| {"event": "hub", "repo": "AdaptiveChunking/ali_ali-noskel_s2", "keep_steps": [0, 1, 2, 4, 6, 8, 16, 32, 64, 128, 256, 512, 1000, 1024, 2048, 3000, 4096, 6000], "t": 3.5} |
| {"event": "start", "run": "ali_ali-noskel_s2", "arch": "ali", "objective": "ratio", "condition": "A", "params": 22785088, "langs": ["en", "tr", "fi"], "premium": {"en": 1.0, "tr": 1.12, "fi": 1.104}, "schedule": {"tau1": 0.4, "tau2": 0.7, "p1_objective": "global", "p2_objective": "global", "p3_objective": "global", "freeze_in_p3": false, "freeze_scope": "router"}, "corpus": "/local/scratch/sas245/beetle-hnet/scratch/data/tier1_corpus", "premium_kind": "flores_per_sent", "doc_filtered": {}, "weights": {"en": 0.74, "tr": 0.1536, "fi": 0.1064}, "device": "cuda", "ali_params": 268096, "ali_config": {"enabled": true, "budget_B": 65536, "staging_capacity": 262144, "policy": "utility", "candidate_types": ["surface", "split", "merge"], "gate": "vector", "retrieval": true, "memory_scope": "persistent", "read_dropout": 0.1, "write_rule": "mean", "eta_min": 0.0, "consolidate_every": 50, "probe_seqs": 2, "probe_reps": 8, "horizon": 32, "predictor": "mlp", "lambda_util": 0.1, "max_span": 64, "byte_classes_path": "/local/scratch/sas245/beetle-hnet/scratch/data/byte_classes.json", "gate_hidden": 0, "type_dim": 16, "gate_bias_init": -2.0}, "t": 3.5} |
| {"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": 3.5} |
| {"event": "probe", "step": 0, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 10933.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.628225, "cps": 85.414062, "mean_p": 0.513249, "boundary_entropy": 0.686295, "bpb": 8.12598}, "tr": {"chunks": 8187.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.452245, "cps": 63.960938, "mean_p": 0.498002, "boundary_entropy": 0.686826, "bpb": 8.097536}, "fi": {"chunks": 13540.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.719792, "cps": 105.78125, "mean_p": 0.517169, "boundary_entropy": 0.686661, "bpb": 8.065117}}, "bpb": {"en": 8.12598, "tr": 8.09754, "fi": 8.06512}, "gini_cpc": 0.099077, "gini_cps": 0.109267, "bpic": 1.737048, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.7198781817505053, "fi": 1.1457544508645294}, "inventory": {"size": 0, "admitted": 0, "staged": 0, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 0.0, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 6.3} |
| {"event": "train", "step": 0, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 5.62735, "bpb": 8.11854, "chunk_loss": 0.062086, "F": 0.59833, "G": 0.50836, "chunks": 612.69, "bytes_per_chunk": 1.671, "boundary_entropy": 0.69105, "router_noise": 0.0, "grad_norm": 11.4934, "lr": 1e-06, "router_trainable": true, "sec_per_step": 0.042, "probe_sec": 0.0694, "probe_extra_seqs": 0, "n_rows": 0, "extra_fwd_seqs": 0, "wall_s": 0.0, "overhead_frac_est": 0.0, "chunk_mismatch": 0, "n_loo": 0, "mean_delta_loo": NaN, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "read_hit_rate": {"surface": 0.0, "split": 0.0, "merge": 0.0}, "mean_gate": {"surface": null, "split": null, "merge": null}, "n_reads": 0, "n_cand": 54182, "read_suppressed_frac": 0.0, "dropout_reads_logged": 0, "inventory": {"size": 4056, "admitted": 0, "staged": 4056, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 2006, "admitted_split": 0, "staged_split": 1198, "admitted_merge": 0, "staged_merge": 852, "admitted_skel": 0, "staged_skel": 0, "mean_count": 0.0, "n_util_obs": 0, "util_count_spearman": NaN}, "probe_overhead_sec_total": 0.069, "probe_overhead_seq_frac": 0.0, "util_loss": -0.0, "alpha": 0.03, "t": 9.7} |
| {"event": "probe", "step": 1, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 10908.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.626788, "cps": 85.21875, "mean_p": 0.513143, "boundary_entropy": 0.686304, "bpb": 8.09432}, "tr": {"chunks": 8204.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.453185, "cps": 64.09375, "mean_p": 0.498082, "boundary_entropy": 0.686828, "bpb": 8.076211}, "fi": {"chunks": 13506.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.717984, "cps": 105.515625, "mean_p": 0.517038, "boundary_entropy": 0.686672, "bpb": 8.036117}}, "bpb": {"en": 8.09432, "tr": 8.07621, "fi": 8.03612}, "gini_cpc": 0.098185, "gini_cps": 0.108366, "bpic": 1.739285, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.723026293680575, "fi": 1.145496727653734}, "inventory": {"size": 4056, "admitted": 0, "staged": 4056, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 2006, "admitted_split": 0, "staged_split": 1198, "admitted_merge": 0, "staged_merge": 852, "admitted_skel": 0, "staged_skel": 0, "mean_count": 0.0, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 11.5} |
| {"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": 0.512978, "boundary_entropy": 0.686318, "bpb": 8.031168}, "tr": {"chunks": 8248.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.455615, "cps": 64.4375, "mean_p": 0.498244, "boundary_entropy": 0.686832, "bpb": 8.032003}, "fi": {"chunks": 13468.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.715964, "cps": 105.21875, "mean_p": 0.516798, "boundary_entropy": 0.686692, "bpb": 7.976288}}, "bpb": {"en": 8.03117, "tr": 8.032, "fi": 7.97629}, "gini_cpc": 0.096603, "gini_cps": 0.106765, "bpic": 1.740512, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.7288417534704552, "fi": 1.1453187423917284}, "inventory": {"size": 7189, "admitted": 0, "staged": 7189, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 3534, "admitted_split": 0, "staged_split": 2218, "admitted_merge": 0, "staged_merge": 1437, "admitted_skel": 0, "staged_skel": 0, "mean_count": 0.0, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 16.9} |
| {"event": "probe", "step": 4, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 10715.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.615698, "cps": 83.710938, "mean_p": 0.512102, "boundary_entropy": 0.68638, "bpb": 7.819661}, "tr": {"chunks": 8387.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.463293, "cps": 65.523438, "mean_p": 0.49867, "boundary_entropy": 0.686846, "bpb": 7.878755}, "fi": {"chunks": 13214.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.702461, "cps": 103.234375, "mean_p": 0.515685, "boundary_entropy": 0.686774, "bpb": 7.773652}}, "bpb": {"en": 7.81966, "tr": 7.87875, "fi": 7.77365}, "gini_cpc": 0.089503, "gini_cps": 0.099579, "bpic": 1.755539, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.7524680015163916, "fi": 1.1409178211148336}, "inventory": {"size": 12288, "admitted": 0, "staged": 12288, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 4, "admitted_surface": 0, "staged_surface": 5973, "admitted_split": 0, "staged_split": 3779, "admitted_merge": 0, "staged_merge": 2536, "admitted_skel": 0, "staged_skel": 0, "mean_count": 0.0, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 23.8} |
| {"event": "probe", "step": 6, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 10329.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.593518, "cps": 80.695312, "mean_p": 0.510353, "boundary_entropy": 0.686483, "bpb": 7.542821}, "tr": {"chunks": 8506.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.469867, "cps": 66.453125, "mean_p": 0.499187, "boundary_entropy": 0.686869, "bpb": 7.674391}, "fi": {"chunks": 12734.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.676944, "cps": 99.484375, "mean_p": 0.513643, "boundary_entropy": 0.686908, "bpb": 7.518254}}, "bpb": {"en": 7.54282, "tr": 7.67439, "fi": 7.51825}, "gini_cpc": 0.079325, "gini_cps": 0.089286, "bpic": 1.797079, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.7916635868887795, "fi": 1.1405617550356804}, "inventory": {"size": 16100, "admitted": 0, "staged": 16100, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 7, "admitted_surface": 0, "staged_surface": 7706, "admitted_split": 0, "staged_split": 4788, "admitted_merge": 0, "staged_merge": 3606, "admitted_skel": 0, "staged_skel": 0, "mean_count": 0.0, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 30.0} |
| {"event": "probe", "step": 8, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 9628.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.553238, "cps": 75.21875, "mean_p": 0.507286, "boundary_entropy": 0.686614, "bpb": 7.293067}, "tr": {"chunks": 8638.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.477158, "cps": 67.484375, "mean_p": 0.499595, "boundary_entropy": 0.686902, "bpb": 7.478032}, "fi": {"chunks": 11875.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.63128, "cps": 92.773438, "mean_p": 0.510289, "boundary_entropy": 0.687081, "bpb": 7.295834}}, "bpb": {"en": 7.29307, "tr": 7.47803, "fi": 7.29583}, "gini_cpc": 0.061834, "gini_cps": 0.071597, "bpic": 1.88222, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.8624832844397914, "fi": 1.141063394763872}, "inventory": {"size": 20465, "admitted": 0, "staged": 20465, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 9, "admitted_surface": 0, "staged_surface": 9814, "admitted_split": 0, "staged_split": 6098, "admitted_merge": 0, "staged_merge": 4553, "admitted_skel": 0, "staged_skel": 0, "mean_count": 0.0, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 35.8} |
| {"event": "probe", "step": 16, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3750.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.21548, "cps": 29.296875, "mean_p": 0.484384, "boundary_entropy": 0.68616, "bpb": 6.724705}, "tr": {"chunks": 7274.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.401812, "cps": 56.828125, "mean_p": 0.494379, "boundary_entropy": 0.686996, "bpb": 7.013502}, "fi": {"chunks": 4587.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.243847, "cps": 35.835938, "mean_p": 0.486534, "boundary_entropy": 0.686903, "bpb": 6.787243}}, "bpb": {"en": 6.72471, "tr": 7.0135, "fi": 6.78724}, "gini_cpc": 0.144252, "gini_cps": 0.150492, "bpic": 3.634104, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.8647284538474285, "fi": 1.1316436978363722}, "inventory": {"size": 48205, "admitted": 0, "staged": 48205, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 23, "admitted_surface": 0, "staged_surface": 23788, "admitted_split": 0, "staged_split": 17658, "admitted_merge": 0, "staged_merge": 6759, "admitted_skel": 0, "staged_skel": 0, "mean_count": 0.0, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 43.9} |
| {"event": "probe", "step": 32, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 2597.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.149227, "cps": 20.289062, "mean_p": 0.486706, "boundary_entropy": 0.686945, "bpb": 6.304665}, "tr": {"chunks": 4632.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.255869, "cps": 36.1875, "mean_p": 0.490155, "boundary_entropy": 0.687558, "bpb": 6.651791}, "fi": {"chunks": 1390.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.073893, "cps": 10.859375, "mean_p": 0.479382, "boundary_entropy": 0.686971, "bpb": 6.42083}}, "bpb": {"en": 6.30467, "tr": 6.65179, "fi": 6.42083}, "gini_cpc": 0.253278, "gini_cps": 0.250764, "bpic": 6.582202, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.7146290199976852, "fi": 0.49517086928815257}, "inventory": {"size": 102604, "admitted": 0, "staged": 102604, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 47, "admitted_surface": 0, "staged_surface": 50340, "admitted_split": 0, "staged_split": 43557, "admitted_merge": 0, "staged_merge": 8707, "admitted_skel": 0, "staged_skel": 0, "mean_count": 0.0, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 64.6} |
| {"event": "consolidate", "step": 50, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 3, "swapped": 0, "uncertain_staged": 167764, "t": 84.9} |
| {"event": "train", "step": 50, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 4.21234, "bpb": 6.07712, "chunk_loss": 0.034216, "F": 0.22592, "G": 0.48849, "chunks": 231.34, "bytes_per_chunk": 4.426, "boundary_entropy": 0.69161, "router_noise": 0.0, "grad_norm": 2.479, "lr": 5.1e-05, "router_trainable": true, "sec_per_step": 1.513, "probe_sec": 0.0541, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03823, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.00415, "n_shadow": 3, "mean_delta_shadow": 0.00405, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": -0.01538, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -0.11439, "trunc_abs_frac": 0.43947, "mean_delta_full": 0.01725, "read_hit_rate": {"surface": 0.02484, "split": 0.04349, "merge": 0.0}, "mean_gate": {"surface": 0.11955, "split": 0.11961, "merge": null}, "n_reads": 585, "n_cand": 23936, "read_suppressed_frac": 0.1043, "dropout_reads_logged": 585, "inventory": {"size": 171707, "admitted": 3, "staged": 171704, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 82, "admitted_surface": 2, "staged_surface": 84458, "admitted_split": 1, "staged_split": 76815, "admitted_merge": 0, "staged_merge": 10431, "admitted_skel": 0, "staged_skel": 0, "mean_count": 26166.0, "n_util_obs": 3, "util_count_spearman": -1.0}, "probe_overhead_sec_total": 5.047, "probe_overhead_seq_frac": 0.489, "util_loss": 0.075457, "alpha": 0.03, "t": 85.3} |
| {"event": "probe", "step": 64, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 4801.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.275872, "cps": 37.507812, "mean_p": 0.485097, "boundary_entropy": 0.685501, "bpb": 5.552077}, "tr": {"chunks": 5408.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.298735, "cps": 42.25, "mean_p": 0.486591, "boundary_entropy": 0.686482, "bpb": 5.956325}, "fi": {"chunks": 4719.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.250864, "cps": 36.867188, "mean_p": 0.484059, "boundary_entropy": 0.686335, "bpb": 5.717691}}, "bpb": {"en": 5.55208, "tr": 5.95633, "fi": 5.71769}, "gini_cpc": 0.038662, "gini_cps": 0.03077, "bpic": 3.800375, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0828755443851386, "fi": 0.9093488211609911}, "inventory": {"size": 222822, "admitted": 3, "staged": 222819, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 91, "admitted_surface": 2, "staged_surface": 110907, "admitted_split": 1, "staged_split": 100012, "admitted_merge": 0, "staged_merge": 11900, "admitted_skel": 0, "staged_skel": 0, "mean_count": 27352.66796875, "n_util_obs": 3, "util_count_spearman": -0.5}, "t": 102.3} |
| {"event": "consolidate", "step": 100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 65533, "swapped": 3, "uncertain_staged": 262139, "t": 147.8} |
| {"event": "train", "step": 100, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 3.24839, "bpb": 4.68644, "chunk_loss": 0.029467, "F": 0.15787, "G": 0.4519, "chunks": 161.66, "bytes_per_chunk": 6.334, "boundary_entropy": 0.68182, "router_noise": 0.0, "grad_norm": 1.657, "lr": 0.000101, "router_trainable": true, "sec_per_step": 1.255, "probe_sec": 0.0727, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06339, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.01721, "n_shadow": 6, "mean_delta_shadow": 0.01762, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.66899, "trunc_abs_frac": 0.87379, "mean_delta_full": 0.08984, "read_hit_rate": {"surface": 0.19011, "split": 0.0, "merge": 0.0}, "mean_gate": {"surface": 0.11978, "split": null, "merge": null}, "n_reads": 977, "n_cand": 18913, "read_suppressed_frac": 0.1095, "dropout_reads_logged": 977, "inventory": {"size": 265991, "admitted": 65536, "staged": 200455, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 118, "admitted_surface": 64983, "staged_surface": 68243, "admitted_split": 545, "staged_split": 117125, "admitted_merge": 8, "staged_merge": 15087, "admitted_skel": 0, "staged_skel": 0, "mean_count": 3.934478759765625, "n_util_obs": 0, "util_count_spearman": NaN}, "probe_overhead_sec_total": 10.669, "probe_overhead_seq_frac": 0.4944, "util_loss": 0.002791, "alpha": 0.03, "t": 148.1} |
| {"event": "probe", "step": 128, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3881.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.223008, "cps": 30.320312, "mean_p": 0.415241, "boundary_entropy": 0.643199, "bpb": 4.150659}, "tr": {"chunks": 4902.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.270784, "cps": 38.296875, "mean_p": 0.437762, "boundary_entropy": 0.659583, "bpb": 4.387585}, "fi": {"chunks": 3837.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.203976, "cps": 29.976562, "mean_p": 0.42882, "boundary_entropy": 0.659956, "bpb": 4.342376}}, "bpb": {"en": 4.15066, "tr": 4.38759, "fi": 4.34238}, "gini_cpc": 0.06383, "gini_cps": 0.05626, "bpic": 4.495404, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.2142363582616083, "fi": 0.9146614875859131}, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 152, "admitted_surface": 64983, "staged_surface": 110480, "admitted_split": 545, "staged_split": 133030, "admitted_merge": 8, "staged_merge": 18634, "admitted_skel": 0, "staged_skel": 0, "mean_count": 4.4665374755859375, "n_util_obs": 5, "util_count_spearman": -0.3}, "t": 174.3} |
| {"event": "consolidate", "step": 150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 12, "uncertain_staged": 262135, "t": 198.3} |
| {"event": "train", "step": 150, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.78736, "bpb": 4.02131, "chunk_loss": 0.031289, "F": 0.19339, "G": 0.38246, "chunks": 198.03, "bytes_per_chunk": 5.171, "boundary_entropy": 0.61899, "router_noise": 0.0, "grad_norm": 1.3213, "lr": 0.000151, "router_trainable": true, "sec_per_step": 1.039, "probe_sec": 0.2657, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.15304, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.00994, "n_shadow": 3, "mean_delta_shadow": 0.00366, "n_split": 2, "mean_delta_split": 0.00433, "n_merge": 1, "mean_delta_merge": -0.02049, "n_dropped_unpaired": 0, "n_full": 0, "read_hit_rate": {"surface": 0.12117, "split": 0.0, "merge": 0.0}, "mean_gate": {"surface": 0.12195, "split": null, "merge": null}, "n_reads": 764, "n_cand": 22250, "read_suppressed_frac": 0.0825, "dropout_reads_logged": 764, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 177, "admitted_surface": 64983, "staged_surface": 123116, "admitted_split": 545, "staged_split": 117996, "admitted_merge": 8, "staged_merge": 21032, "admitted_skel": 0, "staged_skel": 0, "mean_count": 4.16943359375, "n_util_obs": 0, "util_count_spearman": NaN}, "probe_overhead_sec_total": 15.735, "probe_overhead_seq_frac": 0.4963, "util_loss": 0.001036, "alpha": 0.03, "t": 200.0} |
| {"event": "consolidate", "step": 200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 14, "uncertain_staged": 262135, "t": 247.8} |
| {"event": "probe", "step": 200, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3599.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.206803, "cps": 28.117188, "mean_p": 0.368987, "boundary_entropy": 0.586958, "bpb": 3.629008}, "tr": {"chunks": 4177.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.230735, "cps": 32.632812, "mean_p": 0.382859, "boundary_entropy": 0.598527, "bpb": 3.802831}, "fi": {"chunks": 3694.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.196374, "cps": 28.859375, "mean_p": 0.399595, "boundary_entropy": 0.631834, "bpb": 3.860708}}, "bpb": {"en": 3.62901, "tr": 3.80283, "fi": 3.86071}, "gini_cpc": 0.036136, "gini_cps": 0.033595, "bpic": 4.94612, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.1157225156772865, "fi": 0.9495706468059035}, "inventory": {"size": 327666, "admitted": 65536, "staged": 262130, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 239, "admitted_surface": 64974, "staged_surface": 122868, "admitted_split": 546, "staged_split": 113196, "admitted_merge": 16, "staged_merge": 26066, "admitted_skel": 0, "staged_skel": 0, "mean_count": 3.5219268798828125, "n_util_obs": 2, "util_count_spearman": NaN}, "t": 248.8} |
| {"event": "eval", "step": 200, "phase": 1, "bpb": {"en": 3.63632, "tr": 3.81359, "fi": 3.85969}, "bytes_per_chunk": {"en": 5.309, "tr": 5.399, "fi": 5.885}, "bpb_mean": 3.76987, "frozen_inventory": true, "t": 251.1} |
| {"event": "train", "step": 200, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.52842, "bpb": 3.64773, "chunk_loss": 0.030427, "F": 0.17639, "G": 0.35968, "chunks": 180.62, "bytes_per_chunk": 5.669, "boundary_entropy": 0.58882, "router_noise": 0.0, "grad_norm": 1.7291, "lr": 0.000201, "router_trainable": true, "sec_per_step": 1.143, "probe_sec": 0.0903, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05575, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.06412, "n_shadow": 4, "mean_delta_shadow": -0.04898, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": -0.052, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.51452, "trunc_abs_frac": 0.4097, "mean_delta_full": -0.09039, "read_hit_rate": {"surface": 0.04088, "split": 0.0, "merge": 0.0}, "mean_gate": {"surface": 0.13445, "split": null, "merge": null}, "n_reads": 235, "n_cand": 21154, "read_suppressed_frac": 0.1106, "dropout_reads_logged": 235, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 191, "admitted_surface": 64974, "staged_surface": 123192, "admitted_split": 546, "staged_split": 113549, "admitted_merge": 16, "staged_merge": 25403, "admitted_skel": 0, "staged_skel": 0, "mean_count": 3.5255126953125, "n_util_obs": 2, "util_count_spearman": NaN}, "probe_overhead_sec_total": 20.74, "probe_overhead_seq_frac": 0.4972, "util_loss": 0.000774, "alpha": 0.03, "t": 257.2} |
| {"event": "consolidate", "step": 250, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262129, "t": 316.2} |
| {"event": "train", "step": 250, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.38709, "bpb": 3.44385, "chunk_loss": 0.031181, "F": 0.1994, "G": 0.32941, "chunks": 204.19, "bytes_per_chunk": 5.015, "boundary_entropy": 0.53193, "router_noise": 0.0, "grad_norm": 2.2443, "lr": 0.000251, "router_trainable": true, "sec_per_step": 1.212, "probe_sec": 0.0922, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0587, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": 0.0294, "n_shadow": 6, "mean_delta_shadow": -0.02063, "n_split": 1, "mean_delta_split": 0.08264, "n_merge": 1, "mean_delta_merge": -0.03882, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -0.74935, "trunc_abs_frac": 1.60813, "mean_delta_full": 0.03117, "read_hit_rate": {"surface": 0.03522, "split": 0.01253, "merge": 0.00309}, "mean_gate": {"surface": 0.13629, "split": 0.13509, "merge": 0.137}, "n_reads": 386, "n_cand": 23902, "read_suppressed_frac": 0.0933, "dropout_reads_logged": 386, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 153, "admitted_surface": 63949, "staged_surface": 128584, "admitted_split": 1046, "staged_split": 106620, "admitted_merge": 541, "staged_merge": 26940, "admitted_skel": 0, "staged_skel": 0, "mean_count": 4.261810302734375, "n_util_obs": 13, "util_count_spearman": 0.489010989010989}, "probe_overhead_sec_total": 26.515, "probe_overhead_seq_frac": 0.4978, "util_loss": 0.000882, "alpha": 0.03, "t": 317.8} |
| {"event": "probe", "step": 256, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3652.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.209849, "cps": 28.53125, "mean_p": 0.330024, "boundary_entropy": 0.522512, "bpb": 3.350428}, "tr": {"chunks": 4499.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.248522, "cps": 35.148438, "mean_p": 0.323535, "boundary_entropy": 0.520775, "bpb": 3.519069}, "fi": {"chunks": 4418.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.234863, "cps": 34.515625, "mean_p": 0.371387, "boundary_entropy": 0.589606, "bpb": 3.650777}}, "bpb": {"en": 3.35043, "tr": 3.51907, "fi": 3.65078}, "gini_cpc": 0.037191, "gini_cps": 0.044925, "bpic": 4.513645, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.184291993139658, "fi": 1.119198654652148}, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 160, "admitted_surface": 63949, "staged_surface": 128290, "admitted_split": 1046, "staged_split": 106207, "admitted_merge": 541, "staged_merge": 27647, "admitted_skel": 0, "staged_skel": 0, "mean_count": 4.2892303466796875, "n_util_obs": 14, "util_count_spearman": 0.2043956043956044}, "t": 324.2} |
| {"event": "consolidate", "step": 300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262128, "t": 379.2} |
| {"event": "train", "step": 300, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.27732, "bpb": 3.28548, "chunk_loss": 0.030652, "F": 0.18829, "G": 0.30206, "chunks": 192.81, "bytes_per_chunk": 5.311, "boundary_entropy": 0.48002, "router_noise": 0.0, "grad_norm": 1.3753, "lr": 0.0003, "router_trainable": true, "sec_per_step": 1.261, "probe_sec": 0.1267, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07362, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.03274, "n_shadow": 8, "mean_delta_shadow": -0.00155, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.03955, "trunc_abs_frac": 0.03955, "mean_delta_full": 0.57104, "read_hit_rate": {"surface": 0.0233, "split": 0.06098, "merge": 0.00295}, "mean_gate": {"surface": 0.13013, "split": 0.1291, "merge": 0.12953}, "n_reads": 813, "n_cand": 22936, "read_suppressed_frac": 0.1009, "dropout_reads_logged": 813, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 192, "admitted_surface": 62927, "staged_surface": 127171, "admitted_split": 1050, "staged_split": 102619, "admitted_merge": 1559, "staged_merge": 32354, "admitted_skel": 0, "staged_skel": 0, "mean_count": 6.463409423828125, "n_util_obs": 19, "util_count_spearman": 0.20877192982456141}, "probe_overhead_sec_total": 32.008, "probe_overhead_seq_frac": 0.4981, "util_loss": 0.001175, "alpha": 0.03, "t": 380.8} |
| {"event": "consolidate", "step": 350, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262128, "t": 436.4} |
| {"event": "train", "step": 350, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.17458, "bpb": 3.13725, "chunk_loss": 0.030718, "F": 0.19165, "G": 0.29313, "chunks": 196.25, "bytes_per_chunk": 5.218, "boundary_entropy": 0.44505, "router_noise": 0.0, "grad_norm": 1.2424, "lr": 0.0002999, "router_trainable": true, "sec_per_step": 1.145, "probe_sec": 0.1096, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06885, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.048, "n_shadow": 4, "mean_delta_shadow": -0.06696, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.01062, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -4.3909, "trunc_abs_frac": 0.6488, "mean_delta_full": -0.02063, "read_hit_rate": {"surface": 0.04145, "split": 0.0724, "merge": 0.00113}, "mean_gate": {"surface": 0.12598, "split": 0.12633, "merge": 0.12558}, "n_reads": 1052, "n_cand": 23320, "read_suppressed_frac": 0.1141, "dropout_reads_logged": 1052, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 199, "admitted_surface": 63041, "staged_surface": 122633, "admitted_split": 508, "staged_split": 97214, "admitted_merge": 1987, "staged_merge": 42297, "admitted_skel": 0, "staged_skel": 0, "mean_count": 7.709320068359375, "n_util_obs": 19, "util_count_spearman": -0.3385964912280702}, "probe_overhead_sec_total": 37.181, "probe_overhead_seq_frac": 0.4984, "util_loss": 0.001326, "alpha": 0.03, "t": 438.1} |
| {"event": "consolidate", "step": 400, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 611, "uncertain_staged": 262123, "t": 496.1} |
| {"event": "probe", "step": 400, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3217.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.184853, "cps": 25.132812, "mean_p": 0.282175, "boundary_entropy": 0.408346, "bpb": 2.905633}, "tr": {"chunks": 3172.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.17522, "cps": 24.78125, "mean_p": 0.273707, "boundary_entropy": 0.426086, "bpb": 3.185322}, "fi": {"chunks": 4105.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.218223, "cps": 32.070312, "mean_p": 0.280931, "boundary_entropy": 0.44476, "bpb": 3.369582}}, "bpb": {"en": 2.90563, "tr": 3.18532, "fi": 3.36958}, "gini_cpc": 0.049575, "gini_cps": 0.059272, "bpic": 5.406137, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9478850780833078, "fi": 1.1805226860579183}, "inventory": {"size": 327068, "admitted": 65536, "staged": 261532, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 232, "admitted_surface": 62999, "staged_surface": 115856, "admitted_split": 549, "staged_split": 90667, "admitted_merge": 1988, "staged_merge": 55009, "admitted_skel": 0, "staged_skel": 0, "mean_count": 19.0716552734375, "n_util_obs": 26, "util_count_spearman": -0.3593162393162393}, "t": 497.5} |
| {"event": "eval", "step": 400, "phase": 1, "bpb": {"en": 2.91722, "tr": 3.19148, "fi": 3.36164}, "bytes_per_chunk": {"en": 5.489, "tr": 6.44, "fi": 5.12}, "bpb_mean": 3.15678, "frozen_inventory": true, "t": 499.9} |
| {"event": "train", "step": 400, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.08404, "bpb": 3.00663, "chunk_loss": 0.03017, "F": 0.17203, "G": 0.27136, "chunks": 176.16, "bytes_per_chunk": 5.813, "boundary_entropy": 0.40561, "router_noise": 0.0, "grad_norm": 0.9069, "lr": 0.0002998, "router_trainable": true, "sec_per_step": 1.363, "probe_sec": 0.1284, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.1081, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.03645, "n_shadow": 5, "mean_delta_shadow": 0.00996, "n_split": 1, "mean_delta_split": -0.05164, "n_merge": 1, "mean_delta_merge": 0.08813, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": -0.63134, "trunc_abs_frac": 0.63134, "mean_delta_full": 0.08179, "read_hit_rate": {"surface": 0.12542, "split": 0.23034, "merge": 0.00126}, "mean_gate": {"surface": 0.12454, "split": 0.12617, "merge": 0.12289}, "n_reads": 3006, "n_cand": 21146, "read_suppressed_frac": 0.0898, "dropout_reads_logged": 3006, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 233, "admitted_surface": 62999, "staged_surface": 116043, "admitted_split": 549, "staged_split": 90815, "admitted_merge": 1988, "staged_merge": 55286, "admitted_skel": 0, "staged_skel": 0, "mean_count": 19.117523193359375, "n_util_obs": 26, "util_count_spearman": -0.3935042735042735}, "probe_overhead_sec_total": 42.476, "probe_overhead_seq_frac": 0.4986, "util_loss": 0.001396, "alpha": 0.03, "t": 506.2} |
| {"event": "consolidate", "step": 450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262127, "t": 578.2} |
| {"event": "train", "step": 450, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.01516, "bpb": 2.90726, "chunk_loss": 0.030554, "F": 0.18964, "G": 0.26954, "chunks": 194.19, "bytes_per_chunk": 5.273, "boundary_entropy": 0.37423, "router_noise": 0.0, "grad_norm": 1.7231, "lr": 0.0002995, "router_trainable": true, "sec_per_step": 1.466, "probe_sec": 0.106, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06057, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": 0.05422, "n_shadow": 3, "mean_delta_shadow": -0.02089, "n_split": 2, "mean_delta_split": -0.01938, "n_merge": 2, "mean_delta_merge": -0.10367, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.94774, "trunc_abs_frac": 0.41077, "mean_delta_full": 0.10978, "read_hit_rate": {"surface": 0.14057, "split": 0.25546, "merge": 0.00455}, "mean_gate": {"surface": 0.12046, "split": 0.12102, "merge": 0.11776}, "n_reads": 3682, "n_cand": 23234, "read_suppressed_frac": 0.1065, "dropout_reads_logged": 3682, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 216, "admitted_surface": 61989, "staged_surface": 110056, "admitted_split": 553, "staged_split": 86362, "admitted_merge": 2994, "staged_merge": 65726, "admitted_skel": 0, "staged_skel": 0, "mean_count": 21.959030151367188, "n_util_obs": 37, "util_count_spearman": -0.2887624466571835}, "probe_overhead_sec_total": 48.953, "probe_overhead_seq_frac": 0.4988, "util_loss": 0.001445, "alpha": 0.03, "t": 579.5} |
| {"event": "consolidate", "step": 500, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 1024, "uncertain_staged": 262124, "t": 646.5} |
| {"event": "train", "step": 500, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.91999, "bpb": 2.76996, "chunk_loss": 0.030895, "F": 0.20377, "G": 0.26891, "chunks": 208.66, "bytes_per_chunk": 4.908, "boundary_entropy": 0.33726, "router_noise": 0.0, "grad_norm": 1.5172, "lr": 0.0002992, "router_trainable": true, "sec_per_step": 1.379, "probe_sec": 0.1134, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0525, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.04055, "n_shadow": 1, "mean_delta_shadow": 0.04077, "n_split": 2, "mean_delta_split": 0.05639, "n_merge": 1, "mean_delta_merge": -0.03027, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -0.91704, "trunc_abs_frac": 1.80385, "mean_delta_full": -0.04138, "read_hit_rate": {"surface": 0.15214, "split": 0.26721, "merge": 0.01013}, "mean_gate": {"surface": 0.12381, "split": 0.1235, "merge": 0.1211}, "n_reads": 4191, "n_cand": 24908, "read_suppressed_frac": 0.0971, "dropout_reads_logged": 4191, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 234, "admitted_surface": 60963, "staged_surface": 104006, "admitted_split": 553, "staged_split": 79407, "admitted_merge": 4020, "staged_merge": 78731, "admitted_skel": 0, "staged_skel": 0, "mean_count": 24.546951293945312, "n_util_obs": 43, "util_count_spearman": -0.27740863787375414}, "probe_overhead_sec_total": 54.955, "probe_overhead_seq_frac": 0.4989, "util_loss": 0.001653, "alpha": 0.03, "t": 648.5} |
| {"event": "probe", "step": 512, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3455.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.198529, "cps": 26.992188, "mean_p": 0.265985, "boundary_entropy": 0.345007, "bpb": 2.719836}, "tr": {"chunks": 3757.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.207535, "cps": 29.351562, "mean_p": 0.264448, "boundary_entropy": 0.354402, "bpb": 3.037176}, "fi": {"chunks": 4300.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.22859, "cps": 33.59375, "mean_p": 0.273895, "boundary_entropy": 0.352336, "bpb": 3.222976}}, "bpb": {"en": 2.71984, "tr": 3.03718, "fi": 3.22298}, "gini_cpc": 0.031577, "gini_cps": 0.048934, "bpic": 4.928075, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0453620075443837, "fi": 1.1514170089074802}, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 238, "admitted_surface": 60963, "staged_surface": 102552, "admitted_split": 553, "staged_split": 78402, "admitted_merge": 4020, "staged_merge": 81190, "admitted_skel": 0, "staged_skel": 0, "mean_count": 25.166534423828125, "n_util_obs": 45, "util_count_spearman": -0.25612648221343876}, "t": 659.5} |
| {"event": "consolidate", "step": 550, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262123, "t": 723.8} |
| {"event": "train", "step": 550, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.91259, "bpb": 2.75929, "chunk_loss": 0.030532, "F": 0.19162, "G": 0.25688, "chunks": 196.22, "bytes_per_chunk": 5.219, "boundary_entropy": 0.32284, "router_noise": 0.0, "grad_norm": 1.4428, "lr": 0.0002987, "router_trainable": true, "sec_per_step": 1.527, "probe_sec": 0.0833, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05008, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.04372, "n_shadow": 4, "mean_delta_shadow": -0.0499, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.36407, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -7.5, "trunc_abs_frac": 1.29412, "mean_delta_full": -0.00244, "read_hit_rate": {"surface": 0.12502, "split": 0.2397, "merge": 0.01094}, "mean_gate": {"surface": 0.12758, "split": 0.12448, "merge": 0.12201}, "n_reads": 3501, "n_cand": 23526, "read_suppressed_frac": 0.1023, "dropout_reads_logged": 3501, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 256, "admitted_surface": 59940, "staged_surface": 97796, "admitted_split": 551, "staged_split": 75418, "admitted_merge": 5045, "staged_merge": 88930, "admitted_skel": 0, "staged_skel": 0, "mean_count": 27.108963012695312, "n_util_obs": 50, "util_count_spearman": -0.2690516206482593}, "probe_overhead_sec_total": 61.169, "probe_overhead_seq_frac": 0.499, "util_loss": 0.001669, "alpha": 0.03, "t": 724.8} |
| {"event": "consolidate", "step": 600, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262122, "t": 787.6} |
| {"event": "probe", "step": 600, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3628.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.20847, "cps": 28.34375, "mean_p": 0.261271, "boundary_entropy": 0.307252, "bpb": 2.600995}, "tr": {"chunks": 3915.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.216262, "cps": 30.585938, "mean_p": 0.25258, "boundary_entropy": 0.298763, "bpb": 2.920506}, "fi": {"chunks": 4483.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.238318, "cps": 35.023438, "mean_p": 0.271287, "boundary_entropy": 0.30653, "bpb": 3.139255}}, "bpb": {"en": 2.601, "tr": 2.92051, "fi": 3.13926}, "gini_cpc": 0.030011, "gini_cps": 0.047397, "bpic": 4.717446, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0373804441703518, "fi": 1.1431775767269345}, "inventory": {"size": 326655, "admitted": 65536, "staged": 261119, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 270, "admitted_surface": 60814, "staged_surface": 91396, "admitted_split": 519, "staged_split": 70373, "admitted_merge": 4203, "staged_merge": 99350, "admitted_skel": 0, "staged_skel": 0, "mean_count": 29.604812622070312, "n_util_obs": 59, "util_count_spearman": -0.190239625949737}, "t": 789.0} |
| {"event": "eval", "step": 600, "phase": 1, "bpb": {"en": 2.61369, "tr": 2.92785, "fi": 3.14074}, "bytes_per_chunk": {"en": 4.955, "tr": 5.079, "fi": 4.536}, "bpb_mean": 2.89409, "frozen_inventory": true, "t": 791.7} |
| {"event": "train", "step": 600, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.89433, "bpb": 2.73293, "chunk_loss": 0.030721, "F": 0.20203, "G": 0.25243, "chunks": 206.88, "bytes_per_chunk": 4.95, "boundary_entropy": 0.29955, "router_noise": 0.0, "grad_norm": 1.189, "lr": 0.0002982, "router_trainable": true, "sec_per_step": 1.456, "probe_sec": 0.1018, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05556, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": 0.1092, "n_shadow": 4, "mean_delta_shadow": 0.12019, "n_split": 1, "mean_delta_split": -0.00757, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 1.52107, "trunc_abs_frac": 0.66206, "mean_delta_full": 0.08016, "read_hit_rate": {"surface": 0.12113, "split": 0.26464, "merge": 0.01083}, "mean_gate": {"surface": 0.13199, "split": 0.12676, "merge": 0.13068}, "n_reads": 3933, "n_cand": 24722, "read_suppressed_frac": 0.1058, "dropout_reads_logged": 3933, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 270, "admitted_surface": 60814, "staged_surface": 91724, "admitted_split": 519, "staged_split": 70537, "admitted_merge": 4203, "staged_merge": 99883, "admitted_skel": 0, "staged_skel": 0, "mean_count": 29.664825439453125, "n_util_obs": 59, "util_count_spearman": -0.19532437171244885}, "probe_overhead_sec_total": 67.053, "probe_overhead_seq_frac": 0.4991, "util_loss": 0.001809, "alpha": 0.03, "t": 797.6} |
| {"event": "consolidate", "step": 650, "retired": 0, "split_retired": 3, "merge_admitted": 0, "admitted": 3, "swapped": 1024, "uncertain_staged": 262122, "t": 861.0} |
| {"event": "train", "step": 650, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.84484, "bpb": 2.66154, "chunk_loss": 0.030772, "F": 0.20413, "G": 0.25385, "chunks": 209.03, "bytes_per_chunk": 4.899, "boundary_entropy": 0.28704, "router_noise": 0.0, "grad_norm": 1.1765, "lr": 0.0002975, "router_trainable": true, "sec_per_step": 1.295, "probe_sec": 0.1941, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.13589, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 3, "mean_delta_loo": 0.18397, "n_shadow": 9, "mean_delta_shadow": -0.10433, "n_split": 2, "mean_delta_split": -0.01318, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.71065, "trunc_abs_frac": 0.57848, "mean_delta_full": -0.22571, "read_hit_rate": {"surface": 0.12183, "split": 0.2392, "merge": 0.01208}, "mean_gate": {"surface": 0.13445, "split": 0.12801, "merge": 0.1293}, "n_reads": 3682, "n_cand": 24950, "read_suppressed_frac": 0.0942, "dropout_reads_logged": 3682, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 247, "admitted_surface": 61837, "staged_surface": 87109, "admitted_split": 227, "staged_split": 67109, "admitted_merge": 3472, "staged_merge": 107926, "admitted_skel": 0, "staged_skel": 0, "mean_count": 30.866134643554688, "n_util_obs": 58, "util_count_spearman": -0.19720692731243655}, "probe_overhead_sec_total": 73.533, "probe_overhead_seq_frac": 0.4991, "util_loss": 0.001811, "alpha": 0.03, "t": 862.4} |
| {"event": "consolidate", "step": 700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262119, "t": 932.9} |
| {"event": "train", "step": 700, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.99547, "bpb": 2.87886, "chunk_loss": 0.030913, "F": 0.20941, "G": 0.25186, "chunks": 214.44, "bytes_per_chunk": 4.775, "boundary_entropy": 0.27688, "router_noise": 0.0, "grad_norm": 1.2627, "lr": 0.0002967, "router_trainable": true, "sec_per_step": 1.438, "probe_sec": 0.0934, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05289, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": -0.02731, "n_shadow": 2, "mean_delta_shadow": 0.2739, "n_split": 2, "mean_delta_split": -0.0799, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 1.24632, "trunc_abs_frac": 2.41556, "mean_delta_full": -0.08782, "read_hit_rate": {"surface": 0.09693, "split": 0.22264, "merge": 0.01192}, "mean_gate": {"surface": 0.13289, "split": 0.12744, "merge": 0.12494}, "n_reads": 3335, "n_cand": 25270, "read_suppressed_frac": 0.0972, "dropout_reads_logged": 3335, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 258, "admitted_surface": 62837, "staged_surface": 82964, "admitted_split": 145, "staged_split": 63979, "admitted_merge": 2554, "staged_merge": 115201, "admitted_skel": 0, "staged_skel": 0, "mean_count": 32.44267272949219, "n_util_obs": 59, "util_count_spearman": -0.16230274693161895}, "probe_overhead_sec_total": 79.068, "probe_overhead_seq_frac": 0.4992, "util_loss": 0.001988, "alpha": 0.03, "t": 934.3} |
| {"event": "consolidate", "step": 750, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 1024, "uncertain_staged": 262116, "t": 996.0} |
| {"event": "train", "step": 750, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.78723, "bpb": 2.57843, "chunk_loss": 0.030686, "F": 0.20361, "G": 0.24614, "chunks": 208.5, "bytes_per_chunk": 4.911, "boundary_entropy": 0.26586, "router_noise": 0.0, "grad_norm": 1.1427, "lr": 0.0002959, "router_trainable": true, "sec_per_step": 1.283, "probe_sec": 0.1516, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08257, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.05184, "n_shadow": 5, "mean_delta_shadow": -0.01174, "n_split": 1, "mean_delta_split": 0.28613, "n_merge": 1, "mean_delta_merge": 0.01123, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.76893, "trunc_abs_frac": 1.00751, "mean_delta_full": 0.04622, "read_hit_rate": {"surface": 0.11762, "split": 0.23626, "merge": 0.00696}, "mean_gate": {"surface": 0.13354, "split": 0.12764, "merge": 0.12613}, "n_reads": 3561, "n_cand": 24820, "read_suppressed_frac": 0.0955, "dropout_reads_logged": 3561, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 287, "admitted_surface": 62837, "staged_surface": 79879, "admitted_split": 145, "staged_split": 61755, "admitted_merge": 2554, "staged_merge": 120510, "admitted_skel": 0, "staged_skel": 0, "mean_count": 34.52928161621094, "n_util_obs": 62, "util_count_spearman": -0.10271713127345068}, "probe_overhead_sec_total": 84.324, "probe_overhead_seq_frac": 0.4993, "util_loss": 0.002012, "alpha": 0.03, "t": 998.4} |
| {"event": "consolidate", "step": 800, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262114, "t": 1061.9} |
| {"event": "probe", "step": 800, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3513.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.201862, "cps": 27.445312, "mean_p": 0.25145, "boundary_entropy": 0.277387, "bpb": 2.450474}, "tr": {"chunks": 4018.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.221952, "cps": 31.390625, "mean_p": 0.232926, "boundary_entropy": 0.255345, "bpb": 2.764882}, "fi": {"chunks": 4377.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.232683, "cps": 34.195312, "mean_p": 0.256656, "boundary_entropy": 0.270115, "bpb": 2.980066}}, "bpb": {"en": 2.45047, "tr": 2.76488, "fi": 2.98007}, "gini_cpc": 0.031299, "gini_cps": 0.048371, "bpic": 4.764192, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0995256151900128, "fi": 1.152684978505}, "inventory": {"size": 326655, "admitted": 65536, "staged": 261119, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 313, "admitted_surface": 62699, "staged_surface": 77239, "admitted_split": 231, "staged_split": 60758, "admitted_merge": 2606, "staged_merge": 123122, "admitted_skel": 0, "staged_skel": 0, "mean_count": 37.24101257324219, "n_util_obs": 69, "util_count_spearman": -0.09824625502374863}, "t": 1063.4} |
| {"event": "eval", "step": 800, "phase": 1, "bpb": {"en": 2.45912, "tr": 2.77219, "fi": 2.98113}, "bytes_per_chunk": {"en": 5.048, "tr": 4.971, "fi": 4.687}, "bpb_mean": 2.73748, "frozen_inventory": true, "t": 1066.3} |
| {"event": "train", "step": 800, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.86694, "bpb": 2.69343, "chunk_loss": 0.030855, "F": 0.20709, "G": 0.24868, "chunks": 212.06, "bytes_per_chunk": 4.829, "boundary_entropy": 0.26537, "router_noise": 0.0, "grad_norm": 1.1859, "lr": 0.0002949, "router_trainable": true, "sec_per_step": 1.466, "probe_sec": 0.1244, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05794, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": 0.06045, "n_shadow": 6, "mean_delta_shadow": -0.12783, "n_split": 1, "mean_delta_split": -0.03223, "n_merge": 1, "mean_delta_merge": 0.17126, "n_dropped_unpaired": 0, "n_full": 7, "trunc_ratio_sum": 0.55383, "trunc_abs_frac": 0.75721, "mean_delta_full": 0.10925, "read_hit_rate": {"surface": 0.10912, "split": 0.24021, "merge": 0.01413}, "mean_gate": {"surface": 0.13254, "split": 0.127, "merge": 0.12415}, "n_reads": 3579, "n_cand": 24912, "read_suppressed_frac": 0.0997, "dropout_reads_logged": 3579, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 314, "admitted_surface": 62699, "staged_surface": 77455, "admitted_split": 231, "staged_split": 60903, "admitted_merge": 2606, "staged_merge": 123786, "admitted_skel": 0, "staged_skel": 0, "mean_count": 37.29566955566406, "n_util_obs": 69, "util_count_spearman": -0.09645597369382536}, "probe_overhead_sec_total": 90.356, "probe_overhead_seq_frac": 0.4993, "util_loss": 0.002955, "alpha": 0.03, "t": 1071.7} |
| {"event": "consolidate", "step": 850, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 1024, "uncertain_staged": 262111, "t": 1137.2} |
| {"event": "train", "step": 850, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.81332, "bpb": 2.61607, "chunk_loss": 0.031043, "F": 0.21469, "G": 0.24968, "chunks": 219.84, "bytes_per_chunk": 4.658, "boundary_entropy": 0.25702, "router_noise": 0.0, "grad_norm": 0.9959, "lr": 0.0002938, "router_trainable": true, "sec_per_step": 1.331, "probe_sec": 0.0804, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0466, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.02982, "n_shadow": 3, "mean_delta_shadow": 0.10903, "n_split": 1, "mean_delta_split": -0.11316, "n_merge": 1, "mean_delta_merge": 0.14117, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 1.6898, "trunc_abs_frac": 1.99978, "mean_delta_full": 0.0365, "read_hit_rate": {"surface": 0.1081, "split": 0.25772, "merge": 0.02582}, "mean_gate": {"surface": 0.13263, "split": 0.12528, "merge": 0.12531}, "n_reads": 4077, "n_cand": 26158, "read_suppressed_frac": 0.1047, "dropout_reads_logged": 4077, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 322, "admitted_surface": 61674, "staged_surface": 76155, "admitted_split": 230, "staged_split": 59578, "admitted_merge": 3632, "staged_merge": 126411, "admitted_skel": 0, "staged_skel": 0, "mean_count": 39.319244384765625, "n_util_obs": 71, "util_count_spearman": 0.001458764744893191}, "probe_overhead_sec_total": 96.177, "probe_overhead_seq_frac": 0.4993, "util_loss": 0.002989, "alpha": 0.03, "t": 1138.3} |
| {"event": "consolidate", "step": 900, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262109, "t": 1209.4} |
| {"event": "train", "step": 900, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.61977, "bpb": 2.33683, "chunk_loss": 0.030688, "F": 0.20374, "G": 0.24476, "chunks": 208.62, "bytes_per_chunk": 4.908, "boundary_entropy": 0.25163, "router_noise": 0.0, "grad_norm": 1.3818, "lr": 0.0002927, "router_trainable": true, "sec_per_step": 1.45, "probe_sec": 0.1014, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05556, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.04053, "n_shadow": 2, "mean_delta_shadow": -0.10596, "n_split": 1, "mean_delta_split": 0.14122, "n_merge": 1, "mean_delta_merge": 0.12628, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.9422, "trunc_abs_frac": 0.9422, "mean_delta_full": 0.18164, "read_hit_rate": {"surface": 0.11695, "split": 0.25358, "merge": 0.07985}, "mean_gate": {"surface": 0.13632, "split": 0.12716, "merge": 0.13344}, "n_reads": 4283, "n_cand": 25000, "read_suppressed_frac": 0.0995, "dropout_reads_logged": 4283, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 351, "admitted_surface": 60651, "staged_surface": 74994, "admitted_split": 228, "staged_split": 58308, "admitted_merge": 4657, "staged_merge": 128842, "admitted_skel": 0, "staged_skel": 0, "mean_count": 42.57585144042969, "n_util_obs": 75, "util_count_spearman": -0.002162177540435488}, "probe_overhead_sec_total": 102.009, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.002945, "alpha": 0.03, "t": 1210.8} |
| {"event": "consolidate", "step": 950, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262107, "t": 1268.4} |
| {"event": "train", "step": 950, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.69, "bpb": 2.43815, "chunk_loss": 0.03075, "F": 0.20596, "G": 0.24302, "chunks": 210.91, "bytes_per_chunk": 4.855, "boundary_entropy": 0.2413, "router_noise": 0.0, "grad_norm": 1.0869, "lr": 0.0002914, "router_trainable": true, "sec_per_step": 1.169, "probe_sec": 0.1257, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08459, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.03357, "n_shadow": 4, "mean_delta_shadow": 0.04442, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.13922, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -0.42885, "trunc_abs_frac": 0.56361, "mean_delta_full": -0.08218, "read_hit_rate": {"surface": 0.10228, "split": 0.26587, "merge": 0.13598}, "mean_gate": {"surface": 0.12752, "split": 0.12163, "merge": 0.13014}, "n_reads": 4670, "n_cand": 24964, "read_suppressed_frac": 0.0991, "dropout_reads_logged": 4670, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 387, "admitted_surface": 59626, "staged_surface": 74157, "admitted_split": 228, "staged_split": 57714, "admitted_merge": 5682, "staged_merge": 130273, "admitted_skel": 0, "staged_skel": 0, "mean_count": 46.2138671875, "n_util_obs": 75, "util_count_spearman": -0.04591749644381223}, "probe_overhead_sec_total": 107.502, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.002803, "alpha": 0.03, "t": 1269.2} |
| {"event": "consolidate", "step": 1000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262106, "t": 1336.7} |
| {"event": "probe", "step": 1000, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3473.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.199563, "cps": 27.132812, "mean_p": 0.23812, "boundary_entropy": 0.247501, "bpb": 2.350857}, "tr": {"chunks": 3990.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.220405, "cps": 31.171875, "mean_p": 0.228232, "boundary_entropy": 0.218958, "bpb": 2.655603}, "fi": {"chunks": 4372.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.232417, "cps": 34.15625, "mean_p": 0.2503, "boundary_entropy": 0.240224, "bpb": 2.864891}}, "bpb": {"en": 2.35086, "tr": 2.6556, "fi": 2.86489}, "gini_cpc": 0.033573, "gini_cps": 0.050641, "bpic": 4.793578, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.104438865430099, "fi": 1.1646290177361442}, "inventory": {"size": 326656, "admitted": 65536, "staged": 261120, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 419, "admitted_surface": 58616, "staged_surface": 73421, "admitted_split": 215, "staged_split": 57577, "admitted_merge": 6705, "staged_merge": 130122, "admitted_skel": 0, "staged_skel": 0, "mean_count": 50.396575927734375, "n_util_obs": 77, "util_count_spearman": 0.011488511488511488}, "t": 1337.9} |
| {"event": "eval", "step": 1000, "phase": 1, "bpb": {"en": 2.36047, "tr": 2.66443, "fi": 2.86909}, "bytes_per_chunk": {"en": 5.087, "tr": 5.072, "fi": 4.682}, "bpb_mean": 2.63133, "frozen_inventory": true, "t": 1340.5} |
| {"event": "train", "step": 1000, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.74728, "bpb": 2.52079, "chunk_loss": 0.030572, "F": 0.20142, "G": 0.23405, "chunks": 206.25, "bytes_per_chunk": 4.965, "boundary_entropy": 0.22959, "router_noise": 0.0, "grad_norm": 1.0702, "lr": 0.0002901, "router_trainable": true, "sec_per_step": 1.566, "probe_sec": 1.2322, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 1.16269, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.05703, "n_shadow": 5, "mean_delta_shadow": 0.01584, "n_split": 2, "mean_delta_split": 0.00092, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.03474, "trunc_abs_frac": 0.03474, "mean_delta_full": 0.28992, "read_hit_rate": {"surface": 0.10551, "split": 0.25205, "merge": 0.1452}, "mean_gate": {"surface": 0.12212, "split": 0.11756, "merge": 0.12474}, "n_reads": 4536, "n_cand": 24586, "read_suppressed_frac": 0.0966, "dropout_reads_logged": 4536, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 419, "admitted_surface": 58616, "staged_surface": 73708, "admitted_split": 215, "staged_split": 57736, "admitted_merge": 6705, "staged_merge": 130700, "admitted_skel": 0, "staged_skel": 0, "mean_count": 50.465789794921875, "n_util_obs": 77, "util_count_spearman": 0.010804984489195016}, "probe_overhead_sec_total": 114.522, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.003022, "alpha": 0.03, "t": 1347.5} |
| {"event": "probe", "step": 1024, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3404.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.195598, "cps": 26.59375, "mean_p": 0.233289, "boundary_entropy": 0.247228, "bpb": 2.341816}, "tr": {"chunks": 3943.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.217809, "cps": 30.804688, "mean_p": 0.223567, "boundary_entropy": 0.225111, "bpb": 2.643093}, "fi": {"chunks": 4223.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.224496, "cps": 32.992188, "mean_p": 0.247346, "boundary_entropy": 0.245053, "bpb": 2.850462}}, "bpb": {"en": 2.34182, "tr": 2.64309, "fi": 2.85046}, "gini_cpc": 0.030201, "gini_cps": 0.047191, "bpic": 4.903371, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.113552749110992, "fi": 1.1477406586552947}, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 443, "admitted_surface": 58616, "staged_surface": 73382, "admitted_split": 215, "staged_split": 57120, "admitted_merge": 6705, "staged_merge": 131642, "admitted_skel": 0, "staged_skel": 0, "mean_count": 52.08955383300781, "n_util_obs": 77, "util_count_spearman": -0.0024449234975550765}, "t": 1380.6} |
| {"event": "consolidate", "step": 1050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262103, "t": 1427.3} |
| {"event": "train", "step": 1050, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.64728, "bpb": 2.37652, "chunk_loss": 0.030748, "F": 0.20175, "G": 0.23866, "chunks": 206.59, "bytes_per_chunk": 4.957, "boundary_entropy": 0.23503, "router_noise": 0.0, "grad_norm": 1.0009, "lr": 0.0002886, "router_trainable": true, "sec_per_step": 1.639, "probe_sec": 0.0936, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04289, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.11607, "n_shadow": 3, "mean_delta_shadow": 0.04783, "n_split": 2, "mean_delta_split": -0.14633, "n_merge": 1, "mean_delta_merge": -0.08202, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.48942, "trunc_abs_frac": 0.9123, "mean_delta_full": 0.24294, "read_hit_rate": {"surface": 0.12054, "split": 0.28842, "merge": 0.18237}, "mean_gate": {"surface": 0.12414, "split": 0.11831, "merge": 0.12516}, "n_reads": 5230, "n_cand": 24370, "read_suppressed_frac": 0.0983, "dropout_reads_logged": 5230, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 476, "admitted_surface": 57606, "staged_surface": 73420, "admitted_split": 214, "staged_split": 57462, "admitted_merge": 7716, "staged_merge": 131262, "admitted_skel": 0, "staged_skel": 0, "mean_count": 56.29264831542969, "n_util_obs": 82, "util_count_spearman": -0.03564393073649612}, "probe_overhead_sec_total": 121.171, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.002939, "alpha": 0.03, "t": 1429.5} |
| {"event": "consolidate", "step": 1100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262105, "t": 1501.0} |
| {"event": "train", "step": 1100, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.75017, "bpb": 2.52496, "chunk_loss": 0.030602, "F": 0.20111, "G": 0.23414, "chunks": 205.94, "bytes_per_chunk": 4.972, "boundary_entropy": 0.22945, "router_noise": 0.0, "grad_norm": 1.0149, "lr": 0.0002871, "router_trainable": true, "sec_per_step": 1.456, "probe_sec": 0.0701, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04207, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.101, "n_shadow": 4, "mean_delta_shadow": 0.07381, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": -1.26639, "trunc_abs_frac": 1.26639, "mean_delta_full": 0.07446, "read_hit_rate": {"surface": 0.26685, "split": 0.43294, "merge": 0.16181}, "mean_gate": {"surface": 0.12545, "split": 0.12185, "merge": 0.12272}, "n_reads": 7687, "n_cand": 24358, "read_suppressed_frac": 0.1071, "dropout_reads_logged": 7687, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 518, "admitted_surface": 57319, "staged_surface": 72122, "admitted_split": 310, "staged_split": 56980, "admitted_merge": 7907, "staged_merge": 133042, "admitted_skel": 0, "staged_skel": 0, "mean_count": 90.52177429199219, "n_util_obs": 94, "util_count_spearman": -0.09726546978289925}, "probe_overhead_sec_total": 127.594, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.003208, "alpha": 0.03, "t": 1502.3} |
| {"event": "consolidate", "step": 1150, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 1024, "uncertain_staged": 262116, "t": 1564.2} |
| {"event": "train", "step": 1150, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.65256, "bpb": 2.38414, "chunk_loss": 0.030545, "F": 0.20166, "G": 0.23136, "chunks": 206.5, "bytes_per_chunk": 4.959, "boundary_entropy": 0.22351, "router_noise": 0.0, "grad_norm": 0.8766, "lr": 0.0002855, "router_trainable": true, "sec_per_step": 1.276, "probe_sec": 0.0949, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04761, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.07736, "n_shadow": 6, "mean_delta_shadow": 0.04191, "n_split": 1, "mean_delta_split": 0.18835, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.55296, "trunc_abs_frac": 0.66301, "mean_delta_full": 0.22896, "read_hit_rate": {"surface": 0.30368, "split": 0.45277, "merge": 0.17023}, "mean_gate": {"surface": 0.11151, "split": 0.10804, "merge": 0.11059}, "n_reads": 8355, "n_cand": 24702, "read_suppressed_frac": 0.11, "dropout_reads_logged": 8355, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 570, "admitted_surface": 57291, "staged_surface": 71459, "admitted_split": 336, "staged_split": 57434, "admitted_merge": 7909, "staged_merge": 133251, "admitted_skel": 0, "staged_skel": 0, "mean_count": 97.25643920898438, "n_util_obs": 97, "util_count_spearman": -0.06573479907426888}, "probe_overhead_sec_total": 133.284, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.003315, "alpha": 0.03, "t": 1566.1} |
| {"event": "consolidate", "step": 1200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262117, "t": 1642.8} |
| {"event": "probe", "step": 1200, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3442.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.197782, "cps": 26.890625, "mean_p": 0.231738, "boundary_entropy": 0.231633, "bpb": 2.280694}, "tr": {"chunks": 3998.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.220847, "cps": 31.234375, "mean_p": 0.219149, "boundary_entropy": 0.196714, "bpb": 2.564924}, "fi": {"chunks": 4515.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.240019, "cps": 35.273438, "mean_p": 0.252877, "boundary_entropy": 0.21821, "bpb": 2.745714}}, "bpb": {"en": 2.28069, "tr": 2.56492, "fi": 2.74571}, "gini_cpc": 0.042751, "gini_cps": 0.059835, "bpic": 4.745462, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.116620232029497, "fi": 1.213554054056976}, "inventory": {"size": 326656, "admitted": 65536, "staged": 261120, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 625, "admitted_surface": 57292, "staged_surface": 69847, "admitted_split": 335, "staged_split": 57041, "admitted_merge": 7909, "staged_merge": 134232, "admitted_skel": 0, "staged_skel": 0, "mean_count": 103.76167297363281, "n_util_obs": 102, "util_count_spearman": -0.06239185750735877}, "t": 1644.5} |
| {"event": "eval", "step": 1200, "phase": 1, "bpb": {"en": 2.29121, "tr": 2.5736, "fi": 2.75345}, "bytes_per_chunk": {"en": 5.043, "tr": 5.012, "fi": 4.517}, "bpb_mean": 2.53942, "frozen_inventory": true, "t": 1647.7} |
| {"event": "train", "step": 1200, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.62836, "bpb": 2.34923, "chunk_loss": 0.030573, "F": 0.20282, "G": 0.23102, "chunks": 207.69, "bytes_per_chunk": 4.93, "boundary_entropy": 0.21751, "router_noise": 0.0, "grad_norm": 0.9164, "lr": 0.0002837, "router_trainable": true, "sec_per_step": 1.78, "probe_sec": 0.1602, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.11834, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": -0.01847, "n_shadow": 4, "mean_delta_shadow": 0.06216, "n_split": 1, "mean_delta_split": -0.00543, "n_merge": 1, "mean_delta_merge": -0.06555, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.13611, "trunc_abs_frac": 0.81663, "mean_delta_full": -0.03707, "read_hit_rate": {"surface": 0.36922, "split": 0.46078, "merge": 0.17107}, "mean_gate": {"surface": 0.10867, "split": 0.10567, "merge": 0.10822}, "n_reads": 8879, "n_cand": 24722, "read_suppressed_frac": 0.0984, "dropout_reads_logged": 8879, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 627, "admitted_surface": 57292, "staged_surface": 70062, "admitted_split": 335, "staged_split": 57211, "admitted_merge": 7909, "staged_merge": 134871, "admitted_skel": 0, "staged_skel": 0, "mean_count": 103.89718627929688, "n_util_obs": 102, "util_count_spearman": -0.06420659198986718}, "probe_overhead_sec_total": 140.088, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.003475, "alpha": 0.03, "t": 1655.1} |
| {"event": "consolidate", "step": 1250, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262116, "t": 1726.1} |
| {"event": "train", "step": 1250, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.64483, "bpb": 2.37299, "chunk_loss": 0.030395, "F": 0.19489, "G": 0.22296, "chunks": 199.56, "bytes_per_chunk": 5.131, "boundary_entropy": 0.21446, "router_noise": 0.0, "grad_norm": 0.9718, "lr": 0.0002819, "router_trainable": true, "sec_per_step": 1.45, "probe_sec": 0.0895, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04429, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": -0.02457, "n_shadow": 5, "mean_delta_shadow": -0.04406, "n_split": 3, "mean_delta_split": -0.12345, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.04417, "trunc_abs_frac": 1.39687, "mean_delta_full": 0.30951, "read_hit_rate": {"surface": 0.3711, "split": 0.43979, "merge": 0.16593}, "mean_gate": {"surface": 0.1055, "split": 0.10302, "merge": 0.10622}, "n_reads": 8272, "n_cand": 23738, "read_suppressed_frac": 0.1002, "dropout_reads_logged": 8272, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 673, "admitted_surface": 57310, "staged_surface": 68870, "admitted_split": 336, "staged_split": 57092, "admitted_merge": 7890, "staged_merge": 136182, "admitted_skel": 0, "staged_skel": 0, "mean_count": 110.75323486328125, "n_util_obs": 106, "util_count_spearman": -0.043222409753885686}, "probe_overhead_sec_total": 146.27, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.003231, "alpha": 0.03, "t": 1727.6} |
| {"event": "consolidate", "step": 1300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262117, "t": 1794.1} |
| {"event": "train", "step": 1300, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.6805, "bpb": 2.42445, "chunk_loss": 0.030761, "F": 0.21161, "G": 0.23587, "chunks": 216.69, "bytes_per_chunk": 4.726, "boundary_entropy": 0.20418, "router_noise": 0.0, "grad_norm": 0.9704, "lr": 0.00028, "router_trainable": true, "sec_per_step": 1.355, "probe_sec": 0.1634, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10214, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.01693, "n_shadow": 5, "mean_delta_shadow": 0.03202, "n_split": 1, "mean_delta_split": 0.1012, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -1.28812, "trunc_abs_frac": 0.78717, "mean_delta_full": -0.00861, "read_hit_rate": {"surface": 0.40496, "split": 0.45258, "merge": 0.17555}, "mean_gate": {"surface": 0.10352, "split": 0.09962, "merge": 0.10485}, "n_reads": 9403, "n_cand": 25708, "read_suppressed_frac": 0.0952, "dropout_reads_logged": 9403, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 725, "admitted_surface": 57617, "staged_surface": 68040, "admitted_split": 346, "staged_split": 56725, "admitted_merge": 7573, "staged_merge": 137379, "admitted_skel": 0, "staged_skel": 0, "mean_count": 117.56037902832031, "n_util_obs": 111, "util_count_spearman": -0.041286416286416286}, "probe_overhead_sec_total": 152.833, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.003297, "alpha": 0.03, "t": 1795.4} |
| {"event": "consolidate", "step": 1350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262117, "t": 1876.4} |
| {"event": "train", "step": 1350, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.61064, "bpb": 2.32366, "chunk_loss": 0.030522, "F": 0.19934, "G": 0.22844, "chunks": 204.12, "bytes_per_chunk": 5.017, "boundary_entropy": 0.21416, "router_noise": 0.0, "grad_norm": 1.0022, "lr": 0.000278, "router_trainable": true, "sec_per_step": 1.645, "probe_sec": 0.125, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04768, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.0076, "n_shadow": 4, "mean_delta_shadow": -0.03154, "n_split": 1, "mean_delta_split": -0.06854, "n_merge": 1, "mean_delta_merge": 0.005, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.03889, "trunc_abs_frac": 0.05463, "mean_delta_full": 0.28522, "read_hit_rate": {"surface": 0.44585, "split": 0.49439, "merge": 0.18429}, "mean_gate": {"surface": 0.10524, "split": 0.10119, "merge": 0.10773}, "n_reads": 9643, "n_cand": 24200, "read_suppressed_frac": 0.1054, "dropout_reads_logged": 9643, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 780, "admitted_surface": 57068, "staged_surface": 67657, "admitted_split": 832, "staged_split": 56764, "admitted_merge": 7636, "staged_merge": 137723, "admitted_skel": 0, "staged_skel": 0, "mean_count": 125.52116394042969, "n_util_obs": 118, "util_count_spearman": -0.08211628849714593}, "probe_overhead_sec_total": 159.376, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.003869, "alpha": 0.03, "t": 1877.6} |
| {"event": "consolidate", "step": 1400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262116, "t": 1949.7} |
| {"event": "probe", "step": 1400, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3469.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.199333, "cps": 27.101562, "mean_p": 0.226914, "boundary_entropy": 0.212756, "bpb": 2.224149}, "tr": {"chunks": 3804.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.210131, "cps": 29.71875, "mean_p": 0.213563, "boundary_entropy": 0.194858, "bpb": 2.493935}, "fi": {"chunks": 4325.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.229919, "cps": 33.789062, "mean_p": 0.243836, "boundary_entropy": 0.206761, "bpb": 2.655635}}, "bpb": {"en": 2.22415, "tr": 2.49393, "fi": 2.65564}, "gini_cpc": 0.03189, "gini_cps": 0.049204, "bpic": 4.891533, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0541678748142873, "fi": 1.1534374517807802}, "inventory": {"size": 326656, "admitted": 65536, "staged": 261120, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 838, "admitted_surface": 56072, "staged_surface": 67425, "admitted_split": 1755, "staged_split": 55586, "admitted_merge": 7709, "staged_merge": 138109, "admitted_skel": 0, "staged_skel": 0, "mean_count": 132.8387908935547, "n_util_obs": 121, "util_count_spearman": -0.10621546236206528}, "t": 1951.5} |
| {"event": "eval", "step": 1400, "phase": 1, "bpb": {"en": 2.2337, "tr": 2.50054, "fi": 2.66058}, "bytes_per_chunk": {"en": 5.089, "tr": 5.292, "fi": 4.733}, "bpb_mean": 2.46494, "frozen_inventory": true, "t": 1954.9} |
| {"event": "train", "step": 1400, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.60506, "bpb": 2.31562, "chunk_loss": 0.030464, "F": 0.19714, "G": 0.22122, "chunks": 201.88, "bytes_per_chunk": 5.072, "boundary_entropy": 0.19756, "router_noise": 0.0, "grad_norm": 0.8976, "lr": 0.0002759, "router_trainable": true, "sec_per_step": 1.674, "probe_sec": 0.1048, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04182, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": -0.00155, "n_shadow": 3, "mean_delta_shadow": 0.09922, "n_split": 2, "mean_delta_split": 0.02765, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -0.10029, "trunc_abs_frac": 0.61461, "mean_delta_full": -0.11401, "read_hit_rate": {"surface": 0.43871, "split": 0.50153, "merge": 0.18293}, "mean_gate": {"surface": 0.10188, "split": 0.09674, "merge": 0.10288}, "n_reads": 9574, "n_cand": 23958, "read_suppressed_frac": 0.1028, "dropout_reads_logged": 9574, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 840, "admitted_surface": 56072, "staged_surface": 67699, "admitted_split": 1755, "staged_split": 55861, "admitted_merge": 7709, "staged_merge": 138584, "admitted_skel": 0, "staged_skel": 0, "mean_count": 132.98487854003906, "n_util_obs": 121, "util_count_spearman": -0.1066490086863213}, "probe_overhead_sec_total": 165.306, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.003917, "alpha": 0.03, "t": 1961.3} |
| {"event": "consolidate", "step": 1450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262116, "t": 2032.4} |
| {"event": "train", "step": 1450, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.65731, "bpb": 2.391, "chunk_loss": 0.03045, "F": 0.19919, "G": 0.22085, "chunks": 203.97, "bytes_per_chunk": 5.02, "boundary_entropy": 0.19089, "router_noise": 0.0, "grad_norm": 0.8848, "lr": 0.0002738, "router_trainable": true, "sec_per_step": 1.447, "probe_sec": 0.0882, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04717, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": -0.06998, "n_shadow": 7, "mean_delta_shadow": 0.04168, "n_split": 1, "mean_delta_split": -0.06465, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -0.39639, "trunc_abs_frac": 0.37601, "mean_delta_full": -0.06445, "read_hit_rate": {"surface": 0.44742, "split": 0.51264, "merge": 0.1654}, "mean_gate": {"surface": 0.10242, "split": 0.09647, "merge": 0.10205}, "n_reads": 9733, "n_cand": 24190, "read_suppressed_frac": 0.1011, "dropout_reads_logged": 9733, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 878, "admitted_surface": 55567, "staged_surface": 67492, "admitted_split": 2579, "staged_split": 55532, "admitted_merge": 7390, "staged_merge": 139120, "admitted_skel": 0, "staged_skel": 0, "mean_count": 138.77272033691406, "n_util_obs": 115, "util_count_spearman": -0.16542255188195376}, "probe_overhead_sec_total": 171.138, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.00378, "alpha": 0.03, "t": 2033.7} |
| {"event": "consolidate", "step": 1500, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262115, "t": 2104.6} |
| {"event": "train", "step": 1500, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.67209, "bpb": 2.41231, "chunk_loss": 0.030666, "F": 0.20712, "G": 0.23055, "chunks": 212.09, "bytes_per_chunk": 4.828, "boundary_entropy": 0.19149, "router_noise": 0.0, "grad_norm": 0.8664, "lr": 0.0002715, "router_trainable": true, "sec_per_step": 1.429, "probe_sec": 0.1175, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08913, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.00512, "n_shadow": 3, "mean_delta_shadow": -0.02997, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": -0.05185, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -21.5, "trunc_abs_frac": 2.1958, "mean_delta_full": -0.00301, "read_hit_rate": {"surface": 0.4487, "split": 0.50087, "merge": 0.17775}, "mean_gate": {"surface": 0.10124, "split": 0.09451, "merge": 0.10139}, "n_reads": 9978, "n_cand": 24962, "read_suppressed_frac": 0.1027, "dropout_reads_logged": 9978, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 905, "admitted_surface": 56296, "staged_surface": 67131, "admitted_split": 2816, "staged_split": 55596, "admitted_merge": 6424, "staged_merge": 139417, "admitted_skel": 0, "staged_skel": 0, "mean_count": 141.05361938476562, "n_util_obs": 107, "util_count_spearman": -0.1811947726248555}, "probe_overhead_sec_total": 176.941, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.003677, "alpha": 0.03, "t": 2105.1} |
| {"event": "consolidate", "step": 1550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262113, "t": 2180.9} |
| {"event": "train", "step": 1550, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.58881, "bpb": 2.29217, "chunk_loss": 0.030598, "F": 0.20364, "G": 0.22575, "chunks": 208.53, "bytes_per_chunk": 4.911, "boundary_entropy": 0.19217, "router_noise": 0.0, "grad_norm": 0.9099, "lr": 0.0002692, "router_trainable": true, "sec_per_step": 1.538, "probe_sec": 0.0894, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0441, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.00012, "n_shadow": 4, "mean_delta_shadow": 0.08568, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.08052, "trunc_abs_frac": 0.45424, "mean_delta_full": -0.28805, "read_hit_rate": {"surface": 0.45084, "split": 0.50916, "merge": 0.1787}, "mean_gate": {"surface": 0.10252, "split": 0.0949, "merge": 0.10362}, "n_reads": 10010, "n_cand": 24710, "read_suppressed_frac": 0.1002, "dropout_reads_logged": 10010, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 946, "admitted_surface": 55277, "staged_surface": 67166, "admitted_split": 3832, "staged_split": 54512, "admitted_merge": 6427, "staged_merge": 140466, "admitted_skel": 0, "staged_skel": 0, "mean_count": 147.95985412597656, "n_util_obs": 108, "util_count_spearman": -0.2035496870444997}, "probe_overhead_sec_total": 183.156, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.00348, "alpha": 0.03, "t": 2182.0} |
| {"event": "consolidate", "step": 1600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262113, "t": 2240.1} |
| {"event": "probe", "step": 1600, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3377.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.194047, "cps": 26.382812, "mean_p": 0.220536, "boundary_entropy": 0.205299, "bpb": 2.186315}, "tr": {"chunks": 3855.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.212948, "cps": 30.117188, "mean_p": 0.211783, "boundary_entropy": 0.177418, "bpb": 2.424854}, "fi": {"chunks": 4425.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.235235, "cps": 34.570312, "mean_p": 0.24512, "boundary_entropy": 0.194502, "bpb": 2.60447}}, "bpb": {"en": 2.18632, "tr": 2.42485, "fi": 2.60447}, "gini_cpc": 0.042755, "gini_cps": 0.059935, "bpic": 4.866775, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0974048886286873, "fi": 1.2122563038306946}, "inventory": {"size": 326656, "admitted": 65536, "staged": 261120, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 990, "admitted_surface": 54265, "staged_surface": 67254, "admitted_split": 4842, "staged_split": 54405, "admitted_merge": 6429, "staged_merge": 139461, "admitted_skel": 0, "staged_skel": 0, "mean_count": 154.438720703125, "n_util_obs": 112, "util_count_spearman": -0.19449379847609935}, "t": 2241.9} |
| {"event": "eval", "step": 1600, "phase": 1, "bpb": {"en": 2.1971, "tr": 2.43261, "fi": 2.60939}, "bytes_per_chunk": {"en": 5.163, "tr": 5.179, "fi": 4.639}, "bpb_mean": 2.41303, "frozen_inventory": true, "t": 2246.0} |
| {"event": "train", "step": 1600, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.52351, "bpb": 2.19796, "chunk_loss": 0.030398, "F": 0.19711, "G": 0.22124, "chunks": 201.84, "bytes_per_chunk": 5.073, "boundary_entropy": 0.19584, "router_noise": 0.0, "grad_norm": 0.9847, "lr": 0.0002668, "router_trainable": true, "sec_per_step": 1.42, "probe_sec": 0.1728, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.1223, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": -0.0373, "n_shadow": 7, "mean_delta_shadow": 0.04372, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -1.65855, "trunc_abs_frac": 1.99805, "mean_delta_full": -0.03092, "read_hit_rate": {"surface": 0.45589, "split": 0.50909, "merge": 0.16544}, "mean_gate": {"surface": 0.10368, "split": 0.09444, "merge": 0.1058}, "n_reads": 9813, "n_cand": 24264, "read_suppressed_frac": 0.0958, "dropout_reads_logged": 9813, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 990, "admitted_surface": 54265, "staged_surface": 67471, "admitted_split": 4842, "staged_split": 54507, "admitted_merge": 6429, "staged_merge": 140166, "admitted_skel": 0, "staged_skel": 0, "mean_count": 154.5884552001953, "n_util_obs": 112, "util_count_spearman": -0.1937591826972358}, "probe_overhead_sec_total": 188.925, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.004009, "alpha": 0.03, "t": 2253.0} |
| {"event": "consolidate", "step": 1650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262109, "t": 2322.4} |
| {"event": "train", "step": 1650, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.58754, "bpb": 2.29033, "chunk_loss": 0.030503, "F": 0.20374, "G": 0.22364, "chunks": 208.62, "bytes_per_chunk": 4.908, "boundary_entropy": 0.18569, "router_noise": 0.0, "grad_norm": 0.9258, "lr": 0.0002643, "router_trainable": true, "sec_per_step": 1.418, "probe_sec": 0.1067, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04982, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": -0.00164, "n_shadow": 3, "mean_delta_shadow": 0.12271, "n_split": 1, "mean_delta_split": 0.0235, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -0.30502, "trunc_abs_frac": 12.69378, "mean_delta_full": 0.05026, "read_hit_rate": {"surface": 0.44567, "split": 0.51826, "merge": 0.16833}, "mean_gate": {"surface": 0.10368, "split": 0.09344, "merge": 0.10482}, "n_reads": 10035, "n_cand": 24758, "read_suppressed_frac": 0.102, "dropout_reads_logged": 10035, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1034, "admitted_surface": 53256, "staged_surface": 67210, "admitted_split": 5851, "staged_split": 53748, "admitted_merge": 6429, "staged_merge": 141186, "admitted_skel": 0, "staged_skel": 0, "mean_count": 161.55714416503906, "n_util_obs": 114, "util_count_spearman": -0.21961078147465624}, "probe_overhead_sec_total": 194.821, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.004154, "alpha": 0.03, "t": 2323.9} |
| {"event": "consolidate", "step": 1700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262103, "t": 2377.0} |
| {"event": "train", "step": 1700, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.54246, "bpb": 2.2253, "chunk_loss": 0.03043, "F": 0.19928, "G": 0.21911, "chunks": 204.06, "bytes_per_chunk": 5.018, "boundary_entropy": 0.17808, "router_noise": 0.0, "grad_norm": 0.8226, "lr": 0.0002618, "router_trainable": true, "sec_per_step": 1.086, "probe_sec": 0.1722, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.13461, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 3, "mean_delta_loo": 0.10742, "n_shadow": 7, "mean_delta_shadow": 0.06614, "n_split": 1, "mean_delta_split": -0.16583, "n_merge": 1, "mean_delta_merge": -0.05194, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.76644, "trunc_abs_frac": 0.86918, "mean_delta_full": -0.05475, "read_hit_rate": {"surface": 0.4506, "split": 0.5086, "merge": 0.16564}, "mean_gate": {"surface": 0.1057, "split": 0.09415, "merge": 0.10669}, "n_reads": 9736, "n_cand": 24244, "read_suppressed_frac": 0.1051, "dropout_reads_logged": 9736, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1087, "admitted_surface": 52277, "staged_surface": 67345, "admitted_split": 6830, "staged_split": 54006, "admitted_merge": 6429, "staged_merge": 140793, "admitted_skel": 0, "staged_skel": 0, "mean_count": 169.47711181640625, "n_util_obs": 120, "util_count_spearman": -0.20213209250642405}, "probe_overhead_sec_total": 199.629, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.004123, "alpha": 0.03, "t": 2378.2} |
| {"event": "consolidate", "step": 1750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262104, "t": 2445.6} |
| {"event": "train", "step": 1750, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.57365, "bpb": 2.27029, "chunk_loss": 0.030587, "F": 0.20569, "G": 0.2256, "chunks": 210.62, "bytes_per_chunk": 4.862, "boundary_entropy": 0.17733, "router_noise": 0.0, "grad_norm": 0.8574, "lr": 0.0002591, "router_trainable": true, "sec_per_step": 1.366, "probe_sec": 0.0932, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05751, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.03266, "n_shadow": 5, "mean_delta_shadow": 0.00469, "n_split": 1, "mean_delta_split": -0.00745, "n_merge": 1, "mean_delta_merge": -0.05667, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -0.02549, "trunc_abs_frac": 0.61502, "mean_delta_full": -0.07922, "read_hit_rate": {"surface": 0.476, "split": 0.54868, "merge": 0.17166}, "mean_gate": {"surface": 0.10736, "split": 0.09333, "merge": 0.10888}, "n_reads": 10752, "n_cand": 25072, "read_suppressed_frac": 0.1, "dropout_reads_logged": 10752, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1122, "admitted_surface": 51352, "staged_surface": 67322, "admitted_split": 7755, "staged_split": 53269, "admitted_merge": 6429, "staged_merge": 141553, "admitted_skel": 0, "staged_skel": 0, "mean_count": 178.5097198486328, "n_util_obs": 129, "util_count_spearman": -0.1616591057114592}, "probe_overhead_sec_total": 205.476, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.004287, "alpha": 0.03, "t": 2446.6} |
| {"event": "consolidate", "step": 1800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262107, "t": 2512.8} |
| {"event": "probe", "step": 1800, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3437.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.197495, "cps": 26.851562, "mean_p": 0.220546, "boundary_entropy": 0.189643, "bpb": 2.143493}, "tr": {"chunks": 3855.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.212948, "cps": 30.117188, "mean_p": 0.209827, "boundary_entropy": 0.165977, "bpb": 2.365823}, "fi": {"chunks": 4421.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.235022, "cps": 34.539062, "mean_p": 0.240451, "boundary_entropy": 0.177118, "bpb": 2.547581}}, "bpb": {"en": 2.14349, "tr": 2.36582, "fi": 2.54758}, "gini_cpc": 0.03876, "gini_cps": 0.056006, "bpic": 4.843507, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0782473985740695, "fi": 1.1900171478975576}, "inventory": {"size": 326656, "admitted": 65536, "staged": 261120, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1174, "admitted_surface": 50330, "staged_surface": 67328, "admitted_split": 8777, "staged_split": 52313, "admitted_merge": 6429, "staged_merge": 141479, "admitted_skel": 0, "staged_skel": 0, "mean_count": 186.31724548339844, "n_util_obs": 131, "util_count_spearman": -0.13050232210537555}, "t": 2514.0} |
| {"event": "eval", "step": 1800, "phase": 1, "bpb": {"en": 2.15685, "tr": 2.37282, "fi": 2.55217}, "bytes_per_chunk": {"en": 5.091, "tr": 5.226, "fi": 4.625}, "bpb_mean": 2.36061, "frozen_inventory": true, "t": 2516.4} |
| {"event": "train", "step": 1800, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.55027, "bpb": 2.23657, "chunk_loss": 0.030329, "F": 0.19183, "G": 0.21182, "chunks": 196.44, "bytes_per_chunk": 5.213, "boundary_entropy": 0.17301, "router_noise": 0.0, "grad_norm": 0.7902, "lr": 0.0002564, "router_trainable": true, "sec_per_step": 1.477, "probe_sec": 0.0972, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08073, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.08558, "n_shadow": 2, "mean_delta_shadow": -0.02112, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": 0.15994, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.25606, "trunc_abs_frac": 0.7528, "mean_delta_full": -0.20158, "read_hit_rate": {"surface": 0.4641, "split": 0.54206, "merge": 0.17186}, "mean_gate": {"surface": 0.11144, "split": 0.09505, "merge": 0.11058}, "n_reads": 9861, "n_cand": 23339, "read_suppressed_frac": 0.0963, "dropout_reads_logged": 9861, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1176, "admitted_surface": 50330, "staged_surface": 67622, "admitted_split": 8777, "staged_split": 52564, "admitted_merge": 6429, "staged_merge": 141958, "admitted_skel": 0, "staged_skel": 0, "mean_count": 186.4677276611328, "n_util_obs": 131, "util_count_spearman": -0.13303279377686775}, "probe_overhead_sec_total": 211.172, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.004144, "alpha": 0.03, "t": 2520.4} |
| {"event": "consolidate", "step": 1850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262105, "t": 2585.7} |
| {"event": "train", "step": 1850, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.55932, "bpb": 2.24962, "chunk_loss": 0.030554, "F": 0.20462, "G": 0.22238, "chunks": 209.53, "bytes_per_chunk": 4.887, "boundary_entropy": 0.17427, "router_noise": 0.0, "grad_norm": 0.8094, "lr": 0.0002537, "router_trainable": true, "sec_per_step": 1.342, "probe_sec": 0.0864, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04093, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.05348, "n_shadow": 6, "mean_delta_shadow": 0.02205, "n_split": 1, "mean_delta_split": -0.38818, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.74954, "trunc_abs_frac": 4.05854, "mean_delta_full": -0.24524, "read_hit_rate": {"surface": 0.47865, "split": 0.56529, "merge": 0.16729}, "mean_gate": {"surface": 0.11172, "split": 0.09372, "merge": 0.11293}, "n_reads": 10781, "n_cand": 24770, "read_suppressed_frac": 0.0971, "dropout_reads_logged": 10781, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1229, "admitted_surface": 49308, "staged_surface": 67292, "admitted_split": 9799, "staged_split": 52395, "admitted_merge": 6429, "staged_merge": 142457, "admitted_skel": 0, "staged_skel": 0, "mean_count": 194.50439453125, "n_util_obs": 132, "util_count_spearman": -0.15540064595910316}, "probe_overhead_sec_total": 217.349, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.004175, "alpha": 0.03, "t": 2587.5} |
| {"event": "consolidate", "step": 1900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262103, "t": 2655.6} |
| {"event": "train", "step": 1900, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.54279, "bpb": 2.22578, "chunk_loss": 0.030668, "F": 0.20276, "G": 0.22176, "chunks": 207.62, "bytes_per_chunk": 4.932, "boundary_entropy": 0.17228, "router_noise": 0.0, "grad_norm": 0.9251, "lr": 0.0002508, "router_trainable": true, "sec_per_step": 1.384, "probe_sec": 0.0795, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04996, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 3, "mean_delta_loo": 0.03052, "n_shadow": 7, "mean_delta_shadow": 0.01653, "n_split": 2, "mean_delta_split": -0.09473, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.17781, "trunc_abs_frac": 0.05607, "mean_delta_full": 0.07123, "read_hit_rate": {"surface": 0.47837, "split": 0.56771, "merge": 0.16702}, "mean_gate": {"surface": 0.11225, "split": 0.0933, "merge": 0.11024}, "n_reads": 10693, "n_cand": 24520, "read_suppressed_frac": 0.0986, "dropout_reads_logged": 10693, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1261, "admitted_surface": 48285, "staged_surface": 67645, "admitted_split": 10820, "staged_split": 52048, "admitted_merge": 6431, "staged_merge": 142451, "admitted_skel": 0, "staged_skel": 0, "mean_count": 202.2084503173828, "n_util_obs": 133, "util_count_spearman": -0.1420052845818753}, "probe_overhead_sec_total": 223.424, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.004109, "alpha": 0.03, "t": 2656.7} |
| {"event": "consolidate", "step": 1950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262101, "t": 2724.0} |
| {"event": "train", "step": 1950, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.48964, "bpb": 2.14909, "chunk_loss": 0.03043, "F": 0.19635, "G": 0.21567, "chunks": 201.06, "bytes_per_chunk": 5.093, "boundary_entropy": 0.17079, "router_noise": 0.0, "grad_norm": 0.7555, "lr": 0.0002479, "router_trainable": true, "sec_per_step": 1.359, "probe_sec": 0.0903, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05623, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.03093, "n_shadow": 6, "mean_delta_shadow": -0.05357, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.55085, "trunc_abs_frac": 1.77426, "mean_delta_full": 0.09363, "read_hit_rate": {"surface": 0.46454, "split": 0.56, "merge": 0.18289}, "mean_gate": {"surface": 0.11426, "split": 0.0924, "merge": 0.11543}, "n_reads": 10290, "n_cand": 23756, "read_suppressed_frac": 0.1001, "dropout_reads_logged": 10290, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1303, "admitted_surface": 47262, "staged_surface": 67824, "admitted_split": 11843, "staged_split": 51868, "admitted_merge": 6431, "staged_merge": 142452, "admitted_skel": 0, "staged_skel": 0, "mean_count": 210.36569213867188, "n_util_obs": 139, "util_count_spearman": -0.15046695562804416}, "probe_overhead_sec_total": 229.707, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.004066, "alpha": 0.03, "t": 2724.7} |
| {"event": "consolidate", "step": 2000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262098, "t": 2796.0} |
| {"event": "probe", "step": 2000, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3462.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.198931, "cps": 27.046875, "mean_p": 0.218903, "boundary_entropy": 0.176746, "bpb": 2.115423}, "tr": {"chunks": 3796.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.209689, "cps": 29.65625, "mean_p": 0.207896, "boundary_entropy": 0.158967, "bpb": 2.310461}, "fi": {"chunks": 4454.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.236776, "cps": 34.796875, "mean_p": 0.242594, "boundary_entropy": 0.173426, "bpb": 2.491777}}, "bpb": {"en": 2.11542, "tr": 2.31046, "fi": 2.49178}, "gini_cpc": 0.039092, "gini_cps": 0.056466, "bpic": 4.843921, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0540779025776916, "fi": 1.1902423146160563}, "inventory": {"size": 326656, "admitted": 65536, "staged": 261120, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1353, "admitted_surface": 46240, "staged_surface": 67397, "admitted_split": 12865, "staged_split": 51461, "admitted_merge": 6431, "staged_merge": 142262, "admitted_skel": 0, "staged_skel": 0, "mean_count": 218.34925842285156, "n_util_obs": 147, "util_count_spearman": -0.1135067133110186}, "t": 2797.3} |
| {"event": "eval", "step": 2000, "phase": 1, "bpb": {"en": 2.12719, "tr": 2.32131, "fi": 2.4996}, "bytes_per_chunk": {"en": 5.064, "tr": 5.257, "fi": 4.618}, "bpb_mean": 2.31604, "frozen_inventory": true, "t": 2799.5} |
| {"event": "train", "step": 2000, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.65015, "bpb": 2.38067, "chunk_loss": 0.030709, "F": 0.21088, "G": 0.22735, "chunks": 215.94, "bytes_per_chunk": 4.742, "boundary_entropy": 0.1676, "router_noise": 0.0, "grad_norm": 0.8498, "lr": 0.000245, "router_trainable": true, "sec_per_step": 1.605, "probe_sec": 0.0891, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04866, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": 0.13238, "n_shadow": 4, "mean_delta_shadow": 0.01308, "n_split": 1, "mean_delta_split": -0.09802, "n_merge": 3, "mean_delta_merge": -0.03695, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.01215, "trunc_abs_frac": 1.22164, "mean_delta_full": 0.0837, "read_hit_rate": {"surface": 0.47397, "split": 0.56678, "merge": 0.15513}, "mean_gate": {"surface": 0.11421, "split": 0.08949, "merge": 0.11517}, "n_reads": 10934, "n_cand": 25390, "read_suppressed_frac": 0.1006, "dropout_reads_logged": 10934, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1356, "admitted_surface": 46240, "staged_surface": 67676, "admitted_split": 12865, "staged_split": 51587, "admitted_merge": 6431, "staged_merge": 142881, "admitted_skel": 0, "staged_skel": 0, "mean_count": 218.51611328125, "n_util_obs": 147, "util_count_spearman": -0.11172732699151486}, "probe_overhead_sec_total": 236.137, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.00414, "alpha": 0.03, "t": 2804.9} |
| {"event": "probe", "step": 2048, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3476.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.199736, "cps": 27.15625, "mean_p": 0.216879, "boundary_entropy": 0.17403, "bpb": 2.108603}, "tr": {"chunks": 3865.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.213501, "cps": 30.195312, "mean_p": 0.210692, "boundary_entropy": 0.157643, "bpb": 2.304642}, "fi": {"chunks": 4474.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.23784, "cps": 34.953125, "mean_p": 0.240454, "boundary_entropy": 0.162375, "bpb": 2.482725}}, "bpb": {"en": 2.1086, "tr": 2.30464, "fi": 2.48273}, "gini_cpc": 0.039016, "gini_cps": 0.056313, "bpic": 4.801693, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0689153143433747, "fi": 1.1907715468246642}, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1406, "admitted_surface": 46240, "staged_surface": 67222, "admitted_split": 12865, "staged_split": 52344, "admitted_merge": 6431, "staged_merge": 142578, "admitted_skel": 0, "staged_skel": 0, "mean_count": 226.11659240722656, "n_util_obs": 149, "util_count_spearman": -0.10078360239434064}, "t": 2855.9} |
| {"event": "consolidate", "step": 2050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262095, "t": 2871.8} |
| {"event": "train", "step": 2050, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.48507, "bpb": 2.1425, "chunk_loss": 0.030423, "F": 0.19312, "G": 0.20858, "chunks": 197.75, "bytes_per_chunk": 5.178, "boundary_entropy": 0.16344, "router_noise": 0.0, "grad_norm": 0.8465, "lr": 0.0002419, "router_trainable": true, "sec_per_step": 1.36, "probe_sec": 0.0811, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04436, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.01111, "n_shadow": 4, "mean_delta_shadow": -0.09058, "n_split": 1, "mean_delta_split": -0.01068, "n_merge": 1, "mean_delta_merge": -0.23889, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -0.66822, "trunc_abs_frac": 1.54481, "mean_delta_full": 0.14441, "read_hit_rate": {"surface": 0.48158, "split": 0.55984, "merge": 0.16334}, "mean_gate": {"surface": 0.11551, "split": 0.08909, "merge": 0.11544}, "n_reads": 10211, "n_cand": 23555, "read_suppressed_frac": 0.0966, "dropout_reads_logged": 10211, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1411, "admitted_surface": 45216, "staged_surface": 67440, "admitted_split": 13889, "staged_split": 51630, "admitted_merge": 6431, "staged_merge": 143074, "admitted_skel": 0, "staged_skel": 0, "mean_count": 226.74951171875, "n_util_obs": 150, "util_count_spearman": -0.11146628739055069}, "probe_overhead_sec_total": 241.496, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.004084, "alpha": 0.03, "t": 2872.9} |
| {"event": "consolidate", "step": 2100, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262093, "t": 2940.4} |
| {"event": "train", "step": 2100, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.55034, "bpb": 2.23667, "chunk_loss": 0.030354, "F": 0.19434, "G": 0.21159, "chunks": 199.0, "bytes_per_chunk": 5.146, "boundary_entropy": 0.16519, "router_noise": 0.0, "grad_norm": 0.7532, "lr": 0.0002388, "router_trainable": true, "sec_per_step": 1.368, "probe_sec": 0.0765, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04658, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.0482, "n_shadow": 6, "mean_delta_shadow": 0.00714, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.39414, "trunc_abs_frac": 0.56716, "mean_delta_full": 0.08744, "read_hit_rate": {"surface": 0.46275, "split": 0.563, "merge": 0.16529}, "mean_gate": {"surface": 0.11503, "split": 0.08615, "merge": 0.11379}, "n_reads": 10158, "n_cand": 23624, "read_suppressed_frac": 0.0988, "dropout_reads_logged": 10158, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1451, "admitted_surface": 44192, "staged_surface": 67446, "admitted_split": 14913, "staged_split": 51396, "admitted_merge": 6431, "staged_merge": 143302, "admitted_skel": 0, "staged_skel": 0, "mean_count": 235.01805114746094, "n_util_obs": 154, "util_count_spearman": -0.09212482647017751}, "probe_overhead_sec_total": 247.596, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.004064, "alpha": 0.03, "t": 2941.4} |
| {"event": "consolidate", "step": 2150, "retired": 1, "split_retired": 0, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262093, "t": 3009.7} |
| {"event": "train", "step": 2150, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.50503, "bpb": 2.17131, "chunk_loss": 0.030394, "F": 0.19846, "G": 0.21454, "chunks": 203.22, "bytes_per_chunk": 5.039, "boundary_entropy": 0.16566, "router_noise": 0.0, "grad_norm": 0.7161, "lr": 0.0002357, "router_trainable": true, "sec_per_step": 1.392, "probe_sec": 0.0685, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04304, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": -0.05635, "n_shadow": 3, "mean_delta_shadow": 0.0021, "n_split": 3, "mean_delta_split": 0.0051, "n_merge": 1, "mean_delta_merge": 0.13263, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -30.40789, "trunc_abs_frac": 1.30599, "mean_delta_full": 0.00232, "read_hit_rate": {"surface": 0.46608, "split": 0.57071, "merge": 0.17378}, "mean_gate": {"surface": 0.11797, "split": 0.08681, "merge": 0.11681}, "n_reads": 10527, "n_cand": 24110, "read_suppressed_frac": 0.0968, "dropout_reads_logged": 10527, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1486, "admitted_surface": 43577, "staged_surface": 67381, "admitted_split": 15936, "staged_split": 50975, "admitted_merge": 6023, "staged_merge": 143788, "admitted_skel": 0, "staged_skel": 0, "mean_count": 243.06094360351562, "n_util_obs": 155, "util_count_spearman": -0.11664464567690375}, "probe_overhead_sec_total": 253.331, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.00412, "alpha": 0.03, "t": 3011.0} |
| {"event": "consolidate", "step": 2200, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262095, "t": 3081.7} |
| {"event": "probe", "step": 2200, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3436.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.197437, "cps": 26.84375, "mean_p": 0.215697, "boundary_entropy": 0.168045, "bpb": 2.092283}, "tr": {"chunks": 3900.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.215434, "cps": 30.46875, "mean_p": 0.209123, "boundary_entropy": 0.151563, "bpb": 2.288006}, "fi": {"chunks": 4390.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.233374, "cps": 34.296875, "mean_p": 0.238821, "boundary_entropy": 0.16173, "bpb": 2.447667}}, "bpb": {"en": 2.09228, "tr": 2.28801, "fi": 2.44767}, "gini_cpc": 0.037072, "gini_cps": 0.054238, "bpic": 4.838137, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.091151416126978, "fi": 1.1820166710719442}, "inventory": {"size": 326655, "admitted": 65536, "staged": 261119, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1527, "admitted_surface": 42553, "staged_surface": 67260, "admitted_split": 16960, "staged_split": 50445, "admitted_merge": 6023, "staged_merge": 143414, "admitted_skel": 0, "staged_skel": 0, "mean_count": 250.25640869140625, "n_util_obs": 157, "util_count_spearman": -0.08767419792978126}, "t": 3083.6} |
| {"event": "eval", "step": 2200, "phase": 1, "bpb": {"en": 2.10176, "tr": 2.29512, "fi": 2.45418}, "bytes_per_chunk": {"en": 5.158, "tr": 5.187, "fi": 4.673}, "bpb_mean": 2.28369, "frozen_inventory": true, "t": 3086.6} |
| {"event": "train", "step": 2200, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.48538, "bpb": 2.14296, "chunk_loss": 0.030353, "F": 0.19525, "G": 0.21217, "chunks": 199.94, "bytes_per_chunk": 5.122, "boundary_entropy": 0.1584, "router_noise": 0.0, "grad_norm": 0.6751, "lr": 0.0002325, "router_trainable": true, "sec_per_step": 1.657, "probe_sec": 0.1324, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05808, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": -0.01594, "n_shadow": 4, "mean_delta_shadow": 0.07194, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": -0.00889, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.31839, "trunc_abs_frac": 1.36253, "mean_delta_full": -0.13853, "read_hit_rate": {"surface": 0.48413, "split": 0.56819, "merge": 0.16625}, "mean_gate": {"surface": 0.12259, "split": 0.08794, "merge": 0.12122}, "n_reads": 10451, "n_cand": 23816, "read_suppressed_frac": 0.1035, "dropout_reads_logged": 10451, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1529, "admitted_surface": 42553, "staged_surface": 67437, "admitted_split": 16960, "staged_split": 50664, "admitted_merge": 6023, "staged_merge": 144043, "admitted_skel": 0, "staged_skel": 0, "mean_count": 250.4158935546875, "n_util_obs": 157, "util_count_spearman": -0.08701678874217776}, "probe_overhead_sec_total": 259.668, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.004087, "alpha": 0.03, "t": 3093.8} |
| {"event": "consolidate", "step": 2250, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262095, "t": 3155.8} |
| {"event": "train", "step": 2250, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.50519, "bpb": 2.17153, "chunk_loss": 0.030422, "F": 0.20029, "G": 0.2149, "chunks": 205.09, "bytes_per_chunk": 4.993, "boundary_entropy": 0.15366, "router_noise": 0.0, "grad_norm": 0.7639, "lr": 0.0002293, "router_trainable": true, "sec_per_step": 1.271, "probe_sec": 0.14, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09044, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": -0.02362, "n_shadow": 3, "mean_delta_shadow": -0.041, "n_split": 1, "mean_delta_split": 0.28415, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 7, "trunc_ratio_sum": -0.35335, "trunc_abs_frac": 8.01031, "mean_delta_full": 0.08552, "read_hit_rate": {"surface": 0.47726, "split": 0.56095, "merge": 0.16464}, "mean_gate": {"surface": 0.12335, "split": 0.08536, "merge": 0.12081}, "n_reads": 10528, "n_cand": 24334, "read_suppressed_frac": 0.1019, "dropout_reads_logged": 10528, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1555, "admitted_surface": 41532, "staged_surface": 67301, "admitted_split": 17984, "staged_split": 50397, "admitted_merge": 6020, "staged_merge": 144446, "admitted_skel": 0, "staged_skel": 0, "mean_count": 258.2886962890625, "n_util_obs": 159, "util_count_spearman": -0.07356699307379985}, "probe_overhead_sec_total": 265.61, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.004331, "alpha": 0.03, "t": 3157.3} |
| {"event": "consolidate", "step": 2300, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262095, "t": 3224.8} |
| {"event": "train", "step": 2300, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.52092, "bpb": 2.19422, "chunk_loss": 0.030321, "F": 0.19336, "G": 0.21138, "chunks": 198.0, "bytes_per_chunk": 5.172, "boundary_entropy": 0.15844, "router_noise": 0.0, "grad_norm": 0.763, "lr": 0.000226, "router_trainable": true, "sec_per_step": 1.379, "probe_sec": 0.1803, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09408, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.03072, "n_shadow": 4, "mean_delta_shadow": 0.03427, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.12982, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.36336, "trunc_abs_frac": 0.32691, "mean_delta_full": -0.11469, "read_hit_rate": {"surface": 0.45431, "split": 0.5623, "merge": 0.15386}, "mean_gate": {"surface": 0.12279, "split": 0.08349, "merge": 0.11978}, "n_reads": 10039, "n_cand": 23620, "read_suppressed_frac": 0.0962, "dropout_reads_logged": 10039, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1586, "admitted_surface": 40611, "staged_surface": 67735, "admitted_split": 19009, "staged_split": 49982, "admitted_merge": 5916, "staged_merge": 144427, "admitted_skel": 0, "staged_skel": 0, "mean_count": 266.12152099609375, "n_util_obs": 167, "util_count_spearman": -0.06314994795263174}, "probe_overhead_sec_total": 272.275, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.004378, "alpha": 0.03, "t": 3226.3} |
| {"event": "consolidate", "step": 2350, "retired": 0, "split_retired": 3, "merge_admitted": 0, "admitted": 3, "swapped": 1024, "uncertain_staged": 262093, "t": 3293.8} |
| {"event": "train", "step": 2350, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.5691, "bpb": 2.26373, "chunk_loss": 0.030488, "F": 0.20474, "G": 0.21937, "chunks": 209.66, "bytes_per_chunk": 4.884, "boundary_entropy": 0.15584, "router_noise": 0.0, "grad_norm": 0.7839, "lr": 0.0002226, "router_trainable": true, "sec_per_step": 1.375, "probe_sec": 0.1892, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07383, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 11, "mean_delta_loo": 0.02792, "n_shadow": 2, "mean_delta_shadow": -0.01471, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.52521, "n_dropped_unpaired": 0, "n_full": 0, "read_hit_rate": {"surface": 0.46638, "split": 0.57729, "merge": 0.1684}, "mean_gate": {"surface": 0.12358, "split": 0.08263, "merge": 0.11979}, "n_reads": 10851, "n_cand": 24786, "read_suppressed_frac": 0.1015, "dropout_reads_logged": 10851, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1621, "admitted_surface": 40004, "staged_surface": 67527, "admitted_split": 20034, "staged_split": 50398, "admitted_merge": 5498, "staged_merge": 144219, "admitted_skel": 0, "staged_skel": 0, "mean_count": 271.87054443359375, "n_util_obs": 164, "util_count_spearman": -0.06780841483819189}, "probe_overhead_sec_total": 278.48, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.004437, "alpha": 0.03, "t": 3295.0} |
| {"event": "consolidate", "step": 2400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262093, "t": 3364.7} |
| {"event": "probe", "step": 2400, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3430.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.197092, "cps": 26.796875, "mean_p": 0.213872, "boundary_entropy": 0.161093, "bpb": 2.072434}, "tr": {"chunks": 3883.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.214495, "cps": 30.335938, "mean_p": 0.20719, "boundary_entropy": 0.13822, "bpb": 2.254139}, "fi": {"chunks": 4331.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.230238, "cps": 33.835938, "mean_p": 0.235581, "boundary_entropy": 0.154169, "bpb": 2.403394}}, "bpb": {"en": 2.07243, "tr": 2.25414, "fi": 2.40339}, "gini_cpc": 0.034428, "gini_cps": 0.051586, "bpic": 4.872209, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0882955147005418, "fi": 1.168170676762697}, "inventory": {"size": 326656, "admitted": 65536, "staged": 261120, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1652, "admitted_surface": 39240, "staged_surface": 67424, "admitted_split": 21053, "staged_split": 49547, "admitted_merge": 5243, "staged_merge": 144149, "admitted_skel": 0, "staged_skel": 0, "mean_count": 280.385986328125, "n_util_obs": 170, "util_count_spearman": -0.07167884905441467}, "t": 3366.7} |
| {"event": "eval", "step": 2400, "phase": 1, "bpb": {"en": 2.08451, "tr": 2.26225, "fi": 2.41151}, "bytes_per_chunk": {"en": 5.161, "tr": 5.195, "fi": 4.727}, "bpb_mean": 2.25276, "frozen_inventory": true, "t": 3369.3} |
| {"event": "phase_transition", "step": 2400, "phase": 2, "objective": "global", "router_trainable": true, "encoder_trainable": true, "frozen_router_params": 0, "t": 3375.1} |
| {"event": "train", "step": 2400, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.50369, "bpb": 2.16936, "chunk_loss": 0.03048, "F": 0.20471, "G": 0.21692, "chunks": 209.62, "bytes_per_chunk": 4.885, "boundary_entropy": 0.14571, "router_noise": 0.0, "grad_norm": 0.764, "lr": 0.0002192, "router_trainable": true, "sec_per_step": 1.629, "probe_sec": 0.1662, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09654, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": -0.00886, "n_shadow": 1, "mean_delta_shadow": -0.02399, "n_split": 1, "mean_delta_split": -0.05115, "n_merge": 1, "mean_delta_merge": -0.2124, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -0.32777, "trunc_abs_frac": 0.48309, "mean_delta_full": 0.14214, "read_hit_rate": {"surface": 0.46869, "split": 0.57713, "merge": 0.16331}, "mean_gate": {"surface": 0.12463, "split": 0.08098, "merge": 0.12125}, "n_reads": 10948, "n_cand": 24988, "read_suppressed_frac": 0.0976, "dropout_reads_logged": 10948, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1653, "admitted_surface": 39240, "staged_surface": 67655, "admitted_split": 21053, "staged_split": 49708, "admitted_merge": 5243, "staged_merge": 144781, "admitted_skel": 0, "staged_skel": 0, "mean_count": 280.5530700683594, "n_util_obs": 170, "util_count_spearman": -0.07195368369374006}, "probe_overhead_sec_total": 284.529, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.004362, "alpha": 0.03, "t": 3376.5} |
| {"event": "consolidate", "step": 2450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262093, "t": 3441.4} |
| {"event": "train", "step": 2450, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.46628, "bpb": 2.1154, "chunk_loss": 0.030368, "F": 0.19824, "G": 0.21119, "chunks": 203.0, "bytes_per_chunk": 5.044, "boundary_entropy": 0.14713, "router_noise": 0.0, "grad_norm": 0.72, "lr": 0.0002158, "router_trainable": true, "sec_per_step": 1.32, "probe_sec": 0.1905, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07777, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.20275, "n_shadow": 6, "mean_delta_shadow": -0.13408, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.15146, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 3.60424, "trunc_abs_frac": 0.92839, "mean_delta_full": -0.05925, "read_hit_rate": {"surface": 0.47679, "split": 0.56823, "merge": 0.16869}, "mean_gate": {"surface": 0.13225, "split": 0.08381, "merge": 0.1277}, "n_reads": 10555, "n_cand": 24138, "read_suppressed_frac": 0.0991, "dropout_reads_logged": 10555, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1677, "admitted_surface": 38370, "staged_surface": 67863, "admitted_split": 22073, "staged_split": 49673, "admitted_merge": 5093, "staged_merge": 144608, "admitted_skel": 0, "staged_skel": 0, "mean_count": 288.18109130859375, "n_util_obs": 170, "util_count_spearman": -0.061080514335240366}, "probe_overhead_sec_total": 290.648, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.004374, "alpha": 0.03, "t": 3442.5} |
| {"event": "consolidate", "step": 2500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262091, "t": 3511.3} |
| {"event": "train", "step": 2500, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.44874, "bpb": 2.09009, "chunk_loss": 0.030389, "F": 0.19745, "G": 0.21336, "chunks": 202.19, "bytes_per_chunk": 5.065, "boundary_entropy": 0.14602, "router_noise": 0.0, "grad_norm": 0.7549, "lr": 0.0002123, "router_trainable": true, "sec_per_step": 1.415, "probe_sec": 0.1014, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04383, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.04946, "n_shadow": 4, "mean_delta_shadow": 0.07437, "n_split": 1, "mean_delta_split": -0.04065, "n_merge": 1, "mean_delta_merge": 0.31067, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.2183, "trunc_abs_frac": 8.76879, "mean_delta_full": -0.08807, "read_hit_rate": {"surface": 0.47328, "split": 0.5739, "merge": 0.17002}, "mean_gate": {"surface": 0.12873, "split": 0.07959, "merge": 0.12513}, "n_reads": 10613, "n_cand": 24129, "read_suppressed_frac": 0.1027, "dropout_reads_logged": 10613, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1729, "admitted_surface": 37358, "staged_surface": 67736, "admitted_split": 23083, "staged_split": 49689, "admitted_merge": 5095, "staged_merge": 144719, "admitted_skel": 0, "staged_skel": 0, "mean_count": 296.8852233886719, "n_util_obs": 181, "util_count_spearman": -0.022463724121182686}, "probe_overhead_sec_total": 296.897, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.004679, "alpha": 0.03, "t": 3513.2} |
| {"event": "consolidate", "step": 2550, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262093, "t": 3583.2} |
| {"event": "train", "step": 2550, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.46783, "bpb": 2.11763, "chunk_loss": 0.030405, "F": 0.20193, "G": 0.21379, "chunks": 206.78, "bytes_per_chunk": 4.952, "boundary_entropy": 0.13807, "router_noise": 0.0, "grad_norm": 0.7202, "lr": 0.0002088, "router_trainable": true, "sec_per_step": 1.415, "probe_sec": 0.1265, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05713, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": -0.06465, "n_shadow": 5, "mean_delta_shadow": 0.10538, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": -0.49138, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.10802, "trunc_abs_frac": 0.12711, "mean_delta_full": -0.19818, "read_hit_rate": {"surface": 0.48246, "split": 0.56866, "merge": 0.17473}, "mean_gate": {"surface": 0.13053, "split": 0.07936, "merge": 0.12725}, "n_reads": 10898, "n_cand": 24702, "read_suppressed_frac": 0.1004, "dropout_reads_logged": 10898, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1758, "admitted_surface": 36687, "staged_surface": 67796, "admitted_split": 24101, "staged_split": 49459, "admitted_merge": 4748, "staged_merge": 144889, "admitted_skel": 0, "staged_skel": 0, "mean_count": 304.22479248046875, "n_util_obs": 183, "util_count_spearman": -0.035463645406624736}, "probe_overhead_sec_total": 303.202, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.004492, "alpha": 0.03, "t": 3584.0} |
| {"event": "consolidate", "step": 2600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262093, "t": 3651.0} |
| {"event": "probe", "step": 2600, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3502.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.20123, "cps": 27.359375, "mean_p": 0.216483, "boundary_entropy": 0.150364, "bpb": 2.050454}, "tr": {"chunks": 3905.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.21571, "cps": 30.507812, "mean_p": 0.208175, "boundary_entropy": 0.135011, "bpb": 2.241208}, "fi": {"chunks": 4432.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.235607, "cps": 34.625, "mean_p": 0.237578, "boundary_entropy": 0.144771, "bpb": 2.377779}}, "bpb": {"en": 2.05045, "tr": 2.24121, "fi": 2.37778}, "gini_cpc": 0.035121, "gini_cps": 0.052369, "bpic": 4.791959, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0719597166452148, "fi": 1.170835405277803}, "inventory": {"size": 326656, "admitted": 65536, "staged": 261120, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1797, "admitted_surface": 35669, "staged_surface": 67266, "admitted_split": 25119, "staged_split": 48554, "admitted_merge": 4748, "staged_merge": 145300, "admitted_skel": 0, "staged_skel": 0, "mean_count": 314.17523193359375, "n_util_obs": 192, "util_count_spearman": -0.04987385725524238}, "t": 3652.8} |
| {"event": "eval", "step": 2600, "phase": 2, "bpb": {"en": 2.06301, "tr": 2.25092, "fi": 2.38301}, "bytes_per_chunk": {"en": 5.079, "tr": 5.135, "fi": 4.665}, "bpb_mean": 2.23231, "frozen_inventory": true, "t": 3655.7} |
| {"event": "train", "step": 2600, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.42736, "bpb": 2.05924, "chunk_loss": 0.030372, "F": 0.19867, "G": 0.21276, "chunks": 203.44, "bytes_per_chunk": 5.033, "boundary_entropy": 0.14323, "router_noise": 0.0, "grad_norm": 0.7539, "lr": 0.0002053, "router_trainable": true, "sec_per_step": 1.561, "probe_sec": 0.1456, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.1259, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.05392, "n_shadow": 2, "mean_delta_shadow": 0.06894, "n_split": 2, "mean_delta_split": -0.14626, "n_merge": 1, "mean_delta_merge": -0.05945, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.54924, "trunc_abs_frac": 0.54924, "mean_delta_full": 0.28198, "read_hit_rate": {"surface": 0.47206, "split": 0.58347, "merge": 0.16956}, "mean_gate": {"surface": 0.13015, "split": 0.07805, "merge": 0.12763}, "n_reads": 10708, "n_cand": 24162, "read_suppressed_frac": 0.102, "dropout_reads_logged": 10708, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1799, "admitted_surface": 35669, "staged_surface": 67565, "admitted_split": 25119, "staged_split": 48745, "admitted_merge": 4748, "staged_merge": 145834, "admitted_skel": 0, "staged_skel": 0, "mean_count": 314.3386535644531, "n_util_obs": 192, "util_count_spearman": -0.049670401215310744}, "probe_overhead_sec_total": 309.169, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.004199, "alpha": 0.03, "t": 3662.0} |
| {"event": "consolidate", "step": 2650, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262097, "t": 3730.8} |
| {"event": "train", "step": 2650, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.47447, "bpb": 2.1272, "chunk_loss": 0.030269, "F": 0.1922, "G": 0.2057, "chunks": 196.81, "bytes_per_chunk": 5.203, "boundary_entropy": 0.13877, "router_noise": 0.0, "grad_norm": 0.7197, "lr": 0.0002017, "router_trainable": true, "sec_per_step": 1.405, "probe_sec": 0.1087, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05998, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.01953, "n_shadow": 6, "mean_delta_shadow": 0.00935, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.08997, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.58605, "trunc_abs_frac": 0.48024, "mean_delta_full": 0.15747, "read_hit_rate": {"surface": 0.4751, "split": 0.56222, "merge": 0.14359}, "mean_gate": {"surface": 0.13726, "split": 0.08066, "merge": 0.13201}, "n_reads": 10071, "n_cand": 23525, "read_suppressed_frac": 0.0979, "dropout_reads_logged": 10071, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1843, "admitted_surface": 34860, "staged_surface": 67597, "admitted_split": 26134, "staged_split": 49022, "admitted_merge": 4542, "staged_merge": 145525, "admitted_skel": 0, "staged_skel": 0, "mean_count": 321.3775939941406, "n_util_obs": 198, "util_count_spearman": -0.046745804085803574}, "probe_overhead_sec_total": 315.635, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.004409, "alpha": 0.03, "t": 3732.3} |
| {"event": "consolidate", "step": 2700, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262097, "t": 3798.4} |
| {"event": "train", "step": 2700, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.45778, "bpb": 2.10313, "chunk_loss": 0.03038, "F": 0.19852, "G": 0.21172, "chunks": 203.28, "bytes_per_chunk": 5.037, "boundary_entropy": 0.13683, "router_noise": 0.0, "grad_norm": 0.7381, "lr": 0.0001981, "router_trainable": true, "sec_per_step": 1.355, "probe_sec": 0.1119, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05465, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.00101, "n_shadow": 4, "mean_delta_shadow": -0.04343, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -0.04693, "trunc_abs_frac": 15.26795, "mean_delta_full": -0.12954, "read_hit_rate": {"surface": 0.48849, "split": 0.59227, "merge": 0.16022}, "mean_gate": {"surface": 0.13212, "split": 0.07488, "merge": 0.12822}, "n_reads": 10902, "n_cand": 24240, "read_suppressed_frac": 0.1028, "dropout_reads_logged": 10902, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1876, "admitted_surface": 33861, "staged_surface": 67459, "admitted_split": 27131, "staged_split": 48927, "admitted_merge": 4544, "staged_merge": 145758, "admitted_skel": 0, "staged_skel": 0, "mean_count": 330.718994140625, "n_util_obs": 205, "util_count_spearman": -0.07535624242593082}, "probe_overhead_sec_total": 321.961, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.004308, "alpha": 0.03, "t": 3800.0} |
| {"event": "consolidate", "step": 2750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262101, "t": 3861.9} |
| {"event": "train", "step": 2750, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.46129, "bpb": 2.1082, "chunk_loss": 0.030402, "F": 0.19739, "G": 0.21156, "chunks": 202.12, "bytes_per_chunk": 5.066, "boundary_entropy": 0.14091, "router_noise": 0.0, "grad_norm": 0.6951, "lr": 0.0001945, "router_trainable": true, "sec_per_step": 1.279, "probe_sec": 0.0975, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05193, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 3, "mean_delta_loo": -0.05054, "n_shadow": 5, "mean_delta_shadow": 0.03501, "n_split": 0, "mean_delta_split": NaN, "n_merge": 3, "mean_delta_merge": -0.11364, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -2.04888, "trunc_abs_frac": 1.76416, "mean_delta_full": 0.0448, "read_hit_rate": {"surface": 0.48462, "split": 0.58341, "merge": 0.15162}, "mean_gate": {"surface": 0.13545, "split": 0.07627, "merge": 0.1314}, "n_reads": 10588, "n_cand": 23978, "read_suppressed_frac": 0.103, "dropout_reads_logged": 10588, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1913, "admitted_surface": 32865, "staged_surface": 67478, "admitted_split": 28127, "staged_split": 49397, "admitted_merge": 4544, "staged_merge": 145269, "admitted_skel": 0, "staged_skel": 0, "mean_count": 342.2073669433594, "n_util_obs": 212, "util_count_spearman": -0.04718051670480722}, "probe_overhead_sec_total": 327.35, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.004298, "alpha": 0.03, "t": 3864.0} |
| {"event": "consolidate", "step": 2800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262101, "t": 3939.0} |
| {"event": "probe", "step": 2800, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3451.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.198299, "cps": 26.960938, "mean_p": 0.213104, "boundary_entropy": 0.146496, "bpb": 2.031274}, "tr": {"chunks": 3833.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.211733, "cps": 29.945312, "mean_p": 0.20448, "boundary_entropy": 0.138099, "bpb": 2.200909}, "fi": {"chunks": 4413.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.234597, "cps": 34.476562, "mean_p": 0.237787, "boundary_entropy": 0.144068, "bpb": 2.338495}}, "bpb": {"en": 2.03127, "tr": 2.20091, "fi": 2.3385}, "gini_cpc": 0.037539, "gini_cps": 0.054829, "bpic": 4.850133, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0677447107013598, "fi": 1.1830448379887593}, "inventory": {"size": 326656, "admitted": 65536, "staged": 261120, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1952, "admitted_surface": 32102, "staged_surface": 67226, "admitted_split": 28881, "staged_split": 48922, "admitted_merge": 4553, "staged_merge": 144972, "admitted_skel": 0, "staged_skel": 0, "mean_count": 355.375, "n_util_obs": 216, "util_count_spearman": -0.04901344383714976}, "t": 3940.3} |
| {"event": "eval", "step": 2800, "phase": 2, "bpb": {"en": 2.04328, "tr": 2.20992, "fi": 2.34274}, "bytes_per_chunk": {"en": 5.132, "tr": 5.275, "fi": 4.685}, "bpb_mean": 2.19865, "frozen_inventory": true, "t": 3943.0} |
| {"event": "train", "step": 2800, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.41516, "bpb": 2.04164, "chunk_loss": 0.030374, "F": 0.19733, "G": 0.21124, "chunks": 202.06, "bytes_per_chunk": 5.068, "boundary_entropy": 0.1391, "router_noise": 0.0, "grad_norm": 0.7014, "lr": 0.0001909, "router_trainable": true, "sec_per_step": 1.709, "probe_sec": 0.1321, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07348, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.06559, "n_shadow": 4, "mean_delta_shadow": 0.05576, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.40105, "trunc_abs_frac": 0.21254, "mean_delta_full": -0.12126, "read_hit_rate": {"surface": 0.53839, "split": 0.60013, "merge": 0.16682}, "mean_gate": {"surface": 0.13838, "split": 0.07625, "merge": 0.13438}, "n_reads": 11233, "n_cand": 24002, "read_suppressed_frac": 0.1, "dropout_reads_logged": 11233, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1952, "admitted_surface": 32102, "staged_surface": 67412, "admitted_split": 28881, "staged_split": 49103, "admitted_merge": 4553, "staged_merge": 145629, "admitted_skel": 0, "staged_skel": 0, "mean_count": 355.5464172363281, "n_util_obs": 216, "util_count_spearman": -0.05311923218899963}, "probe_overhead_sec_total": 333.959, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.004319, "alpha": 0.03, "t": 3949.4} |
| {"event": "consolidate", "step": 2850, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262106, "t": 4014.7} |
| {"event": "train", "step": 2850, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.46257, "bpb": 2.11004, "chunk_loss": 0.030552, "F": 0.20734, "G": 0.22099, "chunks": 212.31, "bytes_per_chunk": 4.823, "boundary_entropy": 0.14123, "router_noise": 0.0, "grad_norm": 0.7298, "lr": 0.0001872, "router_trainable": true, "sec_per_step": 1.333, "probe_sec": 0.15, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10812, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 3, "mean_delta_loo": -0.04441, "n_shadow": 6, "mean_delta_shadow": -0.11658, "n_split": 3, "mean_delta_split": 0.03712, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 2.98291, "trunc_abs_frac": 5.30066, "mean_delta_full": -0.04196, "read_hit_rate": {"surface": 0.54584, "split": 0.62438, "merge": 0.1575}, "mean_gate": {"surface": 0.12945, "split": 0.07013, "merge": 0.12833}, "n_reads": 12095, "n_cand": 25254, "read_suppressed_frac": 0.0992, "dropout_reads_logged": 12095, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 1977, "admitted_surface": 31473, "staged_surface": 67411, "admitted_split": 29510, "staged_split": 49192, "admitted_merge": 4553, "staged_merge": 145541, "admitted_skel": 0, "staged_skel": 0, "mean_count": 364.02886962890625, "n_util_obs": 221, "util_count_spearman": -0.03204220263043792}, "probe_overhead_sec_total": 339.334, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.004739, "alpha": 0.03, "t": 4016.0} |
| {"event": "consolidate", "step": 2900, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262107, "t": 4077.5} |
| {"event": "train", "step": 2900, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.39523, "bpb": 2.01289, "chunk_loss": 0.030229, "F": 0.1897, "G": 0.20308, "chunks": 194.25, "bytes_per_chunk": 5.272, "boundary_entropy": 0.13972, "router_noise": 0.0, "grad_norm": 0.6565, "lr": 0.0001835, "router_trainable": true, "sec_per_step": 1.255, "probe_sec": 0.0808, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04508, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": -0.02515, "n_shadow": 2, "mean_delta_shadow": -0.09268, "n_split": 2, "mean_delta_split": 0.11365, "n_merge": 1, "mean_delta_merge": 0.10187, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -1.54352, "trunc_abs_frac": 0.38753, "mean_delta_full": -0.01236, "read_hit_rate": {"surface": 0.52895, "split": 0.59791, "merge": 0.16239}, "mean_gate": {"surface": 0.13745, "split": 0.07365, "merge": 0.13552}, "n_reads": 10743, "n_cand": 23162, "read_suppressed_frac": 0.101, "dropout_reads_logged": 10743, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2033, "admitted_surface": 30731, "staged_surface": 67045, "admitted_split": 30252, "staged_split": 49431, "admitted_merge": 4553, "staged_merge": 145668, "admitted_skel": 0, "staged_skel": 0, "mean_count": 372.13922119140625, "n_util_obs": 228, "util_count_spearman": -0.03279614691847638}, "probe_overhead_sec_total": 345.323, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.004795, "alpha": 0.03, "t": 4078.8} |
| {"event": "consolidate", "step": 2950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262107, "t": 4152.8} |
| {"event": "train", "step": 2950, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.54808, "bpb": 2.23341, "chunk_loss": 0.030445, "F": 0.20215, "G": 0.21315, "chunks": 207.0, "bytes_per_chunk": 4.947, "boundary_entropy": 0.13178, "router_noise": 0.0, "grad_norm": 0.7649, "lr": 0.0001799, "router_trainable": true, "sec_per_step": 1.54, "probe_sec": 0.1289, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06339, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": 0.07142, "n_shadow": 3, "mean_delta_shadow": -0.41207, "n_split": 2, "mean_delta_split": -0.02534, "n_merge": 2, "mean_delta_merge": -0.01921, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.79964, "trunc_abs_frac": 0.7889, "mean_delta_full": -0.42293, "read_hit_rate": {"surface": 0.54688, "split": 0.64853, "merge": 0.14634}, "mean_gate": {"surface": 0.13447, "split": 0.07252, "merge": 0.12973}, "n_reads": 12014, "n_cand": 24638, "read_suppressed_frac": 0.1022, "dropout_reads_logged": 12014, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2064, "admitted_surface": 30531, "staged_surface": 66576, "admitted_split": 30449, "staged_split": 50327, "admitted_merge": 4556, "staged_merge": 145241, "admitted_skel": 0, "staged_skel": 0, "mean_count": 392.2527770996094, "n_util_obs": 241, "util_count_spearman": -0.014285778583934903}, "probe_overhead_sec_total": 351.771, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.004901, "alpha": 0.03, "t": 4155.8} |
| {"event": "consolidate", "step": 3000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262113, "t": 4213.8} |
| {"event": "probe", "step": 3000, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3439.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.19761, "cps": 26.867188, "mean_p": 0.212028, "boundary_entropy": 0.142007, "bpb": 2.011933}, "tr": {"chunks": 3802.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.21002, "cps": 29.703125, "mean_p": 0.201363, "boundary_entropy": 0.131377, "bpb": 2.174314}, "fi": {"chunks": 4385.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.233108, "cps": 34.257812, "mean_p": 0.234619, "boundary_entropy": 0.134597, "bpb": 2.323738}}, "bpb": {"en": 2.01193, "tr": 2.17431, "fi": 2.32374}, "gini_cpc": 0.036935, "gini_cps": 0.054246, "bpic": 4.879752, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.062804795864805, "fi": 1.179640456793476}, "inventory": {"size": 326656, "admitted": 65536, "staged": 261120, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2086, "admitted_surface": 29919, "staged_surface": 66190, "admitted_split": 31059, "staged_split": 49257, "admitted_merge": 4558, "staged_merge": 145673, "admitted_skel": 0, "staged_skel": 0, "mean_count": 400.9408874511719, "n_util_obs": 241, "util_count_spearman": -0.019823141315271988}, "t": 4215.8} |
| {"event": "eval", "step": 3000, "phase": 2, "bpb": {"en": 2.02642, "tr": 2.18463, "fi": 2.32541}, "bytes_per_chunk": {"en": 5.121, "tr": 5.268, "fi": 4.67}, "bpb_mean": 2.17882, "frozen_inventory": true, "t": 4219.3} |
| {"event": "train", "step": 3000, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.49303, "bpb": 2.15399, "chunk_loss": 0.030418, "F": 0.19785, "G": 0.20928, "chunks": 202.59, "bytes_per_chunk": 5.054, "boundary_entropy": 0.13305, "router_noise": 0.0, "grad_norm": 0.7754, "lr": 0.0001761, "router_trainable": true, "sec_per_step": 1.473, "probe_sec": 0.1706, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.1, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.01372, "n_shadow": 3, "mean_delta_shadow": -0.02902, "n_split": 1, "mean_delta_split": 0.03189, "n_merge": 1, "mean_delta_merge": 0.13434, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -0.16747, "trunc_abs_frac": 0.25528, "mean_delta_full": -0.12933, "read_hit_rate": {"surface": 0.55557, "split": 0.62935, "merge": 0.14784}, "mean_gate": {"surface": 0.1267, "split": 0.06649, "merge": 0.12529}, "n_reads": 11627, "n_cand": 24142, "read_suppressed_frac": 0.1, "dropout_reads_logged": 11627, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2087, "admitted_surface": 29919, "staged_surface": 66468, "admitted_split": 31059, "staged_split": 49478, "admitted_merge": 4558, "staged_merge": 146198, "admitted_skel": 0, "staged_skel": 0, "mean_count": 401.11834716796875, "n_util_obs": 241, "util_count_spearman": -0.019384631607556223}, "probe_overhead_sec_total": 357.508, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.004737, "alpha": 0.03, "t": 4229.4} |
| {"event": "consolidate", "step": 3050, "retired": 1, "split_retired": 0, "merge_admitted": 0, "admitted": 1, "swapped": 500, "uncertain_staged": 262113, "t": 4298.1} |
| {"event": "train", "step": 3050, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.44438, "bpb": 2.08379, "chunk_loss": 0.030484, "F": 0.20093, "G": 0.21371, "chunks": 205.75, "bytes_per_chunk": 4.977, "boundary_entropy": 0.13312, "router_noise": 0.0, "grad_norm": 0.7053, "lr": 0.0001724, "router_trainable": true, "sec_per_step": 1.4, "probe_sec": 0.0952, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04382, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 11, "mean_delta_loo": -0.01236, "n_shadow": 3, "mean_delta_shadow": -0.01168, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.56866, "trunc_abs_frac": 0.64467, "mean_delta_full": -0.16669, "read_hit_rate": {"surface": 0.56548, "split": 0.65115, "merge": 0.16442}, "mean_gate": {"surface": 0.13056, "split": 0.06706, "merge": 0.13008}, "n_reads": 12148, "n_cand": 24392, "read_suppressed_frac": 0.1006, "dropout_reads_logged": 12148, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2124, "admitted_surface": 29591, "staged_surface": 66271, "admitted_split": 31387, "staged_split": 50212, "admitted_merge": 4558, "staged_merge": 145661, "admitted_skel": 0, "staged_skel": 0, "mean_count": 413.17559814453125, "n_util_obs": 244, "util_count_spearman": -0.028270743398788628}, "probe_overhead_sec_total": 363.439, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.004963, "alpha": 0.03, "t": 4299.5} |
| {"event": "consolidate", "step": 3100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 188, "uncertain_staged": 262115, "t": 4365.1} |
| {"event": "train", "step": 3100, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.48904, "bpb": 2.14824, "chunk_loss": 0.030417, "F": 0.20053, "G": 0.21193, "chunks": 205.34, "bytes_per_chunk": 4.987, "boundary_entropy": 0.12632, "router_noise": 0.0, "grad_norm": 0.7126, "lr": 0.0001687, "router_trainable": true, "sec_per_step": 1.348, "probe_sec": 0.2188, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.11609, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": -0.13231, "n_shadow": 5, "mean_delta_shadow": -0.01988, "n_split": 1, "mean_delta_split": 0.03939, "n_merge": 1, "mean_delta_merge": 0.15894, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -1.18455, "trunc_abs_frac": 0.69193, "mean_delta_full": 0.04849, "read_hit_rate": {"surface": 0.56369, "split": 0.65685, "merge": 0.15798}, "mean_gate": {"surface": 0.13602, "split": 0.06901, "merge": 0.13382}, "n_reads": 12164, "n_cand": 24388, "read_suppressed_frac": 0.0989, "dropout_reads_logged": 12164, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2150, "admitted_surface": 29430, "staged_surface": 66130, "admitted_split": 31548, "staged_split": 49969, "admitted_merge": 4558, "staged_merge": 146045, "admitted_skel": 0, "staged_skel": 0, "mean_count": 422.4054870605469, "n_util_obs": 246, "util_count_spearman": -0.02322383283322216}, "probe_overhead_sec_total": 370.529, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.005041, "alpha": 0.03, "t": 4366.9} |
| {"event": "consolidate", "step": 3150, "retired": 1, "split_retired": 0, "merge_admitted": 0, "admitted": 1, "swapped": 113, "uncertain_staged": 262115, "t": 4430.6} |
| {"event": "train", "step": 3150, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.45255, "bpb": 2.09559, "chunk_loss": 0.030325, "F": 0.19504, "G": 0.20751, "chunks": 199.72, "bytes_per_chunk": 5.127, "boundary_entropy": 0.13172, "router_noise": 0.0, "grad_norm": 0.7035, "lr": 0.000165, "router_trainable": true, "sec_per_step": 1.3, "probe_sec": 0.0936, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04666, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": 0.1353, "n_shadow": 6, "mean_delta_shadow": -0.06509, "n_split": 1, "mean_delta_split": 0.18311, "n_merge": 1, "mean_delta_merge": 0.16403, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 18.52287, "trunc_abs_frac": 2.5728, "mean_delta_full": 0.00774, "read_hit_rate": {"surface": 0.55716, "split": 0.64891, "merge": 0.15489}, "mean_gate": {"surface": 0.13147, "split": 0.06605, "merge": 0.13144}, "n_reads": 11748, "n_cand": 23820, "read_suppressed_frac": 0.0984, "dropout_reads_logged": 11748, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2186, "admitted_surface": 29368, "staged_surface": 66074, "admitted_split": 31610, "staged_split": 49934, "admitted_merge": 4558, "staged_merge": 146136, "admitted_skel": 0, "staged_skel": 0, "mean_count": 433.8033447265625, "n_util_obs": 249, "util_count_spearman": -0.024098722133448207}, "probe_overhead_sec_total": 376.99, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.005126, "alpha": 0.03, "t": 4431.8} |
| {"event": "consolidate", "step": 3200, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 86, "uncertain_staged": 262117, "t": 4506.6} |
| {"event": "probe", "step": 3200, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3464.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.199046, "cps": 27.0625, "mean_p": 0.212745, "boundary_entropy": 0.13555, "bpb": 1.998438}, "tr": {"chunks": 3870.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.213777, "cps": 30.234375, "mean_p": 0.203166, "boundary_entropy": 0.116853, "bpb": 2.150099}, "fi": {"chunks": 4424.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.235182, "cps": 34.5625, "mean_p": 0.23439, "boundary_entropy": 0.132062, "bpb": 2.301478}}, "bpb": {"en": 1.99844, "tr": 2.1501, "fi": 2.30148}, "gini_cpc": 0.037176, "gini_cps": 0.054431, "bpic": 4.82497, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.074005858700005, "fi": 1.1815428371415035}, "inventory": {"size": 327593, "admitted": 65536, "staged": 262057, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2217, "admitted_surface": 29335, "staged_surface": 66298, "admitted_split": 31642, "staged_split": 50013, "admitted_merge": 4559, "staged_merge": 145746, "admitted_skel": 0, "staged_skel": 0, "mean_count": 442.53759765625, "n_util_obs": 250, "util_count_spearman": -0.004745931934910959}, "t": 4508.5} |
| {"event": "eval", "step": 3200, "phase": 2, "bpb": {"en": 2.01096, "tr": 2.15904, "fi": 2.3066}, "bytes_per_chunk": {"en": 5.109, "tr": 5.211, "fi": 4.646}, "bpb_mean": 2.15887, "frozen_inventory": true, "t": 4512.0} |
| {"event": "train", "step": 3200, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.43938, "bpb": 2.07659, "chunk_loss": 0.030333, "F": 0.1969, "G": 0.20928, "chunks": 201.62, "bytes_per_chunk": 5.079, "boundary_entropy": 0.12729, "router_noise": 0.0, "grad_norm": 0.6507, "lr": 0.0001613, "router_trainable": true, "sec_per_step": 1.727, "probe_sec": 0.1199, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04781, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.10099, "n_shadow": 4, "mean_delta_shadow": -0.15516, "n_split": 2, "mean_delta_split": 0.00525, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.32132, "trunc_abs_frac": 0.43701, "mean_delta_full": 0.26514, "read_hit_rate": {"surface": 0.5662, "split": 0.65032, "merge": 0.15592}, "mean_gate": {"surface": 0.13478, "split": 0.06637, "merge": 0.13459}, "n_reads": 11847, "n_cand": 23904, "read_suppressed_frac": 0.101, "dropout_reads_logged": 11847, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2217, "admitted_surface": 29335, "staged_surface": 66285, "admitted_split": 31642, "staged_split": 50071, "admitted_merge": 4559, "staged_merge": 145788, "admitted_skel": 0, "staged_skel": 0, "mean_count": 442.7183837890625, "n_util_obs": 250, "util_count_spearman": -0.0047274996399942395}, "probe_overhead_sec_total": 383.943, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.004566, "alpha": 0.03, "t": 4518.2} |
| {"event": "consolidate", "step": 3250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 654, "uncertain_staged": 262117, "t": 4570.5} |
| {"event": "train", "step": 3250, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.43944, "bpb": 2.07667, "chunk_loss": 0.030498, "F": 0.20331, "G": 0.2131, "chunks": 208.19, "bytes_per_chunk": 4.919, "boundary_entropy": 0.12301, "router_noise": 0.0, "grad_norm": 0.7317, "lr": 0.0001576, "router_trainable": true, "sec_per_step": 1.08, "probe_sec": 0.1047, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04398, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": 0.03561, "n_shadow": 5, "mean_delta_shadow": 0.00166, "n_split": 2, "mean_delta_split": 0.04608, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 0.61822, "trunc_abs_frac": 0.50414, "mean_delta_full": 0.02964, "read_hit_rate": {"surface": 0.59774, "split": 0.65869, "merge": 0.16702}, "mean_gate": {"surface": 0.13756, "split": 0.06795, "merge": 0.13709}, "n_reads": 12611, "n_cand": 24684, "read_suppressed_frac": 0.1017, "dropout_reads_logged": 12611, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2236, "admitted_surface": 29277, "staged_surface": 65653, "admitted_split": 31700, "staged_split": 50367, "admitted_merge": 4559, "staged_merge": 146124, "admitted_skel": 0, "staged_skel": 0, "mean_count": 452.15484619140625, "n_util_obs": 251, "util_count_spearman": -0.0014441282489091253}, "probe_overhead_sec_total": 389.826, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.005277, "alpha": 0.03, "t": 4572.2} |
| {"event": "consolidate", "step": 3300, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 384, "uncertain_staged": 262117, "t": 4649.4} |
| {"event": "train", "step": 3300, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.41762, "bpb": 2.04519, "chunk_loss": 0.030442, "F": 0.2002, "G": 0.21183, "chunks": 205.0, "bytes_per_chunk": 4.995, "boundary_entropy": 0.1307, "router_noise": 0.0, "grad_norm": 0.6882, "lr": 0.0001539, "router_trainable": true, "sec_per_step": 1.56, "probe_sec": 0.1088, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08163, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": 0.0469, "n_shadow": 3, "mean_delta_shadow": -0.07784, "n_split": 1, "mean_delta_split": -0.03247, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.22817, "trunc_abs_frac": 1.63724, "mean_delta_full": 0.04941, "read_hit_rate": {"surface": 0.59452, "split": 0.65803, "merge": 0.16118}, "mean_gate": {"surface": 0.12924, "split": 0.06231, "merge": 0.13138}, "n_reads": 12394, "n_cand": 24370, "read_suppressed_frac": 0.0967, "dropout_reads_logged": 12394, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2274, "admitted_surface": 29241, "staged_surface": 65389, "admitted_split": 31736, "staged_split": 50071, "admitted_merge": 4559, "staged_merge": 146684, "admitted_skel": 0, "staged_skel": 0, "mean_count": 461.2489013671875, "n_util_obs": 253, "util_count_spearman": -0.011648741535766527}, "probe_overhead_sec_total": 397.137, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.005353, "alpha": 0.03, "t": 4650.2} |
| {"event": "consolidate", "step": 3350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 84, "uncertain_staged": 262118, "t": 4717.4} |
| {"event": "train", "step": 3350, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.35683, "bpb": 1.9575, "chunk_loss": 0.03031, "F": 0.19467, "G": 0.20763, "chunks": 199.34, "bytes_per_chunk": 5.137, "boundary_entropy": 0.1264, "router_noise": 0.0, "grad_norm": 0.7463, "lr": 0.0001501, "router_trainable": true, "sec_per_step": 1.365, "probe_sec": 0.1223, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09093, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": -0.0206, "n_shadow": 4, "mean_delta_shadow": 0.0388, "n_split": 1, "mean_delta_split": -0.05493, "n_merge": 3, "mean_delta_merge": 0.163, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 34.96602, "trunc_abs_frac": 1.97306, "mean_delta_full": 0.00629, "read_hit_rate": {"surface": 0.58106, "split": 0.65822, "merge": 0.16374}, "mean_gate": {"surface": 0.13282, "split": 0.06182, "merge": 0.13284}, "n_reads": 12123, "n_cand": 23906, "read_suppressed_frac": 0.1023, "dropout_reads_logged": 12123, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2309, "admitted_surface": 29227, "staged_surface": 65742, "admitted_split": 31750, "staged_split": 50159, "admitted_merge": 4559, "staged_merge": 146243, "admitted_skel": 0, "staged_skel": 0, "mean_count": 470.6470947265625, "n_util_obs": 257, "util_count_spearman": -0.024058205998433422}, "probe_overhead_sec_total": 404.191, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.005035, "alpha": 0.03, "t": 4718.5} |
| {"event": "consolidate", "step": 3400, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 624, "uncertain_staged": 262118, "t": 4785.2} |
| {"event": "probe", "step": 3400, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3478.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.199851, "cps": 27.171875, "mean_p": 0.211467, "boundary_entropy": 0.129173, "bpb": 1.984304}, "tr": {"chunks": 3909.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.215931, "cps": 30.539062, "mean_p": 0.203148, "boundary_entropy": 0.113727, "bpb": 2.147462}, "fi": {"chunks": 4451.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.236617, "cps": 34.773438, "mean_p": 0.235955, "boundary_entropy": 0.129216, "bpb": 2.267547}}, "bpb": {"en": 1.9843, "tr": 2.14746, "fi": 2.26755}, "gini_cpc": 0.03757, "gini_cps": 0.054795, "bpic": 4.792364, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.080462408624192, "fi": 1.1839687874256828}, "inventory": {"size": 327055, "admitted": 65536, "staged": 261519, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2338, "admitted_surface": 29195, "staged_surface": 64835, "admitted_split": 31777, "staged_split": 50042, "admitted_merge": 4564, "staged_merge": 146642, "admitted_skel": 0, "staged_skel": 0, "mean_count": 479.63470458984375, "n_util_obs": 263, "util_count_spearman": -0.01293923474140001}, "t": 4786.5} |
| {"event": "eval", "step": 3400, "phase": 2, "bpb": {"en": 1.9967, "tr": 2.15637, "fi": 2.27121}, "bytes_per_chunk": {"en": 5.098, "tr": 5.176, "fi": 4.606}, "bpb_mean": 2.14142, "frozen_inventory": true, "t": 4789.7} |
| {"event": "train", "step": 3400, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.42573, "bpb": 2.05689, "chunk_loss": 0.030365, "F": 0.19849, "G": 0.20905, "chunks": 203.25, "bytes_per_chunk": 5.038, "boundary_entropy": 0.12106, "router_noise": 0.0, "grad_norm": 0.7918, "lr": 0.0001465, "router_trainable": true, "sec_per_step": 1.558, "probe_sec": 0.1064, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04681, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": -0.13358, "n_shadow": 6, "mean_delta_shadow": 0.00971, "n_split": 2, "mean_delta_split": -0.12131, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 1.05752, "trunc_abs_frac": 1.37241, "mean_delta_full": -0.10017, "read_hit_rate": {"surface": 0.58684, "split": 0.65338, "merge": 0.16398}, "mean_gate": {"surface": 0.13306, "split": 0.06285, "merge": 0.13141}, "n_reads": 12249, "n_cand": 24230, "read_suppressed_frac": 0.1012, "dropout_reads_logged": 12249, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2338, "admitted_surface": 29195, "staged_surface": 64971, "admitted_split": 31777, "staged_split": 50116, "admitted_merge": 4564, "staged_merge": 147057, "admitted_skel": 0, "staged_skel": 0, "mean_count": 479.8216552734375, "n_util_obs": 263, "util_count_spearman": -0.01447558333751118}, "probe_overhead_sec_total": 410.948, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.005245, "alpha": 0.03, "t": 4796.4} |
| {"event": "consolidate", "step": 3450, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 72, "uncertain_staged": 262118, "t": 4861.7} |
| {"event": "train", "step": 3450, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.38945, "bpb": 2.00455, "chunk_loss": 0.030385, "F": 0.20074, "G": 0.20986, "chunks": 205.56, "bytes_per_chunk": 4.981, "boundary_entropy": 0.12028, "router_noise": 0.0, "grad_norm": 0.7933, "lr": 0.0001428, "router_trainable": true, "sec_per_step": 1.325, "probe_sec": 0.1137, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05368, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.04129, "n_shadow": 6, "mean_delta_shadow": -0.0721, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.05646, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.18481, "trunc_abs_frac": 0.94573, "mean_delta_full": 0.13165, "read_hit_rate": {"surface": 0.58005, "split": 0.65801, "merge": 0.17086}, "mean_gate": {"surface": 0.13295, "split": 0.06115, "merge": 0.13254}, "n_reads": 12331, "n_cand": 24338, "read_suppressed_frac": 0.0991, "dropout_reads_logged": 12331, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2375, "admitted_surface": 29180, "staged_surface": 64987, "admitted_split": 31787, "staged_split": 50446, "admitted_merge": 4569, "staged_merge": 146711, "admitted_skel": 0, "staged_skel": 0, "mean_count": 488.63726806640625, "n_util_obs": 267, "util_count_spearman": -0.024770550091881144}, "probe_overhead_sec_total": 417.191, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.004865, "alpha": 0.03, "t": 4862.6} |
| {"event": "consolidate", "step": 3500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 312, "uncertain_staged": 262118, "t": 4924.3} |
| {"event": "train", "step": 3500, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.36839, "bpb": 1.97416, "chunk_loss": 0.030323, "F": 0.19666, "G": 0.20775, "chunks": 201.38, "bytes_per_chunk": 5.085, "boundary_entropy": 0.12236, "router_noise": 0.0, "grad_norm": 0.7221, "lr": 0.0001391, "router_trainable": true, "sec_per_step": 1.247, "probe_sec": 0.1196, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09215, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.09998, "n_shadow": 4, "mean_delta_shadow": -0.02642, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": 0.18173, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -6.51277, "trunc_abs_frac": 0.66016, "mean_delta_full": -0.02669, "read_hit_rate": {"surface": 0.59357, "split": 0.66217, "merge": 0.17947}, "mean_gate": {"surface": 0.13324, "split": 0.06129, "merge": 0.13267}, "n_reads": 12370, "n_cand": 23996, "read_suppressed_frac": 0.0964, "dropout_reads_logged": 12370, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2410, "admitted_surface": 29162, "staged_surface": 64813, "admitted_split": 31804, "staged_split": 50158, "admitted_merge": 4570, "staged_merge": 147173, "admitted_skel": 0, "staged_skel": 0, "mean_count": 498.07708740234375, "n_util_obs": 272, "util_count_spearman": -0.02701783642334742}, "probe_overhead_sec_total": 422.775, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.005017, "alpha": 0.03, "t": 4925.0} |
| {"event": "consolidate", "step": 3550, "retired": 1, "split_retired": 0, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262118, "t": 4994.0} |
| {"event": "train", "step": 3550, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.41021, "bpb": 2.0345, "chunk_loss": 0.030312, "F": 0.19443, "G": 0.20441, "chunks": 199.09, "bytes_per_chunk": 5.143, "boundary_entropy": 0.1209, "router_noise": 0.0, "grad_norm": 0.6696, "lr": 0.0001355, "router_trainable": true, "sec_per_step": 1.418, "probe_sec": 0.1015, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05406, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": -0.0099, "n_shadow": 5, "mean_delta_shadow": 0.04006, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -13.69792, "trunc_abs_frac": 0.95315, "mean_delta_full": -0.00117, "read_hit_rate": {"surface": 0.59016, "split": 0.65635, "merge": 0.15824}, "mean_gate": {"surface": 0.13616, "split": 0.06224, "merge": 0.13443}, "n_reads": 12010, "n_cand": 23724, "read_suppressed_frac": 0.0951, "dropout_reads_logged": 12010, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2431, "admitted_surface": 29135, "staged_surface": 63950, "admitted_split": 31831, "staged_split": 50264, "admitted_merge": 4570, "staged_merge": 147930, "admitted_skel": 0, "staged_skel": 0, "mean_count": 507.5592956542969, "n_util_obs": 275, "util_count_spearman": -0.04761547560658954}, "probe_overhead_sec_total": 428.861, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.005568, "alpha": 0.03, "t": 4995.9} |
| {"event": "consolidate", "step": 3600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 102, "uncertain_staged": 262119, "t": 5062.6} |
| {"event": "probe", "step": 3600, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3469.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.199333, "cps": 27.101562, "mean_p": 0.209549, "boundary_entropy": 0.124458, "bpb": 1.96764}, "tr": {"chunks": 3866.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.213556, "cps": 30.203125, "mean_p": 0.200322, "boundary_entropy": 0.117635, "bpb": 2.105334}, "fi": {"chunks": 4418.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.234863, "cps": 34.515625, "mean_p": 0.233529, "boundary_entropy": 0.123345, "bpb": 2.247283}}, "bpb": {"en": 1.96764, "tr": 2.10533, "fi": 2.24728}, "gini_cpc": 0.036567, "gini_cps": 0.05383, "bpic": 4.827023, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0713493701451196, "fi": 1.1782396906283206}, "inventory": {"size": 327578, "admitted": 65536, "staged": 262042, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2458, "admitted_surface": 29121, "staged_surface": 64508, "admitted_split": 31844, "staged_split": 49822, "admitted_merge": 4571, "staged_merge": 147712, "admitted_skel": 0, "staged_skel": 0, "mean_count": 516.8505859375, "n_util_obs": 278, "util_count_spearman": -0.054701233265336546}, "t": 5064.2} |
| {"event": "eval", "step": 3600, "phase": 2, "bpb": {"en": 1.98159, "tr": 2.11446, "fi": 2.25051}, "bytes_per_chunk": {"en": 5.114, "tr": 5.215, "fi": 4.646}, "bpb_mean": 2.11552, "frozen_inventory": true, "t": 5067.1} |
| {"event": "train", "step": 3600, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.40714, "bpb": 2.03007, "chunk_loss": 0.030406, "F": 0.19943, "G": 0.20887, "chunks": 204.22, "bytes_per_chunk": 5.014, "boundary_entropy": 0.11794, "router_noise": 0.0, "grad_norm": 0.634, "lr": 0.0001319, "router_trainable": true, "sec_per_step": 1.539, "probe_sec": 0.0795, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05077, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.00936, "n_shadow": 6, "mean_delta_shadow": -0.02653, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.08374, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -0.142, "trunc_abs_frac": 0.48005, "mean_delta_full": 0.11761, "read_hit_rate": {"surface": 0.58419, "split": 0.66005, "merge": 0.15917}, "mean_gate": {"surface": 0.13188, "split": 0.05891, "merge": 0.13204}, "n_reads": 12372, "n_cand": 24402, "read_suppressed_frac": 0.0994, "dropout_reads_logged": 12372, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2458, "admitted_surface": 29121, "staged_surface": 64559, "admitted_split": 31844, "staged_split": 49894, "admitted_merge": 4571, "staged_merge": 147691, "admitted_skel": 0, "staged_skel": 0, "mean_count": 517.0393676757812, "n_util_obs": 278, "util_count_spearman": -0.05267540945699246}, "probe_overhead_sec_total": 435.046, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.00522, "alpha": 0.03, "t": 5072.8} |
| {"event": "consolidate", "step": 3650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 676, "uncertain_staged": 262119, "t": 5135.5} |
| {"event": "train", "step": 3650, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.4385, "bpb": 2.07531, "chunk_loss": 0.030552, "F": 0.20483, "G": 0.21272, "chunks": 209.75, "bytes_per_chunk": 4.882, "boundary_entropy": 0.11878, "router_noise": 0.0, "grad_norm": 0.6858, "lr": 0.0001283, "router_trainable": true, "sec_per_step": 1.274, "probe_sec": 0.0761, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04826, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": -0.05151, "n_shadow": 5, "mean_delta_shadow": -0.05164, "n_split": 1, "mean_delta_split": 0.01025, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -0.35075, "trunc_abs_frac": 0.18739, "mean_delta_full": 0.19401, "read_hit_rate": {"surface": 0.59746, "split": 0.67238, "merge": 0.16291}, "mean_gate": {"surface": 0.12955, "split": 0.05829, "merge": 0.13006}, "n_reads": 12879, "n_cand": 24936, "read_suppressed_frac": 0.0985, "dropout_reads_logged": 12879, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2492, "admitted_surface": 29022, "staged_surface": 64228, "admitted_split": 31942, "staged_split": 49979, "admitted_merge": 4572, "staged_merge": 147937, "admitted_skel": 0, "staged_skel": 0, "mean_count": 526.2046508789062, "n_util_obs": 283, "util_count_spearman": -0.062072003481682514}, "probe_overhead_sec_total": 441.136, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.005355, "alpha": 0.03, "t": 5136.5} |
| {"event": "consolidate", "step": 3700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 301, "uncertain_staged": 262120, "t": 5209.2} |
| {"event": "train", "step": 3700, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.39876, "bpb": 2.01798, "chunk_loss": 0.030361, "F": 0.19946, "G": 0.20903, "chunks": 204.25, "bytes_per_chunk": 5.013, "boundary_entropy": 0.12122, "router_noise": 0.0, "grad_norm": 0.6887, "lr": 0.0001247, "router_trainable": true, "sec_per_step": 1.475, "probe_sec": 0.0971, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05672, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": 0.02605, "n_shadow": 6, "mean_delta_shadow": -0.06824, "n_split": 2, "mean_delta_split": 0.18022, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 2.34254, "trunc_abs_frac": 1.05852, "mean_delta_full": -0.03479, "read_hit_rate": {"surface": 0.59732, "split": 0.6628, "merge": 0.15745}, "mean_gate": {"surface": 0.13724, "split": 0.06071, "merge": 0.1379}, "n_reads": 12334, "n_cand": 24186, "read_suppressed_frac": 0.0958, "dropout_reads_logged": 12334, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2515, "admitted_surface": 28998, "staged_surface": 64159, "admitted_split": 31967, "staged_split": 49823, "admitted_merge": 4571, "staged_merge": 148162, "admitted_skel": 0, "staged_skel": 0, "mean_count": 535.51953125, "n_util_obs": 289, "util_count_spearman": -0.06445905227834987}, "probe_overhead_sec_total": 448.418, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.00533, "alpha": 0.03, "t": 5210.3} |
| {"event": "consolidate", "step": 3750, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 758, "uncertain_staged": 262120, "t": 5278.7} |
| {"event": "train", "step": 3750, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.37751, "bpb": 1.98732, "chunk_loss": 0.030302, "F": 0.19446, "G": 0.20503, "chunks": 199.12, "bytes_per_chunk": 5.142, "boundary_entropy": 0.1175, "router_noise": 0.0, "grad_norm": 0.6903, "lr": 0.0001212, "router_trainable": true, "sec_per_step": 1.399, "probe_sec": 0.0841, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05346, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": -0.00617, "n_shadow": 2, "mean_delta_shadow": 0.0227, "n_split": 1, "mean_delta_split": -0.02087, "n_merge": 1, "mean_delta_merge": 0.13025, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.84444, "trunc_abs_frac": 0.84444, "mean_delta_full": 0.14282, "read_hit_rate": {"surface": 0.57965, "split": 0.6633, "merge": 0.15536}, "mean_gate": {"surface": 0.13597, "split": 0.06066, "merge": 0.13432}, "n_reads": 12076, "n_cand": 23836, "read_suppressed_frac": 0.0943, "dropout_reads_logged": 12076, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2544, "admitted_surface": 29033, "staged_surface": 63770, "admitted_split": 31961, "staged_split": 50029, "admitted_merge": 4542, "staged_merge": 148345, "admitted_skel": 0, "staged_skel": 0, "mean_count": 539.8145141601562, "n_util_obs": 279, "util_count_spearman": -0.14300911478316766}, "probe_overhead_sec_total": 455.086, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.005721, "alpha": 0.03, "t": 5280.2} |
| {"event": "consolidate", "step": 3800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262120, "t": 5334.2} |
| {"event": "probe", "step": 3800, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3448.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.198127, "cps": 26.9375, "mean_p": 0.209902, "boundary_entropy": 0.125091, "bpb": 1.954955}, "tr": {"chunks": 3867.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.213611, "cps": 30.210938, "mean_p": 0.200903, "boundary_entropy": 0.111559, "bpb": 2.100006}, "fi": {"chunks": 4420.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.234969, "cps": 34.53125, "mean_p": 0.233425, "boundary_entropy": 0.122631, "bpb": 2.22339}}, "bpb": {"en": 1.95496, "tr": 2.10001, "fi": 2.22339}, "gini_cpc": 0.037979, "gini_cps": 0.055219, "bpic": 4.834427, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0781532185061686, "fi": 1.1859523744622447}, "inventory": {"size": 326656, "admitted": 65536, "staged": 261120, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2559, "admitted_surface": 29221, "staged_surface": 63044, "admitted_split": 31850, "staged_split": 50176, "admitted_merge": 4465, "staged_merge": 147900, "admitted_skel": 0, "staged_skel": 0, "mean_count": 540.4419555664062, "n_util_obs": 263, "util_count_spearman": -0.19868737494283903}, "t": 5335.4} |
| {"event": "eval", "step": 3800, "phase": 2, "bpb": {"en": 1.97057, "tr": 2.10982, "fi": 2.22855}, "bytes_per_chunk": {"en": 5.135, "tr": 5.2, "fi": 4.674}, "bpb_mean": 2.10298, "frozen_inventory": true, "t": 5337.6} |
| {"event": "train", "step": 3800, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.45233, "bpb": 2.09527, "chunk_loss": 0.030368, "F": 0.19943, "G": 0.20886, "chunks": 204.22, "bytes_per_chunk": 5.014, "boundary_entropy": 0.11578, "router_noise": 0.0, "grad_norm": 0.7069, "lr": 0.0001177, "router_trainable": true, "sec_per_step": 1.261, "probe_sec": 0.0993, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05959, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": 0.02308, "n_shadow": 3, "mean_delta_shadow": 0.15148, "n_split": 1, "mean_delta_split": 0.05054, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.61371, "trunc_abs_frac": 0.75726, "mean_delta_full": 0.14861, "read_hit_rate": {"surface": 0.57727, "split": 0.64964, "merge": 0.13738}, "mean_gate": {"surface": 0.1313, "split": 0.05762, "merge": 0.13107}, "n_reads": 12097, "n_cand": 24448, "read_suppressed_frac": 0.0971, "dropout_reads_logged": 12097, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2558, "admitted_surface": 29221, "staged_surface": 63347, "admitted_split": 31850, "staged_split": 50359, "admitted_merge": 4465, "staged_merge": 148438, "admitted_skel": 0, "staged_skel": 0, "mean_count": 540.6265869140625, "n_util_obs": 263, "util_count_spearman": -0.19991859793183336}, "probe_overhead_sec_total": 461.089, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.004813, "alpha": 0.03, "t": 5343.3} |
| {"event": "consolidate", "step": 3850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 355, "uncertain_staged": 262120, "t": 5402.5} |
| {"event": "train", "step": 3850, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.54376, "bpb": 2.22718, "chunk_loss": 0.030652, "F": 0.20276, "G": 0.21161, "chunks": 207.62, "bytes_per_chunk": 4.932, "boundary_entropy": 0.1115, "router_noise": 0.0, "grad_norm": 0.7104, "lr": 0.0001142, "router_trainable": true, "sec_per_step": 1.203, "probe_sec": 0.13, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09867, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": -0.15335, "n_shadow": 1, "mean_delta_shadow": 0.04596, "n_split": 2, "mean_delta_split": 0.10294, "n_merge": 2, "mean_delta_merge": -0.08486, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.67032, "trunc_abs_frac": 0.68791, "mean_delta_full": -0.18079, "read_hit_rate": {"surface": 0.58606, "split": 0.64216, "merge": 0.15395}, "mean_gate": {"surface": 0.13801, "split": 0.06172, "merge": 0.13565}, "n_reads": 12016, "n_cand": 24292, "read_suppressed_frac": 0.1014, "dropout_reads_logged": 12016, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2595, "admitted_surface": 29185, "staged_surface": 63325, "admitted_split": 31900, "staged_split": 50346, "admitted_merge": 4451, "staged_merge": 148473, "admitted_skel": 0, "staged_skel": 0, "mean_count": 550.0289306640625, "n_util_obs": 266, "util_count_spearman": -0.2055311056951261}, "probe_overhead_sec_total": 468.169, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.005623, "alpha": 0.03, "t": 5403.4} |
| {"event": "consolidate", "step": 3900, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262120, "t": 5458.7} |
| {"event": "train", "step": 3900, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.43979, "bpb": 2.07718, "chunk_loss": 0.030458, "F": 0.20047, "G": 0.21146, "chunks": 205.28, "bytes_per_chunk": 4.988, "boundary_entropy": 0.12054, "router_noise": 0.0, "grad_norm": 0.7613, "lr": 0.0001108, "router_trainable": true, "sec_per_step": 1.126, "probe_sec": 0.0757, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04315, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.03277, "n_shadow": 5, "mean_delta_shadow": -0.01482, "n_split": 1, "mean_delta_split": 0.13736, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": -37.90909, "trunc_abs_frac": 37.90909, "mean_delta_full": -0.00269, "read_hit_rate": {"surface": 0.59339, "split": 0.64004, "merge": 0.14789}, "mean_gate": {"surface": 0.12603, "split": 0.05421, "merge": 0.12913}, "n_reads": 12085, "n_cand": 24360, "read_suppressed_frac": 0.0991, "dropout_reads_logged": 12085, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2594, "admitted_surface": 29956, "staged_surface": 62727, "admitted_split": 31397, "staged_split": 50202, "admitted_merge": 4183, "staged_merge": 149215, "admitted_skel": 0, "staged_skel": 0, "mean_count": 554.6531372070312, "n_util_obs": 264, "util_count_spearman": -0.21597418230964135}, "probe_overhead_sec_total": 473.838, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.005561, "alpha": 0.03, "t": 5459.7} |
| {"event": "consolidate", "step": 3950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262120, "t": 5518.3} |
| {"event": "train", "step": 3950, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.49546, "bpb": 2.15749, "chunk_loss": 0.030398, "F": 0.19809, "G": 0.2071, "chunks": 202.84, "bytes_per_chunk": 5.048, "boundary_entropy": 0.11302, "router_noise": 0.0, "grad_norm": 0.8002, "lr": 0.0001074, "router_trainable": true, "sec_per_step": 1.187, "probe_sec": 0.0849, "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": 9, "mean_delta_loo": 0.02889, "n_shadow": 3, "mean_delta_shadow": -0.02561, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.276, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -0.26385, "trunc_abs_frac": 0.40368, "mean_delta_full": -0.18604, "read_hit_rate": {"surface": 0.58136, "split": 0.61442, "merge": 0.1503}, "mean_gate": {"surface": 0.13175, "split": 0.05658, "merge": 0.13505}, "n_reads": 11643, "n_cand": 24152, "read_suppressed_frac": 0.105, "dropout_reads_logged": 11643, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2631, "admitted_surface": 29878, "staged_surface": 62482, "admitted_split": 31472, "staged_split": 50695, "admitted_merge": 4186, "staged_merge": 148967, "admitted_skel": 0, "staged_skel": 0, "mean_count": 549.5555419921875, "n_util_obs": 249, "util_count_spearman": -0.2762666148464827}, "probe_overhead_sec_total": 479.547, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.005253, "alpha": 0.03, "t": 5519.1} |
| {"event": "consolidate", "step": 4000, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 1024, "uncertain_staged": 262120, "t": 5571.5} |
| {"event": "probe", "step": 4000, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3446.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.198012, "cps": 26.921875, "mean_p": 0.20853, "boundary_entropy": 0.122314, "bpb": 1.943638}, "tr": {"chunks": 3932.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.217202, "cps": 30.71875, "mean_p": 0.204375, "boundary_entropy": 0.112863, "bpb": 2.081324}, "fi": {"chunks": 4439.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.235979, "cps": 34.679688, "mean_p": 0.233597, "boundary_entropy": 0.122056, "bpb": 2.206203}}, "bpb": {"en": 1.94364, "tr": 2.08132, "fi": 2.2062}, "gini_cpc": 0.038869, "gini_cps": 0.056021, "bpic": 4.80088, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0969120434821458, "fi": 1.1917416252261976}, "inventory": {"size": 326654, "admitted": 65536, "staged": 261118, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2637, "admitted_surface": 30147, "staged_surface": 61828, "admitted_split": 31377, "staged_split": 50049, "admitted_merge": 4012, "staged_merge": 149241, "admitted_skel": 0, "staged_skel": 0, "mean_count": 554.2196655273438, "n_util_obs": 250, "util_count_spearman": -0.27214937839005426}, "t": 5572.7} |
| {"event": "eval", "step": 4000, "phase": 2, "bpb": {"en": 1.95806, "tr": 2.08965, "fi": 2.21058}, "bytes_per_chunk": {"en": 5.15, "tr": 5.184, "fi": 4.664}, "bpb_mean": 2.0861, "frozen_inventory": true, "t": 5574.8} |
| {"event": "train", "step": 4000, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.41533, "bpb": 2.04189, "chunk_loss": 0.030373, "F": 0.19873, "G": 0.20821, "chunks": 203.5, "bytes_per_chunk": 5.032, "boundary_entropy": 0.11697, "router_noise": 0.0, "grad_norm": 0.6925, "lr": 0.000104, "router_trainable": true, "sec_per_step": 1.2, "probe_sec": 0.0808, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04929, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.0658, "n_shadow": 4, "mean_delta_shadow": -0.10275, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.01944, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 1.01416, "trunc_abs_frac": 4.18507, "mean_delta_full": -0.07974, "read_hit_rate": {"surface": 0.59645, "split": 0.6274, "merge": 0.1478}, "mean_gate": {"surface": 0.13261, "split": 0.05612, "merge": 0.13262}, "n_reads": 11939, "n_cand": 24278, "read_suppressed_frac": 0.0977, "dropout_reads_logged": 11939, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2636, "admitted_surface": 30147, "staged_surface": 62034, "admitted_split": 31377, "staged_split": 50179, "admitted_merge": 4012, "staged_merge": 149931, "admitted_skel": 0, "staged_skel": 0, "mean_count": 554.4019165039062, "n_util_obs": 250, "util_count_spearman": -0.27214937839005426}, "probe_overhead_sec_total": 485.928, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.005771, "alpha": 0.03, "t": 5579.1} |
| {"event": "consolidate", "step": 4050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262119, "t": 5612.7} |
| {"event": "train", "step": 4050, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.42295, "bpb": 2.05288, "chunk_loss": 0.030387, "F": 0.19968, "G": 0.20939, "chunks": 204.47, "bytes_per_chunk": 5.008, "boundary_entropy": 0.10939, "router_noise": 0.0, "grad_norm": 0.7171, "lr": 0.0001007, "router_trainable": true, "sec_per_step": 0.681, "probe_sec": 0.0675, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04551, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": -0.01761, "n_shadow": 2, "mean_delta_shadow": -0.21648, "n_split": 4, "mean_delta_split": 0.0683, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 0, "read_hit_rate": {"surface": 0.59376, "split": 0.64367, "merge": 0.15296}, "mean_gate": {"surface": 0.13839, "split": 0.05923, "merge": 0.13678}, "n_reads": 12070, "n_cand": 24196, "read_suppressed_frac": 0.0952, "dropout_reads_logged": 12070, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2667, "admitted_surface": 30195, "staged_surface": 61914, "admitted_split": 31320, "staged_split": 51015, "admitted_merge": 4021, "staged_merge": 149215, "admitted_skel": 0, "staged_skel": 0, "mean_count": 562.0087890625, "n_util_obs": 250, "util_count_spearman": -0.2957565401046417}, "probe_overhead_sec_total": 490.144, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.005944, "alpha": 0.03, "t": 5613.1} |
| {"event": "probe", "step": 4096, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3467.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.199219, "cps": 27.085938, "mean_p": 0.210961, "boundary_entropy": 0.119618, "bpb": 1.936921}, "tr": {"chunks": 3917.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.216373, "cps": 30.601562, "mean_p": 0.20306, "boundary_entropy": 0.109079, "bpb": 2.074629}, "fi": {"chunks": 4458.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.236989, "cps": 34.828125, "mean_p": 0.23487, "boundary_entropy": 0.120479, "bpb": 2.198448}}, "bpb": {"en": 1.93692, "tr": 2.07463, "fi": 2.19845}, "gini_cpc": 0.038586, "gini_cps": 0.05579, "bpic": 4.790745, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.086108715374022, "fi": 1.1895931623631772}, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2693, "admitted_surface": 30195, "staged_surface": 62365, "admitted_split": 31320, "staged_split": 51240, "admitted_merge": 4021, "staged_merge": 148539, "admitted_skel": 0, "staged_skel": 0, "mean_count": 570.33837890625, "n_util_obs": 253, "util_count_spearman": -0.24431577141635089}, "t": 5644.5} |
| {"event": "consolidate", "step": 4100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262120, "t": 5656.5} |
| {"event": "train", "step": 4100, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.36101, "bpb": 1.96352, "chunk_loss": 0.030457, "F": 0.20209, "G": 0.21195, "chunks": 206.94, "bytes_per_chunk": 4.948, "boundary_entropy": 0.11527, "router_noise": 0.0, "grad_norm": 0.7214, "lr": 9.75e-05, "router_trainable": true, "sec_per_step": 0.877, "probe_sec": 0.0637, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04367, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.03761, "n_shadow": 2, "mean_delta_shadow": 0.02109, "n_split": 0, "mean_delta_split": NaN, "n_merge": 3, "mean_delta_merge": -0.13391, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.14196, "trunc_abs_frac": 0.14196, "mean_delta_full": -0.36157, "read_hit_rate": {"surface": 0.61775, "split": 0.62104, "merge": 0.15066}, "mean_gate": {"surface": 0.13569, "split": 0.05718, "merge": 0.13556}, "n_reads": 12155, "n_cand": 24574, "read_suppressed_frac": 0.1001, "dropout_reads_logged": 12155, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2639, "admitted_surface": 30876, "staged_surface": 61853, "admitted_split": 30804, "staged_split": 51147, "admitted_merge": 3856, "staged_merge": 149144, "admitted_skel": 0, "staged_skel": 0, "mean_count": 558.6668701171875, "n_util_obs": 246, "util_count_spearman": -0.2729765192215086}, "probe_overhead_sec_total": 494.737, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.005819, "alpha": 0.03, "t": 5657.0} |
| {"event": "consolidate", "step": 4150, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 1024, "uncertain_staged": 262120, "t": 5702.0} |
| {"event": "train", "step": 4150, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.4308, "bpb": 2.06421, "chunk_loss": 0.030512, "F": 0.20551, "G": 0.21457, "chunks": 210.44, "bytes_per_chunk": 4.866, "boundary_entropy": 0.11488, "router_noise": 0.0, "grad_norm": 0.7628, "lr": 9.43e-05, "router_trainable": true, "sec_per_step": 0.932, "probe_sec": 0.1095, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06681, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.0107, "n_shadow": 3, "mean_delta_shadow": -0.15053, "n_split": 2, "mean_delta_split": 0.3799, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 1.09323, "trunc_abs_frac": 0.62844, "mean_delta_full": 0.29602, "read_hit_rate": {"surface": 0.60668, "split": 0.59372, "merge": 0.14393}, "mean_gate": {"surface": 0.12978, "split": 0.05435, "merge": 0.12926}, "n_reads": 11969, "n_cand": 25066, "read_suppressed_frac": 0.0962, "dropout_reads_logged": 11969, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2627, "admitted_surface": 31488, "staged_surface": 61468, "admitted_split": 30293, "staged_split": 51942, "admitted_merge": 3755, "staged_merge": 148734, "admitted_skel": 0, "staged_skel": 0, "mean_count": 555.5972900390625, "n_util_obs": 236, "util_count_spearman": -0.2844463109459663}, "probe_overhead_sec_total": 499.409, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.005661, "alpha": 0.03, "t": 5703.6} |
| {"event": "consolidate", "step": 4200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262120, "t": 5739.2} |
| {"event": "probe", "step": 4200, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3473.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.199563, "cps": 27.132812, "mean_p": 0.210104, "boundary_entropy": 0.116285, "bpb": 1.934724}, "tr": {"chunks": 3873.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.213942, "cps": 30.257812, "mean_p": 0.20092, "boundary_entropy": 0.111472, "bpb": 2.066109}, "fi": {"chunks": 4431.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.235554, "cps": 34.617188, "mean_p": 0.233808, "boundary_entropy": 0.118666, "bpb": 2.195391}}, "bpb": {"en": 1.93472, "tr": 2.06611, "fi": 2.19539}, "gini_cpc": 0.036967, "gini_cps": 0.05423, "bpic": 4.817186, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.072053064112976, "fi": 1.180345649036792}, "inventory": {"size": 326656, "admitted": 65536, "staged": 261120, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2662, "admitted_surface": 32368, "staged_surface": 61061, "admitted_split": 29419, "staged_split": 51985, "admitted_merge": 3749, "staged_merge": 148074, "admitted_skel": 0, "staged_skel": 0, "mean_count": 563.7941284179688, "n_util_obs": 242, "util_count_spearman": -0.30411876847184277}, "t": 5740.5} |
| {"event": "eval", "step": 4200, "phase": 2, "bpb": {"en": 1.94817, "tr": 2.07545, "fi": 2.20147}, "bytes_per_chunk": {"en": 5.074, "tr": 5.223, "fi": 4.663}, "bpb_mean": 2.07503, "frozen_inventory": true, "t": 5742.9} |
| {"event": "phase_transition", "step": 4200, "phase": 3, "objective": "global", "router_trainable": true, "encoder_trainable": true, "frozen_router_params": 0, "t": 5751.1} |
| {"event": "train", "step": 4200, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35504, "bpb": 1.95491, "chunk_loss": 0.030283, "F": 0.19391, "G": 0.20296, "chunks": 198.56, "bytes_per_chunk": 5.157, "boundary_entropy": 0.10823, "router_noise": 0.0, "grad_norm": 0.7192, "lr": 9.12e-05, "router_trainable": true, "sec_per_step": 0.97, "probe_sec": 0.1132, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08738, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.0026, "n_shadow": 6, "mean_delta_shadow": 0.01636, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.02917, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.05951, "trunc_abs_frac": 0.05951, "mean_delta_full": 0.18872, "read_hit_rate": {"surface": 0.60187, "split": 0.58409, "merge": 0.14897}, "mean_gate": {"surface": 0.13852, "split": 0.05776, "merge": 0.13955}, "n_reads": 11174, "n_cand": 23624, "read_suppressed_frac": 0.0923, "dropout_reads_logged": 11174, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2662, "admitted_surface": 32368, "staged_surface": 61313, "admitted_split": 29419, "staged_split": 52176, "admitted_merge": 3749, "staged_merge": 148655, "admitted_skel": 0, "staged_skel": 0, "mean_count": 563.9646606445312, "n_util_obs": 242, "util_count_spearman": -0.3036064996331562}, "probe_overhead_sec_total": 504.105, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.005183, "alpha": 0.03, "t": 5752.1} |
| {"event": "consolidate", "step": 4250, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262118, "t": 5793.8} |
| {"event": "train", "step": 4250, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38502, "bpb": 1.99816, "chunk_loss": 0.030397, "F": 0.20081, "G": 0.21063, "chunks": 205.62, "bytes_per_chunk": 4.98, "boundary_entropy": 0.11046, "router_noise": 0.0, "grad_norm": 0.6912, "lr": 8.81e-05, "router_trainable": true, "sec_per_step": 0.852, "probe_sec": 0.1133, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08572, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 8, "mean_delta_loo": 0.0342, "n_shadow": 5, "mean_delta_shadow": -0.06248, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.17657, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 5.319, "trunc_abs_frac": 0.9056, "mean_delta_full": -0.01135, "read_hit_rate": {"surface": 0.60522, "split": 0.56977, "merge": 0.14764}, "mean_gate": {"surface": 0.13541, "split": 0.05506, "merge": 0.13524}, "n_reads": 11450, "n_cand": 24516, "read_suppressed_frac": 0.0997, "dropout_reads_logged": 11450, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2685, "admitted_surface": 33321, "staged_surface": 60723, "admitted_split": 28530, "staged_split": 52389, "admitted_merge": 3685, "staged_merge": 149032, "admitted_skel": 0, "staged_skel": 0, "mean_count": 562.88134765625, "n_util_obs": 244, "util_count_spearman": -0.298905038639621}, "probe_overhead_sec_total": 509.045, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.005488, "alpha": 0.03, "t": 5794.7} |
| {"event": "consolidate", "step": 4300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262117, "t": 5829.4} |
| {"event": "train", "step": 4300, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38914, "bpb": 2.0041, "chunk_loss": 0.030301, "F": 0.19629, "G": 0.20467, "chunks": 201.0, "bytes_per_chunk": 5.095, "boundary_entropy": 0.11275, "router_noise": 0.0, "grad_norm": 0.7334, "lr": 8.5e-05, "router_trainable": true, "sec_per_step": 0.705, "probe_sec": 0.0966, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07568, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.03633, "n_shadow": 4, "mean_delta_shadow": 0.04874, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": -0.01318, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 10.7057, "trunc_abs_frac": 0.59121, "mean_delta_full": 0.00386, "read_hit_rate": {"surface": 0.60688, "split": 0.56325, "merge": 0.14541}, "mean_gate": {"surface": 0.13331, "split": 0.05471, "merge": 0.13091}, "n_reads": 11061, "n_cand": 23866, "read_suppressed_frac": 0.095, "dropout_reads_logged": 11061, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2718, "admitted_surface": 34263, "staged_surface": 60329, "admitted_split": 27573, "staged_split": 53227, "admitted_merge": 3700, "staged_merge": 148588, "admitted_skel": 0, "staged_skel": 0, "mean_count": 570.1074829101562, "n_util_obs": 246, "util_count_spearman": -0.3093924121599449}, "probe_overhead_sec_total": 513.401, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.005872, "alpha": 0.03, "t": 5829.9} |
| {"event": "consolidate", "step": 4350, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 1024, "uncertain_staged": 262117, "t": 5872.0} |
| {"event": "train", "step": 4350, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.41387, "bpb": 2.03979, "chunk_loss": 0.030363, "F": 0.19833, "G": 0.2077, "chunks": 203.09, "bytes_per_chunk": 5.042, "boundary_entropy": 0.11252, "router_noise": 0.0, "grad_norm": 0.6978, "lr": 8.21e-05, "router_trainable": true, "sec_per_step": 0.857, "probe_sec": 0.076, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04844, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": -0.00104, "n_shadow": 7, "mean_delta_shadow": -0.05944, "n_split": 1, "mean_delta_split": -0.24347, "n_merge": 1, "mean_delta_merge": 0.302, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.80028, "trunc_abs_frac": 0.74758, "mean_delta_full": 0.05244, "read_hit_rate": {"surface": 0.60987, "split": 0.58207, "merge": 0.13582}, "mean_gate": {"surface": 0.1302, "split": 0.05293, "merge": 0.12845}, "n_reads": 11414, "n_cand": 24234, "read_suppressed_frac": 0.0953, "dropout_reads_logged": 11414, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2713, "admitted_surface": 35010, "staged_surface": 60412, "admitted_split": 27026, "staged_split": 53192, "admitted_merge": 3500, "staged_merge": 148540, "admitted_skel": 0, "staged_skel": 0, "mean_count": 575.5140380859375, "n_util_obs": 251, "util_count_spearman": -0.3155021817491937}, "probe_overhead_sec_total": 518.182, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006709, "alpha": 0.03, "t": 5872.8} |
| {"event": "consolidate", "step": 4400, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262115, "t": 5926.0} |
| {"event": "probe", "step": 4400, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3459.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.198759, "cps": 27.023438, "mean_p": 0.210203, "boundary_entropy": 0.116201, "bpb": 1.925292}, "tr": {"chunks": 3915.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.216262, "cps": 30.585938, "mean_p": 0.203023, "boundary_entropy": 0.106167, "bpb": 2.052324}, "fi": {"chunks": 4483.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.238318, "cps": 35.023438, "mean_p": 0.234432, "boundary_entropy": 0.117917, "bpb": 2.175507}}, "bpb": {"en": 1.92529, "tr": 2.05232, "fi": 2.17551}, "gini_cpc": 0.040366, "gini_cps": 0.057575, "bpic": 4.784684, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0880648312951824, "fi": 1.199031005598531}, "inventory": {"size": 326655, "admitted": 65536, "staged": 261119, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2738, "admitted_surface": 35346, "staged_surface": 59354, "admitted_split": 26688, "staged_split": 53184, "admitted_merge": 3502, "staged_merge": 148581, "admitted_skel": 0, "staged_skel": 0, "mean_count": 586.849853515625, "n_util_obs": 257, "util_count_spearman": -0.28156812731879466}, "t": 5927.6} |
| {"event": "eval", "step": 4400, "phase": 3, "bpb": {"en": 1.93939, "tr": 2.06248, "fi": 2.18083}, "bytes_per_chunk": {"en": 5.092, "tr": 5.166, "fi": 4.619}, "bpb_mean": 2.0609, "frozen_inventory": true, "t": 5930.1} |
| {"event": "train", "step": 4400, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.36606, "bpb": 1.97082, "chunk_loss": 0.030369, "F": 0.19934, "G": 0.21001, "chunks": 204.12, "bytes_per_chunk": 5.017, "boundary_entropy": 0.11161, "router_noise": 0.0, "grad_norm": 0.6576, "lr": 7.92e-05, "router_trainable": true, "sec_per_step": 1.278, "probe_sec": 0.1166, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07103, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.05139, "n_shadow": 4, "mean_delta_shadow": 0.04057, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 1.23851, "trunc_abs_frac": 19.65438, "mean_delta_full": -0.01506, "read_hit_rate": {"surface": 0.61246, "split": 0.60048, "merge": 0.15337}, "mean_gate": {"surface": 0.13306, "split": 0.05371, "merge": 0.13331}, "n_reads": 11780, "n_cand": 24304, "read_suppressed_frac": 0.0961, "dropout_reads_logged": 11780, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2738, "admitted_surface": 35346, "staged_surface": 59676, "admitted_split": 26688, "staged_split": 53189, "admitted_merge": 3502, "staged_merge": 149279, "admitted_skel": 0, "staged_skel": 0, "mean_count": 587.0296630859375, "n_util_obs": 257, "util_count_spearman": -0.28173850217174917}, "probe_overhead_sec_total": 523.517, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.005632, "alpha": 0.03, "t": 5936.7} |
| {"event": "consolidate", "step": 4450, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262120, "t": 5986.0} |
| {"event": "train", "step": 4450, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.36224, "bpb": 1.9653, "chunk_loss": 0.030275, "F": 0.19257, "G": 0.2013, "chunks": 197.19, "bytes_per_chunk": 5.193, "boundary_entropy": 0.10728, "router_noise": 0.0, "grad_norm": 0.7071, "lr": 7.63e-05, "router_trainable": true, "sec_per_step": 1.012, "probe_sec": 0.0947, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05022, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 3, "mean_delta_loo": -0.04486, "n_shadow": 7, "mean_delta_shadow": 0.00789, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": 0.04145, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.43915, "trunc_abs_frac": 4.93751, "mean_delta_full": 0.06226, "read_hit_rate": {"surface": 0.61198, "split": 0.59675, "merge": 0.14617}, "mean_gate": {"surface": 0.13662, "split": 0.05474, "merge": 0.13327}, "n_reads": 11256, "n_cand": 23418, "read_suppressed_frac": 0.0947, "dropout_reads_logged": 11256, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2771, "admitted_surface": 35170, "staged_surface": 59749, "admitted_split": 26862, "staged_split": 53384, "admitted_merge": 3504, "staged_merge": 149011, "admitted_skel": 0, "staged_skel": 0, "mean_count": 595.8592529296875, "n_util_obs": 261, "util_count_spearman": -0.2675467903159401}, "probe_overhead_sec_total": 528.104, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006086, "alpha": 0.03, "t": 5987.3} |
| {"event": "consolidate", "step": 4500, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 135, "uncertain_staged": 262120, "t": 6033.4} |
| {"event": "train", "step": 4500, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.33711, "bpb": 1.92904, "chunk_loss": 0.030341, "F": 0.19843, "G": 0.20726, "chunks": 203.19, "bytes_per_chunk": 5.04, "boundary_entropy": 0.10936, "router_noise": 0.0, "grad_norm": 0.6815, "lr": 7.36e-05, "router_trainable": true, "sec_per_step": 0.943, "probe_sec": 0.1139, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0601, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.04949, "n_shadow": 3, "mean_delta_shadow": 0.05997, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.02756, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.38248, "trunc_abs_frac": 2.06059, "mean_delta_full": 0.19897, "read_hit_rate": {"surface": 0.62519, "split": 0.6085, "merge": 0.13855}, "mean_gate": {"surface": 0.12902, "split": 0.05118, "merge": 0.12765}, "n_reads": 11824, "n_cand": 24226, "read_suppressed_frac": 0.1072, "dropout_reads_logged": 11824, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2800, "admitted_surface": 35136, "staged_surface": 59824, "admitted_split": 26896, "staged_split": 53311, "admitted_merge": 3504, "staged_merge": 149009, "admitted_skel": 0, "staged_skel": 0, "mean_count": 603.744140625, "n_util_obs": 267, "util_count_spearman": -0.28069965287190807}, "probe_overhead_sec_total": 532.634, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.005895, "alpha": 0.03, "t": 6034.4} |
| {"event": "consolidate", "step": 4550, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 70, "uncertain_staged": 262120, "t": 6076.1} |
| {"event": "train", "step": 4550, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.36739, "bpb": 1.97273, "chunk_loss": 0.030297, "F": 0.1965, "G": 0.20503, "chunks": 201.22, "bytes_per_chunk": 5.089, "boundary_entropy": 0.1055, "router_noise": 0.0, "grad_norm": 0.6849, "lr": 7.09e-05, "router_trainable": true, "sec_per_step": 0.844, "probe_sec": 0.1455, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.12592, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.00607, "n_shadow": 1, "mean_delta_shadow": -0.07056, "n_split": 2, "mean_delta_split": 0.08385, "n_merge": 2, "mean_delta_merge": 0.13248, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -2.68216, "trunc_abs_frac": 0.89206, "mean_delta_full": -0.05927, "read_hit_rate": {"surface": 0.6162, "split": 0.59517, "merge": 0.14183}, "mean_gate": {"surface": 0.13161, "split": 0.05208, "merge": 0.13049}, "n_reads": 11512, "n_cand": 23971, "read_suppressed_frac": 0.1002, "dropout_reads_logged": 11512, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2822, "admitted_surface": 35114, "staged_surface": 59887, "admitted_split": 26915, "staged_split": 53242, "admitted_merge": 3507, "staged_merge": 149015, "admitted_skel": 0, "staged_skel": 0, "mean_count": 612.2061767578125, "n_util_obs": 274, "util_count_spearman": -0.27110353458272446}, "probe_overhead_sec_total": 537.538, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.005989, "alpha": 0.03, "t": 6076.6} |
| {"event": "consolidate", "step": 4600, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 149, "uncertain_staged": 262120, "t": 6121.1} |
| {"event": "probe", "step": 4600, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3471.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.199448, "cps": 27.117188, "mean_p": 0.209836, "boundary_entropy": 0.114081, "bpb": 1.914885}, "tr": {"chunks": 3897.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.215268, "cps": 30.445312, "mean_p": 0.201737, "boundary_entropy": 0.105637, "bpb": 2.043477}, "fi": {"chunks": 4501.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.239275, "cps": 35.164062, "mean_p": 0.23553, "boundary_entropy": 0.114831, "bpb": 2.16487}}, "bpb": {"en": 1.91488, "tr": 2.04348, "fi": 2.16487}, "gini_cpc": 0.040598, "gini_cps": 0.057854, "bpic": 4.779847, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.079317853265557, "fi": 1.1996833625960501}, "inventory": {"size": 327530, "admitted": 65536, "staged": 261994, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2846, "admitted_surface": 35080, "staged_surface": 59956, "admitted_split": 26949, "staged_split": 53169, "admitted_merge": 3507, "staged_merge": 148869, "admitted_skel": 0, "staged_skel": 0, "mean_count": 621.3414306640625, "n_util_obs": 279, "util_count_spearman": -0.28623565933921236}, "t": 6122.5} |
| {"event": "eval", "step": 4600, "phase": 3, "bpb": {"en": 1.92939, "tr": 2.05271, "fi": 2.1691}, "bytes_per_chunk": {"en": 5.086, "tr": 5.193, "fi": 4.61}, "bpb_mean": 2.0504, "frozen_inventory": true, "t": 6124.8} |
| {"event": "train", "step": 4600, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.36665, "bpb": 1.97165, "chunk_loss": 0.030399, "F": 0.20294, "G": 0.21166, "chunks": 207.81, "bytes_per_chunk": 4.928, "boundary_entropy": 0.11, "router_noise": 0.0, "grad_norm": 0.6922, "lr": 6.82e-05, "router_trainable": true, "sec_per_step": 1.041, "probe_sec": 0.075, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05733, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.16274, "n_shadow": 8, "mean_delta_shadow": 0.1011, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 7, "trunc_ratio_sum": 0.93883, "trunc_abs_frac": 2.518, "mean_delta_full": -0.07448, "read_hit_rate": {"surface": 0.62345, "split": 0.61162, "merge": 0.15487}, "mean_gate": {"surface": 0.13047, "split": 0.052, "merge": 0.13143}, "n_reads": 12198, "n_cand": 24734, "read_suppressed_frac": 0.1016, "dropout_reads_logged": 12198, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2845, "admitted_surface": 35080, "staged_surface": 59983, "admitted_split": 26949, "staged_split": 53144, "admitted_merge": 3507, "staged_merge": 149017, "admitted_skel": 0, "staged_skel": 0, "mean_count": 621.5274658203125, "n_util_obs": 279, "util_count_spearman": -0.28657019085192675}, "probe_overhead_sec_total": 541.997, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.00582, "alpha": 0.03, "t": 6128.7} |
| {"event": "consolidate", "step": 4650, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 30, "uncertain_staged": 262121, "t": 6160.3} |
| {"event": "train", "step": 4650, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.39489, "bpb": 2.01241, "chunk_loss": 0.030484, "F": 0.20526, "G": 0.21292, "chunks": 210.19, "bytes_per_chunk": 4.872, "boundary_entropy": 0.10345, "router_noise": 0.0, "grad_norm": 0.6666, "lr": 6.57e-05, "router_trainable": true, "sec_per_step": 0.64, "probe_sec": 0.0487, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03315, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": -0.03725, "n_shadow": 4, "mean_delta_shadow": 0.08929, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 7, "trunc_ratio_sum": 0.51369, "trunc_abs_frac": 0.52878, "mean_delta_full": -0.12706, "read_hit_rate": {"surface": 0.62907, "split": 0.60921, "merge": 0.1486}, "mean_gate": {"surface": 0.13491, "split": 0.05406, "merge": 0.13217}, "n_reads": 12202, "n_cand": 24848, "read_suppressed_frac": 0.0998, "dropout_reads_logged": 12202, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2895, "admitted_surface": 35072, "staged_surface": 60243, "admitted_split": 26957, "staged_split": 53440, "admitted_merge": 3507, "staged_merge": 148461, "admitted_skel": 0, "staged_skel": 0, "mean_count": 629.72900390625, "n_util_obs": 283, "util_count_spearman": -0.26054008435244197}, "probe_overhead_sec_total": 545.612, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.00538, "alpha": 0.03, "t": 6160.7} |
| {"event": "consolidate", "step": 4700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 31, "uncertain_staged": 262121, "t": 6192.1} |
| {"event": "train", "step": 4700, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.42154, "bpb": 2.05085, "chunk_loss": 0.030684, "F": 0.21234, "G": 0.22003, "chunks": 217.44, "bytes_per_chunk": 4.709, "boundary_entropy": 0.10854, "router_noise": 0.0, "grad_norm": 0.6729, "lr": 6.32e-05, "router_trainable": true, "sec_per_step": 0.641, "probe_sec": 0.077, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05169, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 10, "mean_delta_loo": 0.00225, "n_shadow": 2, "mean_delta_shadow": 0.10858, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.42904, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -0.07626, "trunc_abs_frac": 1.98617, "mean_delta_full": 0.14496, "read_hit_rate": {"surface": 0.64814, "split": 0.62625, "merge": 0.1555}, "mean_gate": {"surface": 0.12907, "split": 0.05088, "merge": 0.12804}, "n_reads": 12957, "n_cand": 25630, "read_suppressed_frac": 0.1023, "dropout_reads_logged": 12957, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2919, "admitted_surface": 35065, "staged_surface": 60261, "admitted_split": 26964, "staged_split": 53263, "admitted_merge": 3507, "staged_merge": 148620, "admitted_skel": 0, "staged_skel": 0, "mean_count": 638.7697143554688, "n_util_obs": 288, "util_count_spearman": -0.24916760104328675}, "probe_overhead_sec_total": 549.272, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.00605, "alpha": 0.03, "t": 6192.7} |
| {"event": "consolidate", "step": 4750, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 38, "uncertain_staged": 262121, "t": 6223.0} |
| {"event": "train", "step": 4750, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.41216, "bpb": 2.03732, "chunk_loss": 0.030349, "F": 0.1994, "G": 0.20831, "chunks": 204.19, "bytes_per_chunk": 5.015, "boundary_entropy": 0.10427, "router_noise": 0.0, "grad_norm": 0.6753, "lr": 6.08e-05, "router_trainable": true, "sec_per_step": 0.612, "probe_sec": 0.07, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04738, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": -0.04532, "n_shadow": 8, "mean_delta_shadow": -0.03054, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.36847, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": -0.00651, "trunc_abs_frac": 0.00651, "mean_delta_full": 0.97095, "read_hit_rate": {"surface": 0.61227, "split": 0.61008, "merge": 0.14343}, "mean_gate": {"surface": 0.13113, "split": 0.05117, "merge": 0.12835}, "n_reads": 11831, "n_cand": 24318, "read_suppressed_frac": 0.1018, "dropout_reads_logged": 11831, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2922, "admitted_surface": 35061, "staged_surface": 60334, "admitted_split": 26968, "staged_split": 53443, "admitted_merge": 3507, "staged_merge": 148367, "admitted_skel": 0, "staged_skel": 0, "mean_count": 646.9087524414062, "n_util_obs": 293, "util_count_spearman": -0.23536845660701047}, "probe_overhead_sec_total": 552.788, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.005521, "alpha": 0.03, "t": 6223.3} |
| {"event": "consolidate", "step": 4800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 65, "uncertain_staged": 262122, "t": 6249.3} |
| {"event": "probe", "step": 4800, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3445.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.197954, "cps": 26.914062, "mean_p": 0.20866, "boundary_entropy": 0.114179, "bpb": 1.907314}, "tr": {"chunks": 3863.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.21339, "cps": 30.179688, "mean_p": 0.20096, "boundary_entropy": 0.105522, "bpb": 2.030542}, "fi": {"chunks": 4462.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.237202, "cps": 34.859375, "mean_p": 0.233131, "boundary_entropy": 0.11208, "bpb": 2.156167}}, "bpb": {"en": 1.90731, "tr": 2.03054, "fi": 2.15617}, "gini_cpc": 0.040344, "gini_cps": 0.057604, "bpic": 4.820051, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0779758977956087, "fi": 1.198264178410881}, "inventory": {"size": 327615, "admitted": 65536, "staged": 262079, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2957, "admitted_surface": 35046, "staged_surface": 60380, "admitted_split": 26982, "staged_split": 52971, "admitted_merge": 3508, "staged_merge": 148728, "admitted_skel": 0, "staged_skel": 0, "mean_count": 655.5662231445312, "n_util_obs": 299, "util_count_spearman": -0.22638282265076376}, "t": 6250.5} |
| {"event": "eval", "step": 4800, "phase": 3, "bpb": {"en": 1.92192, "tr": 2.04076, "fi": 2.15886}, "bytes_per_chunk": {"en": 5.127, "tr": 5.249, "fi": 4.627}, "bpb_mean": 2.04051, "frozen_inventory": true, "t": 6252.2} |
| {"event": "train", "step": 4800, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38369, "bpb": 1.99624, "chunk_loss": 0.030384, "F": 0.19757, "G": 0.20775, "chunks": 202.31, "bytes_per_chunk": 5.061, "boundary_entropy": 0.10975, "router_noise": 0.0, "grad_norm": 0.6857, "lr": 5.85e-05, "router_trainable": true, "sec_per_step": 0.651, "probe_sec": 0.0532, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03453, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": 0.03254, "n_shadow": 3, "mean_delta_shadow": -0.01878, "n_split": 2, "mean_delta_split": -0.0217, "n_merge": 2, "mean_delta_merge": 0.26794, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.80236, "trunc_abs_frac": 0.82425, "mean_delta_full": 0.11385, "read_hit_rate": {"surface": 0.62682, "split": 0.61399, "merge": 0.13853}, "mean_gate": {"surface": 0.12399, "split": 0.04836, "merge": 0.12534}, "n_reads": 11772, "n_cand": 24002, "read_suppressed_frac": 0.1013, "dropout_reads_logged": 11772, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2959, "admitted_surface": 35046, "staged_surface": 60373, "admitted_split": 26982, "staged_split": 52972, "admitted_merge": 3508, "staged_merge": 148799, "admitted_skel": 0, "staged_skel": 0, "mean_count": 655.745849609375, "n_util_obs": 300, "util_count_spearman": -0.21912201459149844}, "probe_overhead_sec_total": 555.885, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006161, "alpha": 0.03, "t": 6255.8} |
| {"event": "consolidate", "step": 4850, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 57, "uncertain_staged": 262122, "t": 6283.5} |
| {"event": "train", "step": 4850, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35598, "bpb": 1.95627, "chunk_loss": 0.030396, "F": 0.1998, "G": 0.20799, "chunks": 204.59, "bytes_per_chunk": 5.005, "boundary_entropy": 0.10275, "router_noise": 0.0, "grad_norm": 0.7337, "lr": 5.62e-05, "router_trainable": true, "sec_per_step": 0.567, "probe_sec": 0.0598, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03538, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": -0.03253, "n_shadow": 4, "mean_delta_shadow": -0.03, "n_split": 0, "mean_delta_split": NaN, "n_merge": 2, "mean_delta_merge": -0.09543, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.34629, "trunc_abs_frac": 1.1538, "mean_delta_full": -0.27737, "read_hit_rate": {"surface": 0.63147, "split": 0.61584, "merge": 0.15579}, "mean_gate": {"surface": 0.13331, "split": 0.05276, "merge": 0.12987}, "n_reads": 12035, "n_cand": 24220, "read_suppressed_frac": 0.1043, "dropout_reads_logged": 12035, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2967, "admitted_surface": 35017, "staged_surface": 60278, "admitted_split": 27008, "staged_split": 53014, "admitted_merge": 3511, "staged_merge": 148852, "admitted_skel": 0, "staged_skel": 0, "mean_count": 663.56298828125, "n_util_obs": 303, "util_count_spearman": -0.22954793300870693}, "probe_overhead_sec_total": 559.396, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.005845, "alpha": 0.03, "t": 6284.2} |
| {"event": "consolidate", "step": 4900, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 735, "uncertain_staged": 262122, "t": 6312.7} |
| {"event": "train", "step": 4900, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38345, "bpb": 1.9959, "chunk_loss": 0.030377, "F": 0.19937, "G": 0.20616, "chunks": 204.16, "bytes_per_chunk": 5.016, "boundary_entropy": 0.10024, "router_noise": 0.0, "grad_norm": 0.6753, "lr": 5.41e-05, "router_trainable": true, "sec_per_step": 0.585, "probe_sec": 0.1385, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.11432, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": 0.0519, "n_shadow": 6, "mean_delta_shadow": 0.0641, "n_split": 1, "mean_delta_split": -0.27576, "n_merge": 1, "mean_delta_merge": 0.19421, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -0.37219, "trunc_abs_frac": 0.67926, "mean_delta_full": 0.04883, "read_hit_rate": {"surface": 0.62067, "split": 0.62478, "merge": 0.1365}, "mean_gate": {"surface": 0.12918, "split": 0.05038, "merge": 0.12697}, "n_reads": 11943, "n_cand": 24214, "read_suppressed_frac": 0.1037, "dropout_reads_logged": 11943, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 2999, "admitted_surface": 34330, "staged_surface": 60426, "admitted_split": 27694, "staged_split": 52767, "admitted_merge": 3512, "staged_merge": 148951, "admitted_skel": 0, "staged_skel": 0, "mean_count": 672.8995361328125, "n_util_obs": 310, "util_count_spearman": -0.23771032873461548}, "probe_overhead_sec_total": 562.832, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006318, "alpha": 0.03, "t": 6313.5} |
| {"event": "consolidate", "step": 4950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1012, "uncertain_staged": 262123, "t": 6345.9} |
| {"event": "train", "step": 4950, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.34609, "bpb": 1.942, "chunk_loss": 0.030335, "F": 0.19595, "G": 0.20437, "chunks": 200.66, "bytes_per_chunk": 5.103, "boundary_entropy": 0.10274, "router_noise": 0.0, "grad_norm": 0.6777, "lr": 5.2e-05, "router_trainable": true, "sec_per_step": 0.656, "probe_sec": 0.1103, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09072, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.06658, "n_shadow": 3, "mean_delta_shadow": 0.0139, "n_split": 1, "mean_delta_split": 0.03197, "n_merge": 1, "mean_delta_merge": 0.01758, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.43914, "trunc_abs_frac": 11.65541, "mean_delta_full": 0.11808, "read_hit_rate": {"surface": 0.61512, "split": 0.63, "merge": 0.14976}, "mean_gate": {"surface": 0.13281, "split": 0.05139, "merge": 0.1274}, "n_reads": 11880, "n_cand": 23854, "read_suppressed_frac": 0.1013, "dropout_reads_logged": 11880, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3030, "admitted_surface": 33356, "staged_surface": 60533, "admitted_split": 28668, "staged_split": 51950, "admitted_merge": 3512, "staged_merge": 149661, "admitted_skel": 0, "staged_skel": 0, "mean_count": 681.25, "n_util_obs": 312, "util_count_spearman": -0.2317720289634038}, "probe_overhead_sec_total": 566.46, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006314, "alpha": 0.03, "t": 6346.3} |
| {"event": "consolidate", "step": 5000, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 860, "uncertain_staged": 262123, "t": 6375.4} |
| {"event": "probe", "step": 5000, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3486.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.20031, "cps": 27.234375, "mean_p": 0.209843, "boundary_entropy": 0.109953, "bpb": 1.899511}, "tr": {"chunks": 3884.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.21455, "cps": 30.34375, "mean_p": 0.201383, "boundary_entropy": 0.104277, "bpb": 2.023399}, "fi": {"chunks": 4449.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.236511, "cps": 34.757812, "mean_p": 0.23316, "boundary_entropy": 0.111318, "bpb": 2.14117}}, "bpb": {"en": 1.89951, "tr": 2.0234, "fi": 2.14117}, "gini_cpc": 0.03705, "gini_cps": 0.054319, "bpic": 4.800068, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.071088625300834, "fi": 1.1807209243575303}, "inventory": {"size": 326819, "admitted": 65536, "staged": 261283, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3047, "admitted_surface": 32573, "staged_surface": 60479, "admitted_split": 29451, "staged_split": 50821, "admitted_merge": 3512, "staged_merge": 149983, "admitted_skel": 0, "staged_skel": 0, "mean_count": 688.7015991210938, "n_util_obs": 320, "util_count_spearman": -0.24408767212884924}, "t": 6376.5} |
| {"event": "eval", "step": 5000, "phase": 3, "bpb": {"en": 1.91466, "tr": 2.03296, "fi": 2.14511}, "bytes_per_chunk": {"en": 5.092, "tr": 5.2, "fi": 4.636}, "bpb_mean": 2.03091, "frozen_inventory": true, "t": 6378.3} |
| {"event": "train", "step": 5000, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.33508, "bpb": 1.92611, "chunk_loss": 0.030354, "F": 0.19907, "G": 0.20775, "chunks": 203.84, "bytes_per_chunk": 5.023, "boundary_entropy": 0.10662, "router_noise": 0.0, "grad_norm": 0.646, "lr": 5e-05, "router_trainable": true, "sec_per_step": 0.717, "probe_sec": 0.0542, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03595, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": -0.02349, "n_shadow": 8, "mean_delta_shadow": -0.11387, "n_split": 1, "mean_delta_split": -0.09888, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 51.05263, "trunc_abs_frac": 0.44248, "mean_delta_full": -0.00116, "read_hit_rate": {"surface": 0.63719, "split": 0.63188, "merge": 0.14646}, "mean_gate": {"surface": 0.12399, "split": 0.04732, "merge": 0.12334}, "n_reads": 12197, "n_cand": 24210, "read_suppressed_frac": 0.1004, "dropout_reads_logged": 12197, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3047, "admitted_surface": 32573, "staged_surface": 60690, "admitted_split": 29451, "staged_split": 50954, "admitted_merge": 3512, "staged_merge": 150500, "admitted_skel": 0, "staged_skel": 0, "mean_count": 688.8876953125, "n_util_obs": 320, "util_count_spearman": -0.23801233410482525}, "probe_overhead_sec_total": 569.883, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.005933, "alpha": 0.03, "t": 6382.1} |
| {"event": "consolidate", "step": 5050, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 157, "uncertain_staged": 262123, "t": 6411.5} |
| {"event": "train", "step": 5050, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35589, "bpb": 1.95614, "chunk_loss": 0.030354, "F": 0.19833, "G": 0.20754, "chunks": 203.09, "bytes_per_chunk": 5.042, "boundary_entropy": 0.10368, "router_noise": 0.0, "grad_norm": 0.6905, "lr": 4.81e-05, "router_trainable": true, "sec_per_step": 0.602, "probe_sec": 0.0828, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06714, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": -0.0655, "n_shadow": 2, "mean_delta_shadow": 0.08774, "n_split": 4, "mean_delta_split": 0.0146, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 0, "read_hit_rate": {"surface": 0.62023, "split": 0.63765, "merge": 0.14468}, "mean_gate": {"surface": 0.12375, "split": 0.04787, "merge": 0.12255}, "n_reads": 12173, "n_cand": 24242, "read_suppressed_frac": 0.1029, "dropout_reads_logged": 12173, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3084, "admitted_surface": 32433, "staged_surface": 60630, "admitted_split": 29592, "staged_split": 51439, "admitted_merge": 3511, "staged_merge": 150075, "admitted_skel": 0, "staged_skel": 0, "mean_count": 697.5260009765625, "n_util_obs": 320, "util_count_spearman": -0.25020585837127585}, "probe_overhead_sec_total": 573.781, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006084, "alpha": 0.03, "t": 6412.2} |
| {"event": "consolidate", "step": 5100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 156, "uncertain_staged": 262123, "t": 6438.8} |
| {"event": "train", "step": 5100, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38337, "bpb": 1.99579, "chunk_loss": 0.030421, "F": 0.2023, "G": 0.21058, "chunks": 207.16, "bytes_per_chunk": 4.943, "boundary_entropy": 0.10173, "router_noise": 0.0, "grad_norm": 0.7202, "lr": 4.63e-05, "router_trainable": true, "sec_per_step": 0.546, "probe_sec": 0.0565, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03872, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.01687, "n_shadow": 4, "mean_delta_shadow": 0.04456, "n_split": 1, "mean_delta_split": -0.03564, "n_merge": 1, "mean_delta_merge": -0.08557, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -0.3239, "trunc_abs_frac": 0.76988, "mean_delta_full": 0.12954, "read_hit_rate": {"surface": 0.63059, "split": 0.64511, "merge": 0.14212}, "mean_gate": {"surface": 0.12627, "split": 0.04937, "merge": 0.12428}, "n_reads": 12384, "n_cand": 24464, "read_suppressed_frac": 0.0974, "dropout_reads_logged": 12384, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3117, "admitted_surface": 32300, "staged_surface": 60683, "admitted_split": 29725, "staged_split": 51058, "admitted_merge": 3511, "staged_merge": 150403, "admitted_skel": 0, "staged_skel": 0, "mean_count": 706.2293090820312, "n_util_obs": 322, "util_count_spearman": -0.23799725411216074}, "probe_overhead_sec_total": 577.804, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.00625, "alpha": 0.03, "t": 6439.6} |
| {"event": "consolidate", "step": 5150, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 279, "uncertain_staged": 262123, "t": 6470.1} |
| {"event": "train", "step": 5150, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.37404, "bpb": 1.98232, "chunk_loss": 0.030515, "F": 0.20679, "G": 0.21431, "chunks": 211.75, "bytes_per_chunk": 4.836, "boundary_entropy": 0.10181, "router_noise": 0.0, "grad_norm": 0.7078, "lr": 4.45e-05, "router_trainable": true, "sec_per_step": 0.642, "probe_sec": 0.0802, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03891, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": 0.02659, "n_shadow": 4, "mean_delta_shadow": 0.00847, "n_split": 2, "mean_delta_split": 0.04707, "n_merge": 1, "mean_delta_merge": 0.09351, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.32283, "trunc_abs_frac": 0.32954, "mean_delta_full": -0.21716, "read_hit_rate": {"surface": 0.63256, "split": 0.6472, "merge": 0.16701}, "mean_gate": {"surface": 0.12739, "split": 0.04967, "merge": 0.12699}, "n_reads": 12822, "n_cand": 24944, "read_suppressed_frac": 0.0987, "dropout_reads_logged": 12822, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3137, "admitted_surface": 32036, "staged_surface": 60448, "admitted_split": 29988, "staged_split": 51299, "admitted_merge": 3512, "staged_merge": 150397, "admitted_skel": 0, "staged_skel": 0, "mean_count": 714.5103759765625, "n_util_obs": 329, "util_count_spearman": -0.22246072557437374}, "probe_overhead_sec_total": 581.179, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006225, "alpha": 0.03, "t": 6471.6} |
| {"event": "consolidate", "step": 5200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 82, "uncertain_staged": 262123, "t": 6502.8} |
| {"event": "probe", "step": 5200, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3469.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.199333, "cps": 27.101562, "mean_p": 0.208951, "boundary_entropy": 0.107033, "bpb": 1.893362}, "tr": {"chunks": 3891.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.214937, "cps": 30.398438, "mean_p": 0.201381, "boundary_entropy": 0.104754, "bpb": 2.015822}, "fi": {"chunks": 4453.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.236723, "cps": 34.789062, "mean_p": 0.232902, "boundary_entropy": 0.111128, "bpb": 2.131712}}, "bpb": {"en": 1.89336, "tr": 2.01582, "fi": 2.13171}, "gini_cpc": 0.03829, "gini_cps": 0.055532, "bpic": 4.802506, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.078277392455939, "fi": 1.1875738665386852}, "inventory": {"size": 327598, "admitted": 65536, "staged": 262062, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3163, "admitted_surface": 31961, "staged_surface": 60606, "admitted_split": 30063, "staged_split": 51680, "admitted_merge": 3512, "staged_merge": 149776, "admitted_skel": 0, "staged_skel": 0, "mean_count": 723.7036743164062, "n_util_obs": 336, "util_count_spearman": -0.200598849752931}, "t": 6503.9} |
| {"event": "eval", "step": 5200, "phase": 3, "bpb": {"en": 1.90866, "tr": 2.02504, "fi": 2.13562}, "bytes_per_chunk": {"en": 5.088, "tr": 5.206, "fi": 4.645}, "bpb_mean": 2.02311, "frozen_inventory": true, "t": 6505.6} |
| {"event": "train", "step": 5200, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.33063, "bpb": 1.9197, "chunk_loss": 0.030242, "F": 0.19254, "G": 0.20103, "chunks": 197.16, "bytes_per_chunk": 5.194, "boundary_entropy": 0.10051, "router_noise": 0.0, "grad_norm": 0.6684, "lr": 4.29e-05, "router_trainable": true, "sec_per_step": 0.76, "probe_sec": 0.0777, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0451, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.03712, "n_shadow": 4, "mean_delta_shadow": 0.01217, "n_split": 1, "mean_delta_split": -0.11899, "n_merge": 1, "mean_delta_merge": 0.50647, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.65696, "trunc_abs_frac": 0.51606, "mean_delta_full": 0.28168, "read_hit_rate": {"surface": 0.62291, "split": 0.63512, "merge": 0.13563}, "mean_gate": {"surface": 0.12091, "split": 0.04664, "merge": 0.12255}, "n_reads": 11864, "n_cand": 23712, "read_suppressed_frac": 0.1, "dropout_reads_logged": 11864, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3165, "admitted_surface": 31961, "staged_surface": 60628, "admitted_split": 30063, "staged_split": 51686, "admitted_merge": 3512, "staged_merge": 149830, "admitted_skel": 0, "staged_skel": 0, "mean_count": 723.884765625, "n_util_obs": 337, "util_count_spearman": -0.19365811659237817}, "probe_overhead_sec_total": 585.017, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.00609, "alpha": 0.03, "t": 6509.6} |
| {"event": "consolidate", "step": 5250, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262123, "t": 6536.5} |
| {"event": "train", "step": 5250, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32496, "bpb": 1.91151, "chunk_loss": 0.030337, "F": 0.19791, "G": 0.20614, "chunks": 202.66, "bytes_per_chunk": 5.053, "boundary_entropy": 0.10032, "router_noise": 0.0, "grad_norm": 0.681, "lr": 4.14e-05, "router_trainable": true, "sec_per_step": 0.545, "probe_sec": 0.1158, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09536, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": 0.02453, "n_shadow": 4, "mean_delta_shadow": 0.13025, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.63336, "n_dropped_unpaired": 0, "n_full": 7, "trunc_ratio_sum": 0.97517, "trunc_abs_frac": 4.21627, "mean_delta_full": 0.04442, "read_hit_rate": {"surface": 0.6335, "split": 0.64879, "merge": 0.14562}, "mean_gate": {"surface": 0.12456, "split": 0.04721, "merge": 0.12633}, "n_reads": 12379, "n_cand": 24212, "read_suppressed_frac": 0.1012, "dropout_reads_logged": 12379, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3183, "admitted_surface": 30960, "staged_surface": 60763, "admitted_split": 31064, "staged_split": 50678, "admitted_merge": 3512, "staged_merge": 150703, "admitted_skel": 0, "staged_skel": 0, "mean_count": 731.1713256835938, "n_util_obs": 344, "util_count_spearman": -0.18157606717449332}, "probe_overhead_sec_total": 588.442, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.00707, "alpha": 0.03, "t": 6536.9} |
| {"event": "consolidate", "step": 5300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262122, "t": 6564.4} |
| {"event": "train", "step": 5300, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38033, "bpb": 1.9914, "chunk_loss": 0.03041, "F": 0.20285, "G": 0.20981, "chunks": 207.72, "bytes_per_chunk": 4.93, "boundary_entropy": 0.09954, "router_noise": 0.0, "grad_norm": 0.6829, "lr": 3.99e-05, "router_trainable": true, "sec_per_step": 0.559, "probe_sec": 0.0603, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03896, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.08416, "n_shadow": 5, "mean_delta_shadow": -0.00266, "n_split": 1, "mean_delta_split": 0.19751, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 0.47289, "trunc_abs_frac": 1.26486, "mean_delta_full": 0.16266, "read_hit_rate": {"surface": 0.62464, "split": 0.66073, "merge": 0.13216}, "mean_gate": {"surface": 0.12352, "split": 0.04687, "merge": 0.1184}, "n_reads": 12714, "n_cand": 24870, "read_suppressed_frac": 0.0969, "dropout_reads_logged": 12714, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3204, "admitted_surface": 29954, "staged_surface": 61025, "admitted_split": 32070, "staged_split": 50028, "admitted_merge": 3512, "staged_merge": 151091, "admitted_skel": 0, "staged_skel": 0, "mean_count": 739.6748046875, "n_util_obs": 345, "util_count_spearman": -0.21074771991529131}, "probe_overhead_sec_total": 591.958, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006409, "alpha": 0.03, "t": 6564.8} |
| {"event": "consolidate", "step": 5350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262122, "t": 6591.3} |
| {"event": "train", "step": 5350, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.31438, "bpb": 1.89625, "chunk_loss": 0.030275, "F": 0.19324, "G": 0.20175, "chunks": 197.88, "bytes_per_chunk": 5.175, "boundary_entropy": 0.09991, "router_noise": 0.0, "grad_norm": 0.7147, "lr": 3.86e-05, "router_trainable": true, "sec_per_step": 0.537, "probe_sec": 0.0554, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03469, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": -0.01762, "n_shadow": 4, "mean_delta_shadow": 0.03605, "n_split": 1, "mean_delta_split": -0.22461, "n_merge": 1, "mean_delta_merge": -0.45877, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 0.2963, "trunc_abs_frac": 0.2963, "mean_delta_full": -0.0824, "read_hit_rate": {"surface": 0.62889, "split": 0.65212, "merge": 0.13561}, "mean_gate": {"surface": 0.12093, "split": 0.04586, "merge": 0.11982}, "n_reads": 12074, "n_cand": 23704, "read_suppressed_frac": 0.0987, "dropout_reads_logged": 12074, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3223, "admitted_surface": 28965, "staged_surface": 61093, "admitted_split": 33059, "staged_split": 49731, "admitted_merge": 3512, "staged_merge": 151320, "admitted_skel": 0, "staged_skel": 0, "mean_count": 749.313720703125, "n_util_obs": 354, "util_count_spearman": -0.20932538139481727}, "probe_overhead_sec_total": 595.32, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006937, "alpha": 0.03, "t": 6591.7} |
| {"event": "consolidate", "step": 5400, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 1024, "uncertain_staged": 262123, "t": 6619.4} |
| {"event": "probe", "step": 5400, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3484.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.200195, "cps": 27.21875, "mean_p": 0.20935, "boundary_entropy": 0.107417, "bpb": 1.887495}, "tr": {"chunks": 3931.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.217146, "cps": 30.710938, "mean_p": 0.203699, "boundary_entropy": 0.099638, "bpb": 2.008851}, "fi": {"chunks": 4480.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.238159, "cps": 35.0, "mean_p": 0.234246, "boundary_entropy": 0.109379, "bpb": 2.124952}}, "bpb": {"en": 1.8875, "tr": 2.00885, "fi": 2.12495}, "gini_cpc": 0.03861, "gini_cps": 0.055822, "bpic": 4.769399, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0846720922685489, "fi": 1.1896305389870998}, "inventory": {"size": 326654, "admitted": 65536, "staged": 261118, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3233, "admitted_surface": 28081, "staged_surface": 60953, "admitted_split": 33950, "staged_split": 49211, "admitted_merge": 3505, "staged_merge": 150954, "admitted_skel": 0, "staged_skel": 0, "mean_count": 753.4554443359375, "n_util_obs": 350, "util_count_spearman": -0.2682464813123839}, "t": 6620.6} |
| {"event": "eval", "step": 5400, "phase": 3, "bpb": {"en": 1.90249, "tr": 2.0185, "fi": 2.13}, "bytes_per_chunk": {"en": 5.078, "tr": 5.163, "fi": 4.622}, "bpb_mean": 2.01699, "frozen_inventory": true, "t": 6622.6} |
| {"event": "train", "step": 5400, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.3874, "bpb": 2.00159, "chunk_loss": 0.030577, "F": 0.20709, "G": 0.21435, "chunks": 212.06, "bytes_per_chunk": 4.829, "boundary_entropy": 0.10223, "router_noise": 0.0, "grad_norm": 0.7342, "lr": 3.73e-05, "router_trainable": true, "sec_per_step": 0.69, "probe_sec": 0.0749, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05228, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.02869, "n_shadow": 2, "mean_delta_shadow": 0.07874, "n_split": 2, "mean_delta_split": -0.01524, "n_merge": 1, "mean_delta_merge": -0.00232, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.39973, "trunc_abs_frac": 2.40029, "mean_delta_full": -0.06909, "read_hit_rate": {"surface": 0.63607, "split": 0.66553, "merge": 0.15174}, "mean_gate": {"surface": 0.12026, "split": 0.04511, "merge": 0.12222}, "n_reads": 13096, "n_cand": 25166, "read_suppressed_frac": 0.0983, "dropout_reads_logged": 13096, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3235, "admitted_surface": 28081, "staged_surface": 61157, "admitted_split": 33950, "staged_split": 49396, "admitted_merge": 3505, "staged_merge": 151591, "admitted_skel": 0, "staged_skel": 0, "mean_count": 753.6553344726562, "n_util_obs": 350, "util_count_spearman": -0.2682328044519766}, "probe_overhead_sec_total": 598.954, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006079, "alpha": 0.03, "t": 6626.2} |
| {"event": "consolidate", "step": 5450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262122, "t": 6652.6} |
| {"event": "train", "step": 5450, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.31937, "bpb": 1.90345, "chunk_loss": 0.030493, "F": 0.20422, "G": 0.21232, "chunks": 209.12, "bytes_per_chunk": 4.897, "boundary_entropy": 0.09994, "router_noise": 0.0, "grad_norm": 0.7025, "lr": 3.62e-05, "router_trainable": true, "sec_per_step": 0.537, "probe_sec": 0.0814, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05993, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 3, "mean_delta_loo": 0.01379, "n_shadow": 5, "mean_delta_shadow": -0.00264, "n_split": 2, "mean_delta_split": -0.13489, "n_merge": 1, "mean_delta_merge": 0.0181, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -24.63775, "trunc_abs_frac": 0.57575, "mean_delta_full": 0.00399, "read_hit_rate": {"surface": 0.64309, "split": 0.65896, "merge": 0.16219}, "mean_gate": {"surface": 0.12696, "split": 0.04814, "merge": 0.12544}, "n_reads": 12940, "n_cand": 24794, "read_suppressed_frac": 0.0991, "dropout_reads_logged": 12940, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3263, "admitted_surface": 27187, "staged_surface": 61439, "admitted_split": 34842, "staged_split": 48827, "admitted_merge": 3507, "staged_merge": 151878, "admitted_skel": 0, "staged_skel": 0, "mean_count": 762.6207275390625, "n_util_obs": 356, "util_count_spearman": -0.2621032207787372}, "probe_overhead_sec_total": 602.335, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006538, "alpha": 0.03, "t": 6653.0} |
| {"event": "consolidate", "step": 5500, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262123, "t": 6680.4} |
| {"event": "train", "step": 5500, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.30854, "bpb": 1.88782, "chunk_loss": 0.030454, "F": 0.203, "G": 0.21122, "chunks": 207.88, "bytes_per_chunk": 4.926, "boundary_entropy": 0.10243, "router_noise": 0.0, "grad_norm": 0.6998, "lr": 3.51e-05, "router_trainable": true, "sec_per_step": 0.556, "probe_sec": 0.0567, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03882, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": -0.0015, "n_shadow": 5, "mean_delta_shadow": 0.03025, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": 0.32898, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.55661, "trunc_abs_frac": 0.22063, "mean_delta_full": 0.20504, "read_hit_rate": {"surface": 0.64698, "split": 0.6563, "merge": 0.15149}, "mean_gate": {"surface": 0.12128, "split": 0.04601, "merge": 0.1219}, "n_reads": 12818, "n_cand": 24692, "read_suppressed_frac": 0.101, "dropout_reads_logged": 12818, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3284, "admitted_surface": 26425, "staged_surface": 61210, "admitted_split": 35611, "staged_split": 48561, "admitted_merge": 3500, "staged_merge": 152373, "admitted_skel": 0, "staged_skel": 0, "mean_count": 770.1604614257812, "n_util_obs": 358, "util_count_spearman": -0.2716668292564301}, "probe_overhead_sec_total": 605.953, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006738, "alpha": 0.03, "t": 6680.8} |
| {"event": "consolidate", "step": 5550, "retired": 0, "split_retired": 2, "merge_admitted": 0, "admitted": 2, "swapped": 1024, "uncertain_staged": 262123, "t": 6706.6} |
| {"event": "train", "step": 5550, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.36347, "bpb": 1.96707, "chunk_loss": 0.030427, "F": 0.2019, "G": 0.20907, "chunks": 206.75, "bytes_per_chunk": 4.953, "boundary_entropy": 0.09899, "router_noise": 0.0, "grad_norm": 0.7583, "lr": 3.41e-05, "router_trainable": true, "sec_per_step": 0.531, "probe_sec": 0.0798, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05317, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.07841, "n_shadow": 2, "mean_delta_shadow": -0.03708, "n_split": 2, "mean_delta_split": -0.1554, "n_merge": 1, "mean_delta_merge": -0.12927, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 7.14233, "trunc_abs_frac": 0.15065, "mean_delta_full": -0.00616, "read_hit_rate": {"surface": 0.61741, "split": 0.65026, "merge": 0.15446}, "mean_gate": {"surface": 0.1258, "split": 0.04853, "merge": 0.12452}, "n_reads": 12464, "n_cand": 24496, "read_suppressed_frac": 0.1004, "dropout_reads_logged": 12464, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3307, "admitted_surface": 25684, "staged_surface": 61836, "admitted_split": 36364, "staged_split": 48526, "admitted_merge": 3488, "staged_merge": 151782, "admitted_skel": 0, "staged_skel": 0, "mean_count": 765.2186279296875, "n_util_obs": 342, "util_count_spearman": -0.30806235680637106}, "probe_overhead_sec_total": 609.768, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006527, "alpha": 0.03, "t": 6707.3} |
| {"event": "consolidate", "step": 5600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262123, "t": 6734.7} |
| {"event": "probe", "step": 5600, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3487.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.200368, "cps": 27.242188, "mean_p": 0.209844, "boundary_entropy": 0.106088, "bpb": 1.882745}, "tr": {"chunks": 3919.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.216483, "cps": 30.617188, "mean_p": 0.202695, "boundary_entropy": 0.100094, "bpb": 2.003218}, "fi": {"chunks": 4479.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.238105, "cps": 34.992188, "mean_p": 0.233929, "boundary_entropy": 0.107841, "bpb": 2.116125}}, "bpb": {"en": 1.88275, "tr": 2.00322, "fi": 2.11612}, "gini_cpc": 0.038412, "gini_cps": 0.055644, "bpic": 4.773412, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0804306225848674, "fi": 1.1883417400782819}, "inventory": {"size": 326656, "admitted": 65536, "staged": 261120, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3310, "admitted_surface": 25518, "staged_surface": 61385, "admitted_split": 36542, "staged_split": 48035, "admitted_merge": 3476, "staged_merge": 151700, "admitted_skel": 0, "staged_skel": 0, "mean_count": 760.2388916015625, "n_util_obs": 332, "util_count_spearman": -0.3227970791957092}, "t": 6735.8} |
| {"event": "eval", "step": 5600, "phase": 3, "bpb": {"en": 1.89852, "tr": 2.01212, "fi": 2.12198}, "bytes_per_chunk": {"en": 5.06, "tr": 5.187, "fi": 4.627}, "bpb_mean": 2.01087, "frozen_inventory": true, "t": 6737.6} |
| {"event": "train", "step": 5600, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35166, "bpb": 1.95003, "chunk_loss": 0.030383, "F": 0.20029, "G": 0.20649, "chunks": 205.09, "bytes_per_chunk": 4.993, "boundary_entropy": 0.09852, "router_noise": 0.0, "grad_norm": 0.7443, "lr": 3.33e-05, "router_trainable": true, "sec_per_step": 0.684, "probe_sec": 0.0611, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03398, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.00947, "n_shadow": 5, "mean_delta_shadow": 0.00065, "n_split": 1, "mean_delta_split": 0.21729, "n_merge": 1, "mean_delta_merge": 0.091, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.73293, "trunc_abs_frac": 0.4895, "mean_delta_full": 0.05005, "read_hit_rate": {"surface": 0.63114, "split": 0.65151, "merge": 0.13125}, "mean_gate": {"surface": 0.12249, "split": 0.04583, "merge": 0.12001}, "n_reads": 12526, "n_cand": 24620, "read_suppressed_frac": 0.1017, "dropout_reads_logged": 12526, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3312, "admitted_surface": 25518, "staged_surface": 61573, "admitted_split": 36542, "staged_split": 48274, "admitted_merge": 3476, "staged_merge": 152297, "admitted_skel": 0, "staged_skel": 0, "mean_count": 760.4300537109375, "n_util_obs": 332, "util_count_spearman": -0.322755105136006}, "probe_overhead_sec_total": 613.63, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.007198, "alpha": 0.03, "t": 6741.5} |
| {"event": "consolidate", "step": 5650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262123, "t": 6770.1} |
| {"event": "train", "step": 5650, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.33991, "bpb": 1.93308, "chunk_loss": 0.030402, "F": 0.20111, "G": 0.20901, "chunks": 205.94, "bytes_per_chunk": 4.972, "boundary_entropy": 0.1016, "router_noise": 0.0, "grad_norm": 0.7262, "lr": 3.25e-05, "router_trainable": true, "sec_per_step": 0.578, "probe_sec": 0.0508, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03522, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": -0.11626, "n_shadow": 8, "mean_delta_shadow": 0.05474, "n_split": 1, "mean_delta_split": -0.19345, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.63449, "trunc_abs_frac": 0.75695, "mean_delta_full": 0.0344, "read_hit_rate": {"surface": 0.63846, "split": 0.65879, "merge": 0.13837}, "mean_gate": {"surface": 0.1176, "split": 0.04427, "merge": 0.11548}, "n_reads": 12728, "n_cand": 24678, "read_suppressed_frac": 0.0962, "dropout_reads_logged": 12728, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3338, "admitted_surface": 25415, "staged_surface": 61725, "admitted_split": 36646, "staged_split": 48133, "admitted_merge": 3475, "staged_merge": 152286, "admitted_skel": 0, "staged_skel": 0, "mean_count": 766.9449462890625, "n_util_obs": 333, "util_count_spearman": -0.2920575563345442}, "probe_overhead_sec_total": 618.314, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006738, "alpha": 0.03, "t": 6770.4} |
| {"event": "consolidate", "step": 5700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262123, "t": 6797.2} |
| {"event": "train", "step": 5700, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.3495, "bpb": 1.94691, "chunk_loss": 0.030265, "F": 0.1944, "G": 0.20262, "chunks": 199.06, "bytes_per_chunk": 5.144, "boundary_entropy": 0.10139, "router_noise": 0.0, "grad_norm": 0.7544, "lr": 3.18e-05, "router_trainable": true, "sec_per_step": 0.544, "probe_sec": 0.0558, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03731, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": 0.02422, "n_shadow": 6, "mean_delta_shadow": -0.13334, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.11597, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 1.38508, "trunc_abs_frac": 1.41362, "mean_delta_full": -0.16278, "read_hit_rate": {"surface": 0.61676, "split": 0.63226, "merge": 0.1367}, "mean_gate": {"surface": 0.12142, "split": 0.04549, "merge": 0.11914}, "n_reads": 11779, "n_cand": 23728, "read_suppressed_frac": 0.101, "dropout_reads_logged": 11779, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3353, "admitted_surface": 25437, "staged_surface": 61618, "admitted_split": 36619, "staged_split": 48158, "admitted_merge": 3480, "staged_merge": 152368, "admitted_skel": 0, "staged_skel": 0, "mean_count": 763.0870361328125, "n_util_obs": 334, "util_count_spearman": -0.2902289008107222}, "probe_overhead_sec_total": 621.649, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006724, "alpha": 0.03, "t": 6797.6} |
| {"event": "consolidate", "step": 5750, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262123, "t": 6826.3} |
| {"event": "train", "step": 5750, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.39941, "bpb": 2.01892, "chunk_loss": 0.03045, "F": 0.20294, "G": 0.2106, "chunks": 207.81, "bytes_per_chunk": 4.928, "boundary_entropy": 0.09741, "router_noise": 0.0, "grad_norm": 0.7368, "lr": 3.13e-05, "router_trainable": true, "sec_per_step": 0.588, "probe_sec": 0.0737, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0407, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 7, "mean_delta_loo": 0.01235, "n_shadow": 6, "mean_delta_shadow": 0.0667, "n_split": 0, "mean_delta_split": NaN, "n_merge": 1, "mean_delta_merge": -0.28619, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -1.02208, "trunc_abs_frac": 0.54172, "mean_delta_full": -0.10447, "read_hit_rate": {"surface": 0.61831, "split": 0.6313, "merge": 0.11904}, "mean_gate": {"surface": 0.11983, "split": 0.04491, "merge": 0.12146}, "n_reads": 12165, "n_cand": 24750, "read_suppressed_frac": 0.1011, "dropout_reads_logged": 12165, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3361, "admitted_surface": 25306, "staged_surface": 61644, "admitted_split": 36792, "staged_split": 48309, "admitted_merge": 3438, "staged_merge": 152191, "admitted_skel": 0, "staged_skel": 0, "mean_count": 761.2725830078125, "n_util_obs": 332, "util_count_spearman": -0.29548552665795047}, "probe_overhead_sec_total": 624.93, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.007019, "alpha": 0.03, "t": 6827.0} |
| {"event": "consolidate", "step": 5800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262123, "t": 6854.4} |
| {"event": "probe", "step": 5800, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3493.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.200713, "cps": 27.289062, "mean_p": 0.210009, "boundary_entropy": 0.105215, "bpb": 1.879503}, "tr": {"chunks": 3926.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.21687, "cps": 30.671875, "mean_p": 0.202351, "boundary_entropy": 0.098746, "bpb": 1.999144}, "fi": {"chunks": 4478.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.238052, "cps": 34.984375, "mean_p": 0.233854, "boundary_entropy": 0.108271, "bpb": 2.113814}}, "bpb": {"en": 1.8795, "tr": 1.99914, "fi": 2.11381}, "gini_cpc": 0.037968, "gini_cps": 0.055196, "bpic": 4.768597, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0805012618334893, "fi": 1.1860356419910913}, "inventory": {"size": 326656, "admitted": 65536, "staged": 261120, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3345, "admitted_surface": 25189, "staged_surface": 61316, "admitted_split": 37141, "staged_split": 48138, "admitted_merge": 3206, "staged_merge": 151666, "admitted_skel": 0, "staged_skel": 0, "mean_count": 672.5001220703125, "n_util_obs": 323, "util_count_spearman": -0.27543196185191365}, "t": 6855.6} |
| {"event": "eval", "step": 5800, "phase": 3, "bpb": {"en": 1.895, "tr": 2.00781, "fi": 2.11878}, "bytes_per_chunk": {"en": 5.058, "tr": 5.188, "fi": 4.626}, "bpb_mean": 2.0072, "frozen_inventory": true, "t": 6857.6} |
| {"event": "train", "step": 5800, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.3261, "bpb": 1.91316, "chunk_loss": 0.030465, "F": 0.19907, "G": 0.20661, "chunks": 203.84, "bytes_per_chunk": 5.023, "boundary_entropy": 0.10068, "router_noise": 0.0, "grad_norm": 0.7502, "lr": 3.08e-05, "router_trainable": true, "sec_per_step": 0.686, "probe_sec": 0.0714, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05309, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": -0.08548, "n_shadow": 5, "mean_delta_shadow": 0.04102, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.91732, "trunc_abs_frac": 0.35218, "mean_delta_full": -0.13041, "read_hit_rate": {"surface": 0.6338, "split": 0.53761, "merge": 0.14058}, "mean_gate": {"surface": 0.12218, "split": 0.0465, "merge": 0.11895}, "n_reads": 11112, "n_cand": 24278, "read_suppressed_frac": 0.1012, "dropout_reads_logged": 11112, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3344, "admitted_surface": 25189, "staged_surface": 61531, "admitted_split": 37141, "staged_split": 48415, "admitted_merge": 3206, "staged_merge": 152198, "admitted_skel": 0, "staged_skel": 0, "mean_count": 672.669677734375, "n_util_obs": 323, "util_count_spearman": -0.2780818912868987}, "probe_overhead_sec_total": 628.224, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006688, "alpha": 0.03, "t": 6861.3} |
| {"event": "consolidate", "step": 5850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262122, "t": 6889.0} |
| {"event": "train", "step": 5850, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.34378, "bpb": 1.93867, "chunk_loss": 0.030486, "F": 0.20401, "G": 0.21172, "chunks": 208.91, "bytes_per_chunk": 4.902, "boundary_entropy": 0.09976, "router_noise": 0.0, "grad_norm": 0.7411, "lr": 3.05e-05, "router_trainable": true, "sec_per_step": 0.567, "probe_sec": 0.1262, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10075, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 6, "mean_delta_loo": -0.00282, "n_shadow": 4, "mean_delta_shadow": 0.01539, "n_split": 1, "mean_delta_split": 0.01495, "n_merge": 1, "mean_delta_merge": -0.07275, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": -0.01051, "trunc_abs_frac": 0.01051, "mean_delta_full": -0.45862, "read_hit_rate": {"surface": 0.61446, "split": 0.55316, "merge": 0.14409}, "mean_gate": {"surface": 0.11517, "split": 0.04251, "merge": 0.11476}, "n_reads": 11566, "n_cand": 25068, "read_suppressed_frac": 0.0993, "dropout_reads_logged": 11566, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3336, "admitted_surface": 24924, "staged_surface": 61683, "admitted_split": 37638, "staged_split": 48457, "admitted_merge": 2974, "staged_merge": 152004, "admitted_skel": 0, "staged_skel": 0, "mean_count": 677.9632568359375, "n_util_obs": 322, "util_count_spearman": -0.30572735700612624}, "probe_overhead_sec_total": 631.995, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.006817, "alpha": 0.03, "t": 6889.6} |
| {"event": "consolidate", "step": 5900, "retired": 0, "split_retired": 1, "merge_admitted": 0, "admitted": 1, "swapped": 1024, "uncertain_staged": 262122, "t": 6915.9} |
| {"event": "train", "step": 5900, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.27918, "bpb": 1.84546, "chunk_loss": 0.030286, "F": 0.19626, "G": 0.20279, "chunks": 200.97, "bytes_per_chunk": 5.095, "boundary_entropy": 0.09541, "router_noise": 0.0, "grad_norm": 0.7109, "lr": 3.02e-05, "router_trainable": true, "sec_per_step": 0.532, "probe_sec": 0.0562, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03699, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 4, "mean_delta_loo": 0.1301, "n_shadow": 6, "mean_delta_shadow": -0.06179, "n_split": 1, "mean_delta_split": 0.28052, "n_merge": 1, "mean_delta_merge": 0.05515, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -0.11809, "trunc_abs_frac": 25.17041, "mean_delta_full": -0.23834, "read_hit_rate": {"surface": 0.62619, "split": 0.54775, "merge": 0.13381}, "mean_gate": {"surface": 0.11914, "split": 0.04419, "merge": 0.11937}, "n_reads": 11030, "n_cand": 24032, "read_suppressed_frac": 0.0942, "dropout_reads_logged": 11030, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3351, "admitted_surface": 24424, "staged_surface": 61771, "admitted_split": 38135, "staged_split": 48251, "admitted_merge": 2977, "staged_merge": 152122, "admitted_skel": 0, "staged_skel": 0, "mean_count": 684.6046142578125, "n_util_obs": 331, "util_count_spearman": -0.3507106150501484}, "probe_overhead_sec_total": 635.215, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.007035, "alpha": 0.03, "t": 6916.3} |
| {"event": "consolidate", "step": 5950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262122, "t": 6944.1} |
| {"event": "train", "step": 5950, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38036, "bpb": 1.99144, "chunk_loss": 0.030349, "F": 0.19897, "G": 0.20592, "chunks": 203.75, "bytes_per_chunk": 5.026, "boundary_entropy": 0.09896, "router_noise": 0.0, "grad_norm": 0.7506, "lr": 3.01e-05, "router_trainable": true, "sec_per_step": 0.565, "probe_sec": 0.0705, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05172, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 5, "mean_delta_loo": -0.04075, "n_shadow": 3, "mean_delta_shadow": 0.06555, "n_split": 3, "mean_delta_split": 0.09951, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -0.14262, "trunc_abs_frac": 0.13569, "mean_delta_full": 0.06034, "read_hit_rate": {"surface": 0.61221, "split": 0.58409, "merge": 0.13011}, "mean_gate": {"surface": 0.12237, "split": 0.04505, "merge": 0.11937}, "n_reads": 11522, "n_cand": 24432, "read_suppressed_frac": 0.1016, "dropout_reads_logged": 11522, "inventory": {"size": 327680, "admitted": 65536, "staged": 262144, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3373, "admitted_surface": 24159, "staged_surface": 61709, "admitted_split": 38400, "staged_split": 48168, "admitted_merge": 2977, "staged_merge": 152267, "admitted_skel": 0, "staged_skel": 0, "mean_count": 692.3763427734375, "n_util_obs": 338, "util_count_spearman": -0.3500363636307144}, "probe_overhead_sec_total": 638.68, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.007917, "alpha": 0.03, "t": 6944.5} |
| {"event": "consolidate", "step": 6000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 1024, "uncertain_staged": 262122, "t": 6971.6} |
| {"event": "train", "step": 5999, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38288, "bpb": 1.99507, "chunk_loss": 0.030471, "F": 0.20184, "G": 0.20954, "chunks": 206.69, "bytes_per_chunk": 4.954, "boundary_entropy": 0.09878, "router_noise": 0.0, "grad_norm": 0.7111, "lr": 3e-05, "router_trainable": true, "sec_per_step": 0.541, "probe_sec": 0.07, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04944, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 9, "mean_delta_loo": 0.06328, "n_shadow": 5, "mean_delta_shadow": -0.04992, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.43403, "trunc_abs_frac": 0.45205, "mean_delta_full": 0.16561, "read_hit_rate": {"surface": 0.62109, "split": 0.55998, "merge": 0.15573}, "mean_gate": {"surface": 0.12311, "split": 0.04491, "merge": 0.12374}, "n_reads": 11419, "n_cand": 24402, "read_suppressed_frac": 0.0999, "dropout_reads_logged": 11419, "inventory": {"size": 326656, "admitted": 65536, "staged": 261120, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3386, "admitted_surface": 23702, "staged_surface": 61512, "admitted_split": 38857, "staged_split": 48033, "admitted_merge": 2977, "staged_merge": 151575, "admitted_skel": 0, "staged_skel": 0, "mean_count": 699.0468139648438, "n_util_obs": 339, "util_count_spearman": -0.3538895853675816}, "probe_overhead_sec_total": 642.035, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008012, "alpha": 0.03, "t": 6971.6} |
| {"event": "probe", "step": 6000, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3489.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.200483, "cps": 27.257812, "mean_p": 0.209511, "boundary_entropy": 0.104183, "bpb": 1.877471}, "tr": {"chunks": 3922.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.216649, "cps": 30.640625, "mean_p": 0.202895, "boundary_entropy": 0.096176, "bpb": 1.994715}, "fi": {"chunks": 4488.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.238584, "cps": 35.0625, "mean_p": 0.23381, "boundary_entropy": 0.106138, "bpb": 2.110797}}, "bpb": {"en": 1.87747, "tr": 1.99472, "fi": 2.1108}, "gini_cpc": 0.038737, "gini_cps": 0.055971, "bpic": 4.767796, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0806378842307198, "fi": 1.1900470046591445}, "inventory": {"size": 326656, "admitted": 65536, "staged": 261120, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3386, "admitted_surface": 23702, "staged_surface": 61512, "admitted_split": 38857, "staged_split": 48033, "admitted_merge": 2977, "staged_merge": 151575, "admitted_skel": 0, "staged_skel": 0, "mean_count": 699.0468139648438, "n_util_obs": 339, "util_count_spearman": -0.3538895853675816}, "t": 6972.6} |
| {"event": "eval", "step": 6000, "phase": 3, "bpb": {"en": 1.89289, "tr": 2.00403, "fi": 2.11628}, "bytes_per_chunk": {"en": 5.062, "tr": 5.178, "fi": 4.621}, "bpb_mean": 2.0044, "frozen_inventory": true, "t": 6974.4} |
| {"event": "reps_dumped", "dirs": ["/local/scratch/sas245/beetle-hnet/runs/ali_ali-noskel_s2/probe/reps/th0.30", "/local/scratch/sas245/beetle-hnet/runs/ali_ali-noskel_s2/probe/reps/th0.40", "/local/scratch/sas245/beetle-hnet/runs/ali_ali-noskel_s2/probe/reps/th0.50", "/local/scratch/sas245/beetle-hnet/runs/ali_ali-noskel_s2/probe/reps/th0.60", "/local/scratch/sas245/beetle-hnet/runs/ali_ali-noskel_s2/probe/reps/th0.70"], "t": 6978.2} |
| {"event": "final", "run": "ali_ali-noskel_s2", "arch": "ali", "objective": "ratio", "condition": "A", "step": 6000, "params": 22785088, "seed": 2, "per_lang": {"en": {"cpc": 0.20048267540079295, "cps": 27.2578125, "bpb": 1.898725968492211, "bytes_per_chunk": 4.981074713080489, "reads_per_chunk": 1.7221346051531503}, "tr": {"cpc": 0.21664917417002708, "cps": 30.640625, "bpb": 1.9965886436244815, "bytes_per_chunk": 5.1881450695166, "reads_per_chunk": 1.7412300232546634}, "fi": {"cpc": 0.23858380734676518, "cps": 35.0625, "bpb": 2.134636172269162, "bytes_per_chunk": 4.599663110612016, "reads_per_chunk": 1.8876684446939922}}, "bpb_mean": 2.0099835947952847, "gini_cpc": 0.03873745328890421, "gini_cps": 0.05597109000756384, "bpic": 4.767795613076729, "mean_boundary_entropy": 0.09877879917621613, "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": 642.0346233844757, "probe_overhead_seq_frac": 0.49990625, "wall_sec": 6983.717223644257, "inventory": {"size": 326656, "admitted": 65536, "staged": 261120, "budget_B": 65536, "staging_capacity": 262144, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 3386, "admitted_surface": 23702, "staged_surface": 61512, "admitted_split": 38857, "staged_split": 48033, "admitted_merge": 2977, "staged_merge": 151575, "admitted_skel": 0, "staged_skel": 0, "mean_count": 699.0468139648438, "n_util_obs": 339, "util_count_spearman": -0.3538895853675816}, "ali_params": 268096, "ali_config": {"enabled": true, "budget_B": 65536, "staging_capacity": 262144, "policy": "utility", "candidate_types": ["surface", "split", "merge"], "gate": "vector", "retrieval": true, "memory_scope": "persistent", "read_dropout": 0.1, "write_rule": "mean", "eta_min": 0.0, "consolidate_every": 50, "probe_seqs": 2, "probe_reps": 8, "horizon": 32, "predictor": "mlp", "lambda_util": 0.1, "max_span": 64, "byte_classes_path": "/local/scratch/sas245/beetle-hnet/scratch/data/byte_classes.json", "gate_hidden": 0, "type_dim": 16, "gate_bias_init": -2.0}, "util_vs_frequency_spearman": -0.3538895853675816, "paired_vs_dropout": {"n": 1023, "pearson": -0.040930230288563525, "spearman": -0.02528153098655959, "mean_abs_discrepancy": 0.4435354769229889, "mean_paired": 0.025582697242498398, "mean_dropout": 0.005726948380470276}, "predictor_calibration": {"n": 2607, "all": {"n": 2607, "pearson": 0.10252948259589209, "spearman": 0.12719968041127563, "mae": 0.07177101820707321, "reliability": [{"pred": 0.0018613574793562293, "obs": 0.0011005308479070663, "n": 261}, {"pred": 0.005867928732186556, "obs": 0.0032962211407721043, "n": 261}, {"pred": 0.009169374592602253, "obs": 0.006755373906344175, "n": 260}, {"pred": 0.012885394506156445, "obs": 0.011649513617157936, "n": 261}, {"pred": 0.015666397288441658, "obs": 0.012515993788838387, "n": 260}, {"pred": 0.018621260300278664, "obs": 0.013944972306489944, "n": 261}, {"pred": 0.021858371794223785, "obs": 0.022404633462429047, "n": 261}, {"pred": 0.024777401238679886, "obs": 0.03755471110343933, "n": 260}, {"pred": 0.02776263654232025, "obs": 0.023750539869070053, "n": 261}, {"pred": 0.03260844573378563, "obs": 0.037009917199611664, "n": 261}]}, "admitted": {"n": 2570, "pearson": 0.10141530063650674, "spearman": 0.12712036604241447, "mae": 0.07223286479711533, "reliability": [{"pred": 0.002404953818768263, "obs": 0.002259984612464905, "n": 257}, {"pred": 0.005994603969156742, "obs": 0.004191071726381779, "n": 257}, {"pred": 0.009257971309125423, "obs": 0.00717716570943594, "n": 257}, {"pred": 0.012950187548995018, "obs": 0.010941623710095882, "n": 257}, {"pred": 0.015743210911750793, "obs": 0.012395243160426617, "n": 257}, {"pred": 0.018728459253907204, "obs": 0.01735404133796692, "n": 257}, {"pred": 0.021917937323451042, "obs": 0.021629782393574715, "n": 257}, {"pred": 0.02479337342083454, "obs": 0.03687406703829765, "n": 257}, {"pred": 0.027748797088861465, "obs": 0.02369796112179756, "n": 257}, {"pred": 0.03250524401664734, "obs": 0.039038389921188354, "n": 257}]}, "staged": {"n": 37, "pearson": 0.0011803028758304447, "spearman": -0.14817449027975343, "mae": 0.0396917462348938, "reliability": [{"pred": -0.013648699969053268, "obs": -0.059147514402866364, "n": 4}, {"pred": -0.004886574111878872, "obs": -0.004800606518983841, "n": 4}, {"pred": -0.0032589684706181288, "obs": -0.02864583395421505, "n": 3}, {"pred": 0.0003166003152728081, "obs": -0.014926524832844734, "n": 4}, {"pred": 0.009122427552938461, "obs": -0.012634028680622578, "n": 3}, {"pred": 0.013752173632383347, "obs": -0.009141568094491959, "n": 4}, {"pred": 0.015248742885887623, "obs": -0.00426017539575696, "n": 4}, {"pred": 0.018152182921767235, "obs": -0.026346325874328613, "n": 3}, {"pred": 0.03130064532160759, "obs": -0.026265069842338562, "n": 4}, {"pred": 0.03650445491075516, "obs": -0.03328035771846771, "n": 4}]}}, "t": 6984.1} |
| {"event": "hub_done", "repo": "AdaptiveChunking/ali_ali-noskel_s2", "pushed": [0, 1, 2, 4, 6, 8, 16, 32, 64, 128, 256, 512, 1000, 1024, 2048, 3000, 4096, 6000], "failed": [], "ok": true, "t": 7000.0} |
|
|
