| {"event": "hub", "repo": "AdaptiveChunking/ali_mem-shuffled_s0", "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-shuffled_s0", "arch": "ali", "objective": "ratio", "condition": "A", "params": 22785088, "langs": ["en", "tr", "fi"], "premium": {"en": 1.0, "tr": 1.12, "fi": 1.104}, "schedule": {"tau1": 0.4, "tau2": 0.7, "p1_objective": "global", "p2_objective": "global", "p3_objective": "global", "freeze_in_p3": false, "freeze_scope": "router"}, "corpus": "/local/scratch/sas245/beetle-hnet/scratch/data/tier1_corpus", "premium_kind": "flores_per_sent", "doc_filtered": {}, "weights": {"en": 0.74, "tr": 0.1536, "fi": 0.1064}, "device": "cuda", "ali_params": 268096, "ali_config": {"enabled": true, "budget_B": 65536, "staging_capacity": 262144, "policy": "frequency", "value_mode": "shuffled", "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": 6384.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.366833, "cps": 49.875, "mean_p": 0.496725, "boundary_entropy": 0.686995, "bpb": 8.160238}, "tr": {"chunks": 5290.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.292217, "cps": 41.328125, "mean_p": 0.489567, "boundary_entropy": 0.687288, "bpb": 8.171434}, "fi": {"chunks": 8895.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.472862, "cps": 69.492188, "mean_p": 0.500688, "boundary_entropy": 0.687512, "bpb": 8.214758}}, "bpb": {"en": 8.16024, "tr": 8.17143, "fi": 8.21476}, "gini_cpc": 0.106395, "gini_cps": 0.116842, "bpic": 2.758131, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.7965927738475616, "fi": 1.2890367847879758}, "inventory": {"size": 0, "admitted": 0, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 0.0, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 3.5} |
| {"event": "train", "step": 0, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 5.66332, "bpb": 8.17044, "chunk_loss": 0.038537, "F": 0.28732, "G": 0.4892, "chunks": 294.22, "bytes_per_chunk": 3.48, "boundary_entropy": 0.69137, "router_noise": 0.0, "grad_norm": 11.0317, "lr": 1e-06, "router_trainable": true, "sec_per_step": 0.011, "probe_sec": 0.0055, "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": 9379, "read_suppressed_frac": 0.0, "dropout_reads_logged": 0, "inventory": {"size": 3696, "admitted": 3696, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 3696, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 2.5376081466674805, "n_util_obs": 0, "util_count_spearman": NaN}, "probe_overhead_sec_total": 0.005, "probe_overhead_seq_frac": 0.0, "util_loss": 0.0, "alpha": 0.03, "t": 4.7} |
| {"event": "probe", "step": 1, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 6301.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.362064, "cps": 49.226562, "mean_p": 0.496454, "boundary_entropy": 0.686986, "bpb": 8.128596}, "tr": {"chunks": 5229.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.288847, "cps": 40.851562, "mean_p": 0.489372, "boundary_entropy": 0.687277, "bpb": 8.149555}, "fi": {"chunks": 8822.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.468981, "cps": 68.921875, "mean_p": 0.500418, "boundary_entropy": 0.687508, "bpb": 8.187548}}, "bpb": {"en": 8.1286, "tr": 8.14956, "fi": 8.18755}, "gini_cpc": 0.107233, "gini_cps": 0.117695, "bpic": 2.787539, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.7977792403294754, "fi": 1.2952983448795534}, "inventory": {"size": 3696, "admitted": 3696, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 3696, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 2.5376081466674805, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 5.3} |
| {"event": "probe", "step": 2, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 6172.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.354651, "cps": 48.21875, "mean_p": 0.495926, "boundary_entropy": 0.686967, "bpb": 8.064973}, "tr": {"chunks": 5129.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.283323, "cps": 40.070312, "mean_p": 0.488973, "boundary_entropy": 0.687254, "bpb": 8.105949}, "fi": {"chunks": 8680.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.461432, "cps": 67.8125, "mean_p": 0.49989, "boundary_entropy": 0.687499, "bpb": 8.132375}}, "bpb": {"en": 8.06497, "tr": 8.10595, "fi": 8.13237}, "gini_cpc": 0.108003, "gini_cps": 0.118479, "bpic": 2.839297, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.7988777958086672, "fi": 1.3010861221318328}, "inventory": {"size": 6594, "admitted": 6594, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 6594, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 3.03078556060791, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 7.6} |
| {"event": "probe", "step": 4, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 5669.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.325748, "cps": 44.289062, "mean_p": 0.494031, "boundary_entropy": 0.686889, "bpb": 7.851683}, "tr": {"chunks": 4749.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.262332, "cps": 37.101562, "mean_p": 0.487549, "boundary_entropy": 0.687167, "bpb": 7.953323}, "fi": {"chunks": 8076.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.429323, "cps": 63.09375, "mean_p": 0.497944, "boundary_entropy": 0.687458, "bpb": 7.944822}}, "bpb": {"en": 7.85168, "tr": 7.95332, "fi": 7.94482}, "gini_cpc": 0.109423, "gini_cps": 0.119931, "bpic": 3.067589, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.8053214769639014, "fi": 1.3179595763332066}, "inventory": {"size": 11957, "admitted": 11957, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 11957, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 3.2203731536865234, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 10.1} |
| {"event": "probe", "step": 6, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 5005.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.287594, "cps": 39.101562, "mean_p": 0.491402, "boundary_entropy": 0.686756, "bpb": 7.564955}, "tr": {"chunks": 4211.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.232613, "cps": 32.898438, "mean_p": 0.485581, "boundary_entropy": 0.687032, "bpb": 7.739574}, "fi": {"chunks": 7241.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.384934, "cps": 56.570312, "mean_p": 0.495185, "boundary_entropy": 0.687373, "bpb": 7.686993}}, "bpb": {"en": 7.56496, "tr": 7.73957, "fi": 7.68699}, "gini_cpc": 0.112189, "gini_cps": 0.122744, "bpic": 3.447287, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.8088253016386475, "fi": 1.3384640238821301}, "inventory": {"size": 16821, "admitted": 16821, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 16821, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 3.144878387451172, "n_util_obs": 1, "util_count_spearman": NaN}, "t": 12.4} |
| {"event": "probe", "step": 8, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 4371.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.251164, "cps": 34.148438, "mean_p": 0.488781, "boundary_entropy": 0.686598, "bpb": 7.288793}, "tr": {"chunks": 3767.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.208087, "cps": 29.429688, "mean_p": 0.483666, "boundary_entropy": 0.686885, "bpb": 7.525963}, "fi": {"chunks": 6361.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.338153, "cps": 49.695312, "mean_p": 0.492185, "boundary_entropy": 0.68725, "bpb": 7.435888}}, "bpb": {"en": 7.28879, "tr": 7.52596, "fi": 7.43589}, "gini_cpc": 0.108741, "gini_cps": 0.119273, "bpic": 3.912822, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.8284921207559686, "fi": 1.3463464385296837}, "inventory": {"size": 21453, "admitted": 21453, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 21453, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 3.016920804977417, "n_util_obs": 2, "util_count_spearman": NaN}, "t": 14.2} |
| {"event": "probe", "step": 16, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 4439.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.255071, "cps": 34.679688, "mean_p": 0.491031, "boundary_entropy": 0.686931, "bpb": 6.660825}, "tr": {"chunks": 4547.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.251174, "cps": 35.523438, "mean_p": 0.488438, "boundary_entropy": 0.687357, "bpb": 6.978296}, "fi": {"chunks": 5611.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.298283, "cps": 43.835938, "mean_p": 0.491238, "boundary_entropy": 0.68741, "bpb": 6.839254}}, "bpb": {"en": 6.66082, "tr": 6.9783, "fi": 6.83925}, "gini_cpc": 0.039037, "gini_cps": 0.053527, "bpic": 3.886552, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9847214041421036, "fi": 1.169411500168457}, "inventory": {"size": 37483, "admitted": 37483, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 37483, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 2.6035802364349365, "n_util_obs": 3, "util_count_spearman": -0.5}, "t": 19.2} |
| {"event": "probe", "step": 32, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 4017.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.230822, "cps": 31.382812, "mean_p": 0.494628, "boundary_entropy": 0.687553, "bpb": 6.306956}, "tr": {"chunks": 4054.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.223941, "cps": 31.671875, "mean_p": 0.491857, "boundary_entropy": 0.687956, "bpb": 6.636628}, "fi": {"chunks": 4290.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.228058, "cps": 33.515625, "mean_p": 0.492831, "boundary_entropy": 0.687995, "bpb": 6.431118}}, "bpb": {"en": 6.30696, "tr": 6.63663, "fi": 6.43112}, "gini_cpc": 0.006719, "gini_cps": 0.014724, "bpic": 4.589596, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9701870678847622, "fi": 0.9880244620349693}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 2.7000732421875, "n_util_obs": 5, "util_count_spearman": 0.2}, "t": 27.8} |
| {"event": "consolidate", "step": 50, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 39.7} |
| {"event": "train", "step": 50, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 4.2257, "bpb": 6.0964, "chunk_loss": 0.030679, "F": 0.17596, "G": 0.49135, "chunks": 180.19, "bytes_per_chunk": 5.683, "boundary_entropy": 0.6921, "router_noise": 0.0, "grad_norm": 2.4392, "lr": 5.1e-05, "router_trainable": true, "sec_per_step": 0.716, "probe_sec": 0.0816, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04827, "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": -0.34766, "trunc_abs_frac": 0.45613, "mean_delta_full": -0.0125, "read_hit_rate": {"surface": 0.63549}, "mean_gate": {"surface": 0.12044}, "n_reads": 3642, "n_cand": 5731, "read_suppressed_frac": 0.0983, "dropout_reads_logged": 3642, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 3.7676544189453125, "n_util_obs": 11, "util_count_spearman": -0.17272727272727273}, "probe_overhead_sec_total": 2.639, "probe_overhead_seq_frac": 0.4902, "util_loss": 0.041128, "alpha": 0.03, "t": 40.5} |
| {"event": "probe", "step": 64, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 4092.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.235132, "cps": 31.96875, "mean_p": 0.48847, "boundary_entropy": 0.686631, "bpb": 5.623084}, "tr": {"chunks": 3910.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.215986, "cps": 30.546875, "mean_p": 0.488232, "boundary_entropy": 0.687522, "bpb": 6.010526}, "fi": {"chunks": 5395.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.2868, "cps": 42.148438, "mean_p": 0.493043, "boundary_entropy": 0.687704, "bpb": 5.816911}}, "bpb": {"en": 5.62308, "tr": 6.01053, "fi": 5.81691}, "gini_cpc": 0.063976, "gini_cps": 0.073897, "bpic": 4.234679, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9185751685168108, "fi": 1.2197422313736377}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 4.444732666015625, "n_util_obs": 12, "util_count_spearman": -0.3776223776223776}, "t": 46.8} |
| {"event": "consolidate", "step": 100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 66.1} |
| {"event": "train", "step": 100, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 3.21798, "bpb": 4.64256, "chunk_loss": 0.03236, "F": 0.20956, "G": 0.42149, "chunks": 214.59, "bytes_per_chunk": 4.772, "boundary_entropy": 0.66606, "router_noise": 0.0, "grad_norm": 3.9816, "lr": 0.000101, "router_trainable": true, "sec_per_step": 0.515, "probe_sec": 0.034, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02322, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.00536, "n_shadow": 0, "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": -0.40157, "trunc_abs_frac": 0.53917, "mean_delta_full": -0.03752, "read_hit_rate": {"surface": 0.75099}, "mean_gate": {"surface": 0.12022}, "n_reads": 5133, "n_cand": 6835, "read_suppressed_frac": 0.0953, "dropout_reads_logged": 5133, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 7.277191162109375, "n_util_obs": 14, "util_count_spearman": -0.34505494505494505}, "probe_overhead_sec_total": 5.218, "probe_overhead_seq_frac": 0.495, "util_loss": 0.000974, "alpha": 0.03, "t": 66.3} |
| {"event": "probe", "step": 128, "phase": 1, "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": 0.381228, "boundary_entropy": 0.614267, "bpb": 4.185078}, "tr": {"chunks": 3237.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.17881, "cps": 25.289062, "mean_p": 0.416921, "boundary_entropy": 0.654109, "bpb": 4.514139}, "fi": {"chunks": 4039.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.214715, "cps": 31.554688, "mean_p": 0.400137, "boundary_entropy": 0.639144, "bpb": 4.428807}}, "bpb": {"en": 4.18508, "tr": 4.51414, "fi": 4.42881}, "gini_cpc": 0.039876, "gini_cps": 0.049169, "bpic": 5.217215, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.8648785160156214, "fi": 1.0385440715104397}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 9.3436279296875, "n_util_obs": 25, "util_count_spearman": -0.29}, "t": 82.3} |
| {"event": "consolidate", "step": 150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 93.0} |
| {"event": "train", "step": 150, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.80713, "bpb": 4.04984, "chunk_loss": 0.031555, "F": 0.20349, "G": 0.36243, "chunks": 208.38, "bytes_per_chunk": 4.914, "boundary_entropy": 0.59276, "router_noise": 0.0, "grad_norm": 2.3785, "lr": 0.000151, "router_trainable": true, "sec_per_step": 0.542, "probe_sec": 0.0296, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01844, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.00996, "n_shadow": 0, "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": 0.7494, "trunc_abs_frac": 0.72865, "mean_delta_full": 0.0341, "read_hit_rate": {"surface": 0.71504}, "mean_gate": {"surface": 0.12047}, "n_reads": 4745, "n_cand": 6636, "read_suppressed_frac": 0.1031, "dropout_reads_logged": 4745, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 11.089508056640625, "n_util_obs": 28, "util_count_spearman": -0.3125342090859332}, "probe_overhead_sec_total": 7.907, "probe_overhead_seq_frac": 0.4967, "util_loss": 0.000511, "alpha": 0.03, "t": 93.4} |
| {"event": "consolidate", "step": 200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 116.6} |
| {"event": "probe", "step": 200, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 2995.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.172097, "cps": 23.398438, "mean_p": 0.366968, "boundary_entropy": 0.571284, "bpb": 3.64046}, "tr": {"chunks": 3563.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.196818, "cps": 27.835938, "mean_p": 0.378706, "boundary_entropy": 0.594748, "bpb": 3.789223}, "fi": {"chunks": 3425.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.182074, "cps": 26.757812, "mean_p": 0.372125, "boundary_entropy": 0.600921, "bpb": 3.878583}}, "bpb": {"en": 3.64046, "tr": 3.78922, "fi": 3.87858}, "gini_cpc": 0.029912, "gini_cps": 0.037931, "bpic": 5.682861, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.1436484992157196, "fi": 1.0579764140063326}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 14.020751953125, "n_util_obs": 39, "util_count_spearman": -0.3311740890688259}, "t": 117.2} |
| {"event": "eval", "step": 200, "phase": 1, "bpb": {"en": 3.64665, "tr": 3.79886, "fi": 3.8785}, "bytes_per_chunk": {"en": 5.894, "tr": 5.536, "fi": 5.957}, "bpb_mean": 3.77467, "frozen_inventory": true, "t": 118.1} |
| {"event": "train", "step": 200, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.55156, "bpb": 3.68112, "chunk_loss": 0.030258, "F": 0.17136, "G": 0.37026, "chunks": 175.47, "bytes_per_chunk": 5.836, "boundary_entropy": 0.5794, "router_noise": 0.0, "grad_norm": 1.1396, "lr": 0.000201, "router_trainable": true, "sec_per_step": 0.526, "probe_sec": 0.0527, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03061, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.03958, "n_shadow": 0, "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": -0.21915, "trunc_abs_frac": 0.25541, "mean_delta_full": 0.06991, "read_hit_rate": {"surface": 0.6321}, "mean_gate": {"surface": 0.12289}, "n_reads": 3529, "n_cand": 5583, "read_suppressed_frac": 0.0966, "dropout_reads_logged": 3529, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 14.07562255859375, "n_util_obs": 39, "util_count_spearman": -0.3066801619433198}, "probe_overhead_sec_total": 10.528, "probe_overhead_seq_frac": 0.4975, "util_loss": 0.00039, "alpha": 0.03, "t": 119.7} |
| {"event": "consolidate", "step": 250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 132.8} |
| {"event": "train", "step": 250, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.38684, "bpb": 3.44348, "chunk_loss": 0.031865, "F": 0.21381, "G": 0.35151, "chunks": 218.94, "bytes_per_chunk": 4.677, "boundary_entropy": 0.53416, "router_noise": 0.0, "grad_norm": 1.2786, "lr": 0.000251, "router_trainable": true, "sec_per_step": 0.268, "probe_sec": 0.0812, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07155, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.02105, "n_shadow": 0, "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": -0.44987, "trunc_abs_frac": 0.27276, "mean_delta_full": 0.02374, "read_hit_rate": {"surface": 0.67465}, "mean_gate": {"surface": 0.12417}, "n_reads": 4705, "n_cand": 6974, "read_suppressed_frac": 0.0982, "dropout_reads_logged": 4705, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 17.179351806640625, "n_util_obs": 51, "util_count_spearman": -0.2179185520361991}, "probe_overhead_sec_total": 13.251, "probe_overhead_seq_frac": 0.498, "util_loss": 0.000493, "alpha": 0.03, "t": 133.0} |
| {"event": "probe", "step": 256, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3291.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.189105, "cps": 25.710938, "mean_p": 0.331147, "boundary_entropy": 0.512113, "bpb": 3.332948}, "tr": {"chunks": 3414.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.188588, "cps": 26.671875, "mean_p": 0.350701, "boundary_entropy": 0.538806, "bpb": 3.536269}, "fi": {"chunks": 3861.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.205252, "cps": 30.164062, "mean_p": 0.35547, "boundary_entropy": 0.556234, "bpb": 3.64095}}, "bpb": {"en": 3.33295, "tr": 3.53627, "fi": 3.64095}, "gini_cpc": 0.019058, "gini_cps": 0.035964, "bpic": 5.369298, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9972618447734831, "fi": 1.0853858515937478}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 17.520751953125, "n_util_obs": 51, "util_count_spearman": -0.22570135746606335}, "t": 134.8} |
| {"event": "consolidate", "step": 300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 146.6} |
| {"event": "train", "step": 300, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.28639, "bpb": 3.29856, "chunk_loss": 0.031405, "F": 0.21411, "G": 0.29975, "chunks": 219.25, "bytes_per_chunk": 4.67, "boundary_entropy": 0.42306, "router_noise": 0.0, "grad_norm": 1.799, "lr": 0.0003, "router_trainable": true, "sec_per_step": 0.276, "probe_sec": 0.0486, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03896, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.00356, "n_shadow": 0, "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": 0.09052, "trunc_abs_frac": 0.5754, "mean_delta_full": 0.11505, "read_hit_rate": {"surface": 0.68184}, "mean_gate": {"surface": 0.12603}, "n_reads": 4762, "n_cand": 6984, "read_suppressed_frac": 0.1029, "dropout_reads_logged": 4762, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 20.632766723632812, "n_util_obs": 63, "util_count_spearman": -0.15255376344086022}, "probe_overhead_sec_total": 16.287, "probe_overhead_seq_frac": 0.4983, "util_loss": 0.000605, "alpha": 0.03, "t": 146.9} |
| {"event": "consolidate", "step": 350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 158.2} |
| {"event": "train", "step": 350, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.1975, "bpb": 3.17032, "chunk_loss": 0.030759, "F": 0.19434, "G": 0.28676, "chunks": 199.0, "bytes_per_chunk": 5.146, "boundary_entropy": 0.40711, "router_noise": 0.0, "grad_norm": 1.5057, "lr": 0.0002999, "router_trainable": true, "sec_per_step": 0.231, "probe_sec": 0.0387, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03278, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.02255, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 3.07116, "trunc_abs_frac": 0.64095, "mean_delta_full": 0.04671, "read_hit_rate": {"surface": 0.65972}, "mean_gate": {"surface": 0.14066}, "n_reads": 4180, "n_cand": 6336, "read_suppressed_frac": 0.0981, "dropout_reads_logged": 4180, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 24.077362060546875, "n_util_obs": 78, "util_count_spearman": -0.07573439219008839}, "probe_overhead_sec_total": 19.096, "probe_overhead_seq_frac": 0.4986, "util_loss": 0.000692, "alpha": 0.03, "t": 158.4} |
| {"event": "consolidate", "step": 400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 169.8} |
| {"event": "probe", "step": 400, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3505.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.201402, "cps": 27.382812, "mean_p": 0.262327, "boundary_entropy": 0.331335, "bpb": 2.898186}, "tr": {"chunks": 3731.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.206098, "cps": 29.148438, "mean_p": 0.297282, "boundary_entropy": 0.436589, "bpb": 3.199116}, "fi": {"chunks": 4247.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.225772, "cps": 33.179688, "mean_p": 0.300211, "boundary_entropy": 0.417593, "bpb": 3.387228}}, "bpb": {"en": 2.89819, "tr": 3.19912, "fi": 3.38723}, "gini_cpc": 0.025655, "gini_cps": 0.043078, "bpic": 4.940521, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0233184291850965, "fi": 1.1210022272002047}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 27.586395263671875, "n_util_obs": 96, "util_count_spearman": 0.09865708084644602}, "t": 170.4} |
| {"event": "eval", "step": 400, "phase": 1, "bpb": {"en": 2.91289, "tr": 3.2065, "fi": 3.37989}, "bytes_per_chunk": {"en": 4.7, "tr": 5.59, "fi": 4.784}, "bpb_mean": 3.16643, "frozen_inventory": true, "t": 171.4} |
| {"event": "train", "step": 400, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.04626, "bpb": 2.95212, "chunk_loss": 0.031103, "F": 0.21146, "G": 0.27529, "chunks": 216.53, "bytes_per_chunk": 4.729, "boundary_entropy": 0.34501, "router_noise": 0.0, "grad_norm": 1.2276, "lr": 0.0002998, "router_trainable": true, "sec_per_step": 0.277, "probe_sec": 0.0577, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04805, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.04649, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": -5.51815, "trunc_abs_frac": 0.7366, "mean_delta_full": -0.01514, "read_hit_rate": {"surface": 0.71364}, "mean_gate": {"surface": 0.16555}, "n_reads": 4922, "n_cand": 6897, "read_suppressed_frac": 0.1032, "dropout_reads_logged": 4922, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 27.662841796875, "n_util_obs": 96, "util_count_spearman": 0.10113944655453065}, "probe_overhead_sec_total": 21.612, "probe_overhead_seq_frac": 0.4988, "util_loss": 0.000996, "alpha": 0.03, "t": 172.3} |
| {"event": "consolidate", "step": 450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 183.3} |
| {"event": "train", "step": 450, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.02208, "bpb": 2.91724, "chunk_loss": 0.030947, "F": 0.20648, "G": 0.26903, "chunks": 211.44, "bytes_per_chunk": 4.843, "boundary_entropy": 0.33119, "router_noise": 0.0, "grad_norm": 1.4071, "lr": 0.0002995, "router_trainable": true, "sec_per_step": 0.225, "probe_sec": 0.0346, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02998, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.01769, "n_shadow": 0, "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": 0.48382, "trunc_abs_frac": 0.77673, "mean_delta_full": 0.31689, "read_hit_rate": {"surface": 0.70953}, "mean_gate": {"surface": 0.18469}, "n_reads": 4778, "n_cand": 6734, "read_suppressed_frac": 0.0977, "dropout_reads_logged": 4778, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 31.328323364257812, "n_util_obs": 104, "util_count_spearman": 0.12790496080532215}, "probe_overhead_sec_total": 24.369, "probe_overhead_seq_frac": 0.4989, "util_loss": 0.001444, "alpha": 0.03, "t": 183.5} |
| {"event": "consolidate", "step": 500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 193.9} |
| {"event": "train", "step": 500, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.99028, "bpb": 2.87136, "chunk_loss": 0.031011, "F": 0.20966, "G": 0.26825, "chunks": 214.69, "bytes_per_chunk": 4.77, "boundary_entropy": 0.31691, "router_noise": 0.0, "grad_norm": 1.1348, "lr": 0.0002992, "router_trainable": true, "sec_per_step": 0.212, "probe_sec": 0.0321, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02508, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.04747, "n_shadow": 0, "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": 0.32448, "trunc_abs_frac": 1.34429, "mean_delta_full": -0.06112, "read_hit_rate": {"surface": 0.6964}, "mean_gate": {"surface": 0.18995}, "n_reads": 4762, "n_cand": 6838, "read_suppressed_frac": 0.1002, "dropout_reads_logged": 4762, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 35.008026123046875, "n_util_obs": 117, "util_count_spearman": 0.06645923522624624}, "probe_overhead_sec_total": 26.667, "probe_overhead_seq_frac": 0.499, "util_loss": 0.001962, "alpha": 0.03, "t": 194.1} |
| {"event": "probe", "step": 512, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3467.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.199219, "cps": 27.085938, "mean_p": 0.258201, "boundary_entropy": 0.308675, "bpb": 2.704498}, "tr": {"chunks": 3705.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.204662, "cps": 28.945312, "mean_p": 0.278253, "boundary_entropy": 0.40163, "bpb": 3.032239}, "fi": {"chunks": 4085.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.21716, "cps": 31.914062, "mean_p": 0.287899, "boundary_entropy": 0.383154, "bpb": 3.245036}}, "bpb": {"en": 2.7045, "tr": 3.03224, "fi": 3.24504}, "gini_cpc": 0.01926, "gini_cps": 0.036599, "bpic": 5.039709, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0273251954201563, "fi": 1.090060131954594}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 35.83857727050781, "n_util_obs": 121, "util_count_spearman": 0.07861752915864337}, "t": 197.3} |
| {"event": "consolidate", "step": 550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 206.8} |
| {"event": "train", "step": 550, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.98421, "bpb": 2.86261, "chunk_loss": 0.031016, "F": 0.20908, "G": 0.27086, "chunks": 214.09, "bytes_per_chunk": 4.783, "boundary_entropy": 0.31537, "router_noise": 0.0, "grad_norm": 1.3908, "lr": 0.0002987, "router_trainable": true, "sec_per_step": 0.259, "probe_sec": 0.0289, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01836, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.03827, "n_shadow": 0, "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": 1.6686, "trunc_abs_frac": 1.27494, "mean_delta_full": -0.04462, "read_hit_rate": {"surface": 0.70128}, "mean_gate": {"surface": 0.19225}, "n_reads": 4782, "n_cand": 6819, "read_suppressed_frac": 0.0914, "dropout_reads_logged": 4782, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 38.74851989746094, "n_util_obs": 130, "util_count_spearman": 0.06376981796829125}, "probe_overhead_sec_total": 28.934, "probe_overhead_seq_frac": 0.4991, "util_loss": 0.002912, "alpha": 0.03, "t": 207.1} |
| {"event": "consolidate", "step": 600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 218.2} |
| {"event": "probe", "step": 600, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3552.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.204103, "cps": 27.75, "mean_p": 0.254166, "boundary_entropy": 0.282994, "bpb": 2.597261}, "tr": {"chunks": 3629.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.200464, "cps": 28.351562, "mean_p": 0.261625, "boundary_entropy": 0.348867, "bpb": 2.952565}, "fi": {"chunks": 4348.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.231141, "cps": 33.96875, "mean_p": 0.275383, "boundary_entropy": 0.332848, "bpb": 3.152393}}, "bpb": {"en": 2.59726, "tr": 2.95257, "fi": 3.15239}, "gini_cpc": 0.032171, "gini_cps": 0.046029, "bpic": 4.920808, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9821720698077518, "fi": 1.132475499528022}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 42.351959228515625, "n_util_obs": 143, "util_count_spearman": 0.12172100200269215}, "t": 218.9} |
| {"event": "eval", "step": 600, "phase": 1, "bpb": {"en": 2.61206, "tr": 2.96024, "fi": 3.15482}, "bytes_per_chunk": {"en": 4.71, "tr": 5.521, "fi": 4.602}, "bpb_mean": 2.90904, "frozen_inventory": true, "t": 220.0} |
| {"event": "train", "step": 600, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.88275, "bpb": 2.71623, "chunk_loss": 0.030901, "F": 0.20807, "G": 0.2627, "chunks": 213.06, "bytes_per_chunk": 4.806, "boundary_entropy": 0.29442, "router_noise": 0.0, "grad_norm": 1.3725, "lr": 0.0002982, "router_trainable": true, "sec_per_step": 0.279, "probe_sec": 0.0712, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06566, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.06882, "n_shadow": 0, "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": 0.09724, "trunc_abs_frac": 0.4255, "mean_delta_full": -0.2061, "read_hit_rate": {"surface": 0.70763}, "mean_gate": {"surface": 0.20967}, "n_reads": 4802, "n_cand": 6786, "read_suppressed_frac": 0.0895, "dropout_reads_logged": 4802, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 42.42478942871094, "n_util_obs": 143, "util_count_spearman": 0.12247431480523092}, "probe_overhead_sec_total": 31.8, "probe_overhead_seq_frac": 0.4992, "util_loss": 0.003494, "alpha": 0.03, "t": 221.0} |
| {"event": "consolidate", "step": 650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 231.7} |
| {"event": "train", "step": 650, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.91054, "bpb": 2.75633, "chunk_loss": 0.031026, "F": 0.21146, "G": 0.26615, "chunks": 216.53, "bytes_per_chunk": 4.729, "boundary_entropy": 0.29583, "router_noise": 0.0, "grad_norm": 1.1298, "lr": 0.0002975, "router_trainable": true, "sec_per_step": 0.217, "probe_sec": 0.0273, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02023, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.02879, "n_shadow": 0, "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": 0.79983, "trunc_abs_frac": 0.57191, "mean_delta_full": 0.38586, "read_hit_rate": {"surface": 0.709}, "mean_gate": {"surface": 0.2048}, "n_reads": 4890, "n_cand": 6897, "read_suppressed_frac": 0.0998, "dropout_reads_logged": 4890, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 46.152801513671875, "n_util_obs": 150, "util_count_spearman": 0.20110778355293385}, "probe_overhead_sec_total": 34.552, "probe_overhead_seq_frac": 0.4992, "util_loss": 0.004441, "alpha": 0.03, "t": 231.9} |
| {"event": "consolidate", "step": 700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 243.4} |
| {"event": "train", "step": 700, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.85328, "bpb": 2.67371, "chunk_loss": 0.030847, "F": 0.20706, "G": 0.25815, "chunks": 212.03, "bytes_per_chunk": 4.829, "boundary_entropy": 0.27588, "router_noise": 0.0, "grad_norm": 1.1337, "lr": 0.0002967, "router_trainable": true, "sec_per_step": 0.234, "probe_sec": 0.0835, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07802, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.14241, "n_shadow": 0, "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": 0.59181, "trunc_abs_frac": 0.73239, "mean_delta_full": 0.39938, "read_hit_rate": {"surface": 0.71242}, "mean_gate": {"surface": 0.21761}, "n_reads": 4811, "n_cand": 6753, "read_suppressed_frac": 0.0962, "dropout_reads_logged": 4811, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 49.93736267089844, "n_util_obs": 160, "util_count_spearman": 0.15745792618216325}, "probe_overhead_sec_total": 37.247, "probe_overhead_seq_frac": 0.4993, "util_loss": 0.004783, "alpha": 0.03, "t": 243.6} |
| {"event": "consolidate", "step": 750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 255.0} |
| {"event": "train", "step": 750, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.81976, "bpb": 2.62535, "chunk_loss": 0.030688, "F": 0.19992, "G": 0.25503, "chunks": 204.72, "bytes_per_chunk": 5.002, "boundary_entropy": 0.27857, "router_noise": 0.0, "grad_norm": 1.2298, "lr": 0.0002959, "router_trainable": true, "sec_per_step": 0.232, "probe_sec": 0.0315, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02468, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.05305, "n_shadow": 0, "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": 0.59033, "trunc_abs_frac": 0.69719, "mean_delta_full": 0.30185, "read_hit_rate": {"surface": 0.71422}, "mean_gate": {"surface": 0.21953}, "n_reads": 4656, "n_cand": 6519, "read_suppressed_frac": 0.1052, "dropout_reads_logged": 4656, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 53.76716613769531, "n_util_obs": 173, "util_count_spearman": 0.1064475662648022}, "probe_overhead_sec_total": 39.483, "probe_overhead_seq_frac": 0.4993, "util_loss": 0.005245, "alpha": 0.03, "t": 255.2} |
| {"event": "consolidate", "step": 800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 270.0} |
| {"event": "probe", "step": 800, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3570.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.205137, "cps": 27.890625, "mean_p": 0.245231, "boundary_entropy": 0.246147, "bpb": 2.434395}, "tr": {"chunks": 3500.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.193338, "cps": 27.34375, "mean_p": 0.253101, "boundary_entropy": 0.310316, "bpb": 2.774465}, "fi": {"chunks": 4547.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.24172, "cps": 35.523438, "mean_p": 0.278281, "boundary_entropy": 0.287625, "bpb": 2.962964}}, "bpb": {"en": 2.4344, "tr": 2.77447, "fi": 2.96296}, "gini_cpc": 0.050383, "gini_cps": 0.060084, "bpic": 4.883533, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.942482721420889, "fi": 1.1783355349475924}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 57.66709899902344, "n_util_obs": 186, "util_count_spearman": 0.0962128594007152}, "t": 270.7} |
| {"event": "eval", "step": 800, "phase": 1, "bpb": {"en": 2.44851, "tr": 2.78476, "fi": 2.96753}, "bytes_per_chunk": {"en": 4.646, "tr": 5.716, "fi": 4.448}, "bpb_mean": 2.7336, "frozen_inventory": true, "t": 271.6} |
| {"event": "train", "step": 800, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.80332, "bpb": 2.60164, "chunk_loss": 0.030959, "F": 0.21249, "G": 0.25729, "chunks": 217.59, "bytes_per_chunk": 4.706, "boundary_entropy": 0.25832, "router_noise": 0.0, "grad_norm": 0.9812, "lr": 0.0002949, "router_trainable": true, "sec_per_step": 0.35, "probe_sec": 0.0286, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02198, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.08039, "n_shadow": 0, "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": 0.90447, "trunc_abs_frac": 1.06778, "mean_delta_full": -0.07779, "read_hit_rate": {"surface": 0.72472}, "mean_gate": {"surface": 0.22215}, "n_reads": 5023, "n_cand": 6931, "read_suppressed_frac": 0.0974, "dropout_reads_logged": 5023, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 57.74342346191406, "n_util_obs": 186, "util_count_spearman": 0.09506972155669696}, "probe_overhead_sec_total": 42.681, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.005864, "alpha": 0.03, "t": 272.7} |
| {"event": "consolidate", "step": 850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 285.4} |
| {"event": "train", "step": 850, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.77761, "bpb": 2.56455, "chunk_loss": 0.030877, "F": 0.2113, "G": 0.2526, "chunks": 216.38, "bytes_per_chunk": 4.733, "boundary_entropy": 0.25129, "router_noise": 0.0, "grad_norm": 1.1388, "lr": 0.0002938, "router_trainable": true, "sec_per_step": 0.258, "probe_sec": 0.0232, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01771, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.17512, "n_shadow": 0, "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": 0.18336, "trunc_abs_frac": 4.68712, "mean_delta_full": 0.1757, "read_hit_rate": {"surface": 0.7159}, "mean_gate": {"surface": 0.23007}, "n_reads": 4934, "n_cand": 6892, "read_suppressed_frac": 0.0983, "dropout_reads_logged": 4934, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 61.637725830078125, "n_util_obs": 197, "util_count_spearman": 0.09172686148755764}, "probe_overhead_sec_total": 45.288, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.006422, "alpha": 0.03, "t": 285.6} |
| {"event": "consolidate", "step": 900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 297.4} |
| {"event": "train", "step": 900, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.61833, "bpb": 2.33475, "chunk_loss": 0.030878, "F": 0.21063, "G": 0.25129, "chunks": 215.69, "bytes_per_chunk": 4.748, "boundary_entropy": 0.24121, "router_noise": 0.0, "grad_norm": 0.9289, "lr": 0.0002927, "router_trainable": true, "sec_per_step": 0.245, "probe_sec": 0.0517, "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": 0.00949, "n_shadow": 0, "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": 0.08184, "trunc_abs_frac": 1.19191, "mean_delta_full": 0.0977, "read_hit_rate": {"surface": 0.74934}, "mean_gate": {"surface": 0.22618}, "n_reads": 5148, "n_cand": 6870, "read_suppressed_frac": 0.0979, "dropout_reads_logged": 5148, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 65.56536865234375, "n_util_obs": 216, "util_count_spearman": 0.04961719573775022}, "probe_overhead_sec_total": 48.111, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.006845, "alpha": 0.03, "t": 297.9} |
| {"event": "consolidate", "step": 950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 316.8} |
| {"event": "train", "step": 950, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.77122, "bpb": 2.55533, "chunk_loss": 0.030967, "F": 0.21521, "G": 0.25113, "chunks": 220.38, "bytes_per_chunk": 4.647, "boundary_entropy": 0.23551, "router_noise": 0.0, "grad_norm": 0.9134, "lr": 0.0002914, "router_trainable": true, "sec_per_step": 0.39, "probe_sec": 0.062, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04526, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.1297, "n_shadow": 0, "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": 0.77791, "trunc_abs_frac": 0.86777, "mean_delta_full": 0.2746, "read_hit_rate": {"surface": 0.72479}, "mean_gate": {"surface": 0.22746}, "n_reads": 5088, "n_cand": 7020, "read_suppressed_frac": 0.0994, "dropout_reads_logged": 5088, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 69.46760559082031, "n_util_obs": 224, "util_count_spearman": 0.042927610506085845}, "probe_overhead_sec_total": 50.528, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.007358, "alpha": 0.03, "t": 317.3} |
| {"event": "consolidate", "step": 1000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 342.6} |
| {"event": "probe", "step": 1000, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3542.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.203528, "cps": 27.671875, "mean_p": 0.235501, "boundary_entropy": 0.220257, "bpb": 2.341186}, "tr": {"chunks": 3703.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.204552, "cps": 28.929688, "mean_p": 0.242811, "boundary_entropy": 0.273689, "bpb": 2.65353}, "fi": {"chunks": 4675.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.248525, "cps": 36.523438, "mean_p": 0.27796, "boundary_entropy": 0.252105, "bpb": 2.825358}}, "bpb": {"en": 2.34119, "tr": 2.65353, "fi": 2.82536}, "gini_cpc": 0.045686, "gini_cps": 0.063367, "bpic": 4.759396, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.005029302024275, "fi": 1.2210833095872606}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 73.3236083984375, "n_util_obs": 240, "util_count_spearman": 0.027346828938002395}, "t": 343.3} |
| {"event": "eval", "step": 1000, "phase": 1, "bpb": {"en": 2.35144, "tr": 2.66185, "fi": 2.83369}, "bytes_per_chunk": {"en": 4.681, "tr": 5.367, "fi": 4.313}, "bpb_mean": 2.61566, "frozen_inventory": true, "t": 344.3} |
| {"event": "train", "step": 1000, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.71834, "bpb": 2.47904, "chunk_loss": 0.031168, "F": 0.22177, "G": 0.2538, "chunks": 227.09, "bytes_per_chunk": 4.509, "boundary_entropy": 0.22932, "router_noise": 0.0, "grad_norm": 0.9569, "lr": 0.0002901, "router_trainable": true, "sec_per_step": 0.589, "probe_sec": 0.0666, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03172, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.05032, "n_shadow": 0, "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": 1.42405, "trunc_abs_frac": 0.43776, "mean_delta_full": -0.07263, "read_hit_rate": {"surface": 0.73559}, "mean_gate": {"surface": 0.21992}, "n_reads": 5322, "n_cand": 7235, "read_suppressed_frac": 0.0904, "dropout_reads_logged": 5322, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 73.406494140625, "n_util_obs": 240, "util_count_spearman": 0.027568626965882647}, "probe_overhead_sec_total": 53.338, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.007855, "alpha": 0.03, "t": 346.8} |
| {"event": "probe", "step": 1024, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3573.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.205309, "cps": 27.914062, "mean_p": 0.235183, "boundary_entropy": 0.217621, "bpb": 2.335922}, "tr": {"chunks": 3642.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.201182, "cps": 28.453125, "mean_p": 0.239286, "boundary_entropy": 0.268217, "bpb": 2.651824}, "fi": {"chunks": 4643.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.246824, "cps": 36.273438, "mean_p": 0.27578, "boundary_entropy": 0.244048, "bpb": 2.836126}}, "bpb": {"en": 2.33592, "tr": 2.65182, "fi": 2.83613}, "gini_cpc": 0.046574, "gini_cps": 0.060156, "bpic": 4.784281, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9798971488505948, "fi": 1.2022032680826433}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 75.21173095703125, "n_util_obs": 246, "util_count_spearman": 0.05279242565889224}, "t": 357.2} |
| {"event": "consolidate", "step": 1050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 371.0} |
| {"event": "train", "step": 1050, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.73922, "bpb": 2.50917, "chunk_loss": 0.030804, "F": 0.211, "G": 0.24514, "chunks": 216.06, "bytes_per_chunk": 4.739, "boundary_entropy": 0.22633, "router_noise": 0.0, "grad_norm": 0.9223, "lr": 0.0002886, "router_trainable": true, "sec_per_step": 0.494, "probe_sec": 0.0413, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03064, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.29235, "n_shadow": 0, "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": 1.15782, "trunc_abs_frac": 0.93042, "mean_delta_full": 0.16243, "read_hit_rate": {"surface": 0.71316}, "mean_gate": {"surface": 0.2381}, "n_reads": 4908, "n_cand": 6882, "read_suppressed_frac": 0.1023, "dropout_reads_logged": 4908, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 77.29933166503906, "n_util_obs": 254, "util_count_spearman": 0.04293564504707856}, "probe_overhead_sec_total": 56.461, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.008247, "alpha": 0.03, "t": 371.5} |
| {"event": "consolidate", "step": 1100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 396.5} |
| {"event": "train", "step": 1100, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.73184, "bpb": 2.49852, "chunk_loss": 0.030884, "F": 0.21307, "G": 0.24755, "chunks": 218.19, "bytes_per_chunk": 4.693, "boundary_entropy": 0.22129, "router_noise": 0.0, "grad_norm": 0.9366, "lr": 0.0002871, "router_trainable": true, "sec_per_step": 0.519, "probe_sec": 0.0658, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02779, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.06136, "n_shadow": 0, "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": 0.71122}, "mean_gate": {"surface": 0.23718}, "n_reads": 4943, "n_cand": 6950, "read_suppressed_frac": 0.0961, "dropout_reads_logged": 4943, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 81.15406799316406, "n_util_obs": 268, "util_count_spearman": 0.042198583052520645}, "probe_overhead_sec_total": 59.62, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.008636, "alpha": 0.03, "t": 397.4} |
| {"event": "consolidate", "step": 1150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 412.4} |
| {"event": "train", "step": 1150, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.58705, "bpb": 2.28963, "chunk_loss": 0.030861, "F": 0.21533, "G": 0.24284, "chunks": 220.5, "bytes_per_chunk": 4.644, "boundary_entropy": 0.20841, "router_noise": 0.0, "grad_norm": 0.9556, "lr": 0.0002855, "router_trainable": true, "sec_per_step": 0.304, "probe_sec": 0.0427, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0366, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.14737, "n_shadow": 0, "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": 0.66798, "trunc_abs_frac": 0.67381, "mean_delta_full": -0.48389, "read_hit_rate": {"surface": 0.73946}, "mean_gate": {"surface": 0.2329}, "n_reads": 5194, "n_cand": 7024, "read_suppressed_frac": 0.1032, "dropout_reads_logged": 5194, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 85.04243469238281, "n_util_obs": 281, "util_count_spearman": 0.053490886995188176}, "probe_overhead_sec_total": 62.691, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.009267, "alpha": 0.03, "t": 412.6} |
| {"event": "consolidate", "step": 1200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 423.8} |
| {"event": "probe", "step": 1200, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3534.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.203068, "cps": 27.609375, "mean_p": 0.229431, "boundary_entropy": 0.200783, "bpb": 2.272633}, "tr": {"chunks": 3741.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.206651, "cps": 29.226562, "mean_p": 0.235989, "boundary_entropy": 0.246284, "bpb": 2.558455}, "fi": {"chunks": 4689.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.249269, "cps": 36.632812, "mean_p": 0.267878, "boundary_entropy": 0.209188, "bpb": 2.736555}}, "bpb": {"en": 2.27263, "tr": 2.55845, "fi": 2.73655}, "gini_cpc": 0.046739, "gini_cps": 0.06436, "bpic": 4.741892, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0176413180679842, "fi": 1.227512502844171}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 88.81231689453125, "n_util_obs": 296, "util_count_spearman": 0.061202756471157124}, "t": 424.4} |
| {"event": "eval", "step": 1200, "phase": 1, "bpb": {"en": 2.28372, "tr": 2.56609, "fi": 2.74474}, "bytes_per_chunk": {"en": 4.722, "tr": 5.328, "fi": 4.29}, "bpb_mean": 2.53152, "frozen_inventory": true, "t": 425.4} |
| {"event": "train", "step": 1200, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.62932, "bpb": 2.35061, "chunk_loss": 0.030814, "F": 0.21317, "G": 0.24157, "chunks": 218.28, "bytes_per_chunk": 4.691, "boundary_entropy": 0.20444, "router_noise": 0.0, "grad_norm": 0.8328, "lr": 0.0002837, "router_trainable": true, "sec_per_step": 0.28, "probe_sec": 0.1117, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10348, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.14016, "n_shadow": 0, "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": 0.76677, "trunc_abs_frac": 0.58073, "mean_delta_full": 0.29326, "read_hit_rate": {"surface": 0.74328}, "mean_gate": {"surface": 0.24243}, "n_reads": 5168, "n_cand": 6953, "read_suppressed_frac": 0.1037, "dropout_reads_logged": 5168, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 88.8919677734375, "n_util_obs": 296, "util_count_spearman": 0.06127379702232005}, "probe_overhead_sec_total": 65.809, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.009464, "alpha": 0.03, "t": 426.6} |
| {"event": "consolidate", "step": 1250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 438.0} |
| {"event": "train", "step": 1250, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.67932, "bpb": 2.42275, "chunk_loss": 0.030887, "F": 0.21619, "G": 0.24307, "chunks": 221.38, "bytes_per_chunk": 4.626, "boundary_entropy": 0.20387, "router_noise": 0.0, "grad_norm": 0.9203, "lr": 0.0002819, "router_trainable": true, "sec_per_step": 0.233, "probe_sec": 0.0657, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05802, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.0646, "n_shadow": 0, "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": 0.7604, "trunc_abs_frac": 0.7604, "mean_delta_full": -0.427, "read_hit_rate": {"surface": 0.72433}, "mean_gate": {"surface": 0.2312}, "n_reads": 5108, "n_cand": 7052, "read_suppressed_frac": 0.0938, "dropout_reads_logged": 5108, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 92.76849365234375, "n_util_obs": 310, "util_count_spearman": 0.04933321759243342}, "probe_overhead_sec_total": 68.453, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.009886, "alpha": 0.03, "t": 438.3} |
| {"event": "consolidate", "step": 1300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 449.9} |
| {"event": "train", "step": 1300, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.62488, "bpb": 2.34421, "chunk_loss": 0.030763, "F": 0.21133, "G": 0.23769, "chunks": 216.41, "bytes_per_chunk": 4.732, "boundary_entropy": 0.19972, "router_noise": 0.0, "grad_norm": 0.9329, "lr": 0.00028, "router_trainable": true, "sec_per_step": 0.238, "probe_sec": 0.0301, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02392, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.09533, "n_shadow": 0, "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": 1.38506, "trunc_abs_frac": 1.27238, "mean_delta_full": -0.11603, "read_hit_rate": {"surface": 0.73393}, "mean_gate": {"surface": 0.2394}, "n_reads": 5059, "n_cand": 6893, "read_suppressed_frac": 0.1012, "dropout_reads_logged": 5059, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 96.66156005859375, "n_util_obs": 324, "util_count_spearman": 0.032882970252963564}, "probe_overhead_sec_total": 71.109, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.010816, "alpha": 0.03, "t": 450.2} |
| {"event": "consolidate", "step": 1350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 461.5} |
| {"event": "train", "step": 1350, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.60889, "bpb": 2.32114, "chunk_loss": 0.030971, "F": 0.21906, "G": 0.24127, "chunks": 224.31, "bytes_per_chunk": 4.565, "boundary_entropy": 0.1891, "router_noise": 0.0, "grad_norm": 0.8922, "lr": 0.000278, "router_trainable": true, "sec_per_step": 0.232, "probe_sec": 0.093, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0777, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.3171, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 7, "trunc_ratio_sum": 1.01189, "trunc_abs_frac": 1.058, "mean_delta_full": 0.19359, "read_hit_rate": {"surface": 0.75315}, "mean_gate": {"surface": 0.23159}, "n_reads": 5382, "n_cand": 7146, "read_suppressed_frac": 0.0979, "dropout_reads_logged": 5382, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 100.58291625976562, "n_util_obs": 340, "util_count_spearman": 0.010507726114299428}, "probe_overhead_sec_total": 73.813, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.011416, "alpha": 0.03, "t": 461.8} |
| {"event": "consolidate", "step": 1400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 473.2} |
| {"event": "probe", "step": 1400, "phase": 1, "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": 0.225183, "boundary_entropy": 0.184033, "bpb": 2.227816}, "tr": {"chunks": 3729.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.205988, "cps": 29.132812, "mean_p": 0.23045, "boundary_entropy": 0.227201, "bpb": 2.499224}, "fi": {"chunks": 4760.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.253043, "cps": 37.1875, "mean_p": 0.268505, "boundary_entropy": 0.19799, "bpb": 2.657383}}, "bpb": {"en": 2.22782, "tr": 2.49922, "fi": 2.65738}, "gini_cpc": 0.047118, "gini_cps": 0.064091, "bpic": 4.693638, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9963336380050207, "fi": 1.2239340877419393}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 104.40937805175781, "n_util_obs": 349, "util_count_spearman": 0.0049368364132336955}, "t": 474.0} |
| {"event": "eval", "step": 1400, "phase": 1, "bpb": {"en": 2.23765, "tr": 2.50917, "fi": 2.66658}, "bytes_per_chunk": {"en": 4.66, "tr": 5.447, "fi": 4.246}, "bpb_mean": 2.47113, "frozen_inventory": true, "t": 475.2} |
| {"event": "train", "step": 1400, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.6966, "bpb": 2.44768, "chunk_loss": 0.030786, "F": 0.21362, "G": 0.23682, "chunks": 218.75, "bytes_per_chunk": 4.681, "boundary_entropy": 0.19029, "router_noise": 0.0, "grad_norm": 0.9499, "lr": 0.0002759, "router_trainable": true, "sec_per_step": 0.292, "probe_sec": 0.0486, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04139, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.06189, "n_shadow": 0, "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": -5.37217, "trunc_abs_frac": 5.37217, "mean_delta_full": 0.03772, "read_hit_rate": {"surface": 0.72991}, "mean_gate": {"surface": 0.237}, "n_reads": 5086, "n_cand": 6968, "read_suppressed_frac": 0.1032, "dropout_reads_logged": 5086, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 104.48838806152344, "n_util_obs": 349, "util_count_spearman": 0.005918248450269417}, "probe_overhead_sec_total": 76.525, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.01108, "alpha": 0.03, "t": 476.4} |
| {"event": "consolidate", "step": 1450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 487.6} |
| {"event": "train", "step": 1450, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.64489, "bpb": 2.37307, "chunk_loss": 0.030858, "F": 0.21793, "G": 0.23843, "chunks": 223.16, "bytes_per_chunk": 4.589, "boundary_entropy": 0.19363, "router_noise": 0.0, "grad_norm": 0.8423, "lr": 0.0002738, "router_trainable": true, "sec_per_step": 0.229, "probe_sec": 0.0545, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04763, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.22363, "n_shadow": 0, "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": 0.80121, "trunc_abs_frac": 2.35063, "mean_delta_full": 0.38748, "read_hit_rate": {"surface": 0.73752}, "mean_gate": {"surface": 0.23528}, "n_reads": 5243, "n_cand": 7109, "read_suppressed_frac": 0.0996, "dropout_reads_logged": 5243, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 108.42985534667969, "n_util_obs": 362, "util_count_spearman": 0.0010772478537098712}, "probe_overhead_sec_total": 79.256, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.009804, "alpha": 0.03, "t": 487.9} |
| {"event": "consolidate", "step": 1500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 499.0} |
| {"event": "train", "step": 1500, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.60677, "bpb": 2.31809, "chunk_loss": 0.030639, "F": 0.20859, "G": 0.22824, "chunks": 213.59, "bytes_per_chunk": 4.794, "boundary_entropy": 0.1858, "router_noise": 0.0, "grad_norm": 0.8528, "lr": 0.0002715, "router_trainable": true, "sec_per_step": 0.23, "probe_sec": 0.165, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.15694, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.23313, "n_shadow": 0, "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": 0.72556}, "mean_gate": {"surface": 0.23873}, "n_reads": 4936, "n_cand": 6803, "read_suppressed_frac": 0.0964, "dropout_reads_logged": 4936, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 112.38142395019531, "n_util_obs": 371, "util_count_spearman": 0.004468943696526857}, "probe_overhead_sec_total": 82.135, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.011323, "alpha": 0.03, "t": 499.4} |
| {"event": "consolidate", "step": 1550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 509.9} |
| {"event": "train", "step": 1550, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.57251, "bpb": 2.26865, "chunk_loss": 0.030758, "F": 0.21442, "G": 0.23053, "chunks": 219.56, "bytes_per_chunk": 4.664, "boundary_entropy": 0.17023, "router_noise": 0.0, "grad_norm": 0.7401, "lr": 0.0002692, "router_trainable": true, "sec_per_step": 0.215, "probe_sec": 0.0547, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04752, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.17365, "n_shadow": 0, "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": 0.50929, "trunc_abs_frac": 14.35949, "mean_delta_full": -0.35642, "read_hit_rate": {"surface": 0.73963}, "mean_gate": {"surface": 0.2287}, "n_reads": 5173, "n_cand": 6994, "read_suppressed_frac": 0.0996, "dropout_reads_logged": 5173, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 116.31553649902344, "n_util_obs": 381, "util_count_spearman": 0.03031800156634148}, "probe_overhead_sec_total": 84.749, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.011964, "alpha": 0.03, "t": 510.1} |
| {"event": "consolidate", "step": 1600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 521.6} |
| {"event": "probe", "step": 1600, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3629.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.208527, "cps": 28.351562, "mean_p": 0.221649, "boundary_entropy": 0.169255, "bpb": 2.180155}, "tr": {"chunks": 3787.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.209192, "cps": 29.585938, "mean_p": 0.223699, "boundary_entropy": 0.200629, "bpb": 2.438107}, "fi": {"chunks": 4810.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.255701, "cps": 37.578125, "mean_p": 0.265418, "boundary_entropy": 0.177549, "bpb": 2.594762}}, "bpb": {"en": 2.18016, "tr": 2.43811, "fi": 2.59476}, "gini_cpc": 0.046701, "gini_cps": 0.064398, "bpic": 4.640275, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0031870232409272, "fi": 1.2262255047120039}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 120.16352844238281, "n_util_obs": 393, "util_count_spearman": 0.04019368267637655}, "t": 522.3} |
| {"event": "eval", "step": 1600, "phase": 1, "bpb": {"en": 2.1933, "tr": 2.44587, "fi": 2.60243}, "bytes_per_chunk": {"en": 4.607, "tr": 5.286, "fi": 4.175}, "bpb_mean": 2.41387, "frozen_inventory": true, "t": 523.4} |
| {"event": "train", "step": 1600, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.5761, "bpb": 2.27384, "chunk_loss": 0.030826, "F": 0.21786, "G": 0.23245, "chunks": 223.09, "bytes_per_chunk": 4.59, "boundary_entropy": 0.17126, "router_noise": 0.0, "grad_norm": 0.8701, "lr": 0.0002668, "router_trainable": true, "sec_per_step": 0.293, "probe_sec": 0.0889, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07639, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.19312, "n_shadow": 0, "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": 0.97624, "trunc_abs_frac": 0.82569, "mean_delta_full": 0.50293, "read_hit_rate": {"surface": 0.74884}, "mean_gate": {"surface": 0.23634}, "n_reads": 5322, "n_cand": 7107, "read_suppressed_frac": 0.1047, "dropout_reads_logged": 5322, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 120.24594116210938, "n_util_obs": 393, "util_count_spearman": 0.04033968208764131}, "probe_overhead_sec_total": 87.445, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.013836, "alpha": 0.03, "t": 524.8} |
| {"event": "consolidate", "step": 1650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 537.3} |
| {"event": "train", "step": 1650, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.59435, "bpb": 2.30016, "chunk_loss": 0.030686, "F": 0.21198, "G": 0.22677, "chunks": 217.06, "bytes_per_chunk": 4.718, "boundary_entropy": 0.16994, "router_noise": 0.0, "grad_norm": 0.7499, "lr": 0.0002643, "router_trainable": true, "sec_per_step": 0.255, "probe_sec": 0.0481, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03746, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.02923, "n_shadow": 0, "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": -0.20111, "trunc_abs_frac": 0.50435, "mean_delta_full": 0.05383, "read_hit_rate": {"surface": 0.72693}, "mean_gate": {"surface": 0.23342}, "n_reads": 5026, "n_cand": 6914, "read_suppressed_frac": 0.0987, "dropout_reads_logged": 5026, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 124.12843322753906, "n_util_obs": 406, "util_count_spearman": 0.04512260493769344}, "probe_overhead_sec_total": 90.217, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.013061, "alpha": 0.03, "t": 537.5} |
| {"event": "consolidate", "step": 1700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 549.9} |
| {"event": "train", "step": 1700, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.61344, "bpb": 2.32771, "chunk_loss": 0.030735, "F": 0.21591, "G": 0.22899, "chunks": 221.09, "bytes_per_chunk": 4.632, "boundary_entropy": 0.16749, "router_noise": 0.0, "grad_norm": 0.7633, "lr": 0.0002618, "router_trainable": true, "sec_per_step": 0.251, "probe_sec": 0.0357, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02625, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.01719, "n_shadow": 0, "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": -14.3, "trunc_abs_frac": 0.94975, "mean_delta_full": -0.0061, "read_hit_rate": {"surface": 0.71873}, "mean_gate": {"surface": 0.22889}, "n_reads": 5062, "n_cand": 7043, "read_suppressed_frac": 0.1043, "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": 128.0264129638672, "n_util_obs": 419, "util_count_spearman": 0.03663884760265557}, "probe_overhead_sec_total": 92.944, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.013459, "alpha": 0.03, "t": 550.1} |
| {"event": "consolidate", "step": 1750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 561.1} |
| {"event": "train", "step": 1750, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.56553, "bpb": 2.25858, "chunk_loss": 0.030627, "F": 0.21063, "G": 0.22644, "chunks": 215.69, "bytes_per_chunk": 4.748, "boundary_entropy": 0.16631, "router_noise": 0.0, "grad_norm": 0.7432, "lr": 0.0002591, "router_trainable": true, "sec_per_step": 0.225, "probe_sec": 0.1117, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10462, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.37409, "n_shadow": 0, "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": 1.00152, "trunc_abs_frac": 0.63948, "mean_delta_full": 0.96602, "read_hit_rate": {"surface": 0.73508}, "mean_gate": {"surface": 0.23167}, "n_reads": 5050, "n_cand": 6870, "read_suppressed_frac": 0.1038, "dropout_reads_logged": 5050, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 131.93040466308594, "n_util_obs": 437, "util_count_spearman": 0.054503643961754784}, "probe_overhead_sec_total": 95.748, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.012717, "alpha": 0.03, "t": 561.3} |
| {"event": "consolidate", "step": 1800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 575.2} |
| {"event": "probe", "step": 1800, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3567.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.204965, "cps": 27.867188, "mean_p": 0.220096, "boundary_entropy": 0.17121, "bpb": 2.153778}, "tr": {"chunks": 3769.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.208198, "cps": 29.445312, "mean_p": 0.218179, "boundary_entropy": 0.197841, "bpb": 2.38981}, "fi": {"chunks": 4797.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.25501, "cps": 37.476562, "mean_p": 0.262788, "boundary_entropy": 0.171807, "bpb": 2.522259}}, "bpb": {"en": 2.15378, "tr": 2.38981, "fi": 2.52226}, "gini_cpc": 0.049933, "gini_cps": 0.067584, "bpic": 4.675843, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0157728412043838, "fi": 1.24416748087601}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 135.8058319091797, "n_util_obs": 448, "util_count_spearman": 0.04674098269290779}, "t": 575.8} |
| {"event": "eval", "step": 1800, "phase": 1, "bpb": {"en": 2.16536, "tr": 2.40014, "fi": 2.52995}, "bytes_per_chunk": {"en": 4.676, "tr": 5.3, "fi": 4.207}, "bpb_mean": 2.36515, "frozen_inventory": true, "t": 576.8} |
| {"event": "train", "step": 1800, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.5366, "bpb": 2.21685, "chunk_loss": 0.03062, "F": 0.20963, "G": 0.22334, "chunks": 214.66, "bytes_per_chunk": 4.77, "boundary_entropy": 0.16573, "router_noise": 0.0, "grad_norm": 0.8224, "lr": 0.0002564, "router_trainable": true, "sec_per_step": 0.332, "probe_sec": 0.0245, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0182, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.32128, "n_shadow": 0, "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": 0.94856, "trunc_abs_frac": 1.06226, "mean_delta_full": 0.71185, "read_hit_rate": {"surface": 0.73819}, "mean_gate": {"surface": 0.23219}, "n_reads": 5047, "n_cand": 6837, "read_suppressed_frac": 0.1086, "dropout_reads_logged": 5047, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 135.8839111328125, "n_util_obs": 448, "util_count_spearman": 0.045832533132329184}, "probe_overhead_sec_total": 98.464, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.013439, "alpha": 0.03, "t": 578.0} |
| {"event": "consolidate", "step": 1850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 592.2} |
| {"event": "train", "step": 1850, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.5703, "bpb": 2.26546, "chunk_loss": 0.030684, "F": 0.21478, "G": 0.22587, "chunks": 219.94, "bytes_per_chunk": 4.656, "boundary_entropy": 0.16073, "router_noise": 0.0, "grad_norm": 0.8637, "lr": 0.0002537, "router_trainable": true, "sec_per_step": 0.289, "probe_sec": 0.0862, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07898, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.04018, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 7, "trunc_ratio_sum": -1.20974, "trunc_abs_frac": 1.90534, "mean_delta_full": 0.02122, "read_hit_rate": {"surface": 0.74022}, "mean_gate": {"surface": 0.22091}, "n_reads": 5186, "n_cand": 7006, "read_suppressed_frac": 0.1018, "dropout_reads_logged": 5186, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 139.8312225341797, "n_util_obs": 458, "util_count_spearman": 0.043418323124420594}, "probe_overhead_sec_total": 101.396, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.01305, "alpha": 0.03, "t": 592.4} |
| {"event": "consolidate", "step": 1900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 605.4} |
| {"event": "train", "step": 1900, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.49604, "bpb": 2.15833, "chunk_loss": 0.030698, "F": 0.21439, "G": 0.22602, "chunks": 219.53, "bytes_per_chunk": 4.664, "boundary_entropy": 0.15825, "router_noise": 0.0, "grad_norm": 0.7015, "lr": 0.0002508, "router_trainable": true, "sec_per_step": 0.267, "probe_sec": 0.1248, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.11113, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.12144, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 2.27893, "trunc_abs_frac": 5.79708, "mean_delta_full": -0.16805, "read_hit_rate": {"surface": 0.74746}, "mean_gate": {"surface": 0.23052}, "n_reads": 5227, "n_cand": 6993, "read_suppressed_frac": 0.0972, "dropout_reads_logged": 5227, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 143.79115295410156, "n_util_obs": 474, "util_count_spearman": 0.07030466321232072}, "probe_overhead_sec_total": 103.985, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.017348, "alpha": 0.03, "t": 605.8} |
| {"event": "consolidate", "step": 1950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 623.8} |
| {"event": "train", "step": 1950, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.57768, "bpb": 2.27611, "chunk_loss": 0.030899, "F": 0.22107, "G": 0.23336, "chunks": 226.38, "bytes_per_chunk": 4.523, "boundary_entropy": 0.15727, "router_noise": 0.0, "grad_norm": 0.8038, "lr": 0.0002479, "router_trainable": true, "sec_per_step": 0.375, "probe_sec": 0.0614, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04192, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.07972, "n_shadow": 0, "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": 0.23412, "trunc_abs_frac": 0.73562, "mean_delta_full": 0.19829, "read_hit_rate": {"surface": 0.74099}, "mean_gate": {"surface": 0.22282}, "n_reads": 5344, "n_cand": 7212, "read_suppressed_frac": 0.0988, "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": 147.7613067626953, "n_util_obs": 486, "util_count_spearman": 0.0953247821571146}, "probe_overhead_sec_total": 106.966, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.015793, "alpha": 0.03, "t": 624.5} |
| {"event": "consolidate", "step": 2000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 646.4} |
| {"event": "probe", "step": 2000, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3548.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.203873, "cps": 27.71875, "mean_p": 0.216787, "boundary_entropy": 0.160863, "bpb": 2.120502}, "tr": {"chunks": 3721.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.205546, "cps": 29.070312, "mean_p": 0.213312, "boundary_entropy": 0.184366, "bpb": 2.342676}, "fi": {"chunks": 4818.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.256127, "cps": 37.640625, "mean_p": 0.25936, "boundary_entropy": 0.152429, "bpb": 2.474172}}, "bpb": {"en": 2.1205, "tr": 2.34268, "fi": 2.47417}, "gini_cpc": 0.052342, "gini_cps": 0.070048, "bpic": 4.693638, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0082068124394785, "fi": 1.2563059633665614}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 151.6230926513672, "n_util_obs": 498, "util_count_spearman": 0.09128448396183818}, "t": 647.0} |
| {"event": "eval", "step": 2000, "phase": 1, "bpb": {"en": 2.13274, "tr": 2.35281, "fi": 2.48529}, "bytes_per_chunk": {"en": 4.706, "tr": 5.432, "fi": 4.192}, "bpb_mean": 2.32361, "frozen_inventory": true, "t": 648.1} |
| {"event": "train", "step": 2000, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.54358, "bpb": 2.22692, "chunk_loss": 0.030782, "F": 0.22031, "G": 0.22971, "chunks": 225.59, "bytes_per_chunk": 4.539, "boundary_entropy": 0.15678, "router_noise": 0.0, "grad_norm": 0.8168, "lr": 0.000245, "router_trainable": true, "sec_per_step": 0.508, "probe_sec": 0.0566, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04531, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.05285, "n_shadow": 0, "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": -0.36111, "trunc_abs_frac": 2.17249, "mean_delta_full": 0.12363, "read_hit_rate": {"surface": 0.74885}, "mean_gate": {"surface": 0.2241}, "n_reads": 5382, "n_cand": 7187, "read_suppressed_frac": 0.1033, "dropout_reads_logged": 5382, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 151.706298828125, "n_util_obs": 498, "util_count_spearman": 0.09151602473387696}, "probe_overhead_sec_total": 109.697, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.014686, "alpha": 0.03, "t": 649.9} |
| {"event": "probe", "step": 2048, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3552.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.204103, "cps": 27.75, "mean_p": 0.217042, "boundary_entropy": 0.157432, "bpb": 2.110848}, "tr": {"chunks": 3821.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.21107, "cps": 29.851562, "mean_p": 0.216235, "boundary_entropy": 0.18452, "bpb": 2.347537}, "fi": {"chunks": 4838.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.25719, "cps": 37.796875, "mean_p": 0.261142, "boundary_entropy": 0.150806, "bpb": 2.470048}}, "bpb": {"en": 2.11085, "tr": 2.34754, "fi": 2.47005}, "gini_cpc": 0.052637, "gini_cps": 0.07021, "bpic": 4.645975, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0341359820158225, "fi": 1.260100383329478}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 155.44207763671875, "n_util_obs": 507, "util_count_spearman": 0.09475376350365813}, "t": 675.8} |
| {"event": "consolidate", "step": 2050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 677.6} |
| {"event": "train", "step": 2050, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.56044, "bpb": 2.25124, "chunk_loss": 0.030505, "F": 0.20474, "G": 0.21965, "chunks": 209.66, "bytes_per_chunk": 4.884, "boundary_entropy": 0.16169, "router_noise": 0.0, "grad_norm": 0.8576, "lr": 0.0002419, "router_trainable": true, "sec_per_step": 0.561, "probe_sec": 0.1094, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09056, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.02608, "n_shadow": 0, "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": -1.18501, "trunc_abs_frac": 1.06045, "mean_delta_full": 0.0416, "read_hit_rate": {"surface": 0.71529}, "mean_gate": {"surface": 0.25073}, "n_reads": 4776, "n_cand": 6677, "read_suppressed_frac": 0.1003, "dropout_reads_logged": 4776, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 155.67457580566406, "n_util_obs": 508, "util_count_spearman": 0.09983908887266726}, "probe_overhead_sec_total": 112.532, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.0159, "alpha": 0.03, "t": 677.9} |
| {"event": "consolidate", "step": 2100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 700.3} |
| {"event": "train", "step": 2100, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.48124, "bpb": 2.13697, "chunk_loss": 0.030581, "F": 0.21033, "G": 0.22136, "chunks": 215.38, "bytes_per_chunk": 4.754, "boundary_entropy": 0.1482, "router_noise": 0.0, "grad_norm": 0.7355, "lr": 0.0002388, "router_trainable": true, "sec_per_step": 0.452, "probe_sec": 0.0465, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03769, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.14603, "n_shadow": 0, "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": -0.54511, "trunc_abs_frac": 35.5985, "mean_delta_full": 0.07442, "read_hit_rate": {"surface": 0.74577}, "mean_gate": {"surface": 0.22998}, "n_reads": 5116, "n_cand": 6860, "read_suppressed_frac": 0.1007, "dropout_reads_logged": 5116, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 159.6670379638672, "n_util_obs": 518, "util_count_spearman": 0.10239344216721051}, "probe_overhead_sec_total": 115.162, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.015605, "alpha": 0.03, "t": 700.6} |
| {"event": "consolidate", "step": 2150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 723.8} |
| {"event": "train", "step": 2150, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.49174, "bpb": 2.15212, "chunk_loss": 0.030728, "F": 0.21725, "G": 0.22536, "chunks": 222.47, "bytes_per_chunk": 4.603, "boundary_entropy": 0.14176, "router_noise": 0.0, "grad_norm": 0.7543, "lr": 0.0002357, "router_trainable": true, "sec_per_step": 0.477, "probe_sec": 0.036, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02261, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.10612, "n_shadow": 0, "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": 0.422, "trunc_abs_frac": 0.43197, "mean_delta_full": 0.28778, "read_hit_rate": {"surface": 0.74376}, "mean_gate": {"surface": 0.22797}, "n_reads": 5271, "n_cand": 7087, "read_suppressed_frac": 0.1002, "dropout_reads_logged": 5271, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 163.65013122558594, "n_util_obs": 533, "util_count_spearman": 0.11079077455279929}, "probe_overhead_sec_total": 118.478, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.014203, "alpha": 0.03, "t": 724.4} |
| {"event": "consolidate", "step": 2200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 736.4} |
| {"event": "probe", "step": 2200, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3527.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.202666, "cps": 27.554688, "mean_p": 0.211871, "boundary_entropy": 0.146896, "bpb": 2.095312}, "tr": {"chunks": 3727.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.205877, "cps": 29.117188, "mean_p": 0.209663, "boundary_entropy": 0.170309, "bpb": 2.30272}, "fi": {"chunks": 4837.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.257137, "cps": 37.789062, "mean_p": 0.258316, "boundary_entropy": 0.141631, "bpb": 2.450062}}, "bpb": {"en": 2.09531, "tr": 2.30272, "fi": 2.45006}, "gini_cpc": 0.054551, "gini_cps": 0.07223, "bpic": 4.692085, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0158451275277478, "fi": 1.268769892658979}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 167.52195739746094, "n_util_obs": 545, "util_count_spearman": 0.11280371061604752}, "t": 737.1} |
| {"event": "eval", "step": 2200, "phase": 1, "bpb": {"en": 2.10974, "tr": 2.31143, "fi": 2.4617}, "bytes_per_chunk": {"en": 4.759, "tr": 5.335, "fi": 4.155}, "bpb_mean": 2.29429, "frozen_inventory": true, "t": 738.2} |
| {"event": "train", "step": 2200, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.53016, "bpb": 2.20756, "chunk_loss": 0.030776, "F": 0.21555, "G": 0.22489, "chunks": 220.72, "bytes_per_chunk": 4.639, "boundary_entropy": 0.14392, "router_noise": 0.0, "grad_norm": 0.7504, "lr": 0.0002325, "router_trainable": true, "sec_per_step": 0.301, "probe_sec": 0.075, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06921, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.166, "n_shadow": 0, "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": 0.01595, "trunc_abs_frac": 1.77221, "mean_delta_full": 0.24036, "read_hit_rate": {"surface": 0.73958}, "mean_gate": {"surface": 0.22265}, "n_reads": 5200, "n_cand": 7031, "read_suppressed_frac": 0.101, "dropout_reads_logged": 5200, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 167.60194396972656, "n_util_obs": 545, "util_count_spearman": 0.11296764258350867}, "probe_overhead_sec_total": 121.5, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.020378, "alpha": 0.03, "t": 739.5} |
| {"event": "consolidate", "step": 2250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 750.7} |
| {"event": "train", "step": 2250, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.4979, "bpb": 2.16102, "chunk_loss": 0.030654, "F": 0.2153, "G": 0.22205, "chunks": 220.47, "bytes_per_chunk": 4.645, "boundary_entropy": 0.142, "router_noise": 0.0, "grad_norm": 0.755, "lr": 0.0002293, "router_trainable": true, "sec_per_step": 0.228, "probe_sec": 0.0465, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03933, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.04244, "n_shadow": 0, "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": 0.95167, "trunc_abs_frac": 0.92415, "mean_delta_full": 0.33215, "read_hit_rate": {"surface": 0.75367}, "mean_gate": {"surface": 0.22104}, "n_reads": 5293, "n_cand": 7023, "read_suppressed_frac": 0.1024, "dropout_reads_logged": 5293, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 171.5417938232422, "n_util_obs": 554, "util_count_spearman": 0.10475385042691701}, "probe_overhead_sec_total": 124.222, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.014547, "alpha": 0.03, "t": 750.9} |
| {"event": "consolidate", "step": 2300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 762.6} |
| {"event": "train", "step": 2300, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.5423, "bpb": 2.22507, "chunk_loss": 0.0306, "F": 0.20984, "G": 0.22069, "chunks": 214.88, "bytes_per_chunk": 4.766, "boundary_entropy": 0.14713, "router_noise": 0.0, "grad_norm": 0.7006, "lr": 0.000226, "router_trainable": true, "sec_per_step": 0.244, "probe_sec": 0.0439, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03758, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.12908, "n_shadow": 0, "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": 0.44625, "trunc_abs_frac": 0.974, "mean_delta_full": 0.20093, "read_hit_rate": {"surface": 0.7333}, "mean_gate": {"surface": 0.22627}, "n_reads": 5018, "n_cand": 6843, "read_suppressed_frac": 0.1, "dropout_reads_logged": 5018, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 175.49810791015625, "n_util_obs": 565, "util_count_spearman": 0.09771158125091886}, "probe_overhead_sec_total": 127.224, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.013712, "alpha": 0.03, "t": 763.1} |
| {"event": "consolidate", "step": 2350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 773.9} |
| {"event": "train", "step": 2350, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.46276, "bpb": 2.11032, "chunk_loss": 0.030629, "F": 0.21353, "G": 0.22072, "chunks": 218.66, "bytes_per_chunk": 4.683, "boundary_entropy": 0.13676, "router_noise": 0.0, "grad_norm": 0.72, "lr": 0.0002226, "router_trainable": true, "sec_per_step": 0.22, "probe_sec": 0.0377, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0315, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.20201, "n_shadow": 0, "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": 0.47092, "trunc_abs_frac": 0.3795, "mean_delta_full": 0.69055, "read_hit_rate": {"surface": 0.7476}, "mean_gate": {"surface": 0.21442}, "n_reads": 5207, "n_cand": 6965, "read_suppressed_frac": 0.101, "dropout_reads_logged": 5207, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 179.41920471191406, "n_util_obs": 582, "util_count_spearman": 0.09281651442463318}, "probe_overhead_sec_total": 130.025, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.015263, "alpha": 0.03, "t": 774.1} |
| {"event": "consolidate", "step": 2400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 785.9} |
| {"event": "probe", "step": 2400, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3550.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.203988, "cps": 27.734375, "mean_p": 0.213208, "boundary_entropy": 0.145607, "bpb": 2.071286}, "tr": {"chunks": 3643.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.201237, "cps": 28.460938, "mean_p": 0.209314, "boundary_entropy": 0.169601, "bpb": 2.269354}, "fi": {"chunks": 4797.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.25501, "cps": 37.476562, "mean_p": 0.261231, "boundary_entropy": 0.149232, "bpb": 2.405839}}, "bpb": {"en": 2.07129, "tr": 2.26935, "fi": 2.40584}, "gini_cpc": 0.054297, "gini_cps": 0.069336, "bpic": 4.73161, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9865165761180351, "fi": 1.250125465995698}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 183.2310028076172, "n_util_obs": 590, "util_count_spearman": 0.10733149694716586}, "t": 786.6} |
| {"event": "eval", "step": 2400, "phase": 1, "bpb": {"en": 2.08349, "tr": 2.2776, "fi": 2.41897}, "bytes_per_chunk": {"en": 4.699, "tr": 5.507, "fi": 4.192}, "bpb_mean": 2.26002, "frozen_inventory": true, "t": 787.6} |
| {"event": "phase_transition", "step": 2400, "phase": 2, "objective": "global", "router_trainable": true, "encoder_trainable": true, "frozen_router_params": 0, "t": 788.6} |
| {"event": "train", "step": 2400, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.46268, "bpb": 2.1102, "chunk_loss": 0.03055, "F": 0.20987, "G": 0.21985, "chunks": 214.91, "bytes_per_chunk": 4.765, "boundary_entropy": 0.14621, "router_noise": 0.0, "grad_norm": 0.6558, "lr": 0.0002192, "router_trainable": true, "sec_per_step": 0.296, "probe_sec": 0.0534, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04334, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.01488, "n_shadow": 0, "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": -0.10135, "trunc_abs_frac": 0.35584, "mean_delta_full": -0.20455, "read_hit_rate": {"surface": 0.74054}, "mean_gate": {"surface": 0.21605}, "n_reads": 5069, "n_cand": 6845, "read_suppressed_frac": 0.0971, "dropout_reads_logged": 5069, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 183.30706787109375, "n_util_obs": 591, "util_count_spearman": 0.10793251136282839}, "probe_overhead_sec_total": 132.55, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.016737, "alpha": 0.03, "t": 788.9} |
| {"event": "consolidate", "step": 2450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 800.6} |
| {"event": "train", "step": 2450, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.48139, "bpb": 2.1372, "chunk_loss": 0.030832, "F": 0.22141, "G": 0.22973, "chunks": 226.72, "bytes_per_chunk": 4.517, "boundary_entropy": 0.14567, "router_noise": 0.0, "grad_norm": 0.8831, "lr": 0.0002158, "router_trainable": true, "sec_per_step": 0.237, "probe_sec": 0.0299, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02356, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.12859, "n_shadow": 0, "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": 0.72089, "trunc_abs_frac": 0.91184, "mean_delta_full": 0.26335, "read_hit_rate": {"surface": 0.74664}, "mean_gate": {"surface": 0.21595}, "n_reads": 5393, "n_cand": 7223, "read_suppressed_frac": 0.0957, "dropout_reads_logged": 5393, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 187.18777465820312, "n_util_obs": 610, "util_count_spearman": 0.11778570518958603}, "probe_overhead_sec_total": 135.387, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.013944, "alpha": 0.03, "t": 800.7} |
| {"event": "consolidate", "step": 2500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 812.0} |
| {"event": "train", "step": 2500, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.57974, "bpb": 2.27908, "chunk_loss": 0.030715, "F": 0.21796, "G": 0.22496, "chunks": 223.19, "bytes_per_chunk": 4.588, "boundary_entropy": 0.14071, "router_noise": 0.0, "grad_norm": 0.7409, "lr": 0.0002123, "router_trainable": true, "sec_per_step": 0.23, "probe_sec": 0.0316, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02514, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.39804, "n_shadow": 0, "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": 1.0588, "trunc_abs_frac": 0.71842, "mean_delta_full": 0.54785, "read_hit_rate": {"surface": 0.73446}, "mean_gate": {"surface": 0.21357}, "n_reads": 5222, "n_cand": 7110, "read_suppressed_frac": 0.0963, "dropout_reads_logged": 5222, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 191.11904907226562, "n_util_obs": 625, "util_count_spearman": 0.11281774706752543}, "probe_overhead_sec_total": 138.118, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.015619, "alpha": 0.03, "t": 812.2} |
| {"event": "consolidate", "step": 2550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 823.3} |
| {"event": "train", "step": 2550, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.50016, "bpb": 2.16427, "chunk_loss": 0.030567, "F": 0.21161, "G": 0.22134, "chunks": 216.69, "bytes_per_chunk": 4.726, "boundary_entropy": 0.13599, "router_noise": 0.0, "grad_norm": 0.724, "lr": 0.0002088, "router_trainable": true, "sec_per_step": 0.226, "probe_sec": 0.0708, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0621, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.01636, "n_shadow": 0, "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": 0.05101, "trunc_abs_frac": 0.68096, "mean_delta_full": -0.17334, "read_hit_rate": {"surface": 0.73703}, "mean_gate": {"surface": 0.21771}, "n_reads": 5087, "n_cand": 6902, "read_suppressed_frac": 0.0957, "dropout_reads_logged": 5087, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 195.03367614746094, "n_util_obs": 637, "util_count_spearman": 0.10746174986458987}, "probe_overhead_sec_total": 141.122, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.015396, "alpha": 0.03, "t": 823.6} |
| {"event": "consolidate", "step": 2600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 835.4} |
| {"event": "probe", "step": 2600, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3558.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.204448, "cps": 27.796875, "mean_p": 0.21211, "boundary_entropy": 0.135314, "bpb": 2.052752}, "tr": {"chunks": 3803.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.210076, "cps": 29.710938, "mean_p": 0.208574, "boundary_entropy": 0.158239, "bpb": 2.244388}, "fi": {"chunks": 4847.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.257668, "cps": 37.867188, "mean_p": 0.256968, "boundary_entropy": 0.130408, "bpb": 2.36419}}, "bpb": {"en": 2.05275, "tr": 2.24439, "fi": 2.36419}, "gini_cpc": 0.052783, "gini_cps": 0.070391, "bpic": 4.647117, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.027528674917064, "fi": 1.260315602357811}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 198.8527069091797, "n_util_obs": 648, "util_count_spearman": 0.10701024429851806}, "t": 836.0} |
| {"event": "eval", "step": 2600, "phase": 2, "bpb": {"en": 2.06562, "tr": 2.25247, "fi": 2.37574}, "bytes_per_chunk": {"en": 4.726, "tr": 5.244, "fi": 4.149}, "bpb_mean": 2.23127, "frozen_inventory": true, "t": 837.0} |
| {"event": "train", "step": 2600, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.50581, "bpb": 2.17242, "chunk_loss": 0.030693, "F": 0.21442, "G": 0.2226, "chunks": 219.56, "bytes_per_chunk": 4.664, "boundary_entropy": 0.12958, "router_noise": 0.0, "grad_norm": 0.6738, "lr": 0.0002053, "router_trainable": true, "sec_per_step": 0.295, "probe_sec": 0.0739, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05969, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.18369, "n_shadow": 0, "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": 0.83503, "trunc_abs_frac": 1.12722, "mean_delta_full": 0.94586, "read_hit_rate": {"surface": 0.72905}, "mean_gate": {"surface": 0.21631}, "n_reads": 5099, "n_cand": 6994, "read_suppressed_frac": 0.1004, "dropout_reads_logged": 5099, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 198.928955078125, "n_util_obs": 648, "util_count_spearman": 0.10752403738470118}, "probe_overhead_sec_total": 143.78, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.017886, "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": 849.2} |
| {"event": "train", "step": 2650, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.44208, "bpb": 2.08049, "chunk_loss": 0.030461, "F": 0.20691, "G": 0.21485, "chunks": 211.88, "bytes_per_chunk": 4.833, "boundary_entropy": 0.13529, "router_noise": 0.0, "grad_norm": 0.6881, "lr": 0.0002017, "router_trainable": true, "sec_per_step": 0.222, "probe_sec": 0.0291, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0225, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.01805, "n_shadow": 0, "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": 0.51359, "trunc_abs_frac": 0.9247, "mean_delta_full": 0.09769, "read_hit_rate": {"surface": 0.73592}, "mean_gate": {"surface": 0.21469}, "n_reads": 4966, "n_cand": 6748, "read_suppressed_frac": 0.0954, "dropout_reads_logged": 4966, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 202.84385681152344, "n_util_obs": 664, "util_count_spearman": 0.11534680524844598}, "probe_overhead_sec_total": 146.124, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.014376, "alpha": 0.03, "t": 849.4} |
| {"event": "consolidate", "step": 2700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 861.0} |
| {"event": "train", "step": 2700, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.52493, "bpb": 2.20001, "chunk_loss": 0.030625, "F": 0.21088, "G": 0.22007, "chunks": 215.94, "bytes_per_chunk": 4.742, "boundary_entropy": 0.13724, "router_noise": 0.0, "grad_norm": 0.7583, "lr": 0.0001981, "router_trainable": true, "sec_per_step": 0.235, "probe_sec": 0.0235, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01779, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.1054, "n_shadow": 0, "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": -0.49166, "trunc_abs_frac": 0.49166, "mean_delta_full": 0.09509, "read_hit_rate": {"surface": 0.71736}, "mean_gate": {"surface": 0.21272}, "n_reads": 4934, "n_cand": 6878, "read_suppressed_frac": 0.0973, "dropout_reads_logged": 4934, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 206.73373413085938, "n_util_obs": 681, "util_count_spearman": 0.11714138663307866}, "probe_overhead_sec_total": 148.7, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.018317, "alpha": 0.03, "t": 861.1} |
| {"event": "consolidate", "step": 2750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 875.0} |
| {"event": "train", "step": 2750, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.45992, "bpb": 2.10622, "chunk_loss": 0.030557, "F": 0.21085, "G": 0.2191, "chunks": 215.91, "bytes_per_chunk": 4.743, "boundary_entropy": 0.13623, "router_noise": 0.0, "grad_norm": 0.6976, "lr": 0.0001945, "router_trainable": true, "sec_per_step": 0.282, "probe_sec": 0.0477, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03997, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.09961, "n_shadow": 0, "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": -0.06634, "trunc_abs_frac": 0.06634, "mean_delta_full": 0.31628, "read_hit_rate": {"surface": 0.73142}, "mean_gate": {"surface": 0.21718}, "n_reads": 5030, "n_cand": 6877, "read_suppressed_frac": 0.1044, "dropout_reads_logged": 5030, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 210.67190551757812, "n_util_obs": 691, "util_count_spearman": 0.11789747116879867}, "probe_overhead_sec_total": 151.833, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.016584, "alpha": 0.03, "t": 875.3} |
| {"event": "consolidate", "step": 2800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 892.8} |
| {"event": "probe", "step": 2800, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3509.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.201632, "cps": 27.414062, "mean_p": 0.208378, "boundary_entropy": 0.132401, "bpb": 2.033659}, "tr": {"chunks": 3611.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.19947, "cps": 28.210938, "mean_p": 0.20452, "boundary_entropy": 0.158795, "bpb": 2.236396}, "fi": {"chunks": 4826.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.256552, "cps": 37.703125, "mean_p": 0.255353, "boundary_entropy": 0.126055, "bpb": 2.336903}}, "bpb": {"en": 2.03366, "tr": 2.2364, "fi": 2.3369}, "gini_cpc": 0.057865, "gini_cps": 0.073497, "bpic": 4.749037, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9892764916477192, "fi": 1.2723781018664173}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 214.4933319091797, "n_util_obs": 700, "util_count_spearman": 0.10828371310503596}, "t": 893.5} |
| {"event": "eval", "step": 2800, "phase": 2, "bpb": {"en": 2.04626, "tr": 2.24406, "fi": 2.34971}, "bytes_per_chunk": {"en": 4.761, "tr": 5.495, "fi": 4.164}, "bpb_mean": 2.21334, "frozen_inventory": true, "t": 894.4} |
| {"event": "train", "step": 2800, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.4525, "bpb": 2.09552, "chunk_loss": 0.030655, "F": 0.21579, "G": 0.22117, "chunks": 220.97, "bytes_per_chunk": 4.634, "boundary_entropy": 0.13172, "router_noise": 0.0, "grad_norm": 0.7021, "lr": 0.0001909, "router_trainable": true, "sec_per_step": 0.415, "probe_sec": 0.108, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09569, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.23455, "n_shadow": 0, "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": 0.13889, "trunc_abs_frac": 0.37825, "mean_delta_full": 0.18178, "read_hit_rate": {"surface": 0.75593}, "mean_gate": {"surface": 0.20017}, "n_reads": 5321, "n_cand": 7039, "read_suppressed_frac": 0.0977, "dropout_reads_logged": 5321, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 214.5753631591797, "n_util_obs": 701, "util_count_spearman": 0.1107989699868368}, "probe_overhead_sec_total": 155.16, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.0152, "alpha": 0.03, "t": 896.0} |
| {"event": "consolidate", "step": 2850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 914.7} |
| {"event": "train", "step": 2850, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.55559, "bpb": 2.24424, "chunk_loss": 0.030746, "F": 0.21555, "G": 0.22367, "chunks": 220.72, "bytes_per_chunk": 4.639, "boundary_entropy": 0.13653, "router_noise": 0.0, "grad_norm": 0.6922, "lr": 0.0001872, "router_trainable": true, "sec_per_step": 0.378, "probe_sec": 0.0435, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03678, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.23254, "n_shadow": 0, "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": 2.04124, "trunc_abs_frac": 2.04124, "mean_delta_full": 0.28711, "read_hit_rate": {"surface": 0.72223}, "mean_gate": {"surface": 0.20222}, "n_reads": 5078, "n_cand": 7031, "read_suppressed_frac": 0.1067, "dropout_reads_logged": 5078, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 218.4311065673828, "n_util_obs": 713, "util_count_spearman": 0.11127123724446122}, "probe_overhead_sec_total": 157.934, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.020724, "alpha": 0.03, "t": 914.9} |
| {"event": "consolidate", "step": 2900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 937.7} |
| {"event": "train", "step": 2900, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.50203, "bpb": 2.16698, "chunk_loss": 0.030832, "F": 0.22168, "G": 0.22706, "chunks": 227.0, "bytes_per_chunk": 4.511, "boundary_entropy": 0.1325, "router_noise": 0.0, "grad_norm": 0.674, "lr": 0.0001835, "router_trainable": true, "sec_per_step": 0.469, "probe_sec": 0.1163, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10649, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.1712, "n_shadow": 0, "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": 0.31947, "trunc_abs_frac": 1.01948, "mean_delta_full": 0.15239, "read_hit_rate": {"surface": 0.75194}, "mean_gate": {"surface": 0.20668}, "n_reads": 5438, "n_cand": 7232, "read_suppressed_frac": 0.101, "dropout_reads_logged": 5438, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 222.33689880371094, "n_util_obs": 728, "util_count_spearman": 0.10530748230355959}, "probe_overhead_sec_total": 161.026, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.019566, "alpha": 0.03, "t": 938.3} |
| {"event": "consolidate", "step": 2950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 962.1} |
| {"event": "train", "step": 2950, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.43403, "bpb": 2.06887, "chunk_loss": 0.030547, "F": 0.20941, "G": 0.21589, "chunks": 214.44, "bytes_per_chunk": 4.775, "boundary_entropy": 0.12927, "router_noise": 0.0, "grad_norm": 0.6482, "lr": 0.0001799, "router_trainable": true, "sec_per_step": 0.483, "probe_sec": 0.0501, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03709, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.11526, "n_shadow": 0, "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": -0.64664, "trunc_abs_frac": 0.64243, "mean_delta_full": -0.08533, "read_hit_rate": {"surface": 0.72723}, "mean_gate": {"surface": 0.21196}, "n_reads": 4967, "n_cand": 6830, "read_suppressed_frac": 0.0932, "dropout_reads_logged": 4967, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 226.2337188720703, "n_util_obs": 736, "util_count_spearman": 0.10502286915471695}, "probe_overhead_sec_total": 163.749, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.020585, "alpha": 0.03, "t": 962.5} |
| {"event": "consolidate", "step": 3000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 987.3} |
| {"event": "probe", "step": 3000, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3564.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.204792, "cps": 27.84375, "mean_p": 0.209367, "boundary_entropy": 0.128508, "bpb": 2.014435}, "tr": {"chunks": 3636.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.200851, "cps": 28.40625, "mean_p": 0.203628, "boundary_entropy": 0.1514, "bpb": 2.195659}, "fi": {"chunks": 4852.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.257934, "cps": 37.90625, "mean_p": 0.257306, "boundary_entropy": 0.120853, "bpb": 2.315047}}, "bpb": {"en": 2.01443, "tr": 2.19566, "fi": 2.31505}, "gini_cpc": 0.057349, "gini_cps": 0.071247, "bpic": 4.707269, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9807532319270705, "fi": 1.2594917688214164}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 230.05821228027344, "n_util_obs": 749, "util_count_spearman": 0.10541797922801682}, "t": 988.0} |
| {"event": "eval", "step": 3000, "phase": 2, "bpb": {"en": 2.02878, "tr": 2.20359, "fi": 2.32494}, "bytes_per_chunk": {"en": 4.708, "tr": 5.471, "fi": 4.154}, "bpb_mean": 2.18577, "frozen_inventory": true, "t": 988.9} |
| {"event": "train", "step": 3000, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.46229, "bpb": 2.10964, "chunk_loss": 0.030566, "F": 0.21262, "G": 0.21873, "chunks": 217.72, "bytes_per_chunk": 4.703, "boundary_entropy": 0.12567, "router_noise": 0.0, "grad_norm": 0.6753, "lr": 0.0001761, "router_trainable": true, "sec_per_step": 0.576, "probe_sec": 0.05, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02971, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.22114, "n_shadow": 0, "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": 0.33747, "trunc_abs_frac": 0.6389, "mean_delta_full": -0.12665, "read_hit_rate": {"surface": 0.73641}, "mean_gate": {"surface": 0.20829}, "n_reads": 5107, "n_cand": 6935, "read_suppressed_frac": 0.1005, "dropout_reads_logged": 5107, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 230.13714599609375, "n_util_obs": 749, "util_count_spearman": 0.10553611703945603}, "probe_overhead_sec_total": 166.429, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.024366, "alpha": 0.03, "t": 991.3} |
| {"event": "consolidate", "step": 3050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1014.1} |
| {"event": "train", "step": 3050, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.45645, "bpb": 2.10122, "chunk_loss": 0.030568, "F": 0.21182, "G": 0.21987, "chunks": 216.91, "bytes_per_chunk": 4.721, "boundary_entropy": 0.12719, "router_noise": 0.0, "grad_norm": 0.7542, "lr": 0.0001724, "router_trainable": true, "sec_per_step": 0.471, "probe_sec": 0.0562, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02609, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.09865, "n_shadow": 0, "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": 0.136, "trunc_abs_frac": 1.02917, "mean_delta_full": 0.08667, "read_hit_rate": {"surface": 0.73368}, "mean_gate": {"surface": 0.21574}, "n_reads": 5069, "n_cand": 6909, "read_suppressed_frac": 0.0996, "dropout_reads_logged": 5069, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 234.02969360351562, "n_util_obs": 759, "util_count_spearman": 0.1093965210607253}, "probe_overhead_sec_total": 169.199, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.015718, "alpha": 0.03, "t": 1014.8} |
| {"event": "consolidate", "step": 3100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1035.8} |
| {"event": "train", "step": 3100, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.54654, "bpb": 2.23118, "chunk_loss": 0.030609, "F": 0.2095, "G": 0.21719, "chunks": 214.53, "bytes_per_chunk": 4.773, "boundary_entropy": 0.12899, "router_noise": 0.0, "grad_norm": 0.7197, "lr": 0.0001687, "router_trainable": true, "sec_per_step": 0.423, "probe_sec": 0.0534, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04561, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.15336, "n_shadow": 0, "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": 1.56689, "trunc_abs_frac": 0.67622, "mean_delta_full": 0.12775, "read_hit_rate": {"surface": 0.72223}, "mean_gate": {"surface": 0.2128}, "n_reads": 4935, "n_cand": 6833, "read_suppressed_frac": 0.0936, "dropout_reads_logged": 4935, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 237.9192657470703, "n_util_obs": 771, "util_count_spearman": 0.125964250859845}, "probe_overhead_sec_total": 172.296, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.018972, "alpha": 0.03, "t": 1036.0} |
| {"event": "consolidate", "step": 3150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1046.9} |
| {"event": "train", "step": 3150, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.46202, "bpb": 2.10925, "chunk_loss": 0.030594, "F": 0.20996, "G": 0.21797, "chunks": 215.0, "bytes_per_chunk": 4.763, "boundary_entropy": 0.12435, "router_noise": 0.0, "grad_norm": 0.6752, "lr": 0.000165, "router_trainable": true, "sec_per_step": 0.225, "probe_sec": 0.0857, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07846, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.01027, "n_shadow": 0, "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": 0.72649}, "mean_gate": {"surface": 0.20779}, "n_reads": 4975, "n_cand": 6848, "read_suppressed_frac": 0.0999, "dropout_reads_logged": 4975, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 241.8436737060547, "n_util_obs": 784, "util_count_spearman": 0.1380409814317178}, "probe_overhead_sec_total": 174.908, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.021563, "alpha": 0.03, "t": 1047.2} |
| {"event": "consolidate", "step": 3200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1058.2} |
| {"event": "probe", "step": 3200, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3563.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.204735, "cps": 27.835938, "mean_p": 0.208807, "boundary_entropy": 0.121485, "bpb": 1.998}, "tr": {"chunks": 3722.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.205601, "cps": 29.078125, "mean_p": 0.20467, "boundary_entropy": 0.147717, "bpb": 2.180172}, "fi": {"chunks": 4871.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.258944, "cps": 38.054688, "mean_p": 0.258854, "boundary_entropy": 0.121981, "bpb": 2.290483}}, "bpb": {"en": 1.998, "tr": 2.18017, "fi": 2.29048}, "gini_cpc": 0.053998, "gini_cps": 0.071734, "bpic": 4.666996, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0042321366576255, "fi": 1.264778702627894}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 245.6544952392578, "n_util_obs": 791, "util_count_spearman": 0.13062685333405621}, "t": 1058.8} |
| {"event": "eval", "step": 3200, "phase": 2, "bpb": {"en": 2.01247, "tr": 2.18916, "fi": 2.30066}, "bytes_per_chunk": {"en": 4.71, "tr": 5.39, "fi": 4.146}, "bpb_mean": 2.16743, "frozen_inventory": true, "t": 1059.9} |
| {"event": "train", "step": 3200, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.43111, "bpb": 2.06466, "chunk_loss": 0.03079, "F": 0.21973, "G": 0.22534, "chunks": 225.0, "bytes_per_chunk": 4.551, "boundary_entropy": 0.12204, "router_noise": 0.0, "grad_norm": 0.671, "lr": 0.0001613, "router_trainable": true, "sec_per_step": 0.281, "probe_sec": 0.0548, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04567, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.0948, "n_shadow": 0, "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": -0.83133, "trunc_abs_frac": 0.93987, "mean_delta_full": 0.07186, "read_hit_rate": {"surface": 0.75181}, "mean_gate": {"surface": 0.20351}, "n_reads": 5389, "n_cand": 7168, "read_suppressed_frac": 0.0987, "dropout_reads_logged": 5389, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 245.73727416992188, "n_util_obs": 791, "util_count_spearman": 0.1301663948854617}, "probe_overhead_sec_total": 177.904, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.019063, "alpha": 0.03, "t": 1061.3} |
| {"event": "consolidate", "step": 3250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1073.0} |
| {"event": "train", "step": 3250, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.40952, "bpb": 2.03351, "chunk_loss": 0.030455, "F": 0.20624, "G": 0.21278, "chunks": 211.19, "bytes_per_chunk": 4.849, "boundary_entropy": 0.12605, "router_noise": 0.0, "grad_norm": 0.6693, "lr": 0.0001576, "router_trainable": true, "sec_per_step": 0.239, "probe_sec": 0.0363, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02822, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.25376, "n_shadow": 0, "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": 0.8313, "trunc_abs_frac": 0.82246, "mean_delta_full": 1.20294, "read_hit_rate": {"surface": 0.73788}, "mean_gate": {"surface": 0.1991}, "n_reads": 4963, "n_cand": 6726, "read_suppressed_frac": 0.1018, "dropout_reads_logged": 4963, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 249.65707397460938, "n_util_obs": 801, "util_count_spearman": 0.13671844239284614}, "probe_overhead_sec_total": 180.819, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.024089, "alpha": 0.03, "t": 1073.2} |
| {"event": "consolidate", "step": 3300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1084.8} |
| {"event": "train", "step": 3300, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.45481, "bpb": 2.09885, "chunk_loss": 0.030636, "F": 0.21347, "G": 0.21848, "chunks": 218.59, "bytes_per_chunk": 4.684, "boundary_entropy": 0.12465, "router_noise": 0.0, "grad_norm": 0.6685, "lr": 0.0001539, "router_trainable": true, "sec_per_step": 0.235, "probe_sec": 0.028, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02278, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.19346, "n_shadow": 0, "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": 0.64641, "trunc_abs_frac": 0.94023, "mean_delta_full": 0.24503, "read_hit_rate": {"surface": 0.73374}, "mean_gate": {"surface": 0.19674}, "n_reads": 5109, "n_cand": 6963, "read_suppressed_frac": 0.1026, "dropout_reads_logged": 5109, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 253.5690460205078, "n_util_obs": 816, "util_count_spearman": 0.15299590896883977}, "probe_overhead_sec_total": 183.58, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.024639, "alpha": 0.03, "t": 1085.0} |
| {"event": "consolidate", "step": 3350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1096.3} |
| {"event": "train", "step": 3350, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.48383, "bpb": 2.14071, "chunk_loss": 0.030728, "F": 0.21725, "G": 0.22259, "chunks": 222.47, "bytes_per_chunk": 4.603, "boundary_entropy": 0.12758, "router_noise": 0.0, "grad_norm": 0.74, "lr": 0.0001501, "router_trainable": true, "sec_per_step": 0.231, "probe_sec": 0.0244, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01779, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.04142, "n_shadow": 0, "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": 0.69045, "trunc_abs_frac": 1.35403, "mean_delta_full": -0.07365, "read_hit_rate": {"surface": 0.73783}, "mean_gate": {"surface": 0.20214}, "n_reads": 5229, "n_cand": 7087, "read_suppressed_frac": 0.1017, "dropout_reads_logged": 5229, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 257.4782409667969, "n_util_obs": 826, "util_count_spearman": 0.15700204424183542}, "probe_overhead_sec_total": 186.466, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.016688, "alpha": 0.03, "t": 1096.5} |
| {"event": "consolidate", "step": 3400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1108.7} |
| {"event": "probe", "step": 3400, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3530.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.202839, "cps": 27.578125, "mean_p": 0.207809, "boundary_entropy": 0.122798, "bpb": 1.984326}, "tr": {"chunks": 3637.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.200906, "cps": 28.414062, "mean_p": 0.201166, "boundary_entropy": 0.149424, "bpb": 2.160456}, "fi": {"chunks": 4837.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.257137, "cps": 37.789062, "mean_p": 0.254437, "boundary_entropy": 0.115926, "bpb": 2.268058}}, "bpb": {"en": 1.98433, "tr": 2.16046, "fi": 2.26806}, "gini_cpc": 0.056723, "gini_cps": 0.072587, "bpic": 4.726091, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9904719124543706, "fi": 1.267691617962668}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 261.33135986328125, "n_util_obs": 834, "util_count_spearman": 0.15313714290061448}, "t": 1109.5} |
| {"event": "eval", "step": 3400, "phase": 2, "bpb": {"en": 1.99753, "tr": 2.17002, "fi": 2.27618}, "bytes_per_chunk": {"en": 4.72, "tr": 5.479, "fi": 4.156}, "bpb_mean": 2.14791, "frozen_inventory": true, "t": 1110.5} |
| {"event": "train", "step": 3400, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.39987, "bpb": 2.01958, "chunk_loss": 0.030524, "F": 0.20877, "G": 0.21302, "chunks": 213.78, "bytes_per_chunk": 4.79, "boundary_entropy": 0.11989, "router_noise": 0.0, "grad_norm": 0.7021, "lr": 0.0001465, "router_trainable": true, "sec_per_step": 0.307, "probe_sec": 0.0284, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02227, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.10048, "n_shadow": 0, "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": 1.0828, "trunc_abs_frac": 1.18914, "mean_delta_full": -0.30863, "read_hit_rate": {"surface": 0.74357}, "mean_gate": {"surface": 0.19529}, "n_reads": 5063, "n_cand": 6809, "read_suppressed_frac": 0.1021, "dropout_reads_logged": 5063, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 261.40972900390625, "n_util_obs": 834, "util_count_spearman": 0.15321266632617458}, "probe_overhead_sec_total": 189.4, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.018694, "alpha": 0.03, "t": 1111.8} |
| {"event": "consolidate", "step": 3450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1122.9} |
| {"event": "train", "step": 3450, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.39164, "bpb": 2.00771, "chunk_loss": 0.030467, "F": 0.20853, "G": 0.21311, "chunks": 213.53, "bytes_per_chunk": 4.796, "boundary_entropy": 0.1169, "router_noise": 0.0, "grad_norm": 0.5993, "lr": 0.0001428, "router_trainable": true, "sec_per_step": 0.229, "probe_sec": 0.1041, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09384, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.12472, "n_shadow": 0, "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": 0.65201, "trunc_abs_frac": 0.54516, "mean_delta_full": -1.25861, "read_hit_rate": {"surface": 0.74239}, "mean_gate": {"surface": 0.18734}, "n_reads": 5049, "n_cand": 6801, "read_suppressed_frac": 0.0883, "dropout_reads_logged": 5049, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 265.32269287109375, "n_util_obs": 851, "util_count_spearman": 0.15116504873698622}, "probe_overhead_sec_total": 192.356, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.01795, "alpha": 0.03, "t": 1123.3} |
| {"event": "consolidate", "step": 3500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1134.5} |
| {"event": "train", "step": 3500, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.48195, "bpb": 2.138, "chunk_loss": 0.030701, "F": 0.21375, "G": 0.2188, "chunks": 218.88, "bytes_per_chunk": 4.678, "boundary_entropy": 0.11865, "router_noise": 0.0, "grad_norm": 0.6686, "lr": 0.0001391, "router_trainable": true, "sec_per_step": 0.23, "probe_sec": 0.0439, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03365, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.11904, "n_shadow": 0, "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": 1.38898, "trunc_abs_frac": 1.93011, "mean_delta_full": 0.38989, "read_hit_rate": {"surface": 0.72791}, "mean_gate": {"surface": 0.19326}, "n_reads": 5075, "n_cand": 6972, "read_suppressed_frac": 0.1027, "dropout_reads_logged": 5075, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 269.2698669433594, "n_util_obs": 863, "util_count_spearman": 0.1549037943466712}, "probe_overhead_sec_total": 195.174, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.022181, "alpha": 0.03, "t": 1134.8} |
| {"event": "consolidate", "step": 3550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1146.3} |
| {"event": "train", "step": 3550, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.48364, "bpb": 2.14045, "chunk_loss": 0.030885, "F": 0.22375, "G": 0.22825, "chunks": 229.12, "bytes_per_chunk": 4.469, "boundary_entropy": 0.11947, "router_noise": 0.0, "grad_norm": 0.7194, "lr": 0.0001355, "router_trainable": true, "sec_per_step": 0.235, "probe_sec": 0.0817, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07309, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.06243, "n_shadow": 0, "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": 0.08582, "trunc_abs_frac": 0.0858, "mean_delta_full": -0.44061, "read_hit_rate": {"surface": 0.73863}, "mean_gate": {"surface": 0.19411}, "n_reads": 5392, "n_cand": 7300, "read_suppressed_frac": 0.1026, "dropout_reads_logged": 5392, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 273.19354248046875, "n_util_obs": 875, "util_count_spearman": 0.15146965211291}, "probe_overhead_sec_total": 198.101, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.022212, "alpha": 0.03, "t": 1146.6} |
| {"event": "consolidate", "step": 3600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1159.7} |
| {"event": "probe", "step": 3600, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3554.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.204218, "cps": 27.765625, "mean_p": 0.207566, "boundary_entropy": 0.120583, "bpb": 1.969455}, "tr": {"chunks": 3699.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.204331, "cps": 28.898438, "mean_p": 0.203122, "boundary_entropy": 0.13969, "bpb": 2.14038}, "fi": {"chunks": 4856.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.258147, "cps": 37.9375, "mean_p": 0.254447, "boundary_entropy": 0.110724, "bpb": 2.249636}}, "bpb": {"en": 1.96946, "tr": 2.14038, "fi": 2.24964}, "gini_cpc": 0.053927, "gini_cps": 0.071682, "bpic": 4.68511, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.00055387120613, "fi": 1.264076889361811}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 276.9776916503906, "n_util_obs": 881, "util_count_spearman": 0.1475132757081797}, "t": 1160.3} |
| {"event": "eval", "step": 3600, "phase": 2, "bpb": {"en": 1.98369, "tr": 2.14781, "fi": 2.26021}, "bytes_per_chunk": {"en": 4.72, "tr": 5.398, "fi": 4.148}, "bpb_mean": 2.13057, "frozen_inventory": true, "t": 1161.3} |
| {"event": "train", "step": 3600, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.46315, "bpb": 2.11088, "chunk_loss": 0.030555, "F": 0.21268, "G": 0.21582, "chunks": 217.78, "bytes_per_chunk": 4.702, "boundary_entropy": 0.1175, "router_noise": 0.0, "grad_norm": 0.6694, "lr": 0.0001319, "router_trainable": true, "sec_per_step": 0.323, "probe_sec": 0.0773, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06912, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.08751, "n_shadow": 0, "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": 0.03517, "trunc_abs_frac": 1.19004, "mean_delta_full": 0.24597, "read_hit_rate": {"surface": 0.72899}, "mean_gate": {"surface": 0.1811}, "n_reads": 5057, "n_cand": 6937, "read_suppressed_frac": 0.102, "dropout_reads_logged": 5057, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 277.05523681640625, "n_util_obs": 882, "util_count_spearman": 0.1453501369447884}, "probe_overhead_sec_total": 200.624, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.016571, "alpha": 0.03, "t": 1162.7} |
| {"event": "consolidate", "step": 3650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1176.8} |
| {"event": "train", "step": 3650, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.48272, "bpb": 2.13912, "chunk_loss": 0.030715, "F": 0.21545, "G": 0.22205, "chunks": 220.62, "bytes_per_chunk": 4.641, "boundary_entropy": 0.11997, "router_noise": 0.0, "grad_norm": 0.7042, "lr": 0.0001283, "router_trainable": true, "sec_per_step": 0.285, "probe_sec": 0.049, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03686, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.12265, "n_shadow": 0, "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": 19.43807, "trunc_abs_frac": 3.28461, "mean_delta_full": 0.00665, "read_hit_rate": {"surface": 0.73534}, "mean_gate": {"surface": 0.19716}, "n_reads": 5168, "n_cand": 7028, "read_suppressed_frac": 0.1, "dropout_reads_logged": 5168, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 280.93524169921875, "n_util_obs": 895, "util_count_spearman": 0.13688519930178844}, "probe_overhead_sec_total": 203.134, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.02538, "alpha": 0.03, "t": 1177.0} |
| {"event": "consolidate", "step": 3700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1193.9} |
| {"event": "train", "step": 3700, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.37892, "bpb": 1.98936, "chunk_loss": 0.030489, "F": 0.20779, "G": 0.21268, "chunks": 212.78, "bytes_per_chunk": 4.812, "boundary_entropy": 0.12099, "router_noise": 0.0, "grad_norm": 0.6884, "lr": 0.0001247, "router_trainable": true, "sec_per_step": 0.342, "probe_sec": 0.025, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01887, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.10172, "n_shadow": 0, "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": 0.98074, "trunc_abs_frac": 1.00214, "mean_delta_full": 0.39043, "read_hit_rate": {"surface": 0.73779}, "mean_gate": {"surface": 0.19226}, "n_reads": 5000, "n_cand": 6777, "read_suppressed_frac": 0.0962, "dropout_reads_logged": 5000, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 284.844970703125, "n_util_obs": 904, "util_count_spearman": 0.13684463012636908}, "probe_overhead_sec_total": 205.641, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.01697, "alpha": 0.03, "t": 1194.1} |
| {"event": "consolidate", "step": 3750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1214.9} |
| {"event": "train", "step": 3750, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.40234, "bpb": 2.02315, "chunk_loss": 0.030737, "F": 0.22061, "G": 0.22243, "chunks": 225.91, "bytes_per_chunk": 4.533, "boundary_entropy": 0.11302, "router_noise": 0.0, "grad_norm": 0.6474, "lr": 0.0001212, "router_trainable": true, "sec_per_step": 0.435, "probe_sec": 0.0478, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02816, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.1231, "n_shadow": 0, "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": 0.51206, "trunc_abs_frac": 1.6562, "mean_delta_full": 0.22109, "read_hit_rate": {"surface": 0.76032}, "mean_gate": {"surface": 0.17583}, "n_reads": 5472, "n_cand": 7197, "read_suppressed_frac": 0.1014, "dropout_reads_logged": 5472, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 288.74273681640625, "n_util_obs": 913, "util_count_spearman": 0.13019139105889097}, "probe_overhead_sec_total": 208.561, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.017929, "alpha": 0.03, "t": 1215.8} |
| {"event": "consolidate", "step": 3800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1240.7} |
| {"event": "probe", "step": 3800, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3542.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.203528, "cps": 27.671875, "mean_p": 0.207812, "boundary_entropy": 0.12049, "bpb": 1.956734}, "tr": {"chunks": 3720.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.205491, "cps": 29.0625, "mean_p": 0.20251, "boundary_entropy": 0.140235, "bpb": 2.122277}, "fi": {"chunks": 4844.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.257509, "cps": 37.84375, "mean_p": 0.254107, "boundary_entropy": 0.110887, "bpb": 2.226912}}, "bpb": {"en": 1.95673, "tr": 2.12228, "fi": 2.22691}, "gini_cpc": 0.053992, "gini_cps": 0.0717, "bpic": 4.686271, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0096432631731849, "fi": 1.2652251447359768}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 292.57061767578125, "n_util_obs": 923, "util_count_spearman": 0.13244345033354515}, "t": 1241.4} |
| {"event": "eval", "step": 3800, "phase": 2, "bpb": {"en": 1.96985, "tr": 2.12918, "fi": 2.23598}, "bytes_per_chunk": {"en": 4.73, "tr": 5.361, "fi": 4.148}, "bpb_mean": 2.11167, "frozen_inventory": true, "t": 1242.5} |
| {"event": "train", "step": 3800, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.38346, "bpb": 1.99591, "chunk_loss": 0.030599, "F": 0.21042, "G": 0.2146, "chunks": 215.47, "bytes_per_chunk": 4.752, "boundary_entropy": 0.11923, "router_noise": 0.0, "grad_norm": 0.6302, "lr": 0.0001177, "router_trainable": true, "sec_per_step": 0.578, "probe_sec": 0.0435, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02433, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.12891, "n_shadow": 0, "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": 0.30446, "trunc_abs_frac": 1.40911, "mean_delta_full": 0.20744, "read_hit_rate": {"surface": 0.75273}, "mean_gate": {"surface": 0.19178}, "n_reads": 5166, "n_cand": 6863, "read_suppressed_frac": 0.0989, "dropout_reads_logged": 5166, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 292.65057373046875, "n_util_obs": 923, "util_count_spearman": 0.13257937501643582}, "probe_overhead_sec_total": 211.325, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.018506, "alpha": 0.03, "t": 1244.7} |
| {"event": "consolidate", "step": 3850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1271.1} |
| {"event": "train", "step": 3850, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.37118, "bpb": 1.97819, "chunk_loss": 0.030391, "F": 0.20364, "G": 0.20667, "chunks": 208.53, "bytes_per_chunk": 4.911, "boundary_entropy": 0.12006, "router_noise": 0.0, "grad_norm": 0.7196, "lr": 0.0001142, "router_trainable": true, "sec_per_step": 0.545, "probe_sec": 0.0407, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02495, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.07911, "n_shadow": 0, "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": 1.08213, "trunc_abs_frac": 1.03155, "mean_delta_full": 0.40048, "read_hit_rate": {"surface": 0.73257}, "mean_gate": {"surface": 0.19191}, "n_reads": 4865, "n_cand": 6641, "read_suppressed_frac": 0.0976, "dropout_reads_logged": 4865, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 296.5716552734375, "n_util_obs": 935, "util_count_spearman": 0.12692735843831215}, "probe_overhead_sec_total": 214.133, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.019636, "alpha": 0.03, "t": 1272.0} |
| {"event": "consolidate", "step": 3900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1298.2} |
| {"event": "train", "step": 3900, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.40494, "bpb": 2.0269, "chunk_loss": 0.030533, "F": 0.20917, "G": 0.21336, "chunks": 214.19, "bytes_per_chunk": 4.781, "boundary_entropy": 0.11991, "router_noise": 0.0, "grad_norm": 0.6668, "lr": 0.0001108, "router_trainable": true, "sec_per_step": 0.529, "probe_sec": 0.026, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01853, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.13656, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 5.27312, "trunc_abs_frac": 11.91518, "mean_delta_full": 0.03564, "read_hit_rate": {"surface": 0.72984}, "mean_gate": {"surface": 0.19532}, "n_reads": 4979, "n_cand": 6822, "read_suppressed_frac": 0.1006, "dropout_reads_logged": 4979, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 300.4981994628906, "n_util_obs": 950, "util_count_spearman": 0.1155595244811355}, "probe_overhead_sec_total": 216.527, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.020769, "alpha": 0.03, "t": 1298.4} |
| {"event": "consolidate", "step": 3950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1318.0} |
| {"event": "train", "step": 3950, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.43744, "bpb": 2.07378, "chunk_loss": 0.030606, "F": 0.21509, "G": 0.21769, "chunks": 220.25, "bytes_per_chunk": 4.649, "boundary_entropy": 0.11317, "router_noise": 0.0, "grad_norm": 0.6632, "lr": 0.0001074, "router_trainable": true, "sec_per_step": 0.395, "probe_sec": 0.0279, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02124, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.10751, "n_shadow": 0, "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": 0.90877, "trunc_abs_frac": 0.56656, "mean_delta_full": 0.57098, "read_hit_rate": {"surface": 0.74701}, "mean_gate": {"surface": 0.18043}, "n_reads": 5241, "n_cand": 7016, "read_suppressed_frac": 0.0967, "dropout_reads_logged": 5241, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 304.3955078125, "n_util_obs": 960, "util_count_spearman": 0.12007428068761872}, "probe_overhead_sec_total": 219.391, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.017548, "alpha": 0.03, "t": 1318.2} |
| {"event": "consolidate", "step": 4000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1329.2} |
| {"event": "probe", "step": 4000, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3535.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.203126, "cps": 27.617188, "mean_p": 0.206447, "boundary_entropy": 0.113072, "bpb": 1.944078}, "tr": {"chunks": 3699.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.204331, "cps": 28.898438, "mean_p": 0.200422, "boundary_entropy": 0.135118, "bpb": 2.10816}, "fi": {"chunks": 4852.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.257934, "cps": 37.90625, "mean_p": 0.253534, "boundary_entropy": 0.110624, "bpb": 2.210913}}, "bpb": {"en": 1.94408, "tr": 2.10816, "fi": 2.21091}, "gini_cpc": 0.054913, "gini_cps": 0.072646, "bpic": 4.694026, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.005931671362542, "fi": 1.2698242331200928}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 308.24737548828125, "n_util_obs": 967, "util_count_spearman": 0.12324660015761582}, "t": 1329.9} |
| {"event": "eval", "step": 4000, "phase": 2, "bpb": {"en": 1.95837, "tr": 2.11552, "fi": 2.21889}, "bytes_per_chunk": {"en": 4.718, "tr": 5.381, "fi": 4.158}, "bpb_mean": 2.09759, "frozen_inventory": true, "t": 1331.0} |
| {"event": "train", "step": 4000, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.45111, "bpb": 2.09351, "chunk_loss": 0.030649, "F": 0.21704, "G": 0.22014, "chunks": 222.25, "bytes_per_chunk": 4.607, "boundary_entropy": 0.11437, "router_noise": 0.0, "grad_norm": 0.7024, "lr": 0.000104, "router_trainable": true, "sec_per_step": 0.287, "probe_sec": 0.0669, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05872, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.00691, "n_shadow": 0, "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": -0.00366, "trunc_abs_frac": 0.00366, "mean_delta_full": -0.26709, "read_hit_rate": {"surface": 0.74379}, "mean_gate": {"surface": 0.18246}, "n_reads": 5266, "n_cand": 7080, "read_suppressed_frac": 0.0997, "dropout_reads_logged": 5266, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 308.32647705078125, "n_util_obs": 967, "util_count_spearman": 0.12303392761698068}, "probe_overhead_sec_total": 222.312, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.017603, "alpha": 0.03, "t": 1332.5} |
| {"event": "consolidate", "step": 4050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1343.8} |
| {"event": "train", "step": 4050, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.37856, "bpb": 1.98884, "chunk_loss": 0.030517, "F": 0.20898, "G": 0.21231, "chunks": 214.0, "bytes_per_chunk": 4.785, "boundary_entropy": 0.11019, "router_noise": 0.0, "grad_norm": 0.6702, "lr": 0.0001007, "router_trainable": true, "sec_per_step": 0.229, "probe_sec": 0.0514, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04226, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.18195, "n_shadow": 0, "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": -0.07018, "trunc_abs_frac": 0.42933, "mean_delta_full": 0.23568, "read_hit_rate": {"surface": 0.74384}, "mean_gate": {"surface": 0.17407}, "n_reads": 5070, "n_cand": 6816, "read_suppressed_frac": 0.0943, "dropout_reads_logged": 5070, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 312.2601318359375, "n_util_obs": 974, "util_count_spearman": 0.12969108547495045}, "probe_overhead_sec_total": 225.675, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.020004, "alpha": 0.03, "t": 1344.0} |
| {"event": "probe", "step": 4096, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3528.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.202724, "cps": 27.5625, "mean_p": 0.20529, "boundary_entropy": 0.112291, "bpb": 1.935528}, "tr": {"chunks": 3648.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.201514, "cps": 28.5, "mean_p": 0.199418, "boundary_entropy": 0.139937, "bpb": 2.094162}, "fi": {"chunks": 4834.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.256977, "cps": 37.765625, "mean_p": 0.252431, "boundary_entropy": 0.109707, "bpb": 2.197489}}, "bpb": {"en": 1.93553, "tr": 2.09416, "fi": 2.19749}, "gini_cpc": 0.055921, "gini_cps": 0.072495, "bpic": 4.72373, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9940307559802356, "fi": 1.2676235716762916}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 315.8087158203125, "n_util_obs": 982, "util_count_spearman": 0.12616691909149957}, "t": 1355.0} |
| {"event": "consolidate", "step": 4100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1357.5} |
| {"event": "train", "step": 4100, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.44603, "bpb": 2.08618, "chunk_loss": 0.030757, "F": 0.22003, "G": 0.22279, "chunks": 225.31, "bytes_per_chunk": 4.545, "boundary_entropy": 0.11205, "router_noise": 0.0, "grad_norm": 0.7019, "lr": 9.75e-05, "router_trainable": true, "sec_per_step": 0.276, "probe_sec": 0.0921, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08563, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.18398, "n_shadow": 0, "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": 0.48768, "trunc_abs_frac": 0.90996, "mean_delta_full": 0.04907, "read_hit_rate": {"surface": 0.74004}, "mean_gate": {"surface": 0.18467}, "n_reads": 5312, "n_cand": 7178, "read_suppressed_frac": 0.0971, "dropout_reads_logged": 5312, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 316.20367431640625, "n_util_obs": 985, "util_count_spearman": 0.11935158245170029}, "probe_overhead_sec_total": 228.47, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.017798, "alpha": 0.03, "t": 1357.8} |
| {"event": "consolidate", "step": 4150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1369.0} |
| {"event": "train", "step": 4150, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.39655, "bpb": 2.0148, "chunk_loss": 0.030855, "F": 0.22507, "G": 0.22598, "chunks": 230.47, "bytes_per_chunk": 4.443, "boundary_entropy": 0.10962, "router_noise": 0.0, "grad_norm": 0.6883, "lr": 9.43e-05, "router_trainable": true, "sec_per_step": 0.23, "probe_sec": 0.1457, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.13918, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.0237, "n_shadow": 0, "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": -0.91888, "trunc_abs_frac": 0.54777, "mean_delta_full": 0.04138, "read_hit_rate": {"surface": 0.75991}, "mean_gate": {"surface": 0.17527}, "n_reads": 5580, "n_cand": 7343, "read_suppressed_frac": 0.1142, "dropout_reads_logged": 5580, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 320.1389465332031, "n_util_obs": 993, "util_count_spearman": 0.12526520164884852}, "probe_overhead_sec_total": 231.02, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.015579, "alpha": 0.03, "t": 1369.3} |
| {"event": "consolidate", "step": 4200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1382.4} |
| {"event": "probe", "step": 4200, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3516.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.202034, "cps": 27.46875, "mean_p": 0.204822, "boundary_entropy": 0.110975, "bpb": 1.934422}, "tr": {"chunks": 3666.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.202508, "cps": 28.640625, "mean_p": 0.197879, "boundary_entropy": 0.136669, "bpb": 2.085209}, "fi": {"chunks": 4848.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.257722, "cps": 37.875, "mean_p": 0.252525, "boundary_entropy": 0.109848, "bpb": 2.186059}}, "bpb": {"en": 1.93442, "tr": 2.08521, "fi": 2.18606}, "gini_cpc": 0.056058, "gini_cps": 0.073815, "bpic": 4.715877, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0023448492217175, "fi": 1.2756336926527811}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 324.010498046875, "n_util_obs": 998, "util_count_spearman": 0.1282910650439841}, "t": 1383.1} |
| {"event": "eval", "step": 4200, "phase": 2, "bpb": {"en": 1.94718, "tr": 2.0933, "fi": 2.19488}, "bytes_per_chunk": {"en": 4.73, "tr": 5.465, "fi": 4.149}, "bpb_mean": 2.07845, "frozen_inventory": true, "t": 1384.3} |
| {"event": "phase_transition", "step": 4200, "phase": 3, "objective": "global", "router_trainable": true, "encoder_trainable": true, "frozen_router_params": 0, "t": 1385.5} |
| {"event": "train", "step": 4200, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38505, "bpb": 1.99821, "chunk_loss": 0.0304, "F": 0.20612, "G": 0.20903, "chunks": 211.06, "bytes_per_chunk": 4.852, "boundary_entropy": 0.11041, "router_noise": 0.0, "grad_norm": 0.674, "lr": 9.12e-05, "router_trainable": true, "sec_per_step": 0.33, "probe_sec": 0.0981, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08961, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.19833, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 3.77943, "trunc_abs_frac": 1.50794, "mean_delta_full": 0.05322, "read_hit_rate": {"surface": 0.73237}, "mean_gate": {"surface": 0.17063}, "n_reads": 4923, "n_cand": 6722, "read_suppressed_frac": 0.0981, "dropout_reads_logged": 4923, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 324.086181640625, "n_util_obs": 998, "util_count_spearman": 0.12800969743912746}, "probe_overhead_sec_total": 234.268, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.015073, "alpha": 0.03, "t": 1385.8} |
| {"event": "consolidate", "step": 4250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1396.3} |
| {"event": "train", "step": 4250, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38211, "bpb": 1.99397, "chunk_loss": 0.030415, "F": 0.20206, "G": 0.20666, "chunks": 206.91, "bytes_per_chunk": 4.949, "boundary_entropy": 0.11486, "router_noise": 0.0, "grad_norm": 0.6897, "lr": 8.81e-05, "router_trainable": true, "sec_per_step": 0.215, "probe_sec": 0.0427, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03336, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.19333, "n_shadow": 0, "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": 0.79499, "trunc_abs_frac": 0.7452, "mean_delta_full": 0.25151, "read_hit_rate": {"surface": 0.72333}, "mean_gate": {"surface": 0.18187}, "n_reads": 4766, "n_cand": 6589, "read_suppressed_frac": 0.0906, "dropout_reads_logged": 4766, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 328.00518798828125, "n_util_obs": 1007, "util_count_spearman": 0.13288571790584877}, "probe_overhead_sec_total": 236.726, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.019001, "alpha": 0.03, "t": 1396.6} |
| {"event": "consolidate", "step": 4300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1407.8} |
| {"event": "train", "step": 4300, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.47449, "bpb": 2.12723, "chunk_loss": 0.030575, "F": 0.21048, "G": 0.21566, "chunks": 215.53, "bytes_per_chunk": 4.751, "boundary_entropy": 0.11782, "router_noise": 0.0, "grad_norm": 0.7769, "lr": 8.5e-05, "router_trainable": true, "sec_per_step": 0.229, "probe_sec": 0.0431, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03613, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.03139, "n_shadow": 0, "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": 0.91967, "trunc_abs_frac": 0.6318, "mean_delta_full": -0.67749, "read_hit_rate": {"surface": 0.72454}, "mean_gate": {"surface": 0.18843}, "n_reads": 4974, "n_cand": 6865, "read_suppressed_frac": 0.0943, "dropout_reads_logged": 4974, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 331.9141845703125, "n_util_obs": 1018, "util_count_spearman": 0.12562914980438739}, "probe_overhead_sec_total": 239.628, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.018838, "alpha": 0.03, "t": 1408.0} |
| {"event": "consolidate", "step": 4350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1419.4} |
| {"event": "train", "step": 4350, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.3265, "bpb": 1.91373, "chunk_loss": 0.030553, "F": 0.21164, "G": 0.21274, "chunks": 216.72, "bytes_per_chunk": 4.725, "boundary_entropy": 0.10538, "router_noise": 0.0, "grad_norm": 0.6965, "lr": 8.21e-05, "router_trainable": true, "sec_per_step": 0.231, "probe_sec": 0.03, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02353, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.21694, "n_shadow": 0, "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": 0.95237, "trunc_abs_frac": 1.26935, "mean_delta_full": 0.26935, "read_hit_rate": {"surface": 0.7433}, "mean_gate": {"surface": 0.17389}, "n_reads": 5131, "n_cand": 6903, "read_suppressed_frac": 0.0955, "dropout_reads_logged": 5131, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 335.85015869140625, "n_util_obs": 1029, "util_count_spearman": 0.11896180593206053}, "probe_overhead_sec_total": 242.693, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.014898, "alpha": 0.03, "t": 1419.6} |
| {"event": "consolidate", "step": 4400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1431.0} |
| {"event": "probe", "step": 4400, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3547.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.203815, "cps": 27.710938, "mean_p": 0.205511, "boundary_entropy": 0.106716, "bpb": 1.921159}, "tr": {"chunks": 3661.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.202232, "cps": 28.601562, "mean_p": 0.198177, "boundary_entropy": 0.135166, "bpb": 2.070845}, "fi": {"chunks": 4849.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.257775, "cps": 37.882812, "mean_p": 0.252239, "boundary_entropy": 0.107154, "bpb": 2.170189}}, "bpb": {"en": 1.92116, "tr": 2.07084, "fi": 2.17019}, "gini_cpc": 0.055781, "gini_cps": 0.071991, "bpic": 4.705316, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9922294412135078, "fi": 1.264745760796523}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 339.7069396972656, "n_util_obs": 1038, "util_count_spearman": 0.12375068078257762}, "t": 1431.7} |
| {"event": "eval", "step": 4400, "phase": 3, "bpb": {"en": 1.93542, "tr": 2.07901, "fi": 2.17754}, "bytes_per_chunk": {"en": 4.71, "tr": 5.443, "fi": 4.151}, "bpb_mean": 2.06399, "frozen_inventory": true, "t": 1432.8} |
| {"event": "train", "step": 4400, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.44401, "bpb": 2.08327, "chunk_loss": 0.030756, "F": 0.21585, "G": 0.21945, "chunks": 221.03, "bytes_per_chunk": 4.633, "boundary_entropy": 0.10838, "router_noise": 0.0, "grad_norm": 0.7359, "lr": 7.92e-05, "router_trainable": true, "sec_per_step": 0.295, "probe_sec": 0.0252, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0181, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.20781, "n_shadow": 0, "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": 1.26435, "trunc_abs_frac": 1.16745, "mean_delta_full": 0.76654, "read_hit_rate": {"surface": 0.73654}, "mean_gate": {"surface": 0.16951}, "n_reads": 5186, "n_cand": 7041, "read_suppressed_frac": 0.0966, "dropout_reads_logged": 5186, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 339.7852783203125, "n_util_obs": 1038, "util_count_spearman": 0.12441906045346489}, "probe_overhead_sec_total": 245.584, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.020547, "alpha": 0.03, "t": 1434.4} |
| {"event": "consolidate", "step": 4450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1452.5} |
| {"event": "train", "step": 4450, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38032, "bpb": 1.99138, "chunk_loss": 0.030645, "F": 0.21545, "G": 0.21705, "chunks": 220.62, "bytes_per_chunk": 4.641, "boundary_entropy": 0.10622, "router_noise": 0.0, "grad_norm": 0.6845, "lr": 7.63e-05, "router_trainable": true, "sec_per_step": 0.367, "probe_sec": 0.0396, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03303, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.01229, "n_shadow": 0, "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": 3.08103, "trunc_abs_frac": 1.16051, "mean_delta_full": -0.00885, "read_hit_rate": {"surface": 0.74132}, "mean_gate": {"surface": 0.17913}, "n_reads": 5210, "n_cand": 7028, "read_suppressed_frac": 0.0994, "dropout_reads_logged": 5210, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 343.72137451171875, "n_util_obs": 1052, "util_count_spearman": 0.12686858828746317}, "probe_overhead_sec_total": 248.334, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.01832, "alpha": 0.03, "t": 1452.7} |
| {"event": "consolidate", "step": 4500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1470.1} |
| {"event": "train", "step": 4500, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.39279, "bpb": 2.00937, "chunk_loss": 0.03067, "F": 0.21606, "G": 0.21882, "chunks": 221.25, "bytes_per_chunk": 4.628, "boundary_entropy": 0.10974, "router_noise": 0.0, "grad_norm": 0.7165, "lr": 7.36e-05, "router_trainable": true, "sec_per_step": 0.351, "probe_sec": 0.0256, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01812, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.07572, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -3.32835, "trunc_abs_frac": 1.86037, "mean_delta_full": -0.03996, "read_hit_rate": {"surface": 0.73666}, "mean_gate": {"surface": 0.17776}, "n_reads": 5192, "n_cand": 7048, "read_suppressed_frac": 0.0948, "dropout_reads_logged": 5192, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 347.6392822265625, "n_util_obs": 1064, "util_count_spearman": 0.12860689775928583}, "probe_overhead_sec_total": 250.689, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.018776, "alpha": 0.03, "t": 1470.3} |
| {"event": "consolidate", "step": 4550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1489.8} |
| {"event": "train", "step": 4550, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.36885, "bpb": 1.97483, "chunk_loss": 0.030655, "F": 0.21582, "G": 0.21851, "chunks": 221.0, "bytes_per_chunk": 4.633, "boundary_entropy": 0.11236, "router_noise": 0.0, "grad_norm": 0.6679, "lr": 7.09e-05, "router_trainable": true, "sec_per_step": 0.394, "probe_sec": 0.0551, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04715, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.09903, "n_shadow": 0, "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": 1.39142, "trunc_abs_frac": 0.42491, "mean_delta_full": 0.11945, "read_hit_rate": {"surface": 0.75412}, "mean_gate": {"surface": 0.17641}, "n_reads": 5309, "n_cand": 7040, "read_suppressed_frac": 0.1027, "dropout_reads_logged": 5309, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 351.52593994140625, "n_util_obs": 1070, "util_count_spearman": 0.12871014387163673}, "probe_overhead_sec_total": 253.686, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.021002, "alpha": 0.03, "t": 1490.0} |
| {"event": "consolidate", "step": 4600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1516.3} |
| {"event": "probe", "step": 4600, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3546.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.203758, "cps": 27.703125, "mean_p": 0.205097, "boundary_entropy": 0.10726, "bpb": 1.910086}, "tr": {"chunks": 3633.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.200685, "cps": 28.382812, "mean_p": 0.196322, "boundary_entropy": 0.135861, "bpb": 2.06102}, "fi": {"chunks": 4851.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.257881, "cps": 37.898438, "mean_p": 0.251381, "boundary_entropy": 0.107152, "bpb": 2.161621}}, "bpb": {"en": 1.91009, "tr": 2.06102, "fi": 2.16162}, "gini_cpc": 0.057571, "gini_cps": 0.072319, "bpic": 4.715877, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9849183647660834, "fi": 1.2656242283986014}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 355.35382080078125, "n_util_obs": 1084, "util_count_spearman": 0.12658712358218166}, "t": 1516.8} |
| {"event": "eval", "step": 4600, "phase": 3, "bpb": {"en": 1.92397, "tr": 2.069, "fi": 2.17007}, "bytes_per_chunk": {"en": 4.712, "tr": 5.496, "fi": 4.153}, "bpb_mean": 2.05435, "frozen_inventory": true, "t": 1517.8} |
| {"event": "train", "step": 4600, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.42997, "bpb": 2.06301, "chunk_loss": 0.030499, "F": 0.20779, "G": 0.21051, "chunks": 212.78, "bytes_per_chunk": 4.812, "boundary_entropy": 0.10971, "router_noise": 0.0, "grad_norm": 0.7188, "lr": 6.82e-05, "router_trainable": true, "sec_per_step": 0.602, "probe_sec": 0.0318, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02014, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.02078, "n_shadow": 0, "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": -0.15656, "trunc_abs_frac": 0.14245, "mean_delta_full": 0.25122, "read_hit_rate": {"surface": 0.72569}, "mean_gate": {"surface": 0.17857}, "n_reads": 4918, "n_cand": 6777, "read_suppressed_frac": 0.0978, "dropout_reads_logged": 4918, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 355.4293518066406, "n_util_obs": 1084, "util_count_spearman": 0.12665788538570122}, "probe_overhead_sec_total": 256.456, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.019199, "alpha": 0.03, "t": 1520.0} |
| {"event": "consolidate", "step": 4650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1546.7} |
| {"event": "train", "step": 4650, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.34488, "bpb": 1.94025, "chunk_loss": 0.030518, "F": 0.21027, "G": 0.21221, "chunks": 215.31, "bytes_per_chunk": 4.756, "boundary_entropy": 0.11076, "router_noise": 0.0, "grad_norm": 0.6729, "lr": 6.57e-05, "router_trainable": true, "sec_per_step": 0.537, "probe_sec": 0.0502, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04233, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.13125, "n_shadow": 0, "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": 0.90383, "trunc_abs_frac": 0.62369, "mean_delta_full": 0.10649, "read_hit_rate": {"surface": 0.74001}, "mean_gate": {"surface": 0.17888}, "n_reads": 5075, "n_cand": 6858, "read_suppressed_frac": 0.1005, "dropout_reads_logged": 5075, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 359.3502502441406, "n_util_obs": 1091, "util_count_spearman": 0.12809296749022947}, "probe_overhead_sec_total": 258.993, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.015972, "alpha": 0.03, "t": 1546.9} |
| {"event": "consolidate", "step": 4700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1570.5} |
| {"event": "train", "step": 4700, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.36273, "bpb": 1.966, "chunk_loss": 0.030578, "F": 0.21191, "G": 0.21336, "chunks": 217.0, "bytes_per_chunk": 4.719, "boundary_entropy": 0.10621, "router_noise": 0.0, "grad_norm": 0.6673, "lr": 6.32e-05, "router_trainable": true, "sec_per_step": 0.478, "probe_sec": 0.0336, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02622, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.0327, "n_shadow": 0, "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": -0.65424, "trunc_abs_frac": 0.48339, "mean_delta_full": -0.15527, "read_hit_rate": {"surface": 0.74248}, "mean_gate": {"surface": 0.17481}, "n_reads": 5132, "n_cand": 6912, "read_suppressed_frac": 0.1023, "dropout_reads_logged": 5132, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 363.3149719238281, "n_util_obs": 1099, "util_count_spearman": 0.12757906837871874}, "probe_overhead_sec_total": 261.592, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.014474, "alpha": 0.03, "t": 1570.8} |
| {"event": "consolidate", "step": 4750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1594.7} |
| {"event": "train", "step": 4750, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.3276, "bpb": 1.91532, "chunk_loss": 0.03056, "F": 0.21207, "G": 0.21462, "chunks": 217.16, "bytes_per_chunk": 4.715, "boundary_entropy": 0.10858, "router_noise": 0.0, "grad_norm": 0.689, "lr": 6.08e-05, "router_trainable": true, "sec_per_step": 0.491, "probe_sec": 0.041, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02441, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.26541, "n_shadow": 0, "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": 1.13423, "trunc_abs_frac": 1.53889, "mean_delta_full": 0.72141, "read_hit_rate": {"surface": 0.7483}, "mean_gate": {"surface": 0.17224}, "n_reads": 5176, "n_cand": 6917, "read_suppressed_frac": 0.1028, "dropout_reads_logged": 5176, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 367.2387390136719, "n_util_obs": 1114, "util_count_spearman": 0.13448690742947123}, "probe_overhead_sec_total": 264.607, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.015888, "alpha": 0.03, "t": 1595.4} |
| {"event": "consolidate", "step": 4800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1608.3} |
| {"event": "probe", "step": 4800, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3536.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.203183, "cps": 27.625, "mean_p": 0.204994, "boundary_entropy": 0.103961, "bpb": 1.900028}, "tr": {"chunks": 3665.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.202453, "cps": 28.632812, "mean_p": 0.195462, "boundary_entropy": 0.131838, "bpb": 2.047635}, "fi": {"chunks": 4864.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.258572, "cps": 38.0, "mean_p": 0.251208, "boundary_entropy": 0.10272, "bpb": 2.147651}}, "bpb": {"en": 1.90003, "tr": 2.04763, "fi": 2.14765}, "gini_cpc": 0.056327, "gini_cps": 0.07338, "bpic": 4.702196, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9964036078867955, "fi": 1.2726047698576288}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 371.0771179199219, "n_util_obs": 1121, "util_count_spearman": 0.13427711008099386}, "t": 1609.0} |
| {"event": "eval", "step": 4800, "phase": 3, "bpb": {"en": 1.91406, "tr": 2.05602, "fi": 2.15565}, "bytes_per_chunk": {"en": 4.708, "tr": 5.47, "fi": 4.135}, "bpb_mean": 2.04191, "frozen_inventory": true, "t": 1610.1} |
| {"event": "train", "step": 4800, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.40231, "bpb": 2.02311, "chunk_loss": 0.030835, "F": 0.22159, "G": 0.22427, "chunks": 226.91, "bytes_per_chunk": 4.513, "boundary_entropy": 0.10506, "router_noise": 0.0, "grad_norm": 0.7391, "lr": 5.85e-05, "router_trainable": true, "sec_per_step": 0.324, "probe_sec": 0.0706, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06134, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.00544, "n_shadow": 0, "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": -3.21477, "trunc_abs_frac": 6.79348, "mean_delta_full": -0.09052, "read_hit_rate": {"surface": 0.74893}, "mean_gate": {"surface": 0.16955}, "n_reads": 5414, "n_cand": 7229, "read_suppressed_frac": 0.0983, "dropout_reads_logged": 5414, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 371.1612548828125, "n_util_obs": 1121, "util_count_spearman": 0.13435985313078871}, "probe_overhead_sec_total": 267.304, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.017323, "alpha": 0.03, "t": 1611.6} |
| {"event": "consolidate", "step": 4850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1622.9} |
| {"event": "train", "step": 4850, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35507, "bpb": 1.95495, "chunk_loss": 0.030695, "F": 0.21683, "G": 0.21896, "chunks": 222.03, "bytes_per_chunk": 4.612, "boundary_entropy": 0.10811, "router_noise": 0.0, "grad_norm": 0.692, "lr": 5.62e-05, "router_trainable": true, "sec_per_step": 0.23, "probe_sec": 0.028, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01872, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.00431, "n_shadow": 0, "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": 0.42606, "trunc_abs_frac": 0.71886, "mean_delta_full": -0.26162, "read_hit_rate": {"surface": 0.74735}, "mean_gate": {"surface": 0.16678}, "n_reads": 5286, "n_cand": 7073, "read_suppressed_frac": 0.1033, "dropout_reads_logged": 5286, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 375.08428955078125, "n_util_obs": 1131, "util_count_spearman": 0.1375966955698324}, "probe_overhead_sec_total": 270.341, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.016127, "alpha": 0.03, "t": 1623.1} |
| {"event": "consolidate", "step": 4900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1634.8} |
| {"event": "train", "step": 4900, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35692, "bpb": 1.95762, "chunk_loss": 0.030458, "F": 0.20792, "G": 0.2098, "chunks": 212.91, "bytes_per_chunk": 4.81, "boundary_entropy": 0.10306, "router_noise": 0.0, "grad_norm": 0.7076, "lr": 5.41e-05, "router_trainable": true, "sec_per_step": 0.238, "probe_sec": 0.0294, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02305, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.02225, "n_shadow": 0, "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": 0.78422, "trunc_abs_frac": 0.66639, "mean_delta_full": -0.90283, "read_hit_rate": {"surface": 0.73396}, "mean_gate": {"surface": 0.15931}, "n_reads": 4977, "n_cand": 6781, "read_suppressed_frac": 0.1005, "dropout_reads_logged": 4977, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 378.965576171875, "n_util_obs": 1141, "util_count_spearman": 0.1360836977361113}, "probe_overhead_sec_total": 272.992, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.016893, "alpha": 0.03, "t": 1635.0} |
| {"event": "consolidate", "step": 4950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1646.3} |
| {"event": "train", "step": 4950, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.34294, "bpb": 1.93745, "chunk_loss": 0.030706, "F": 0.2196, "G": 0.21883, "chunks": 224.88, "bytes_per_chunk": 4.554, "boundary_entropy": 0.10319, "router_noise": 0.0, "grad_norm": 0.7164, "lr": 5.2e-05, "router_trainable": true, "sec_per_step": 0.232, "probe_sec": 0.0417, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0342, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.00236, "n_shadow": 0, "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": 0.42364, "trunc_abs_frac": 0.84637, "mean_delta_full": 0.18864, "read_hit_rate": {"surface": 0.74986}, "mean_gate": {"surface": 0.15769}, "n_reads": 5372, "n_cand": 7164, "read_suppressed_frac": 0.0998, "dropout_reads_logged": 5372, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 382.87152099609375, "n_util_obs": 1146, "util_count_spearman": 0.13935279855534988}, "probe_overhead_sec_total": 275.689, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.015868, "alpha": 0.03, "t": 1646.6} |
| {"event": "consolidate", "step": 5000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1659.2} |
| {"event": "probe", "step": 5000, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3536.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.203183, "cps": 27.625, "mean_p": 0.203818, "boundary_entropy": 0.106037, "bpb": 1.89175}, "tr": {"chunks": 3664.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.202397, "cps": 28.625, "mean_p": 0.19608, "boundary_entropy": 0.133135, "bpb": 2.037299}, "fi": {"chunks": 4851.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.257881, "cps": 37.898438, "mean_p": 0.250765, "boundary_entropy": 0.106334, "bpb": 2.134195}}, "bpb": {"en": 1.89175, "tr": 2.0373, "fi": 2.13419}, "gini_cpc": 0.055752, "gini_cps": 0.072746, "bpic": 4.707659, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.996131737871001, "fi": 1.2692034824381901}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 386.7155456542969, "n_util_obs": 1156, "util_count_spearman": 0.1436153632259528}, "t": 1660.0} |
| {"event": "eval", "step": 5000, "phase": 3, "bpb": {"en": 1.9067, "tr": 2.04546, "fi": 2.14215}, "bytes_per_chunk": {"en": 4.713, "tr": 5.456, "fi": 4.14}, "bpb_mean": 2.03144, "frozen_inventory": true, "t": 1661.0} |
| {"event": "train", "step": 5000, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.34796, "bpb": 1.94469, "chunk_loss": 0.030598, "F": 0.21356, "G": 0.21422, "chunks": 218.69, "bytes_per_chunk": 4.682, "boundary_entropy": 0.10723, "router_noise": 0.0, "grad_norm": 0.6829, "lr": 5e-05, "router_trainable": true, "sec_per_step": 0.319, "probe_sec": 0.111, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10315, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.27345, "n_shadow": 0, "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": -0.34617, "trunc_abs_frac": 2.96971, "mean_delta_full": 0.13312, "read_hit_rate": {"surface": 0.74462}, "mean_gate": {"surface": 0.16835}, "n_reads": 5187, "n_cand": 6966, "read_suppressed_frac": 0.0981, "dropout_reads_logged": 5187, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 386.79522705078125, "n_util_obs": 1156, "util_count_spearman": 0.14366026880521557}, "probe_overhead_sec_total": 278.661, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.018337, "alpha": 0.03, "t": 1662.5} |
| {"event": "consolidate", "step": 5050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1673.8} |
| {"event": "train", "step": 5050, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35759, "bpb": 1.9586, "chunk_loss": 0.030584, "F": 0.21429, "G": 0.21516, "chunks": 219.44, "bytes_per_chunk": 4.666, "boundary_entropy": 0.10514, "router_noise": 0.0, "grad_norm": 0.6575, "lr": 4.81e-05, "router_trainable": true, "sec_per_step": 0.233, "probe_sec": 0.0512, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04281, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.06695, "n_shadow": 0, "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": 0.58544, "trunc_abs_frac": 0.24543, "mean_delta_full": 0.25647, "read_hit_rate": {"surface": 0.74506}, "mean_gate": {"surface": 0.16101}, "n_reads": 5208, "n_cand": 6990, "read_suppressed_frac": 0.0968, "dropout_reads_logged": 5208, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 390.7185363769531, "n_util_obs": 1169, "util_count_spearman": 0.14265914497619867}, "probe_overhead_sec_total": 281.758, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.014827, "alpha": 0.03, "t": 1674.1} |
| {"event": "consolidate", "step": 5100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1685.5} |
| {"event": "train", "step": 5100, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32307, "bpb": 1.90878, "chunk_loss": 0.03043, "F": 0.20776, "G": 0.20752, "chunks": 212.75, "bytes_per_chunk": 4.813, "boundary_entropy": 0.10416, "router_noise": 0.0, "grad_norm": 0.6818, "lr": 4.63e-05, "router_trainable": true, "sec_per_step": 0.233, "probe_sec": 0.076, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06802, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.02525, "n_shadow": 0, "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": 0.43761, "trunc_abs_frac": 0.8906, "mean_delta_full": 0.2251, "read_hit_rate": {"surface": 0.74439}, "mean_gate": {"surface": 0.1554}, "n_reads": 5044, "n_cand": 6776, "read_suppressed_frac": 0.1011, "dropout_reads_logged": 5044, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 394.6433410644531, "n_util_obs": 1177, "util_count_spearman": 0.1419872208605609}, "probe_overhead_sec_total": 284.992, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.016361, "alpha": 0.03, "t": 1685.8} |
| {"event": "consolidate", "step": 5150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1697.6} |
| {"event": "train", "step": 5150, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.31806, "bpb": 1.90156, "chunk_loss": 0.030604, "F": 0.21484, "G": 0.21428, "chunks": 220.0, "bytes_per_chunk": 4.655, "boundary_entropy": 0.10556, "router_noise": 0.0, "grad_norm": 0.6836, "lr": 4.45e-05, "router_trainable": true, "sec_per_step": 0.241, "probe_sec": 0.0619, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04835, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.16279, "n_shadow": 0, "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": 2.14344, "trunc_abs_frac": 0.63924, "mean_delta_full": 0.05957, "read_hit_rate": {"surface": 0.75357}, "mean_gate": {"surface": 0.15893}, "n_reads": 5281, "n_cand": 7008, "read_suppressed_frac": 0.0988, "dropout_reads_logged": 5281, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 398.57720947265625, "n_util_obs": 1189, "util_count_spearman": 0.1440056052684137}, "probe_overhead_sec_total": 287.891, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.017224, "alpha": 0.03, "t": 1697.9} |
| {"event": "consolidate", "step": 5200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1710.0} |
| {"event": "probe", "step": 5200, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3528.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.202724, "cps": 27.5625, "mean_p": 0.203127, "boundary_entropy": 0.105833, "bpb": 1.88481}, "tr": {"chunks": 3696.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.204165, "cps": 28.875, "mean_p": 0.196391, "boundary_entropy": 0.130976, "bpb": 2.027339}, "fi": {"chunks": 4850.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.257828, "cps": 37.890625, "mean_p": 0.250786, "boundary_entropy": 0.105349, "bpb": 2.128452}}, "bpb": {"en": 1.88481, "tr": 2.02734, "fi": 2.12845}, "gini_cpc": 0.055266, "gini_cps": 0.072994, "bpic": 4.698691, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.007110108032607, "fi": 1.271819264093921}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 402.4277648925781, "n_util_obs": 1205, "util_count_spearman": 0.14242621329952534}, "t": 1710.8} |
| {"event": "eval", "step": 5200, "phase": 3, "bpb": {"en": 1.89952, "tr": 2.03479, "fi": 2.13624}, "bytes_per_chunk": {"en": 4.718, "tr": 5.416, "fi": 4.14}, "bpb_mean": 2.02352, "frozen_inventory": true, "t": 1711.8} |
| {"event": "train", "step": 5200, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35474, "bpb": 1.95448, "chunk_loss": 0.030795, "F": 0.2207, "G": 0.22097, "chunks": 226.0, "bytes_per_chunk": 4.531, "boundary_entropy": 0.10792, "router_noise": 0.0, "grad_norm": 0.7056, "lr": 4.29e-05, "router_trainable": true, "sec_per_step": 0.306, "probe_sec": 0.0456, "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": 0.30905, "n_shadow": 0, "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": -0.51563, "trunc_abs_frac": 2.32591, "mean_delta_full": -0.34607, "read_hit_rate": {"surface": 0.75222}, "mean_gate": {"surface": 0.1635}, "n_reads": 5416, "n_cand": 7200, "read_suppressed_frac": 0.1006, "dropout_reads_logged": 5416, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 402.51068115234375, "n_util_obs": 1205, "util_count_spearman": 0.14232344001194847}, "probe_overhead_sec_total": 290.926, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.017739, "alpha": 0.03, "t": 1713.2} |
| {"event": "consolidate", "step": 5250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1729.1} |
| {"event": "train", "step": 5250, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38565, "bpb": 1.99907, "chunk_loss": 0.030628, "F": 0.21689, "G": 0.2173, "chunks": 222.09, "bytes_per_chunk": 4.611, "boundary_entropy": 0.10654, "router_noise": 0.0, "grad_norm": 0.6775, "lr": 4.14e-05, "router_trainable": true, "sec_per_step": 0.324, "probe_sec": 0.0553, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04519, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.00306, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 5.53162, "trunc_abs_frac": 0.34082, "mean_delta_full": -0.02767, "read_hit_rate": {"surface": 0.74615}, "mean_gate": {"surface": 0.16088}, "n_reads": 5279, "n_cand": 7075, "read_suppressed_frac": 0.1038, "dropout_reads_logged": 5279, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 406.440673828125, "n_util_obs": 1212, "util_count_spearman": 0.14537176446982053}, "probe_overhead_sec_total": 293.735, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.014514, "alpha": 0.03, "t": 1729.4} |
| {"event": "consolidate", "step": 5300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1747.0} |
| {"event": "train", "step": 5300, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.37661, "bpb": 1.98603, "chunk_loss": 0.030523, "F": 0.21149, "G": 0.21226, "chunks": 216.56, "bytes_per_chunk": 4.728, "boundary_entropy": 0.10395, "router_noise": 0.0, "grad_norm": 0.6826, "lr": 3.99e-05, "router_trainable": true, "sec_per_step": 0.355, "probe_sec": 0.0438, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03232, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.08047, "n_shadow": 0, "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": 0.87679, "trunc_abs_frac": 1.35213, "mean_delta_full": 0.43449, "read_hit_rate": {"surface": 0.73355}, "mean_gate": {"surface": 0.15955}, "n_reads": 5060, "n_cand": 6898, "read_suppressed_frac": 0.1018, "dropout_reads_logged": 5060, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 410.40380859375, "n_util_obs": 1218, "util_count_spearman": 0.1486487218128781}, "probe_overhead_sec_total": 296.256, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.016205, "alpha": 0.03, "t": 1747.1} |
| {"event": "consolidate", "step": 5350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1764.3} |
| {"event": "train", "step": 5350, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.33928, "bpb": 1.93217, "chunk_loss": 0.03049, "F": 0.20886, "G": 0.21149, "chunks": 213.88, "bytes_per_chunk": 4.788, "boundary_entropy": 0.10721, "router_noise": 0.0, "grad_norm": 0.7516, "lr": 3.86e-05, "router_trainable": true, "sec_per_step": 0.346, "probe_sec": 0.0251, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01815, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.23183, "n_shadow": 0, "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": -0.92214, "trunc_abs_frac": 0.81764, "mean_delta_full": 0.04272, "read_hit_rate": {"surface": 0.72798}, "mean_gate": {"surface": 0.16526}, "n_reads": 4959, "n_cand": 6812, "read_suppressed_frac": 0.0992, "dropout_reads_logged": 4959, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 414.327880859375, "n_util_obs": 1230, "util_count_spearman": 0.1470069455371154}, "probe_overhead_sec_total": 298.976, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.018222, "alpha": 0.03, "t": 1764.5} |
| {"event": "consolidate", "step": 5400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1783.5} |
| {"event": "probe", "step": 5400, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3528.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.202724, "cps": 27.5625, "mean_p": 0.203237, "boundary_entropy": 0.102963, "bpb": 1.880321}, "tr": {"chunks": 3710.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.204938, "cps": 28.984375, "mean_p": 0.19596, "boundary_entropy": 0.129452, "bpb": 2.020017}, "fi": {"chunks": 4847.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.257668, "cps": 37.867188, "mean_p": 0.251359, "boundary_entropy": 0.104126, "bpb": 2.119226}}, "bpb": {"en": 1.88032, "tr": 2.02002, "fi": 2.11923}, "gini_cpc": 0.055055, "gini_cps": 0.072762, "bpic": 4.694415, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.010924919047882, "fi": 1.2710325717656155}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 418.175048828125, "n_util_obs": 1242, "util_count_spearman": 0.14515919738058033}, "t": 1784.0} |
| {"event": "eval", "step": 5400, "phase": 3, "bpb": {"en": 1.89497, "tr": 2.02744, "fi": 2.12653}, "bytes_per_chunk": {"en": 4.718, "tr": 5.432, "fi": 4.136}, "bpb_mean": 2.01632, "frozen_inventory": true, "t": 1785.0} |
| {"event": "train", "step": 5400, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.26463, "bpb": 1.82447, "chunk_loss": 0.030471, "F": 0.20557, "G": 0.2056, "chunks": 210.5, "bytes_per_chunk": 4.865, "boundary_entropy": 0.10311, "router_noise": 0.0, "grad_norm": 0.6767, "lr": 3.73e-05, "router_trainable": true, "sec_per_step": 0.434, "probe_sec": 0.0393, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02992, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.05142, "n_shadow": 0, "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": -1.10065, "trunc_abs_frac": 1.73874, "mean_delta_full": 0.04199, "read_hit_rate": {"surface": 0.75343}, "mean_gate": {"surface": 0.16198}, "n_reads": 5051, "n_cand": 6704, "read_suppressed_frac": 0.0974, "dropout_reads_logged": 5051, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 418.25286865234375, "n_util_obs": 1242, "util_count_spearman": 0.14480414087341595}, "probe_overhead_sec_total": 301.566, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.015636, "alpha": 0.03, "t": 1786.2} |
| {"event": "consolidate", "step": 5450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1813.1} |
| {"event": "train", "step": 5450, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.31245, "bpb": 1.89346, "chunk_loss": 0.030557, "F": 0.21472, "G": 0.21419, "chunks": 219.88, "bytes_per_chunk": 4.657, "boundary_entropy": 0.10205, "router_noise": 0.0, "grad_norm": 0.6754, "lr": 3.62e-05, "router_trainable": true, "sec_per_step": 0.552, "probe_sec": 0.0674, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04653, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.12149, "n_shadow": 0, "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": 0.55759, "trunc_abs_frac": 0.9465, "mean_delta_full": 0.28497, "read_hit_rate": {"surface": 0.74786}, "mean_gate": {"surface": 0.15524}, "n_reads": 5238, "n_cand": 7004, "read_suppressed_frac": 0.0949, "dropout_reads_logged": 5238, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 422.1920166015625, "n_util_obs": 1253, "util_count_spearman": 0.14184230283265434}, "probe_overhead_sec_total": 304.372, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.014844, "alpha": 0.03, "t": 1813.8} |
| {"event": "consolidate", "step": 5500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1838.0} |
| {"event": "train", "step": 5500, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32394, "bpb": 1.91004, "chunk_loss": 0.030587, "F": 0.21497, "G": 0.21467, "chunks": 220.12, "bytes_per_chunk": 4.652, "boundary_entropy": 0.10692, "router_noise": 0.0, "grad_norm": 0.6711, "lr": 3.51e-05, "router_trainable": true, "sec_per_step": 0.501, "probe_sec": 0.0422, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02142, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.01568, "n_shadow": 0, "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": 2.49056, "trunc_abs_frac": 0.77917, "mean_delta_full": 0.03848, "read_hit_rate": {"surface": 0.75456}, "mean_gate": {"surface": 0.16042}, "n_reads": 5291, "n_cand": 7012, "read_suppressed_frac": 0.0956, "dropout_reads_logged": 5291, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 426.12652587890625, "n_util_obs": 1261, "util_count_spearman": 0.14154874970158934}, "probe_overhead_sec_total": 307.169, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.015649, "alpha": 0.03, "t": 1838.8} |
| {"event": "consolidate", "step": 5550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1861.5} |
| {"event": "train", "step": 5550, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.31059, "bpb": 1.89077, "chunk_loss": 0.030651, "F": 0.21985, "G": 0.21895, "chunks": 225.12, "bytes_per_chunk": 4.549, "boundary_entropy": 0.09928, "router_noise": 0.0, "grad_norm": 0.6714, "lr": 3.41e-05, "router_trainable": true, "sec_per_step": 0.47, "probe_sec": 0.0477, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02322, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.03863, "n_shadow": 0, "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": 1.04613, "trunc_abs_frac": 0.8457, "mean_delta_full": -0.30525, "read_hit_rate": {"surface": 0.76004}, "mean_gate": {"surface": 0.15287}, "n_reads": 5451, "n_cand": 7172, "read_suppressed_frac": 0.1048, "dropout_reads_logged": 5451, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 430.041015625, "n_util_obs": 1270, "util_count_spearman": 0.14381195038728092}, "probe_overhead_sec_total": 310.401, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.017034, "alpha": 0.03, "t": 1862.3} |
| {"event": "consolidate", "step": 5600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1888.2} |
| {"event": "probe", "step": 5600, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3523.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.202436, "cps": 27.523438, "mean_p": 0.202895, "boundary_entropy": 0.103212, "bpb": 1.87429}, "tr": {"chunks": 3653.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.20179, "cps": 28.539062, "mean_p": 0.194272, "boundary_entropy": 0.131673, "bpb": 2.014097}, "fi": {"chunks": 4826.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.256552, "cps": 37.703125, "mean_p": 0.250128, "boundary_entropy": 0.10418, "bpb": 2.110061}}, "bpb": {"en": 1.87429, "tr": 2.0141, "fi": 2.11006}, "gini_cpc": 0.05525, "gini_cps": 0.072377, "bpic": 4.726879, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9968058952533402, "fi": 1.2673218164772233}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 433.88995361328125, "n_util_obs": 1282, "util_count_spearman": 0.14658114017570942}, "t": 1888.9} |
| {"event": "eval", "step": 5600, "phase": 3, "bpb": {"en": 1.88967, "tr": 2.0216, "fi": 2.11804}, "bytes_per_chunk": {"en": 4.728, "tr": 5.482, "fi": 4.15}, "bpb_mean": 2.00977, "frozen_inventory": true, "t": 1890.2} |
| {"event": "train", "step": 5600, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.34032, "bpb": 1.93368, "chunk_loss": 0.030632, "F": 0.2153, "G": 0.21667, "chunks": 220.47, "bytes_per_chunk": 4.645, "boundary_entropy": 0.10609, "router_noise": 0.0, "grad_norm": 0.7033, "lr": 3.33e-05, "router_trainable": true, "sec_per_step": 0.579, "probe_sec": 0.0272, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01903, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.07279, "n_shadow": 0, "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": 0.19775, "trunc_abs_frac": 0.19775, "mean_delta_full": 0.14075, "read_hit_rate": {"surface": 0.74356}, "mean_gate": {"surface": 0.15972}, "n_reads": 5222, "n_cand": 7023, "read_suppressed_frac": 0.0957, "dropout_reads_logged": 5222, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 433.9700927734375, "n_util_obs": 1282, "util_count_spearman": 0.1467843822998327}, "probe_overhead_sec_total": 312.966, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.013708, "alpha": 0.03, "t": 1891.2} |
| {"event": "consolidate", "step": 5650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1902.8} |
| {"event": "train", "step": 5650, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32134, "bpb": 1.90629, "chunk_loss": 0.030532, "F": 0.21146, "G": 0.21236, "chunks": 216.53, "bytes_per_chunk": 4.729, "boundary_entropy": 0.10179, "router_noise": 0.0, "grad_norm": 0.6522, "lr": 3.25e-05, "router_trainable": true, "sec_per_step": 0.235, "probe_sec": 0.0361, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02082, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.30117, "n_shadow": 0, "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": 0.60152, "trunc_abs_frac": 0.75212, "mean_delta_full": 0.65924, "read_hit_rate": {"surface": 0.7467}, "mean_gate": {"surface": 0.15972}, "n_reads": 5150, "n_cand": 6897, "read_suppressed_frac": 0.0988, "dropout_reads_logged": 5150, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 437.89520263671875, "n_util_obs": 1287, "util_count_spearman": 0.14895766808915314}, "probe_overhead_sec_total": 315.82, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.023484, "alpha": 0.03, "t": 1903.0} |
| {"event": "consolidate", "step": 5700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1914.2} |
| {"event": "train", "step": 5700, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.40025, "bpb": 2.02013, "chunk_loss": 0.030606, "F": 0.21512, "G": 0.21611, "chunks": 220.28, "bytes_per_chunk": 4.649, "boundary_entropy": 0.10524, "router_noise": 0.0, "grad_norm": 0.6993, "lr": 3.18e-05, "router_trainable": true, "sec_per_step": 0.235, "probe_sec": 0.032, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02495, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.00835, "n_shadow": 0, "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": 0.80684, "trunc_abs_frac": 1.35602, "mean_delta_full": -0.33728, "read_hit_rate": {"surface": 0.73279}, "mean_gate": {"surface": 0.16212}, "n_reads": 5142, "n_cand": 7017, "read_suppressed_frac": 0.1023, "dropout_reads_logged": 5142, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 441.84490966796875, "n_util_obs": 1291, "util_count_spearman": 0.1498657732781095}, "probe_overhead_sec_total": 318.641, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.01659, "alpha": 0.03, "t": 1914.7} |
| {"event": "consolidate", "step": 5750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1927.4} |
| {"event": "train", "step": 5750, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.3846, "bpb": 1.99756, "chunk_loss": 0.030539, "F": 0.21106, "G": 0.21269, "chunks": 216.12, "bytes_per_chunk": 4.738, "boundary_entropy": 0.10419, "router_noise": 0.0, "grad_norm": 0.7119, "lr": 3.13e-05, "router_trainable": true, "sec_per_step": 0.258, "probe_sec": 0.0397, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03296, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.14797, "n_shadow": 0, "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": 0.41905, "trunc_abs_frac": 0.99812, "mean_delta_full": 0.66811, "read_hit_rate": {"surface": 0.73053}, "mean_gate": {"surface": 0.15969}, "n_reads": 5029, "n_cand": 6884, "read_suppressed_frac": 0.0984, "dropout_reads_logged": 5029, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 445.7874755859375, "n_util_obs": 1301, "util_count_spearman": 0.15730295549845574}, "probe_overhead_sec_total": 321.677, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.019213, "alpha": 0.03, "t": 1927.6} |
| {"event": "consolidate", "step": 5800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1940.2} |
| {"event": "probe", "step": 5800, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3540.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.203413, "cps": 27.65625, "mean_p": 0.203334, "boundary_entropy": 0.101595, "bpb": 1.869856}, "tr": {"chunks": 3708.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.204828, "cps": 28.96875, "mean_p": 0.196151, "boundary_entropy": 0.128085, "bpb": 2.007538}, "fi": {"chunks": 4870.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.258891, "cps": 38.046875, "mean_p": 0.2511, "boundary_entropy": 0.100899, "bpb": 2.103464}}, "bpb": {"en": 1.86986, "tr": 2.00754, "fi": 2.10346}, "gini_cpc": 0.055439, "gini_cps": 0.073169, "bpic": 4.681631, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0069549292794433, "fi": 1.2727348495155646}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 449.6476135253906, "n_util_obs": 1309, "util_count_spearman": 0.16061767132672036}, "t": 1940.9} |
| {"event": "eval", "step": 5800, "phase": 3, "bpb": {"en": 1.88543, "tr": 2.01533, "fi": 2.11142}, "bytes_per_chunk": {"en": 4.706, "tr": 5.44, "fi": 4.128}, "bpb_mean": 2.00406, "frozen_inventory": true, "t": 1942.2} |
| {"event": "train", "step": 5800, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.29957, "bpb": 1.87488, "chunk_loss": 0.030519, "F": 0.21094, "G": 0.21123, "chunks": 216.0, "bytes_per_chunk": 4.741, "boundary_entropy": 0.10126, "router_noise": 0.0, "grad_norm": 0.7144, "lr": 3.08e-05, "router_trainable": true, "sec_per_step": 0.329, "probe_sec": 0.0978, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0864, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.04959, "n_shadow": 0, "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": 0.91773, "trunc_abs_frac": 0.91773, "mean_delta_full": 0.3717, "read_hit_rate": {"surface": 0.75276}, "mean_gate": {"surface": 0.15308}, "n_reads": 5179, "n_cand": 6880, "read_suppressed_frac": 0.1031, "dropout_reads_logged": 5179, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 449.7276611328125, "n_util_obs": 1309, "util_count_spearman": 0.16067517718436397}, "probe_overhead_sec_total": 325.073, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.01645, "alpha": 0.03, "t": 1944.1} |
| {"event": "consolidate", "step": 5850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1956.1} |
| {"event": "train", "step": 5850, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35266, "bpb": 1.95148, "chunk_loss": 0.030431, "F": 0.20566, "G": 0.20637, "chunks": 210.59, "bytes_per_chunk": 4.862, "boundary_entropy": 0.10168, "router_noise": 0.0, "grad_norm": 0.6969, "lr": 3.05e-05, "router_trainable": true, "sec_per_step": 0.243, "probe_sec": 0.0247, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01855, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.05887, "n_shadow": 0, "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": -0.13931, "trunc_abs_frac": 26.09694, "mean_delta_full": -0.06887, "read_hit_rate": {"surface": 0.72655}, "mean_gate": {"surface": 0.15319}, "n_reads": 4873, "n_cand": 6707, "read_suppressed_frac": 0.0999, "dropout_reads_logged": 4873, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 453.6658935546875, "n_util_obs": 1319, "util_count_spearman": 0.16520041722001244}, "probe_overhead_sec_total": 328.292, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.019016, "alpha": 0.03, "t": 1956.3} |
| {"event": "consolidate", "step": 5900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1968.2} |
| {"event": "train", "step": 5900, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38407, "bpb": 1.9968, "chunk_loss": 0.030629, "F": 0.21667, "G": 0.21811, "chunks": 221.88, "bytes_per_chunk": 4.615, "boundary_entropy": 0.1034, "router_noise": 0.0, "grad_norm": 0.7425, "lr": 3.02e-05, "router_trainable": true, "sec_per_step": 0.244, "probe_sec": 0.1025, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09058, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.12013, "n_shadow": 0, "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": 1.06446, "trunc_abs_frac": 0.52071, "mean_delta_full": -0.35051, "read_hit_rate": {"surface": 0.7384}, "mean_gate": {"surface": 0.16248}, "n_reads": 5219, "n_cand": 7068, "read_suppressed_frac": 0.0991, "dropout_reads_logged": 5219, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 457.6285400390625, "n_util_obs": 1332, "util_count_spearman": 0.16860452248536448}, "probe_overhead_sec_total": 331.241, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.013527, "alpha": 0.03, "t": 1968.5} |
| {"event": "consolidate", "step": 5950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1979.7} |
| {"event": "train", "step": 5950, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.31753, "bpb": 1.9008, "chunk_loss": 0.030547, "F": 0.21143, "G": 0.21163, "chunks": 216.5, "bytes_per_chunk": 4.73, "boundary_entropy": 0.10526, "router_noise": 0.0, "grad_norm": 0.7798, "lr": 3.01e-05, "router_trainable": true, "sec_per_step": 0.231, "probe_sec": 0.0308, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0246, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.1319, "n_shadow": 0, "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": -0.65091, "trunc_abs_frac": 0.65091, "mean_delta_full": 0.06714, "read_hit_rate": {"surface": 0.75116}, "mean_gate": {"surface": 0.1607}, "n_reads": 5180, "n_cand": 6896, "read_suppressed_frac": 0.0913, "dropout_reads_logged": 5180, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 461.569580078125, "n_util_obs": 1339, "util_count_spearman": 0.1681564226437379}, "probe_overhead_sec_total": 333.945, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.015729, "alpha": 0.03, "t": 1980.0} |
| {"event": "consolidate", "step": 6000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1991.4} |
| {"event": "train", "step": 5999, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.33832, "bpb": 1.93079, "chunk_loss": 0.030637, "F": 0.21533, "G": 0.21632, "chunks": 220.5, "bytes_per_chunk": 4.644, "boundary_entropy": 0.10378, "router_noise": 0.0, "grad_norm": 0.7352, "lr": 3e-05, "router_trainable": true, "sec_per_step": 0.228, "probe_sec": 0.03, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02308, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.10367, "n_shadow": 0, "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": 1.94567, "trunc_abs_frac": 0.96884, "mean_delta_full": 0.04887, "read_hit_rate": {"surface": 0.749}, "mean_gate": {"surface": 0.1598}, "n_reads": 5261, "n_cand": 7024, "read_suppressed_frac": 0.0937, "dropout_reads_logged": 5261, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 465.41375732421875, "n_util_obs": 1348, "util_count_spearman": 0.16434094444072728}, "probe_overhead_sec_total": 336.684, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.014685, "alpha": 0.03, "t": 1991.4} |
| {"event": "probe", "step": 6000, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3523.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.202436, "cps": 27.523438, "mean_p": 0.202214, "boundary_entropy": 0.101511, "bpb": 1.867456}, "tr": {"chunks": 3689.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.203778, "cps": 28.820312, "mean_p": 0.195053, "boundary_entropy": 0.129705, "bpb": 2.001755}, "fi": {"chunks": 4833.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.256924, "cps": 37.757812, "mean_p": 0.249719, "boundary_entropy": 0.102733, "bpb": 2.099956}}, "bpb": {"en": 1.86746, "tr": 2.00175, "fi": 2.09996}, "gini_cpc": 0.054778, "gini_cps": 0.072506, "bpic": 4.710004, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0066293313959958, "fi": 1.2691600370978908}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 465.41375732421875, "n_util_obs": 1348, "util_count_spearman": 0.16434094444072728}, "t": 1992.1} |
| {"event": "eval", "step": 6000, "phase": 3, "bpb": {"en": 1.88302, "tr": 2.00958, "fi": 2.10701}, "bytes_per_chunk": {"en": 4.73, "tr": 5.467, "fi": 4.145}, "bpb_mean": 1.99987, "frozen_inventory": true, "t": 1993.2} |
| {"event": "reps_dumped", "dirs": ["/local/scratch/sas245/beetle-hnet/runs/ali_mem-shuffled_s0/probe/reps/th0.30", "/local/scratch/sas245/beetle-hnet/runs/ali_mem-shuffled_s0/probe/reps/th0.40", "/local/scratch/sas245/beetle-hnet/runs/ali_mem-shuffled_s0/probe/reps/th0.50", "/local/scratch/sas245/beetle-hnet/runs/ali_mem-shuffled_s0/probe/reps/th0.60", "/local/scratch/sas245/beetle-hnet/runs/ali_mem-shuffled_s0/probe/reps/th0.70"], "t": 1995.6} |
| {"event": "final", "run": "ali_mem-shuffled_s0", "arch": "ali", "objective": "ratio", "condition": "A", "step": 6000, "params": 22785088, "seed": 0, "per_lang": {"en": {"cpc": 0.20243636154685973, "cps": 27.5234375, "bpb": 1.8868516198438707, "bytes_per_chunk": 4.682397070643923, "reads_per_chunk": 0.7544252924890595}, "tr": {"cpc": 0.20377837927415346, "cps": 28.8203125, "bpb": 2.0020402679296034, "bytes_per_chunk": 5.461219557925877, "reads_per_chunk": 0.6578925439053354}, "fi": {"cpc": 0.2569241401307745, "cps": 37.7578125, "bpb": 2.125636330898427, "bytes_per_chunk": 4.126302534235794, "reads_per_chunk": 0.7199748150480089}}, "bpb_mean": 2.0048427395573003, "gini_cpc": 0.05477764427034226, "gini_cps": 0.07250588072505892, "bpic": 4.7100041511000414, "mean_boundary_entropy": 0.10378013551235199, "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": 336.6841495037079, "probe_overhead_seq_frac": 0.4999166666666667, "wall_sec": 1997.9117681980133, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 65536, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 465.41375732421875, "n_util_obs": 1348, "util_count_spearman": 0.16434094444072728}, "ali_params": 268096, "ali_config": {"enabled": true, "budget_B": 65536, "staging_capacity": 262144, "policy": "frequency", "value_mode": "shuffled", "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": 0.16434094444072728, "paired_vs_dropout": {"n": 2820, "pearson": 0.034973164959911436, "spearman": 0.0430551992947645, "mean_abs_discrepancy": 0.9360973238945007, "mean_paired": 0.07086104899644852, "mean_dropout": 0.08221510052680969}, "predictor_calibration": {"n": 6569, "all": {"n": 6569, "pearson": 0.1483633496701249, "spearman": 0.20082913916984332, "mae": 0.11744236201047897, "reliability": [{"pred": -0.00343409669585526, "obs": 0.009722440503537655, "n": 657}, {"pred": 0.005598395597189665, "obs": 0.015091556124389172, "n": 657}, {"pred": 0.011477815918624401, "obs": 0.024840466678142548, "n": 657}, {"pred": 0.016125883907079697, "obs": 0.037111762911081314, "n": 657}, {"pred": 0.020868105813860893, "obs": 0.04083925858139992, "n": 656}, {"pred": 0.02569163776934147, "obs": 0.04500530660152435, "n": 657}, {"pred": 0.031187789514660835, "obs": 0.05046120285987854, "n": 657}, {"pred": 0.038196127861738205, "obs": 0.059076108038425446, "n": 657}, {"pred": 0.0493241585791111, "obs": 0.09118574857711792, "n": 657}, {"pred": 0.07556474208831787, "obs": 0.10364902764558792, "n": 657}]}, "admitted": {"n": 6569, "pearson": 0.1483633496701249, "spearman": 0.20082913916984332, "mae": 0.11744236201047897, "reliability": [{"pred": -0.00343409669585526, "obs": 0.009722440503537655, "n": 657}, {"pred": 0.005598395597189665, "obs": 0.015091556124389172, "n": 657}, {"pred": 0.011477815918624401, "obs": 0.024840466678142548, "n": 657}, {"pred": 0.016125883907079697, "obs": 0.037111762911081314, "n": 657}, {"pred": 0.020868105813860893, "obs": 0.04083925858139992, "n": 656}, {"pred": 0.02569163776934147, "obs": 0.04500530660152435, "n": 657}, {"pred": 0.031187789514660835, "obs": 0.05046120285987854, "n": 657}, {"pred": 0.038196127861738205, "obs": 0.059076108038425446, "n": 657}, {"pred": 0.0493241585791111, "obs": 0.09118574857711792, "n": 657}, {"pred": 0.07556474208831787, "obs": 0.10364902764558792, "n": 657}]}, "staged": {"n": 0, "pearson": NaN, "spearman": NaN, "mae": NaN}}, "t": 1998.0} |
|
|
