717592363}, "probe_overhead_sec_total": 315.899, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.018361, "alpha": 0.03, "t": 3510.9} |
| {"event": "consolidate", "step": 5500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3535.8} |
| {"event": "train", "step": 5500, "arch": "lexmem_v1", "phase": 3, "objective": "global", "lm_loss": 1.29735, "bpb": 1.87168, "chunk_loss": 0.030517, "F": 0.20773, "G": 0.21227, "chunks": 212.72, "bytes_per_chunk": 4.814, "boundary_entropy": 0.08559, "router_noise": 0.0, "grad_norm": 0.6557, "lr": 3.51e-05, "router_trainable": true, "sec_per_step": 0.511, "probe_sec": 0.0485, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03171, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.15186, "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": -699.64062, "trunc_abs_frac": 0.40709, "mean_delta_full": -0.00016, "read_hit_rate": {"surface": 0.73506}, "mean_gate": {"surface": 0.23739}, "n_reads": 4980, "n_cand": 6775, "read_suppressed_frac": 0.095, "dropout_reads_logged": 4980, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 410.630615234375, "n_util_obs": 1272, "util_count_spearman": 0.16773282673662562}, "probe_overhead_sec_total": 318.648, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.02183, "alpha": 0.03, "t": 3536.5} |
| {"event": "consolidate", "step": 5550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3568.8} |
| {"event": "train", "step": 5550, "arch": "lexmem_v1", "phase": 3, "objective": "global", "lm_loss": 1.35641, "bpb": 1.95688, "chunk_loss": 0.030416, "F": 0.20209, "G": 0.20739, "chunks": 206.94, "bytes_per_chunk": 4.948, "boundary_entropy": 0.08724, "router_noise": 0.0, "grad_norm": 0.7144, "lr": 3.41e-05, "router_trainable": true, "sec_per_step": 0.657, "probe_sec": 0.0369, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01821, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.0516, "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": -0.02149, "trunc_abs_frac": 0.72188, "mean_delta_full": 0.41331, "read_hit_rate": {"surface": 0.72155}, "mean_gate": {"surface": 0.23934}, "n_reads": 4755, "n_cand": 6590, "read_suppressed_frac": 0.1054, "dropout_reads_logged": 4755, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 414.305908203125, "n_util_obs": 1284, "util_count_spearman": 0.16433645391572702}, "probe_overhead_sec_total": 321.307, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.025845, "alpha": 0.03, "t": 3569.4} |
| {"event": "consolidate", "step": 5600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3595.1} |
| {"event": "probe", "step": 5600, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3553.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.20416, "cps": 27.757812, "mean_p": 0.209161, "boundary_entropy": 0.086981, "bpb": 1.867413}, "tr": {"chunks": 3447.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.19041, "cps": 26.929688, "mean_p": 0.187648, "boundary_entropy": 0.103305, "bpb": 2.005112}, "fi": {"chunks": 4499.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.239169, "cps": 35.148438, "mean_p": 0.235894, "boundary_entropy": 0.086298, "bpb": 2.102249}}, "bpb": {"en": 1.86741, "tr": 2.00511, "fi": 2.10225}, "gini_cpc": 0.051291, "gini_cps": 0.060991, "bpic": 4.933646, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9326520404031974, "fi": 1.171474993305577}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 417.9156188964844, "n_util_obs": 1294, "util_count_spearman": 0.16706960546300634}, "t": 3595.6} |
| {"event": "eval", "step": 5600, "phase": 3, "bpb": {"en": 1.88209, "tr": 2.01509, "fi": 2.1066}, "bytes_per_chunk": {"en": 4.836, "tr": 5.756, "fi": 4.618}, "bpb_mean": 2.00126, "frozen_inventory": true, "t": 3596.4} |
| {"event": "train", "step": 5600, "arch": "lexmem_v1", "phase": 3, "objective": "global", "lm_loss": 1.34246, "bpb": 1.93676, "chunk_loss": 0.030416, "F": 0.2034, "G": 0.20887, "chunks": 208.28, "bytes_per_chunk": 4.916, "boundary_entropy": 0.08682, "router_noise": 0.0, "grad_norm": 0.7402, "lr": 3.33e-05, "router_trainable": true, "sec_per_step": 0.576, "probe_sec": 0.0334, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02336, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.00556, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.79119, "trunc_abs_frac": 0.78864, "mean_delta_full": -0.57889, "read_hit_rate": {"surface": 0.71235}, "mean_gate": {"surface": 0.23933}, "n_reads": 4725, "n_cand": 6633, "read_suppressed_frac": 0.0935, "dropout_reads_logged": 4725, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 417.98809814453125, "n_util_obs": 1295, "util_count_spearman": 0.1667561083028685}, "probe_overhead_sec_total": 323.788, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.016506, "alpha": 0.03, "t": 3598.2} |
| {"event": "consolidate", "step": 5650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3617.5} |
| {"event": "train", "step": 5650, "arch": "lexmem_v1", "phase": 3, "objective": "global", "lm_loss": 1.32737, "bpb": 1.91499, "chunk_loss": 0.03053, "F": 0.20911, "G": 0.21239, "chunks": 214.12, "bytes_per_chunk": 4.782, "boundary_entropy": 0.08316, "router_noise": 0.0, "grad_norm": 0.6926, "lr": 3.25e-05, "router_trainable": true, "sec_per_step": 0.391, "probe_sec": 0.0701, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0634, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.10624, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.16736, "trunc_abs_frac": 14.86624, "mean_delta_full": 0.3427, "read_hit_rate": {"surface": 0.72889}, "mean_gate": {"surface": 0.23258}, "n_reads": 4971, "n_cand": 6820, "read_suppressed_frac": 0.0923, "dropout_reads_logged": 4971, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 421.67535400390625, "n_util_obs": 1302, "util_count_spearman": 0.16588004624385805}, "probe_overhead_sec_total": 326.555, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.014573, "alpha": 0.03, "t": 3617.7} |
| {"event": "consolidate", "step": 5700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3639.5} |
| {"event": "train", "step": 5700, "arch": "lexmem_v1", "phase": 3, "objective": "global", "lm_loss": 1.33712, "bpb": 1.92906, "chunk_loss": 0.03038, "F": 0.20041, "G": 0.20582, "chunks": 205.22, "bytes_per_chunk": 4.99, "boundary_entropy": 0.0844, "router_noise": 0.0, "grad_norm": 0.7052, "lr": 3.18e-05, "router_trainable": true, "sec_per_step": 0.446, "probe_sec": 0.0412, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02402, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.10324, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 28.14545, "trunc_abs_frac": 0.69747, "mean_delta_full": 0.00168, "read_hit_rate": {"surface": 0.71186}, "mean_gate": {"surface": 0.23521}, "n_reads": 4652, "n_cand": 6535, "read_suppressed_frac": 0.0959, "dropout_reads_logged": 4652, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 425.3564453125, "n_util_obs": 1313, "util_count_spearman": 0.16228357736475468}, "probe_overhead_sec_total": 328.793, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.021133, "alpha": 0.03, "t": 3640.0} |
| {"event": "consolidate", "step": 5750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3661.0} |
| {"event": "train", "step": 5750, "arch": "lexmem_v1", "phase": 3, "objective": "global", "lm_loss": 1.39681, "bpb": 2.01517, "chunk_loss": 0.030561, "F": 0.20959, "G": 0.21373, "chunks": 214.62, "bytes_per_chunk": 4.771, "boundary_entropy": 0.08266, "router_noise": 0.0, "grad_norm": 0.7307, "lr": 3.13e-05, "router_trainable": true, "sec_per_step": 0.425, "probe_sec": 0.027, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02119, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.24757, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 1.37151, "trunc_abs_frac": 1.54134, "mean_delta_full": 0.21324, "read_hit_rate": {"surface": 0.71592}, "mean_gate": {"surface": 0.24202}, "n_reads": 4894, "n_cand": 6836, "read_suppressed_frac": 0.0971, "dropout_reads_logged": 4894, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 429.04205322265625, "n_util_obs": 1322, "util_count_spearman": 0.16345513272958134}, "probe_overhead_sec_total": 331.374, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.02013, "alpha": 0.03, "t": 3661.3} |
| {"event": "consolidate", "step": 5800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3681.8} |
| {"event": "probe", "step": 5800, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3529.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.202781, "cps": 27.570312, "mean_p": 0.207939, "boundary_entropy": 0.08456, "bpb": 1.865351}, "tr": {"chunks": 3467.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.191515, "cps": 27.085938, "mean_p": 0.187476, "boundary_entropy": 0.098166, "bpb": 2.001765}, "fi": {"chunks": 4492.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.238796, "cps": 35.09375, "mean_p": 0.235304, "boundary_entropy": 0.082142, "bpb": 2.098258}}, "bpb": {"en": 1.86535, "tr": 2.00177, "fi": 2.09826}, "gini_cpc": 0.049789, "gini_cps": 0.059482, "bpic": 4.93837, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9444430000197072, "fi": 1.1776068574576661}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 432.63671875, "n_util_obs": 1332, "util_count_spearman": 0.16159331897397838}, "t": 3682.3} |
| {"event": "eval", "step": 5800, "phase": 3, "bpb": {"en": 1.87965, "tr": 2.01049, "fi": 2.10242}, "bytes_per_chunk": {"en": 4.848, "tr": 5.715, "fi": 4.613}, "bpb_mean": 1.99752, "frozen_inventory": true, "t": 3683.2} |
| {"event": "train", "step": 5800, "arch": "lexmem_v1", "phase": 3, "objective": "global", "lm_loss": 1.31511, "bpb": 1.8973, "chunk_loss": 0.030576, "F": 0.20505, "G": 0.20946, "chunks": 209.97, "bytes_per_chunk": 4.877, "boundary_entropy": 0.08615, "router_noise": 0.0, "grad_norm": 0.7074, "lr": 3.08e-05, "router_trainable": true, "sec_per_step": 0.468, "probe_sec": 0.0242, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01822, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.02096, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 1.80621, "trunc_abs_frac": 2.00912, "mean_delta_full": 0.20951, "read_hit_rate": {"surface": 0.73381}, "mean_gate": {"surface": 0.24509}, "n_reads": 4907, "n_cand": 6687, "read_suppressed_frac": 0.105, "dropout_reads_logged": 4907, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 432.7130126953125, "n_util_obs": 1333, "util_count_spearman": 0.16083553302143264}, "probe_overhead_sec_total": 333.993, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.018802, "alpha": 0.03, "t": 3684.7} |
| {"event": "consolidate", "step": 5850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3699.3} |
| {"event": "train", "step": 5850, "arch": "lexmem_v1", "phase": 3, "objective": "global", "lm_loss": 1.33904, "bpb": 1.93183, "chunk_loss": 0.030505, "F": 0.20804, "G": 0.21211, "chunks": 213.03, "bytes_per_chunk": 4.807, "boundary_entropy": 0.08012, "router_noise": 0.0, "grad_norm": 0.7099, "lr": 3.05e-05, "router_trainable": true, "sec_per_step": 0.306, "probe_sec": 0.0397, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02362, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.64903, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.97151, "trunc_abs_frac": 0.74421, "mean_delta_full": 1.12798, "read_hit_rate": {"surface": 0.7126}, "mean_gate": {"surface": 0.23176}, "n_reads": 4835, "n_cand": 6785, "read_suppressed_frac": 0.0978, "dropout_reads_logged": 4835, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 436.38079833984375, "n_util_obs": 1344, "util_count_spearman": 0.15906385940227552}, "probe_overhead_sec_total": 336.127, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.017376, "alpha": 0.03, "t": 3700.0} |
| {"event": "consolidate", "step": 5900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3715.8} |
| {"event": "train", "step": 5900, "arch": "lexmem_v1", "phase": 3, "objective": "global", "lm_loss": 1.27051, "bpb": 1.83296, "chunk_loss": 0.030343, "F": 0.2012, "G": 0.20603, "chunks": 206.03, "bytes_per_chunk": 4.97, "boundary_entropy": 0.08127, "router_noise": 0.0, "grad_norm": 0.6807, "lr": 3.02e-05, "router_trainable": true, "sec_per_step": 0.32, "probe_sec": 0.08, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07398, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.08733, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.39202, "trunc_abs_frac": 0.44062, "mean_delta_full": 0.47369, "read_hit_rate": {"surface": 0.71849}, "mean_gate": {"surface": 0.23764}, "n_reads": 4714, "n_cand": 6561, "read_suppressed_frac": 0.0991, "dropout_reads_logged": 4714, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 440.07696533203125, "n_util_obs": 1353, "util_count_spearman": 0.15569331121739177}, "probe_overhead_sec_total": 338.499, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.016816, "alpha": 0.03, "t": 3716.0} |
| {"event": "consolidate", "step": 5950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3736.9} |
| {"event": "train", "step": 5950, "arch": "lexmem_v1", "phase": 3, "objective": "global", "lm_loss": 1.37203, "bpb": 1.97942, "chunk_loss": 0.030486, "F": 0.20593, "G": 0.20996, "chunks": 210.88, "bytes_per_chunk": 4.856, "boundary_entropy": 0.0847, "router_noise": 0.0, "grad_norm": 0.7018, "lr": 3.01e-05, "router_trainable": true, "sec_per_step": 0.425, "probe_sec": 0.0349, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02302, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.1281, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.93335, "trunc_abs_frac": 0.78792, "mean_delta_full": 0.32227, "read_hit_rate": {"surface": 0.7165}, "mean_gate": {"surface": 0.23609}, "n_reads": 4812, "n_cand": 6716, "read_suppressed_frac": 0.1016, "dropout_reads_logged": 4812, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 443.762939453125, "n_util_obs": 1368, "util_count_spearman": 0.16388774397557482}, "probe_overhead_sec_total": 341.1, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.020013, "alpha": 0.03, "t": 3737.3} |
| {"event": "consolidate", "step": 6000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 3754.3} |
| {"event": "train", "step": 5999, "arch": "lexmem_v1", "phase": 3, "objective": "global", "lm_loss": 1.37472, "bpb": 1.98331, "chunk_loss": 0.03054, "F": 0.20837, "G": 0.21212, "chunks": 213.38, "bytes_per_chunk": 4.799, "boundary_entropy": 0.08332, "router_noise": 0.0, "grad_norm": 0.6855, "lr": 3e-05, "router_trainable": true, "sec_per_step": 0.342, "probe_sec": 0.0265, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02005, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.1338, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -49.17391, "trunc_abs_frac": 0.65152, "mean_delta_full": 0.00056, "read_hit_rate": {"surface": 0.72719}, "mean_gate": {"surface": 0.22903}, "n_reads": 4942, "n_cand": 6796, "read_suppressed_frac": 0.1034, "dropout_reads_logged": 4942, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 447.3790283203125, "n_util_obs": 1378, "util_count_spearman": 0.1662517158296253}, "probe_overhead_sec_total": 343.291, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.021259, "alpha": 0.03, "t": 3754.3} |
| {"event": "probe", "step": 6000, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3539.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.203356, "cps": 27.648438, "mean_p": 0.207701, "boundary_entropy": 0.085018, "bpb": 1.861703}, "tr": {"chunks": 3433.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.189637, "cps": 26.820312, "mean_p": 0.186103, "boundary_entropy": 0.103428, "bpb": 1.99843}, "fi": {"chunks": 4495.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.238956, "cps": 35.117188, "mean_p": 0.234828, "boundary_entropy": 0.081968, "bpb": 2.095376}}, "bpb": {"en": 1.8617, "tr": 1.99843, "fi": 2.09538}, "gini_cpc": 0.052028, "gini_cps": 0.061742, "bpic": 4.947414, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.932538583297731, "fi": 1.175063591551954}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 447.3790283203125, "n_util_obs": 1378, "util_count_spearman": 0.1662517158296253}, "t": 3754.9} |
| {"event": "eval", "step": 6000, "phase": 3, "bpb": {"en": 1.87645, "tr": 2.00775, "fi": 2.09939}, "bytes_per_chunk": {"en": 4.852, "tr": 5.822, "fi": 4.628}, "bpb_mean": 1.99453, "frozen_inventory": true, "t": 3755.7} |
| {"event": "reps_dumped", "dirs": ["/local/scratch/sas245/beetle-hnet/runs/ali_lexmem-v1_s2/probe/reps/th0.30", "/local/scratch/sas245/beetle-hnet/runs/ali_lexmem-v1_s2/probe/reps/th0.40", "/local/scratch/sas245/beetle-hnet/runs/ali_lexmem-v1_s2/probe/reps/th0.50", "/local/scratch/sas245/beetle-hnet/runs/ali_lexmem-v1_s2/probe/reps/th0.60", "/local/scratch/sas245/beetle-hnet/runs/ali_lexmem-v1_s2/probe/reps/th0.70"], "t": 3758.2} |
| {"event": "final", "run": "ali_lexmem-v1_s2", "arch": "lexmem_v1", "objective": "ratio", "condition": "A", "step": 6000, "params": 22719553, "seed": 2, "per_lang": {"en": {"cpc": 0.20335574326265587, "cps": 27.6484375, "bpb": 1.8823098697531386, "bytes_per_chunk": 4.791913061757959, "reads_per_chunk": 0.7401266783047409}, "tr": {"cpc": 0.1896370767276142, "cps": 26.8203125, "bpb": 1.9991002460109348, "bytes_per_chunk": 5.833459432996573, "reads_per_chunk": 0.6492389514442121}, "fi": {"cpc": 0.2389559300409335, "cps": 35.1171875, "bpb": 2.1171532866446188, "bytes_per_chunk": 4.599864887391537, "reads_per_chunk": 0.6566735977680099}}, "bpb_mean": 1.9995211341362307, "gini_cpc": 0.05202832593717366, "gini_cps": 0.061742391209558045, "bpic": 4.947414319351181, "mean_boundary_entropy": 0.08331762999296188, "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": 343.29130029678345, "probe_overhead_seq_frac": 0.499890625, "wall_sec": 3760.2309215068817, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 447.3790283203125, "n_util_obs": 1378, "util_count_spearman": 0.1662517158296253}, "ali_params": 202561, "ali_config": {"enabled": true, "budget_B": 65536, "staging_capacity": 262144, "policy": "frequency", "candidate_types": ["surface"], "gate": "scalar", "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": 0.1662517158296253, "paired_vs_dropout": {"n": 2906, "pearson": 0.08378102282529734, "spearman": 0.09772621039636144, "mean_abs_discrepancy": 0.9626404643058777, "mean_paired": 0.09258974343538284, "mean_dropout": 0.07520633190870285}, "predictor_calibration": {"n": 6655, "all": {"n": 6655, "pearson": 0.12455224372004542, "spearman": 0.21281410091092712, "mae": 0.13819727301597595, "reliability": [{"pred": 0.002805632073432207, "obs": 0.026666948571801186, "n": 666}, {"pred": 0.012120313942432404, "obs": 0.031520191580057144, "n": 665}, {"pred": 0.01786142587661743, "obs": 0.04358869418501854, "n": 666}, {"pred": 0.022935178130865097, "obs": 0.06163495406508446, "n": 665}, {"pred": 0.028229784220457077, "obs": 0.06935030221939087, "n": 665}, {"pred": 0.03458526358008385, "obs": 0.06276576966047287, "n": 666}, {"pred": 0.041279543191194534, "obs": 0.09939464181661606, "n": 665}, {"pred": 0.04989882931113243, "obs": 0.08988390117883682, "n": 666}, {"pred": 0.06324664503335953, "obs": 0.10193153470754623, "n": 665}, {"pred": 0.09470333904027939, "obs": 0.1347000002861023, "n": 666}]}, "admitted": {"n": 6655, "pearson": 0.12455224372004542, "spearman": 0.21281410091092712, "mae": 0.13819727301597595, "reliability": [{"pred": 0.002805632073432207, "obs": 0.026666948571801186, "n": 666}, {"pred": 0.012120313942432404, "obs": 0.031520191580057144, "n": 665}, {"pred": 0.01786142587661743, "obs": 0.04358869418501854, "n": 666}, {"pred": 0.022935178130865097, "obs": 0.06163495406508446, "n": 665}, {"pred": 0.028229784220457077, "obs": 0.06935030221939087, "n": 665}, {"pred": 0.03458526358008385, "obs": 0.06276576966047287, "n": 666}, {"pred": 0.041279543191194534, "obs": 0.09939464181661606, "n": 665}, {"pred": 0.04989882931113243, "obs": 0.08988390117883682, "n": 666}, {"pred": 0.06324664503335953, "obs": 0.10193153470754623, "n": 665}, {"pred": 0.09470333904027939, "obs": 0.1347000002861023, "n": 666}]}, "staged": {"n": 0, "pearson": NaN, "spearman": NaN, "mae": NaN}}, "t": 3762.3} |
| {"event": "hub_done", "repo": "AdaptiveChunking/ali_lexmem-v1_s2", "pushed": [0, 1, 2, 4, 6, 8, 16, 32, 64, 128, 256, 512, 1000, 1024, 2048, 3000, 4096, 6000], "failed": [], "ok": true, "t": 3777.0} |
|
|
