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-random_s1", "keep_steps": [0, 1, 2, 4, 6, 8, 16, 32, 64, 128, 256, 512, 1000, 1024, 2048, 3000, 4096, 6000], "t": 2.2}
+{"event": "start", "run": "ali_mem-random_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": "random", "value_mode": "learned", "candidate_types": ["surface"], "gate": "vector", "retrieval": true, "memory_scope": "persistent", "read_dropout": 0.1, "write_rule": "mean", "eta_min": 0.0, "consolidate_every": 50, "probe_seqs": 2, "probe_reps": 8, "horizon": 32, "predictor": "mlp", "lambda_util": 0.1, "max_span": 64, "byte_classes_path": "/local/scratch/sas245/beetle-hnet/scratch/data/byte_classes.json", "gate_hidden": 0, "type_dim": 16, "gate_bias_init": -2.0}, "t": 2.2}
+{"event": "probe_schedule", "n": 45, "steps": [0, 1, 2, 4, 6, 8, 16, 32, 64, 128, 200, 256, 400, 512, 600, 800, 1000, 1024, 1200, 1400, 1600, 1800, 2000, 2048, 2200, 2400, 2600, 2800, 3000, 3200, 3400, 3600, 3800, 4000, 4096, 4200, 4400, 4600, 4800, 5000, 5200, 5400, 5600, 5800, 6000], "t": 2.2}
+{"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": 3.7}
+{"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": 11.3986, "lr": 1e-06, "router_trainable": true, "sec_per_step": 0.027, "probe_sec": 0.0268, "probe_extra_seqs": 0, "n_rows": 0, "extra_fwd_seqs": 0, "wall_s": 0.0, "overhead_frac_est": 0.0, "chunk_mismatch": 0, "n_loo": 0, "mean_delta_loo": NaN, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "read_hit_rate": {"surface": 0.0}, "mean_gate": {"surface": null}, "n_reads": 0, "n_cand": 5434, "read_suppressed_frac": 0.0, "dropout_reads_logged": 0, "inventory": {"size": 3161, "admitted": 3161, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 3161, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 1.7190762758255005, "n_util_obs": 0, "util_count_spearman": NaN}, "probe_overhead_sec_total": 0.027, "probe_overhead_seq_frac": 0.0, "util_loss": 0.0, "alpha": 0.03, "t": 5.3}
+{"event": "probe", "step": 1, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 2660.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.152847, "cps": 20.78125, "mean_p": 0.478104, "boundary_entropy": 0.685439, "bpb": 7.998223}, "tr": {"chunks": 5624.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.310667, "cps": 43.9375, "mean_p": 0.485999, "boundary_entropy": 0.686373, "bpb": 7.980111}, "fi": {"chunks": 4708.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.250279, "cps": 36.78125, "mean_p": 0.484401, "boundary_entropy": 0.686587, "bpb": 8.032287}}, "bpb": {"en": 7.99822, "tr": 7.98011, "fi": 8.03229}, "gini_cpc": 0.1474, "gini_cps": 0.152094, "bpic": 4.366687, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 2.0325313089385344, "fi": 1.637446255050736}, "inventory": {"size": 3161, "admitted": 3161, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 3161, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 1.7190762758255005, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 6.1}
+{"event": "probe", "step": 2, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 2658.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.152732, "cps": 20.765625, "mean_p": 0.478082, "boundary_entropy": 0.685441, "bpb": 7.933695}, "tr": {"chunks": 5589.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.308733, "cps": 43.664062, "mean_p": 0.485909, "boundary_entropy": 0.686371, "bpb": 7.931291}, "fi": {"chunks": 4671.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.248312, "cps": 36.492188, "mean_p": 0.484236, "boundary_entropy": 0.686576, "bpb": 7.979139}}, "bpb": {"en": 7.9337, "tr": 7.93129, "fi": 7.97914}, "gini_cpc": 0.146526, "gini_cps": 0.151262, "bpic": 4.391702, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 2.0214020498953253, "fi": 1.625800030792223}, "inventory": {"size": 5574, "admitted": 5574, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 5574, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 1.8426623344421387, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 7.9}
+{"event": "probe", "step": 4, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 2621.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.150606, "cps": 20.476562, "mean_p": 0.478058, "boundary_entropy": 0.685456, "bpb": 7.718957}, "tr": {"chunks": 5471.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.302215, "cps": 42.742188, "mean_p": 0.485639, "boundary_entropy": 0.686369, "bpb": 7.776379}, "fi": {"chunks": 4561.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.242465, "cps": 35.632812, "mean_p": 0.483757, "boundary_entropy": 0.686544, "bpb": 7.80156}}, "bpb": {"en": 7.71896, "tr": 7.77638, "fi": 7.80156}, "gini_cpc": 0.145368, "gini_cps": 0.150162, "bpic": 4.48368, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 2.006657546078427, "fi": 1.6099236788462903}, "inventory": {"size": 10497, "admitted": 10497, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 10497, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 1.9930455684661865, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 9.5}
+{"event": "probe", "step": 6, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 2591.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.148882, "cps": 20.242188, "mean_p": 0.478116, "boundary_entropy": 0.685492, "bpb": 7.436782}, "tr": {"chunks": 5283.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.29183, "cps": 41.273438, "mean_p": 0.485225, "boundary_entropy": 0.686368, "bpb": 7.577645}, "fi": {"chunks": 4331.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.230238, "cps": 33.835938, "mean_p": 0.483009, "boundary_entropy": 0.686496, "bpb": 7.566418}}, "bpb": {"en": 7.43678, "tr": 7.57765, "fi": 7.56642}, "gini_cpc": 0.142035, "gini_cps": 0.147044, "bpic": 4.648259, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.9601385340069035, "fi": 1.5464397612103629}, "inventory": {"size": 15126, "admitted": 15126, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 15126, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 2.0404601097106934, "n_util_obs": 1, "util_count_spearman": NaN}, "t": 11.1}
+{"event": "probe", "step": 8, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 2643.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.15187, "cps": 20.648438, "mean_p": 0.478884, "boundary_entropy": 0.685615, "bpb": 7.161997}, "tr": {"chunks": 5208.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.287687, "cps": 40.6875, "mean_p": 0.485133, "boundary_entropy": 0.686403, "bpb": 7.387748}, "fi": {"chunks": 4172.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.221785, "cps": 32.59375, "mean_p": 0.482573, "boundary_entropy": 0.686484, "bpb": 7.331005}}, "bpb": {"en": 7.162, "tr": 7.38775, "fi": 7.331}, "gini_cpc": 0.13691, "gini_cps": 0.142227, "bpic": 4.718623, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.8942939891877373, "fi": 1.460358132039414}, "inventory": {"size": 19631, "admitted": 19631, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 19631, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 2.055779218673706, "n_util_obs": 1, "util_count_spearman": NaN}, "t": 13.0}
+{"event": "probe", "step": 16, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 2600.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.1494, "cps": 20.3125, "mean_p": 0.481058, "boundary_entropy": 0.686028, "bpb": 6.512752}, "tr": {"chunks": 4391.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.242556, "cps": 34.304688, "mean_p": 0.482986, "boundary_entropy": 0.68644, "bpb": 6.895595}, "fi": {"chunks": 2955.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.157089, "cps": 23.085938, "mean_p": 0.478361, "boundary_entropy": 0.686277, "bpb": 6.690498}}, "bpb": {"en": 6.51275, "tr": 6.89559, "fi": 6.6905}, "gini_cpc": 0.113114, "gini_cps": 0.120048, "bpic": 5.704002, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.6235424855208869, "fi": 1.0514687600953614}, "inventory": {"size": 38465, "admitted": 38465, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 38465, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 2.3182373046875, "n_util_obs": 3, "util_count_spearman": 1.0}, "t": 16.0}
+{"event": "probe", "step": 32, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3612.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.20755, "cps": 28.21875, "mean_p": 0.49026, "boundary_entropy": 0.687159, "bpb": 6.160973}, "tr": {"chunks": 3907.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.215821, "cps": 30.523438, "mean_p": 0.486187, "boundary_entropy": 0.687283, "bpb": 6.576628}, "fi": {"chunks": 3190.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.169582, "cps": 24.921875, "mean_p": 0.485052, "boundary_entropy": 0.687297, "bpb": 6.341077}}, "bpb": {"en": 6.16097, "tr": 6.57663, "fi": 6.34108}, "gini_cpc": 0.051987, "gini_cps": 0.044635, "bpic": 5.2976, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0398465095357812, "fi": 0.8170623112121963}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 2.5149993896484375, "n_util_obs": 4, "util_count_spearman": 0.8}, "t": 21.0}
+{"event": "consolidate", "step": 50, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 26.9}
+{"event": "train", "step": 50, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 4.09784, "bpb": 5.91193, "chunk_loss": 0.030439, "F": 0.17267, "G": 0.48617, "chunks": 176.81, "bytes_per_chunk": 5.791, "boundary_entropy": 0.69163, "router_noise": 0.0, "grad_norm": 2.3625, "lr": 5.1e-05, "router_trainable": true, "sec_per_step": 0.437, "probe_sec": 0.0306, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01969, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.00428, "n_shadow": 0, "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": 0.09905, "trunc_abs_frac": 1.70732, "mean_delta_full": -0.05135, "read_hit_rate": {"surface": 0.5904}, "mean_gate": {"surface": 0.1193}, "n_reads": 3321, "n_cand": 5625, "read_suppressed_frac": 0.1087, "dropout_reads_logged": 3321, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 3.190582275390625, "n_util_obs": 8, "util_count_spearman": -0.047619047619047616}, "probe_overhead_sec_total": 3.055, "probe_overhead_seq_frac": 0.4902, "util_loss": 0.00291, "alpha": 0.03, "t": 27.2}
+{"event": "probe", "step": 64, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3766.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.216399, "cps": 29.421875, "mean_p": 0.484548, "boundary_entropy": 0.686176, "bpb": 5.426114}, "tr": {"chunks": 4529.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.25018, "cps": 35.382812, "mean_p": 0.481453, "boundary_entropy": 0.686085, "bpb": 5.919935}, "fi": {"chunks": 3908.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.207751, "cps": 30.53125, "mean_p": 0.482267, "boundary_entropy": 0.686503, "bpb": 5.652945}}, "bpb": {"en": 5.42611, "tr": 5.91993, "fi": 5.65294}, "gini_cpc": 0.041947, "gini_cps": 0.041684, "bpic": 4.649021, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.1561004594321589, "fi": 0.9600336951580263}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 3.5869140625, "n_util_obs": 11, "util_count_spearman": 0.19090909090909092}, "t": 31.8}
+{"event": "consolidate", "step": 100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 43.8}
+{"event": "train", "step": 100, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 3.20792, "bpb": 4.62805, "chunk_loss": 0.033103, "F": 0.21494, "G": 0.4628, "chunks": 220.09, "bytes_per_chunk": 4.653, "boundary_entropy": 0.68413, "router_noise": 0.0, "grad_norm": 1.7925, "lr": 0.000101, "router_trainable": true, "sec_per_step": 0.339, "probe_sec": 0.0696, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06287, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.00591, "n_shadow": 0, "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": -0.05349, "trunc_abs_frac": 0.61421, "mean_delta_full": -0.08521, "read_hit_rate": {"surface": 0.68378}, "mean_gate": {"surface": 0.11978}, "n_reads": 4794, "n_cand": 7011, "read_suppressed_frac": 0.1047, "dropout_reads_logged": 4794, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 4.3182220458984375, "n_util_obs": 18, "util_count_spearman": -0.04437564499484004}, "probe_overhead_sec_total": 6.177, "probe_overhead_seq_frac": 0.495, "util_loss": 0.000599, "alpha": 0.03, "t": 44.1}
+{"event": "probe", "step": 128, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3691.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.21209, "cps": 28.835938, "mean_p": 0.425763, "boundary_entropy": 0.621112, "bpb": 4.114108}, "tr": {"chunks": 4055.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.223996, "cps": 31.679688, "mean_p": 0.408528, "boundary_entropy": 0.641, "bpb": 4.365995}, "fi": {"chunks": 4137.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.219925, "cps": 32.320312, "mean_p": 0.413927, "boundary_entropy": 0.642791, "bpb": 4.329959}}, "bpb": {"en": 4.11411, "tr": 4.366, "fi": 4.32996}, "gini_cpc": 0.0121, "gini_cps": 0.025022, "bpic": 4.774215, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0561373026466916, "fi": 1.036940202315664}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 5.614501953125, "n_util_obs": 24, "util_count_spearman": 0.11652173913043479}, "t": 53.4}
+{"event": "consolidate", "step": 150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 60.6}
+{"event": "train", "step": 150, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.75742, "bpb": 3.97812, "chunk_loss": 0.031999, "F": 0.20609, "G": 0.39796, "chunks": 211.03, "bytes_per_chunk": 4.852, "boundary_entropy": 0.58145, "router_noise": 0.0, "grad_norm": 1.5859, "lr": 0.000151, "router_trainable": true, "sec_per_step": 0.336, "probe_sec": 0.0307, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02301, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.01292, "n_shadow": 0, "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": 0.43285, "trunc_abs_frac": 3.91859, "mean_delta_full": 0.03394, "read_hit_rate": {"surface": 0.75212}, "mean_gate": {"surface": 0.1209}, "n_reads": 5055, "n_cand": 6721, "read_suppressed_frac": 0.1007, "dropout_reads_logged": 5055, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 6.7596282958984375, "n_util_obs": 30, "util_count_spearman": -0.0028921023359288096}, "probe_overhead_sec_total": 9.546, "probe_overhead_seq_frac": 0.4967, "util_loss": 0.00035, "alpha": 0.03, "t": 60.9}
+{"event": "consolidate", "step": 200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 74.3}
+{"event": "probe", "step": 200, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3266.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.187669, "cps": 25.515625, "mean_p": 0.326312, "boundary_entropy": 0.468606, "bpb": 3.544976}, "tr": {"chunks": 4365.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.24112, "cps": 34.101562, "mean_p": 0.308297, "boundary_entropy": 0.488258, "bpb": 3.791648}, "fi": {"chunks": 3880.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.206262, "cps": 30.3125, "mean_p": 0.325178, "boundary_entropy": 0.518971, "bpb": 3.883648}}, "bpb": {"en": 3.54498, "tr": 3.79165, "fi": 3.88365}, "gini_cpc": 0.056112, "gini_cps": 0.063649, "bpic": 4.928503, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.2848180712131732, "fi": 1.0990761454313174}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 9.200775146484375, "n_util_obs": 37, "util_count_spearman": 0.2546230440967283}, "t": 75.1}
+{"event": "eval", "step": 200, "phase": 1, "bpb": {"en": 3.55462, "tr": 3.80037, "fi": 3.88342}, "bytes_per_chunk": {"en": 5.081, "tr": 4.534, "fi": 5.202}, "bpb_mean": 3.74614, "frozen_inventory": true, "t": 76.4}
+{"event": "train", "step": 200, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.48723, "bpb": 3.58831, "chunk_loss": 0.031056, "F": 0.19727, "G": 0.32465, "chunks": 202.0, "bytes_per_chunk": 5.069, "boundary_entropy": 0.47586, "router_noise": 0.0, "grad_norm": 1.752, "lr": 0.000201, "router_trainable": true, "sec_per_step": 0.327, "probe_sec": 0.0371, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02697, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.00342, "n_shadow": 0, "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": 2.20536, "trunc_abs_frac": 2.27746, "mean_delta_full": -0.02734, "read_hit_rate": {"surface": 0.787}, "mean_gate": {"surface": 0.12891}, "n_reads": 5062, "n_cand": 6432, "read_suppressed_frac": 0.1021, "dropout_reads_logged": 5062, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 9.25128173828125, "n_util_obs": 37, "util_count_spearman": 0.24466571834992887}, "probe_overhead_sec_total": 12.873, "probe_overhead_seq_frac": 0.4975, "util_loss": 0.000809, "alpha": 0.03, "t": 77.3}
+{"event": "consolidate", "step": 250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 90.1}
+{"event": "train", "step": 250, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.35705, "bpb": 3.4005, "chunk_loss": 0.030588, "F": 0.18671, "G": 0.29685, "chunks": 191.19, "bytes_per_chunk": 5.356, "boundary_entropy": 0.40687, "router_noise": 0.0, "grad_norm": 1.6876, "lr": 0.000251, "router_trainable": true, "sec_per_step": 0.263, "probe_sec": 0.0381, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02863, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.07912, "n_shadow": 0, "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": 0.75744, "trunc_abs_frac": 0.76081, "mean_delta_full": 0.15308, "read_hit_rate": {"surface": 0.77867}, "mean_gate": {"surface": 0.15342}, "n_reads": 4739, "n_cand": 6086, "read_suppressed_frac": 0.0977, "dropout_reads_logged": 4739, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 10.584335327148438, "n_util_obs": 44, "util_count_spearman": 0.14517265680056377}, "probe_overhead_sec_total": 15.647, "probe_overhead_seq_frac": 0.498, "util_loss": 0.002972, "alpha": 0.03, "t": 90.4}
+{"event": "probe", "step": 256, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3182.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.182842, "cps": 24.859375, "mean_p": 0.285905, "boundary_entropy": 0.38579, "bpb": 3.270601}, "tr": {"chunks": 3850.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.212672, "cps": 30.078125, "mean_p": 0.283734, "boundary_entropy": 0.410974, "bpb": 3.496841}, "fi": {"chunks": 3404.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.180958, "cps": 26.59375, "mean_p": 0.281598, "boundary_entropy": 0.438019, "bpb": 3.650202}}, "bpb": {"en": 3.2706, "tr": 3.49684, "fi": 3.6502}, "gini_cpc": 0.036676, "gini_cps": 0.042673, "bpic": 5.436182, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.163145709308558, "fi": 0.9896955393491933}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 10.661758422851562, "n_util_obs": 43, "util_count_spearman": 0.1587133796436122}, "t": 92.6}
+{"event": "consolidate", "step": 300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 105.9}
+{"event": "train", "step": 300, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.20036, "bpb": 3.17444, "chunk_loss": 0.030595, "F": 0.18881, "G": 0.29149, "chunks": 193.34, "bytes_per_chunk": 5.296, "boundary_entropy": 0.37226, "router_noise": 0.0, "grad_norm": 1.4942, "lr": 0.0003, "router_trainable": true, "sec_per_step": 0.315, "probe_sec": 0.0419, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0355, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.04609, "n_shadow": 0, "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": 0.1359, "trunc_abs_frac": 0.53818, "mean_delta_full": 0.14421, "read_hit_rate": {"surface": 0.77189}, "mean_gate": {"surface": 0.19681}, "n_reads": 4751, "n_cand": 6155, "read_suppressed_frac": 0.104, "dropout_reads_logged": 4751, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 12.041458129882812, "n_util_obs": 52, "util_count_spearman": 0.2556134209852301}, "probe_overhead_sec_total": 19.372, "probe_overhead_seq_frac": 0.4983, "util_loss": 0.004739, "alpha": 0.03, "t": 106.1}
+{"event": "consolidate", "step": 350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 118.4}
+{"event": "train", "step": 350, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.19694, "bpb": 3.16951, "chunk_loss": 0.030698, "F": 0.1936, "G": 0.28498, "chunks": 198.25, "bytes_per_chunk": 5.165, "boundary_entropy": 0.35805, "router_noise": 0.0, "grad_norm": 1.9155, "lr": 0.0002999, "router_trainable": true, "sec_per_step": 0.25, "probe_sec": 0.0313, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02548, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.00673, "n_shadow": 0, "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": 0.56727, "trunc_abs_frac": 0.90241, "mean_delta_full": -0.16704, "read_hit_rate": {"surface": 0.76141}, "mean_gate": {"surface": 0.21189}, "n_reads": 4806, "n_cand": 6312, "read_suppressed_frac": 0.0961, "dropout_reads_logged": 4806, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 13.792999267578125, "n_util_obs": 58, "util_count_spearman": 0.1698914146851641}, "probe_overhead_sec_total": 22.62, "probe_overhead_seq_frac": 0.4986, "util_loss": 0.005036, "alpha": 0.03, "t": 118.6}
+{"event": "consolidate", "step": 400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 131.0}
+{"event": "probe", "step": 400, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3385.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.194507, "cps": 26.445312, "mean_p": 0.264621, "boundary_entropy": 0.291154, "bpb": 2.863267}, "tr": {"chunks": 3422.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.189029, "cps": 26.734375, "mean_p": 0.293623, "boundary_entropy": 0.395091, "bpb": 3.129498}, "fi": {"chunks": 3428.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.182234, "cps": 26.78125, "mean_p": 0.289541, "boundary_entropy": 0.408446, "bpb": 3.351644}}, "bpb": {"en": 2.86327, "tr": 3.1295, "fi": 3.35164}, "gini_cpc": 0.014462, "gini_cps": 0.002801, "bpic": 5.542941, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9718402925129476, "fi": 0.936902455091057}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 15.824691772460938, "n_util_obs": 65, "util_count_spearman": 0.06888111888111888}, "t": 131.7}
+{"event": "eval", "step": 400, "phase": 1, "bpb": {"en": 2.87518, "tr": 3.13684, "fi": 3.34454}, "bytes_per_chunk": {"en": 4.905, "tr": 5.835, "fi": 5.696}, "bpb_mean": 3.11885, "frozen_inventory": true, "t": 132.9}
+{"event": "train", "step": 400, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.02346, "bpb": 2.91924, "chunk_loss": 0.030655, "F": 0.19629, "G": 0.2704, "chunks": 201.0, "bytes_per_chunk": 5.095, "boundary_entropy": 0.29829, "router_noise": 0.0, "grad_norm": 1.1488, "lr": 0.0002998, "router_trainable": true, "sec_per_step": 0.304, "probe_sec": 0.0299, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0239, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.001, "n_shadow": 0, "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": 3.12185, "trunc_abs_frac": 2.5058, "mean_delta_full": -0.02783, "read_hit_rate": {"surface": 0.80281}, "mean_gate": {"surface": 0.24153}, "n_reads": 5138, "n_cand": 6400, "read_suppressed_frac": 0.0967, "dropout_reads_logged": 5138, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 15.879241943359375, "n_util_obs": 65, "util_count_spearman": 0.06547202797202797}, "probe_overhead_sec_total": 25.362, "probe_overhead_seq_frac": 0.4988, "util_loss": 0.007006, "alpha": 0.03, "t": 133.9}
+{"event": "consolidate", "step": 450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 147.5}
+{"event": "train", "step": 450, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.05196, "bpb": 2.96035, "chunk_loss": 0.030755, "F": 0.20139, "G": 0.26291, "chunks": 206.22, "bytes_per_chunk": 4.966, "boundary_entropy": 0.27994, "router_noise": 0.0, "grad_norm": 0.998, "lr": 0.0002995, "router_trainable": true, "sec_per_step": 0.279, "probe_sec": 0.0319, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02215, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.1663, "n_shadow": 0, "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": 0.92021, "trunc_abs_frac": 1.04581, "mean_delta_full": 0.43732, "read_hit_rate": {"surface": 0.79504}, "mean_gate": {"surface": 0.25937}, "n_reads": 5221, "n_cand": 6567, "read_suppressed_frac": 0.1009, "dropout_reads_logged": 5221, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 17.858993530273438, "n_util_obs": 72, "util_count_spearman": 0.14312817544536627}, "probe_overhead_sec_total": 28.884, "probe_overhead_seq_frac": 0.4989, "util_loss": 0.007213, "alpha": 0.03, "t": 147.8}
+{"event": "consolidate", "step": 500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 160.2}
+{"event": "train", "step": 500, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.01952, "bpb": 2.91355, "chunk_loss": 0.030617, "F": 0.19727, "G": 0.26183, "chunks": 202.0, "bytes_per_chunk": 5.069, "boundary_entropy": 0.28181, "router_noise": 0.0, "grad_norm": 1.2252, "lr": 0.0002992, "router_trainable": true, "sec_per_step": 0.254, "probe_sec": 0.0885, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08137, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.04313, "n_shadow": 0, "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": 1.02694, "trunc_abs_frac": 0.98358, "mean_delta_full": 0.0623, "read_hit_rate": {"surface": 0.77254}, "mean_gate": {"surface": 0.27153}, "n_reads": 4969, "n_cand": 6432, "read_suppressed_frac": 0.1008, "dropout_reads_logged": 4969, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 19.7735595703125, "n_util_obs": 81, "util_count_spearman": 0.0248193315266486}, "probe_overhead_sec_total": 32.03, "probe_overhead_seq_frac": 0.499, "util_loss": 0.016845, "alpha": 0.03, "t": 160.6}
+{"event": "probe", "step": 512, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3579.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.205654, "cps": 27.960938, "mean_p": 0.255039, "boundary_entropy": 0.250461, "bpb": 2.668275}, "tr": {"chunks": 3290.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.181738, "cps": 25.703125, "mean_p": 0.284875, "boundary_entropy": 0.377633, "bpb": 2.970651}, "fi": {"chunks": 3486.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.185317, "cps": 27.234375, "mean_p": 0.26567, "boundary_entropy": 0.350546, "bpb": 3.231011}}, "bpb": {"en": 2.66828, "tr": 2.97065, "fi": 3.23101}, "gini_cpc": 0.02784, "gini_cps": 0.018606, "bpic": 5.478706, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.8837059280648839, "fi": 0.9011102321353146}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 20.139144897460938, "n_util_obs": 80, "util_count_spearman": 0.043248945147679324}, "t": 164.4}
+{"event": "consolidate", "step": 550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 175.5}
+{"event": "train", "step": 550, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.94229, "bpb": 2.80214, "chunk_loss": 0.031014, "F": 0.21353, "G": 0.26232, "chunks": 218.66, "bytes_per_chunk": 4.683, "boundary_entropy": 0.24708, "router_noise": 0.0, "grad_norm": 1.3394, "lr": 0.0002987, "router_trainable": true, "sec_per_step": 0.306, "probe_sec": 0.0381, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03053, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.05899, "n_shadow": 0, "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": 0.32841, "trunc_abs_frac": 1.12757, "mean_delta_full": 0.09137, "read_hit_rate": {"surface": 0.81192}, "mean_gate": {"surface": 0.29859}, "n_reads": 5655, "n_cand": 6965, "read_suppressed_frac": 0.1052, "dropout_reads_logged": 5655, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 21.669052124023438, "n_util_obs": 84, "util_count_spearman": 0.058884276602207145}, "probe_overhead_sec_total": 35.521, "probe_overhead_seq_frac": 0.4991, "util_loss": 0.014888, "alpha": 0.03, "t": 175.9}
+{"event": "consolidate", "step": 600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 190.5}
+{"event": "probe", "step": 600, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3608.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.207321, "cps": 28.1875, "mean_p": 0.247221, "boundary_entropy": 0.219134, "bpb": 2.561626}, "tr": {"chunks": 3506.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.19367, "cps": 27.390625, "mean_p": 0.299062, "boundary_entropy": 0.384138, "bpb": 2.888931}, "fi": {"chunks": 3624.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.192653, "cps": 28.3125, "mean_p": 0.267466, "boundary_entropy": 0.335567, "bpb": 3.126252}}, "bpb": {"en": 2.56163, "tr": 2.88893, "fi": 3.12625}, "gini_cpc": 0.016472, "gini_cps": 0.007326, "bpic": 5.283293, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9341550193258508, "fi": 0.9292528396309137}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 23.76239013671875, "n_util_obs": 93, "util_count_spearman": -0.007288678675256891}, "t": 191.3}
+{"event": "eval", "step": 600, "phase": 1, "bpb": {"en": 2.57636, "tr": 2.8993, "fi": 3.1261}, "bytes_per_chunk": {"en": 4.621, "tr": 5.557, "fi": 5.447}, "bpb_mean": 2.86725, "frozen_inventory": true, "t": 192.4}
+{"event": "train", "step": 600, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.91526, "bpb": 2.76314, "chunk_loss": 0.030808, "F": 0.20557, "G": 0.26364, "chunks": 210.5, "bytes_per_chunk": 4.865, "boundary_entropy": 0.25635, "router_noise": 0.0, "grad_norm": 1.1325, "lr": 0.0002982, "router_trainable": true, "sec_per_step": 0.355, "probe_sec": 0.0635, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04184, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.21752, "n_shadow": 0, "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": 1.09763, "trunc_abs_frac": 1.2165, "mean_delta_full": 0.40387, "read_hit_rate": {"surface": 0.79564}, "mean_gate": {"surface": 0.30414}, "n_reads": 5334, "n_cand": 6704, "read_suppressed_frac": 0.0971, "dropout_reads_logged": 5334, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 23.81195068359375, "n_util_obs": 93, "util_count_spearman": -0.004297096475784071}, "probe_overhead_sec_total": 38.435, "probe_overhead_seq_frac": 0.4992, "util_loss": 0.021809, "alpha": 0.03, "t": 193.6}
+{"event": "consolidate", "step": 650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 213.1}
+{"event": "train", "step": 650, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.84984, "bpb": 2.66876, "chunk_loss": 0.03078, "F": 0.2067, "G": 0.26051, "chunks": 211.66, "bytes_per_chunk": 4.838, "boundary_entropy": 0.24469, "router_noise": 0.0, "grad_norm": 1.2927, "lr": 0.0002975, "router_trainable": true, "sec_per_step": 0.409, "probe_sec": 0.0641, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03644, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.20188, "n_shadow": 0, "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": -0.9527, "trunc_abs_frac": 1.1803, "mean_delta_full": -0.14323, "read_hit_rate": {"surface": 0.80359}, "mean_gate": {"surface": 0.31532}, "n_reads": 5417, "n_cand": 6741, "read_suppressed_frac": 0.1015, "dropout_reads_logged": 5417, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 26.046279907226562, "n_util_obs": 98, "util_count_spearman": 0.041422136647994795}, "probe_overhead_sec_total": 41.701, "probe_overhead_seq_frac": 0.4992, "util_loss": 0.021188, "alpha": 0.03, "t": 214.1}
+{"event": "consolidate", "step": 700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 227.9}
+{"event": "train", "step": 700, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.86406, "bpb": 2.68927, "chunk_loss": 0.030544, "F": 0.19495, "G": 0.26699, "chunks": 199.62, "bytes_per_chunk": 5.13, "boundary_entropy": 0.27334, "router_noise": 0.0, "grad_norm": 1.045, "lr": 0.0002967, "router_trainable": true, "sec_per_step": 0.282, "probe_sec": 0.0434, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03165, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.00218, "n_shadow": 0, "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": 1.1077, "trunc_abs_frac": 1.69922, "mean_delta_full": 0.20402, "read_hit_rate": {"surface": 0.75205}, "mean_gate": {"surface": 0.31708}, "n_reads": 4780, "n_cand": 6356, "read_suppressed_frac": 0.1067, "dropout_reads_logged": 4780, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 27.98486328125, "n_util_obs": 103, "util_count_spearman": 0.04571576054294058}, "probe_overhead_sec_total": 45.257, "probe_overhead_seq_frac": 0.4993, "util_loss": 0.024337, "alpha": 0.03, "t": 228.2}
+{"event": "consolidate", "step": 750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 243.7}
+{"event": "train", "step": 750, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.77151, "bpb": 2.55575, "chunk_loss": 0.030627, "F": 0.20352, "G": 0.25, "chunks": 208.41, "bytes_per_chunk": 4.913, "boundary_entropy": 0.21818, "router_noise": 0.0, "grad_norm": 1.0616, "lr": 0.0002959, "router_trainable": true, "sec_per_step": 0.317, "probe_sec": 0.0775, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06854, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.33876, "n_shadow": 0, "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": -0.0156, "trunc_abs_frac": 3.16515, "mean_delta_full": 0.16825, "read_hit_rate": {"surface": 0.7984}, "mean_gate": {"surface": 0.33648}, "n_reads": 5299, "n_cand": 6637, "read_suppressed_frac": 0.1053, "dropout_reads_logged": 5299, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 29.981033325195312, "n_util_obs": 110, "util_count_spearman": 0.0830649033315983}, "probe_overhead_sec_total": 48.436, "probe_overhead_seq_frac": 0.4993, "util_loss": 0.025722, "alpha": 0.03, "t": 244.0}
+{"event": "consolidate", "step": 800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 260.3}
+{"event": "probe", "step": 800, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3642.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.209274, "cps": 28.453125, "mean_p": 0.234842, "boundary_entropy": 0.171629, "bpb": 2.387699}, "tr": {"chunks": 3432.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.189582, "cps": 26.8125, "mean_p": 0.27728, "boundary_entropy": 0.342996, "bpb": 2.703795}, "fi": {"chunks": 3513.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.186752, "cps": 27.445312, "mean_p": 0.245355, "boundary_entropy": 0.27332, "bpb": 2.97461}}, "bpb": {"en": 2.3877, "tr": 2.7038, "fi": 2.97461}, "gini_cpc": 0.025639, "gini_cps": 0.013224, "bpic": 5.358647, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.905901349235261, "fi": 0.8923812673313145}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 31.749313354492188, "n_util_obs": 121, "util_count_spearman": 0.03996070993090367}, "t": 261.0}
+{"event": "eval", "step": 800, "phase": 1, "bpb": {"en": 2.4058, "tr": 2.71375, "fi": 2.97682}, "bytes_per_chunk": {"en": 4.575, "tr": 5.713, "fi": 5.529}, "bpb_mean": 2.69879, "frozen_inventory": true, "t": 262.2}
+{"event": "train", "step": 800, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.78003, "bpb": 2.56805, "chunk_loss": 0.030652, "F": 0.20276, "G": 0.25111, "chunks": 207.62, "bytes_per_chunk": 4.932, "boundary_entropy": 0.22014, "router_noise": 0.0, "grad_norm": 1.0869, "lr": 0.0002949, "router_trainable": true, "sec_per_step": 0.409, "probe_sec": 0.0444, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02507, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.0546, "n_shadow": 0, "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": 1.23462, "trunc_abs_frac": 1.72515, "mean_delta_full": 0.27859, "read_hit_rate": {"surface": 0.7866}, "mean_gate": {"surface": 0.34068}, "n_reads": 5201, "n_cand": 6612, "read_suppressed_frac": 0.1042, "dropout_reads_logged": 5201, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 31.758438110351562, "n_util_obs": 121, "util_count_spearman": 0.03820965690571952}, "probe_overhead_sec_total": 51.416, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.030533, "alpha": 0.03, "t": 264.4}
+{"event": "consolidate", "step": 850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 285.1}
+{"event": "train", "step": 850, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.70882, "bpb": 2.4653, "chunk_loss": 0.030694, "F": 0.20996, "G": 0.23746, "chunks": 215.0, "bytes_per_chunk": 4.763, "boundary_entropy": 0.17167, "router_noise": 0.0, "grad_norm": 1.1603, "lr": 0.0002938, "router_trainable": true, "sec_per_step": 0.42, "probe_sec": 0.1514, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.14151, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.28703, "n_shadow": 0, "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": {"surface": 0.83455}, "mean_gate": {"surface": 0.37595}, "n_reads": 5715, "n_cand": 6848, "read_suppressed_frac": 0.0955, "dropout_reads_logged": 5715, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 33.74562072753906, "n_util_obs": 131, "util_count_spearman": -0.01697541367563258}, "probe_overhead_sec_total": 54.637, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.026938, "alpha": 0.03, "t": 285.4}
+{"event": "consolidate", "step": 900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 306.5}
+{"event": "train", "step": 900, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.68981, "bpb": 2.43788, "chunk_loss": 0.030711, "F": 0.2099, "G": 0.24495, "chunks": 214.94, "bytes_per_chunk": 4.764, "boundary_entropy": 0.18886, "router_noise": 0.0, "grad_norm": 1.0367, "lr": 0.0002927, "router_trainable": true, "sec_per_step": 0.426, "probe_sec": 0.0633, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05701, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.15635, "n_shadow": 0, "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": -0.45007, "trunc_abs_frac": 0.69801, "mean_delta_full": -0.11837, "read_hit_rate": {"surface": 0.82428}, "mean_gate": {"surface": 0.36363}, "n_reads": 5643, "n_cand": 6846, "read_suppressed_frac": 0.1024, "dropout_reads_logged": 5643, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 35.85246276855469, "n_util_obs": 136, "util_count_spearman": -0.001011403334053644}, "probe_overhead_sec_total": 57.97, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.028595, "alpha": 0.03, "t": 306.7}
+{"event": "consolidate", "step": 950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 320.7}
+{"event": "train", "step": 950, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.69459, "bpb": 2.44477, "chunk_loss": 0.030717, "F": 0.21121, "G": 0.24128, "chunks": 216.28, "bytes_per_chunk": 4.735, "boundary_entropy": 0.17155, "router_noise": 0.0, "grad_norm": 1.0146, "lr": 0.0002914, "router_trainable": true, "sec_per_step": 0.283, "probe_sec": 0.0238, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01758, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.30765, "n_shadow": 0, "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": 0.98827, "trunc_abs_frac": 1.02408, "mean_delta_full": 0.44924, "read_hit_rate": {"surface": 0.83786}, "mean_gate": {"surface": 0.37963}, "n_reads": 5772, "n_cand": 6889, "read_suppressed_frac": 0.1036, "dropout_reads_logged": 5772, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 38.08168029785156, "n_util_obs": 137, "util_count_spearman": 0.01912326983413411}, "probe_overhead_sec_total": 60.905, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.027647, "alpha": 0.03, "t": 320.9}
+{"event": "consolidate", "step": 1000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 335.0}
+{"event": "probe", "step": 1000, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3644.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.209389, "cps": 28.46875, "mean_p": 0.228221, "boundary_entropy": 0.149353, "bpb": 2.294692}, "tr": {"chunks": 3475.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.191957, "cps": 27.148438, "mean_p": 0.259303, "boundary_entropy": 0.302782, "bpb": 2.607632}, "fi": {"chunks": 3576.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.190102, "cps": 27.9375, "mean_p": 0.239588, "boundary_entropy": 0.248789, "bpb": 2.840754}}, "bpb": {"en": 2.29469, "tr": 2.60763, "fi": 2.84075}, "gini_cpc": 0.021741, "gini_cps": 0.010535, "bpic": 5.304535, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9167480807015205, "fi": 0.9078861242700731}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 40.23472595214844, "n_util_obs": 141, "util_count_spearman": -0.0012457725073847339}, "t": 335.7}
+{"event": "eval", "step": 1000, "phase": 1, "bpb": {"en": 2.30958, "tr": 2.61587, "fi": 2.84423}, "bytes_per_chunk": {"en": 4.568, "tr": 5.669, "fi": 5.442}, "bpb_mean": 2.58989, "frozen_inventory": true, "t": 336.7}
+{"event": "train", "step": 1000, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.70213, "bpb": 2.45565, "chunk_loss": 0.030686, "F": 0.20987, "G": 0.24065, "chunks": 214.91, "bytes_per_chunk": 4.765, "boundary_entropy": 0.17609, "router_noise": 0.0, "grad_norm": 0.9069, "lr": 0.0002901, "router_trainable": true, "sec_per_step": 0.343, "probe_sec": 0.0543, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04644, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.07283, "n_shadow": 0, "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": 0.927, "trunc_abs_frac": 0.94635, "mean_delta_full": 0.82609, "read_hit_rate": {"surface": 0.80687}, "mean_gate": {"surface": 0.37757}, "n_reads": 5523, "n_cand": 6845, "read_suppressed_frac": 0.1003, "dropout_reads_logged": 5523, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 40.274261474609375, "n_util_obs": 141, "util_count_spearman": -0.003090885739971745}, "probe_overhead_sec_total": 63.989, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.034826, "alpha": 0.03, "t": 338.0}
+{"event": "probe", "step": 1024, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3644.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.209389, "cps": 28.46875, "mean_p": 0.228396, "boundary_entropy": 0.147233, "bpb": 2.285235}, "tr": {"chunks": 3377.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.186544, "cps": 26.382812, "mean_p": 0.259109, "boundary_entropy": 0.310062, "bpb": 2.578023}, "fi": {"chunks": 3596.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.191165, "cps": 28.09375, "mean_p": 0.236854, "boundary_entropy": 0.245896, "bpb": 2.800156}}, "bpb": {"en": 2.28524, "tr": 2.57802, "fi": 2.80016}, "gini_cpc": 0.025942, "gini_cps": 0.016766, "bpic": 5.343506, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.8908944657637511, "fi": 0.9129637871574897}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 41.09297180175781, "n_util_obs": 145, "util_count_spearman": 0.027755471579278854}, "t": 343.9}
+{"event": "consolidate", "step": 1050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 351.2}
+{"event": "train", "step": 1050, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.72442, "bpb": 2.48781, "chunk_loss": 0.030558, "F": 0.20398, "G": 0.23884, "chunks": 208.88, "bytes_per_chunk": 4.902, "boundary_entropy": 0.18374, "router_noise": 0.0, "grad_norm": 0.9678, "lr": 0.0002886, "router_trainable": true, "sec_per_step": 0.268, "probe_sec": 0.0477, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03958, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.21763, "n_shadow": 0, "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": -0.05961, "trunc_abs_frac": 1.21269, "mean_delta_full": 0.17041, "read_hit_rate": {"surface": 0.78879}, "mean_gate": {"surface": 0.36893}, "n_reads": 5247, "n_cand": 6652, "read_suppressed_frac": 0.0962, "dropout_reads_logged": 5247, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 42.27384948730469, "n_util_obs": 144, "util_count_spearman": 0.05152318945422394}, "probe_overhead_sec_total": 66.751, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.047552, "alpha": 0.03, "t": 351.4}
+{"event": "consolidate", "step": 1100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 365.3}
+{"event": "train", "step": 1100, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.69977, "bpb": 2.45225, "chunk_loss": 0.03074, "F": 0.21384, "G": 0.23704, "chunks": 218.97, "bytes_per_chunk": 4.676, "boundary_entropy": 0.15808, "router_noise": 0.0, "grad_norm": 0.9257, "lr": 0.0002871, "router_trainable": true, "sec_per_step": 0.283, "probe_sec": 0.0687, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05887, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.09909, "n_shadow": 0, "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": 0.95347, "trunc_abs_frac": 0.81766, "mean_delta_full": -0.62345, "read_hit_rate": {"surface": 0.81534}, "mean_gate": {"surface": 0.38043}, "n_reads": 5687, "n_cand": 6975, "read_suppressed_frac": 0.0946, "dropout_reads_logged": 5687, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 44.23329162597656, "n_util_obs": 149, "util_count_spearman": 0.08380570253179492}, "probe_overhead_sec_total": 69.658, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.053229, "alpha": 0.03, "t": 365.6}
+{"event": "consolidate", "step": 1150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 377.9}
+{"event": "train", "step": 1150, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.71452, "bpb": 2.47353, "chunk_loss": 0.030749, "F": 0.20935, "G": 0.24705, "chunks": 214.38, "bytes_per_chunk": 4.777, "boundary_entropy": 0.19969, "router_noise": 0.0, "grad_norm": 0.8908, "lr": 0.0002855, "router_trainable": true, "sec_per_step": 0.253, "probe_sec": 0.1164, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10684, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.33408, "n_shadow": 0, "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": 62.02975, "trunc_abs_frac": 0.69318, "mean_delta_full": -0.00761, "read_hit_rate": {"surface": 0.78427}, "mean_gate": {"surface": 0.37659}, "n_reads": 5355, "n_cand": 6828, "read_suppressed_frac": 0.0984, "dropout_reads_logged": 5355, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 46.18516540527344, "n_util_obs": 152, "util_count_spearman": 0.06254319546950686}, "probe_overhead_sec_total": 72.777, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.036343, "alpha": 0.03, "t": 378.2}
+{"event": "consolidate", "step": 1200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 391.8}
+{"event": "probe", "step": 1200, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3614.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.207665, "cps": 28.234375, "mean_p": 0.224232, "boundary_entropy": 0.140462, "bpb": 2.23469}, "tr": {"chunks": 3405.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.18809, "cps": 26.601562, "mean_p": 0.248334, "boundary_entropy": 0.286417, "bpb": 2.519088}, "fi": {"chunks": 3608.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.191803, "cps": 28.1875, "mean_p": 0.236779, "boundary_entropy": 0.238108, "bpb": 2.726742}}, "bpb": {"en": 2.23469, "tr": 2.51909, "fi": 2.72674}, "gini_cpc": 0.022211, "gini_cps": 0.013111, "bpic": 5.338477, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9057378914684254, "fi": 0.923614234238777}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 48.301361083984375, "n_util_obs": 158, "util_count_spearman": 0.03063622769293491}, "t": 392.5}
+{"event": "eval", "step": 1200, "phase": 1, "bpb": {"en": 2.24893, "tr": 2.52854, "fi": 2.73342}, "bytes_per_chunk": {"en": 4.603, "tr": 5.75, "fi": 5.34}, "bpb_mean": 2.50363, "frozen_inventory": true, "t": 393.8}
+{"event": "train", "step": 1200, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.61709, "bpb": 2.33297, "chunk_loss": 0.030609, "F": 0.20654, "G": 0.2376, "chunks": 211.5, "bytes_per_chunk": 4.842, "boundary_entropy": 0.17282, "router_noise": 0.0, "grad_norm": 0.8778, "lr": 0.0002837, "router_trainable": true, "sec_per_step": 0.333, "probe_sec": 0.0467, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03584, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.05778, "n_shadow": 0, "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": 0.31732, "trunc_abs_frac": 1.67906, "mean_delta_full": -0.3479, "read_hit_rate": {"surface": 0.81354}, "mean_gate": {"surface": 0.38238}, "n_reads": 5480, "n_cand": 6736, "read_suppressed_frac": 0.1038, "dropout_reads_logged": 5480, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 48.35418701171875, "n_util_obs": 158, "util_count_spearman": 0.030840073688806268}, "probe_overhead_sec_total": 75.97, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.032196, "alpha": 0.03, "t": 394.8}
+{"event": "consolidate", "step": 1250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 408.1}
+{"event": "train", "step": 1250, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.65061, "bpb": 2.38133, "chunk_loss": 0.030654, "F": 0.20761, "G": 0.23922, "chunks": 212.59, "bytes_per_chunk": 4.817, "boundary_entropy": 0.17499, "router_noise": 0.0, "grad_norm": 0.8901, "lr": 0.0002819, "router_trainable": true, "sec_per_step": 0.27, "probe_sec": 0.0324, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02656, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.26651, "n_shadow": 0, "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": 0.70138, "trunc_abs_frac": 0.53967, "mean_delta_full": 0.59018, "read_hit_rate": {"surface": 0.79353}, "mean_gate": {"surface": 0.38841}, "n_reads": 5373, "n_cand": 6771, "read_suppressed_frac": 0.1046, "dropout_reads_logged": 5373, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 50.33216857910156, "n_util_obs": 158, "util_count_spearman": 0.039949251474460684}, "probe_overhead_sec_total": 79.701, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.032809, "alpha": 0.03, "t": 408.3}
+{"event": "consolidate", "step": 1300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 420.9}
+{"event": "train", "step": 1300, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.57147, "bpb": 2.26715, "chunk_loss": 0.030688, "F": 0.2121, "G": 0.23557, "chunks": 217.19, "bytes_per_chunk": 4.715, "boundary_entropy": 0.14596, "router_noise": 0.0, "grad_norm": 0.7732, "lr": 0.00028, "router_trainable": true, "sec_per_step": 0.258, "probe_sec": 0.0357, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02934, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.09225, "n_shadow": 0, "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": 1.99179, "trunc_abs_frac": 2.52405, "mean_delta_full": 0.11275, "read_hit_rate": {"surface": 0.82813}, "mean_gate": {"surface": 0.4067}, "n_reads": 5729, "n_cand": 6918, "read_suppressed_frac": 0.1025, "dropout_reads_logged": 5729, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 52.39463806152344, "n_util_obs": 159, "util_count_spearman": 0.017843921662288036}, "probe_overhead_sec_total": 81.973, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.032041, "alpha": 0.03, "t": 421.2}
+{"event": "consolidate", "step": 1350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 433.2}
+{"event": "train", "step": 1350, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.60151, "bpb": 2.31049, "chunk_loss": 0.030624, "F": 0.21198, "G": 0.22752, "chunks": 217.06, "bytes_per_chunk": 4.718, "boundary_entropy": 0.12842, "router_noise": 0.0, "grad_norm": 0.9406, "lr": 0.000278, "router_trainable": true, "sec_per_step": 0.246, "probe_sec": 0.0439, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03699, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.23056, "n_shadow": 0, "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": 0.96018, "trunc_abs_frac": 0.89635, "mean_delta_full": 0.44949, "read_hit_rate": {"surface": 0.83714}, "mean_gate": {"surface": 0.41219}, "n_reads": 5788, "n_cand": 6914, "read_suppressed_frac": 0.1018, "dropout_reads_logged": 5788, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 54.43487548828125, "n_util_obs": 161, "util_count_spearman": 0.07159822866344605}, "probe_overhead_sec_total": 84.646, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.032181, "alpha": 0.03, "t": 433.5}
+{"event": "consolidate", "step": 1400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 446.3}
+{"event": "probe", "step": 1400, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3621.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.208068, "cps": 28.289062, "mean_p": 0.220319, "boundary_entropy": 0.121317, "bpb": 2.178838}, "tr": {"chunks": 3504.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.193559, "cps": 27.375, "mean_p": 0.240415, "boundary_entropy": 0.25713, "bpb": 2.457057}, "fi": {"chunks": 3625.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.192706, "cps": 28.320312, "mean_p": 0.224392, "boundary_entropy": 0.201097, "bpb": 2.660544}}, "bpb": {"en": 2.17884, "tr": 2.45706, "fi": 2.66054}, "gini_cpc": 0.017231, "gini_cps": 0.007504, "bpic": 5.277395, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9302702692559985, "fi": 0.9261721611616747}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 56.545440673828125, "n_util_obs": 163, "util_count_spearman": 0.07396877649758092}, "t": 447.1}
+{"event": "eval", "step": 1400, "phase": 1, "bpb": {"en": 2.19385, "tr": 2.46605, "fi": 2.66637}, "bytes_per_chunk": {"en": 4.588, "tr": 5.567, "fi": 5.294}, "bpb_mean": 2.44209, "frozen_inventory": true, "t": 448.3}
+{"event": "train", "step": 1400, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.56975, "bpb": 2.26467, "chunk_loss": 0.030549, "F": 0.20831, "G": 0.22785, "chunks": 213.31, "bytes_per_chunk": 4.8, "boundary_entropy": 0.14291, "router_noise": 0.0, "grad_norm": 0.8669, "lr": 0.0002759, "router_trainable": true, "sec_per_step": 0.318, "probe_sec": 0.0362, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02963, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.09809, "n_shadow": 0, "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": 0.31511, "trunc_abs_frac": 1.20066, "mean_delta_full": -0.03445, "read_hit_rate": {"surface": 0.82205}, "mean_gate": {"surface": 0.40801}, "n_reads": 5585, "n_cand": 6794, "read_suppressed_frac": 0.0933, "dropout_reads_logged": 5585, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 56.608123779296875, "n_util_obs": 163, "util_count_spearman": 0.07443430262523484}, "probe_overhead_sec_total": 87.221, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.034525, "alpha": 0.03, "t": 449.4}
+{"event": "consolidate", "step": 1450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 462.8}
+{"event": "train", "step": 1450, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.54673, "bpb": 2.23146, "chunk_loss": 0.030682, "F": 0.21118, "G": 0.23526, "chunks": 216.25, "bytes_per_chunk": 4.735, "boundary_entropy": 0.15424, "router_noise": 0.0, "grad_norm": 0.8629, "lr": 0.0002738, "router_trainable": true, "sec_per_step": 0.272, "probe_sec": 0.0357, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03067, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.08442, "n_shadow": 0, "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": 0.80708, "trunc_abs_frac": 1.19369, "mean_delta_full": 0.23844, "read_hit_rate": {"surface": 0.82927}, "mean_gate": {"surface": 0.40489}, "n_reads": 5712, "n_cand": 6888, "read_suppressed_frac": 0.0994, "dropout_reads_logged": 5712, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 58.366485595703125, "n_util_obs": 163, "util_count_spearman": 0.01906163233411475}, "probe_overhead_sec_total": 91.107, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.040117, "alpha": 0.03, "t": 463.0}
+{"event": "consolidate", "step": 1500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 475.7}
+{"event": "train", "step": 1500, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.53951, "bpb": 2.22104, "chunk_loss": 0.030549, "F": 0.2074, "G": 0.22635, "chunks": 212.38, "bytes_per_chunk": 4.822, "boundary_entropy": 0.14143, "router_noise": 0.0, "grad_norm": 0.9476, "lr": 0.0002715, "router_trainable": true, "sec_per_step": 0.261, "probe_sec": 0.0294, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02292, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.18075, "n_shadow": 0, "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": -0.73385, "trunc_abs_frac": 0.73385, "mean_delta_full": -0.57678, "read_hit_rate": {"surface": 0.82436}, "mean_gate": {"surface": 0.40445}, "n_reads": 5576, "n_cand": 6764, "read_suppressed_frac": 0.0898, "dropout_reads_logged": 5576, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 60.375732421875, "n_util_obs": 161, "util_count_spearman": 0.007876121463077985}, "probe_overhead_sec_total": 94.166, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.056139, "alpha": 0.03, "t": 476.0}
+{"event": "consolidate", "step": 1550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 488.1}
+{"event": "train", "step": 1550, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.5282, "bpb": 2.20473, "chunk_loss": 0.030612, "F": 0.21036, "G": 0.22801, "chunks": 215.41, "bytes_per_chunk": 4.754, "boundary_entropy": 0.13532, "router_noise": 0.0, "grad_norm": 0.7511, "lr": 0.0002692, "router_trainable": true, "sec_per_step": 0.247, "probe_sec": 0.0611, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05134, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.22859, "n_shadow": 0, "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": -0.31595, "trunc_abs_frac": 2.73424, "mean_delta_full": 0.11784, "read_hit_rate": {"surface": 0.83049}, "mean_gate": {"surface": 0.41615}, "n_reads": 5698, "n_cand": 6861, "read_suppressed_frac": 0.1021, "dropout_reads_logged": 5698, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 62.428131103515625, "n_util_obs": 162, "util_count_spearman": -0.06127493700257185}, "probe_overhead_sec_total": 97.309, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.041903, "alpha": 0.03, "t": 488.4}
+{"event": "consolidate", "step": 1600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 501.3}
+{"event": "probe", "step": 1600, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3609.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.207378, "cps": 28.195312, "mean_p": 0.215072, "boundary_entropy": 0.109798, "bpb": 2.134289}, "tr": {"chunks": 3565.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.196929, "cps": 27.851562, "mean_p": 0.233537, "boundary_entropy": 0.237881, "bpb": 2.413823}, "fi": {"chunks": 3697.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.196534, "cps": 28.882812, "mean_p": 0.226605, "boundary_entropy": 0.198506, "bpb": 2.586815}}, "bpb": {"en": 2.13429, "tr": 2.41382, "fi": 2.58681}, "gini_cpc": 0.012032, "gini_cps": 0.008095, "bpic": 5.218655, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9496120587150951, "fi": 0.9477085642528978}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 64.228759765625, "n_util_obs": 159, "util_count_spearman": -0.05732484076439508}, "t": 501.9}
+{"event": "eval", "step": 1600, "phase": 1, "bpb": {"en": 2.15065, "tr": 2.42244, "fi": 2.59668}, "bytes_per_chunk": {"en": 4.611, "tr": 5.559, "fi": 5.195}, "bpb_mean": 2.38992, "frozen_inventory": true, "t": 502.8}
+{"event": "train", "step": 1600, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.50347, "bpb": 2.16905, "chunk_loss": 0.030483, "F": 0.20447, "G": 0.22537, "chunks": 209.38, "bytes_per_chunk": 4.891, "boundary_entropy": 0.14163, "router_noise": 0.0, "grad_norm": 0.7745, "lr": 0.0002668, "router_trainable": true, "sec_per_step": 0.328, "probe_sec": 0.0577, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03356, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.36739, "n_shadow": 0, "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": 0.79317, "trunc_abs_frac": 0.74261, "mean_delta_full": 1.03337, "read_hit_rate": {"surface": 0.82079}, "mean_gate": {"surface": 0.41372}, "n_reads": 5473, "n_cand": 6668, "read_suppressed_frac": 0.0976, "dropout_reads_logged": 5473, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 64.27253723144531, "n_util_obs": 160, "util_count_spearman": -0.06166354154459158}, "probe_overhead_sec_total": 100.306, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.038438, "alpha": 0.03, "t": 504.8}
+{"event": "consolidate", "step": 1650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 518.0}
+{"event": "train", "step": 1650, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.56148, "bpb": 2.25273, "chunk_loss": 0.030561, "F": 0.20959, "G": 0.22514, "chunks": 214.62, "bytes_per_chunk": 4.771, "boundary_entropy": 0.13352, "router_noise": 0.0, "grad_norm": 0.7934, "lr": 0.0002643, "router_trainable": true, "sec_per_step": 0.272, "probe_sec": 0.0366, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02592, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.1062, "n_shadow": 0, "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": 0.81366, "trunc_abs_frac": 0.93006, "mean_delta_full": 0.24512, "read_hit_rate": {"surface": 0.8268}, "mean_gate": {"surface": 0.41358}, "n_reads": 5652, "n_cand": 6836, "read_suppressed_frac": 0.1054, "dropout_reads_logged": 5652, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 66.40054321289062, "n_util_obs": 161, "util_count_spearman": -0.02769151138716356}, "probe_overhead_sec_total": 104.234, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.038201, "alpha": 0.03, "t": 518.4}
+{"event": "consolidate", "step": 1700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 531.3}
+{"event": "train", "step": 1700, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.5522, "bpb": 2.23935, "chunk_loss": 0.030552, "F": 0.2067, "G": 0.22861, "chunks": 211.66, "bytes_per_chunk": 4.838, "boundary_entropy": 0.1454, "router_noise": 0.0, "grad_norm": 0.8278, "lr": 0.0002618, "router_trainable": true, "sec_per_step": 0.262, "probe_sec": 0.081, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0722, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.07405, "n_shadow": 0, "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": -167.82143, "trunc_abs_frac": 1.27877, "mean_delta_full": -0.00085, "read_hit_rate": {"surface": 0.80092}, "mean_gate": {"surface": 0.40885}, "n_reads": 5399, "n_cand": 6741, "read_suppressed_frac": 0.0917, "dropout_reads_logged": 5399, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 68.47892761230469, "n_util_obs": 163, "util_count_spearman": -0.03879941282706416}, "probe_overhead_sec_total": 108.106, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.041088, "alpha": 0.03, "t": 531.5}
+{"event": "consolidate", "step": 1750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 544.3}
+{"event": "train", "step": 1750, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.52444, "bpb": 2.1993, "chunk_loss": 0.030653, "F": 0.21243, "G": 0.22763, "chunks": 217.53, "bytes_per_chunk": 4.707, "boundary_entropy": 0.12448, "router_noise": 0.0, "grad_norm": 0.7711, "lr": 0.0002591, "router_trainable": true, "sec_per_step": 0.261, "probe_sec": 0.0367, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02731, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.39472, "n_shadow": 0, "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": 1.41879, "trunc_abs_frac": 3.86092, "mean_delta_full": 0.2361, "read_hit_rate": {"surface": 0.83432}, "mean_gate": {"surface": 0.42354}, "n_reads": 5781, "n_cand": 6929, "read_suppressed_frac": 0.0974, "dropout_reads_logged": 5781, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 69.88323974609375, "n_util_obs": 162, "util_count_spearman": -0.013262090349313608}, "probe_overhead_sec_total": 111.001, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.048146, "alpha": 0.03, "t": 544.6}
+{"event": "consolidate", "step": 1800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 556.9}
+{"event": "probe", "step": 1800, "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": 0.215349, "boundary_entropy": 0.104674, "bpb": 2.101028}, "tr": {"chunks": 3673.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.202895, "cps": 28.695312, "mean_p": 0.238101, "boundary_entropy": 0.236506, "bpb": 2.349152}, "fi": {"chunks": 3827.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.203445, "cps": 29.898438, "mean_p": 0.229591, "boundary_entropy": 0.192364, "bpb": 2.561812}}, "bpb": {"en": 2.10103, "tr": 2.34915, "fi": 2.56181}, "gini_cpc": 0.005428, "gini_cps": 0.012532, "bpic": 5.102716, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9759463284948876, "fi": 0.9785930711829895}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 71.61691284179688, "n_util_obs": 161, "util_count_spearman": -0.00022141722841708937}, "t": 557.7}
+{"event": "eval", "step": 1800, "phase": 1, "bpb": {"en": 2.11857, "tr": 2.3591, "fi": 2.571}, "bytes_per_chunk": {"en": 4.61, "tr": 5.344, "fi": 5.067}, "bpb_mean": 2.34956, "frozen_inventory": true, "t": 559.0}
+{"event": "train", "step": 1800, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.57518, "bpb": 2.27251, "chunk_loss": 0.030616, "F": 0.21082, "G": 0.23044, "chunks": 215.88, "bytes_per_chunk": 4.743, "boundary_entropy": 0.1393, "router_noise": 0.0, "grad_norm": 0.8225, "lr": 0.0002564, "router_trainable": true, "sec_per_step": 0.313, "probe_sec": 0.0621, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05453, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.58136, "n_shadow": 0, "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": {"surface": 0.79828}, "mean_gate": {"surface": 0.40236}, "n_reads": 5489, "n_cand": 6876, "read_suppressed_frac": 0.1037, "dropout_reads_logged": 5489, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 71.64115905761719, "n_util_obs": 161, "util_count_spearman": 0.0005434786515692194}, "probe_overhead_sec_total": 114.235, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.050361, "alpha": 0.03, "t": 560.2}
+{"event": "consolidate", "step": 1850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 573.5}
+{"event": "train", "step": 1850, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.52896, "bpb": 2.20583, "chunk_loss": 0.030648, "F": 0.21298, "G": 0.22884, "chunks": 218.09, "bytes_per_chunk": 4.695, "boundary_entropy": 0.13147, "router_noise": 0.0, "grad_norm": 0.7233, "lr": 0.0002537, "router_trainable": true, "sec_per_step": 0.271, "probe_sec": 0.0314, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0224, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.39395, "n_shadow": 0, "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": 1.67125, "trunc_abs_frac": 1.67125, "mean_delta_full": -0.11548, "read_hit_rate": {"surface": 0.81589}, "mean_gate": {"surface": 0.41072}, "n_reads": 5668, "n_cand": 6947, "read_suppressed_frac": 0.0986, "dropout_reads_logged": 5668, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 73.48139953613281, "n_util_obs": 158, "util_count_spearman": 0.01753836183881258}, "probe_overhead_sec_total": 116.951, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.043369, "alpha": 0.03, "t": 573.8}
+{"event": "consolidate", "step": 1900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 586.3}
+{"event": "train", "step": 1900, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.56705, "bpb": 2.26077, "chunk_loss": 0.030593, "F": 0.21048, "G": 0.22665, "chunks": 215.53, "bytes_per_chunk": 4.751, "boundary_entropy": 0.12731, "router_noise": 0.0, "grad_norm": 0.7705, "lr": 0.0002508, "router_trainable": true, "sec_per_step": 0.254, "probe_sec": 0.0334, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02677, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.21679, "n_shadow": 0, "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": -0.0135, "trunc_abs_frac": 0.90117, "mean_delta_full": -0.20701, "read_hit_rate": {"surface": 0.81238}, "mean_gate": {"surface": 0.42119}, "n_reads": 5577, "n_cand": 6865, "read_suppressed_frac": 0.0979, "dropout_reads_logged": 5577, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 75.47975158691406, "n_util_obs": 156, "util_count_spearman": 0.016285502046751277}, "probe_overhead_sec_total": 119.685, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.042416, "alpha": 0.03, "t": 586.5}
+{"event": "consolidate", "step": 1950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 599.9}
+{"event": "train", "step": 1950, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.52392, "bpb": 2.19855, "chunk_loss": 0.030598, "F": 0.21078, "G": 0.22322, "chunks": 215.84, "bytes_per_chunk": 4.744, "boundary_entropy": 0.11852, "router_noise": 0.0, "grad_norm": 0.7155, "lr": 0.0002479, "router_trainable": true, "sec_per_step": 0.274, "probe_sec": 0.1123, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09948, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.13626, "n_shadow": 0, "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": 1.23927, "trunc_abs_frac": 1.18695, "mean_delta_full": 0.08154, "read_hit_rate": {"surface": 0.82051}, "mean_gate": {"surface": 0.41871}, "n_reads": 5641, "n_cand": 6875, "read_suppressed_frac": 0.1018, "dropout_reads_logged": 5641, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 77.37677001953125, "n_util_obs": 156, "util_count_spearman": -0.029327812109813344}, "probe_overhead_sec_total": 122.766, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.04154, "alpha": 0.03, "t": 600.2}
+{"event": "consolidate", "step": 2000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 612.2}
+{"event": "probe", "step": 2000, "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": 0.214247, "boundary_entropy": 0.101703, "bpb": 2.079444}, "tr": {"chunks": 3570.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.197205, "cps": 27.890625, "mean_p": 0.23054, "boundary_entropy": 0.22436, "bpb": 2.330018}, "fi": {"chunks": 3774.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.200627, "cps": 29.484375, "mean_p": 0.222927, "boundary_entropy": 0.176325, "bpb": 2.501788}}, "bpb": {"en": 2.07944, "tr": 2.33002, "fi": 2.50179}, "gini_cpc": 0.011766, "gini_cps": 0.012406, "bpic": 5.175333, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9485783808131634, "fi": 0.9650405671922139}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 79.63516235351562, "n_util_obs": 160, "util_count_spearman": -0.03885066715302457}, "t": 612.9}
+{"event": "eval", "step": 2000, "phase": 1, "bpb": {"en": 2.09812, "tr": 2.33666, "fi": 2.50933}, "bytes_per_chunk": {"en": 4.589, "tr": 5.386, "fi": 5.094}, "bpb_mean": 2.3147, "frozen_inventory": true, "t": 614.1}
+{"event": "train", "step": 2000, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.55061, "bpb": 2.23705, "chunk_loss": 0.030543, "F": 0.2084, "G": 0.22502, "chunks": 213.41, "bytes_per_chunk": 4.798, "boundary_entropy": 0.12588, "router_noise": 0.0, "grad_norm": 0.8098, "lr": 0.000245, "router_trainable": true, "sec_per_step": 0.302, "probe_sec": 0.0353, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02239, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.1744, "n_shadow": 0, "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": -1.44088, "trunc_abs_frac": 1.34549, "mean_delta_full": -0.15796, "read_hit_rate": {"surface": 0.79785}, "mean_gate": {"surface": 0.41326}, "n_reads": 5423, "n_cand": 6797, "read_suppressed_frac": 0.1058, "dropout_reads_logged": 5423, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 79.68685913085938, "n_util_obs": 160, "util_count_spearman": -0.03673385679128091}, "probe_overhead_sec_total": 126.272, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.037922, "alpha": 0.03, "t": 615.3}
+{"event": "probe", "step": 2048, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3604.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.207091, "cps": 28.15625, "mean_p": 0.213028, "boundary_entropy": 0.09985, "bpb": 2.063156}, "tr": {"chunks": 3411.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.188422, "cps": 26.648438, "mean_p": 0.225307, "boundary_entropy": 0.216107, "bpb": 2.319819}, "fi": {"chunks": 3761.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.199936, "cps": 29.382812, "mean_p": 0.226025, "boundary_entropy": 0.183962, "bpb": 2.482469}}, "bpb": {"en": 2.06316, "tr": 2.31982, "fi": 2.48247}, "gini_cpc": 0.020902, "gini_cps": 0.021653, "bpic": 5.264662, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9098514800282979, "fi": 0.9654522252459198}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 81.2320556640625, "n_util_obs": 160, "util_count_spearman": -0.006562756357670222}, "t": 628.2}
+{"event": "consolidate", "step": 2050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 630.2}
+{"event": "train", "step": 2050, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.60246, "bpb": 2.31186, "chunk_loss": 0.030512, "F": 0.20529, "G": 0.22439, "chunks": 210.22, "bytes_per_chunk": 4.871, "boundary_entropy": 0.13344, "router_noise": 0.0, "grad_norm": 0.8054, "lr": 0.0002419, "router_trainable": true, "sec_per_step": 0.303, "probe_sec": 0.0342, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02657, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.40092, "n_shadow": 0, "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": -0.33845, "trunc_abs_frac": 0.33845, "mean_delta_full": -0.37988, "read_hit_rate": {"surface": 0.78506}, "mean_gate": {"surface": 0.40683}, "n_reads": 5256, "n_cand": 6695, "read_suppressed_frac": 0.1052, "dropout_reads_logged": 5256, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 81.25120544433594, "n_util_obs": 160, "util_count_spearman": -0.0053293097386616665}, "probe_overhead_sec_total": 129.258, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.042364, "alpha": 0.03, "t": 630.4}
+{"event": "consolidate", "step": 2100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 643.1}
+{"event": "train", "step": 2100, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.48649, "bpb": 2.14455, "chunk_loss": 0.030631, "F": 0.21234, "G": 0.22849, "chunks": 217.44, "bytes_per_chunk": 4.709, "boundary_entropy": 0.12374, "router_noise": 0.0, "grad_norm": 0.7523, "lr": 0.0002388, "router_trainable": true, "sec_per_step": 0.26, "probe_sec": 0.0906, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08411, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.31425, "n_shadow": 0, "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": 16.81402, "trunc_abs_frac": 0.40554, "mean_delta_full": 0.01001, "read_hit_rate": {"surface": 0.81692}, "mean_gate": {"surface": 0.41986}, "n_reads": 5658, "n_cand": 6926, "read_suppressed_frac": 0.1016, "dropout_reads_logged": 5658, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 83.37583923339844, "n_util_obs": 165, "util_count_spearman": -0.050297865519728584}, "probe_overhead_sec_total": 132.819, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.041518, "alpha": 0.03, "t": 643.4}
+{"event": "consolidate", "step": 2150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 655.8}
+{"event": "train", "step": 2150, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.50735, "bpb": 2.17465, "chunk_loss": 0.030453, "F": 0.20529, "G": 0.21847, "chunks": 210.22, "bytes_per_chunk": 4.871, "boundary_entropy": 0.12022, "router_noise": 0.0, "grad_norm": 0.707, "lr": 0.0002357, "router_trainable": true, "sec_per_step": 0.263, "probe_sec": 0.1073, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09355, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.56464, "n_shadow": 0, "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": 0.8605, "trunc_abs_frac": 0.80993, "mean_delta_full": 1.32269, "read_hit_rate": {"surface": 0.80807}, "mean_gate": {"surface": 0.41935}, "n_reads": 5410, "n_cand": 6695, "read_suppressed_frac": 0.0982, "dropout_reads_logged": 5410, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 85.32609558105469, "n_util_obs": 170, "util_count_spearman": -0.07891134030691067}, "probe_overhead_sec_total": 135.301, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.040917, "alpha": 0.03, "t": 656.6}
+{"event": "consolidate", "step": 2200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 676.5}
+{"event": "probe", "step": 2200, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3610.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.207435, "cps": 28.203125, "mean_p": 0.212578, "boundary_entropy": 0.095722, "bpb": 2.045961}, "tr": {"chunks": 3429.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.189416, "cps": 26.789062, "mean_p": 0.227345, "boundary_entropy": 0.21717, "bpb": 2.280465}, "fi": {"chunks": 3658.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.194461, "cps": 28.578125, "mean_p": 0.21727, "boundary_entropy": 0.167723, "bpb": 2.454614}}, "bpb": {"en": 2.04596, "tr": 2.28046, "fi": 2.45461}, "gini_cpc": 0.020316, "gini_cps": 0.014272, "bpic": 5.303543, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9131326085283304, "fi": 0.9374513438901121}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 87.03370666503906, "n_util_obs": 172, "util_count_spearman": -0.038363484999634455}, "t": 677.1}
+{"event": "eval", "step": 2200, "phase": 1, "bpb": {"en": 2.06487, "tr": 2.28863, "fi": 2.46406}, "bytes_per_chunk": {"en": 4.606, "tr": 5.593, "fi": 5.235}, "bpb_mean": 2.27252, "frozen_inventory": true, "t": 678.2}
+{"event": "train", "step": 2200, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.46993, "bpb": 2.12067, "chunk_loss": 0.030554, "F": 0.21036, "G": 0.22226, "chunks": 215.41, "bytes_per_chunk": 4.754, "boundary_entropy": 0.11035, "router_noise": 0.0, "grad_norm": 0.6446, "lr": 0.0002325, "router_trainable": true, "sec_per_step": 0.455, "probe_sec": 0.0866, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07102, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.10446, "n_shadow": 0, "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": 0.69899, "trunc_abs_frac": 1.26994, "mean_delta_full": -0.23259, "read_hit_rate": {"surface": 0.82714}, "mean_gate": {"surface": 0.42257}, "n_reads": 5675, "n_cand": 6861, "read_suppressed_frac": 0.1024, "dropout_reads_logged": 5675, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 87.09474182128906, "n_util_obs": 172, "util_count_spearman": -0.0410119262398502}, "probe_overhead_sec_total": 138.347, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.045623, "alpha": 0.03, "t": 679.3}
+{"event": "consolidate", "step": 2250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 705.7}
+{"event": "train", "step": 2250, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.50632, "bpb": 2.17316, "chunk_loss": 0.030481, "F": 0.20605, "G": 0.21997, "chunks": 211.0, "bytes_per_chunk": 4.853, "boundary_entropy": 0.12126, "router_noise": 0.0, "grad_norm": 0.7489, "lr": 0.0002293, "router_trainable": true, "sec_per_step": 0.533, "probe_sec": 0.0899, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07135, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.57928, "n_shadow": 0, "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": 0.96347, "trunc_abs_frac": 0.63661, "mean_delta_full": 0.53739, "read_hit_rate": {"surface": 0.80818}, "mean_gate": {"surface": 0.41619}, "n_reads": 5431, "n_cand": 6720, "read_suppressed_frac": 0.104, "dropout_reads_logged": 5431, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 89.14297485351562, "n_util_obs": 175, "util_count_spearman": -0.021793551276309896}, "probe_overhead_sec_total": 141.767, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.044461, "alpha": 0.03, "t": 706.0}
+{"event": "consolidate", "step": 2300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 729.0}
+{"event": "train", "step": 2300, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.4871, "bpb": 2.14543, "chunk_loss": 0.030567, "F": 0.21194, "G": 0.2218, "chunks": 217.03, "bytes_per_chunk": 4.718, "boundary_entropy": 0.11171, "router_noise": 0.0, "grad_norm": 0.7265, "lr": 0.000226, "router_trainable": true, "sec_per_step": 0.467, "probe_sec": 0.0893, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08241, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.04084, "n_shadow": 0, "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": 0.53069, "trunc_abs_frac": 0.88724, "mean_delta_full": 0.22036, "read_hit_rate": {"surface": 0.83046}, "mean_gate": {"surface": 0.42561}, "n_reads": 5741, "n_cand": 6913, "read_suppressed_frac": 0.0995, "dropout_reads_logged": 5741, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 90.77224731445312, "n_util_obs": 174, "util_count_spearman": -0.029140920869045246}, "probe_overhead_sec_total": 144.861, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.043063, "alpha": 0.03, "t": 729.3}
+{"event": "consolidate", "step": 2350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 742.5}
+{"event": "train", "step": 2350, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.48981, "bpb": 2.14934, "chunk_loss": 0.030409, "F": 0.19839, "G": 0.22345, "chunks": 203.16, "bytes_per_chunk": 5.04, "boundary_entropy": 0.13971, "router_noise": 0.0, "grad_norm": 0.7675, "lr": 0.0002226, "router_trainable": true, "sec_per_step": 0.269, "probe_sec": 0.0711, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05587, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.221, "n_shadow": 0, "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": -0.55459, "trunc_abs_frac": 0.55459, "mean_delta_full": 0.02795, "read_hit_rate": {"surface": 0.77524}, "mean_gate": {"surface": 0.40506}, "n_reads": 5015, "n_cand": 6469, "read_suppressed_frac": 0.1023, "dropout_reads_logged": 5015, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 92.64837646484375, "n_util_obs": 173, "util_count_spearman": -0.05047953238060362}, "probe_overhead_sec_total": 148.008, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.046218, "alpha": 0.03, "t": 742.8}
+{"event": "consolidate", "step": 2400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 760.9}
+{"event": "probe", "step": 2400, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3600.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.206861, "cps": 28.125, "mean_p": 0.212057, "boundary_entropy": 0.096631, "bpb": 2.026165}, "tr": {"chunks": 3511.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.193946, "cps": 27.429688, "mean_p": 0.231372, "boundary_entropy": 0.221347, "bpb": 2.254497}, "fi": {"chunks": 3712.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.197331, "cps": 29.0, "mean_p": 0.220354, "boundary_entropy": 0.169209, "bpb": 2.422158}}, "bpb": {"en": 2.02617, "tr": 2.2545, "fi": 2.42216}, "gini_cpc": 0.014395, "gini_cps": 0.012381, "bpic": 5.2418, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9375661032241434, "fi": 0.9539326280722271}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 94.35458374023438, "n_util_obs": 174, "util_count_spearman": -0.07129701870782197}, "t": 761.6}
+{"event": "eval", "step": 2400, "phase": 1, "bpb": {"en": 2.04235, "tr": 2.26268, "fi": 2.43403}, "bytes_per_chunk": {"en": 4.617, "tr": 5.41, "fi": 5.148}, "bpb_mean": 2.24635, "frozen_inventory": true, "t": 762.6}
+{"event": "phase_transition", "step": 2400, "phase": 2, "objective": "global", "router_trainable": true, "encoder_trainable": true, "frozen_router_params": 0, "t": 764.0}
+{"event": "train", "step": 2400, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.44215, "bpb": 2.08058, "chunk_loss": 0.030674, "F": 0.21576, "G": 0.22761, "chunks": 220.94, "bytes_per_chunk": 4.635, "boundary_entropy": 0.11283, "router_noise": 0.0, "grad_norm": 0.7378, "lr": 0.0002192, "router_trainable": true, "sec_per_step": 0.442, "probe_sec": 0.0547, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02626, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.08382, "n_shadow": 0, "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": 0.32618, "trunc_abs_frac": 0.78144, "mean_delta_full": -0.40743, "read_hit_rate": {"surface": 0.83816}, "mean_gate": {"surface": 0.42046}, "n_reads": 5899, "n_cand": 7038, "read_suppressed_frac": 0.0995, "dropout_reads_logged": 5899, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 94.42155456542969, "n_util_obs": 174, "util_count_spearman": -0.07155902591705134}, "probe_overhead_sec_total": 151.294, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.038961, "alpha": 0.03, "t": 764.9}
+{"event": "consolidate", "step": 2450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 781.8}
+{"event": "train", "step": 2450, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.62788, "bpb": 2.34854, "chunk_loss": 0.030588, "F": 0.21036, "G": 0.22476, "chunks": 215.41, "bytes_per_chunk": 4.754, "boundary_entropy": 0.11269, "router_noise": 0.0, "grad_norm": 0.6968, "lr": 0.0002158, "router_trainable": true, "sec_per_step": 0.348, "probe_sec": 0.0526, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0384, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.12709, "n_shadow": 0, "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": -1.2808, "trunc_abs_frac": 2.47169, "mean_delta_full": -0.13245, "read_hit_rate": {"surface": 0.7786}, "mean_gate": {"surface": 0.40837}, "n_reads": 5342, "n_cand": 6861, "read_suppressed_frac": 0.1, "dropout_reads_logged": 5342, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 95.95594787597656, "n_util_obs": 175, "util_count_spearman": -0.008901706729573008}, "probe_overhead_sec_total": 154.343, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.045221, "alpha": 0.03, "t": 782.3}
+{"event": "consolidate", "step": 2500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 804.4}
+{"event": "train", "step": 2500, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.47722, "bpb": 2.13118, "chunk_loss": 0.030547, "F": 0.21036, "G": 0.22129, "chunks": 215.41, "bytes_per_chunk": 4.754, "boundary_entropy": 0.10726, "router_noise": 0.0, "grad_norm": 0.6786, "lr": 0.0002123, "router_trainable": true, "sec_per_step": 0.45, "probe_sec": 0.0595, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05251, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.63229, "n_shadow": 0, "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": -11.19935, "trunc_abs_frac": 1.47223, "mean_delta_full": -0.0485, "read_hit_rate": {"surface": 0.82043}, "mean_gate": {"surface": 0.42486}, "n_reads": 5629, "n_cand": 6861, "read_suppressed_frac": 0.0974, "dropout_reads_logged": 5629, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 97.67604064941406, "n_util_obs": 177, "util_count_spearman": -0.06317411402157165}, "probe_overhead_sec_total": 157.644, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.053208, "alpha": 0.03, "t": 804.7}
+{"event": "consolidate", "step": 2550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 820.0}
+{"event": "train", "step": 2550, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.46911, "bpb": 2.11948, "chunk_loss": 0.030631, "F": 0.21292, "G": 0.22705, "chunks": 218.03, "bytes_per_chunk": 4.697, "boundary_entropy": 0.13006, "router_noise": 0.0, "grad_norm": 0.691, "lr": 0.0002088, "router_trainable": true, "sec_per_step": 0.323, "probe_sec": 0.0778, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04429, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.30812, "n_shadow": 0, "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": 0.54775, "trunc_abs_frac": 0.757, "mean_delta_full": 0.31964, "read_hit_rate": {"surface": 0.79698}, "mean_gate": {"surface": 0.41293}, "n_reads": 5535, "n_cand": 6945, "read_suppressed_frac": 0.093, "dropout_reads_logged": 5535, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 99.86317443847656, "n_util_obs": 173, "util_count_spearman": -0.09094427782747959}, "probe_overhead_sec_total": 161.118, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.060536, "alpha": 0.03, "t": 820.9}
+{"event": "consolidate", "step": 2600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 835.1}
+{"event": "probe", "step": 2600, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3607.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.207263, "cps": 28.179688, "mean_p": 0.209935, "boundary_entropy": 0.090802, "bpb": 2.004742}, "tr": {"chunks": 3301.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.182345, "cps": 25.789062, "mean_p": 0.221731, "boundary_entropy": 0.202931, "bpb": 2.236975}, "fi": {"chunks": 3532.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.187762, "cps": 27.59375, "mean_p": 0.211312, "boundary_entropy": 0.156163, "bpb": 2.392735}}, "bpb": {"en": 2.00474, "tr": 2.23697, "fi": 2.39273}, "gini_cpc": 0.028771, "gini_cps": 0.01954, "bpic": 5.4341, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.8797777024337571, "fi": 0.90591362046141}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 101.35519409179688, "n_util_obs": 170, "util_count_spearman": -0.09562722237116991}, "t": 835.8}
+{"event": "eval", "step": 2600, "phase": 2, "bpb": {"en": 2.02096, "tr": 2.24252, "fi": 2.40573}, "bytes_per_chunk": {"en": 4.613, "tr": 5.857, "fi": 5.386}, "bpb_mean": 2.22307, "frozen_inventory": true, "t": 837.1}
+{"event": "train", "step": 2600, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.49646, "bpb": 2.15894, "chunk_loss": 0.030454, "F": 0.20309, "G": 0.21817, "chunks": 207.97, "bytes_per_chunk": 4.924, "boundary_entropy": 0.11471, "router_noise": 0.0, "grad_norm": 0.704, "lr": 0.0002053, "router_trainable": true, "sec_per_step": 0.347, "probe_sec": 0.0488, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04157, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.0432, "n_shadow": 0, "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": -0.53052, "trunc_abs_frac": 1.79136, "mean_delta_full": -0.09867, "read_hit_rate": {"surface": 0.79662}, "mean_gate": {"surface": 0.39801}, "n_reads": 5276, "n_cand": 6623, "read_suppressed_frac": 0.097, "dropout_reads_logged": 5276, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 101.39878845214844, "n_util_obs": 170, "util_count_spearman": -0.09541599699080398}, "probe_overhead_sec_total": 164.089, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.054153, "alpha": 0.03, "t": 838.3}
+{"event": "consolidate", "step": 2650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 850.1}
+{"event": "train", "step": 2650, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.45464, "bpb": 2.0986, "chunk_loss": 0.030473, "F": 0.20258, "G": 0.22372, "chunks": 207.44, "bytes_per_chunk": 4.936, "boundary_entropy": 0.1261, "router_noise": 0.0, "grad_norm": 0.7254, "lr": 0.0002017, "router_trainable": true, "sec_per_step": 0.241, "probe_sec": 0.0714, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05824, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.3942, "n_shadow": 0, "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": 0.31345, "trunc_abs_frac": 0.7362, "mean_delta_full": 0.42441, "read_hit_rate": {"surface": 0.79761}, "mean_gate": {"surface": 0.40377}, "n_reads": 5269, "n_cand": 6606, "read_suppressed_frac": 0.1002, "dropout_reads_logged": 5269, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 102.74313354492188, "n_util_obs": 166, "util_count_spearman": -0.11519961172944364}, "probe_overhead_sec_total": 166.709, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.032231, "alpha": 0.03, "t": 850.3}
+{"event": "consolidate", "step": 2700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 862.9}
+{"event": "train", "step": 2700, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.53707, "bpb": 2.21753, "chunk_loss": 0.030548, "F": 0.20764, "G": 0.22403, "chunks": 212.62, "bytes_per_chunk": 4.816, "boundary_entropy": 0.11532, "router_noise": 0.0, "grad_norm": 0.8149, "lr": 0.0001981, "router_trainable": true, "sec_per_step": 0.257, "probe_sec": 0.0389, "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": 0.00313, "n_shadow": 0, "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": 0.50583, "trunc_abs_frac": 1.32433, "mean_delta_full": -0.24946, "read_hit_rate": {"surface": 0.785}, "mean_gate": {"surface": 0.40864}, "n_reads": 5316, "n_cand": 6772, "read_suppressed_frac": 0.1031, "dropout_reads_logged": 5316, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 104.85035705566406, "n_util_obs": 167, "util_count_spearman": -0.11944613355045502}, "probe_overhead_sec_total": 169.393, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.039667, "alpha": 0.03, "t": 863.2}
+{"event": "consolidate", "step": 2750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 876.9}
+{"event": "train", "step": 2750, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.47244, "bpb": 2.12429, "chunk_loss": 0.030507, "F": 0.20728, "G": 0.2221, "chunks": 212.25, "bytes_per_chunk": 4.824, "boundary_entropy": 0.11606, "router_noise": 0.0, "grad_norm": 0.6949, "lr": 0.0001945, "router_trainable": true, "sec_per_step": 0.279, "probe_sec": 0.0551, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04841, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.05013, "n_shadow": 0, "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": 0.80299, "trunc_abs_frac": 1.10836, "mean_delta_full": 0.33439, "read_hit_rate": {"surface": 0.79926}, "mean_gate": {"surface": 0.41044}, "n_reads": 5403, "n_cand": 6760, "read_suppressed_frac": 0.0953, "dropout_reads_logged": 5403, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 106.81401062011719, "n_util_obs": 163, "util_count_spearman": -0.12725212119196858}, "probe_overhead_sec_total": 172.274, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.038816, "alpha": 0.03, "t": 877.1}
+{"event": "consolidate", "step": 2800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 889.3}
+{"event": "probe", "step": 2800, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3610.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.207435, "cps": 28.203125, "mean_p": 0.209436, "boundary_entropy": 0.094114, "bpb": 1.985357}, "tr": {"chunks": 3351.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.185107, "cps": 26.179688, "mean_p": 0.223837, "boundary_entropy": 0.204175, "bpb": 2.212226}, "fi": {"chunks": 3666.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.194886, "cps": 28.640625, "mean_p": 0.215749, "boundary_entropy": 0.16017, "bpb": 2.358171}}, "bpb": {"en": 1.98536, "tr": 2.21223, "fi": 2.35817}, "gini_cpc": 0.02534, "gini_cps": 0.019761, "bpic": 5.338477, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.892361438080617, "fi": 0.9395015381905825}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 108.55421447753906, "n_util_obs": 164, "util_count_spearman": -0.1700216288275543}, "t": 890.1}
+{"event": "eval", "step": 2800, "phase": 2, "bpb": {"en": 2.00493, "tr": 2.21773, "fi": 2.36884}, "bytes_per_chunk": {"en": 4.616, "tr": 5.743, "fi": 5.215}, "bpb_mean": 2.19717, "frozen_inventory": true, "t": 891.3}
+{"event": "train", "step": 2800, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.45048, "bpb": 2.0926, "chunk_loss": 0.030547, "F": 0.21024, "G": 0.22082, "chunks": 215.28, "bytes_per_chunk": 4.757, "boundary_entropy": 0.10703, "router_noise": 0.0, "grad_norm": 0.6564, "lr": 0.0001909, "router_trainable": true, "sec_per_step": 0.309, "probe_sec": 0.0404, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03334, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.07595, "n_shadow": 0, "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": 0.10639, "trunc_abs_frac": 0.11782, "mean_delta_full": -0.18777, "read_hit_rate": {"surface": 0.82573}, "mean_gate": {"surface": 0.42469}, "n_reads": 5662, "n_cand": 6857, "read_suppressed_frac": 0.0998, "dropout_reads_logged": 5662, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 108.6044921875, "n_util_obs": 164, "util_count_spearman": -0.16989376028729614}, "probe_overhead_sec_total": 175.377, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.037235, "alpha": 0.03, "t": 892.6}
+{"event": "consolidate", "step": 2850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 905.5}
+{"event": "train", "step": 2850, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.53249, "bpb": 2.21091, "chunk_loss": 0.030562, "F": 0.21085, "G": 0.22227, "chunks": 215.91, "bytes_per_chunk": 4.743, "boundary_entropy": 0.1053, "router_noise": 0.0, "grad_norm": 0.6648, "lr": 0.0001872, "router_trainable": true, "sec_per_step": 0.262, "probe_sec": 0.0301, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02306, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.0522, "n_shadow": 0, "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": 1.44701, "trunc_abs_frac": 1.51625, "mean_delta_full": -0.26467, "read_hit_rate": {"surface": 0.7986}, "mean_gate": {"surface": 0.42437}, "n_reads": 5492, "n_cand": 6877, "read_suppressed_frac": 0.1007, "dropout_reads_logged": 5492, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 110.67469787597656, "n_util_obs": 168, "util_count_spearman": -0.16122109120727268}, "probe_overhead_sec_total": 178.677, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.037503, "alpha": 0.03, "t": 905.7}
+{"event": "consolidate", "step": 2900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 919.2}
+{"event": "train", "step": 2900, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.41264, "bpb": 2.03801, "chunk_loss": 0.03046, "F": 0.20432, "G": 0.22103, "chunks": 209.22, "bytes_per_chunk": 4.894, "boundary_entropy": 0.11586, "router_noise": 0.0, "grad_norm": 0.697, "lr": 0.0001835, "router_trainable": true, "sec_per_step": 0.276, "probe_sec": 0.0302, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02224, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.20448, "n_shadow": 0, "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": 1.80625, "trunc_abs_frac": 2.93494, "mean_delta_full": 0.1633, "read_hit_rate": {"surface": 0.81555}, "mean_gate": {"surface": 0.41991}, "n_reads": 5434, "n_cand": 6663, "read_suppressed_frac": 0.0959, "dropout_reads_logged": 5434, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 112.90226745605469, "n_util_obs": 168, "util_count_spearman": -0.15401319086256904}, "probe_overhead_sec_total": 182.232, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.044746, "alpha": 0.03, "t": 919.5}
+{"event": "consolidate", "step": 2950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 933.2}
+{"event": "train", "step": 2950, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.3512, "bpb": 1.94937, "chunk_loss": 0.030644, "F": 0.21667, "G": 0.22273, "chunks": 221.88, "bytes_per_chunk": 4.615, "boundary_entropy": 0.09536, "router_noise": 0.0, "grad_norm": 0.6423, "lr": 0.0001799, "router_trainable": true, "sec_per_step": 0.281, "probe_sec": 0.0627, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05611, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.07473, "n_shadow": 0, "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": 0.38042, "trunc_abs_frac": 0.38042, "mean_delta_full": 0.49512, "read_hit_rate": {"surface": 0.8663}, "mean_gate": {"surface": 0.43125}, "n_reads": 6123, "n_cand": 7068, "read_suppressed_frac": 0.1006, "dropout_reads_logged": 6123, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 114.65042114257812, "n_util_obs": 166, "util_count_spearman": -0.16730394632421902}, "probe_overhead_sec_total": 185.188, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.044636, "alpha": 0.03, "t": 933.6}
+{"event": "consolidate", "step": 3000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 946.1}
+{"event": "probe", "step": 3000, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3616.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.20778, "cps": 28.25, "mean_p": 0.209668, "boundary_entropy": 0.08674, "bpb": 1.970725}, "tr": {"chunks": 3359.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.185549, "cps": 26.242188, "mean_p": 0.215784, "boundary_entropy": 0.192873, "bpb": 2.195844}, "fi": {"chunks": 3761.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.199936, "cps": 29.382812, "mean_p": 0.220235, "boundary_entropy": 0.156121, "bpb": 2.330348}}, "bpb": {"en": 1.97072, "tr": 2.19584, "fi": 2.33035}, "gini_cpc": 0.024981, "gini_cps": 0.024963, "bpic": 5.284277, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.8930075913932027, "fi": 0.9622482908701037}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 116.27212524414062, "n_util_obs": 165, "util_count_spearman": -0.1950872225042075}, "t": 946.9}
+{"event": "eval", "step": 3000, "phase": 2, "bpb": {"en": 1.98769, "tr": 2.20157, "fi": 2.3398}, "bytes_per_chunk": {"en": 4.605, "tr": 5.74, "fi": 5.109}, "bpb_mean": 2.17635, "frozen_inventory": true, "t": 948.0}
+{"event": "train", "step": 3000, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.44347, "bpb": 2.08249, "chunk_loss": 0.030522, "F": 0.20712, "G": 0.21964, "chunks": 212.09, "bytes_per_chunk": 4.828, "boundary_entropy": 0.11054, "router_noise": 0.0, "grad_norm": 0.7526, "lr": 0.0001761, "router_trainable": true, "sec_per_step": 0.316, "probe_sec": 0.0315, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02087, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.0556, "n_shadow": 0, "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": 0.36306, "trunc_abs_frac": 0.47227, "mean_delta_full": 0.10986, "read_hit_rate": {"surface": 0.79867}, "mean_gate": {"surface": 0.40265}, "n_reads": 5395, "n_cand": 6755, "read_suppressed_frac": 0.0986, "dropout_reads_logged": 5395, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 116.31494140625, "n_util_obs": 165, "util_count_spearman": -0.1947105495151337}, "probe_overhead_sec_total": 187.705, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.043201, "alpha": 0.03, "t": 949.4}
+{"event": "consolidate", "step": 3050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 962.7}
+{"event": "train", "step": 3050, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.4075, "bpb": 2.0306, "chunk_loss": 0.030527, "F": 0.20935, "G": 0.22053, "chunks": 214.38, "bytes_per_chunk": 4.777, "boundary_entropy": 0.09667, "router_noise": 0.0, "grad_norm": 0.7192, "lr": 0.0001724, "router_trainable": true, "sec_per_step": 0.272, "probe_sec": 0.0852, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07088, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.1478, "n_shadow": 0, "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": -7.74659, "trunc_abs_frac": 1.00816, "mean_delta_full": 0.0224, "read_hit_rate": {"surface": 0.82865}, "mean_gate": {"surface": 0.42761}, "n_reads": 5658, "n_cand": 6828, "read_suppressed_frac": 0.106, "dropout_reads_logged": 5658, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 117.98377990722656, "n_util_obs": 167, "util_count_spearman": -0.17359085615344183}, "probe_overhead_sec_total": 191.728, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.046979, "alpha": 0.03, "t": 962.9}
+{"event": "consolidate", "step": 3100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 976.0}
+{"event": "train", "step": 3100, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.49964, "bpb": 2.16353, "chunk_loss": 0.030428, "F": 0.20377, "G": 0.21563, "chunks": 208.66, "bytes_per_chunk": 4.908, "boundary_entropy": 0.10073, "router_noise": 0.0, "grad_norm": 0.7942, "lr": 0.0001687, "router_trainable": true, "sec_per_step": 0.265, "probe_sec": 0.0409, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03345, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.44486, "n_shadow": 0, "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": 2.81283, "trunc_abs_frac": 10.72725, "mean_delta_full": 0.01389, "read_hit_rate": {"surface": 0.80361}, "mean_gate": {"surface": 0.40861}, "n_reads": 5340, "n_cand": 6645, "read_suppressed_frac": 0.1039, "dropout_reads_logged": 5340, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 120.26145935058594, "n_util_obs": 166, "util_count_spearman": -0.1789336988673256}, "probe_overhead_sec_total": 195.013, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.053046, "alpha": 0.03, "t": 976.2}
+{"event": "consolidate", "step": 3150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 990.3}
+{"event": "train", "step": 3150, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.40214, "bpb": 2.02287, "chunk_loss": 0.030464, "F": 0.2067, "G": 0.21581, "chunks": 211.66, "bytes_per_chunk": 4.838, "boundary_entropy": 0.09695, "router_noise": 0.0, "grad_norm": 0.671, "lr": 0.000165, "router_trainable": true, "sec_per_step": 0.288, "probe_sec": 0.0414, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03525, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.22676, "n_shadow": 0, "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": 1.09012, "trunc_abs_frac": 1.13116, "mean_delta_full": 0.25398, "read_hit_rate": {"surface": 0.82377}, "mean_gate": {"surface": 0.41773}, "n_reads": 5553, "n_cand": 6741, "read_suppressed_frac": 0.099, "dropout_reads_logged": 5553, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 122.43319702148438, "n_util_obs": 168, "util_count_spearman": -0.1425939330130947}, "probe_overhead_sec_total": 198.07, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.052937, "alpha": 0.03, "t": 990.6}
+{"event": "consolidate", "step": 3200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1002.7}
+{"event": "probe", "step": 3200, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3605.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.207148, "cps": 28.164062, "mean_p": 0.209193, "boundary_entropy": 0.0798, "bpb": 1.952485}, "tr": {"chunks": 3282.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.181296, "cps": 25.640625, "mean_p": 0.213658, "boundary_entropy": 0.186236, "bpb": 2.163133}, "fi": {"chunks": 3702.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.1968, "cps": 28.921875, "mean_p": 0.215811, "boundary_entropy": 0.148539, "bpb": 2.305786}}, "bpb": {"en": 1.95248, "tr": 2.16313, "fi": 2.30579}, "gini_cpc": 0.029449, "gini_cps": 0.026443, "bpic": 5.357635, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.8751991283044175, "fi": 0.9500432619359626}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 124.22503662109375, "n_util_obs": 170, "util_count_spearman": -0.12742594390605821}, "t": 1003.5}
+{"event": "eval", "step": 3200, "phase": 2, "bpb": {"en": 1.96942, "tr": 2.16989, "fi": 2.31812}, "bytes_per_chunk": {"en": 4.624, "tr": 5.898, "fi": 5.15}, "bpb_mean": 2.15247, "frozen_inventory": true, "t": 1004.6}
+{"event": "train", "step": 3200, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.33396, "bpb": 1.9245, "chunk_loss": 0.03053, "F": 0.21216, "G": 0.21744, "chunks": 217.25, "bytes_per_chunk": 4.713, "boundary_entropy": 0.08555, "router_noise": 0.0, "grad_norm": 0.6683, "lr": 0.0001613, "router_trainable": true, "sec_per_step": 0.303, "probe_sec": 0.033, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02688, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.24287, "n_shadow": 0, "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": 1.30055, "trunc_abs_frac": 1.9045, "mean_delta_full": -0.1584, "read_hit_rate": {"surface": 0.86214}, "mean_gate": {"surface": 0.43012}, "n_reads": 5966, "n_cand": 6920, "read_suppressed_frac": 0.1009, "dropout_reads_logged": 5966, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 124.27215576171875, "n_util_obs": 170, "util_count_spearman": -0.1272891592015193}, "probe_overhead_sec_total": 200.454, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.052755, "alpha": 0.03, "t": 1005.7}
+{"event": "consolidate", "step": 3250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1018.6}
+{"event": "train", "step": 3250, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.33707, "bpb": 1.92899, "chunk_loss": 0.030535, "F": 0.20975, "G": 0.21989, "chunks": 214.78, "bytes_per_chunk": 4.768, "boundary_entropy": 0.09947, "router_noise": 0.0, "grad_norm": 0.6704, "lr": 0.0001576, "router_trainable": true, "sec_per_step": 0.262, "probe_sec": 0.0507, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04287, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.02238, "n_shadow": 0, "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": 0.01104, "trunc_abs_frac": 3.279, "mean_delta_full": -0.18333, "read_hit_rate": {"surface": 0.84242}, "mean_gate": {"surface": 0.40933}, "n_reads": 5763, "n_cand": 6841, "read_suppressed_frac": 0.0977, "dropout_reads_logged": 5763, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 125.9429931640625, "n_util_obs": 171, "util_count_spearman": -0.12332061343509251}, "probe_overhead_sec_total": 203.496, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.059013, "alpha": 0.03, "t": 1018.8}
+{"event": "consolidate", "step": 3300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1031.9}
+{"event": "train", "step": 3300, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.46148, "bpb": 2.10847, "chunk_loss": 0.03053, "F": 0.20905, "G": 0.21936, "chunks": 214.06, "bytes_per_chunk": 4.784, "boundary_entropy": 0.09752, "router_noise": 0.0, "grad_norm": 0.7288, "lr": 0.0001539, "router_trainable": true, "sec_per_step": 0.266, "probe_sec": 0.0694, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06079, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.03342, "n_shadow": 0, "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": 0.02244, "trunc_abs_frac": 1.1593, "mean_delta_full": -0.56946, "read_hit_rate": {"surface": 0.81476}, "mean_gate": {"surface": 0.4067}, "n_reads": 5555, "n_cand": 6818, "read_suppressed_frac": 0.0986, "dropout_reads_logged": 5555, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 127.29962158203125, "n_util_obs": 171, "util_count_spearman": -0.17565459476324188}, "probe_overhead_sec_total": 206.929, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.044051, "alpha": 0.03, "t": 1032.1}
+{"event": "consolidate", "step": 3350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1045.7}
+{"event": "train", "step": 3350, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.41349, "bpb": 2.03924, "chunk_loss": 0.030471, "F": 0.20673, "G": 0.21661, "chunks": 211.69, "bytes_per_chunk": 4.837, "boundary_entropy": 0.0969, "router_noise": 0.0, "grad_norm": 0.6466, "lr": 0.0001501, "router_trainable": true, "sec_per_step": 0.277, "probe_sec": 0.0883, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07806, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.04296, "n_shadow": 0, "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": 0.04162, "trunc_abs_frac": 0.74994, "mean_delta_full": -0.33879, "read_hit_rate": {"surface": 0.81637}, "mean_gate": {"surface": 0.41316}, "n_reads": 5504, "n_cand": 6742, "read_suppressed_frac": 0.0941, "dropout_reads_logged": 5504, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 129.2159881591797, "n_util_obs": 175, "util_count_spearman": -0.1463613972234662}, "probe_overhead_sec_total": 210.279, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.066532, "alpha": 0.03, "t": 1046.0}
+{"event": "consolidate", "step": 3400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1058.6}
+{"event": "probe", "step": 3400, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3603.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.207033, "cps": 28.148438, "mean_p": 0.208983, "boundary_entropy": 0.07778, "bpb": 1.931995}, "tr": {"chunks": 3317.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.183229, "cps": 25.914062, "mean_p": 0.213252, "boundary_entropy": 0.183645, "bpb": 2.138783}, "fi": {"chunks": 3671.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.195152, "cps": 28.679688, "mean_p": 0.214315, "boundary_entropy": 0.14381, "bpb": 2.282423}}, "bpb": {"en": 1.932, "tr": 2.13878, "fi": 2.28242}, "gini_cpc": 0.027108, "gini_cps": 0.022283, "bpic": 5.356624, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.8850234500949625, "fi": 0.9426106865829694}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 131.2031707763672, "n_util_obs": 173, "util_count_spearman": -0.10978023557115112}, "t": 1059.3}
+{"event": "eval", "step": 3400, "phase": 2, "bpb": {"en": 1.95305, "tr": 2.14656, "fi": 2.29287}, "bytes_per_chunk": {"en": 4.62, "tr": 5.734, "fi": 5.127}, "bpb_mean": 2.13083, "frozen_inventory": true, "t": 1060.7}
+{"event": "train", "step": 3400, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.37254, "bpb": 1.98016, "chunk_loss": 0.030524, "F": 0.21106, "G": 0.21854, "chunks": 216.12, "bytes_per_chunk": 4.738, "boundary_entropy": 0.08771, "router_noise": 0.0, "grad_norm": 0.6649, "lr": 0.0001465, "router_trainable": true, "sec_per_step": 0.315, "probe_sec": 0.0412, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0304, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.12226, "n_shadow": 0, "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": -0.82248, "trunc_abs_frac": 0.82248, "mean_delta_full": -0.07495, "read_hit_rate": {"surface": 0.83701}, "mean_gate": {"surface": 0.42771}, "n_reads": 5762, "n_cand": 6884, "read_suppressed_frac": 0.1038, "dropout_reads_logged": 5762, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 131.2534637451172, "n_util_obs": 173, "util_count_spearman": -0.10962958499242112}, "probe_overhead_sec_total": 213.521, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.059385, "alpha": 0.03, "t": 1061.7}
+{"event": "consolidate", "step": 3450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1076.2}
+{"event": "train", "step": 3450, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.52677, "bpb": 2.20266, "chunk_loss": 0.030574, "F": 0.2099, "G": 0.22104, "chunks": 214.94, "bytes_per_chunk": 4.764, "boundary_entropy": 0.09683, "router_noise": 0.0, "grad_norm": 0.7071, "lr": 0.0001428, "router_trainable": true, "sec_per_step": 0.297, "probe_sec": 0.1475, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.1355, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.10729, "n_shadow": 0, "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": 0.42529, "trunc_abs_frac": 0.84757, "mean_delta_full": -0.81047, "read_hit_rate": {"surface": 0.78265}, "mean_gate": {"surface": 0.40503}, "n_reads": 5358, "n_cand": 6846, "read_suppressed_frac": 0.109, "dropout_reads_logged": 5358, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 133.29808044433594, "n_util_obs": 177, "util_count_spearman": -0.07986794715939863}, "probe_overhead_sec_total": 217.215, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.061862, "alpha": 0.03, "t": 1076.5}
+{"event": "consolidate", "step": 3500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1089.7}
+{"event": "train", "step": 3500, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.40295, "bpb": 2.02402, "chunk_loss": 0.030489, "F": 0.20709, "G": 0.21717, "chunks": 212.06, "bytes_per_chunk": 4.829, "boundary_entropy": 0.0955, "router_noise": 0.0, "grad_norm": 0.6978, "lr": 0.0001391, "router_trainable": true, "sec_per_step": 0.271, "probe_sec": 0.0338, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02692, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.1968, "n_shadow": 0, "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": 1.52944, "trunc_abs_frac": 2.82918, "mean_delta_full": 0.29858, "read_hit_rate": {"surface": 0.81907}, "mean_gate": {"surface": 0.41394}, "n_reads": 5532, "n_cand": 6754, "read_suppressed_frac": 0.101, "dropout_reads_logged": 5532, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 135.17160034179688, "n_util_obs": 176, "util_count_spearman": -0.10021791767554479}, "probe_overhead_sec_total": 219.505, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.060325, "alpha": 0.03, "t": 1090.1}
+{"event": "consolidate", "step": 3550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1104.3}
+{"event": "train", "step": 3550, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.39294, "bpb": 2.00959, "chunk_loss": 0.030462, "F": 0.20789, "G": 0.21622, "chunks": 212.88, "bytes_per_chunk": 4.81, "boundary_entropy": 0.09344, "router_noise": 0.0, "grad_norm": 0.6351, "lr": 0.0001355, "router_trainable": true, "sec_per_step": 0.29, "probe_sec": 0.0325, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02483, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.1047, "n_shadow": 0, "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": 1.14672, "trunc_abs_frac": 0.65604, "mean_delta_full": -0.17139, "read_hit_rate": {"surface": 0.81431}, "mean_gate": {"surface": 0.41281}, "n_reads": 5521, "n_cand": 6780, "read_suppressed_frac": 0.0915, "dropout_reads_logged": 5521, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 137.31979370117188, "n_util_obs": 177, "util_count_spearman": -0.12082330638319974}, "probe_overhead_sec_total": 223.135, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.051639, "alpha": 0.03, "t": 1104.5}
+{"event": "consolidate", "step": 3600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1118.0}
+{"event": "probe", "step": 3600, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3603.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.207033, "cps": 28.148438, "mean_p": 0.207976, "boundary_entropy": 0.082872, "bpb": 1.916079}, "tr": {"chunks": 3309.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.182787, "cps": 25.851562, "mean_p": 0.213244, "boundary_entropy": 0.185091, "bpb": 2.128114}, "fi": {"chunks": 3509.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.18654, "cps": 27.414062, "mean_p": 0.206869, "boundary_entropy": 0.139345, "bpb": 2.2668}}, "bpb": {"en": 1.91608, "tr": 2.12811, "fi": 2.2668}, "gini_cpc": 0.028045, "gini_cps": 0.018808, "bpic": 5.444007, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.8828889346892468, "fi": 0.9010135928138491}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 139.08848571777344, "n_util_obs": 175, "util_count_spearman": -0.09811688311688312}, "t": 1118.8}
+{"event": "eval", "step": 3600, "phase": 2, "bpb": {"en": 1.93517, "tr": 2.1337, "fi": 2.27931}, "bytes_per_chunk": {"en": 4.624, "tr": 5.781, "fi": 5.336}, "bpb_mean": 2.11606, "frozen_inventory": true, "t": 1120.1}
+{"event": "train", "step": 3600, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.41699, "bpb": 2.04428, "chunk_loss": 0.030441, "F": 0.20367, "G": 0.21845, "chunks": 208.56, "bytes_per_chunk": 4.91, "boundary_entropy": 0.1113, "router_noise": 0.0, "grad_norm": 0.6758, "lr": 0.0001319, "router_trainable": true, "sec_per_step": 0.344, "probe_sec": 0.0281, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01846, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.03813, "n_shadow": 0, "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": 1.97097, "trunc_abs_frac": 0.87311, "mean_delta_full": 0.28452, "read_hit_rate": {"surface": 0.78997}, "mean_gate": {"surface": 0.40415}, "n_reads": 5247, "n_cand": 6642, "read_suppressed_frac": 0.1031, "dropout_reads_logged": 5247, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 139.1229248046875, "n_util_obs": 175, "util_count_spearman": -0.09814044902624076}, "probe_overhead_sec_total": 226.622, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.052799, "alpha": 0.03, "t": 1121.8}
+{"event": "consolidate", "step": 3650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1134.5}
+{"event": "train", "step": 3650, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.34258, "bpb": 1.93694, "chunk_loss": 0.030471, "F": 0.20883, "G": 0.21461, "chunks": 213.84, "bytes_per_chunk": 4.789, "boundary_entropy": 0.08621, "router_noise": 0.0, "grad_norm": 0.6465, "lr": 0.0001283, "router_trainable": true, "sec_per_step": 0.261, "probe_sec": 0.0545, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04499, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.2716, "n_shadow": 0, "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": 1.03447, "trunc_abs_frac": 1.05258, "mean_delta_full": 0.27285, "read_hit_rate": {"surface": 0.85024}, "mean_gate": {"surface": 0.41407}, "n_reads": 5791, "n_cand": 6811, "read_suppressed_frac": 0.0964, "dropout_reads_logged": 5791, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 140.9792022705078, "n_util_obs": 176, "util_count_spearman": -0.08675544794188862}, "probe_overhead_sec_total": 229.574, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.054885, "alpha": 0.03, "t": 1134.8}
+{"event": "consolidate", "step": 3700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1148.5}
+{"event": "train", "step": 3700, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.44972, "bpb": 2.0915, "chunk_loss": 0.030448, "F": 0.20322, "G": 0.21699, "chunks": 208.09, "bytes_per_chunk": 4.921, "boundary_entropy": 0.10547, "router_noise": 0.0, "grad_norm": 0.7272, "lr": 0.0001247, "router_trainable": true, "sec_per_step": 0.28, "probe_sec": 0.0613, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04506, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.2655, "n_shadow": 0, "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": 0.74855, "trunc_abs_frac": 0.67494, "mean_delta_full": 0.75936, "read_hit_rate": {"surface": 0.77893}, "mean_gate": {"surface": 0.39475}, "n_reads": 5162, "n_cand": 6627, "read_suppressed_frac": 0.0978, "dropout_reads_logged": 5162, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 142.87374877929688, "n_util_obs": 177, "util_count_spearman": -0.09092423953878914}, "probe_overhead_sec_total": 233.684, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.037139, "alpha": 0.03, "t": 1148.8}
+{"event": "consolidate", "step": 3750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1169.8}
+{"event": "train", "step": 3750, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.43512, "bpb": 2.07044, "chunk_loss": 0.030663, "F": 0.21106, "G": 0.22261, "chunks": 216.12, "bytes_per_chunk": 4.738, "boundary_entropy": 0.10107, "router_noise": 0.0, "grad_norm": 0.7436, "lr": 0.0001212, "router_trainable": true, "sec_per_step": 0.426, "probe_sec": 0.0515, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02531, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.28375, "n_shadow": 0, "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": {"surface": 0.79663}, "mean_gate": {"surface": 0.41064}, "n_reads": 5484, "n_cand": 6884, "read_suppressed_frac": 0.0985, "dropout_reads_logged": 5484, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 144.8294219970703, "n_util_obs": 176, "util_count_spearman": -0.09291877613911512}, "probe_overhead_sec_total": 237.127, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.034555, "alpha": 0.03, "t": 1170.1}
+{"event": "consolidate", "step": 3800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1191.9}
+{"event": "probe", "step": 3800, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3604.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.207091, "cps": 28.15625, "mean_p": 0.207879, "boundary_entropy": 0.078948, "bpb": 1.900594}, "tr": {"chunks": 3367.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.185991, "cps": 26.304688, "mean_p": 0.212809, "boundary_entropy": 0.176168, "bpb": 2.105797}, "fi": {"chunks": 3639.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.193451, "cps": 28.429688, "mean_p": 0.210799, "boundary_entropy": 0.13622, "bpb": 2.246675}}, "bpb": {"en": 1.90059, "tr": 2.1058, "fi": 2.24667}, "gini_cpc": 0.023982, "gini_cps": 0.017091, "bpic": 5.347031, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.8981149027426791, "fi": 0.9341347108933534}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 146.7423095703125, "n_util_obs": 177, "util_count_spearman": -0.09209934038538115}, "t": 1192.7}
+{"event": "eval", "step": 3800, "phase": 2, "bpb": {"en": 1.92248, "tr": 2.11307, "fi": 2.25788}, "bytes_per_chunk": {"en": 4.624, "tr": 5.755, "fi": 5.2}, "bpb_mean": 2.09781, "frozen_inventory": true, "t": 1193.7}
+{"event": "train", "step": 3800, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.32848, "bpb": 1.9166, "chunk_loss": 0.030523, "F": 0.21136, "G": 0.21716, "chunks": 216.44, "bytes_per_chunk": 4.731, "boundary_entropy": 0.08531, "router_noise": 0.0, "grad_norm": 0.6936, "lr": 0.0001177, "router_trainable": true, "sec_per_step": 0.519, "probe_sec": 0.0628, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02954, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.25815, "n_shadow": 0, "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": 0.39719, "trunc_abs_frac": 0.5838, "mean_delta_full": -0.2288, "read_hit_rate": {"surface": 0.84842}, "mean_gate": {"surface": 0.4248}, "n_reads": 5849, "n_cand": 6894, "read_suppressed_frac": 0.0898, "dropout_reads_logged": 5849, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 146.78602600097656, "n_util_obs": 177, "util_count_spearman": -0.09228009182240006}, "probe_overhead_sec_total": 240.461, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.036423, "alpha": 0.03, "t": 1196.1}
+{"event": "consolidate", "step": 3850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1219.0}
+{"event": "train", "step": 3850, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.34493, "bpb": 1.94032, "chunk_loss": 0.030553, "F": 0.21194, "G": 0.21707, "chunks": 217.03, "bytes_per_chunk": 4.718, "boundary_entropy": 0.08852, "router_noise": 0.0, "grad_norm": 0.6799, "lr": 0.0001142, "router_trainable": true, "sec_per_step": 0.463, "probe_sec": 0.0695, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05773, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.13858, "n_shadow": 0, "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": 0.84073, "trunc_abs_frac": 0.78074, "mean_delta_full": 0.52808, "read_hit_rate": {"surface": 0.83828}, "mean_gate": {"surface": 0.42024}, "n_reads": 5795, "n_cand": 6913, "read_suppressed_frac": 0.1023, "dropout_reads_logged": 5795, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 148.27456665039062, "n_util_obs": 180, "util_count_spearman": -0.09053365844624835}, "probe_overhead_sec_total": 244.25, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.042969, "alpha": 0.03, "t": 1219.2}
+{"event": "consolidate", "step": 3900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1234.1}
+{"event": "train", "step": 3900, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.41095, "bpb": 2.03557, "chunk_loss": 0.030448, "F": 0.20663, "G": 0.21684, "chunks": 211.59, "bytes_per_chunk": 4.839, "boundary_entropy": 0.09709, "router_noise": 0.0, "grad_norm": 0.6954, "lr": 0.0001108, "router_trainable": true, "sec_per_step": 0.303, "probe_sec": 0.1132, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10537, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.20497, "n_shadow": 0, "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": 1.41616, "trunc_abs_frac": 0.97901, "mean_delta_full": 0.19983, "read_hit_rate": {"surface": 0.81169}, "mean_gate": {"surface": 0.40986}, "n_reads": 5470, "n_cand": 6739, "read_suppressed_frac": 0.0967, "dropout_reads_logged": 5470, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 150.18377685546875, "n_util_obs": 178, "util_count_spearman": -0.09281658508249027}, "probe_overhead_sec_total": 247.651, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.046961, "alpha": 0.03, "t": 1234.4}
+{"event": "consolidate", "step": 3950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1248.5}
+{"event": "train", "step": 3950, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.40904, "bpb": 2.03282, "chunk_loss": 0.030517, "F": 0.21051, "G": 0.21589, "chunks": 215.56, "bytes_per_chunk": 4.75, "boundary_entropy": 0.08539, "router_noise": 0.0, "grad_norm": 0.7121, "lr": 0.0001074, "router_trainable": true, "sec_per_step": 0.298, "probe_sec": 0.044, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02548, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.19696, "n_shadow": 0, "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": 0.28296, "read_hit_rate": {"surface": 0.82057}, "mean_gate": {"surface": 0.40829}, "n_reads": 5634, "n_cand": 6866, "read_suppressed_frac": 0.1006, "dropout_reads_logged": 5634, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 151.90072631835938, "n_util_obs": 176, "util_count_spearman": -0.07514417785604226}, "probe_overhead_sec_total": 251.336, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.041798, "alpha": 0.03, "t": 1249.3}
+{"event": "consolidate", "step": 4000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1271.1}
+{"event": "probe", "step": 4000, "phase": 2, "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": 0.20858, "boundary_entropy": 0.078855, "bpb": 1.893541}, "tr": {"chunks": 3374.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.186378, "cps": 26.359375, "mean_p": 0.21171, "boundary_entropy": 0.173042, "bpb": 2.088253}, "fi": {"chunks": 3718.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.19765, "cps": 29.046875, "mean_p": 0.215246, "boundary_entropy": 0.137134, "bpb": 2.240499}}, "bpb": {"en": 1.89354, "tr": 2.08825, "fi": 2.2405}, "gini_cpc": 0.023922, "gini_cps": 0.021423, "bpic": 5.29958, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.8977402258830781, "fi": 0.9520366349649991}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 153.7180633544922, "n_util_obs": 173, "util_count_spearman": -0.08820475499580496}, "t": 1271.9}
+{"event": "eval", "step": 4000, "phase": 2, "bpb": {"en": 1.91336, "tr": 2.09405, "fi": 2.25065}, "bytes_per_chunk": {"en": 4.618, "tr": 5.716, "fi": 5.065}, "bpb_mean": 2.08602, "frozen_inventory": true, "t": 1273.0}
+{"event": "train", "step": 4000, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.35221, "bpb": 1.95082, "chunk_loss": 0.030427, "F": 0.20392, "G": 0.21559, "chunks": 208.81, "bytes_per_chunk": 4.904, "boundary_entropy": 0.1014, "router_noise": 0.0, "grad_norm": 0.6775, "lr": 0.000104, "router_trainable": true, "sec_per_step": 0.514, "probe_sec": 0.107, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09216, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.54646, "n_shadow": 0, "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": 0.45741, "trunc_abs_frac": 15.04691, "mean_delta_full": -0.25653, "read_hit_rate": {"surface": 0.82211}, "mean_gate": {"surface": 0.40546}, "n_reads": 5467, "n_cand": 6650, "read_suppressed_frac": 0.0924, "dropout_reads_logged": 5467, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 153.77635192871094, "n_util_obs": 173, "util_count_spearman": -0.08791509328464543}, "probe_overhead_sec_total": 254.388, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.050524, "alpha": 0.03, "t": 1275.0}
+{"event": "consolidate", "step": 4050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1293.3}
+{"event": "train", "step": 4050, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.4811, "bpb": 2.13677, "chunk_loss": 0.030459, "F": 0.2037, "G": 0.21572, "chunks": 208.59, "bytes_per_chunk": 4.909, "boundary_entropy": 0.10052, "router_noise": 0.0, "grad_norm": 0.6937, "lr": 0.0001007, "router_trainable": true, "sec_per_step": 0.372, "probe_sec": 0.0347, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02437, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.01825, "n_shadow": 0, "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": 1.21071, "trunc_abs_frac": 0.56219, "mean_delta_full": -0.15497, "read_hit_rate": {"surface": 0.77314}, "mean_gate": {"surface": 0.39364}, "n_reads": 5136, "n_cand": 6643, "read_suppressed_frac": 0.1014, "dropout_reads_logged": 5136, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 155.08230590820312, "n_util_obs": 175, "util_count_spearman": -0.041618898343036274}, "probe_overhead_sec_total": 257.553, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.047486, "alpha": 0.03, "t": 1293.6}
+{"event": "probe", "step": 4096, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3588.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.206171, "cps": 28.03125, "mean_p": 0.206813, "boundary_entropy": 0.078003, "bpb": 1.885251}, "tr": {"chunks": 3454.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.190797, "cps": 26.984375, "mean_p": 0.213429, "boundary_entropy": 0.174376, "bpb": 2.080619}, "fi": {"chunks": 3658.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.194461, "cps": 28.578125, "mean_p": 0.211961, "boundary_entropy": 0.13519, "bpb": 2.226228}}, "bpb": {"en": 1.88525, "tr": 2.08062, "fi": 2.22623}, "gini_cpc": 0.01733, "gini_cps": 0.01271, "bpic": 5.302056, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9254297731838088, "fi": 0.9431993733119579}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 156.80892944335938, "n_util_obs": 172, "util_count_spearman": -0.10332116809846284}, "t": 1307.2}
+{"event": "consolidate", "step": 4100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1310.9}
+{"event": "train", "step": 4100, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.42282, "bpb": 2.0527, "chunk_loss": 0.030507, "F": 0.20847, "G": 0.22045, "chunks": 213.47, "bytes_per_chunk": 4.797, "boundary_entropy": 0.10259, "router_noise": 0.0, "grad_norm": 0.6906, "lr": 9.75e-05, "router_trainable": true, "sec_per_step": 0.355, "probe_sec": 0.0404, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03048, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.57401, "n_shadow": 0, "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": 0.47897, "trunc_abs_frac": 0.26523, "mean_delta_full": -0.30782, "read_hit_rate": {"surface": 0.78982}, "mean_gate": {"surface": 0.39456}, "n_reads": 5370, "n_cand": 6799, "read_suppressed_frac": 0.0987, "dropout_reads_logged": 5370, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 156.94723510742188, "n_util_obs": 171, "util_count_spearman": -0.09589969034230872}, "probe_overhead_sec_total": 260.885, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.051982, "alpha": 0.03, "t": 1311.3}
+{"event": "consolidate", "step": 4150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1324.2}
+{"event": "train", "step": 4150, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.3611, "bpb": 1.96365, "chunk_loss": 0.030503, "F": 0.20923, "G": 0.2152, "chunks": 214.25, "bytes_per_chunk": 4.779, "boundary_entropy": 0.08703, "router_noise": 0.0, "grad_norm": 0.658, "lr": 9.43e-05, "router_trainable": true, "sec_per_step": 0.262, "probe_sec": 0.0763, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06833, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.00592, "n_shadow": 0, "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": 1.42731, "trunc_abs_frac": 0.74551, "mean_delta_full": 0.05192, "read_hit_rate": {"surface": 0.83133}, "mean_gate": {"surface": 0.41489}, "n_reads": 5673, "n_cand": 6824, "read_suppressed_frac": 0.1042, "dropout_reads_logged": 5673, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 158.7308349609375, "n_util_obs": 170, "util_count_spearman": -0.0842947954641215}, "probe_overhead_sec_total": 264.079, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.042128, "alpha": 0.03, "t": 1324.4}
+{"event": "consolidate", "step": 4200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1337.0}
+{"event": "probe", "step": 4200, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3591.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.206344, "cps": 28.054688, "mean_p": 0.206661, "boundary_entropy": 0.077835, "bpb": 1.882088}, "tr": {"chunks": 3311.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.182898, "cps": 25.867188, "mean_p": 0.209078, "boundary_entropy": 0.17144, "bpb": 2.069432}, "fi": {"chunks": 3596.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.191165, "cps": 28.09375, "mean_p": 0.208487, "boundary_entropy": 0.13265, "bpb": 2.220252}}, "bpb": {"en": 1.88209, "tr": 2.06943, "fi": 2.22025}, "gini_cpc": 0.02693, "gini_cps": 0.018099, "bpic": 5.404077, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.8863746857246045, "fi": 0.9264383292681406}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 160.25880432128906, "n_util_obs": 169, "util_count_spearman": -0.08087116503406096}, "t": 1337.6}
+{"event": "eval", "step": 4200, "phase": 2, "bpb": {"en": 1.9006, "tr": 2.07904, "fi": 2.22867}, "bytes_per_chunk": {"en": 4.634, "tr": 5.828, "fi": 5.191}, "bpb_mean": 2.06944, "frozen_inventory": true, "t": 1339.0}
+{"event": "phase_transition", "step": 4200, "phase": 3, "objective": "global", "router_trainable": true, "encoder_trainable": true, "frozen_router_params": 0, "t": 1340.2}
+{"event": "train", "step": 4200, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.43715, "bpb": 2.07337, "chunk_loss": 0.030536, "F": 0.21149, "G": 0.21921, "chunks": 216.56, "bytes_per_chunk": 4.728, "boundary_entropy": 0.09113, "router_noise": 0.0, "grad_norm": 0.6843, "lr": 9.12e-05, "router_trainable": true, "sec_per_step": 0.32, "probe_sec": 0.0476, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03436, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.75375, "n_shadow": 0, "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": 0.9113, "trunc_abs_frac": 0.93283, "mean_delta_full": 0.6788, "read_hit_rate": {"surface": 0.81052}, "mean_gate": {"surface": 0.40353}, "n_reads": 5591, "n_cand": 6898, "read_suppressed_frac": 0.1002, "dropout_reads_logged": 5591, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 160.2967987060547, "n_util_obs": 169, "util_count_spearman": -0.08006812192332555}, "probe_overhead_sec_total": 266.426, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.046948, "alpha": 0.03, "t": 1340.4}
+{"event": "consolidate", "step": 4250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1354.3}
+{"event": "train", "step": 4250, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32643, "bpb": 1.91363, "chunk_loss": 0.030439, "F": 0.20703, "G": 0.21285, "chunks": 212.0, "bytes_per_chunk": 4.83, "boundary_entropy": 0.08508, "router_noise": 0.0, "grad_norm": 0.6438, "lr": 8.81e-05, "router_trainable": true, "sec_per_step": 0.283, "probe_sec": 0.0419, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03381, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.31099, "n_shadow": 0, "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": 2.43183, "trunc_abs_frac": 27.47202, "mean_delta_full": 0.05115, "read_hit_rate": {"surface": 0.84108}, "mean_gate": {"surface": 0.41478}, "n_reads": 5679, "n_cand": 6752, "read_suppressed_frac": 0.1018, "dropout_reads_logged": 5679, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 162.34547424316406, "n_util_obs": 168, "util_count_spearman": -0.12260770091262951}, "probe_overhead_sec_total": 269.681, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.052313, "alpha": 0.03, "t": 1354.6}
+{"event": "consolidate", "step": 4300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1367.5}
+{"event": "train", "step": 4300, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.33693, "bpb": 1.92879, "chunk_loss": 0.030359, "F": 0.19897, "G": 0.21321, "chunks": 203.75, "bytes_per_chunk": 5.026, "boundary_entropy": 0.10353, "router_noise": 0.0, "grad_norm": 0.7086, "lr": 8.5e-05, "router_trainable": true, "sec_per_step": 0.265, "probe_sec": 0.0373, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02732, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.10627, "n_shadow": 0, "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": -0.51494, "trunc_abs_frac": 2.24233, "mean_delta_full": 0.30908, "read_hit_rate": {"surface": 0.80009}, "mean_gate": {"surface": 0.39521}, "n_reads": 5191, "n_cand": 6488, "read_suppressed_frac": 0.0988, "dropout_reads_logged": 5191, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 163.80197143554688, "n_util_obs": 173, "util_count_spearman": -0.08101876933472288}, "probe_overhead_sec_total": 272.972, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.053654, "alpha": 0.03, "t": 1367.8}
+{"event": "consolidate", "step": 4350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1379.9}
+{"event": "train", "step": 4350, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.31558, "bpb": 1.89798, "chunk_loss": 0.030425, "F": 0.20425, "G": 0.21393, "chunks": 209.16, "bytes_per_chunk": 4.896, "boundary_entropy": 0.09362, "router_noise": 0.0, "grad_norm": 0.7189, "lr": 8.21e-05, "router_trainable": true, "sec_per_step": 0.248, "probe_sec": 0.1452, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.13839, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.20668, "n_shadow": 0, "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": 0.5379, "trunc_abs_frac": 0.55377, "mean_delta_full": 0.37003, "read_hit_rate": {"surface": 0.82}, "mean_gate": {"surface": 0.41021}, "n_reads": 5462, "n_cand": 6661, "read_suppressed_frac": 0.1022, "dropout_reads_logged": 5462, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 165.7232208251953, "n_util_obs": 176, "util_count_spearman": -0.04898525203609949}, "probe_overhead_sec_total": 277.028, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.042093, "alpha": 0.03, "t": 1380.2}
+{"event": "consolidate", "step": 4400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1393.7}
+{"event": "probe", "step": 4400, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3596.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.206631, "cps": 28.09375, "mean_p": 0.207221, "boundary_entropy": 0.075077, "bpb": 1.87206}, "tr": {"chunks": 3334.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.184168, "cps": 26.046875, "mean_p": 0.207687, "boundary_entropy": 0.165782, "bpb": 2.055491}, "fi": {"chunks": 3600.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.191377, "cps": 28.125, "mean_p": 0.2084, "boundary_entropy": 0.130141, "bpb": 2.204075}}, "bpb": {"en": 1.87206, "tr": 2.05549, "fi": 2.20408}, "gini_cpc": 0.025723, "gini_cps": 0.016841, "bpic": 5.387654, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.8912909179870936, "fi": 0.9261792661608014}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 167.6566925048828, "n_util_obs": 177, "util_count_spearman": -0.027284846176485866}, "t": 1394.6}
+{"event": "eval", "step": 4400, "phase": 3, "bpb": {"en": 1.89273, "tr": 2.06299, "fi": 2.21427}, "bytes_per_chunk": {"en": 4.627, "tr": 5.806, "fi": 5.197}, "bpb_mean": 2.05666, "frozen_inventory": true, "t": 1395.7}
+{"event": "train", "step": 4400, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35212, "bpb": 1.9507, "chunk_loss": 0.030528, "F": 0.2131, "G": 0.21652, "chunks": 218.22, "bytes_per_chunk": 4.693, "boundary_entropy": 0.07869, "router_noise": 0.0, "grad_norm": 0.7602, "lr": 7.92e-05, "router_trainable": true, "sec_per_step": 0.341, "probe_sec": 0.0463, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03652, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.28627, "n_shadow": 0, "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": 50.89474, "trunc_abs_frac": 1.37992, "mean_delta_full": 0.00696, "read_hit_rate": {"surface": 0.8311}, "mean_gate": {"surface": 0.41576}, "n_reads": 5777, "n_cand": 6951, "read_suppressed_frac": 0.0949, "dropout_reads_logged": 5777, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 167.71937561035156, "n_util_obs": 177, "util_count_spearman": -0.027688463933531828}, "probe_overhead_sec_total": 280.159, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.040156, "alpha": 0.03, "t": 1397.3}
+{"event": "consolidate", "step": 4450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1411.0}
+{"event": "train", "step": 4450, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.37904, "bpb": 1.98953, "chunk_loss": 0.030372, "F": 0.20193, "G": 0.21029, "chunks": 206.78, "bytes_per_chunk": 4.952, "boundary_entropy": 0.09075, "router_noise": 0.0, "grad_norm": 0.6732, "lr": 7.63e-05, "router_trainable": true, "sec_per_step": 0.28, "probe_sec": 0.0889, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07907, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.25081, "n_shadow": 0, "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": 0.82516, "trunc_abs_frac": 0.85974, "mean_delta_full": 0.98739, "read_hit_rate": {"surface": 0.81002}, "mean_gate": {"surface": 0.40932}, "n_reads": 5334, "n_cand": 6585, "read_suppressed_frac": 0.0986, "dropout_reads_logged": 5334, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 169.83853149414062, "n_util_obs": 175, "util_count_spearman": -0.06993734322512472}, "probe_overhead_sec_total": 283.524, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.049885, "alpha": 0.03, "t": 1411.3}
+{"event": "consolidate", "step": 4500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1423.9}
+{"event": "train", "step": 4500, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.39633, "bpb": 2.01447, "chunk_loss": 0.030481, "F": 0.20844, "G": 0.21551, "chunks": 213.44, "bytes_per_chunk": 4.798, "boundary_entropy": 0.09137, "router_noise": 0.0, "grad_norm": 0.682, "lr": 7.36e-05, "router_trainable": true, "sec_per_step": 0.259, "probe_sec": 0.1091, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09631, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.11874, "n_shadow": 0, "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": 0.67672, "trunc_abs_frac": 0.74567, "mean_delta_full": 0.62457, "read_hit_rate": {"surface": 0.81142}, "mean_gate": {"surface": 0.39752}, "n_reads": 5516, "n_cand": 6798, "read_suppressed_frac": 0.0986, "dropout_reads_logged": 5516, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 170.12306213378906, "n_util_obs": 174, "util_count_spearman": -0.062157806315669584}, "probe_overhead_sec_total": 286.573, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.040791, "alpha": 0.03, "t": 1424.2}
+{"event": "consolidate", "step": 4550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1437.1}
+{"event": "train", "step": 4550, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.31323, "bpb": 1.89459, "chunk_loss": 0.030366, "F": 0.20007, "G": 0.21003, "chunks": 204.88, "bytes_per_chunk": 4.998, "boundary_entropy": 0.09366, "router_noise": 0.0, "grad_norm": 0.6556, "lr": 7.09e-05, "router_trainable": true, "sec_per_step": 0.263, "probe_sec": 0.055, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04681, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.24585, "n_shadow": 0, "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": 0.51299, "trunc_abs_frac": 0.69356, "mean_delta_full": 0.26196, "read_hit_rate": {"surface": 0.81836}, "mean_gate": {"surface": 0.40414}, "n_reads": 5339, "n_cand": 6524, "read_suppressed_frac": 0.101, "dropout_reads_logged": 5339, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 172.3208770751953, "n_util_obs": 175, "util_count_spearman": -0.0895947156291984}, "probe_overhead_sec_total": 289.674, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.042812, "alpha": 0.03, "t": 1437.4}
+{"event": "consolidate", "step": 4600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1450.5}
+{"event": "probe", "step": 4600, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3594.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.206516, "cps": 28.078125, "mean_p": 0.206139, "boundary_entropy": 0.075472, "bpb": 1.859555}, "tr": {"chunks": 3339.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.184445, "cps": 26.085938, "mean_p": 0.207352, "boundary_entropy": 0.165165, "bpb": 2.04778}, "fi": {"chunks": 3681.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.195683, "cps": 28.757812, "mean_p": 0.211749, "boundary_entropy": 0.131623, "bpb": 2.195512}}, "bpb": {"en": 1.85955, "tr": 2.04778, "fi": 2.19551}, "gini_cpc": 0.025082, "gini_cps": 0.021481, "bpic": 5.345016, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.893124319132119, "fi": 0.9475452992596862}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 173.8086700439453, "n_util_obs": 175, "util_count_spearman": -0.08604567845947156}, "t": 1451.3}
+{"event": "eval", "step": 4600, "phase": 3, "bpb": {"en": 1.87992, "tr": 2.0565, "fi": 2.20474}, "bytes_per_chunk": {"en": 4.633, "tr": 5.785, "fi": 5.072}, "bpb_mean": 2.04706, "frozen_inventory": true, "t": 1452.6}
+{"event": "train", "step": 4600, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.37433, "bpb": 1.98273, "chunk_loss": 0.030443, "F": 0.20651, "G": 0.2127, "chunks": 211.47, "bytes_per_chunk": 4.842, "boundary_entropy": 0.08757, "router_noise": 0.0, "grad_norm": 0.702, "lr": 6.82e-05, "router_trainable": true, "sec_per_step": 0.331, "probe_sec": 0.1203, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10769, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.12603, "n_shadow": 0, "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": 0.33727, "trunc_abs_frac": 0.57531, "mean_delta_full": 0.30319, "read_hit_rate": {"surface": 0.81425}, "mean_gate": {"surface": 0.40702}, "n_reads": 5484, "n_cand": 6735, "read_suppressed_frac": 0.1081, "dropout_reads_logged": 5484, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 173.84300231933594, "n_util_obs": 175, "util_count_spearman": -0.08651141961486788}, "probe_overhead_sec_total": 292.829, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.072611, "alpha": 0.03, "t": 1453.9}
+{"event": "consolidate", "step": 4650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1467.2}
+{"event": "train", "step": 4650, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.37896, "bpb": 1.98942, "chunk_loss": 0.03052, "F": 0.20999, "G": 0.2176, "chunks": 215.03, "bytes_per_chunk": 4.762, "boundary_entropy": 0.09231, "router_noise": 0.0, "grad_norm": 0.7203, "lr": 6.57e-05, "router_trainable": true, "sec_per_step": 0.269, "probe_sec": 0.0382, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03154, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.19158, "n_shadow": 0, "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": 0.55221, "trunc_abs_frac": 1.37589, "mean_delta_full": 0.23872, "read_hit_rate": {"surface": 0.82611}, "mean_gate": {"surface": 0.40651}, "n_reads": 5658, "n_cand": 6849, "read_suppressed_frac": 0.0976, "dropout_reads_logged": 5658, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 174.92535400390625, "n_util_obs": 171, "util_count_spearman": -0.05175558595531236}, "probe_overhead_sec_total": 295.67, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.071479, "alpha": 0.03, "t": 1467.4}
+{"event": "consolidate", "step": 4700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1481.9}
+{"event": "train", "step": 4700, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35519, "bpb": 1.95513, "chunk_loss": 0.030433, "F": 0.20447, "G": 0.21528, "chunks": 209.38, "bytes_per_chunk": 4.891, "boundary_entropy": 0.09754, "router_noise": 0.0, "grad_norm": 0.7093, "lr": 6.32e-05, "router_trainable": true, "sec_per_step": 0.295, "probe_sec": 0.0344, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02807, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.07306, "n_shadow": 0, "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": 0.99305, "trunc_abs_frac": 0.8437, "mean_delta_full": -0.20959, "read_hit_rate": {"surface": 0.80924}, "mean_gate": {"surface": 0.39907}, "n_reads": 5396, "n_cand": 6668, "read_suppressed_frac": 0.1006, "dropout_reads_logged": 5396, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 176.3004913330078, "n_util_obs": 170, "util_count_spearman": -0.05122953572584437}, "probe_overhead_sec_total": 299.057, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.07663, "alpha": 0.03, "t": 1482.1}
+{"event": "consolidate", "step": 4750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1495.1}
+{"event": "train", "step": 4750, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.4494, "bpb": 2.09104, "chunk_loss": 0.030411, "F": 0.20334, "G": 0.21322, "chunks": 208.22, "bytes_per_chunk": 4.918, "boundary_entropy": 0.09531, "router_noise": 0.0, "grad_norm": 0.7999, "lr": 6.08e-05, "router_trainable": true, "sec_per_step": 0.262, "probe_sec": 0.0343, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02714, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.3131, "n_shadow": 0, "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": 0.26566, "trunc_abs_frac": 0.37659, "mean_delta_full": -0.21045, "read_hit_rate": {"surface": 0.78404}, "mean_gate": {"surface": 0.38838}, "n_reads": 5199, "n_cand": 6631, "read_suppressed_frac": 0.1164, "dropout_reads_logged": 5199, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 178.40264892578125, "n_util_obs": 171, "util_count_spearman": -0.04570763433892529}, "probe_overhead_sec_total": 302.089, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.047615, "alpha": 0.03, "t": 1495.3}
+{"event": "consolidate", "step": 4800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1508.6}
+{"event": "probe", "step": 4800, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3592.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.206401, "cps": 28.0625, "mean_p": 0.206242, "boundary_entropy": 0.073292, "bpb": 1.851146}, "tr": {"chunks": 3268.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.180523, "cps": 25.53125, "mean_p": 0.205178, "boundary_entropy": 0.162717, "bpb": 2.034109}, "fi": {"chunks": 3584.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.190527, "cps": 28.0, "mean_p": 0.206285, "boundary_entropy": 0.127249, "bpb": 2.18098}}, "bpb": {"en": 1.85115, "tr": 2.03411, "fi": 2.18098}, "gini_cpc": 0.029877, "gini_cps": 0.020682, "bpic": 5.432018, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.8746197673372867, "fi": 0.9230897099846451}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 179.9802703857422, "n_util_obs": 172, "util_count_spearman": -0.035571183638623376}, "t": 1509.3}
+{"event": "eval", "step": 4800, "phase": 3, "bpb": {"en": 1.87118, "tr": 2.04176, "fi": 2.19041}, "bytes_per_chunk": {"en": 4.636, "tr": 5.865, "fi": 5.21}, "bpb_mean": 2.03445, "frozen_inventory": true, "t": 1510.7}
+{"event": "train", "step": 4800, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32427, "bpb": 1.91052, "chunk_loss": 0.030396, "F": 0.20572, "G": 0.21042, "chunks": 210.66, "bytes_per_chunk": 4.861, "boundary_entropy": 0.08038, "router_noise": 0.0, "grad_norm": 0.6911, "lr": 5.85e-05, "router_trainable": true, "sec_per_step": 0.335, "probe_sec": 0.0469, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03878, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.06287, "n_shadow": 0, "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": 0.70789, "trunc_abs_frac": 0.73007, "mean_delta_full": -0.43848, "read_hit_rate": {"surface": 0.83276}, "mean_gate": {"surface": 0.41642}, "n_reads": 5587, "n_cand": 6709, "read_suppressed_frac": 0.102, "dropout_reads_logged": 5587, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 180.02587890625, "n_util_obs": 172, "util_count_spearman": -0.035571183638623376}, "probe_overhead_sec_total": 304.929, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.042784, "alpha": 0.03, "t": 1512.0}
+{"event": "consolidate", "step": 4850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1525.6}
+{"event": "train", "step": 4850, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32051, "bpb": 1.9051, "chunk_loss": 0.030593, "F": 0.21432, "G": 0.22005, "chunks": 219.47, "bytes_per_chunk": 4.666, "boundary_entropy": 0.0874, "router_noise": 0.0, "grad_norm": 0.6623, "lr": 5.62e-05, "router_trainable": true, "sec_per_step": 0.279, "probe_sec": 0.038, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02547, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.15084, "n_shadow": 0, "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": 0.40432, "trunc_abs_frac": 0.63263, "mean_delta_full": 0.70459, "read_hit_rate": {"surface": 0.83808}, "mean_gate": {"surface": 0.40404}, "n_reads": 5859, "n_cand": 6991, "read_suppressed_frac": 0.1021, "dropout_reads_logged": 5859, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 181.94775390625, "n_util_obs": 171, "util_count_spearman": -0.042458060335517314}, "probe_overhead_sec_total": 308.409, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.040304, "alpha": 0.03, "t": 1526.0}
+{"event": "consolidate", "step": 4900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1539.2}
+{"event": "train", "step": 4900, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.37621, "bpb": 1.98546, "chunk_loss": 0.030431, "F": 0.20578, "G": 0.21336, "chunks": 210.72, "bytes_per_chunk": 4.86, "boundary_entropy": 0.09159, "router_noise": 0.0, "grad_norm": 0.7997, "lr": 5.41e-05, "router_trainable": true, "sec_per_step": 0.27, "probe_sec": 0.1026, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09485, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.22602, "n_shadow": 0, "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": 2.09395, "trunc_abs_frac": 1.98266, "mean_delta_full": 0.20667, "read_hit_rate": {"surface": 0.81955}, "mean_gate": {"surface": 0.39616}, "n_reads": 5500, "n_cand": 6711, "read_suppressed_frac": 0.1009, "dropout_reads_logged": 5500, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 184.08984375, "n_util_obs": 174, "util_count_spearman": -0.06113272017995956}, "probe_overhead_sec_total": 312.027, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.036592, "alpha": 0.03, "t": 1539.4}
+{"event": "consolidate", "step": 4950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1552.4}
+{"event": "train", "step": 4950, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.3116, "bpb": 1.89224, "chunk_loss": 0.03053, "F": 0.20432, "G": 0.21478, "chunks": 209.22, "bytes_per_chunk": 4.894, "boundary_entropy": 0.0989, "router_noise": 0.0, "grad_norm": 0.6968, "lr": 5.2e-05, "router_trainable": true, "sec_per_step": 0.263, "probe_sec": 0.0369, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03108, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.09984, "n_shadow": 0, "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": 1.09933, "trunc_abs_frac": 2.37096, "mean_delta_full": 0.15198, "read_hit_rate": {"surface": 0.81105}, "mean_gate": {"surface": 0.40215}, "n_reads": 5404, "n_cand": 6663, "read_suppressed_frac": 0.1005, "dropout_reads_logged": 5404, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 185.96728515625, "n_util_obs": 172, "util_count_spearman": -0.0034691514375399446}, "probe_overhead_sec_total": 315.682, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.03317, "alpha": 0.03, "t": 1552.6}
+{"event": "consolidate", "step": 5000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1564.4}
+{"event": "probe", "step": 5000, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3593.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.206459, "cps": 28.070312, "mean_p": 0.205467, "boundary_entropy": 0.075934, "bpb": 1.84066}, "tr": {"chunks": 3307.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.182677, "cps": 25.835938, "mean_p": 0.206366, "boundary_entropy": 0.162668, "bpb": 2.023939}, "fi": {"chunks": 3633.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.193132, "cps": 28.382812, "mean_p": 0.208116, "boundary_entropy": 0.126553, "bpb": 2.168334}}, "bpb": {"en": 1.84066, "tr": 2.02394, "fi": 2.16833}, "gini_cpc": 0.027229, "gini_cps": 0.020634, "bpic": 5.38612, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.8848110678913602, "fi": 0.9354496512638805}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 187.71942138671875, "n_util_obs": 171, "util_count_spearman": -0.006043151654786761}, "t": 1565.1}
+{"event": "eval", "step": 5000, "phase": 3, "bpb": {"en": 1.86083, "tr": 2.0313, "fi": 2.17816}, "bytes_per_chunk": {"en": 4.632, "tr": 5.853, "fi": 5.147}, "bpb_mean": 2.02343, "frozen_inventory": true, "t": 1566.2}
+{"event": "train", "step": 5000, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.41919, "bpb": 2.04746, "chunk_loss": 0.030468, "F": 0.20703, "G": 0.21612, "chunks": 212.0, "bytes_per_chunk": 4.83, "boundary_entropy": 0.09624, "router_noise": 0.0, "grad_norm": 0.7301, "lr": 5e-05, "router_trainable": true, "sec_per_step": 0.3, "probe_sec": 0.0334, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02627, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.11418, "n_shadow": 0, "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": 1.72923, "trunc_abs_frac": 0.76342, "mean_delta_full": 0.13992, "read_hit_rate": {"surface": 0.78495}, "mean_gate": {"surface": 0.38206}, "n_reads": 5300, "n_cand": 6752, "read_suppressed_frac": 0.1008, "dropout_reads_logged": 5300, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 187.77005004882812, "n_util_obs": 171, "util_count_spearman": -0.006095951232390141}, "probe_overhead_sec_total": 318.684, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.155903, "alpha": 0.03, "t": 1567.6}
+{"event": "consolidate", "step": 5050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1582.7}
+{"event": "train", "step": 5050, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.2713, "bpb": 1.83409, "chunk_loss": 0.030433, "F": 0.20728, "G": 0.21329, "chunks": 212.25, "bytes_per_chunk": 4.824, "boundary_entropy": 0.08842, "router_noise": 0.0, "grad_norm": 0.6829, "lr": 4.81e-05, "router_trainable": true, "sec_per_step": 0.307, "probe_sec": 0.0424, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03499, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.26348, "n_shadow": 0, "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": 0.4272, "trunc_abs_frac": 0.62881, "mean_delta_full": 0.82806, "read_hit_rate": {"surface": 0.84216}, "mean_gate": {"surface": 0.41039}, "n_reads": 5693, "n_cand": 6760, "read_suppressed_frac": 0.1082, "dropout_reads_logged": 5693, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 189.6731719970703, "n_util_obs": 170, "util_count_spearman": -0.03127850953523733}, "probe_overhead_sec_total": 322.246, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.156475, "alpha": 0.03, "t": 1582.9}
+{"event": "consolidate", "step": 5100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1598.9}
+{"event": "train", "step": 5100, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32594, "bpb": 1.91292, "chunk_loss": 0.030342, "F": 0.2016, "G": 0.20831, "chunks": 206.44, "bytes_per_chunk": 4.96, "boundary_entropy": 0.08769, "router_noise": 0.0, "grad_norm": 0.684, "lr": 4.63e-05, "router_trainable": true, "sec_per_step": 0.324, "probe_sec": 0.0785, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07097, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.18314, "n_shadow": 0, "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": 1.0137, "trunc_abs_frac": 1.13446, "mean_delta_full": 0.70624, "read_hit_rate": {"surface": 0.82492}, "mean_gate": {"surface": 0.4005}, "n_reads": 5423, "n_cand": 6574, "read_suppressed_frac": 0.0955, "dropout_reads_logged": 5423, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 191.06512451171875, "n_util_obs": 167, "util_count_spearman": -0.026330299813453986}, "probe_overhead_sec_total": 325.696, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.150495, "alpha": 0.03, "t": 1599.1}
+{"event": "consolidate", "step": 5150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1613.4}
+{"event": "train", "step": 5150, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.36031, "bpb": 1.96251, "chunk_loss": 0.030483, "F": 0.20789, "G": 0.21449, "chunks": 212.88, "bytes_per_chunk": 4.81, "boundary_entropy": 0.09222, "router_noise": 0.0, "grad_norm": 0.7458, "lr": 4.45e-05, "router_trainable": true, "sec_per_step": 0.29, "probe_sec": 0.0616, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05349, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.21909, "n_shadow": 0, "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": 0.65154, "trunc_abs_frac": 2.09829, "mean_delta_full": -0.28402, "read_hit_rate": {"surface": 0.81608}, "mean_gate": {"surface": 0.39156}, "n_reads": 5533, "n_cand": 6780, "read_suppressed_frac": 0.0958, "dropout_reads_logged": 5533, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 193.27108764648438, "n_util_obs": 171, "util_count_spearman": -0.03260373917008664}, "probe_overhead_sec_total": 329.599, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.154477, "alpha": 0.03, "t": 1613.7}
+{"event": "consolidate", "step": 5200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1640.7}
+{"event": "probe", "step": 5200, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3598.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.206746, "cps": 28.109375, "mean_p": 0.205084, "boundary_entropy": 0.07629, "bpb": 1.835387}, "tr": {"chunks": 3313.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.183008, "cps": 25.882812, "mean_p": 0.206257, "boundary_entropy": 0.160931, "bpb": 2.011723}, "fi": {"chunks": 3615.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.192175, "cps": 28.242188, "mean_p": 0.206336, "boundary_entropy": 0.1239, "bpb": 2.16167}}, "bpb": {"en": 1.83539, "tr": 2.01172, "fi": 2.16167}, "gini_cpc": 0.027194, "gini_cps": 0.019127, "bpic": 5.389702, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.8851845917700816, "fi": 0.9295213712577962}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 195.16024780273438, "n_util_obs": 173, "util_count_spearman": -0.04566798466608878}, "t": 1641.3}
+{"event": "eval", "step": 5200, "phase": 3, "bpb": {"en": 1.85583, "tr": 2.02012, "fi": 2.17076}, "bytes_per_chunk": {"en": 4.629, "tr": 5.819, "fi": 5.161}, "bpb_mean": 2.01557, "frozen_inventory": true, "t": 1642.3}
+{"event": "train", "step": 5200, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.30003, "bpb": 1.87555, "chunk_loss": 0.03051, "F": 0.21069, "G": 0.21484, "chunks": 215.75, "bytes_per_chunk": 4.746, "boundary_entropy": 0.08679, "router_noise": 0.0, "grad_norm": 0.6978, "lr": 4.29e-05, "router_trainable": true, "sec_per_step": 0.634, "probe_sec": 0.0524, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03464, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.049, "n_shadow": 0, "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": 0.56984, "trunc_abs_frac": 5.97631, "mean_delta_full": -0.12833, "read_hit_rate": {"surface": 0.83818}, "mean_gate": {"surface": 0.40094}, "n_reads": 5760, "n_cand": 6872, "read_suppressed_frac": 0.0943, "dropout_reads_logged": 5760, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 195.22406005859375, "n_util_obs": 173, "util_count_spearman": -0.045633219147920324}, "probe_overhead_sec_total": 332.98, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.153494, "alpha": 0.03, "t": 1645.3}
+{"event": "consolidate", "step": 5250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1688.8}
+{"event": "train", "step": 5250, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.29966, "bpb": 1.87501, "chunk_loss": 0.030433, "F": 0.2059, "G": 0.21306, "chunks": 210.84, "bytes_per_chunk": 4.857, "boundary_entropy": 0.09195, "router_noise": 0.0, "grad_norm": 0.6685, "lr": 4.14e-05, "router_trainable": true, "sec_per_step": 0.892, "probe_sec": 0.0672, "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": 0.50766, "n_shadow": 0, "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": 0.33413, "trunc_abs_frac": 14.67922, "mean_delta_full": 0.5741, "read_hit_rate": {"surface": 0.82264}, "mean_gate": {"surface": 0.39161}, "n_reads": 5524, "n_cand": 6715, "read_suppressed_frac": 0.1104, "dropout_reads_logged": 5524, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 197.24386596679688, "n_util_obs": 180, "util_count_spearman": -0.008950893546097101}, "probe_overhead_sec_total": 336.026, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.156573, "alpha": 0.03, "t": 1689.9}
+{"event": "consolidate", "step": 5300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1730.0}
+{"event": "train", "step": 5300, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.2942, "bpb": 1.86714, "chunk_loss": 0.030435, "F": 0.20859, "G": 0.21187, "chunks": 213.59, "bytes_per_chunk": 4.794, "boundary_entropy": 0.0849, "router_noise": 0.0, "grad_norm": 0.6542, "lr": 3.99e-05, "router_trainable": true, "sec_per_step": 0.874, "probe_sec": 0.2811, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.18262, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.07772, "n_shadow": 0, "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": 0.54687, "trunc_abs_frac": 0.54687, "mean_delta_full": -1.17456, "read_hit_rate": {"surface": 0.8314}, "mean_gate": {"surface": 0.40388}, "n_reads": 5656, "n_cand": 6803, "read_suppressed_frac": 0.1031, "dropout_reads_logged": 5656, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 199.10702514648438, "n_util_obs": 180, "util_count_spearman": -0.0016008724549111599}, "probe_overhead_sec_total": 339.732, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.156491, "alpha": 0.03, "t": 1733.6}
+{"event": "consolidate", "step": 5350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1776.4}
+{"event": "train", "step": 5350, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.29323, "bpb": 1.86573, "chunk_loss": 0.03041, "F": 0.20523, "G": 0.21068, "chunks": 210.16, "bytes_per_chunk": 4.873, "boundary_entropy": 0.08493, "router_noise": 0.0, "grad_norm": 0.6956, "lr": 3.86e-05, "router_trainable": true, "sec_per_step": 0.875, "probe_sec": 0.0497, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02997, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.0697, "n_shadow": 0, "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": 1.07418, "trunc_abs_frac": 1.01146, "mean_delta_full": 0.57239, "read_hit_rate": {"surface": 0.82937}, "mean_gate": {"surface": 0.40211}, "n_reads": 5551, "n_cand": 6693, "read_suppressed_frac": 0.1032, "dropout_reads_logged": 5551, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 201.07913208007812, "n_util_obs": 186, "util_count_spearman": 0.0033446936672743123}, "probe_overhead_sec_total": 343.544, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.157366, "alpha": 0.03, "t": 1777.4}
+{"event": "consolidate", "step": 5400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1815.5}
+{"event": "probe", "step": 5400, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3594.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.206516, "cps": 28.078125, "mean_p": 0.205777, "boundary_entropy": 0.075274, "bpb": 1.827959}, "tr": {"chunks": 3358.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.185494, "cps": 26.234375, "mean_p": 0.209033, "boundary_entropy": 0.164625, "bpb": 2.003823}, "fi": {"chunks": 3664.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.19478, "cps": 28.625, "mean_p": 0.208473, "boundary_entropy": 0.124694, "bpb": 2.151373}}, "bpb": {"en": 1.82796, "tr": 2.00382, "fi": 2.15137}, "gini_cpc": 0.023884, "gini_cps": 0.019216, "bpic": 5.344009, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.8982064880639878, "fi": 0.9431692410995627}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 202.8641815185547, "n_util_obs": 186, "util_count_spearman": 0.006227825203157272}, "t": 1816.1}
+{"event": "eval", "step": 5400, "phase": 3, "bpb": {"en": 1.84932, "tr": 2.01181, "fi": 2.16012}, "bytes_per_chunk": {"en": 4.632, "tr": 5.762, "fi": 5.109}, "bpb_mean": 2.00709, "frozen_inventory": true, "t": 1817.1}
+{"event": "train", "step": 5400, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35953, "bpb": 1.96139, "chunk_loss": 0.030448, "F": 0.20486, "G": 0.21815, "chunks": 209.78, "bytes_per_chunk": 4.881, "boundary_entropy": 0.10444, "router_noise": 0.0, "grad_norm": 0.7588, "lr": 3.73e-05, "router_trainable": true, "sec_per_step": 0.867, "probe_sec": 0.1295, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10196, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.42299, "n_shadow": 0, "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": 0.26853, "trunc_abs_frac": 0.54942, "mean_delta_full": -0.23468, "read_hit_rate": {"surface": 0.78985}, "mean_gate": {"surface": 0.37854}, "n_reads": 5277, "n_cand": 6681, "read_suppressed_frac": 0.0968, "dropout_reads_logged": 5277, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 202.90927124023438, "n_util_obs": 186, "util_count_spearman": 0.006298691308178974}, "probe_overhead_sec_total": 346.99, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.05224, "alpha": 0.03, "t": 1820.7}
+{"event": "consolidate", "step": 5450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1856.5}
+{"event": "train", "step": 5450, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.30299, "bpb": 1.87982, "chunk_loss": 0.030434, "F": 0.20636, "G": 0.21219, "chunks": 211.31, "bytes_per_chunk": 4.846, "boundary_entropy": 0.08663, "router_noise": 0.0, "grad_norm": 0.6852, "lr": 3.62e-05, "router_trainable": true, "sec_per_step": 0.732, "probe_sec": 0.1438, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07867, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.05373, "n_shadow": 0, "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": -0.31441, "trunc_abs_frac": 0.16509, "mean_delta_full": -0.06934, "read_hit_rate": {"surface": 0.83715}, "mean_gate": {"surface": 0.39596}, "n_reads": 5634, "n_cand": 6730, "read_suppressed_frac": 0.1019, "dropout_reads_logged": 5634, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 204.50827026367188, "n_util_obs": 184, "util_count_spearman": -0.02000155073180807}, "probe_overhead_sec_total": 350.591, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.038017, "alpha": 0.03, "t": 1857.3}
+{"event": "consolidate", "step": 5500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1876.4}
+{"event": "train", "step": 5500, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.29597, "bpb": 1.86969, "chunk_loss": 0.030542, "F": 0.21292, "G": 0.2174, "chunks": 218.03, "bytes_per_chunk": 4.697, "boundary_entropy": 0.08759, "router_noise": 0.0, "grad_norm": 0.6749, "lr": 3.51e-05, "router_trainable": true, "sec_per_step": 0.385, "probe_sec": 0.0285, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02101, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.00919, "n_shadow": 0, "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": 0.36646, "trunc_abs_frac": 0.36646, "mean_delta_full": 0.91138, "read_hit_rate": {"surface": 0.83744}, "mean_gate": {"surface": 0.40216}, "n_reads": 5816, "n_cand": 6945, "read_suppressed_frac": 0.0977, "dropout_reads_logged": 5816, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 206.35586547851562, "n_util_obs": 180, "util_count_spearman": -0.0047017912075475584}, "probe_overhead_sec_total": 353.739, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.036934, "alpha": 0.03, "t": 1876.6}
+{"event": "consolidate", "step": 5550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1892.0}
+{"event": "train", "step": 5550, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.29343, "bpb": 1.86602, "chunk_loss": 0.030504, "F": 0.20828, "G": 0.21482, "chunks": 213.28, "bytes_per_chunk": 4.801, "boundary_entropy": 0.09096, "router_noise": 0.0, "grad_norm": 0.6707, "lr": 3.41e-05, "router_trainable": true, "sec_per_step": 0.313, "probe_sec": 0.0324, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02009, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.41625, "n_shadow": 0, "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": 1.03665, "trunc_abs_frac": 7.68444, "mean_delta_full": 0.40371, "read_hit_rate": {"surface": 0.82659}, "mean_gate": {"surface": 0.38479}, "n_reads": 5615, "n_cand": 6793, "read_suppressed_frac": 0.0976, "dropout_reads_logged": 5615, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 207.22361755371094, "n_util_obs": 178, "util_count_spearman": -0.006695186551324621}, "probe_overhead_sec_total": 355.992, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.035616, "alpha": 0.03, "t": 1892.2}
+{"event": "consolidate", "step": 5600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1908.8}
+{"event": "probe", "step": 5600, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3595.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.206574, "cps": 28.085938, "mean_p": 0.20534, "boundary_entropy": 0.074355, "bpb": 1.822407}, "tr": {"chunks": 3306.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.182622, "cps": 25.828125, "mean_p": 0.206362, "boundary_entropy": 0.161501, "bpb": 2.003206}, "fi": {"chunks": 3617.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.192281, "cps": 28.257812, "mean_p": 0.20509, "boundary_entropy": 0.121475, "bpb": 2.144937}}, "bpb": {"en": 1.82241, "tr": 2.00321, "fi": 2.14494}, "gini_cpc": 0.027461, "gini_cps": 0.019712, "bpic": 5.393801, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.8840514143415321, "fi": 0.9308117368961684}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 209.392822265625, "n_util_obs": 179, "util_count_spearman": 0.003299646391731007}, "t": 1909.4}
+{"event": "eval", "step": 5600, "phase": 3, "bpb": {"en": 1.84399, "tr": 2.01019, "fi": 2.15446}, "bytes_per_chunk": {"en": 4.633, "tr": 5.818, "fi": 5.16}, "bpb_mean": 2.00288, "frozen_inventory": true, "t": 1910.5}
+{"event": "train", "step": 5600, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.34793, "bpb": 1.94466, "chunk_loss": 0.030374, "F": 0.20151, "G": 0.21098, "chunks": 206.34, "bytes_per_chunk": 4.963, "boundary_entropy": 0.09532, "router_noise": 0.0, "grad_norm": 0.7145, "lr": 3.33e-05, "router_trainable": true, "sec_per_step": 0.404, "probe_sec": 0.1068, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09156, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.29599, "n_shadow": 0, "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": -0.60527, "trunc_abs_frac": 1.5165, "mean_delta_full": -0.14075, "read_hit_rate": {"surface": 0.80597}, "mean_gate": {"surface": 0.38551}, "n_reads": 5296, "n_cand": 6571, "read_suppressed_frac": 0.102, "dropout_reads_logged": 5296, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 209.43231201171875, "n_util_obs": 179, "util_count_spearman": 0.0024417814542831025}, "probe_overhead_sec_total": 358.38, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.039184, "alpha": 0.03, "t": 1912.4}
+{"event": "consolidate", "step": 5650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1929.5}
+{"event": "train", "step": 5650, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.30473, "bpb": 1.88233, "chunk_loss": 0.030413, "F": 0.20441, "G": 0.21156, "chunks": 209.31, "bytes_per_chunk": 4.892, "boundary_entropy": 0.08973, "router_noise": 0.0, "grad_norm": 0.7342, "lr": 3.25e-05, "router_trainable": true, "sec_per_step": 0.346, "probe_sec": 0.0565, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0454, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.24306, "n_shadow": 0, "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": -1.68268, "trunc_abs_frac": 1.34678, "mean_delta_full": -0.12663, "read_hit_rate": {"surface": 0.82283}, "mean_gate": {"surface": 0.38409}, "n_reads": 5485, "n_cand": 6666, "read_suppressed_frac": 0.1043, "dropout_reads_logged": 5485, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 210.6644287109375, "n_util_obs": 179, "util_count_spearman": 0.023122633021572195}, "probe_overhead_sec_total": 361.184, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.043721, "alpha": 0.03, "t": 1929.7}
+{"event": "consolidate", "step": 5700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1956.5}
+{"event": "train", "step": 5700, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.426, "bpb": 2.05728, "chunk_loss": 0.030433, "F": 0.20428, "G": 0.21333, "chunks": 209.19, "bytes_per_chunk": 4.895, "boundary_entropy": 0.09586, "router_noise": 0.0, "grad_norm": 0.7818, "lr": 3.18e-05, "router_trainable": true, "sec_per_step": 0.545, "probe_sec": 0.0326, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02286, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.16809, "n_shadow": 0, "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": 0.93071, "trunc_abs_frac": 3.73495, "mean_delta_full": -0.11274, "read_hit_rate": {"surface": 0.7828}, "mean_gate": {"surface": 0.3714}, "n_reads": 5215, "n_cand": 6662, "read_suppressed_frac": 0.1011, "dropout_reads_logged": 5215, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 212.7618408203125, "n_util_obs": 180, "util_count_spearman": -0.0006810909801742852}, "probe_overhead_sec_total": 363.863, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.060439, "alpha": 0.03, "t": 1957.0}
+{"event": "consolidate", "step": 5750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1988.6}
+{"event": "train", "step": 5750, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.25519, "bpb": 1.81086, "chunk_loss": 0.03054, "F": 0.21021, "G": 0.21736, "chunks": 215.25, "bytes_per_chunk": 4.757, "boundary_entropy": 0.08812, "router_noise": 0.0, "grad_norm": 0.7101, "lr": 3.13e-05, "router_trainable": true, "sec_per_step": 0.635, "probe_sec": 0.0246, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01837, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.17038, "n_shadow": 0, "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": 0.77079, "trunc_abs_frac": 2.57455, "mean_delta_full": -0.29268, "read_hit_rate": {"surface": 0.83795}, "mean_gate": {"surface": 0.39156}, "n_reads": 5745, "n_cand": 6856, "read_suppressed_frac": 0.1015, "dropout_reads_logged": 5745, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 214.16079711914062, "n_util_obs": 177, "util_count_spearman": -0.04813585291113381}, "probe_overhead_sec_total": 366.565, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.059369, "alpha": 0.03, "t": 1988.7}
+{"event": "consolidate", "step": 5800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2000.7}
+{"event": "probe", "step": 5800, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3591.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.206344, "cps": 28.054688, "mean_p": 0.204329, "boundary_entropy": 0.074385, "bpb": 1.819442}, "tr": {"chunks": 3294.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.181959, "cps": 25.734375, "mean_p": 0.20504, "boundary_entropy": 0.158883, "bpb": 1.995564}, "fi": {"chunks": 3586.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.190633, "cps": 28.015625, "mean_p": 0.203286, "boundary_entropy": 0.119756, "bpb": 2.141884}}, "bpb": {"en": 1.81944, "tr": 1.99556, "fi": 2.14188}, "gini_cpc": 0.02808, "gini_cps": 0.018909, "bpic": 5.418012, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.8818236831098906, "fi": 0.9238620269064383}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 216.25363159179688, "n_util_obs": 178, "util_count_spearman": -0.05346467658727415}, "t": 2001.3}
+{"event": "eval", "step": 5800, "phase": 3, "bpb": {"en": 1.84042, "tr": 2.00361, "fi": 2.15104}, "bytes_per_chunk": {"en": 4.638, "tr": 5.76, "fi": 5.193}, "bpb_mean": 1.99836, "frozen_inventory": true, "t": 2002.3}
+{"event": "train", "step": 5800, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.23957, "bpb": 1.78832, "chunk_loss": 0.030465, "F": 0.20786, "G": 0.21358, "chunks": 212.84, "bytes_per_chunk": 4.811, "boundary_entropy": 0.09003, "router_noise": 0.0, "grad_norm": 0.6999, "lr": 3.08e-05, "router_trainable": true, "sec_per_step": 0.302, "probe_sec": 0.0523, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03898, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.10381, "n_shadow": 0, "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": {"surface": 0.84659}, "mean_gate": {"surface": 0.39402}, "n_reads": 5739, "n_cand": 6779, "read_suppressed_frac": 0.0932, "dropout_reads_logged": 5739, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 216.32257080078125, "n_util_obs": 178, "util_count_spearman": -0.05369873681948317}, "probe_overhead_sec_total": 368.864, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.044566, "alpha": 0.03, "t": 2003.8}
+{"event": "consolidate", "step": 5850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2025.8}
+{"event": "train", "step": 5850, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.36054, "bpb": 1.96284, "chunk_loss": 0.030456, "F": 0.20465, "G": 0.21488, "chunks": 209.56, "bytes_per_chunk": 4.886, "boundary_entropy": 0.09822, "router_noise": 0.0, "grad_norm": 0.7472, "lr": 3.05e-05, "router_trainable": true, "sec_per_step": 0.455, "probe_sec": 0.0602, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03142, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.17916, "n_shadow": 0, "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": 0.04007, "trunc_abs_frac": 5.78664, "mean_delta_full": -0.32149, "read_hit_rate": {"surface": 0.78184}, "mean_gate": {"surface": 0.37561}, "n_reads": 5218, "n_cand": 6674, "read_suppressed_frac": 0.1094, "dropout_reads_logged": 5218, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 218.0099334716797, "n_util_obs": 175, "util_count_spearman": -0.047377966860725484}, "probe_overhead_sec_total": 371.383, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.036998, "alpha": 0.03, "t": 2026.6}
+{"event": "consolidate", "step": 5900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2049.5}
+{"event": "train", "step": 5900, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.34185, "bpb": 1.93588, "chunk_loss": 0.030519, "F": 0.21066, "G": 0.21643, "chunks": 215.72, "bytes_per_chunk": 4.747, "boundary_entropy": 0.09085, "router_noise": 0.0, "grad_norm": 0.717, "lr": 3.02e-05, "router_trainable": true, "sec_per_step": 0.464, "probe_sec": 0.0476, "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": 0.05837, "n_shadow": 0, "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": 0.88076, "trunc_abs_frac": 0.54682, "mean_delta_full": -0.30396, "read_hit_rate": {"surface": 0.81808}, "mean_gate": {"surface": 0.38074}, "n_reads": 5621, "n_cand": 6871, "read_suppressed_frac": 0.0961, "dropout_reads_logged": 5621, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 219.35012817382812, "n_util_obs": 172, "util_count_spearman": -0.04674746417057565}, "probe_overhead_sec_total": 374.018, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.032619, "alpha": 0.03, "t": 2049.8}
+{"event": "consolidate", "step": 5950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2061.5}
+{"event": "train", "step": 5950, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.28478, "bpb": 1.85354, "chunk_loss": 0.030409, "F": 0.20331, "G": 0.21147, "chunks": 208.19, "bytes_per_chunk": 4.919, "boundary_entropy": 0.09249, "router_noise": 0.0, "grad_norm": 0.6962, "lr": 3.01e-05, "router_trainable": true, "sec_per_step": 0.246, "probe_sec": 0.046, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.024, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.24764, "n_shadow": 0, "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": 0.66977, "trunc_abs_frac": 0.98348, "mean_delta_full": 0.23258, "read_hit_rate": {"surface": 0.80603}, "mean_gate": {"surface": 0.37704}, "n_reads": 5344, "n_cand": 6630, "read_suppressed_frac": 0.0981, "dropout_reads_logged": 5344, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 220.77706909179688, "n_util_obs": 170, "util_count_spearman": 0.01002070089948156}, "probe_overhead_sec_total": 376.857, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.032846, "alpha": 0.03, "t": 2062.1}
+{"event": "consolidate", "step": 6000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2075.7}
+{"event": "train", "step": 5999, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.33716, "bpb": 1.92911, "chunk_loss": 0.030462, "F": 0.20709, "G": 0.21255, "chunks": 212.06, "bytes_per_chunk": 4.829, "boundary_entropy": 0.08922, "router_noise": 0.0, "grad_norm": 0.7215, "lr": 3e-05, "router_trainable": true, "sec_per_step": 0.273, "probe_sec": 0.0542, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04732, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.22716, "n_shadow": 0, "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": 0.99127, "trunc_abs_frac": 0.99127, "mean_delta_full": 1.75903, "read_hit_rate": {"surface": 0.82099}, "mean_gate": {"surface": 0.39829}, "n_reads": 5545, "n_cand": 6754, "read_suppressed_frac": 0.0994, "dropout_reads_logged": 5545, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 222.83358764648438, "n_util_obs": 171, "util_count_spearman": 0.003163174694602443}, "probe_overhead_sec_total": 379.567, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.035731, "alpha": 0.03, "t": 2075.7}
+{"event": "probe", "step": 6000, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3592.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.206401, "cps": 28.0625, "mean_p": 0.204439, "boundary_entropy": 0.07554, "bpb": 1.816589}, "tr": {"chunks": 3307.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.182677, "cps": 25.835938, "mean_p": 0.205658, "boundary_entropy": 0.15935, "bpb": 1.989709}, "fi": {"chunks": 3618.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.192334, "cps": 28.265625, "mean_p": 0.204769, "boundary_entropy": 0.121402, "bpb": 2.136389}}, "bpb": {"en": 1.81659, "tr": 1.98971, "fi": 2.13639}, "gini_cpc": 0.027203, "gini_cps": 0.019714, "bpic": 5.394314, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.8850573961396596, "fi": 0.9318466994208833}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 222.83358764648438, "n_util_obs": 171, "util_count_spearman": 0.003163174694602443}, "t": 2076.4}
+{"event": "eval", "step": 6000, "phase": 3, "bpb": {"en": 1.83754, "tr": 1.99703, "fi": 2.14495}, "bytes_per_chunk": {"en": 4.637, "tr": 5.825, "fi": 5.166}, "bpb_mean": 1.99317, "frozen_inventory": true, "t": 2077.4}
+{"event": "reps_dumped", "dirs": ["/local/scratch/sas245/beetle-hnet/runs/ali_mem-random_s1/probe/reps/th0.30", "/local/scratch/sas245/beetle-hnet/runs/ali_mem-random_s1/probe/reps/th0.40", "/local/scratch/sas245/beetle-hnet/runs/ali_mem-random_s1/probe/reps/th0.50", "/local/scratch/sas245/beetle-hnet/runs/ali_mem-random_s1/probe/reps/th0.60", "/local/scratch/sas245/beetle-hnet/runs/ali_mem-random_s1/probe/reps/th0.70"], "t": 2080.6}
+{"event": "final", "run": "ali_mem-random_s1", "arch": "ali", "objective": "ratio", "condition": "A", "step": 6000, "params": 22785088, "seed": 1, "per_lang": {"en": {"cpc": 0.20640119519623054, "cps": 28.0625, "bpb": 1.8419399844062314, "bytes_per_chunk": 4.629351981845955, "reads_per_chunk": 0.8604451979197019}, "tr": {"cpc": 0.18267690438048942, "cps": 25.8359375, "bpb": 1.9899282500066862, "bytes_per_chunk": 5.839371832711477, "reads_per_chunk": 0.6357520743999554}, "fi": {"cpc": 0.1923342725001329, "cps": 28.265625, "bpb": 2.163213852181347, "bytes_per_chunk": 5.153366032023747, "reads_per_chunk": 0.6826031827162192}}, "bpb_mean": 1.9983606955314215, "gini_cpc": 0.027203056963964967, "gini_cps": 0.019714113657253396, "bpic": 5.3943139678615575, "mean_boundary_entropy": 0.08922240138053894, "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": 379.56724762916565, "probe_overhead_seq_frac": 0.4999166666666667, "wall_sec": 2082.8154470920563, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 222.83358764648438, "n_util_obs": 171, "util_count_spearman": 0.003163174694602443}, "ali_params": 268096, "ali_config": {"enabled": true, "budget_B": 65536, "staging_capacity": 262144, "policy": "random", "value_mode": "learned", "candidate_types": ["surface"], "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": 0.003163174694602443, "paired_vs_dropout": {"n": 223, "pearson": 0.24694214704998463, "spearman": 0.3293983846113891, "mean_abs_discrepancy": 0.39204078912734985, "mean_paired": 0.21460074186325073, "mean_dropout": 0.27552202343940735}, "predictor_calibration": {"n": 321, "all": {"n": 321, "pearson": 0.24120962335115792, "spearman": 0.28644424449991296, "mae": 0.21345339715480804, "reliability": [{"pred": 0.03801350295543671, "obs": 0.15347807109355927, "n": 32}, {"pred": 0.0771574079990387, "obs": 0.12130757421255112, "n": 32}, {"pred": 0.0957690179347992, "obs": 0.2269809991121292, "n": 33}, {"pred": 0.11215756833553314, "obs": 0.13108080625534058, "n": 31}, {"pred": 0.12409555912017822, "obs": 0.11267541348934174, "n": 32}, {"pred": 0.141658216714859, "obs": 0.15495285391807556, "n": 32}, {"pred": 0.1542840451002121, "obs": 0.2011578381061554, "n": 33}, {"pred": 0.18275374174118042, "obs": 0.3365477919578552, "n": 31}, {"pred": 0.22093304991722107, "obs": 0.33224111795425415, "n": 32}, {"pred": 0.2968534529209137, "obs": 0.4010263979434967, "n": 33}]}, "admitted": {"n": 321, "pearson": 0.24120962335115792, "spearman": 0.28644424449991296, "mae": 0.21345339715480804, "reliability": [{"pred": 0.03801350295543671, "obs": 0.15347807109355927, "n": 32}, {"pred": 0.0771574079990387, "obs": 0.12130757421255112, "n": 32}, {"pred": 0.0957690179347992, "obs": 0.2269809991121292, "n": 33}, {"pred": 0.11215756833553314, "obs": 0.13108080625534058, "n": 31}, {"pred": 0.12409555912017822, "obs": 0.11267541348934174, "n": 32}, {"pred": 0.141658216714859, "obs": 0.15495285391807556, "n": 32}, {"pred": 0.1542840451002121, "obs": 0.2011578381061554, "n": 33}, {"pred": 0.18275374174118042, "obs": 0.3365477919578552, "n": 31}, {"pred": 0.22093304991722107, "obs": 0.33224111795425415, "n": 32}, {"pred": 0.2968534529209137, "obs": 0.4010263979434967, "n": 33}]}, "staged": {"n": 0, "pearson": NaN, "spearman": NaN, "mae": NaN}}, "t": 2083.1}