diff --git "a/log.jsonl" "b/log.jsonl"
new file mode 100644--- /dev/null
+++ "b/log.jsonl"
@@ -0,0 +1,323 @@
+{"event": "hub", "repo": "AdaptiveChunking/ali_mem-split_s1", "keep_steps": [0, 1, 2, 4, 6, 8, 16, 32, 64, 128, 256, 512, 1000, 1024, 2048, 3000, 4096, 6000], "t": 5.5}
+{"event": "start", "run": "ali_mem-split_s1", "arch": "ali", "objective": "ratio", "condition": "A", "params": 22785088, "langs": ["en", "tr", "fi"], "premium": {"en": 1.0, "tr": 1.12, "fi": 1.104}, "schedule": {"tau1": 0.4, "tau2": 0.7, "p1_objective": "global", "p2_objective": "global", "p3_objective": "global", "freeze_in_p3": false, "freeze_scope": "router"}, "corpus": "/local/scratch/sas245/beetle-hnet/scratch/data/tier1_corpus", "premium_kind": "flores_per_sent", "doc_filtered": {}, "weights": {"en": 0.74, "tr": 0.1536, "fi": 0.1064}, "device": "cuda", "ali_params": 268096, "ali_config": {"enabled": true, "budget_B": 65536, "staging_capacity": 262144, "policy": "frequency", "value_mode": "learned", "candidate_types": ["split"], "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": 5.5}
+{"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": 5.5}
+{"event": "probe", "step": 0, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 2674.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.153652, "cps": 20.890625, "mean_p": 0.478126, "boundary_entropy": 0.685438, "bpb": 8.041379}, "tr": {"chunks": 5645.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.311827, "cps": 44.101562, "mean_p": 0.486063, "boundary_entropy": 0.686375, "bpb": 8.01701}, "fi": {"chunks": 4732.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.251555, "cps": 36.96875, "mean_p": 0.48451, "boundary_entropy": 0.686595, "bpb": 8.075294}}, "bpb": {"en": 8.04138, "tr": 8.01701, "fi": 8.07529}, "gini_cpc": 0.147064, "gini_cps": 0.151764, "bpic": 4.346947, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 2.029439514461233, "fi": 1.6371767549398106}, "inventory": {"size": 0, "admitted": 0, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 0.0, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 7.3}
+{"event": "train", "step": 0, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 5.58509, "bpb": 8.05758, "chunk_loss": 0.030141, "F": 0.16724, "G": 0.47713, "chunks": 171.25, "bytes_per_chunk": 5.98, "boundary_entropy": 0.69022, "router_noise": 0.0, "grad_norm": 11.3985, "lr": 1e-06, "router_trainable": true, "sec_per_step": 0.021, "probe_sec": 0.02, "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": {"split": 0.0}, "mean_gate": {"split": null}, "n_reads": 0, "n_cand": 8317, "read_suppressed_frac": 0.0, "dropout_reads_logged": 0, "inventory": {"size": 3293, "admitted": 3293, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 3293, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 2.525660514831543, "n_util_obs": 0, "util_count_spearman": NaN}, "probe_overhead_sec_total": 0.02, "probe_overhead_seq_frac": 0.0, "util_loss": 0.0, "alpha": 0.03, "t": 8.8}
+{"event": "probe", "step": 1, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 2661.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.152905, "cps": 20.789062, "mean_p": 0.478104, "boundary_entropy": 0.685439, "bpb": 7.995304}, "tr": {"chunks": 5624.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.310667, "cps": 43.9375, "mean_p": 0.485998, "boundary_entropy": 0.686373, "bpb": 7.972234}, "fi": {"chunks": 4706.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.250173, "cps": 36.765625, "mean_p": 0.484401, "boundary_entropy": 0.686587, "bpb": 8.027531}}, "bpb": {"en": 7.9953, "tr": 7.97223, "fi": 8.02753}, "gini_cpc": 0.147356, "gini_cps": 0.152054, "bpic": 4.367023, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 2.03176748657516, "fi": 1.6361355648548603}, "inventory": {"size": 3293, "admitted": 3293, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 3293, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 2.525660514831543, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 10.0}
+{"event": "probe", "step": 2, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 2659.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.15279, "cps": 20.773438, "mean_p": 0.478079, "boundary_entropy": 0.685441, "bpb": 7.931845}, "tr": {"chunks": 5589.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.308733, "cps": 43.664062, "mean_p": 0.485916, "boundary_entropy": 0.686372, "bpb": 7.924633}, "fi": {"chunks": 4671.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.248312, "cps": 36.492188, "mean_p": 0.484243, "boundary_entropy": 0.686576, "bpb": 7.975275}}, "bpb": {"en": 7.93184, "tr": 7.92463, "fi": 7.97527}, "gini_cpc": 0.14646, "gini_cps": 0.151198, "bpic": 4.391362, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 2.0206418385189076, "fi": 1.625188597911143}, "inventory": {"size": 5747, "admitted": 5747, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 5747, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 2.7814512252807617, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 12.0}
+{"event": "probe", "step": 4, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 2627.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.150951, "cps": 20.523438, "mean_p": 0.478069, "boundary_entropy": 0.685456, "bpb": 7.718832}, "tr": {"chunks": 5486.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.303044, "cps": 42.859375, "mean_p": 0.485685, "boundary_entropy": 0.686372, "bpb": 7.771181}, "fi": {"chunks": 4570.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.242943, "cps": 35.703125, "mean_p": 0.483805, "boundary_entropy": 0.686547, "bpb": 7.798779}}, "bpb": {"en": 7.71883, "tr": 7.77118, "fi": 7.79878}, "gini_cpc": 0.145487, "gini_cps": 0.15028, "bpic": 4.473074, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 2.007563537841377, "fi": 1.6094161834255472}, "inventory": {"size": 10466, "admitted": 10466, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 10466, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 3.0808331966400146, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 14.5}
+{"event": "probe", "step": 6, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 2602.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.149514, "cps": 20.328125, "mean_p": 0.478167, "boundary_entropy": 0.685497, "bpb": 7.43677}, "tr": {"chunks": 5308.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.293211, "cps": 41.46875, "mean_p": 0.485329, "boundary_entropy": 0.686375, "bpb": 7.573529}, "fi": {"chunks": 4362.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.231886, "cps": 34.078125, "mean_p": 0.483111, "boundary_entropy": 0.686503, "bpb": 7.563642}}, "bpb": {"en": 7.43677, "tr": 7.57353, "fi": 7.56364}, "gini_cpc": 0.142004, "gini_cps": 0.147001, "bpic": 4.622881, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.9610884900108068, "fi": 1.5509243185306516}, "inventory": {"size": 14966, "admitted": 14966, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 14966, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 3.18194580078125, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 16.6}
+{"event": "probe", "step": 8, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 2667.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.153249, "cps": 20.835938, "mean_p": 0.478981, "boundary_entropy": 0.685625, "bpb": 7.160058}, "tr": {"chunks": 5245.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.289731, "cps": 40.976562, "mean_p": 0.485273, "boundary_entropy": 0.686413, "bpb": 7.383668}, "fi": {"chunks": 4207.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.223646, "cps": 32.867188, "mean_p": 0.482708, "boundary_entropy": 0.686493, "bpb": 7.327325}}, "bpb": {"en": 7.16006, "tr": 7.38367, "fi": 7.32732}, "gini_cpc": 0.13649, "gini_cps": 0.141816, "bpic": 4.681244, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.890584293711891, "fi": 1.4593576300012097}, "inventory": {"size": 19318, "admitted": 19318, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 19318, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 3.2594988346099854, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 19.7}
+{"event": "probe", "step": 16, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 2611.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.150032, "cps": 20.398438, "mean_p": 0.481112, "boundary_entropy": 0.686035, "bpb": 6.511824}, "tr": {"chunks": 4405.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.24333, "cps": 34.414062, "mean_p": 0.483041, "boundary_entropy": 0.686446, "bpb": 6.895112}, "fi": {"chunks": 2951.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.156876, "cps": 23.054688, "mean_p": 0.478287, "boundary_entropy": 0.686269, "bpb": 6.689287}}, "bpb": {"en": 6.51182, "tr": 6.89511, "fi": 6.68929}, "gini_cpc": 0.11304, "gini_cps": 0.119996, "bpic": 5.691984, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.6218571871375702, "fi": 1.0456216681484454}, "inventory": {"size": 35575, "admitted": 35575, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 35575, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 3.843822956085205, "n_util_obs": 5, "util_count_spearman": -0.8}, "t": 23.1}
+{"event": "probe", "step": 32, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3520.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.202264, "cps": 27.5, "mean_p": 0.490038, "boundary_entropy": 0.687151, "bpb": 6.160803}, "tr": {"chunks": 3886.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.214661, "cps": 30.359375, "mean_p": 0.48616, "boundary_entropy": 0.687286, "bpb": 6.5768}, "fi": {"chunks": 3148.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.167349, "cps": 24.59375, "mean_p": 0.484909, "boundary_entropy": 0.687291, "bpb": 6.340781}}, "bpb": {"en": 6.1608, "tr": 6.5768, "fi": 6.34078}, "gini_cpc": 0.053983, "gini_cps": 0.046617, "bpic": 5.375403, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0612890944745472, "fi": 0.8273786251757916}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 4.184814453125, "n_util_obs": 11, "util_count_spearman": -0.43636363636363634}, "t": 28.9}
+{"event": "consolidate", "step": 50, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 39.7}
+{"event": "train", "step": 50, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 4.0974, "bpb": 5.9113, "chunk_loss": 0.029787, "F": 0.16321, "G": 0.48556, "chunks": 167.12, "bytes_per_chunk": 6.127, "boundary_entropy": 0.6916, "router_noise": 0.0, "grad_norm": 2.3626, "lr": 5.1e-05, "router_trainable": true, "sec_per_step": 0.649, "probe_sec": 0.1098, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05508, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.01124, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 0.59435, "trunc_abs_frac": 0.58228, "mean_delta_full": 0.02124, "read_hit_rate": {"split": 0.77708}, "mean_gate": {"split": 0.11926}, "n_reads": 6599, "n_cand": 8492, "read_suppressed_frac": 0.1079, "dropout_reads_logged": 6599, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 6.22265625, "n_util_obs": 17, "util_count_spearman": -0.14950980392156862}, "probe_overhead_sec_total": 3.605, "probe_overhead_seq_frac": 0.4902, "util_loss": 0.004027, "alpha": 0.03, "t": 41.2}
+{"event": "probe", "step": 64, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3774.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.216859, "cps": 29.484375, "mean_p": 0.484558, "boundary_entropy": 0.686165, "bpb": 5.424479}, "tr": {"chunks": 4626.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.255538, "cps": 36.140625, "mean_p": 0.481624, "boundary_entropy": 0.686069, "bpb": 5.91892}, "fi": {"chunks": 3874.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.205943, "cps": 30.265625, "mean_p": 0.481903, "boundary_entropy": 0.686447, "bpb": 5.652137}}, "bpb": {"en": 5.42448, "tr": 5.91892, "fi": 5.65214}, "gini_cpc": 0.048741, "gini_cps": 0.046277, "bpic": 4.622128, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.1783581268359546, "fi": 0.9496639612904502}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 7.656463623046875, "n_util_obs": 19, "util_count_spearman": -0.054385964912280704}, "t": 48.9}
+{"event": "consolidate", "step": 100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 61.7}
+{"event": "train", "step": 100, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 3.21187, "bpb": 4.63375, "chunk_loss": 0.033126, "F": 0.21509, "G": 0.46414, "chunks": 220.25, "bytes_per_chunk": 4.649, "boundary_entropy": 0.68493, "router_noise": 0.0, "grad_norm": 1.8778, "lr": 0.000101, "router_trainable": true, "sec_per_step": 0.414, "probe_sec": 0.0405, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03252, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.00213, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 7, "trunc_ratio_sum": -0.12149, "trunc_abs_frac": 0.41492, "mean_delta_full": -0.03973, "read_hit_rate": {"split": 0.87962}, "mean_gate": {"split": 0.11966}, "n_reads": 10902, "n_cand": 12394, "read_suppressed_frac": 0.0997, "dropout_reads_logged": 10902, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 12.756072998046875, "n_util_obs": 28, "util_count_spearman": 0.10016420361247948}, "probe_overhead_sec_total": 6.922, "probe_overhead_seq_frac": 0.495, "util_loss": 0.000452, "alpha": 0.03, "t": 61.9}
+{"event": "probe", "step": 128, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3673.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.211056, "cps": 28.695312, "mean_p": 0.425558, "boundary_entropy": 0.62533, "bpb": 4.113804}, "tr": {"chunks": 4109.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.226979, "cps": 32.101562, "mean_p": 0.408927, "boundary_entropy": 0.643751, "bpb": 4.363621}, "fi": {"chunks": 4125.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.219287, "cps": 32.226562, "mean_p": 0.412546, "boundary_entropy": 0.645361, "bpb": 4.31325}}, "bpb": {"en": 4.1138, "tr": 4.36362, "fi": 4.31325}, "gini_cpc": 0.01615, "gini_cps": 0.025307, "bpic": 4.764592, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0754464286318544, "fi": 1.0389993151674566}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 17.345138549804688, "n_util_obs": 32, "util_count_spearman": -0.03519061583577713}, "t": 74.0}
+{"event": "consolidate", "step": 150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 85.4}
+{"event": "train", "step": 150, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.7491, "bpb": 3.96612, "chunk_loss": 0.031932, "F": 0.20441, "G": 0.40053, "chunks": 209.31, "bytes_per_chunk": 4.892, "boundary_entropy": 0.58328, "router_noise": 0.0, "grad_norm": 1.2116, "lr": 0.000151, "router_trainable": true, "sec_per_step": 0.475, "probe_sec": 0.0474, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03728, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.02282, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 8.55932, "trunc_abs_frac": 2.57374, "mean_delta_full": 0.0048, "read_hit_rate": {"split": 0.91807}, "mean_gate": {"split": 0.11985}, "n_reads": 11015, "n_cand": 11998, "read_suppressed_frac": 0.1025, "dropout_reads_logged": 11015, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 21.318618774414062, "n_util_obs": 44, "util_count_spearman": -0.15193798449612403}, "probe_overhead_sec_total": 10.866, "probe_overhead_seq_frac": 0.4967, "util_loss": 0.000384, "alpha": 0.03, "t": 85.7}
+{"event": "consolidate", "step": 200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 106.0}
+{"event": "probe", "step": 200, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3274.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.188128, "cps": 25.578125, "mean_p": 0.357366, "boundary_entropy": 0.507331, "bpb": 3.553562}, "tr": {"chunks": 4054.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.223941, "cps": 31.671875, "mean_p": 0.330222, "boundary_entropy": 0.513894, "bpb": 3.768716}, "fi": {"chunks": 4696.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.249641, "cps": 36.6875, "mean_p": 0.351506, "boundary_entropy": 0.53267, "bpb": 3.856694}}, "bpb": {"en": 3.55356, "tr": 3.76872, "fi": 3.85669}, "gini_cpc": 0.061973, "gini_cps": 0.078842, "bpic": 4.71823, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.1903608587944683, "fi": 1.326971666554035}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 29.213119506835938, "n_util_obs": 59, "util_count_spearman": -0.2566919929865576}, "t": 106.8}
+{"event": "eval", "step": 200, "phase": 1, "bpb": {"en": 3.56221, "tr": 3.77752, "fi": 3.85564}, "bytes_per_chunk": {"en": 5.369, "tr": 5.032, "fi": 4.483}, "bpb_mean": 3.73179, "frozen_inventory": true, "t": 108.3}
+{"event": "train", "step": 200, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.48675, "bpb": 3.58763, "chunk_loss": 0.030903, "F": 0.18954, "G": 0.34788, "chunks": 194.09, "bytes_per_chunk": 5.276, "boundary_entropy": 0.5087, "router_noise": 0.0, "grad_norm": 0.9301, "lr": 0.000201, "router_trainable": true, "sec_per_step": 0.471, "probe_sec": 0.0474, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03752, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.02159, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.69932, "trunc_abs_frac": 0.73913, "mean_delta_full": 0.05359, "read_hit_rate": {"split": 0.89788}, "mean_gate": {"split": 0.12097}, "n_reads": 9812, "n_cand": 10928, "read_suppressed_frac": 0.1059, "dropout_reads_logged": 9812, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 29.364120483398438, "n_util_obs": 59, "util_count_spearman": -0.25891291642314435}, "probe_overhead_sec_total": 14.751, "probe_overhead_seq_frac": 0.4975, "util_loss": 0.000544, "alpha": 0.03, "t": 109.2}
+{"event": "consolidate", "step": 250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 123.5}
+{"event": "train", "step": 250, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.35955, "bpb": 3.40412, "chunk_loss": 0.031223, "F": 0.20486, "G": 0.31054, "chunks": 209.78, "bytes_per_chunk": 4.881, "boundary_entropy": 0.43504, "router_noise": 0.0, "grad_norm": 1.5112, "lr": 0.000251, "router_trainable": true, "sec_per_step": 0.29, "probe_sec": 0.0475, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03741, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.01495, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 1.07227, "trunc_abs_frac": 1.7056, "mean_delta_full": -0.11092, "read_hit_rate": {"split": 0.9269}, "mean_gate": {"split": 0.12817}, "n_reads": 11564, "n_cand": 12476, "read_suppressed_frac": 0.0965, "dropout_reads_logged": 11564, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 37.65730285644531, "n_util_obs": 70, "util_count_spearman": -0.28622418183390613}, "probe_overhead_sec_total": 18.747, "probe_overhead_seq_frac": 0.498, "util_loss": 0.000849, "alpha": 0.03, "t": 123.7}
+{"event": "probe", "step": 256, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3623.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.208182, "cps": 28.304688, "mean_p": 0.296253, "boundary_entropy": 0.40617, "bpb": 3.284805}, "tr": {"chunks": 3964.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.218969, "cps": 30.96875, "mean_p": 0.290271, "boundary_entropy": 0.427949, "bpb": 3.511084}, "fi": {"chunks": 3941.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.209505, "cps": 30.789062, "mean_p": 0.284134, "boundary_entropy": 0.437834, "bpb": 3.654273}}, "bpb": {"en": 3.2848, "tr": 3.51108, "fi": 3.65427}, "gini_cpc": 0.011295, "gini_cps": 0.01972, "bpic": 4.921235, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.051813838770812, "fi": 1.0063529814634302}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 38.56401062011719, "n_util_obs": 71, "util_count_spearman": -0.29948272538548604}, "t": 125.9}
+{"event": "consolidate", "step": 300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 148.9}
+{"event": "train", "step": 300, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.22058, "bpb": 3.20362, "chunk_loss": 0.031143, "F": 0.20715, "G": 0.29168, "chunks": 212.12, "bytes_per_chunk": 4.827, "boundary_entropy": 0.39633, "router_noise": 0.0, "grad_norm": 2.2442, "lr": 0.0003, "router_trainable": true, "sec_per_step": 0.509, "probe_sec": 0.064, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05347, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.04314, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 0.42043, "trunc_abs_frac": 1.2427, "mean_delta_full": 0.07227, "read_hit_rate": {"split": 0.93373}, "mean_gate": {"split": 0.14634}, "n_reads": 11737, "n_cand": 12570, "read_suppressed_frac": 0.1003, "dropout_reads_logged": 11737, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 46.864898681640625, "n_util_obs": 77, "util_count_spearman": -0.24517587675482413}, "probe_overhead_sec_total": 22.216, "probe_overhead_seq_frac": 0.4983, "util_loss": 0.001087, "alpha": 0.03, "t": 149.2}
+{"event": "consolidate", "step": 350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 163.3}
+{"event": "train", "step": 350, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.20119, "bpb": 3.17565, "chunk_loss": 0.031318, "F": 0.21744, "G": 0.28483, "chunks": 222.66, "bytes_per_chunk": 4.599, "boundary_entropy": 0.369, "router_noise": 0.0, "grad_norm": 2.0235, "lr": 0.0002999, "router_trainable": true, "sec_per_step": 0.287, "probe_sec": 0.0542, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04463, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.06929, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 7, "trunc_ratio_sum": 1.91012, "trunc_abs_frac": 2.78167, "mean_delta_full": 0.05251, "read_hit_rate": {"split": 0.93259}, "mean_gate": {"split": 0.16305}, "n_reads": 12189, "n_cand": 13070, "read_suppressed_frac": 0.0962, "dropout_reads_logged": 12189, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 56.09388732910156, "n_util_obs": 91, "util_count_spearman": -0.10114667940754897}, "probe_overhead_sec_total": 25.705, "probe_overhead_seq_frac": 0.4986, "util_loss": 0.002028, "alpha": 0.03, "t": 163.5}
+{"event": "consolidate", "step": 400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 179.4}
+{"event": "probe", "step": 400, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3698.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.212492, "cps": 28.890625, "mean_p": 0.28001, "boundary_entropy": 0.335503, "bpb": 2.869527}, "tr": {"chunks": 3501.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.193393, "cps": 27.351562, "mean_p": 0.278233, "boundary_entropy": 0.372273, "bpb": 3.134386}, "fi": {"chunks": 4213.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.223965, "cps": 32.914062, "mean_p": 0.282019, "boundary_entropy": 0.374991, "bpb": 3.370696}}, "bpb": {"en": 2.86953, "tr": 3.13439, "fi": 3.3707}, "gini_cpc": 0.032358, "gini_cps": 0.041594, "bpic": 4.971258, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.910120240088811, "fi": 1.0539907247922526}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 65.06787109375, "n_util_obs": 96, "util_count_spearman": -0.09058600108518719}, "t": 180.3}
+{"event": "eval", "step": 400, "phase": 1, "bpb": {"en": 2.88349, "tr": 3.14202, "fi": 3.36304}, "bytes_per_chunk": {"en": 4.491, "tr": 5.749, "fi": 4.746}, "bpb_mean": 3.12952, "frozen_inventory": true, "t": 182.0}
+{"event": "train", "step": 400, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.02815, "bpb": 2.926, "chunk_loss": 0.031116, "F": 0.21143, "G": 0.27796, "chunks": 216.5, "bytes_per_chunk": 4.73, "boundary_entropy": 0.33602, "router_noise": 0.0, "grad_norm": 1.4624, "lr": 0.0002998, "router_trainable": true, "sec_per_step": 0.389, "probe_sec": 0.0442, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03458, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.00765, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -0.08283, "trunc_abs_frac": 0.33381, "mean_delta_full": -0.29352, "read_hit_rate": {"split": 0.94375}, "mean_gate": {"split": 0.18325}, "n_reads": 12114, "n_cand": 12836, "read_suppressed_frac": 0.0994, "dropout_reads_logged": 12114, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 65.25430297851562, "n_util_obs": 96, "util_count_spearman": -0.09096581660336409}, "probe_overhead_sec_total": 29.266, "probe_overhead_seq_frac": 0.4988, "util_loss": 0.002388, "alpha": 0.03, "t": 183.0}
+{"event": "consolidate", "step": 450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 203.6}
+{"event": "train", "step": 450, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.04891, "bpb": 2.95595, "chunk_loss": 0.031249, "F": 0.21796, "G": 0.27507, "chunks": 223.19, "bytes_per_chunk": 4.588, "boundary_entropy": 0.31508, "router_noise": 0.0, "grad_norm": 1.1556, "lr": 0.0002995, "router_trainable": true, "sec_per_step": 0.427, "probe_sec": 0.0745, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04887, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.05328, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.86887, "trunc_abs_frac": 0.8308, "mean_delta_full": -0.16055, "read_hit_rate": {"split": 0.93787}, "mean_gate": {"split": 0.18737}, "n_reads": 12288, "n_cand": 13102, "read_suppressed_frac": 0.1002, "dropout_reads_logged": 12288, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 74.43939208984375, "n_util_obs": 101, "util_count_spearman": 0.030669772859638907}, "probe_overhead_sec_total": 33.188, "probe_overhead_seq_frac": 0.4989, "util_loss": 0.002851, "alpha": 0.03, "t": 204.3}
+{"event": "consolidate", "step": 500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 223.5}
+{"event": "train", "step": 500, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.01709, "bpb": 2.91005, "chunk_loss": 0.031101, "F": 0.21188, "G": 0.27581, "chunks": 216.97, "bytes_per_chunk": 4.72, "boundary_entropy": 0.3078, "router_noise": 0.0, "grad_norm": 1.1783, "lr": 0.0002992, "router_trainable": true, "sec_per_step": 0.388, "probe_sec": 0.0847, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07277, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.00933, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.54674, "trunc_abs_frac": 0.59537, "mean_delta_full": 0.32971, "read_hit_rate": {"split": 0.92632}, "mean_gate": {"split": 0.19401}, "n_reads": 11566, "n_cand": 12486, "read_suppressed_frac": 0.0963, "dropout_reads_logged": 11566, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 83.60923767089844, "n_util_obs": 117, "util_count_spearman": 0.09429183713228131}, "probe_overhead_sec_total": 37.121, "probe_overhead_seq_frac": 0.499, "util_loss": 0.003323, "alpha": 0.03, "t": 223.7}
+{"event": "probe", "step": 512, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3697.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.212435, "cps": 28.882812, "mean_p": 0.26818, "boundary_entropy": 0.290913, "bpb": 2.650595}, "tr": {"chunks": 3326.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.183726, "cps": 25.984375, "mean_p": 0.273193, "boundary_entropy": 0.352513, "bpb": 2.971011}, "fi": {"chunks": 4359.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.231726, "cps": 34.054688, "mean_p": 0.284479, "boundary_entropy": 0.337826, "bpb": 3.204992}}, "bpb": {"en": 2.6506, "tr": 2.97101, "fi": 3.20499}, "gini_cpc": 0.050964, "gini_cps": 0.060505, "bpic": 4.984361, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.8648610987489299, "fi": 1.0908113676717592}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 85.61734008789062, "n_util_obs": 120, "util_count_spearman": 0.07752621709840961}, "t": 227.6}
+{"event": "consolidate", "step": 550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 239.6}
+{"event": "train", "step": 550, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.93795, "bpb": 2.79587, "chunk_loss": 0.031447, "F": 0.22284, "G": 0.27728, "chunks": 228.19, "bytes_per_chunk": 4.488, "boundary_entropy": 0.28649, "router_noise": 0.0, "grad_norm": 1.412, "lr": 0.0002987, "router_trainable": true, "sec_per_step": 0.322, "probe_sec": 0.0557, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04608, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.0078, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.0997, "trunc_abs_frac": 0.69151, "mean_delta_full": -0.14142, "read_hit_rate": {"split": 0.93984}, "mean_gate": {"split": 0.21094}, "n_reads": 12498, "n_cand": 13298, "read_suppressed_frac": 0.1, "dropout_reads_logged": 12498, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 92.76329040527344, "n_util_obs": 129, "util_count_spearman": 0.12531305903398926}, "probe_overhead_sec_total": 41.116, "probe_overhead_seq_frac": 0.4991, "util_loss": 0.003577, "alpha": 0.03, "t": 239.8}
+{"event": "consolidate", "step": 600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 253.6}
+{"event": "probe", "step": 600, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3764.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.216285, "cps": 29.40625, "mean_p": 0.266159, "boundary_entropy": 0.279526, "bpb": 2.552528}, "tr": {"chunks": 3412.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.188477, "cps": 26.65625, "mean_p": 0.278777, "boundary_entropy": 0.358927, "bpb": 2.882196}, "fi": {"chunks": 4602.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.244644, "cps": 35.953125, "mean_p": 0.289834, "boundary_entropy": 0.322232, "bpb": 3.100733}}, "bpb": {"en": 2.55253, "tr": 2.8822, "fi": 3.10073}, "gini_cpc": 0.05766, "gini_cps": 0.067357, "bpic": 4.816777, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.8714309421885852, "fi": 1.1311214451534817}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 101.8697509765625, "n_util_obs": 139, "util_count_spearman": 0.1237277399881687}, "t": 254.5}
+{"event": "eval", "step": 600, "phase": 1, "bpb": {"en": 2.56608, "tr": 2.89317, "fi": 3.09955}, "bytes_per_chunk": {"en": 4.481, "tr": 5.872, "fi": 4.3}, "bpb_mean": 2.85293, "frozen_inventory": true, "t": 256.1}
+{"event": "train", "step": 600, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.90959, "bpb": 2.75495, "chunk_loss": 0.031093, "F": 0.21399, "G": 0.27305, "chunks": 219.12, "bytes_per_chunk": 4.673, "boundary_entropy": 0.29284, "router_noise": 0.0, "grad_norm": 1.1373, "lr": 0.0002982, "router_trainable": true, "sec_per_step": 0.346, "probe_sec": 0.0384, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02947, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.06134, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.1897, "trunc_abs_frac": 0.44079, "mean_delta_full": -0.22015, "read_hit_rate": {"split": 0.94034}, "mean_gate": {"split": 0.21216}, "n_reads": 11963, "n_cand": 12722, "read_suppressed_frac": 0.0991, "dropout_reads_logged": 11963, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 102.05186462402344, "n_util_obs": 139, "util_count_spearman": 0.12912998793511773}, "probe_overhead_sec_total": 44.738, "probe_overhead_seq_frac": 0.4992, "util_loss": 0.004052, "alpha": 0.03, "t": 257.2}
+{"event": "consolidate", "step": 650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 269.6}
+{"event": "train", "step": 650, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.84997, "bpb": 2.66894, "chunk_loss": 0.030888, "F": 0.20834, "G": 0.26432, "chunks": 213.34, "bytes_per_chunk": 4.8, "boundary_entropy": 0.27797, "router_noise": 0.0, "grad_norm": 1.2079, "lr": 0.0002975, "router_trainable": true, "sec_per_step": 0.255, "probe_sec": 0.0601, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04845, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.0593, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -0.05616, "trunc_abs_frac": 1.98528, "mean_delta_full": -0.18803, "read_hit_rate": {"split": 0.93819}, "mean_gate": {"split": 0.21374}, "n_reads": 11748, "n_cand": 12522, "read_suppressed_frac": 0.0996, "dropout_reads_logged": 11748, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 111.33966064453125, "n_util_obs": 145, "util_count_spearman": 0.09272949141867423}, "probe_overhead_sec_total": 48.028, "probe_overhead_seq_frac": 0.4992, "util_loss": 0.005171, "alpha": 0.03, "t": 269.9}
+{"event": "consolidate", "step": 700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 283.0}
+{"event": "train", "step": 700, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.86277, "bpb": 2.68741, "chunk_loss": 0.030781, "F": 0.20248, "G": 0.26931, "chunks": 207.34, "bytes_per_chunk": 4.939, "boundary_entropy": 0.29072, "router_noise": 0.0, "grad_norm": 0.9722, "lr": 0.0002967, "router_trainable": true, "sec_per_step": 0.266, "probe_sec": 0.0485, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03926, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.07959, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.19825, "trunc_abs_frac": 1.22934, "mean_delta_full": 0.13033, "read_hit_rate": {"split": 0.92123}, "mean_gate": {"split": 0.21588}, "n_reads": 11169, "n_cand": 12124, "read_suppressed_frac": 0.1029, "dropout_reads_logged": 11169, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 120.62203979492188, "n_util_obs": 154, "util_count_spearman": 0.08541856625348086}, "probe_overhead_sec_total": 51.596, "probe_overhead_seq_frac": 0.4993, "util_loss": 0.00544, "alpha": 0.03, "t": 283.2}
+{"event": "consolidate", "step": 750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 296.1}
+{"event": "train", "step": 750, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.76968, "bpb": 2.55311, "chunk_loss": 0.030883, "F": 0.21011, "G": 0.2588, "chunks": 215.16, "bytes_per_chunk": 4.759, "boundary_entropy": 0.26005, "router_noise": 0.0, "grad_norm": 1.022, "lr": 0.0002959, "router_trainable": true, "sec_per_step": 0.263, "probe_sec": 0.0501, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04017, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.08935, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -10.73584, "trunc_abs_frac": 1.05679, "mean_delta_full": 0.00826, "read_hit_rate": {"split": 0.94319}, "mean_gate": {"split": 0.23073}, "n_reads": 11954, "n_cand": 12674, "read_suppressed_frac": 0.1017, "dropout_reads_logged": 11954, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 129.96250915527344, "n_util_obs": 160, "util_count_spearman": 0.11289698816360014}, "probe_overhead_sec_total": 55.218, "probe_overhead_seq_frac": 0.4993, "util_loss": 0.00625, "alpha": 0.03, "t": 296.3}
+{"event": "consolidate", "step": 800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 310.3}
+{"event": "probe", "step": 800, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3670.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.210883, "cps": 28.671875, "mean_p": 0.247744, "boundary_entropy": 0.236732, "bpb": 2.396273}, "tr": {"chunks": 3407.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.188201, "cps": 26.617188, "mean_p": 0.268618, "boundary_entropy": 0.334096, "bpb": 2.709593}, "fi": {"chunks": 4618.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.245495, "cps": 36.078125, "mean_p": 0.284597, "boundary_entropy": 0.28642, "bpb": 2.935624}}, "bpb": {"en": 2.39627, "tr": 2.70959, "fi": 2.93562}, "gini_cpc": 0.059257, "gini_cps": 0.069032, "bpic": 4.850962, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.8924412546372176, "fi": 1.1641263003834068}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 139.06015014648438, "n_util_obs": 168, "util_count_spearman": 0.15167476859983536}, "t": 311.1}
+{"event": "eval", "step": 800, "phase": 1, "bpb": {"en": 2.41121, "tr": 2.72385, "fi": 2.93464}, "bytes_per_chunk": {"en": 4.519, "tr": 5.862, "fi": 4.281}, "bpb_mean": 2.6899, "frozen_inventory": true, "t": 312.8}
+{"event": "train", "step": 800, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.77821, "bpb": 2.56541, "chunk_loss": 0.031108, "F": 0.21756, "G": 0.26332, "chunks": 222.78, "bytes_per_chunk": 4.596, "boundary_entropy": 0.25906, "router_noise": 0.0, "grad_norm": 0.9601, "lr": 0.0002949, "router_trainable": true, "sec_per_step": 0.36, "probe_sec": 0.1521, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.14197, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.04098, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 1.06393, "trunc_abs_frac": 2.65017, "mean_delta_full": 0.03088, "read_hit_rate": {"split": 0.94319}, "mean_gate": {"split": 0.22612}, "n_reads": 12235, "n_cand": 12972, "read_suppressed_frac": 0.1034, "dropout_reads_logged": 12235, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 139.2471923828125, "n_util_obs": 168, "util_count_spearman": 0.15241125045637718}, "probe_overhead_sec_total": 59.039, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.007138, "alpha": 0.03, "t": 314.4}
+{"event": "consolidate", "step": 850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 330.6}
+{"event": "train", "step": 850, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.70381, "bpb": 2.45808, "chunk_loss": 0.030947, "F": 0.21863, "G": 0.2486, "chunks": 223.88, "bytes_per_chunk": 4.574, "boundary_entropy": 0.22659, "router_noise": 0.0, "grad_norm": 1.1489, "lr": 0.0002938, "router_trainable": true, "sec_per_step": 0.33, "probe_sec": 0.0489, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03397, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.05141, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 0, "read_hit_rate": {"split": 0.95643}, "mean_gate": {"split": 0.24368}, "n_reads": 12709, "n_cand": 13288, "read_suppressed_frac": 0.0989, "dropout_reads_logged": 12709, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 148.5414276123047, "n_util_obs": 180, "util_count_spearman": 0.12677963311624843}, "probe_overhead_sec_total": 62.988, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.00754, "alpha": 0.03, "t": 330.8}
+{"event": "consolidate", "step": 900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 344.9}
+{"event": "train", "step": 900, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.68955, "bpb": 2.43751, "chunk_loss": 0.030869, "F": 0.21335, "G": 0.25223, "chunks": 218.47, "bytes_per_chunk": 4.687, "boundary_entropy": 0.23494, "router_noise": 0.0, "grad_norm": 0.9678, "lr": 0.0002927, "router_trainable": true, "sec_per_step": 0.287, "probe_sec": 0.0453, "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": 14, "mean_delta_loo": 0.09262, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 1.50741, "trunc_abs_frac": 2.22084, "mean_delta_full": 0.16823, "read_hit_rate": {"split": 0.9505}, "mean_gate": {"split": 0.24793}, "n_reads": 12347, "n_cand": 12990, "read_suppressed_frac": 0.1032, "dropout_reads_logged": 12347, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 157.85409545898438, "n_util_obs": 185, "util_count_spearman": 0.12065691217163868}, "probe_overhead_sec_total": 66.425, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.007918, "alpha": 0.03, "t": 345.2}
+{"event": "consolidate", "step": 950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 359.3}
+{"event": "train", "step": 950, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.70232, "bpb": 2.45593, "chunk_loss": 0.030812, "F": 0.21161, "G": 0.24833, "chunks": 216.69, "bytes_per_chunk": 4.726, "boundary_entropy": 0.22625, "router_noise": 0.0, "grad_norm": 1.0549, "lr": 0.0002914, "router_trainable": true, "sec_per_step": 0.287, "probe_sec": 0.1106, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09939, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.14628, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.35022, "trunc_abs_frac": 1.77421, "mean_delta_full": 0.05801, "read_hit_rate": {"split": 0.94707}, "mean_gate": {"split": 0.25607}, "n_reads": 12257, "n_cand": 12942, "read_suppressed_frac": 0.1008, "dropout_reads_logged": 12257, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 167.1678924560547, "n_util_obs": 193, "util_count_spearman": 0.13812046563754285}, "probe_overhead_sec_total": 70.282, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.008256, "alpha": 0.03, "t": 359.5}
+{"event": "consolidate", "step": 1000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 374.1}
+{"event": "probe", "step": 1000, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3660.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.210309, "cps": 28.59375, "mean_p": 0.240397, "boundary_entropy": 0.221515, "bpb": 2.307924}, "tr": {"chunks": 3493.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.192951, "cps": 27.289062, "mean_p": 0.268477, "boundary_entropy": 0.332696, "bpb": 2.605325}, "fi": {"chunks": 4616.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.245388, "cps": 36.0625, "mean_p": 0.272688, "boundary_entropy": 0.266135, "bpb": 2.808804}}, "bpb": {"en": 2.30792, "tr": 2.60532, "fi": 2.8088}, "gini_cpc": 0.053893, "gini_cps": 0.063613, "bpic": 4.820461, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9174683029621935, "fi": 1.166801426789871}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 176.3306884765625, "n_util_obs": 203, "util_count_spearman": 0.11773312944160252}, "t": 375.1}
+{"event": "eval", "step": 1000, "phase": 1, "bpb": {"en": 2.32176, "tr": 2.61573, "fi": 2.8109}, "bytes_per_chunk": {"en": 4.518, "tr": 5.681, "fi": 4.324}, "bpb_mean": 2.5828, "frozen_inventory": true, "t": 376.8}
+{"event": "train", "step": 1000, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.70046, "bpb": 2.45324, "chunk_loss": 0.030977, "F": 0.21777, "G": 0.25359, "chunks": 223.0, "bytes_per_chunk": 4.592, "boundary_entropy": 0.23721, "router_noise": 0.0, "grad_norm": 0.8834, "lr": 0.0002901, "router_trainable": true, "sec_per_step": 0.378, "probe_sec": 0.0542, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03659, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.00808, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 3.72556, "trunc_abs_frac": 2.72476, "mean_delta_full": -0.01831, "read_hit_rate": {"split": 0.94552}, "mean_gate": {"split": 0.2436}, "n_reads": 12339, "n_cand": 13050, "read_suppressed_frac": 0.0989, "dropout_reads_logged": 12339, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 176.51962280273438, "n_util_obs": 203, "util_count_spearman": 0.11774106147345818}, "probe_overhead_sec_total": 73.82, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.008442, "alpha": 0.03, "t": 378.4}
+{"event": "probe", "step": 1024, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3651.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.209791, "cps": 28.523438, "mean_p": 0.240305, "boundary_entropy": 0.205218, "bpb": 2.292961}, "tr": {"chunks": 3432.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.189582, "cps": 26.8125, "mean_p": 0.264809, "boundary_entropy": 0.321099, "bpb": 2.588784}, "fi": {"chunks": 4693.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.249482, "cps": 36.664062, "mean_p": 0.282445, "boundary_entropy": 0.264846, "bpb": 2.776091}}, "bpb": {"en": 2.29296, "tr": 2.58878, "fi": 2.77609}, "gini_cpc": 0.061544, "gini_cps": 0.071388, "bpic": 4.817595, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9036682316940073, "fi": 1.1891892045367431}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 180.79147338867188, "n_util_obs": 212, "util_count_spearman": 0.11027930976721458}, "t": 385.8}
+{"event": "consolidate", "step": 1050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 394.3}
+{"event": "train", "step": 1050, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.72822, "bpb": 2.49329, "chunk_loss": 0.03079, "F": 0.21054, "G": 0.2468, "chunks": 215.59, "bytes_per_chunk": 4.75, "boundary_entropy": 0.22861, "router_noise": 0.0, "grad_norm": 1.0057, "lr": 0.0002886, "router_trainable": true, "sec_per_step": 0.325, "probe_sec": 0.0559, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0415, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.1688, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.72548, "trunc_abs_frac": 0.59412, "mean_delta_full": 0.70407, "read_hit_rate": {"split": 0.93881}, "mean_gate": {"split": 0.24127}, "n_reads": 11891, "n_cand": 12666, "read_suppressed_frac": 0.0997, "dropout_reads_logged": 11891, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 185.80801391601562, "n_util_obs": 218, "util_count_spearman": 0.0955104886923898}, "probe_overhead_sec_total": 77.516, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.008554, "alpha": 0.03, "t": 394.7}
+{"event": "consolidate", "step": 1100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 408.5}
+{"event": "train", "step": 1100, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.70158, "bpb": 2.45486, "chunk_loss": 0.030873, "F": 0.21542, "G": 0.2465, "chunks": 220.59, "bytes_per_chunk": 4.642, "boundary_entropy": 0.21184, "router_noise": 0.0, "grad_norm": 0.8586, "lr": 0.0002871, "router_trainable": true, "sec_per_step": 0.293, "probe_sec": 0.1281, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10715, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.01712, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.82418, "trunc_abs_frac": 0.98622, "mean_delta_full": -0.05115, "read_hit_rate": {"split": 0.9473}, "mean_gate": {"split": 0.25402}, "n_reads": 12385, "n_cand": 13074, "read_suppressed_frac": 0.0993, "dropout_reads_logged": 12385, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 195.15151977539062, "n_util_obs": 224, "util_count_spearman": 0.08892594967220446}, "probe_overhead_sec_total": 81.046, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.008769, "alpha": 0.03, "t": 409.3}
+{"event": "consolidate", "step": 1150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 422.9}
+{"event": "train", "step": 1150, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.71386, "bpb": 2.47257, "chunk_loss": 0.030911, "F": 0.21155, "G": 0.25583, "chunks": 216.62, "bytes_per_chunk": 4.727, "boundary_entropy": 0.2373, "router_noise": 0.0, "grad_norm": 0.8398, "lr": 0.0002855, "router_trainable": true, "sec_per_step": 0.277, "probe_sec": 0.0545, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03859, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.00745, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.96852, "trunc_abs_frac": 2.01062, "mean_delta_full": 0.27393, "read_hit_rate": {"split": 0.93484}, "mean_gate": {"split": 0.23989}, "n_reads": 11794, "n_cand": 12616, "read_suppressed_frac": 0.0998, "dropout_reads_logged": 11794, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 204.4838104248047, "n_util_obs": 236, "util_count_spearman": 0.0962386376034088}, "probe_overhead_sec_total": 84.772, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.009163, "alpha": 0.03, "t": 423.2}
+{"event": "consolidate", "step": 1200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 437.4}
+{"event": "probe", "step": 1200, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3631.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.208642, "cps": 28.367188, "mean_p": 0.229448, "boundary_entropy": 0.190542, "bpb": 2.240068}, "tr": {"chunks": 3354.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.185273, "cps": 26.203125, "mean_p": 0.25587, "boundary_entropy": 0.309442, "bpb": 2.526578}, "fi": {"chunks": 4537.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.241189, "cps": 35.445312, "mean_p": 0.262478, "boundary_entropy": 0.23606, "bpb": 2.694004}}, "bpb": {"en": 2.24007, "tr": 2.52658, "fi": 2.694}, "gini_cpc": 0.058694, "gini_cps": 0.068449, "bpic": 4.923798, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.8879947090604723, "fi": 1.1559918364037542}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 213.5907745361328, "n_util_obs": 243, "util_count_spearman": 0.08148716842209142}, "t": 438.3}
+{"event": "eval", "step": 1200, "phase": 1, "bpb": {"en": 2.25293, "tr": 2.53456, "fi": 2.69929}, "bytes_per_chunk": {"en": 4.539, "tr": 5.871, "fi": 4.382}, "bpb_mean": 2.49559, "frozen_inventory": true, "t": 439.5}
+{"event": "train", "step": 1200, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.61168, "bpb": 2.32516, "chunk_loss": 0.030679, "F": 0.20813, "G": 0.2438, "chunks": 213.12, "bytes_per_chunk": 4.805, "boundary_entropy": 0.21918, "router_noise": 0.0, "grad_norm": 0.8871, "lr": 0.0002837, "router_trainable": true, "sec_per_step": 0.352, "probe_sec": 0.0591, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0458, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.15267, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.81667, "trunc_abs_frac": 0.74589, "mean_delta_full": 0.75159, "read_hit_rate": {"split": 0.94248}, "mean_gate": {"split": 0.25716}, "n_reads": 11928, "n_cand": 12656, "read_suppressed_frac": 0.0981, "dropout_reads_logged": 11928, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 213.77320861816406, "n_util_obs": 243, "util_count_spearman": 0.08305109445404181}, "probe_overhead_sec_total": 88.35, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.008904, "alpha": 0.03, "t": 440.8}
+{"event": "consolidate", "step": 1250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 455.1}
+{"event": "train", "step": 1250, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.65369, "bpb": 2.38577, "chunk_loss": 0.03079, "F": 0.21027, "G": 0.24927, "chunks": 215.31, "bytes_per_chunk": 4.756, "boundary_entropy": 0.21442, "router_noise": 0.0, "grad_norm": 0.9043, "lr": 0.0002819, "router_trainable": true, "sec_per_step": 0.291, "probe_sec": 0.0389, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02979, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.04448, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.65061, "trunc_abs_frac": 0.89149, "mean_delta_full": -0.36523, "read_hit_rate": {"split": 0.93889}, "mean_gate": {"split": 0.25449}, "n_reads": 11800, "n_cand": 12568, "read_suppressed_frac": 0.1043, "dropout_reads_logged": 11800, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 223.05841064453125, "n_util_obs": 247, "util_count_spearman": 0.08534435252932913}, "probe_overhead_sec_total": 91.995, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.010015, "alpha": 0.03, "t": 455.3}
+{"event": "consolidate", "step": 1300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 469.2}
+{"event": "train", "step": 1300, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.57097, "bpb": 2.26643, "chunk_loss": 0.030811, "F": 0.21451, "G": 0.24527, "chunks": 219.66, "bytes_per_chunk": 4.662, "boundary_entropy": 0.20046, "router_noise": 0.0, "grad_norm": 0.8037, "lr": 0.00028, "router_trainable": true, "sec_per_step": 0.282, "probe_sec": 0.0509, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03995, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.07997, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.84156, "trunc_abs_frac": 12.78436, "mean_delta_full": 0.15043, "read_hit_rate": {"split": 0.95213}, "mean_gate": {"split": 0.25947}, "n_reads": 12490, "n_cand": 13118, "read_suppressed_frac": 0.1034, "dropout_reads_logged": 12490, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 232.39093017578125, "n_util_obs": 257, "util_count_spearman": 0.10508522984345309}, "probe_overhead_sec_total": 95.416, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.010285, "alpha": 0.03, "t": 469.4}
+{"event": "consolidate", "step": 1350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 485.3}
+{"event": "train", "step": 1350, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.6003, "bpb": 2.30875, "chunk_loss": 0.030864, "F": 0.21915, "G": 0.24079, "chunks": 224.41, "bytes_per_chunk": 4.563, "boundary_entropy": 0.18261, "router_noise": 0.0, "grad_norm": 0.9206, "lr": 0.000278, "router_trainable": true, "sec_per_step": 0.323, "probe_sec": 0.0478, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03598, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.06006, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.66883, "trunc_abs_frac": 0.68287, "mean_delta_full": 0.19733, "read_hit_rate": {"split": 0.95642}, "mean_gate": {"split": 0.2681}, "n_reads": 12663, "n_cand": 13240, "read_suppressed_frac": 0.1015, "dropout_reads_logged": 12663, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 241.71746826171875, "n_util_obs": 262, "util_count_spearman": 0.0939494586909247}, "probe_overhead_sec_total": 99.582, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.009834, "alpha": 0.03, "t": 485.6}
+{"event": "consolidate", "step": 1400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 499.2}
+{"event": "probe", "step": 1400, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3613.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.207608, "cps": 28.226562, "mean_p": 0.225819, "boundary_entropy": 0.175922, "bpb": 2.18571}, "tr": {"chunks": 3588.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.198199, "cps": 28.03125, "mean_p": 0.26136, "boundary_entropy": 0.304045, "bpb": 2.463782}, "fi": {"chunks": 4605.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.244804, "cps": 35.976562, "mean_p": 0.263117, "boundary_entropy": 0.225427, "bpb": 2.622411}}, "bpb": {"en": 2.18571, "tr": 2.46378, "fi": 2.62241}, "gini_cpc": 0.047754, "gini_cps": 0.057428, "bpic": 4.805353, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9546804773172745, "fi": 1.1791631802081282}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 250.880859375, "n_util_obs": 268, "util_count_spearman": 0.09077045004158447}, "t": 500.0}
+{"event": "eval", "step": 1400, "phase": 1, "bpb": {"en": 2.19961, "tr": 2.47271, "fi": 2.6273}, "bytes_per_chunk": {"en": 4.573, "tr": 5.551, "fi": 4.326}, "bpb_mean": 2.4332, "frozen_inventory": true, "t": 501.4}
+{"event": "train", "step": 1400, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.56498, "bpb": 2.25779, "chunk_loss": 0.030719, "F": 0.21216, "G": 0.23834, "chunks": 217.25, "bytes_per_chunk": 4.713, "boundary_entropy": 0.19775, "router_noise": 0.0, "grad_norm": 0.8292, "lr": 0.0002759, "router_trainable": true, "sec_per_step": 0.341, "probe_sec": 0.0755, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06641, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.04537, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.02209, "trunc_abs_frac": 0.37201, "mean_delta_full": 0.09883, "read_hit_rate": {"split": 0.95037}, "mean_gate": {"split": 0.26272}, "n_reads": 12294, "n_cand": 12936, "read_suppressed_frac": 0.0996, "dropout_reads_logged": 12294, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 251.06922912597656, "n_util_obs": 268, "util_count_spearman": 0.09165192790485566}, "probe_overhead_sec_total": 103.133, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.010072, "alpha": 0.03, "t": 502.6}
+{"event": "consolidate", "step": 1450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 516.2}
+{"event": "train", "step": 1450, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.55205, "bpb": 2.23913, "chunk_loss": 0.03072, "F": 0.20938, "G": 0.24472, "chunks": 214.41, "bytes_per_chunk": 4.776, "boundary_entropy": 0.20325, "router_noise": 0.0, "grad_norm": 0.8844, "lr": 0.0002738, "router_trainable": true, "sec_per_step": 0.28, "probe_sec": 0.1117, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09531, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.09475, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.0335, "trunc_abs_frac": 0.85154, "mean_delta_full": -0.01245, "read_hit_rate": {"split": 0.94689}, "mean_gate": {"split": 0.25917}, "n_reads": 11946, "n_cand": 12616, "read_suppressed_frac": 0.1018, "dropout_reads_logged": 11946, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 260.360595703125, "n_util_obs": 274, "util_count_spearman": 0.0759972873841487}, "probe_overhead_sec_total": 106.769, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.01067, "alpha": 0.03, "t": 516.6}
+{"event": "consolidate", "step": 1500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 530.8}
+{"event": "train", "step": 1500, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.53155, "bpb": 2.20956, "chunk_loss": 0.030744, "F": 0.21149, "G": 0.23827, "chunks": 216.56, "bytes_per_chunk": 4.728, "boundary_entropy": 0.18971, "router_noise": 0.0, "grad_norm": 0.891, "lr": 0.0002715, "router_trainable": true, "sec_per_step": 0.289, "probe_sec": 0.0402, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0281, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.15015, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": -0.14469, "trunc_abs_frac": 0.14469, "mean_delta_full": -0.35718, "read_hit_rate": {"split": 0.95222}, "mean_gate": {"split": 0.26503}, "n_reads": 12097, "n_cand": 12704, "read_suppressed_frac": 0.0947, "dropout_reads_logged": 12097, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 269.66217041015625, "n_util_obs": 282, "util_count_spearman": 0.07984737720030449}, "probe_overhead_sec_total": 110.322, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.01081, "alpha": 0.03, "t": 531.1}
+{"event": "consolidate", "step": 1550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 546.7}
+{"event": "train", "step": 1550, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.52558, "bpb": 2.20095, "chunk_loss": 0.030736, "F": 0.21436, "G": 0.23853, "chunks": 219.5, "bytes_per_chunk": 4.665, "boundary_entropy": 0.17831, "router_noise": 0.0, "grad_norm": 0.7271, "lr": 0.0002692, "router_trainable": true, "sec_per_step": 0.318, "probe_sec": 0.0531, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04192, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.03531, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.10776, "trunc_abs_frac": 1.23907, "mean_delta_full": -0.22888, "read_hit_rate": {"split": 0.9534}, "mean_gate": {"split": 0.2591}, "n_reads": 12379, "n_cand": 12984, "read_suppressed_frac": 0.1017, "dropout_reads_logged": 12379, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 279.0330505371094, "n_util_obs": 292, "util_count_spearman": 0.060115453424061924}, "probe_overhead_sec_total": 113.798, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.010594, "alpha": 0.03, "t": 547.0}
+{"event": "consolidate", "step": 1600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 560.8}
+{"event": "probe", "step": 1600, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3645.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.209447, "cps": 28.476562, "mean_p": 0.226641, "boundary_entropy": 0.163858, "bpb": 2.147278}, "tr": {"chunks": 3413.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.188532, "cps": 26.664062, "mean_p": 0.255607, "boundary_entropy": 0.297932, "bpb": 2.41452}, "fi": {"chunks": 4614.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.245282, "cps": 36.046875, "mean_p": 0.271682, "boundary_entropy": 0.225849, "bpb": 2.562974}}, "bpb": {"en": 2.14728, "tr": 2.41452, "fi": 2.56297}, "gini_cpc": 0.058815, "gini_cps": 0.068597, "bpic": 4.860521, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9001446910215868, "fi": 1.1710954517681078}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 288.18536376953125, "n_util_obs": 298, "util_count_spearman": 0.07467575457743765}, "t": 561.8}
+{"event": "eval", "step": 1600, "phase": 1, "bpb": {"en": 2.1639, "tr": 2.42405, "fi": 2.57107}, "bytes_per_chunk": {"en": 4.554, "tr": 5.838, "fi": 4.318}, "bpb_mean": 2.38634, "frozen_inventory": true, "t": 563.4}
+{"event": "train", "step": 1600, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.51504, "bpb": 2.18574, "chunk_loss": 0.030536, "F": 0.20337, "G": 0.23777, "chunks": 208.25, "bytes_per_chunk": 4.917, "boundary_entropy": 0.19716, "router_noise": 0.0, "grad_norm": 0.7766, "lr": 0.0002668, "router_trainable": true, "sec_per_step": 0.353, "probe_sec": 0.0463, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03754, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.15324, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.37873, "trunc_abs_frac": 0.30567, "mean_delta_full": 0.11877, "read_hit_rate": {"split": 0.94306}, "mean_gate": {"split": 0.26086}, "n_reads": 11843, "n_cand": 12558, "read_suppressed_frac": 0.1012, "dropout_reads_logged": 11843, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 288.3670654296875, "n_util_obs": 298, "util_count_spearman": 0.07447467955216339}, "probe_overhead_sec_total": 117.194, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.013622, "alpha": 0.03, "t": 564.6}
+{"event": "consolidate", "step": 1650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 579.0}
+{"event": "train", "step": 1650, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.5638, "bpb": 2.25609, "chunk_loss": 0.030722, "F": 0.2128, "G": 0.23653, "chunks": 217.91, "bytes_per_chunk": 4.699, "boundary_entropy": 0.1818, "router_noise": 0.0, "grad_norm": 0.778, "lr": 0.0002643, "router_trainable": true, "sec_per_step": 0.292, "probe_sec": 0.1117, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09876, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.12071, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.85892, "trunc_abs_frac": 3.42416, "mean_delta_full": 0.32841, "read_hit_rate": {"split": 0.94846}, "mean_gate": {"split": 0.25892}, "n_reads": 12201, "n_cand": 12864, "read_suppressed_frac": 0.1034, "dropout_reads_logged": 12201, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 297.6903076171875, "n_util_obs": 304, "util_count_spearman": 0.11852788701762032}, "probe_overhead_sec_total": 120.946, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.013574, "alpha": 0.03, "t": 579.2}
+{"event": "consolidate", "step": 1700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 592.8}
+{"event": "train", "step": 1700, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.56429, "bpb": 2.25679, "chunk_loss": 0.030549, "F": 0.20432, "G": 0.23217, "chunks": 209.22, "bytes_per_chunk": 4.894, "boundary_entropy": 0.18531, "router_noise": 0.0, "grad_norm": 0.8476, "lr": 0.0002618, "router_trainable": true, "sec_per_step": 0.277, "probe_sec": 0.0601, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04987, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.13307, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.42728, "trunc_abs_frac": 2.51889, "mean_delta_full": 0.28147, "read_hit_rate": {"split": 0.94215}, "mean_gate": {"split": 0.26611}, "n_reads": 11775, "n_cand": 12498, "read_suppressed_frac": 0.0976, "dropout_reads_logged": 11775, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 307.01531982421875, "n_util_obs": 313, "util_count_spearman": 0.1300874270128969}, "probe_overhead_sec_total": 124.552, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.012126, "alpha": 0.03, "t": 593.1}
+{"event": "consolidate", "step": 1750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 607.0}
+{"event": "train", "step": 1750, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.52994, "bpb": 2.20724, "chunk_loss": 0.030666, "F": 0.21072, "G": 0.23242, "chunks": 215.78, "bytes_per_chunk": 4.746, "boundary_entropy": 0.16575, "router_noise": 0.0, "grad_norm": 0.7514, "lr": 0.0002591, "router_trainable": true, "sec_per_step": 0.283, "probe_sec": 0.0479, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03777, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.00891, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 8, "trunc_ratio_sum": 5.82721, "trunc_abs_frac": 0.86693, "mean_delta_full": -0.00104, "read_hit_rate": {"split": 0.94835}, "mean_gate": {"split": 0.26586}, "n_reads": 12046, "n_cand": 12702, "read_suppressed_frac": 0.0991, "dropout_reads_logged": 12046, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 316.365966796875, "n_util_obs": 317, "util_count_spearman": 0.13834622154540127}, "probe_overhead_sec_total": 128.21, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.01195, "alpha": 0.03, "t": 607.3}
+{"event": "consolidate", "step": 1800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 619.9}
+{"event": "probe", "step": 1800, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3634.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.208815, "cps": 28.390625, "mean_p": 0.225913, "boundary_entropy": 0.155817, "bpb": 2.113626}, "tr": {"chunks": 3557.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.196487, "cps": 27.789062, "mean_p": 0.254891, "boundary_entropy": 0.282855, "bpb": 2.359305}, "fi": {"chunks": 4613.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.245229, "cps": 36.039062, "mean_p": 0.266136, "boundary_entropy": 0.196774, "bpb": 2.5272}}, "bpb": {"en": 2.11363, "tr": 2.3593, "fi": 2.5272}, "gini_cpc": 0.049951, "gini_cps": 0.059641, "bpic": 4.806167, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9409629226461156, "fi": 1.174385737896716}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 325.4422302246094, "n_util_obs": 324, "util_count_spearman": 0.15350005323137433}, "t": 620.9}
+{"event": "eval", "step": 1800, "phase": 1, "bpb": {"en": 2.12917, "tr": 2.3676, "fi": 2.53589}, "bytes_per_chunk": {"en": 4.549, "tr": 5.701, "fi": 4.317}, "bpb_mean": 2.34422, "frozen_inventory": true, "t": 622.4}
+{"event": "train", "step": 1800, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.57456, "bpb": 2.27161, "chunk_loss": 0.030792, "F": 0.21384, "G": 0.24387, "chunks": 218.97, "bytes_per_chunk": 4.676, "boundary_entropy": 0.18477, "router_noise": 0.0, "grad_norm": 0.8223, "lr": 0.0002564, "router_trainable": true, "sec_per_step": 0.331, "probe_sec": 0.0528, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03909, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.0569, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 0, "read_hit_rate": {"split": 0.93723}, "mean_gate": {"split": 0.25928}, "n_reads": 11974, "n_cand": 12776, "read_suppressed_frac": 0.1031, "dropout_reads_logged": 11974, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 325.62225341796875, "n_util_obs": 324, "util_count_spearman": 0.152594899367125}, "probe_overhead_sec_total": 131.717, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.011799, "alpha": 0.03, "t": 623.8}
+{"event": "consolidate", "step": 1850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 637.5}
+{"event": "train", "step": 1850, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.53013, "bpb": 2.20751, "chunk_loss": 0.030839, "F": 0.21655, "G": 0.23957, "chunks": 221.75, "bytes_per_chunk": 4.618, "boundary_entropy": 0.17408, "router_noise": 0.0, "grad_norm": 0.7261, "lr": 0.0002537, "router_trainable": true, "sec_per_step": 0.281, "probe_sec": 0.0445, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03363, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.04397, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": -7.57991, "trunc_abs_frac": 7.57991, "mean_delta_full": -0.02826, "read_hit_rate": {"split": 0.95348}, "mean_gate": {"split": 0.25577}, "n_reads": 12523, "n_cand": 13134, "read_suppressed_frac": 0.0996, "dropout_reads_logged": 12523, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 334.92669677734375, "n_util_obs": 334, "util_count_spearman": 0.11544301473696415}, "probe_overhead_sec_total": 135.432, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.013715, "alpha": 0.03, "t": 637.9}
+{"event": "consolidate", "step": 1900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 652.0}
+{"event": "train", "step": 1900, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.5654, "bpb": 2.25839, "chunk_loss": 0.030659, "F": 0.21094, "G": 0.23276, "chunks": 216.0, "bytes_per_chunk": 4.741, "boundary_entropy": 0.16476, "router_noise": 0.0, "grad_norm": 0.7049, "lr": 0.0002508, "router_trainable": true, "sec_per_step": 0.288, "probe_sec": 0.0814, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06767, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.13241, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 1.34329, "trunc_abs_frac": 1.25312, "mean_delta_full": 0.13569, "read_hit_rate": {"split": 0.94345}, "mean_gate": {"split": 0.26629}, "n_reads": 12163, "n_cand": 12892, "read_suppressed_frac": 0.0961, "dropout_reads_logged": 12163, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 344.2469482421875, "n_util_obs": 339, "util_count_spearman": 0.12259579750713866}, "probe_overhead_sec_total": 139.083, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.012098, "alpha": 0.03, "t": 652.3}
+{"event": "consolidate", "step": 1950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 665.7}
+{"event": "train", "step": 1950, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.52845, "bpb": 2.20509, "chunk_loss": 0.030713, "F": 0.21411, "G": 0.23321, "chunks": 219.25, "bytes_per_chunk": 4.67, "boundary_entropy": 0.1605, "router_noise": 0.0, "grad_norm": 0.7131, "lr": 0.0002479, "router_trainable": true, "sec_per_step": 0.274, "probe_sec": 0.0525, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03998, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.0659, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.21569, "trunc_abs_frac": 0.83314, "mean_delta_full": -0.21261, "read_hit_rate": {"split": 0.94689}, "mean_gate": {"split": 0.25466}, "n_reads": 12266, "n_cand": 12954, "read_suppressed_frac": 0.0974, "dropout_reads_logged": 12266, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 353.5865783691406, "n_util_obs": 348, "util_count_spearman": 0.1320650008193382}, "probe_overhead_sec_total": 142.994, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.015622, "alpha": 0.03, "t": 666.0}
+{"event": "consolidate", "step": 2000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 680.0}
+{"event": "probe", "step": 2000, "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": 0.223275, "boundary_entropy": 0.140918, "bpb": 2.086231}, "tr": {"chunks": 3357.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.185439, "cps": 26.226562, "mean_p": 0.245205, "boundary_entropy": 0.269084, "bpb": 2.331506}, "fi": {"chunks": 4609.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.245016, "cps": 36.007812, "mean_p": 0.265338, "boundary_entropy": 0.189459, "bpb": 2.465858}}, "bpb": {"en": 2.08623, "tr": 2.33151, "fi": 2.46586}, "gini_cpc": 0.062164, "gini_cps": 0.071991, "bpic": 4.893221, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.8895239210931981, "fi": 1.1753079302106717}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 362.717529296875, "n_util_obs": 358, "util_count_spearman": 0.13472214639422023}, "t": 680.9}
+{"event": "eval", "step": 2000, "phase": 1, "bpb": {"en": 2.10167, "tr": 2.33756, "fi": 2.47191}, "bytes_per_chunk": {"en": 4.566, "tr": 5.948, "fi": 4.329}, "bpb_mean": 2.30371, "frozen_inventory": true, "t": 682.5}
+{"event": "train", "step": 2000, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.54197, "bpb": 2.22459, "chunk_loss": 0.030809, "F": 0.21472, "G": 0.2382, "chunks": 219.88, "bytes_per_chunk": 4.657, "boundary_entropy": 0.16387, "router_noise": 0.0, "grad_norm": 0.7569, "lr": 0.000245, "router_trainable": true, "sec_per_step": 0.355, "probe_sec": 0.0593, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0489, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.16447, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 1.43179, "trunc_abs_frac": 1.90526, "mean_delta_full": 0.15167, "read_hit_rate": {"split": 0.94581}, "mean_gate": {"split": 0.25537}, "n_reads": 12044, "n_cand": 12734, "read_suppressed_frac": 0.0987, "dropout_reads_logged": 12044, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 362.9023132324219, "n_util_obs": 358, "util_count_spearman": 0.13476216178576136}, "probe_overhead_sec_total": 146.898, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.016746, "alpha": 0.03, "t": 683.7}
+{"event": "probe", "step": 2048, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3618.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.207895, "cps": 28.265625, "mean_p": 0.221593, "boundary_entropy": 0.141438, "bpb": 2.078983}, "tr": {"chunks": 3456.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.190908, "cps": 27.0, "mean_p": 0.246804, "boundary_entropy": 0.264062, "bpb": 2.327389}, "fi": {"chunks": 4559.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.242358, "cps": 35.617188, "mean_p": 0.259947, "boundary_entropy": 0.179528, "bpb": 2.450823}}, "bpb": {"en": 2.07898, "tr": 2.32739, "fi": 2.45082}, "gini_cpc": 0.053497, "gini_cps": 0.063211, "bpic": 4.876816, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9182876426023229, "fi": 1.165771050829174}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 371.65069580078125, "n_util_obs": 363, "util_count_spearman": 0.1321546524990154}, "t": 697.3}
+{"event": "consolidate", "step": 2050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 699.2}
+{"event": "train", "step": 2050, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.60619, "bpb": 2.31725, "chunk_loss": 0.030638, "F": 0.2084, "G": 0.23768, "chunks": 213.41, "bytes_per_chunk": 4.798, "boundary_entropy": 0.17315, "router_noise": 0.0, "grad_norm": 0.8013, "lr": 0.0002419, "router_trainable": true, "sec_per_step": 0.316, "probe_sec": 0.1609, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.14973, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.10073, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 1.83665, "trunc_abs_frac": 1.83665, "mean_delta_full": 0.12256, "read_hit_rate": {"split": 0.93568}, "mean_gate": {"split": 0.25961}, "n_reads": 11666, "n_cand": 12468, "read_suppressed_frac": 0.1036, "dropout_reads_logged": 11666, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 372.200927734375, "n_util_obs": 364, "util_count_spearman": 0.13369060997416263}, "probe_overhead_sec_total": 150.624, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.017986, "alpha": 0.03, "t": 699.6}
+{"event": "consolidate", "step": 2100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 713.1}
+{"event": "train", "step": 2100, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.49024, "bpb": 2.14996, "chunk_loss": 0.030707, "F": 0.21317, "G": 0.23466, "chunks": 218.28, "bytes_per_chunk": 4.691, "boundary_entropy": 0.16832, "router_noise": 0.0, "grad_norm": 0.7509, "lr": 0.0002388, "router_trainable": true, "sec_per_step": 0.277, "probe_sec": 0.07, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0595, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.20604, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.33689, "trunc_abs_frac": 0.62946, "mean_delta_full": 0.41284, "read_hit_rate": {"split": 0.94908}, "mean_gate": {"split": 0.24937}, "n_reads": 12245, "n_cand": 12902, "read_suppressed_frac": 0.101, "dropout_reads_logged": 12245, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 381.4478759765625, "n_util_obs": 367, "util_count_spearman": 0.12289911070283414}, "probe_overhead_sec_total": 154.244, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.017932, "alpha": 0.03, "t": 713.4}
+{"event": "consolidate", "step": 2150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 726.9}
+{"event": "train", "step": 2150, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.50607, "bpb": 2.17281, "chunk_loss": 0.030607, "F": 0.20911, "G": 0.23033, "chunks": 214.12, "bytes_per_chunk": 4.782, "boundary_entropy": 0.15572, "router_noise": 0.0, "grad_norm": 0.7105, "lr": 0.0002357, "router_trainable": true, "sec_per_step": 0.276, "probe_sec": 0.058, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04802, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.0977, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.59364, "trunc_abs_frac": 0.72177, "mean_delta_full": -0.70361, "read_hit_rate": {"split": 0.94586}, "mean_gate": {"split": 0.2498}, "n_reads": 12073, "n_cand": 12764, "read_suppressed_frac": 0.102, "dropout_reads_logged": 12073, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 390.7186584472656, "n_util_obs": 375, "util_count_spearman": 0.09970394811696438}, "probe_overhead_sec_total": 158.132, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.018644, "alpha": 0.03, "t": 727.2}
+{"event": "consolidate", "step": 2200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 740.8}
+{"event": "probe", "step": 2200, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3618.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.207895, "cps": 28.265625, "mean_p": 0.221468, "boundary_entropy": 0.141068, "bpb": 2.05507}, "tr": {"chunks": 3356.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.185384, "cps": 26.21875, "mean_p": 0.242568, "boundary_entropy": 0.260312, "bpb": 2.288579}, "fi": {"chunks": 4559.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.242358, "cps": 35.617188, "mean_p": 0.258613, "boundary_entropy": 0.177634, "bpb": 2.427805}}, "bpb": {"en": 2.05507, "tr": 2.28858, "fi": 2.42781}, "gini_cpc": 0.059756, "gini_cps": 0.06954, "bpic": 4.919102, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.8917168196103574, "fi": 1.165771050829174}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 399.8116455078125, "n_util_obs": 384, "util_count_spearman": 0.09096250185239269}, "t": 741.6}
+{"event": "eval", "step": 2200, "phase": 1, "bpb": {"en": 2.07158, "tr": 2.29502, "fi": 2.43491}, "bytes_per_chunk": {"en": 4.574, "tr": 5.947, "fi": 4.36}, "bpb_mean": 2.26717, "frozen_inventory": true, "t": 743.1}
+{"event": "train", "step": 2200, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.47423, "bpb": 2.12686, "chunk_loss": 0.030808, "F": 0.21768, "G": 0.23716, "chunks": 222.91, "bytes_per_chunk": 4.594, "boundary_entropy": 0.15614, "router_noise": 0.0, "grad_norm": 0.6433, "lr": 0.0002325, "router_trainable": true, "sec_per_step": 0.345, "probe_sec": 0.0453, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0329, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.10477, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 2.21516, "trunc_abs_frac": 0.89687, "mean_delta_full": 0.08154, "read_hit_rate": {"split": 0.94895}, "mean_gate": {"split": 0.26455}, "n_reads": 12380, "n_cand": 13046, "read_suppressed_frac": 0.1029, "dropout_reads_logged": 12380, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 400.000732421875, "n_util_obs": 384, "util_count_spearman": 0.09100234933987848}, "probe_overhead_sec_total": 161.618, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.018368, "alpha": 0.03, "t": 744.5}
+{"event": "consolidate", "step": 2250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 757.6}
+{"event": "train", "step": 2250, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.50204, "bpb": 2.16699, "chunk_loss": 0.03062, "F": 0.21039, "G": 0.23082, "chunks": 215.44, "bytes_per_chunk": 4.753, "boundary_entropy": 0.1599, "router_noise": 0.0, "grad_norm": 0.6816, "lr": 0.0002293, "router_trainable": true, "sec_per_step": 0.268, "probe_sec": 0.0482, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03917, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.07865, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -0.07598, "trunc_abs_frac": 0.24622, "mean_delta_full": 0.32987, "read_hit_rate": {"split": 0.94635}, "mean_gate": {"split": 0.25824}, "n_reads": 12136, "n_cand": 12824, "read_suppressed_frac": 0.1059, "dropout_reads_logged": 12136, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 409.2915344238281, "n_util_obs": 387, "util_count_spearman": 0.1016046008114883}, "probe_overhead_sec_total": 165.212, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.015266, "alpha": 0.03, "t": 757.8}
+{"event": "consolidate", "step": 2300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 771.2}
+{"event": "train", "step": 2300, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.49049, "bpb": 2.15032, "chunk_loss": 0.030615, "F": 0.21082, "G": 0.22977, "chunks": 215.88, "bytes_per_chunk": 4.743, "boundary_entropy": 0.15668, "router_noise": 0.0, "grad_norm": 0.7679, "lr": 0.000226, "router_trainable": true, "sec_per_step": 0.272, "probe_sec": 0.0407, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03075, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.03801, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 1.14001, "trunc_abs_frac": 0.97531, "mean_delta_full": 0.01687, "read_hit_rate": {"split": 0.94807}, "mean_gate": {"split": 0.26325}, "n_reads": 12103, "n_cand": 12766, "read_suppressed_frac": 0.0972, "dropout_reads_logged": 12103, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 418.5865783691406, "n_util_obs": 393, "util_count_spearman": 0.09591716611837085}, "probe_overhead_sec_total": 168.752, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.014544, "alpha": 0.03, "t": 771.4}
+{"event": "consolidate", "step": 2350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 785.2}
+{"event": "train", "step": 2350, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.49364, "bpb": 2.15487, "chunk_loss": 0.030526, "F": 0.20148, "G": 0.23578, "chunks": 206.31, "bytes_per_chunk": 4.963, "boundary_entropy": 0.18171, "router_noise": 0.0, "grad_norm": 0.7818, "lr": 0.0002226, "router_trainable": true, "sec_per_step": 0.281, "probe_sec": 0.0573, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0343, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.02788, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 1.1005, "trunc_abs_frac": 1.1005, "mean_delta_full": 0.17249, "read_hit_rate": {"split": 0.93729}, "mean_gate": {"split": 0.2506}, "n_reads": 11523, "n_cand": 12294, "read_suppressed_frac": 0.1045, "dropout_reads_logged": 11523, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 427.86212158203125, "n_util_obs": 399, "util_count_spearman": 0.08077039331998337}, "probe_overhead_sec_total": 172.262, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.019065, "alpha": 0.03, "t": 785.5}
+{"event": "consolidate", "step": 2400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 798.3}
+{"event": "probe", "step": 2400, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3622.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.208125, "cps": 28.296875, "mean_p": 0.219686, "boundary_entropy": 0.128951, "bpb": 2.035567}, "tr": {"chunks": 3435.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.189748, "cps": 26.835938, "mean_p": 0.244404, "boundary_entropy": 0.256491, "bpb": 2.258632}, "fi": {"chunks": 4565.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.242677, "cps": 35.664062, "mean_p": 0.257699, "boundary_entropy": 0.165477, "bpb": 2.39986}}, "bpb": {"en": 2.03557, "tr": 2.25863, "fi": 2.39986}, "gini_cpc": 0.055088, "gini_cps": 0.06482, "bpic": 4.881432, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9116998097853034, "fi": 1.1660161686826271}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 436.97796630859375, "n_util_obs": 405, "util_count_spearman": 0.08043602658455114}, "t": 799.2}
+{"event": "eval", "step": 2400, "phase": 1, "bpb": {"en": 2.05137, "tr": 2.26397, "fi": 2.40935}, "bytes_per_chunk": {"en": 4.58, "tr": 5.831, "fi": 4.367}, "bpb_mean": 2.24157, "frozen_inventory": true, "t": 800.8}
+{"event": "phase_transition", "step": 2400, "phase": 2, "objective": "global", "router_trainable": true, "encoder_trainable": true, "frozen_router_params": 0, "t": 801.8}
+{"event": "train", "step": 2400, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.43907, "bpb": 2.07614, "chunk_loss": 0.030691, "F": 0.21347, "G": 0.23352, "chunks": 218.59, "bytes_per_chunk": 4.684, "boundary_entropy": 0.14729, "router_noise": 0.0, "grad_norm": 0.7086, "lr": 0.0002192, "router_trainable": true, "sec_per_step": 0.333, "probe_sec": 0.0705, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06077, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.07686, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.64251, "trunc_abs_frac": 1.23519, "mean_delta_full": 0.13249, "read_hit_rate": {"split": 0.95209}, "mean_gate": {"split": 0.25942}, "n_reads": 12419, "n_cand": 13044, "read_suppressed_frac": 0.0986, "dropout_reads_logged": 12419, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 437.168212890625, "n_util_obs": 406, "util_count_spearman": 0.08045564957952361}, "probe_overhead_sec_total": 175.795, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.017299, "alpha": 0.03, "t": 802.2}
+{"event": "consolidate", "step": 2450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 815.0}
+{"event": "train", "step": 2450, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.62839, "bpb": 2.34927, "chunk_loss": 0.03073, "F": 0.21249, "G": 0.23452, "chunks": 217.59, "bytes_per_chunk": 4.706, "boundary_entropy": 0.14818, "router_noise": 0.0, "grad_norm": 0.7128, "lr": 0.0002158, "router_trainable": true, "sec_per_step": 0.263, "probe_sec": 0.0481, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03748, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.03178, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 0.37478, "trunc_abs_frac": 0.82198, "mean_delta_full": 0.24567, "read_hit_rate": {"split": 0.9344}, "mean_gate": {"split": 0.25775}, "n_reads": 11893, "n_cand": 12728, "read_suppressed_frac": 0.0992, "dropout_reads_logged": 11893, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 446.44586181640625, "n_util_obs": 418, "util_count_spearman": 0.06607120052272496}, "probe_overhead_sec_total": 179.407, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.027341, "alpha": 0.03, "t": 815.3}
+{"event": "consolidate", "step": 2500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 828.9}
+{"event": "train", "step": 2500, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.47924, "bpb": 2.13409, "chunk_loss": 0.03062, "F": 0.21243, "G": 0.22842, "chunks": 217.53, "bytes_per_chunk": 4.707, "boundary_entropy": 0.13813, "router_noise": 0.0, "grad_norm": 0.7214, "lr": 0.0002123, "router_trainable": true, "sec_per_step": 0.277, "probe_sec": 0.0884, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07482, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.14087, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.08679, "trunc_abs_frac": 1.08162, "mean_delta_full": 0.2762, "read_hit_rate": {"split": 0.95147}, "mean_gate": {"split": 0.25314}, "n_reads": 12331, "n_cand": 12960, "read_suppressed_frac": 0.0992, "dropout_reads_logged": 12331, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 455.75848388671875, "n_util_obs": 426, "util_count_spearman": 0.08784737117931886}, "probe_overhead_sec_total": 182.772, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.023247, "alpha": 0.03, "t": 829.2}
+{"event": "consolidate", "step": 2550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 844.1}
+{"event": "train", "step": 2550, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.4676, "bpb": 2.11731, "chunk_loss": 0.030585, "F": 0.2066, "G": 0.23331, "chunks": 211.56, "bytes_per_chunk": 4.84, "boundary_entropy": 0.16316, "router_noise": 0.0, "grad_norm": 0.7021, "lr": 0.0002088, "router_trainable": true, "sec_per_step": 0.305, "probe_sec": 0.0422, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03321, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.07379, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 0.15992, "trunc_abs_frac": 0.66113, "mean_delta_full": 0.25771, "read_hit_rate": {"split": 0.94218}, "mean_gate": {"split": 0.24619}, "n_reads": 11798, "n_cand": 12522, "read_suppressed_frac": 0.0976, "dropout_reads_logged": 11798, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 465.0734558105469, "n_util_obs": 432, "util_count_spearman": 0.08747187911636028}, "probe_overhead_sec_total": 186.647, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.019611, "alpha": 0.03, "t": 844.4}
+{"event": "consolidate", "step": 2600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 858.2}
+{"event": "probe", "step": 2600, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3602.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.206976, "cps": 28.140625, "mean_p": 0.217124, "boundary_entropy": 0.123958, "bpb": 2.014514}, "tr": {"chunks": 3474.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.191902, "cps": 27.140625, "mean_p": 0.244239, "boundary_entropy": 0.252163, "bpb": 2.235533}, "fi": {"chunks": 4533.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.240976, "cps": 35.414062, "mean_p": 0.252775, "boundary_entropy": 0.164572, "bpb": 2.376178}}, "bpb": {"en": 2.01451, "tr": 2.23553, "fi": 2.37618}, "gini_cpc": 0.051131, "gini_cps": 0.060815, "bpic": 4.886898, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9271706478901975, "fi": 1.1642714484368915}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 474.1438293457031, "n_util_obs": 440, "util_count_spearman": 0.08303722278232646}, "t": 859.1}
+{"event": "eval", "step": 2600, "phase": 2, "bpb": {"en": 2.03061, "tr": 2.24188, "fi": 2.38521}, "bytes_per_chunk": {"en": 4.614, "tr": 5.728, "fi": 4.42}, "bpb_mean": 2.21923, "frozen_inventory": true, "t": 860.6}
+{"event": "train", "step": 2600, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.49675, "bpb": 2.15935, "chunk_loss": 0.030627, "F": 0.20874, "G": 0.22962, "chunks": 213.75, "bytes_per_chunk": 4.791, "boundary_entropy": 0.14673, "router_noise": 0.0, "grad_norm": 0.6962, "lr": 0.0002053, "router_trainable": true, "sec_per_step": 0.353, "probe_sec": 0.0643, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0549, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.07519, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.61135, "trunc_abs_frac": 0.69337, "mean_delta_full": 0.31042, "read_hit_rate": {"split": 0.94257}, "mean_gate": {"split": 0.24669}, "n_reads": 11850, "n_cand": 12572, "read_suppressed_frac": 0.0976, "dropout_reads_logged": 11850, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 474.3232727050781, "n_util_obs": 440, "util_count_spearman": 0.08305814231260623}, "probe_overhead_sec_total": 190.244, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.027644, "alpha": 0.03, "t": 862.1}
+{"event": "consolidate", "step": 2650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 878.7}
+{"event": "train", "step": 2650, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.45653, "bpb": 2.10132, "chunk_loss": 0.030571, "F": 0.20596, "G": 0.23156, "chunks": 210.91, "bytes_per_chunk": 4.855, "boundary_entropy": 0.15891, "router_noise": 0.0, "grad_norm": 0.7211, "lr": 0.0002017, "router_trainable": true, "sec_per_step": 0.338, "probe_sec": 0.0705, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05342, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.01426, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -0.63546, "trunc_abs_frac": 0.55465, "mean_delta_full": 0.01108, "read_hit_rate": {"split": 0.93996}, "mean_gate": {"split": 0.25721}, "n_reads": 11725, "n_cand": 12474, "read_suppressed_frac": 0.0977, "dropout_reads_logged": 11725, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 483.62579345703125, "n_util_obs": 450, "util_count_spearman": 0.06708175794806616}, "probe_overhead_sec_total": 193.667, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.025754, "alpha": 0.03, "t": 879.0}
+{"event": "consolidate", "step": 2700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 895.4}
+{"event": "train", "step": 2700, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.53457, "bpb": 2.21391, "chunk_loss": 0.030743, "F": 0.21219, "G": 0.23438, "chunks": 217.28, "bytes_per_chunk": 4.713, "boundary_entropy": 0.14949, "router_noise": 0.0, "grad_norm": 0.7858, "lr": 0.0001981, "router_trainable": true, "sec_per_step": 0.333, "probe_sec": 0.0512, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03999, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.00223, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 7.73639, "trunc_abs_frac": 0.88601, "mean_delta_full": -0.01436, "read_hit_rate": {"split": 0.93827}, "mean_gate": {"split": 0.24622}, "n_reads": 12038, "n_cand": 12830, "read_suppressed_frac": 0.0998, "dropout_reads_logged": 12038, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 492.8846435546875, "n_util_obs": 458, "util_count_spearman": 0.064557265754962}, "probe_overhead_sec_total": 197.683, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.017375, "alpha": 0.03, "t": 895.6}
+{"event": "consolidate", "step": 2750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 927.7}
+{"event": "train", "step": 2750, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.46787, "bpb": 2.11769, "chunk_loss": 0.030534, "F": 0.20639, "G": 0.23064, "chunks": 211.34, "bytes_per_chunk": 4.845, "boundary_entropy": 0.15262, "router_noise": 0.0, "grad_norm": 0.6932, "lr": 0.0001945, "router_trainable": true, "sec_per_step": 0.648, "probe_sec": 0.1452, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.13204, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.00827, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 2.42808, "trunc_abs_frac": 0.79916, "mean_delta_full": 0.07157, "read_hit_rate": {"split": 0.93986}, "mean_gate": {"split": 0.24797}, "n_reads": 11737, "n_cand": 12488, "read_suppressed_frac": 0.098, "dropout_reads_logged": 11737, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 502.1774597167969, "n_util_obs": 462, "util_count_spearman": 0.07462748956350168}, "probe_overhead_sec_total": 202.279, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.027389, "alpha": 0.03, "t": 928.0}
+{"event": "consolidate", "step": 2800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 941.9}
+{"event": "probe", "step": 2800, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3575.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.205424, "cps": 27.929688, "mean_p": 0.212777, "boundary_entropy": 0.121656, "bpb": 1.991154}, "tr": {"chunks": 3437.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.189858, "cps": 26.851562, "mean_p": 0.239695, "boundary_entropy": 0.246145, "bpb": 2.210525}, "fi": {"chunks": 4563.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.242571, "cps": 35.648438, "mean_p": 0.254754, "boundary_entropy": 0.158022, "bpb": 2.335442}}, "bpb": {"en": 1.99115, "tr": 2.21053, "fi": 2.33544}, "gini_cpc": 0.055094, "gini_cps": 0.064852, "bpic": 4.901253, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9242236016207181, "fi": 1.1808280454859585}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 511.2325439453125, "n_util_obs": 470, "util_count_spearman": 0.07616672246381927}, "t": 942.9}
+{"event": "eval", "step": 2800, "phase": 2, "bpb": {"en": 2.00888, "tr": 2.216, "fi": 2.34431}, "bytes_per_chunk": {"en": 4.636, "tr": 5.806, "fi": 4.378}, "bpb_mean": 2.18973, "frozen_inventory": true, "t": 944.4}
+{"event": "train", "step": 2800, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.44695, "bpb": 2.08751, "chunk_loss": 0.030774, "F": 0.2175, "G": 0.23046, "chunks": 222.72, "bytes_per_chunk": 4.598, "boundary_entropy": 0.13614, "router_noise": 0.0, "grad_norm": 0.6236, "lr": 0.0001909, "router_trainable": true, "sec_per_step": 0.353, "probe_sec": 0.0518, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03997, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.0338, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.58164, "trunc_abs_frac": 1.21992, "mean_delta_full": 0.09476, "read_hit_rate": {"split": 0.95311}, "mean_gate": {"split": 0.24991}, "n_reads": 12602, "n_cand": 13222, "read_suppressed_frac": 0.098, "dropout_reads_logged": 12602, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 511.425537109375, "n_util_obs": 470, "util_count_spearman": 0.07630685225876432}, "probe_overhead_sec_total": 205.581, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.025798, "alpha": 0.03, "t": 945.7}
+{"event": "consolidate", "step": 2850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 965.2}
+{"event": "train", "step": 2850, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.52722, "bpb": 2.20331, "chunk_loss": 0.030702, "F": 0.21368, "G": 0.23261, "chunks": 218.81, "bytes_per_chunk": 4.68, "boundary_entropy": 0.13998, "router_noise": 0.0, "grad_norm": 0.6778, "lr": 0.0001872, "router_trainable": true, "sec_per_step": 0.397, "probe_sec": 0.1079, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09628, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.05557, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 4.96765, "trunc_abs_frac": 2.6786, "mean_delta_full": 0.02905, "read_hit_rate": {"split": 0.94486}, "mean_gate": {"split": 0.25171}, "n_reads": 12304, "n_cand": 13022, "read_suppressed_frac": 0.0998, "dropout_reads_logged": 12304, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 520.663818359375, "n_util_obs": 478, "util_count_spearman": 0.07329595453981606}, "probe_overhead_sec_total": 208.948, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.027631, "alpha": 0.03, "t": 965.5}
+{"event": "consolidate", "step": 2900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 986.7}
+{"event": "train", "step": 2900, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.41137, "bpb": 2.03618, "chunk_loss": 0.030527, "F": 0.20654, "G": 0.22924, "chunks": 211.5, "bytes_per_chunk": 4.842, "boundary_entropy": 0.15167, "router_noise": 0.0, "grad_norm": 0.7112, "lr": 0.0001835, "router_trainable": true, "sec_per_step": 0.432, "probe_sec": 0.0585, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03071, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.01831, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.14456, "trunc_abs_frac": 0.54398, "mean_delta_full": -0.30325, "read_hit_rate": {"split": 0.94598}, "mean_gate": {"split": 0.24036}, "n_reads": 12031, "n_cand": 12718, "read_suppressed_frac": 0.0975, "dropout_reads_logged": 12031, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 529.9652099609375, "n_util_obs": 487, "util_count_spearman": 0.05860629029354867}, "probe_overhead_sec_total": 212.861, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.036997, "alpha": 0.03, "t": 987.1}
+{"event": "consolidate", "step": 2950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1000.9}
+{"event": "train", "step": 2950, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.35509, "bpb": 1.95498, "chunk_loss": 0.030644, "F": 0.215, "G": 0.22571, "chunks": 220.16, "bytes_per_chunk": 4.651, "boundary_entropy": 0.13264, "router_noise": 0.0, "grad_norm": 0.6364, "lr": 0.0001799, "router_trainable": true, "sec_per_step": 0.282, "probe_sec": 0.0561, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04252, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.29891, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": -0.03457, "trunc_abs_frac": 0.03457, "mean_delta_full": 0.24365, "read_hit_rate": {"split": 0.96269}, "mean_gate": {"split": 0.25627}, "n_reads": 12875, "n_cand": 13374, "read_suppressed_frac": 0.1001, "dropout_reads_logged": 12875, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 539.2493286132812, "n_util_obs": 495, "util_count_spearman": 0.04738430427478021}, "probe_overhead_sec_total": 216.346, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.02187, "alpha": 0.03, "t": 1001.2}
+{"event": "consolidate", "step": 3000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1015.5}
+{"event": "probe", "step": 3000, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3563.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.204735, "cps": 27.835938, "mean_p": 0.209902, "boundary_entropy": 0.117782, "bpb": 1.982319}, "tr": {"chunks": 3328.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.183837, "cps": 26.0, "mean_p": 0.230641, "boundary_entropy": 0.236437, "bpb": 2.190625}, "fi": {"chunks": 4431.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.235554, "cps": 34.617188, "mean_p": 0.245047, "boundary_entropy": 0.146904, "bpb": 2.312332}}, "bpb": {"en": 1.98232, "tr": 2.19063, "fi": 2.31233}, "gini_cpc": 0.055242, "gini_cps": 0.064947, "bpic": 5.010775, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.8979270689942445, "fi": 1.1505305751065897}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 548.3526611328125, "n_util_obs": 505, "util_count_spearman": 0.03340018940093653}, "t": 1016.1}
+{"event": "eval", "step": 3000, "phase": 2, "bpb": {"en": 1.99788, "tr": 2.19574, "fi": 2.31904}, "bytes_per_chunk": {"en": 4.648, "tr": 5.982, "fi": 4.503}, "bpb_mean": 2.17088, "frozen_inventory": true, "t": 1017.3}
+{"event": "train", "step": 3000, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.44682, "bpb": 2.08732, "chunk_loss": 0.030543, "F": 0.20703, "G": 0.2241, "chunks": 212.0, "bytes_per_chunk": 4.83, "boundary_entropy": 0.14238, "router_noise": 0.0, "grad_norm": 0.7638, "lr": 0.0001761, "router_trainable": true, "sec_per_step": 0.343, "probe_sec": 0.0753, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06509, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.13415, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.95433, "trunc_abs_frac": 1.00066, "mean_delta_full": 0.35046, "read_hit_rate": {"split": 0.93878}, "mean_gate": {"split": 0.24046}, "n_reads": 11793, "n_cand": 12562, "read_suppressed_frac": 0.0985, "dropout_reads_logged": 11793, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 548.532470703125, "n_util_obs": 505, "util_count_spearman": 0.0336494860960774}, "probe_overhead_sec_total": 219.936, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.01715, "alpha": 0.03, "t": 1018.4}
+{"event": "consolidate", "step": 3050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1029.6}
+{"event": "train", "step": 3050, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.40408, "bpb": 2.02566, "chunk_loss": 0.030611, "F": 0.21265, "G": 0.22579, "chunks": 217.75, "bytes_per_chunk": 4.703, "boundary_entropy": 0.12752, "router_noise": 0.0, "grad_norm": 0.6787, "lr": 0.0001724, "router_trainable": true, "sec_per_step": 0.228, "probe_sec": 0.0433, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03166, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.0799, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -0.49946, "trunc_abs_frac": 5.91516, "mean_delta_full": 0.1564, "read_hit_rate": {"split": 0.9559}, "mean_gate": {"split": 0.24864}, "n_reads": 12551, "n_cand": 13130, "read_suppressed_frac": 0.1036, "dropout_reads_logged": 12551, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 557.7923583984375, "n_util_obs": 509, "util_count_spearman": 0.0567036120447525}, "probe_overhead_sec_total": 222.498, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.031355, "alpha": 0.03, "t": 1029.8}
+{"event": "consolidate", "step": 3100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1040.5}
+{"event": "train", "step": 3100, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.4906, "bpb": 2.15048, "chunk_loss": 0.030517, "F": 0.20511, "G": 0.22366, "chunks": 210.03, "bytes_per_chunk": 4.875, "boundary_entropy": 0.13928, "router_noise": 0.0, "grad_norm": 0.7824, "lr": 0.0001687, "router_trainable": true, "sec_per_step": 0.221, "probe_sec": 0.0453, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03272, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.10411, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.49989, "trunc_abs_frac": 0.72812, "mean_delta_full": 0.22107, "read_hit_rate": {"split": 0.94076}, "mean_gate": {"split": 0.25497}, "n_reads": 11846, "n_cand": 12592, "read_suppressed_frac": 0.1014, "dropout_reads_logged": 11846, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 567.052490234375, "n_util_obs": 518, "util_count_spearman": 0.05315741383207047}, "probe_overhead_sec_total": 225.136, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.022629, "alpha": 0.03, "t": 1040.8}
+{"event": "consolidate", "step": 3150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1054.7}
+{"event": "train", "step": 3150, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.40192, "bpb": 2.02254, "chunk_loss": 0.030562, "F": 0.2084, "G": 0.22458, "chunks": 213.41, "bytes_per_chunk": 4.798, "boundary_entropy": 0.13342, "router_noise": 0.0, "grad_norm": 0.6586, "lr": 0.000165, "router_trainable": true, "sec_per_step": 0.28, "probe_sec": 0.0375, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02745, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.04365, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 1.45129, "trunc_abs_frac": 0.97387, "mean_delta_full": -0.10288, "read_hit_rate": {"split": 0.9443}, "mean_gate": {"split": 0.2419}, "n_reads": 12021, "n_cand": 12730, "read_suppressed_frac": 0.0987, "dropout_reads_logged": 12021, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 576.33203125, "n_util_obs": 522, "util_count_spearman": 0.055507970359764104}, "probe_overhead_sec_total": 228.101, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.031156, "alpha": 0.03, "t": 1054.8}
+{"event": "consolidate", "step": 3200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1070.5}
+{"event": "probe", "step": 3200, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3569.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.20508, "cps": 27.882812, "mean_p": 0.212561, "boundary_entropy": 0.110674, "bpb": 1.960669}, "tr": {"chunks": 3396.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.187593, "cps": 26.53125, "mean_p": 0.235668, "boundary_entropy": 0.238287, "bpb": 2.155907}, "fi": {"chunks": 4460.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.237095, "cps": 34.84375, "mean_p": 0.250212, "boundary_entropy": 0.144762, "bpb": 2.289655}}, "bpb": {"en": 1.96067, "tr": 2.15591, "fi": 2.28965}, "gini_cpc": 0.052402, "gini_cps": 0.062086, "bpic": 4.965602, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9147337484965665, "fi": 1.1561136997112105}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 585.4098510742188, "n_util_obs": 528, "util_count_spearman": 0.05743732120692419}, "t": 1071.4}
+{"event": "eval", "step": 3200, "phase": 2, "bpb": {"en": 1.97576, "tr": 2.16421, "fi": 2.29781}, "bytes_per_chunk": {"en": 4.641, "tr": 5.857, "fi": 4.461}, "bpb_mean": 2.14593, "frozen_inventory": true, "t": 1072.8}
+{"event": "train", "step": 3200, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.33664, "bpb": 1.92836, "chunk_loss": 0.030681, "F": 0.21735, "G": 0.22665, "chunks": 222.56, "bytes_per_chunk": 4.601, "boundary_entropy": 0.11461, "router_noise": 0.0, "grad_norm": 0.6775, "lr": 0.0001613, "router_trainable": true, "sec_per_step": 0.406, "probe_sec": 0.0664, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03765, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.13041, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.60861, "trunc_abs_frac": 0.7922, "mean_delta_full": -0.55164, "read_hit_rate": {"split": 0.9643}, "mean_gate": {"split": 0.25606}, "n_reads": 12856, "n_cand": 13332, "read_suppressed_frac": 0.0979, "dropout_reads_logged": 12856, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 585.6064453125, "n_util_obs": 528, "util_count_spearman": 0.05722009452407854}, "probe_overhead_sec_total": 231.862, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.034404, "alpha": 0.03, "t": 1075.1}
+{"event": "consolidate", "step": 3250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1094.0}
+{"event": "train", "step": 3250, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.33669, "bpb": 1.92844, "chunk_loss": 0.030615, "F": 0.21219, "G": 0.2254, "chunks": 217.28, "bytes_per_chunk": 4.713, "boundary_entropy": 0.12945, "router_noise": 0.0, "grad_norm": 0.6435, "lr": 0.0001576, "router_trainable": true, "sec_per_step": 0.398, "probe_sec": 0.0758, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03547, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.14694, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -5.52873, "trunc_abs_frac": 2.07069, "mean_delta_full": 0.04026, "read_hit_rate": {"split": 0.95585}, "mean_gate": {"split": 0.24286}, "n_reads": 12428, "n_cand": 13002, "read_suppressed_frac": 0.1003, "dropout_reads_logged": 12428, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 594.8131103515625, "n_util_obs": 532, "util_count_spearman": 0.056859089807536754}, "probe_overhead_sec_total": 235.326, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.018624, "alpha": 0.03, "t": 1095.0}
+{"event": "consolidate", "step": 3300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1112.7}
+{"event": "train", "step": 3300, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.46512, "bpb": 2.11372, "chunk_loss": 0.030618, "F": 0.21201, "G": 0.22723, "chunks": 217.09, "bytes_per_chunk": 4.717, "boundary_entropy": 0.12997, "router_noise": 0.0, "grad_norm": 0.7118, "lr": 0.0001539, "router_trainable": true, "sec_per_step": 0.357, "probe_sec": 0.0445, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02922, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.07001, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 2.74591, "trunc_abs_frac": 1.80705, "mean_delta_full": 0.16556, "read_hit_rate": {"split": 0.94961}, "mean_gate": {"split": 0.23054}, "n_reads": 12307, "n_cand": 12960, "read_suppressed_frac": 0.101, "dropout_reads_logged": 12307, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 604.0477294921875, "n_util_obs": 535, "util_count_spearman": 0.051746403351023186}, "probe_overhead_sec_total": 239.332, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.022596, "alpha": 0.03, "t": 1112.9}
+{"event": "consolidate", "step": 3350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1127.6}
+{"event": "train", "step": 3350, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.41547, "bpb": 2.04209, "chunk_loss": 0.03054, "F": 0.20801, "G": 0.22514, "chunks": 213.0, "bytes_per_chunk": 4.808, "boundary_entropy": 0.12784, "router_noise": 0.0, "grad_norm": 0.6681, "lr": 0.0001501, "router_trainable": true, "sec_per_step": 0.297, "probe_sec": 0.0375, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02683, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.09628, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.75984, "trunc_abs_frac": 0.70423, "mean_delta_full": 0.31437, "read_hit_rate": {"split": 0.94955}, "mean_gate": {"split": 0.24797}, "n_reads": 12160, "n_cand": 12806, "read_suppressed_frac": 0.0942, "dropout_reads_logged": 12160, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 613.2840576171875, "n_util_obs": 538, "util_count_spearman": 0.050342101797430545}, "probe_overhead_sec_total": 242.778, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.022293, "alpha": 0.03, "t": 1127.7}
+{"event": "consolidate", "step": 3400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1140.1}
+{"event": "probe", "step": 3400, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3566.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.204907, "cps": 27.859375, "mean_p": 0.21167, "boundary_entropy": 0.110028, "bpb": 1.940716}, "tr": {"chunks": 3398.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.187704, "cps": 26.546875, "mean_p": 0.234944, "boundary_entropy": 0.234003, "bpb": 2.134393}, "fi": {"chunks": 4500.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.239222, "cps": 35.15625, "mean_p": 0.250706, "boundary_entropy": 0.141392, "bpb": 2.265626}}, "bpb": {"en": 1.94072, "tr": 2.13439, "fi": 2.26563}, "gini_cpc": 0.054358, "gini_cps": 0.064085, "bpic": 4.948709, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9160424602175951, "fi": 1.16746376931935}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 622.3712158203125, "n_util_obs": 544, "util_count_spearman": 0.035321030327224216}, "t": 1141.0}
+{"event": "eval", "step": 3400, "phase": 2, "bpb": {"en": 1.95882, "tr": 2.14237, "fi": 2.27278}, "bytes_per_chunk": {"en": 4.65, "tr": 5.887, "fi": 4.441}, "bpb_mean": 2.12465, "frozen_inventory": true, "t": 1142.3}
+{"event": "train", "step": 3400, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.37452, "bpb": 1.98301, "chunk_loss": 0.030536, "F": 0.20975, "G": 0.22267, "chunks": 214.78, "bytes_per_chunk": 4.768, "boundary_entropy": 0.12175, "router_noise": 0.0, "grad_norm": 0.6808, "lr": 0.0001465, "router_trainable": true, "sec_per_step": 0.318, "probe_sec": 0.1093, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08932, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.34151, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": -0.64199, "trunc_abs_frac": 0.64199, "mean_delta_full": -0.28027, "read_hit_rate": {"split": 0.9513}, "mean_gate": {"split": 0.24292}, "n_reads": 12249, "n_cand": 12876, "read_suppressed_frac": 0.1003, "dropout_reads_logged": 12249, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 622.5590209960938, "n_util_obs": 544, "util_count_spearman": 0.03605662066907078}, "probe_overhead_sec_total": 246.56, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.025755, "alpha": 0.03, "t": 1143.6}
+{"event": "consolidate", "step": 3450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1155.7}
+{"event": "train", "step": 3450, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.52164, "bpb": 2.19526, "chunk_loss": 0.030756, "F": 0.21545, "G": 0.23278, "chunks": 220.62, "bytes_per_chunk": 4.641, "boundary_entropy": 0.13395, "router_noise": 0.0, "grad_norm": 0.7478, "lr": 0.0001428, "router_trainable": true, "sec_per_step": 0.247, "probe_sec": 0.0976, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0843, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.0815, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.5762, "trunc_abs_frac": 0.436, "mean_delta_full": -0.38672, "read_hit_rate": {"split": 0.93749}, "mean_gate": {"split": 0.23055}, "n_reads": 12103, "n_cand": 12910, "read_suppressed_frac": 0.1, "dropout_reads_logged": 12103, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 631.8671264648438, "n_util_obs": 552, "util_count_spearman": 0.03329297031576492}, "probe_overhead_sec_total": 250.251, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.017423, "alpha": 0.03, "t": 1156.0}
+{"event": "consolidate", "step": 3500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1168.5}
+{"event": "train", "step": 3500, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.40492, "bpb": 2.02687, "chunk_loss": 0.030582, "F": 0.20932, "G": 0.22545, "chunks": 214.34, "bytes_per_chunk": 4.777, "boundary_entropy": 0.12933, "router_noise": 0.0, "grad_norm": 0.6883, "lr": 0.0001391, "router_trainable": true, "sec_per_step": 0.258, "probe_sec": 0.0938, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08128, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.12769, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -0.85381, "trunc_abs_frac": 1.41491, "mean_delta_full": -0.09895, "read_hit_rate": {"split": 0.95172}, "mean_gate": {"split": 0.23487}, "n_reads": 12144, "n_cand": 12760, "read_suppressed_frac": 0.0956, "dropout_reads_logged": 12144, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 641.152099609375, "n_util_obs": 555, "util_count_spearman": 0.03237248794789361}, "probe_overhead_sec_total": 253.814, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.023878, "alpha": 0.03, "t": 1168.9}
+{"event": "consolidate", "step": 3550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1181.4}
+{"event": "train", "step": 3550, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.39486, "bpb": 2.01235, "chunk_loss": 0.030548, "F": 0.20972, "G": 0.22249, "chunks": 214.75, "bytes_per_chunk": 4.768, "boundary_entropy": 0.12344, "router_noise": 0.0, "grad_norm": 0.6569, "lr": 0.0001355, "router_trainable": true, "sec_per_step": 0.26, "probe_sec": 0.0567, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03685, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.08487, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 2.10677, "trunc_abs_frac": 0.25605, "mean_delta_full": -0.04144, "read_hit_rate": {"split": 0.94671}, "mean_gate": {"split": 0.23648}, "n_reads": 12186, "n_cand": 12872, "read_suppressed_frac": 0.0962, "dropout_reads_logged": 12186, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 650.4235229492188, "n_util_obs": 559, "util_count_spearman": 0.02554959014046019}, "probe_overhead_sec_total": 257.834, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.020009, "alpha": 0.03, "t": 1181.9}
+{"event": "consolidate", "step": 3600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1194.1}
+{"event": "probe", "step": 3600, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3564.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.204792, "cps": 27.84375, "mean_p": 0.211066, "boundary_entropy": 0.107938, "bpb": 1.926443}, "tr": {"chunks": 3419.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.188864, "cps": 26.710938, "mean_p": 0.233904, "boundary_entropy": 0.234685, "bpb": 2.124083}, "fi": {"chunks": 4442.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.236138, "cps": 34.703125, "mean_p": 0.247924, "boundary_entropy": 0.139639, "bpb": 2.248837}}, "bpb": {"en": 1.92644, "tr": 2.12408, "fi": 2.24884}, "gini_cpc": 0.050042, "gini_cps": 0.059694, "bpic": 4.965602, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9222209295815881, "fi": 1.1530631568641245}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 659.473876953125, "n_util_obs": 565, "util_count_spearman": 0.021194728621336735}, "t": 1194.9}
+{"event": "eval", "step": 3600, "phase": 2, "bpb": {"en": 1.94359, "tr": 2.13073, "fi": 2.25721}, "bytes_per_chunk": {"en": 4.652, "tr": 5.832, "fi": 4.483}, "bpb_mean": 2.11051, "frozen_inventory": true, "t": 1196.5}
+{"event": "train", "step": 3600, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.4223, "bpb": 2.05195, "chunk_loss": 0.030477, "F": 0.20316, "G": 0.22485, "chunks": 208.03, "bytes_per_chunk": 4.922, "boundary_entropy": 0.1406, "router_noise": 0.0, "grad_norm": 0.6854, "lr": 0.0001319, "router_trainable": true, "sec_per_step": 0.319, "probe_sec": 0.0489, "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": 14, "mean_delta_loo": 0.00275, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 4.24637, "trunc_abs_frac": 5.9125, "mean_delta_full": -0.02209, "read_hit_rate": {"split": 0.93503}, "mean_gate": {"split": 0.23148}, "n_reads": 11527, "n_cand": 12328, "read_suppressed_frac": 0.1005, "dropout_reads_logged": 11527, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 659.648193359375, "n_util_obs": 565, "util_count_spearman": 0.02075897019912475}, "probe_overhead_sec_total": 261.443, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.019907, "alpha": 0.03, "t": 1197.9}
+{"event": "consolidate", "step": 3650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1210.3}
+{"event": "train", "step": 3650, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.34583, "bpb": 1.94162, "chunk_loss": 0.030549, "F": 0.21143, "G": 0.22164, "chunks": 216.5, "bytes_per_chunk": 4.73, "boundary_entropy": 0.11539, "router_noise": 0.0, "grad_norm": 0.6453, "lr": 0.0001283, "router_trainable": true, "sec_per_step": 0.255, "probe_sec": 0.0671, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05384, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.05275, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": -0.30819, "trunc_abs_frac": 0.82316, "mean_delta_full": 0.05548, "read_hit_rate": {"split": 0.95714}, "mean_gate": {"split": 0.23947}, "n_reads": 12372, "n_cand": 12926, "read_suppressed_frac": 0.0989, "dropout_reads_logged": 12372, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 668.8897094726562, "n_util_obs": 570, "util_count_spearman": 0.025302602570995307}, "probe_overhead_sec_total": 265.099, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.019254, "alpha": 0.03, "t": 1210.6}
+{"event": "consolidate", "step": 3700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1222.9}
+{"event": "train", "step": 3700, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.44967, "bpb": 2.09143, "chunk_loss": 0.030483, "F": 0.20337, "G": 0.22055, "chunks": 208.25, "bytes_per_chunk": 4.917, "boundary_entropy": 0.13035, "router_noise": 0.0, "grad_norm": 0.7614, "lr": 0.0001247, "router_trainable": true, "sec_per_step": 0.25, "probe_sec": 0.0515, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04137, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.14354, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 1.09083, "trunc_abs_frac": 0.97653, "mean_delta_full": 0.09318, "read_hit_rate": {"split": 0.93218}, "mean_gate": {"split": 0.22545}, "n_reads": 11395, "n_cand": 12224, "read_suppressed_frac": 0.0964, "dropout_reads_logged": 11395, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 678.216552734375, "n_util_obs": 578, "util_count_spearman": 0.027887887847764983}, "probe_overhead_sec_total": 268.492, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.024507, "alpha": 0.03, "t": 1223.1}
+{"event": "consolidate", "step": 3750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1236.1}
+{"event": "train", "step": 3750, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.42855, "bpb": 2.06097, "chunk_loss": 0.030698, "F": 0.21448, "G": 0.22935, "chunks": 219.62, "bytes_per_chunk": 4.662, "boundary_entropy": 0.13141, "router_noise": 0.0, "grad_norm": 0.7258, "lr": 0.0001212, "router_trainable": true, "sec_per_step": 0.266, "probe_sec": 0.1244, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10941, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.10932, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 0, "read_hit_rate": {"split": 0.95029}, "mean_gate": {"split": 0.23116}, "n_reads": 12445, "n_cand": 13096, "read_suppressed_frac": 0.0964, "dropout_reads_logged": 12445, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 687.549072265625, "n_util_obs": 582, "util_count_spearman": 0.03227163737636602}, "probe_overhead_sec_total": 272.64, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.019794, "alpha": 0.03, "t": 1236.4}
+{"event": "consolidate", "step": 3800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1249.0}
+{"event": "probe", "step": 3800, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3589.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.206229, "cps": 28.039062, "mean_p": 0.211599, "boundary_entropy": 0.112468, "bpb": 1.907592}, "tr": {"chunks": 3450.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.190576, "cps": 26.953125, "mean_p": 0.233537, "boundary_entropy": 0.230453, "bpb": 2.105896}, "fi": {"chunks": 4522.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.240391, "cps": 35.328125, "mean_p": 0.250187, "boundary_entropy": 0.136867, "bpb": 2.227403}}, "bpb": {"en": 1.90759, "tr": 2.1059, "fi": 2.2274}, "gini_cpc": 0.052119, "gini_cps": 0.061817, "bpic": 4.907188, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9241005006074418, "fi": 1.165653136057599}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 696.5978393554688, "n_util_obs": 591, "util_count_spearman": 0.0463425024647881}, "t": 1249.8}
+{"event": "eval", "step": 3800, "phase": 2, "bpb": {"en": 1.92669, "tr": 2.11331, "fi": 2.23557}, "bytes_per_chunk": {"en": 4.629, "tr": 5.748, "fi": 4.435}, "bpb_mean": 2.09186, "frozen_inventory": true, "t": 1251.2}
+{"event": "train", "step": 3800, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.32467, "bpb": 1.9111, "chunk_loss": 0.030674, "F": 0.216, "G": 0.22367, "chunks": 221.19, "bytes_per_chunk": 4.63, "boundary_entropy": 0.11537, "router_noise": 0.0, "grad_norm": 0.6459, "lr": 0.0001177, "router_trainable": true, "sec_per_step": 0.325, "probe_sec": 0.0684, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0557, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.08698, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -11.76931, "trunc_abs_frac": 2.41409, "mean_delta_full": -0.01554, "read_hit_rate": {"split": 0.96188}, "mean_gate": {"split": 0.23764}, "n_reads": 12895, "n_cand": 13406, "read_suppressed_frac": 0.0946, "dropout_reads_logged": 12895, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 696.7952880859375, "n_util_obs": 591, "util_count_spearman": 0.046321108977869285}, "probe_overhead_sec_total": 275.784, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.02244, "alpha": 0.03, "t": 1252.6}
+{"event": "consolidate", "step": 3850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1264.9}
+{"event": "train", "step": 3850, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.34402, "bpb": 1.93901, "chunk_loss": 0.030546, "F": 0.21057, "G": 0.22035, "chunks": 215.62, "bytes_per_chunk": 4.749, "boundary_entropy": 0.1193, "router_noise": 0.0, "grad_norm": 0.6546, "lr": 0.0001142, "router_trainable": true, "sec_per_step": 0.249, "probe_sec": 0.0426, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03303, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.06491, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.53988, "trunc_abs_frac": 1.23615, "mean_delta_full": 0.33466, "read_hit_rate": {"split": 0.95589}, "mean_gate": {"split": 0.24333}, "n_reads": 12287, "n_cand": 12854, "read_suppressed_frac": 0.0985, "dropout_reads_logged": 12287, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 706.006103515625, "n_util_obs": 599, "util_count_spearman": 0.05274931684144636}, "probe_overhead_sec_total": 278.996, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.031406, "alpha": 0.03, "t": 1265.1}
+{"event": "consolidate", "step": 3900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1277.6}
+{"event": "train", "step": 3900, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.40762, "bpb": 2.03076, "chunk_loss": 0.030513, "F": 0.20813, "G": 0.2234, "chunks": 213.12, "bytes_per_chunk": 4.805, "boundary_entropy": 0.1292, "router_noise": 0.0, "grad_norm": 0.6805, "lr": 0.0001108, "router_trainable": true, "sec_per_step": 0.256, "probe_sec": 0.0826, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07276, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.16913, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 1.15337, "trunc_abs_frac": 1.21198, "mean_delta_full": 0.1178, "read_hit_rate": {"split": 0.94194}, "mean_gate": {"split": 0.23758}, "n_reads": 11989, "n_cand": 12728, "read_suppressed_frac": 0.0961, "dropout_reads_logged": 11989, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 715.3148193359375, "n_util_obs": 603, "util_count_spearman": 0.043619226179066745}, "probe_overhead_sec_total": 282.947, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.022335, "alpha": 0.03, "t": 1277.9}
+{"event": "consolidate", "step": 3950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1290.8}
+{"event": "train", "step": 3950, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.41584, "bpb": 2.04263, "chunk_loss": 0.030628, "F": 0.2142, "G": 0.22356, "chunks": 219.34, "bytes_per_chunk": 4.668, "boundary_entropy": 0.11354, "router_noise": 0.0, "grad_norm": 0.713, "lr": 0.0001074, "router_trainable": true, "sec_per_step": 0.262, "probe_sec": 0.0391, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02877, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.09443, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 1.0, "trunc_abs_frac": 1.0, "mean_delta_full": 0.12341, "read_hit_rate": {"split": 0.95109}, "mean_gate": {"split": 0.23181}, "n_reads": 12446, "n_cand": 13086, "read_suppressed_frac": 0.1015, "dropout_reads_logged": 12446, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 724.6052856445312, "n_util_obs": 613, "util_count_spearman": 0.023773981589051727}, "probe_overhead_sec_total": 286.753, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.027255, "alpha": 0.03, "t": 1291.0}
+{"event": "consolidate", "step": 4000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1302.9}
+{"event": "probe", "step": 4000, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3582.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.205827, "cps": 27.984375, "mean_p": 0.209233, "boundary_entropy": 0.104593, "bpb": 1.900059}, "tr": {"chunks": 3422.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.189029, "cps": 26.734375, "mean_p": 0.226354, "boundary_entropy": 0.218961, "bpb": 2.089884}, "fi": {"chunks": 4485.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.238424, "cps": 35.039062, "mean_p": 0.248459, "boundary_entropy": 0.131934, "bpb": 2.218399}}, "bpb": {"en": 1.90006, "tr": 2.08988, "fi": 2.2184}, "gini_cpc": 0.051999, "gini_cps": 0.061682, "bpic": 4.937941, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9183917895467135, "fi": 1.1583748042215223}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 733.6998901367188, "n_util_obs": 620, "util_count_spearman": 0.025256445376171357}, "t": 1303.7}
+{"event": "eval", "step": 4000, "phase": 2, "bpb": {"en": 1.91884, "tr": 2.09665, "fi": 2.22584}, "bytes_per_chunk": {"en": 4.641, "tr": 5.846, "fi": 4.438}, "bpb_mean": 2.08044, "frozen_inventory": true, "t": 1305.1}
+{"event": "train", "step": 4000, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.35071, "bpb": 1.94866, "chunk_loss": 0.030486, "F": 0.20599, "G": 0.22131, "chunks": 210.94, "bytes_per_chunk": 4.855, "boundary_entropy": 0.12766, "router_noise": 0.0, "grad_norm": 0.6714, "lr": 0.000104, "router_trainable": true, "sec_per_step": 0.307, "probe_sec": 0.0412, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03102, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.01282, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 1.18703, "trunc_abs_frac": 0.36698, "mean_delta_full": 0.07343, "read_hit_rate": {"split": 0.94633}, "mean_gate": {"split": 0.22638}, "n_reads": 11867, "n_cand": 12540, "read_suppressed_frac": 0.0978, "dropout_reads_logged": 11867, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 733.8817138671875, "n_util_obs": 620, "util_count_spearman": 0.025462884114833686}, "probe_overhead_sec_total": 290.025, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.023853, "alpha": 0.03, "t": 1306.3}
+{"event": "consolidate", "step": 4050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1318.6}
+{"event": "train", "step": 4050, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.47542, "bpb": 2.12858, "chunk_loss": 0.030552, "F": 0.20853, "G": 0.22258, "chunks": 213.53, "bytes_per_chunk": 4.796, "boundary_entropy": 0.12825, "router_noise": 0.0, "grad_norm": 0.6986, "lr": 0.0001007, "router_trainable": true, "sec_per_step": 0.251, "probe_sec": 0.0446, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03523, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.09329, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.39748, "trunc_abs_frac": 0.6476, "mean_delta_full": -0.23218, "read_hit_rate": {"split": 0.93739}, "mean_gate": {"split": 0.22662}, "n_reads": 11858, "n_cand": 12650, "read_suppressed_frac": 0.1016, "dropout_reads_logged": 11858, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 743.02783203125, "n_util_obs": 630, "util_count_spearman": 0.02051048445372098}, "probe_overhead_sec_total": 293.426, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.020849, "alpha": 0.03, "t": 1318.9}
+{"event": "probe", "step": 4096, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3584.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.205942, "cps": 28.0, "mean_p": 0.209326, "boundary_entropy": 0.102605, "bpb": 1.892944}, "tr": {"chunks": 3341.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.184555, "cps": 26.101562, "mean_p": 0.223404, "boundary_entropy": 0.215713, "bpb": 2.078849}, "fi": {"chunks": 4475.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.237893, "cps": 34.960938, "mean_p": 0.246826, "boundary_entropy": 0.132416, "bpb": 2.206145}}, "bpb": {"en": 1.89294, "tr": 2.07885, "fi": 2.20614}, "gini_cpc": 0.056587, "gini_cps": 0.066316, "bpic": 4.976491, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.896152753268006, "fi": 1.1551470555079855}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 751.3719482421875, "n_util_obs": 638, "util_count_spearman": 0.01868302253743291}, "t": 1330.9}
+{"event": "consolidate", "step": 4100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1333.7}
+{"event": "train", "step": 4100, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.41853, "bpb": 2.04651, "chunk_loss": 0.03062, "F": 0.211, "G": 0.22659, "chunks": 216.06, "bytes_per_chunk": 4.739, "boundary_entropy": 0.12772, "router_noise": 0.0, "grad_norm": 0.7065, "lr": 9.75e-05, "router_trainable": true, "sec_per_step": 0.304, "probe_sec": 0.1337, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.12188, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.01035, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 1.33839, "trunc_abs_frac": 0.47608, "mean_delta_full": 0.07926, "read_hit_rate": {"split": 0.94088}, "mean_gate": {"split": 0.22551}, "n_reads": 12096, "n_cand": 12856, "read_suppressed_frac": 0.0988, "dropout_reads_logged": 12096, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 752.306640625, "n_util_obs": 638, "util_count_spearman": 0.020362625896623618}, "probe_overhead_sec_total": 297.016, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.024075, "alpha": 0.03, "t": 1334.1}
+{"event": "consolidate", "step": 4150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1346.7}
+{"event": "train", "step": 4150, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.36039, "bpb": 1.96263, "chunk_loss": 0.030576, "F": 0.21213, "G": 0.22191, "chunks": 217.22, "bytes_per_chunk": 4.714, "boundary_entropy": 0.11745, "router_noise": 0.0, "grad_norm": 0.655, "lr": 9.43e-05, "router_trainable": true, "sec_per_step": 0.256, "probe_sec": 0.0364, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02647, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.25994, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.81952, "trunc_abs_frac": 0.66565, "mean_delta_full": 0.95964, "read_hit_rate": {"split": 0.95303}, "mean_gate": {"split": 0.23403}, "n_reads": 12498, "n_cand": 13114, "read_suppressed_frac": 0.1013, "dropout_reads_logged": 12498, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 761.5448608398438, "n_util_obs": 645, "util_count_spearman": 0.014032466334171063}, "probe_overhead_sec_total": 300.835, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.02448, "alpha": 0.03, "t": 1346.8}
+{"event": "consolidate", "step": 4200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1361.9}
+{"event": "probe", "step": 4200, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3569.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.20508, "cps": 27.882812, "mean_p": 0.206859, "boundary_entropy": 0.104362, "bpb": 1.885963}, "tr": {"chunks": 3448.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.190466, "cps": 26.9375, "mean_p": 0.226091, "boundary_entropy": 0.213777, "bpb": 2.070301}, "fi": {"chunks": 4466.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.237414, "cps": 34.890625, "mean_p": 0.247255, "boundary_entropy": 0.129972, "bpb": 2.203292}}, "bpb": {"en": 1.88596, "tr": 2.0703, "fi": 2.20329}, "gini_cpc": 0.049449, "gini_cps": 0.059102, "bpic": 4.940521, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9287402723251358, "fi": 1.157669009621136}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 770.5897216796875, "n_util_obs": 648, "util_count_spearman": 0.01845086743033078}, "t": 1362.6}
+{"event": "eval", "step": 4200, "phase": 2, "bpb": {"en": 1.90485, "tr": 2.07805, "fi": 2.20678}, "bytes_per_chunk": {"en": 4.642, "tr": 5.768, "fi": 4.448}, "bpb_mean": 2.06323, "frozen_inventory": true, "t": 1364.0}
+{"event": "phase_transition", "step": 4200, "phase": 3, "objective": "global", "router_trainable": true, "encoder_trainable": true, "frozen_router_params": 0, "t": 1365.0}
+{"event": "train", "step": 4200, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.44, "bpb": 2.07747, "chunk_loss": 0.030654, "F": 0.21429, "G": 0.22515, "chunks": 219.44, "bytes_per_chunk": 4.666, "boundary_entropy": 0.12283, "router_noise": 0.0, "grad_norm": 0.6797, "lr": 9.12e-05, "router_trainable": true, "sec_per_step": 0.369, "probe_sec": 0.1118, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10017, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.34754, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 1.45362, "trunc_abs_frac": 0.62484, "mean_delta_full": 0.12106, "read_hit_rate": {"split": 0.94408}, "mean_gate": {"split": 0.22349}, "n_reads": 12224, "n_cand": 12948, "read_suppressed_frac": 0.098, "dropout_reads_logged": 12224, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 770.7772827148438, "n_util_obs": 648, "util_count_spearman": 0.018429103143118134}, "probe_overhead_sec_total": 304.599, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.028389, "alpha": 0.03, "t": 1365.3}
+{"event": "consolidate", "step": 4250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1380.9}
+{"event": "train", "step": 4250, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32908, "bpb": 1.91745, "chunk_loss": 0.030536, "F": 0.2114, "G": 0.21871, "chunks": 216.47, "bytes_per_chunk": 4.73, "boundary_entropy": 0.10929, "router_noise": 0.0, "grad_norm": 0.6475, "lr": 8.81e-05, "router_trainable": true, "sec_per_step": 0.315, "probe_sec": 0.0431, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03304, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.06816, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.40716, "trunc_abs_frac": 0.53032, "mean_delta_full": 0.37177, "read_hit_rate": {"split": 0.95737}, "mean_gate": {"split": 0.23063}, "n_reads": 12553, "n_cand": 13112, "read_suppressed_frac": 0.1025, "dropout_reads_logged": 12553, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 780.0394287109375, "n_util_obs": 654, "util_count_spearman": 0.018155599565590003}, "probe_overhead_sec_total": 308.096, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.020198, "alpha": 0.03, "t": 1381.1}
+{"event": "consolidate", "step": 4300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1396.2}
+{"event": "train", "step": 4300, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.34147, "bpb": 1.93533, "chunk_loss": 0.030425, "F": 0.2019, "G": 0.22017, "chunks": 206.75, "bytes_per_chunk": 4.953, "boundary_entropy": 0.13279, "router_noise": 0.0, "grad_norm": 0.6938, "lr": 8.5e-05, "router_trainable": true, "sec_per_step": 0.308, "probe_sec": 0.046, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03507, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.14837, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.21569, "trunc_abs_frac": 0.60308, "mean_delta_full": 0.526, "read_hit_rate": {"split": 0.93972}, "mean_gate": {"split": 0.22117}, "n_reads": 11598, "n_cand": 12342, "read_suppressed_frac": 0.1006, "dropout_reads_logged": 11598, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 789.3436279296875, "n_util_obs": 659, "util_count_spearman": 0.026108035974149492}, "probe_overhead_sec_total": 312.099, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.023266, "alpha": 0.03, "t": 1396.5}
+{"event": "consolidate", "step": 4350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1413.6}
+{"event": "train", "step": 4350, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.31247, "bpb": 1.89349, "chunk_loss": 0.030474, "F": 0.20511, "G": 0.21871, "chunks": 210.03, "bytes_per_chunk": 4.875, "boundary_entropy": 0.12454, "router_noise": 0.0, "grad_norm": 0.6907, "lr": 8.21e-05, "router_trainable": true, "sec_per_step": 0.357, "probe_sec": 0.0647, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0372, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.09005, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 3.43934, "trunc_abs_frac": 0.928, "mean_delta_full": -0.01974, "read_hit_rate": {"split": 0.94669}, "mean_gate": {"split": 0.22361}, "n_reads": 11915, "n_cand": 12586, "read_suppressed_frac": 0.1001, "dropout_reads_logged": 11915, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 798.5413208007812, "n_util_obs": 665, "util_count_spearman": 0.033580411203048804}, "probe_overhead_sec_total": 315.986, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.025055, "alpha": 0.03, "t": 1414.3}
+{"event": "consolidate", "step": 4400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1434.5}
+{"event": "probe", "step": 4400, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3570.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.205137, "cps": 27.890625, "mean_p": 0.208089, "boundary_entropy": 0.100162, "bpb": 1.879263}, "tr": {"chunks": 3349.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.184997, "cps": 26.164062, "mean_p": 0.221573, "boundary_entropy": 0.209513, "bpb": 2.05705}, "fi": {"chunks": 4407.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.234278, "cps": 34.429688, "mean_p": 0.24069, "boundary_entropy": 0.123816, "bpb": 2.187118}}, "bpb": {"en": 1.87926, "tr": 2.05705, "fi": 2.18712}, "gini_cpc": 0.052616, "gini_cps": 0.062276, "bpic": 5.009006, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9018213240110163, "fi": 1.1420551358069144}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 807.5830078125, "n_util_obs": 672, "util_count_spearman": 0.046958225029630395}, "t": 1435.4}
+{"event": "eval", "step": 4400, "phase": 3, "bpb": {"en": 1.89794, "tr": 2.063, "fi": 2.19429}, "bytes_per_chunk": {"en": 4.64, "tr": 5.892, "fi": 4.506}, "bpb_mean": 2.05174, "frozen_inventory": true, "t": 1436.8}
+{"event": "train", "step": 4400, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.36223, "bpb": 1.96528, "chunk_loss": 0.030511, "F": 0.21082, "G": 0.217, "chunks": 215.88, "bytes_per_chunk": 4.743, "boundary_entropy": 0.10388, "router_noise": 0.0, "grad_norm": 0.7439, "lr": 7.92e-05, "router_trainable": true, "sec_per_step": 0.478, "probe_sec": 0.0853, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0724, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.0303, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -1.2722, "trunc_abs_frac": 2.37449, "mean_delta_full": 0.07332, "read_hit_rate": {"split": 0.95297}, "mean_gate": {"split": 0.22496}, "n_reads": 12381, "n_cand": 12992, "read_suppressed_frac": 0.0964, "dropout_reads_logged": 12381, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 807.7728271484375, "n_util_obs": 673, "util_count_spearman": 0.04636963371393479}, "probe_overhead_sec_total": 319.59, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.020622, "alpha": 0.03, "t": 1438.2}
+{"event": "consolidate", "step": 4450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1452.8}
+{"event": "train", "step": 4450, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38494, "bpb": 1.99804, "chunk_loss": 0.0304, "F": 0.20285, "G": 0.21712, "chunks": 207.72, "bytes_per_chunk": 4.93, "boundary_entropy": 0.12488, "router_noise": 0.0, "grad_norm": 0.6694, "lr": 7.63e-05, "router_trainable": true, "sec_per_step": 0.296, "probe_sec": 0.0472, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03422, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.06505, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.70217, "trunc_abs_frac": 0.63568, "mean_delta_full": 0.39478, "read_hit_rate": {"split": 0.94479}, "mean_gate": {"split": 0.22875}, "n_reads": 11774, "n_cand": 12462, "read_suppressed_frac": 0.0978, "dropout_reads_logged": 11774, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 816.9910278320312, "n_util_obs": 681, "util_count_spearman": 0.04969479787670057}, "probe_overhead_sec_total": 323.143, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.02456, "alpha": 0.03, "t": 1453.0}
+{"event": "consolidate", "step": 4500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1468.4}
+{"event": "train", "step": 4500, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.3964, "bpb": 2.01458, "chunk_loss": 0.030611, "F": 0.2135, "G": 0.22303, "chunks": 218.62, "bytes_per_chunk": 4.684, "boundary_entropy": 0.10931, "router_noise": 0.0, "grad_norm": 0.6597, "lr": 7.36e-05, "router_trainable": true, "sec_per_step": 0.311, "probe_sec": 0.0364, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02663, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.10821, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 1.28311, "trunc_abs_frac": 1.46732, "mean_delta_full": 0.1424, "read_hit_rate": {"split": 0.94791}, "mean_gate": {"split": 0.21671}, "n_reads": 12448, "n_cand": 13132, "read_suppressed_frac": 0.0995, "dropout_reads_logged": 12448, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 826.2559204101562, "n_util_obs": 685, "util_count_spearman": 0.04603020673865393}, "probe_overhead_sec_total": 327.186, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.021096, "alpha": 0.03, "t": 1468.5}
+{"event": "consolidate", "step": 4550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1481.5}
+{"event": "train", "step": 4550, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.3107, "bpb": 1.89094, "chunk_loss": 0.030466, "F": 0.20462, "G": 0.21781, "chunks": 209.53, "bytes_per_chunk": 4.887, "boundary_entropy": 0.12377, "router_noise": 0.0, "grad_norm": 0.6732, "lr": 7.09e-05, "router_trainable": true, "sec_per_step": 0.265, "probe_sec": 0.0422, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03221, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.12984, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 1.19004, "trunc_abs_frac": 0.60603, "mean_delta_full": 0.00827, "read_hit_rate": {"split": 0.94477}, "mean_gate": {"split": 0.21282}, "n_reads": 11923, "n_cand": 12620, "read_suppressed_frac": 0.1005, "dropout_reads_logged": 11923, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 835.531005859375, "n_util_obs": 689, "util_count_spearman": 0.040397494368294165}, "probe_overhead_sec_total": 330.601, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.020546, "alpha": 0.03, "t": 1481.8}
+{"event": "consolidate", "step": 4600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1494.2}
+{"event": "probe", "step": 4600, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3581.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.205769, "cps": 27.976562, "mean_p": 0.205784, "boundary_entropy": 0.099149, "bpb": 1.865062}, "tr": {"chunks": 3425.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.189195, "cps": 26.757812, "mean_p": 0.222501, "boundary_entropy": 0.206957, "bpb": 2.042874}, "fi": {"chunks": 4428.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.235394, "cps": 34.59375, "mean_p": 0.240912, "boundary_entropy": 0.122723, "bpb": 2.175791}}, "bpb": {"en": 1.86506, "tr": 2.04287, "fi": 2.17579}, "gini_cpc": 0.04886, "gini_cps": 0.058481, "bpic": 4.961693, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9194536127572956, "fi": 1.1439723508993347}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 844.6511840820312, "n_util_obs": 692, "util_count_spearman": 0.04098098126698722}, "t": 1495.0}
+{"event": "eval", "step": 4600, "phase": 3, "bpb": {"en": 1.88511, "tr": 2.05136, "fi": 2.18149}, "bytes_per_chunk": {"en": 4.644, "tr": 5.769, "fi": 4.495}, "bpb_mean": 2.03932, "frozen_inventory": true, "t": 1496.5}
+{"event": "train", "step": 4600, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.37015, "bpb": 1.97671, "chunk_loss": 0.030546, "F": 0.21121, "G": 0.21793, "chunks": 216.28, "bytes_per_chunk": 4.735, "boundary_entropy": 0.10722, "router_noise": 0.0, "grad_norm": 0.7133, "lr": 6.82e-05, "router_trainable": true, "sec_per_step": 0.32, "probe_sec": 0.0776, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06489, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.04356, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.02316, "trunc_abs_frac": 2.0777, "mean_delta_full": 0.08722, "read_hit_rate": {"split": 0.9476}, "mean_gate": {"split": 0.21874}, "n_reads": 12298, "n_cand": 12978, "read_suppressed_frac": 0.1012, "dropout_reads_logged": 12298, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 844.8397216796875, "n_util_obs": 692, "util_count_spearman": 0.040796531349858445}, "probe_overhead_sec_total": 334.171, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.0235, "alpha": 0.03, "t": 1497.8}
+{"event": "consolidate", "step": 4650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1510.3}
+{"event": "train", "step": 4650, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38473, "bpb": 1.99775, "chunk_loss": 0.030541, "F": 0.2106, "G": 0.22173, "chunks": 215.66, "bytes_per_chunk": 4.748, "boundary_entropy": 0.11844, "router_noise": 0.0, "grad_norm": 0.7743, "lr": 6.57e-05, "router_trainable": true, "sec_per_step": 0.257, "probe_sec": 0.1441, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.13487, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.00569, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.16909, "trunc_abs_frac": 0.48851, "mean_delta_full": -0.34265, "read_hit_rate": {"split": 0.95059}, "mean_gate": {"split": 0.21694}, "n_reads": 12177, "n_cand": 12810, "read_suppressed_frac": 0.1001, "dropout_reads_logged": 12177, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 854.107177734375, "n_util_obs": 699, "util_count_spearman": 0.04056075185194325}, "probe_overhead_sec_total": 338.104, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.019485, "alpha": 0.03, "t": 1510.6}
+{"event": "consolidate", "step": 4700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1522.8}
+{"event": "train", "step": 4700, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35194, "bpb": 1.95044, "chunk_loss": 0.030505, "F": 0.2067, "G": 0.2211, "chunks": 211.66, "bytes_per_chunk": 4.838, "boundary_entropy": 0.12502, "router_noise": 0.0, "grad_norm": 0.6893, "lr": 6.32e-05, "router_trainable": true, "sec_per_step": 0.249, "probe_sec": 0.0751, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06042, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.06643, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.09475, "trunc_abs_frac": 0.97765, "mean_delta_full": 0.20529, "read_hit_rate": {"split": 0.94725}, "mean_gate": {"split": 0.21806}, "n_reads": 11996, "n_cand": 12664, "read_suppressed_frac": 0.0992, "dropout_reads_logged": 11996, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 863.419189453125, "n_util_obs": 708, "util_count_spearman": 0.0448745493765804}, "probe_overhead_sec_total": 341.354, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.024653, "alpha": 0.03, "t": 1523.1}
+{"event": "consolidate", "step": 4750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1535.4}
+{"event": "train", "step": 4750, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.44445, "bpb": 2.0839, "chunk_loss": 0.03049, "F": 0.20605, "G": 0.21928, "chunks": 211.0, "bytes_per_chunk": 4.853, "boundary_entropy": 0.12311, "router_noise": 0.0, "grad_norm": 0.7584, "lr": 6.08e-05, "router_trainable": true, "sec_per_step": 0.253, "probe_sec": 0.1316, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.12139, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.02647, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -13.09301, "trunc_abs_frac": 0.52214, "mean_delta_full": -0.00854, "read_hit_rate": {"split": 0.93956}, "mean_gate": {"split": 0.21129}, "n_reads": 11876, "n_cand": 12640, "read_suppressed_frac": 0.1044, "dropout_reads_logged": 11876, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 872.65673828125, "n_util_obs": 716, "util_count_spearman": 0.05277357500195397}, "probe_overhead_sec_total": 345.195, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.021536, "alpha": 0.03, "t": 1535.7}
+{"event": "consolidate", "step": 4800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1547.6}
+{"event": "probe", "step": 4800, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3575.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.205424, "cps": 27.929688, "mean_p": 0.207715, "boundary_entropy": 0.09523, "bpb": 1.854419}, "tr": {"chunks": 3423.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.189085, "cps": 26.742188, "mean_p": 0.223588, "boundary_entropy": 0.209927, "bpb": 2.032991}, "fi": {"chunks": 4434.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.235713, "cps": 34.640625, "mean_p": 0.24258, "boundary_entropy": 0.122143, "bpb": 2.162241}}, "bpb": {"en": 1.85442, "tr": 2.03299, "fi": 2.16224}, "gini_cpc": 0.049325, "gini_cps": 0.058957, "bpic": 4.962561, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9204589433656439, "fi": 1.1474450040948367}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 881.68505859375, "n_util_obs": 723, "util_count_spearman": 0.06565888252961105}, "t": 1548.5}
+{"event": "eval", "step": 4800, "phase": 3, "bpb": {"en": 1.87387, "tr": 2.03944, "fi": 2.16855}, "bytes_per_chunk": {"en": 4.649, "tr": 5.792, "fi": 4.484}, "bpb_mean": 2.02728, "frozen_inventory": true, "t": 1549.8}
+{"event": "train", "step": 4800, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32197, "bpb": 1.90719, "chunk_loss": 0.03049, "F": 0.20972, "G": 0.21675, "chunks": 214.75, "bytes_per_chunk": 4.768, "boundary_entropy": 0.10433, "router_noise": 0.0, "grad_norm": 0.6756, "lr": 5.85e-05, "router_trainable": true, "sec_per_step": 0.311, "probe_sec": 0.0962, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0864, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.02488, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.10676, "trunc_abs_frac": 0.39743, "mean_delta_full": 0.0343, "read_hit_rate": {"split": 0.95251}, "mean_gate": {"split": 0.22301}, "n_reads": 12236, "n_cand": 12846, "read_suppressed_frac": 0.0997, "dropout_reads_logged": 12236, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 881.8726196289062, "n_util_obs": 723, "util_count_spearman": 0.06568480519917827}, "probe_overhead_sec_total": 348.315, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.022721, "alpha": 0.03, "t": 1551.3}
+{"event": "consolidate", "step": 4850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1563.4}
+{"event": "train", "step": 4850, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.3226, "bpb": 1.90811, "chunk_loss": 0.030666, "F": 0.21594, "G": 0.22473, "chunks": 221.12, "bytes_per_chunk": 4.631, "boundary_entropy": 0.11023, "router_noise": 0.0, "grad_norm": 0.6645, "lr": 5.62e-05, "router_trainable": true, "sec_per_step": 0.249, "probe_sec": 0.057, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04425, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.04764, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 2.34872, "trunc_abs_frac": 1.31002, "mean_delta_full": -0.02612, "read_hit_rate": {"split": 0.95403}, "mean_gate": {"split": 0.21867}, "n_reads": 12618, "n_cand": 13226, "read_suppressed_frac": 0.1011, "dropout_reads_logged": 12618, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 891.1121215820312, "n_util_obs": 729, "util_count_spearman": 0.05555004602721161}, "probe_overhead_sec_total": 351.58, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.022395, "alpha": 0.03, "t": 1563.8}
+{"event": "consolidate", "step": 4900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1577.1}
+{"event": "train", "step": 4900, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.37746, "bpb": 1.98725, "chunk_loss": 0.030459, "F": 0.20633, "G": 0.21676, "chunks": 211.28, "bytes_per_chunk": 4.847, "boundary_entropy": 0.11485, "router_noise": 0.0, "grad_norm": 0.8023, "lr": 5.41e-05, "router_trainable": true, "sec_per_step": 0.273, "probe_sec": 0.0864, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07075, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.00559, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": -0.89459, "trunc_abs_frac": 0.42152, "mean_delta_full": 0.06464, "read_hit_rate": {"split": 0.94481}, "mean_gate": {"split": 0.21917}, "n_reads": 12069, "n_cand": 12774, "read_suppressed_frac": 0.1013, "dropout_reads_logged": 12069, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 900.3713989257812, "n_util_obs": 733, "util_count_spearman": 0.057504384819186245}, "probe_overhead_sec_total": 355.6, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.025594, "alpha": 0.03, "t": 1577.4}
+{"event": "consolidate", "step": 4950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1589.8}
+{"event": "train", "step": 4950, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.31055, "bpb": 1.89073, "chunk_loss": 0.030555, "F": 0.20514, "G": 0.22011, "chunks": 210.06, "bytes_per_chunk": 4.875, "boundary_entropy": 0.12871, "router_noise": 0.0, "grad_norm": 0.7055, "lr": 5.2e-05, "router_trainable": true, "sec_per_step": 0.254, "probe_sec": 0.083, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06851, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.10846, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.76757, "trunc_abs_frac": 0.59954, "mean_delta_full": 0.422, "read_hit_rate": {"split": 0.94238}, "mean_gate": {"split": 0.21773}, "n_reads": 11612, "n_cand": 12322, "read_suppressed_frac": 0.1001, "dropout_reads_logged": 11612, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 909.7161865234375, "n_util_obs": 739, "util_count_spearman": 0.06360005860212334}, "probe_overhead_sec_total": 358.876, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.023409, "alpha": 0.03, "t": 1590.1}
+{"event": "consolidate", "step": 5000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1602.8}
+{"event": "probe", "step": 5000, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3584.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.205942, "cps": 28.0, "mean_p": 0.206234, "boundary_entropy": 0.096156, "bpb": 1.846848}, "tr": {"chunks": 3380.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.186709, "cps": 26.40625, "mean_p": 0.220373, "boundary_entropy": 0.206466, "bpb": 2.019086}, "fi": {"chunks": 4435.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.235766, "cps": 34.648438, "mean_p": 0.241746, "boundary_entropy": 0.118597, "bpb": 2.148602}}, "bpb": {"en": 1.84685, "tr": 2.01909, "fi": 2.1486}, "gini_cpc": 0.052043, "gini_cps": 0.061701, "bpic": 4.976928, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9066136803489556, "fi": 1.1448217186989755}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 918.8258666992188, "n_util_obs": 742, "util_count_spearman": 0.054918255563510474}, "t": 1603.6}
+{"event": "eval", "step": 5000, "phase": 3, "bpb": {"en": 1.86579, "tr": 2.02697, "fi": 2.15447}, "bytes_per_chunk": {"en": 4.644, "tr": 5.831, "fi": 4.487}, "bpb_mean": 2.01574, "frozen_inventory": true, "t": 1605.0}
+{"event": "train", "step": 5000, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.41651, "bpb": 2.04359, "chunk_loss": 0.030649, "F": 0.21509, "G": 0.22601, "chunks": 220.25, "bytes_per_chunk": 4.649, "boundary_entropy": 0.11523, "router_noise": 0.0, "grad_norm": 0.741, "lr": 5e-05, "router_trainable": true, "sec_per_step": 0.325, "probe_sec": 0.1386, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.12906, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.03209, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -1.31954, "trunc_abs_frac": 1.70761, "mean_delta_full": -0.12869, "read_hit_rate": {"split": 0.93787}, "mean_gate": {"split": 0.201}, "n_reads": 12258, "n_cand": 13070, "read_suppressed_frac": 0.1017, "dropout_reads_logged": 12258, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 919.0112915039062, "n_util_obs": 742, "util_count_spearman": 0.0546440299981121}, "probe_overhead_sec_total": 362.55, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.023341, "alpha": 0.03, "t": 1606.4}
+{"event": "consolidate", "step": 5050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1618.8}
+{"event": "train", "step": 5050, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.27323, "bpb": 1.83688, "chunk_loss": 0.030569, "F": 0.21216, "G": 0.22035, "chunks": 217.25, "bytes_per_chunk": 4.713, "boundary_entropy": 0.10901, "router_noise": 0.0, "grad_norm": 0.6747, "lr": 4.81e-05, "router_trainable": true, "sec_per_step": 0.253, "probe_sec": 0.0383, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02775, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.12141, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.53646, "trunc_abs_frac": 4.58019, "mean_delta_full": 0.45236, "read_hit_rate": {"split": 0.96055}, "mean_gate": {"split": 0.21602}, "n_reads": 12637, "n_cand": 13156, "read_suppressed_frac": 0.1048, "dropout_reads_logged": 12637, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 928.314697265625, "n_util_obs": 749, "util_count_spearman": 0.06174935203440469}, "probe_overhead_sec_total": 365.904, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.025956, "alpha": 0.03, "t": 1619.0}
+{"event": "consolidate", "step": 5100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1631.5}
+{"event": "train", "step": 5100, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32199, "bpb": 1.90723, "chunk_loss": 0.030495, "F": 0.20865, "G": 0.21726, "chunks": 213.66, "bytes_per_chunk": 4.793, "boundary_entropy": 0.10962, "router_noise": 0.0, "grad_norm": 0.6827, "lr": 4.63e-05, "router_trainable": true, "sec_per_step": 0.254, "probe_sec": 0.0376, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02768, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.07939, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 1.27644, "trunc_abs_frac": 1.34638, "mean_delta_full": -0.12695, "read_hit_rate": {"split": 0.95255}, "mean_gate": {"split": 0.21575}, "n_reads": 12124, "n_cand": 12728, "read_suppressed_frac": 0.096, "dropout_reads_logged": 12124, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 937.56103515625, "n_util_obs": 752, "util_count_spearman": 0.056363890301324965}, "probe_overhead_sec_total": 368.991, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.022843, "alpha": 0.03, "t": 1631.7}
+{"event": "consolidate", "step": 5150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1643.9}
+{"event": "train", "step": 5150, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.36373, "bpb": 1.96745, "chunk_loss": 0.030568, "F": 0.20959, "G": 0.21993, "chunks": 214.62, "bytes_per_chunk": 4.771, "boundary_entropy": 0.11662, "router_noise": 0.0, "grad_norm": 0.7338, "lr": 4.45e-05, "router_trainable": true, "sec_per_step": 0.249, "probe_sec": 0.0615, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04895, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.00151, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 0.42591, "trunc_abs_frac": 1.30784, "mean_delta_full": -0.17415, "read_hit_rate": {"split": 0.94276}, "mean_gate": {"split": 0.21179}, "n_reads": 11973, "n_cand": 12700, "read_suppressed_frac": 0.0996, "dropout_reads_logged": 11973, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 946.8301391601562, "n_util_obs": 758, "util_count_spearman": 0.06005575131177903}, "probe_overhead_sec_total": 372.452, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.025078, "alpha": 0.03, "t": 1644.2}
+{"event": "consolidate", "step": 5200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1656.4}
+{"event": "probe", "step": 5200, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3586.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.206056, "cps": 28.015625, "mean_p": 0.206686, "boundary_entropy": 0.095495, "bpb": 1.83889}, "tr": {"chunks": 3385.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.186986, "cps": 26.445312, "mean_p": 0.218647, "boundary_entropy": 0.202317, "bpb": 2.009361}, "fi": {"chunks": 4406.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.234225, "cps": 34.421875, "mean_p": 0.238955, "boundary_entropy": 0.115902, "bpb": 2.143784}}, "bpb": {"en": 1.83889, "tr": 2.00936, "fi": 2.14378}, "gini_cpc": 0.050206, "gini_cps": 0.059828, "bpic": 4.986552, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9074484362102352, "fi": 1.13670152946252}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 955.908935546875, "n_util_obs": 761, "util_count_spearman": 0.06550447148489795}, "t": 1657.3}
+{"event": "eval", "step": 5200, "phase": 3, "bpb": {"en": 1.85869, "tr": 2.01719, "fi": 2.15005}, "bytes_per_chunk": {"en": 4.642, "tr": 5.819, "fi": 4.502}, "bpb_mean": 2.00864, "frozen_inventory": true, "t": 1658.7}
+{"event": "train", "step": 5200, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.29597, "bpb": 1.86969, "chunk_loss": 0.030704, "F": 0.2182, "G": 0.2237, "chunks": 223.44, "bytes_per_chunk": 4.583, "boundary_entropy": 0.10563, "router_noise": 0.0, "grad_norm": 0.6888, "lr": 4.29e-05, "router_trainable": true, "sec_per_step": 0.316, "probe_sec": 0.0448, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03405, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.02091, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.34913, "trunc_abs_frac": 7.49245, "mean_delta_full": 0.23569, "read_hit_rate": {"split": 0.95882}, "mean_gate": {"split": 0.20871}, "n_reads": 12854, "n_cand": 13406, "read_suppressed_frac": 0.0953, "dropout_reads_logged": 12854, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 956.1056518554688, "n_util_obs": 761, "util_count_spearman": 0.06544240394758366}, "probe_overhead_sec_total": 375.981, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.023448, "alpha": 0.03, "t": 1660.0}
+{"event": "consolidate", "step": 5250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1672.9}
+{"event": "train", "step": 5250, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.29365, "bpb": 1.86634, "chunk_loss": 0.030472, "F": 0.20743, "G": 0.21822, "chunks": 212.41, "bytes_per_chunk": 4.821, "boundary_entropy": 0.1178, "router_noise": 0.0, "grad_norm": 0.695, "lr": 4.14e-05, "router_trainable": true, "sec_per_step": 0.264, "probe_sec": 0.1478, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.13695, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.0286, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.68896, "trunc_abs_frac": 4.05397, "mean_delta_full": -0.20319, "read_hit_rate": {"split": 0.94958}, "mean_gate": {"split": 0.20938}, "n_reads": 12111, "n_cand": 12754, "read_suppressed_frac": 0.1073, "dropout_reads_logged": 12111, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 965.4287109375, "n_util_obs": 766, "util_count_spearman": 0.061721059993945705}, "probe_overhead_sec_total": 379.347, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.021712, "alpha": 0.03, "t": 1673.2}
+{"event": "consolidate", "step": 5300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1685.5}
+{"event": "train", "step": 5300, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.29102, "bpb": 1.86255, "chunk_loss": 0.030441, "F": 0.20819, "G": 0.21368, "chunks": 213.19, "bytes_per_chunk": 4.803, "boundary_entropy": 0.10729, "router_noise": 0.0, "grad_norm": 0.6485, "lr": 3.99e-05, "router_trainable": true, "sec_per_step": 0.25, "probe_sec": 0.0757, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06215, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.06774, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 1.25671, "trunc_abs_frac": 1.25671, "mean_delta_full": -0.17065, "read_hit_rate": {"split": 0.95349}, "mean_gate": {"split": 0.21023}, "n_reads": 12239, "n_cand": 12836, "read_suppressed_frac": 0.1035, "dropout_reads_logged": 12239, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 974.7064819335938, "n_util_obs": 776, "util_count_spearman": 0.048744193984650755}, "probe_overhead_sec_total": 383.292, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.022176, "alpha": 0.03, "t": 1685.7}
+{"event": "consolidate", "step": 5350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1697.8}
+{"event": "train", "step": 5350, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.28832, "bpb": 1.85865, "chunk_loss": 0.030499, "F": 0.20935, "G": 0.21728, "chunks": 214.38, "bytes_per_chunk": 4.777, "boundary_entropy": 0.10973, "router_noise": 0.0, "grad_norm": 0.6757, "lr": 3.86e-05, "router_trainable": true, "sec_per_step": 0.249, "probe_sec": 0.0924, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08168, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.0887, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 1.26612, "trunc_abs_frac": 1.46917, "mean_delta_full": 0.53632, "read_hit_rate": {"split": 0.95626}, "mean_gate": {"split": 0.21433}, "n_reads": 12221, "n_cand": 12780, "read_suppressed_frac": 0.1015, "dropout_reads_logged": 12221, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 983.983154296875, "n_util_obs": 780, "util_count_spearman": 0.04782826906408147}, "probe_overhead_sec_total": 386.648, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.024135, "alpha": 0.03, "t": 1698.2}
+{"event": "consolidate", "step": 5400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1710.2}
+{"event": "probe", "step": 5400, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3576.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.205482, "cps": 27.9375, "mean_p": 0.205908, "boundary_entropy": 0.095203, "bpb": 1.83392}, "tr": {"chunks": 3416.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.188698, "cps": 26.6875, "mean_p": 0.218964, "boundary_entropy": 0.20106, "bpb": 2.003152}, "fi": {"chunks": 4416.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.234756, "cps": 34.5, "mean_p": 0.239669, "boundary_entropy": 0.114603, "bpb": 2.131421}}, "bpb": {"en": 1.83392, "tr": 2.00315, "fi": 2.13142}, "gini_cpc": 0.048821, "gini_cps": 0.058439, "bpic": 4.973001, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9183197415831185, "fi": 1.1424673340138338}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 993.1072998046875, "n_util_obs": 785, "util_count_spearman": 0.04899248177020963}, "t": 1711.1}
+{"event": "eval", "step": 5400, "phase": 3, "bpb": {"en": 1.85291, "tr": 2.0111, "fi": 2.13774}, "bytes_per_chunk": {"en": 4.646, "tr": 5.796, "fi": 4.489}, "bpb_mean": 2.00058, "frozen_inventory": true, "t": 1712.5}
+{"event": "train", "step": 5400, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.36246, "bpb": 1.96562, "chunk_loss": 0.030556, "F": 0.20908, "G": 0.2247, "chunks": 214.09, "bytes_per_chunk": 4.783, "boundary_entropy": 0.12976, "router_noise": 0.0, "grad_norm": 0.7732, "lr": 3.73e-05, "router_trainable": true, "sec_per_step": 0.315, "probe_sec": 0.1099, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0975, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.10265, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.98159, "trunc_abs_frac": 1.03125, "mean_delta_full": 0.04309, "read_hit_rate": {"split": 0.93687}, "mean_gate": {"split": 0.20518}, "n_reads": 11812, "n_cand": 12608, "read_suppressed_frac": 0.0984, "dropout_reads_logged": 11812, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 993.2837524414062, "n_util_obs": 786, "util_count_spearman": 0.04698533723890864}, "probe_overhead_sec_total": 390.133, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.022247, "alpha": 0.03, "t": 1713.9}
+{"event": "consolidate", "step": 5450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1726.3}
+{"event": "train", "step": 5450, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.30534, "bpb": 1.88321, "chunk_loss": 0.030468, "F": 0.20758, "G": 0.21594, "chunks": 212.56, "bytes_per_chunk": 4.817, "boundary_entropy": 0.11075, "router_noise": 0.0, "grad_norm": 0.6859, "lr": 3.62e-05, "router_trainable": true, "sec_per_step": 0.254, "probe_sec": 0.1149, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10351, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.11132, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.72278, "trunc_abs_frac": 0.76036, "mean_delta_full": -0.24105, "read_hit_rate": {"split": 0.95051}, "mean_gate": {"split": 0.2194}, "n_reads": 12003, "n_cand": 12628, "read_suppressed_frac": 0.0993, "dropout_reads_logged": 12003, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 1002.5706787109375, "n_util_obs": 795, "util_count_spearman": 0.048435385639275234}, "probe_overhead_sec_total": 393.692, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.024719, "alpha": 0.03, "t": 1726.6}
+{"event": "consolidate", "step": 5500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1738.8}
+{"event": "train", "step": 5500, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.29908, "bpb": 1.87417, "chunk_loss": 0.03057, "F": 0.21237, "G": 0.22037, "chunks": 217.47, "bytes_per_chunk": 4.709, "boundary_entropy": 0.10906, "router_noise": 0.0, "grad_norm": 0.6779, "lr": 3.51e-05, "router_trainable": true, "sec_per_step": 0.248, "probe_sec": 0.0889, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07824, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.15814, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 1.43364, "trunc_abs_frac": 1.43364, "mean_delta_full": 0.172, "read_hit_rate": {"split": 0.9553}, "mean_gate": {"split": 0.21284}, "n_reads": 12482, "n_cand": 13066, "read_suppressed_frac": 0.1, "dropout_reads_logged": 12482, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 1011.7981567382812, "n_util_obs": 799, "util_count_spearman": 0.041199222070309974}, "probe_overhead_sec_total": 397.071, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.022146, "alpha": 0.03, "t": 1739.0}
+{"event": "consolidate", "step": 5550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1751.2}
+{"event": "train", "step": 5550, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.29478, "bpb": 1.86797, "chunk_loss": 0.030648, "F": 0.21225, "G": 0.22078, "chunks": 217.34, "bytes_per_chunk": 4.711, "boundary_entropy": 0.10944, "router_noise": 0.0, "grad_norm": 0.6578, "lr": 3.41e-05, "router_trainable": true, "sec_per_step": 0.249, "probe_sec": 0.0839, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07043, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.13782, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.19132, "trunc_abs_frac": 4.63044, "mean_delta_full": 0.28784, "read_hit_rate": {"split": 0.95158}, "mean_gate": {"split": 0.20623}, "n_reads": 12323, "n_cand": 12950, "read_suppressed_frac": 0.0963, "dropout_reads_logged": 12323, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 1021.1416015625, "n_util_obs": 806, "util_count_spearman": 0.034029776550083206}, "probe_overhead_sec_total": 400.868, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.023621, "alpha": 0.03, "t": 1751.5}
+{"event": "consolidate", "step": 5600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1764.2}
+{"event": "probe", "step": 5600, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3581.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.205769, "cps": 27.976562, "mean_p": 0.206222, "boundary_entropy": 0.095808, "bpb": 1.828573}, "tr": {"chunks": 3441.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.190079, "cps": 26.882812, "mean_p": 0.221221, "boundary_entropy": 0.202266, "bpb": 1.997778}, "fi": {"chunks": 4412.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.234544, "cps": 34.46875, "mean_p": 0.239373, "boundary_entropy": 0.113845, "bpb": 2.124419}}, "bpb": {"en": 1.82857, "tr": 1.99778, "fi": 2.12442}, "gini_cpc": 0.047023, "gini_cps": 0.056615, "bpic": 4.961693, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9237488705103224, "fi": 1.1398387561354708}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 1030.1839599609375, "n_util_obs": 815, "util_count_spearman": 0.03840479598735401}, "t": 1765.0}
+{"event": "eval", "step": 5600, "phase": 3, "bpb": {"en": 1.84822, "tr": 2.00474, "fi": 2.13089}, "bytes_per_chunk": {"en": 4.643, "tr": 5.727, "fi": 4.495}, "bpb_mean": 1.99462, "frozen_inventory": true, "t": 1766.5}
+{"event": "train", "step": 5600, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.34637, "bpb": 1.94241, "chunk_loss": 0.03043, "F": 0.20422, "G": 0.2152, "chunks": 209.12, "bytes_per_chunk": 4.897, "boundary_entropy": 0.1171, "router_noise": 0.0, "grad_norm": 0.7108, "lr": 3.33e-05, "router_trainable": true, "sec_per_step": 0.326, "probe_sec": 0.0469, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03701, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.23275, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 0.4861, "trunc_abs_frac": 0.53466, "mean_delta_full": 0.65411, "read_hit_rate": {"split": 0.94208}, "mean_gate": {"split": 0.20747}, "n_reads": 11874, "n_cand": 12604, "read_suppressed_frac": 0.1048, "dropout_reads_logged": 11874, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 1030.365966796875, "n_util_obs": 815, "util_count_spearman": 0.038434954980149615}, "probe_overhead_sec_total": 404.442, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.021711, "alpha": 0.03, "t": 1767.8}
+{"event": "consolidate", "step": 5650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1780.2}
+{"event": "train", "step": 5650, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.29624, "bpb": 1.87009, "chunk_loss": 0.030524, "F": 0.21082, "G": 0.21869, "chunks": 215.88, "bytes_per_chunk": 4.743, "boundary_entropy": 0.10671, "router_noise": 0.0, "grad_norm": 0.6985, "lr": 3.25e-05, "router_trainable": true, "sec_per_step": 0.253, "probe_sec": 0.0448, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03341, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.25824, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.53079, "trunc_abs_frac": 0.35825, "mean_delta_full": 0.45756, "read_hit_rate": {"split": 0.95378}, "mean_gate": {"split": 0.20528}, "n_reads": 12340, "n_cand": 12938, "read_suppressed_frac": 0.1045, "dropout_reads_logged": 12340, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 1039.6109619140625, "n_util_obs": 823, "util_count_spearman": 0.02727273382866045}, "probe_overhead_sec_total": 407.668, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.024697, "alpha": 0.03, "t": 1780.4}
+{"event": "consolidate", "step": 5700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1792.5}
+{"event": "train", "step": 5700, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.42141, "bpb": 2.05066, "chunk_loss": 0.030487, "F": 0.20572, "G": 0.21739, "chunks": 210.66, "bytes_per_chunk": 4.861, "boundary_entropy": 0.11586, "router_noise": 0.0, "grad_norm": 0.784, "lr": 3.18e-05, "router_trainable": true, "sec_per_step": 0.251, "probe_sec": 0.0565, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0336, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.18754, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 0.7761, "trunc_abs_frac": 1.47186, "mean_delta_full": 0.29998, "read_hit_rate": {"split": 0.93329}, "mean_gate": {"split": 0.19378}, "n_reads": 11681, "n_cand": 12516, "read_suppressed_frac": 0.1047, "dropout_reads_logged": 11681, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 1048.855712890625, "n_util_obs": 826, "util_count_spearman": 0.026620755885226745}, "probe_overhead_sec_total": 410.803, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.022794, "alpha": 0.03, "t": 1793.0}
+{"event": "consolidate", "step": 5750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1813.7}
+{"event": "train", "step": 5750, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.25562, "bpb": 1.81148, "chunk_loss": 0.030555, "F": 0.21036, "G": 0.21951, "chunks": 215.41, "bytes_per_chunk": 4.754, "boundary_entropy": 0.11149, "router_noise": 0.0, "grad_norm": 0.7334, "lr": 3.13e-05, "router_trainable": true, "sec_per_step": 0.425, "probe_sec": 0.055, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03978, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.0286, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 1.38322, "trunc_abs_frac": 4.94314, "mean_delta_full": 0.01746, "read_hit_rate": {"split": 0.95631}, "mean_gate": {"split": 0.21229}, "n_reads": 12170, "n_cand": 12726, "read_suppressed_frac": 0.1046, "dropout_reads_logged": 12170, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 1058.11083984375, "n_util_obs": 835, "util_count_spearman": 0.027153910306266172}, "probe_overhead_sec_total": 414.713, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.019844, "alpha": 0.03, "t": 1814.2}
+{"event": "consolidate", "step": 5800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1831.5}
+{"event": "probe", "step": 5800, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3587.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.206114, "cps": 28.023438, "mean_p": 0.205686, "boundary_entropy": 0.09366, "bpb": 1.823819}, "tr": {"chunks": 3423.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.189085, "cps": 26.742188, "mean_p": 0.218523, "boundary_entropy": 0.198216, "bpb": 1.992067}, "fi": {"chunks": 4417.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.234809, "cps": 34.507812, "mean_p": 0.237615, "boundary_entropy": 0.111612, "bpb": 2.120841}}, "bpb": {"en": 1.82382, "tr": 1.99207, "fi": 2.12084}, "gini_cpc": 0.048385, "gini_cps": 0.057991, "bpic": 4.964733, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9173796271347023, "fi": 1.139221727519252}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 1067.2440185546875, "n_util_obs": 842, "util_count_spearman": 0.032078837978881294}, "t": 1832.4}
+{"event": "eval", "step": 5800, "phase": 3, "bpb": {"en": 1.84354, "tr": 1.99953, "fi": 2.12785}, "bytes_per_chunk": {"en": 4.646, "tr": 5.771, "fi": 4.496}, "bpb_mean": 1.99031, "frozen_inventory": true, "t": 1835.7}
+{"event": "train", "step": 5800, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.23574, "bpb": 1.7828, "chunk_loss": 0.030529, "F": 0.20898, "G": 0.21657, "chunks": 214.0, "bytes_per_chunk": 4.785, "boundary_entropy": 0.10948, "router_noise": 0.0, "grad_norm": 0.6884, "lr": 3.08e-05, "router_trainable": true, "sec_per_step": 0.486, "probe_sec": 0.0546, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03944, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.08143, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 0, "read_hit_rate": {"split": 0.95849}, "mean_gate": {"split": 0.20765}, "n_reads": 12491, "n_cand": 13032, "read_suppressed_frac": 0.0973, "dropout_reads_logged": 12491, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 1067.435302734375, "n_util_obs": 842, "util_count_spearman": 0.03211726351297536}, "probe_overhead_sec_total": 418.368, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.021513, "alpha": 0.03, "t": 1838.5}
+{"event": "consolidate", "step": 5850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1858.0}
+{"event": "train", "step": 5850, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35861, "bpb": 1.96006, "chunk_loss": 0.030611, "F": 0.21185, "G": 0.22486, "chunks": 216.94, "bytes_per_chunk": 4.72, "boundary_entropy": 0.11987, "router_noise": 0.0, "grad_norm": 0.7315, "lr": 3.05e-05, "router_trainable": true, "sec_per_step": 0.395, "probe_sec": 0.0794, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06701, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.00017, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 12.4018, "trunc_abs_frac": 0.72918, "mean_delta_full": 0.026, "read_hit_rate": {"split": 0.9368}, "mean_gate": {"split": 0.19504}, "n_reads": 12126, "n_cand": 12944, "read_suppressed_frac": 0.0997, "dropout_reads_logged": 12126, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 1076.6673583984375, "n_util_obs": 853, "util_count_spearman": 0.02770887228278201}, "probe_overhead_sec_total": 422.45, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.021191, "alpha": 0.03, "t": 1858.3}
+{"event": "consolidate", "step": 5900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1873.2}
+{"event": "train", "step": 5900, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.34428, "bpb": 1.93938, "chunk_loss": 0.03062, "F": 0.21252, "G": 0.2201, "chunks": 217.62, "bytes_per_chunk": 4.705, "boundary_entropy": 0.11051, "router_noise": 0.0, "grad_norm": 0.727, "lr": 3.02e-05, "router_trainable": true, "sec_per_step": 0.304, "probe_sec": 0.0722, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06025, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.02965, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -10.04096, "trunc_abs_frac": 2.85134, "mean_delta_full": -0.00715, "read_hit_rate": {"split": 0.94974}, "mean_gate": {"split": 0.20504}, "n_reads": 12225, "n_cand": 12872, "read_suppressed_frac": 0.0995, "dropout_reads_logged": 12225, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 1085.9200439453125, "n_util_obs": 860, "util_count_spearman": 0.03007882774534507}, "probe_overhead_sec_total": 426.107, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.024407, "alpha": 0.03, "t": 1873.4}
+{"event": "consolidate", "step": 5950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1888.7}
+{"event": "train", "step": 5950, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.28317, "bpb": 1.85123, "chunk_loss": 0.03058, "F": 0.21109, "G": 0.22066, "chunks": 216.16, "bytes_per_chunk": 4.737, "boundary_entropy": 0.10905, "router_noise": 0.0, "grad_norm": 0.6966, "lr": 3.01e-05, "router_trainable": true, "sec_per_step": 0.31, "probe_sec": 0.0577, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04595, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.0457, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 10, "trunc_ratio_sum": 0.23431, "trunc_abs_frac": 7.3159, "mean_delta_full": 0.20532, "read_hit_rate": {"split": 0.94871}, "mean_gate": {"split": 0.19511}, "n_reads": 12320, "n_cand": 12986, "read_suppressed_frac": 0.0965, "dropout_reads_logged": 12320, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 1095.1761474609375, "n_util_obs": 868, "util_count_spearman": 0.04106485837645554}, "probe_overhead_sec_total": 429.975, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.025929, "alpha": 0.03, "t": 1888.9}
+{"event": "consolidate", "step": 6000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1909.4}
+{"event": "train", "step": 5999, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.3378, "bpb": 1.93004, "chunk_loss": 0.030489, "F": 0.20813, "G": 0.21596, "chunks": 213.12, "bytes_per_chunk": 4.805, "boundary_entropy": 0.11218, "router_noise": 0.0, "grad_norm": 0.7112, "lr": 3e-05, "router_trainable": true, "sec_per_step": 0.409, "probe_sec": 0.0697, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04336, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.08894, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": -0.96146, "trunc_abs_frac": 0.96146, "mean_delta_full": 0.19006, "read_hit_rate": {"split": 0.94809}, "mean_gate": {"split": 0.21387}, "n_reads": 12109, "n_cand": 12772, "read_suppressed_frac": 0.0998, "dropout_reads_logged": 12109, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 1104.2510986328125, "n_util_obs": 870, "util_count_spearman": 0.04138498818384197}, "probe_overhead_sec_total": 433.984, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.027598, "alpha": 0.03, "t": 1909.4}
+{"event": "probe", "step": 6000, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3585.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.205999, "cps": 28.007812, "mean_p": 0.205651, "boundary_entropy": 0.094929, "bpb": 1.821045}, "tr": {"chunks": 3421.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.188974, "cps": 26.726562, "mean_p": 0.218982, "boundary_entropy": 0.198867, "bpb": 1.985302}, "fi": {"chunks": 4416.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.234756, "cps": 34.5, "mean_p": 0.238807, "boundary_entropy": 0.112021, "bpb": 2.11323}}, "bpb": {"en": 1.82105, "tr": 1.9853, "fi": 2.11323}, "gini_cpc": 0.048467, "gini_cps": 0.058075, "bpic": 4.966906, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9173551067727973, "fi": 1.1395992151836736}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 1104.2510986328125, "n_util_obs": 870, "util_count_spearman": 0.04138498818384197}, "t": 1910.2}
+{"event": "eval", "step": 6000, "phase": 3, "bpb": {"en": 1.83968, "tr": 1.99263, "fi": 2.12006}, "bytes_per_chunk": {"en": 4.644, "tr": 5.773, "fi": 4.492}, "bpb_mean": 1.98412, "frozen_inventory": true, "t": 1911.8}
+{"event": "reps_dumped", "dirs": ["/local/scratch/sas245/beetle-hnet/runs/ali_mem-split_s1/probe/reps/th0.30", "/local/scratch/sas245/beetle-hnet/runs/ali_mem-split_s1/probe/reps/th0.40", "/local/scratch/sas245/beetle-hnet/runs/ali_mem-split_s1/probe/reps/th0.50", "/local/scratch/sas245/beetle-hnet/runs/ali_mem-split_s1/probe/reps/th0.60", "/local/scratch/sas245/beetle-hnet/runs/ali_mem-split_s1/probe/reps/th0.70"], "t": 1914.4}
+{"event": "final", "run": "ali_mem-split_s1", "arch": "ali", "objective": "ratio", "condition": "A", "step": 6000, "params": 22785088, "seed": 1, "per_lang": {"en": {"cpc": 0.20599896569556972, "cps": 28.0078125, "bpb": 1.8429593685735581, "bytes_per_chunk": 4.633320371875994, "reads_per_chunk": 1.8014069072784475}, "tr": {"cpc": 0.18897420317074518, "cps": 26.7265625, "bpb": 1.9840548275014076, "bytes_per_chunk": 5.779634671987477, "reads_per_chunk": 1.590124899408023}, "fi": {"cpc": 0.23475625963531976, "cps": 34.5, "bpb": 2.138425337028634, "bytes_per_chunk": 4.477879129513854, "reads_per_chunk": 1.7251204263605617}}, "bpb_mean": 1.9884798443678668, "gini_cpc": 0.048467436322620516, "gini_cps": 0.058075059826066644, "bpic": 4.966905970933286, "mean_boundary_entropy": 0.11218297481536865, "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": 433.98393273353577, "probe_overhead_seq_frac": 0.4999166666666667, "wall_sec": 1917.5672039985657, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 1104.2510986328125, "n_util_obs": 870, "util_count_spearman": 0.04138498818384197}, "ali_params": 268096, "ali_config": {"enabled": true, "budget_B": 65536, "staging_capacity": 262144, "policy": "frequency", "value_mode": "learned", "candidate_types": ["split"], "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": 0.04138498818384197, "paired_vs_dropout": {"n": 1817, "pearson": 0.08266490306155061, "spearman": 0.07086633564890017, "mean_abs_discrepancy": 0.6155990958213806, "mean_paired": 0.08697881549596786, "mean_dropout": 0.08633054047822952}, "predictor_calibration": {"n": 4058, "all": {"n": 4058, "pearson": 0.13256802448049135, "spearman": 0.16531389590389728, "mae": 0.1590728610754013, "reliability": [{"pred": -0.0028653182089328766, "obs": 0.017683785408735275, "n": 406}, {"pred": 0.013966542668640614, "obs": 0.05502212047576904, "n": 406}, {"pred": 0.022499751299619675, "obs": 0.06273537129163742, "n": 406}, {"pred": 0.02940893918275833, "obs": 0.06503923237323761, "n": 405}, {"pred": 0.035803183913230896, "obs": 0.059896066784858704, "n": 406}, {"pred": 0.04160938039422035, "obs": 0.07249367237091064, "n": 406}, {"pred": 0.04843706637620926, "obs": 0.10839352756738663, "n": 405}, {"pred": 0.056011125445365906, "obs": 0.09082813560962677, "n": 406}, {"pred": 0.06582339853048325, "obs": 0.10861404240131378, "n": 406}, {"pred": 0.08516347408294678, "obs": 0.14782555401325226, "n": 406}]}, "admitted": {"n": 4058, "pearson": 0.13256802448049135, "spearman": 0.16531389590389728, "mae": 0.1590728610754013, "reliability": [{"pred": -0.0028653182089328766, "obs": 0.017683785408735275, "n": 406}, {"pred": 0.013966542668640614, "obs": 0.05502212047576904, "n": 406}, {"pred": 0.022499751299619675, "obs": 0.06273537129163742, "n": 406}, {"pred": 0.02940893918275833, "obs": 0.06503923237323761, "n": 405}, {"pred": 0.035803183913230896, "obs": 0.059896066784858704, "n": 406}, {"pred": 0.04160938039422035, "obs": 0.07249367237091064, "n": 406}, {"pred": 0.04843706637620926, "obs": 0.10839352756738663, "n": 405}, {"pred": 0.056011125445365906, "obs": 0.09082813560962677, "n": 406}, {"pred": 0.06582339853048325, "obs": 0.10861404240131378, "n": 406}, {"pred": 0.08516347408294678, "obs": 0.14782555401325226, "n": 406}]}, "staged": {"n": 0, "pearson": NaN, "spearman": NaN, "mae": NaN}}, "t": 1918.5}