| {"event": "hub", "repo": "AdaptiveChunking/ali_mem-skel_s0", "keep_steps": [0, 1, 2, 4, 6, 8, 16, 32, 64, 128, 256, 512, 1000, 1024, 2048, 3000, 4096, 6000], "t": 3.6} |
| {"event": "start", "run": "ali_mem-skel_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": "learned", "candidate_types": ["skel"], "gate": "vector", "retrieval": true, "memory_scope": "persistent", "read_dropout": 0.1, "write_rule": "mean", "eta_min": 0.0, "consolidate_every": 50, "probe_seqs": 2, "probe_reps": 8, "horizon": 32, "predictor": "mlp", "lambda_util": 0.1, "max_span": 64, "byte_classes_path": "/local/scratch/sas245/beetle-hnet/scratch/data/byte_classes.json", "gate_hidden": 0, "type_dim": 16, "gate_bias_init": -2.0}, "t": 3.6} |
| {"event": "probe_schedule", "n": 45, "steps": [0, 1, 2, 4, 6, 8, 16, 32, 64, 128, 200, 256, 400, 512, 600, 800, 1000, 1024, 1200, 1400, 1600, 1800, 2000, 2048, 2200, 2400, 2600, 2800, 3000, 3200, 3400, 3600, 3800, 4000, 4096, 4200, 4400, 4600, 4800, 5000, 5200, 5400, 5600, 5800, 6000], "t": 3.6} |
| {"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": 5.2} |
| {"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": 11.0316, "lr": 1e-06, "router_trainable": true, "sec_per_step": 0.012, "probe_sec": 0.0034, "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": {"skel": 0.0}, "mean_gate": {"skel": null}, "n_reads": 0, "n_cand": 4181, "read_suppressed_frac": 0.0, "dropout_reads_logged": 0, "inventory": {"size": 2501, "admitted": 2501, "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": 2501, "staged_skel": 0, "mean_count": 1.6717313528060913, "n_util_obs": 0, "util_count_spearman": NaN}, "probe_overhead_sec_total": 0.003, "probe_overhead_seq_frac": 0.0, "util_loss": 0.0, "alpha": 0.03, "t": 6.4} |
| {"event": "probe", "step": 1, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 6300.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.362007, "cps": 49.21875, "mean_p": 0.496454, "boundary_entropy": 0.686986, "bpb": 8.131341}, "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": 8.151598}, "fi": {"chunks": 8819.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.468821, "cps": 68.898438, "mean_p": 0.500418, "boundary_entropy": 0.687508, "bpb": 8.191543}}, "bpb": {"en": 8.13134, "tr": 8.1516, "fi": 8.19154}, "gini_cpc": 0.107159, "gini_cps": 0.11762, "bpic": 2.788087, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.7979058719549246, "fi": 1.295063400087588}, "inventory": {"size": 2501, "admitted": 2501, "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": 2501, "staged_skel": 0, "mean_count": 1.6717313528060913, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 7.1} |
| {"event": "probe", "step": 2, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 6177.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.354939, "cps": 48.257812, "mean_p": 0.495941, "boundary_entropy": 0.686967, "bpb": 8.067232}, "tr": {"chunks": 5130.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.283378, "cps": 40.078125, "mean_p": 0.488986, "boundary_entropy": 0.687255, "bpb": 8.107643}, "fi": {"chunks": 8681.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.461485, "cps": 67.820312, "mean_p": 0.4999, "boundary_entropy": 0.687499, "bpb": 8.135731}}, "bpb": {"en": 8.06723, "tr": 8.10764, "fi": 8.13573}, "gini_cpc": 0.107963, "gini_cps": 0.118438, "bpic": 2.838303, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.7983867715892737, "fi": 1.3001827255910245}, "inventory": {"size": 4372, "admitted": 4372, "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": 4372, "staged_skel": 0, "mean_count": 1.9675205945968628, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 9.0} |
| {"event": "probe", "step": 4, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 5684.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.32661, "cps": 44.40625, "mean_p": 0.494085, "boundary_entropy": 0.68689, "bpb": 7.851187}, "tr": {"chunks": 4765.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.263216, "cps": 37.226562, "mean_p": 0.48759, "boundary_entropy": 0.687169, "bpb": 7.954375}, "fi": {"chunks": 8088.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.429961, "cps": 63.1875, "mean_p": 0.497984, "boundary_entropy": 0.687458, "bpb": 7.946167}}, "bpb": {"en": 7.85119, "tr": 7.95438, "fi": 7.94617}, "gini_cpc": 0.109006, "gini_cps": 0.119509, "bpic": 3.060474, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.8059023171924606, "fi": 1.3164346657934547}, "inventory": {"size": 8071, "admitted": 8071, "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": 8071, "staged_skel": 0, "mean_count": 2.1132450103759766, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 11.0} |
| {"event": "probe", "step": 6, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 5030.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.289031, "cps": 39.296875, "mean_p": 0.491499, "boundary_entropy": 0.686759, "bpb": 7.564359}, "tr": {"chunks": 4229.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.233608, "cps": 33.039062, "mean_p": 0.485659, "boundary_entropy": 0.687037, "bpb": 7.739806}, "fi": {"chunks": 7265.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.38621, "cps": 56.757812, "mean_p": 0.495251, "boundary_entropy": 0.687375, "bpb": 7.687204}}, "bpb": {"en": 7.56436, "tr": 7.73981, "fi": 7.6872}, "gini_cpc": 0.111938, "gini_cps": 0.122489, "bpic": 3.433309, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.8082454507886119, "fi": 1.3362258536389975}, "inventory": {"size": 11659, "admitted": 11659, "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": 11659, "staged_skel": 0, "mean_count": 2.1088428497314453, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 12.6} |
| {"event": "probe", "step": 8, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 4411.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.253462, "cps": 34.460938, "mean_p": 0.488955, "boundary_entropy": 0.686607, "bpb": 7.288457}, "tr": {"chunks": 3808.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.210352, "cps": 29.75, "mean_p": 0.483802, "boundary_entropy": 0.686896, "bpb": 7.525245}, "fi": {"chunks": 6403.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.340386, "cps": 50.023438, "mean_p": 0.492305, "boundary_entropy": 0.687255, "bpb": 7.436075}}, "bpb": {"en": 7.28846, "tr": 7.52525, "fi": 7.43608}, "gini_cpc": 0.107796, "gini_cps": 0.118315, "bpic": 3.879907, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.8299146876522694, "fi": 1.3429464045098423}, "inventory": {"size": 15162, "admitted": 15162, "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": 15162, "staged_skel": 0, "mean_count": 2.06219482421875, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 15.0} |
| {"event": "probe", "step": 16, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 4509.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.259093, "cps": 35.226562, "mean_p": 0.491279, "boundary_entropy": 0.686941, "bpb": 6.660485}, "tr": {"chunks": 4634.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.25598, "cps": 36.203125, "mean_p": 0.488632, "boundary_entropy": 0.687366, "bpb": 6.978171}, "fi": {"chunks": 5659.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.300835, "cps": 44.210938, "mean_p": 0.491376, "boundary_entropy": 0.687415, "bpb": 6.839534}}, "bpb": {"en": 6.66048, "tr": 6.97817, "fi": 6.83953}, "gini_cpc": 0.03665, "gini_cps": 0.051795, "bpic": 3.832725, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9879827522035235, "fi": 1.161105535106669}, "inventory": {"size": 28371, "admitted": 28371, "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": 28371, "staged_skel": 0, "mean_count": 1.8489302396774292, "n_util_obs": 3, "util_count_spearman": -0.5}, "t": 19.4} |
| {"event": "probe", "step": 32, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3926.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.225593, "cps": 30.671875, "mean_p": 0.494328, "boundary_entropy": 0.687539, "bpb": 6.307007}, "tr": {"chunks": 3966.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.21908, "cps": 30.984375, "mean_p": 0.491602, "boundary_entropy": 0.687941, "bpb": 6.636179}, "fi": {"chunks": 4164.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.22136, "cps": 32.53125, "mean_p": 0.492522, "boundary_entropy": 0.687979, "bpb": 6.431252}}, "bpb": {"en": 6.30701, "tr": 6.63618, "fi": 6.43125}, "gini_cpc": 0.00652, "gini_cps": 0.013161, "bpic": 4.705707, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9711268982726313, "fi": 0.9812341674873849}, "inventory": {"size": 52915, "admitted": 52915, "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": 52915, "staged_skel": 0, "mean_count": 1.8330150842666626, "n_util_obs": 4, "util_count_spearman": 0.0}, "t": 24.3} |
| {"event": "consolidate", "step": 50, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 29.2} |
| {"event": "train", "step": 50, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 4.22553, "bpb": 6.09615, "chunk_loss": 0.032403, "F": 0.20044, "G": 0.4921, "chunks": 205.25, "bytes_per_chunk": 4.989, "boundary_entropy": 0.69212, "router_noise": 0.0, "grad_norm": 2.4395, "lr": 5.1e-05, "router_trainable": true, "sec_per_step": 0.464, "probe_sec": 0.1056, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09869, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.00952, "n_shadow": 0, "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": 0.35876, "trunc_abs_frac": 0.35038, "mean_delta_full": -0.04321, "read_hit_rate": {"skel": 0.58396}, "mean_gate": {"skel": 0.12048}, "n_reads": 2097, "n_cand": 3591, "read_suppressed_frac": 0.0944, "dropout_reads_logged": 2097, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 2.1795654296875, "n_util_obs": 11, "util_count_spearman": -0.01818181818181818}, "probe_overhead_sec_total": 2.424, "probe_overhead_seq_frac": 0.4835, "util_loss": 0.049938, "alpha": 0.03, "t": 29.6} |
| {"event": "probe", "step": 64, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3880.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.22295, "cps": 30.3125, "mean_p": 0.487591, "boundary_entropy": 0.686572, "bpb": 5.623575}, "tr": {"chunks": 3673.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.202895, "cps": 28.695312, "mean_p": 0.487377, "boundary_entropy": 0.687466, "bpb": 6.010516}, "fi": {"chunks": 5116.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.271969, "cps": 39.96875, "mean_p": 0.49229, "boundary_entropy": 0.687672, "bpb": 5.817856}}, "bpb": {"en": 5.62358, "tr": 6.01052, "fi": 5.81786}, "gini_cpc": 0.065991, "gini_cps": 0.075933, "bpic": 4.478017, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9100447980655938, "fi": 1.2198629667769518}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 2.5645599365234375, "n_util_obs": 14, "util_count_spearman": -0.11208791208791209}, "t": 33.6} |
| {"event": "consolidate", "step": 100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 45.3} |
| {"event": "train", "step": 100, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 3.20621, "bpb": 4.62558, "chunk_loss": 0.031449, "F": 0.19296, "G": 0.422, "chunks": 197.59, "bytes_per_chunk": 5.182, "boundary_entropy": 0.66654, "router_noise": 0.0, "grad_norm": 2.4704, "lr": 0.000101, "router_trainable": true, "sec_per_step": 0.322, "probe_sec": 0.0384, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03225, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.00648, "n_shadow": 0, "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": 0.08485, "trunc_abs_frac": 1.00576, "mean_delta_full": -0.05719, "read_hit_rate": {"skel": 0.72105}, "mean_gate": {"skel": 0.12023}, "n_reads": 3306, "n_cand": 4585, "read_suppressed_frac": 0.0959, "dropout_reads_logged": 3306, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 4.56658935546875, "n_util_obs": 23, "util_count_spearman": 0.10474308300395258}, "probe_overhead_sec_total": 5.376, "probe_overhead_seq_frac": 0.4916, "util_loss": 0.001006, "alpha": 0.03, "t": 45.7} |
| {"event": "probe", "step": 128, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3556.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.204333, "cps": 27.78125, "mean_p": 0.380831, "boundary_entropy": 0.612388, "bpb": 4.198971}, "tr": {"chunks": 2901.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.16025, "cps": 22.664062, "mean_p": 0.410013, "boundary_entropy": 0.650235, "bpb": 4.518261}, "fi": {"chunks": 3500.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.186061, "cps": 27.34375, "mean_p": 0.393379, "boundary_entropy": 0.635628, "bpb": 4.444545}}, "bpb": {"en": 4.19897, "tr": 4.51826, "fi": 4.44454}, "gini_cpc": 0.053371, "gini_cps": 0.043855, "bpic": 5.6977, "worst_lang": "en", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.7842590614001235, "fi": 0.9105808839441825}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 5.9349822998046875, "n_util_obs": 26, "util_count_spearman": 0.11863247863247864}, "t": 56.8} |
| {"event": "consolidate", "step": 150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 65.5} |
| {"event": "train", "step": 150, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.81636, "bpb": 4.06315, "chunk_loss": 0.031261, "F": 0.19705, "G": 0.3567, "chunks": 201.78, "bytes_per_chunk": 5.075, "boundary_entropy": 0.58942, "router_noise": 0.0, "grad_norm": 3.0267, "lr": 0.000151, "router_trainable": true, "sec_per_step": 0.403, "probe_sec": 0.0307, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02238, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.00924, "n_shadow": 0, "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": -0.05675, "trunc_abs_frac": 0.40002, "mean_delta_full": -0.07988, "read_hit_rate": {"skel": 0.66208}, "mean_gate": {"skel": 0.12134}, "n_reads": 2937, "n_cand": 4436, "read_suppressed_frac": 0.11, "dropout_reads_logged": 2937, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 7.064605712890625, "n_util_obs": 33, "util_count_spearman": -0.03442513368983957}, "probe_overhead_sec_total": 8.052, "probe_overhead_seq_frac": 0.4944, "util_loss": 0.000425, "alpha": 0.03, "t": 65.8} |
| {"event": "consolidate", "step": 200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 86.5} |
| {"event": "probe", "step": 200, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3179.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.18267, "cps": 24.835938, "mean_p": 0.382683, "boundary_entropy": 0.587685, "bpb": 3.626851}, "tr": {"chunks": 3535.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.195272, "cps": 27.617188, "mean_p": 0.383595, "boundary_entropy": 0.607699, "bpb": 3.774552}, "fi": {"chunks": 3300.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.175429, "cps": 25.78125, "mean_p": 0.38269, "boundary_entropy": 0.616837, "bpb": 3.864512}}, "bpb": {"en": 3.62685, "tr": 3.77455, "fi": 3.86451}, "gini_cpc": 0.023905, "gini_cps": 0.0237, "bpic": 5.665269, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0689870867025164, "fi": 0.9603635066650064}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 9.022705078125, "n_util_obs": 46, "util_count_spearman": -0.010545790934320074}, "t": 87.2} |
| {"event": "eval", "step": 200, "phase": 1, "bpb": {"en": 3.6338, "tr": 3.78429, "fi": 3.86397}, "bytes_per_chunk": {"en": 5.598, "tr": 5.413, "fi": 6.043}, "bpb_mean": 3.76069, "frozen_inventory": true, "t": 88.5} |
| {"event": "train", "step": 200, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.54471, "bpb": 3.67124, "chunk_loss": 0.030645, "F": 0.17932, "G": 0.38461, "chunks": 183.62, "bytes_per_chunk": 5.577, "boundary_entropy": 0.59575, "router_noise": 0.0, "grad_norm": 1.1953, "lr": 0.000201, "router_trainable": true, "sec_per_step": 0.473, "probe_sec": 0.0409, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03386, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.03018, "n_shadow": 0, "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": 0.60593, "trunc_abs_frac": 0.72916, "mean_delta_full": 0.11523, "read_hit_rate": {"skel": 0.64133}, "mean_gate": {"skel": 0.12483}, "n_reads": 2725, "n_cand": 4249, "read_suppressed_frac": 0.0965, "dropout_reads_logged": 2725, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 9.064956665039062, "n_util_obs": 46, "util_count_spearman": 0.004748689485044711}, "probe_overhead_sec_total": 10.524, "probe_overhead_seq_frac": 0.4958, "util_loss": 0.000462, "alpha": 0.03, "t": 89.5} |
| {"event": "consolidate", "step": 250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 105.5} |
| {"event": "train", "step": 250, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.39612, "bpb": 3.45687, "chunk_loss": 0.031868, "F": 0.21353, "G": 0.35392, "chunks": 218.66, "bytes_per_chunk": 4.683, "boundary_entropy": 0.54185, "router_noise": 0.0, "grad_norm": 1.6996, "lr": 0.000251, "router_trainable": true, "sec_per_step": 0.341, "probe_sec": 0.098, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08437, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.02182, "n_shadow": 0, "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": 0.19322, "trunc_abs_frac": 0.49194, "mean_delta_full": 0.08089, "read_hit_rate": {"skel": 0.6912}, "mean_gate": {"skel": 0.1331}, "n_reads": 3494, "n_cand": 5055, "read_suppressed_frac": 0.0944, "dropout_reads_logged": 3494, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 11.272476196289062, "n_util_obs": 54, "util_count_spearman": 0.24551172098341909}, "probe_overhead_sec_total": 13.402, "probe_overhead_seq_frac": 0.4966, "util_loss": 0.000512, "alpha": 0.03, "t": 106.5} |
| {"event": "probe", "step": 256, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3230.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.1856, "cps": 25.234375, "mean_p": 0.325672, "boundary_entropy": 0.516737, "bpb": 3.346801}, "tr": {"chunks": 3317.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.183229, "cps": 25.914062, "mean_p": 0.342857, "boundary_entropy": 0.528743, "bpb": 3.531546}, "fi": {"chunks": 4067.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.216203, "cps": 31.773438, "mean_p": 0.374913, "boundary_entropy": 0.575947, "bpb": 3.63793}}, "bpb": {"en": 3.3468, "tr": 3.53155, "fi": 3.63793}, "gini_cpc": 0.037575, "gini_cps": 0.052572, "bpic": 5.345016, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9872258485115016, "fi": 1.1648872366195064}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 11.538253784179688, "n_util_obs": 56, "util_count_spearman": 0.22850307587149693}, "t": 110.1} |
| {"event": "consolidate", "step": 300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 136.0} |
| {"event": "train", "step": 300, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.28037, "bpb": 3.28988, "chunk_loss": 0.031057, "F": 0.20181, "G": 0.29935, "chunks": 206.66, "bytes_per_chunk": 4.955, "boundary_entropy": 0.43781, "router_noise": 0.0, "grad_norm": 1.2711, "lr": 0.0003, "router_trainable": true, "sec_per_step": 0.596, "probe_sec": 0.0964, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08817, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.03924, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 8, "trunc_ratio_sum": -0.76317, "trunc_abs_frac": 0.68655, "mean_delta_full": 0.04495, "read_hit_rate": {"skel": 0.67235}, "mean_gate": {"skel": 0.15388}, "n_reads": 3271, "n_cand": 4865, "read_suppressed_frac": 0.1042, "dropout_reads_logged": 3271, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 13.824920654296875, "n_util_obs": 71, "util_count_spearman": 0.18319919517102615}, "probe_overhead_sec_total": 16.2, "probe_overhead_seq_frac": 0.4972, "util_loss": 0.001061, "alpha": 0.03, "t": 136.3} |
| {"event": "consolidate", "step": 350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 164.1} |
| {"event": "train", "step": 350, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.19601, "bpb": 3.16817, "chunk_loss": 0.030989, "F": 0.20175, "G": 0.2946, "chunks": 206.59, "bytes_per_chunk": 4.957, "boundary_entropy": 0.40965, "router_noise": 0.0, "grad_norm": 1.222, "lr": 0.0002999, "router_trainable": true, "sec_per_step": 0.561, "probe_sec": 0.0393, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03188, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.01111, "n_shadow": 0, "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": -0.01738, "trunc_abs_frac": 0.46973, "mean_delta_full": 0.28215, "read_hit_rate": {"skel": 0.7071}, "mean_gate": {"skel": 0.18591}, "n_reads": 3416, "n_cand": 4831, "read_suppressed_frac": 0.0934, "dropout_reads_logged": 3416, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 16.540756225585938, "n_util_obs": 85, "util_count_spearman": 0.30953683799101034}, "probe_overhead_sec_total": 19.003, "probe_overhead_seq_frac": 0.4976, "util_loss": 0.001291, "alpha": 0.03, "t": 164.3} |
| {"event": "consolidate", "step": 400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 191.2} |
| {"event": "probe", "step": 400, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3659.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.210251, "cps": 28.585938, "mean_p": 0.270341, "boundary_entropy": 0.335788, "bpb": 2.901514}, "tr": {"chunks": 3893.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.215047, "cps": 30.414062, "mean_p": 0.295185, "boundary_entropy": 0.425355, "bpb": 3.191674}, "fi": {"chunks": 4385.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.233108, "cps": 34.257812, "mean_p": 0.293733, "boundary_entropy": 0.412056, "bpb": 3.376031}}, "bpb": {"en": 2.90151, "tr": 3.19167, "fi": 3.37603}, "gini_cpc": 0.023144, "gini_cps": 0.040546, "bpic": 4.752618, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.022811407264649, "fi": 1.1087137280439365}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 19.353195190429688, "n_util_obs": 92, "util_count_spearman": 0.2647997164126196}, "t": 191.9} |
| {"event": "eval", "step": 400, "phase": 1, "bpb": {"en": 2.91668, "tr": 3.19821, "fi": 3.36889}, "bytes_per_chunk": {"en": 4.494, "tr": 5.134, "fi": 4.546}, "bpb_mean": 3.16126, "frozen_inventory": true, "t": 193.1} |
| {"event": "train", "step": 400, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.04587, "bpb": 2.95156, "chunk_loss": 0.03143, "F": 0.22134, "G": 0.28236, "chunks": 226.66, "bytes_per_chunk": 4.518, "boundary_entropy": 0.34418, "router_noise": 0.0, "grad_norm": 1.1251, "lr": 0.0002998, "router_trainable": true, "sec_per_step": 0.612, "probe_sec": 0.0707, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05959, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.02973, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.63364, "trunc_abs_frac": 0.61512, "mean_delta_full": -0.07153, "read_hit_rate": {"skel": 0.73464}, "mean_gate": {"skel": 0.22857}, "n_reads": 3671, "n_cand": 4997, "read_suppressed_frac": 0.1041, "dropout_reads_logged": 3671, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 19.409866333007812, "n_util_obs": 94, "util_count_spearman": 0.30324135116099726}, "probe_overhead_sec_total": 22.135, "probe_overhead_seq_frac": 0.4979, "util_loss": 0.001884, "alpha": 0.03, "t": 194.9} |
| {"event": "consolidate", "step": 450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 217.3} |
| {"event": "train", "step": 450, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.01438, "bpb": 2.90614, "chunk_loss": 0.031303, "F": 0.21896, "G": 0.27579, "chunks": 224.22, "bytes_per_chunk": 4.567, "boundary_entropy": 0.32896, "router_noise": 0.0, "grad_norm": 1.489, "lr": 0.0002995, "router_trainable": true, "sec_per_step": 0.453, "probe_sec": 0.039, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03414, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.02162, "n_shadow": 0, "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": -8.12111, "trunc_abs_frac": 0.7402, "mean_delta_full": -0.00916, "read_hit_rate": {"skel": 0.7411}, "mean_gate": {"skel": 0.2448}, "n_reads": 3790, "n_cand": 5114, "read_suppressed_frac": 0.0987, "dropout_reads_logged": 3790, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 22.277908325195312, "n_util_obs": 99, "util_count_spearman": 0.2667408781694496}, "probe_overhead_sec_total": 25.049, "probe_overhead_seq_frac": 0.4981, "util_loss": 0.002537, "alpha": 0.03, "t": 217.6} |
| {"event": "consolidate", "step": 500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 230.8} |
| {"event": "train", "step": 500, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.99133, "bpb": 2.87289, "chunk_loss": 0.031511, "F": 0.22699, "G": 0.27581, "chunks": 232.44, "bytes_per_chunk": 4.405, "boundary_entropy": 0.31106, "router_noise": 0.0, "grad_norm": 1.1314, "lr": 0.0002992, "router_trainable": true, "sec_per_step": 0.269, "probe_sec": 0.0511, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0398, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.06594, "n_shadow": 0, "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": {"skel": 0.74319}, "mean_gate": {"skel": 0.24071}, "n_reads": 3982, "n_cand": 5358, "read_suppressed_frac": 0.1025, "dropout_reads_logged": 3982, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 25.232574462890625, "n_util_obs": 104, "util_count_spearman": 0.41678224687933424}, "probe_overhead_sec_total": 27.663, "probe_overhead_seq_frac": 0.4983, "util_loss": 0.003599, "alpha": 0.03, "t": 231.0} |
| {"event": "probe", "step": 512, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3818.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.219387, "cps": 29.828125, "mean_p": 0.262463, "boundary_entropy": 0.298952, "bpb": 2.695847}, "tr": {"chunks": 3667.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.202563, "cps": 28.648438, "mean_p": 0.273839, "boundary_entropy": 0.363753, "bpb": 3.03183}, "fi": {"chunks": 4571.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.242996, "cps": 35.710938, "mean_p": 0.282131, "boundary_entropy": 0.345329, "bpb": 3.245886}}, "bpb": {"en": 2.69585, "tr": 3.03183, "fi": 3.24589}, "gini_cpc": 0.040538, "gini_cps": 0.049989, "bpic": 4.705707, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.923312158784549, "fi": 1.1076116982810371}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 25.908309936523438, "n_util_obs": 106, "util_count_spearman": 0.4011547501202998}, "t": 234.7} |
| {"event": "consolidate", "step": 550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 245.7} |
| {"event": "train", "step": 550, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.96793, "bpb": 2.83913, "chunk_loss": 0.031468, "F": 0.22626, "G": 0.27494, "chunks": 231.69, "bytes_per_chunk": 4.42, "boundary_entropy": 0.30154, "router_noise": 0.0, "grad_norm": 1.1988, "lr": 0.0002987, "router_trainable": true, "sec_per_step": 0.299, "probe_sec": 0.0572, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04524, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.03528, "n_shadow": 0, "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": {"skel": 0.7378}, "mean_gate": {"skel": 0.24038}, "n_reads": 4038, "n_cand": 5473, "read_suppressed_frac": 0.0944, "dropout_reads_logged": 4038, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 28.26239013671875, "n_util_obs": 113, "util_count_spearman": 0.32504792837432733}, "probe_overhead_sec_total": 30.498, "probe_overhead_seq_frac": 0.4985, "util_loss": 0.004264, "alpha": 0.03, "t": 246.0} |
| {"event": "consolidate", "step": 600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 258.9} |
| {"event": "probe", "step": 600, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3845.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.220939, "cps": 30.039062, "mean_p": 0.261719, "boundary_entropy": 0.281194, "bpb": 2.581158}, "tr": {"chunks": 3715.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.205215, "cps": 29.023438, "mean_p": 0.266203, "boundary_entropy": 0.344661, "bpb": 2.935861}, "fi": {"chunks": 4626.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.24592, "cps": 36.140625, "mean_p": 0.282516, "boundary_entropy": 0.318371, "bpb": 3.138163}}, "bpb": {"en": 2.58116, "tr": 2.93586, "fi": 3.13816}, "gini_cpc": 0.040378, "gini_cps": 0.049839, "bpic": 4.655506, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9288295907100584, "fi": 1.1130675484608066}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 31.27252197265625, "n_util_obs": 122, "util_count_spearman": 0.32706586786772895}, "t": 259.7} |
| {"event": "eval", "step": 600, "phase": 1, "bpb": {"en": 2.59509, "tr": 2.94496, "fi": 3.1385}, "bytes_per_chunk": {"en": 4.327, "tr": 5.525, "fi": 4.311}, "bpb_mean": 2.89285, "frozen_inventory": true, "t": 260.9} |
| {"event": "train", "step": 600, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.87032, "bpb": 2.6983, "chunk_loss": 0.031322, "F": 0.22406, "G": 0.26933, "chunks": 229.44, "bytes_per_chunk": 4.463, "boundary_entropy": 0.28727, "router_noise": 0.0, "grad_norm": 1.2282, "lr": 0.0002982, "router_trainable": true, "sec_per_step": 0.324, "probe_sec": 0.0627, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05283, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.08353, "n_shadow": 0, "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": -71.65909, "trunc_abs_frac": 1.06076, "mean_delta_full": -0.00179, "read_hit_rate": {"skel": 0.74972}, "mean_gate": {"skel": 0.24642}, "n_reads": 4062, "n_cand": 5418, "read_suppressed_frac": 0.0894, "dropout_reads_logged": 4062, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 31.334152221679688, "n_util_obs": 122, "util_count_spearman": 0.32925396660218326}, "probe_overhead_sec_total": 33.088, "probe_overhead_seq_frac": 0.4986, "util_loss": 0.005471, "alpha": 0.03, "t": 262.2} |
| {"event": "consolidate", "step": 650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 275.1} |
| {"event": "train", "step": 650, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.89786, "bpb": 2.73804, "chunk_loss": 0.031423, "F": 0.22827, "G": 0.26798, "chunks": 233.75, "bytes_per_chunk": 4.381, "boundary_entropy": 0.2794, "router_noise": 0.0, "grad_norm": 0.981, "lr": 0.0002975, "router_trainable": true, "sec_per_step": 0.274, "probe_sec": 0.0695, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05845, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.01925, "n_shadow": 0, "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": 2.63889, "trunc_abs_frac": 1.40293, "mean_delta_full": 0.01831, "read_hit_rate": {"skel": 0.76374}, "mean_gate": {"skel": 0.2518}, "n_reads": 4238, "n_cand": 5549, "read_suppressed_frac": 0.1017, "dropout_reads_logged": 4238, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 34.488525390625, "n_util_obs": 132, "util_count_spearman": 0.32233338986034743}, "probe_overhead_sec_total": 35.783, "probe_overhead_seq_frac": 0.4987, "util_loss": 0.006188, "alpha": 0.03, "t": 275.9} |
| {"event": "consolidate", "step": 700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 288.3} |
| {"event": "train", "step": 700, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.84704, "bpb": 2.66471, "chunk_loss": 0.031384, "F": 0.22916, "G": 0.2651, "chunks": 234.66, "bytes_per_chunk": 4.364, "boundary_entropy": 0.26632, "router_noise": 0.0, "grad_norm": 1.1914, "lr": 0.0002967, "router_trainable": true, "sec_per_step": 0.254, "probe_sec": 0.1141, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10611, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.43837, "n_shadow": 0, "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": 1.01987, "trunc_abs_frac": 0.94157, "mean_delta_full": 1.03988, "read_hit_rate": {"skel": 0.76529}, "mean_gate": {"skel": 0.25257}, "n_reads": 4118, "n_cand": 5381, "read_suppressed_frac": 0.0932, "dropout_reads_logged": 4118, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 37.64903259277344, "n_util_obs": 144, "util_count_spearman": 0.21769194134748024}, "probe_overhead_sec_total": 38.855, "probe_overhead_seq_frac": 0.4988, "util_loss": 0.007164, "alpha": 0.03, "t": 288.6} |
| {"event": "consolidate", "step": 750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 301.3} |
| {"event": "train", "step": 750, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.80285, "bpb": 2.60096, "chunk_loss": 0.031111, "F": 0.21909, "G": 0.25763, "chunks": 224.34, "bytes_per_chunk": 4.564, "boundary_entropy": 0.2621, "router_noise": 0.0, "grad_norm": 1.0274, "lr": 0.0002959, "router_trainable": true, "sec_per_step": 0.264, "probe_sec": 0.1626, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.1522, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.23081, "n_shadow": 0, "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": -0.01622, "trunc_abs_frac": 0.53852, "mean_delta_full": 0.39921, "read_hit_rate": {"skel": 0.75704}, "mean_gate": {"skel": 0.25651}, "n_reads": 4116, "n_cand": 5437, "read_suppressed_frac": 0.1062, "dropout_reads_logged": 4116, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 40.877349853515625, "n_util_obs": 151, "util_count_spearman": 0.22537489140379016}, "probe_overhead_sec_total": 41.834, "probe_overhead_seq_frac": 0.4989, "util_loss": 0.00832, "alpha": 0.03, "t": 301.8} |
| {"event": "consolidate", "step": 800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 315.6} |
| {"event": "probe", "step": 800, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3825.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.21979, "cps": 29.882812, "mean_p": 0.248711, "boundary_entropy": 0.239831, "bpb": 2.423592}, "tr": {"chunks": 3578.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.197647, "cps": 27.953125, "mean_p": 0.256043, "boundary_entropy": 0.307658, "bpb": 2.772964}, "fi": {"chunks": 4706.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.250173, "cps": 36.765625, "mean_p": 0.28355, "boundary_entropy": 0.27753, "bpb": 2.953911}}, "bpb": {"en": 2.42359, "tr": 2.77296, "fi": 2.95391}, "gini_cpc": 0.052452, "gini_cps": 0.062103, "bpic": 4.68511, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.8992541805983841, "fi": 1.1382370557068715}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 44.067840576171875, "n_util_obs": 164, "util_count_spearman": 0.28672674233203965}, "t": 316.3} |
| {"event": "eval", "step": 800, "phase": 1, "bpb": {"en": 2.43687, "tr": 2.7824, "fi": 2.96021}, "bytes_per_chunk": {"en": 4.285, "tr": 5.668, "fi": 4.275}, "bpb_mean": 2.72649, "frozen_inventory": true, "t": 317.6} |
| {"event": "train", "step": 800, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.79874, "bpb": 2.59503, "chunk_loss": 0.031275, "F": 0.22675, "G": 0.26012, "chunks": 232.19, "bytes_per_chunk": 4.41, "boundary_entropy": 0.25222, "router_noise": 0.0, "grad_norm": 1.0353, "lr": 0.0002949, "router_trainable": true, "sec_per_step": 0.343, "probe_sec": 0.0299, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02459, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.03546, "n_shadow": 0, "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": {"skel": 0.76462}, "mean_gate": {"skel": 0.26578}, "n_reads": 4145, "n_cand": 5421, "read_suppressed_frac": 0.0984, "dropout_reads_logged": 4145, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 44.130889892578125, "n_util_obs": 165, "util_count_spearman": 0.2887522890931557}, "probe_overhead_sec_total": 44.601, "probe_overhead_seq_frac": 0.4989, "util_loss": 0.009573, "alpha": 0.03, "t": 319.0} |
| {"event": "consolidate", "step": 850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 331.3} |
| {"event": "train", "step": 850, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.76691, "bpb": 2.54912, "chunk_loss": 0.031109, "F": 0.22116, "G": 0.25635, "chunks": 226.47, "bytes_per_chunk": 4.522, "boundary_entropy": 0.24547, "router_noise": 0.0, "grad_norm": 1.0776, "lr": 0.0002938, "router_trainable": true, "sec_per_step": 0.253, "probe_sec": 0.0286, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02011, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.18251, "n_shadow": 0, "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": 0.80212, "trunc_abs_frac": 0.86299, "mean_delta_full": 0.41007, "read_hit_rate": {"skel": 0.7467}, "mean_gate": {"skel": 0.26164}, "n_reads": 3965, "n_cand": 5310, "read_suppressed_frac": 0.1014, "dropout_reads_logged": 3965, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 47.29450988769531, "n_util_obs": 177, "util_count_spearman": 0.22682380132348312}, "probe_overhead_sec_total": 47.674, "probe_overhead_seq_frac": 0.499, "util_loss": 0.009481, "alpha": 0.03, "t": 331.6} |
| {"event": "consolidate", "step": 900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 343.8} |
| {"event": "train", "step": 900, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.61571, "bpb": 2.33098, "chunk_loss": 0.030922, "F": 0.21396, "G": 0.24995, "chunks": 219.09, "bytes_per_chunk": 4.674, "boundary_entropy": 0.2376, "router_noise": 0.0, "grad_norm": 0.9517, "lr": 0.0002927, "router_trainable": true, "sec_per_step": 0.248, "probe_sec": 0.0423, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03664, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.09338, "n_shadow": 0, "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": 2.4603, "trunc_abs_frac": 2.95499, "mean_delta_full": 0.13375, "read_hit_rate": {"skel": 0.76748}, "mean_gate": {"skel": 0.2597}, "n_reads": 4017, "n_cand": 5234, "read_suppressed_frac": 0.1001, "dropout_reads_logged": 4017, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 50.44950866699219, "n_util_obs": 188, "util_count_spearman": 0.2614794467109861}, "probe_overhead_sec_total": 50.167, "probe_overhead_seq_frac": 0.4991, "util_loss": 0.009887, "alpha": 0.03, "t": 344.0} |
| {"event": "consolidate", "step": 950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 359.8} |
| {"event": "train", "step": 950, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.77503, "bpb": 2.56082, "chunk_loss": 0.031073, "F": 0.22076, "G": 0.25288, "chunks": 226.06, "bytes_per_chunk": 4.53, "boundary_entropy": 0.23527, "router_noise": 0.0, "grad_norm": 0.95, "lr": 0.0002914, "router_trainable": true, "sec_per_step": 0.324, "probe_sec": 0.1337, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.12281, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.22139, "n_shadow": 0, "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": 1.29825, "trunc_abs_frac": 1.76013, "mean_delta_full": 0.23043, "read_hit_rate": {"skel": 0.76067}, "mean_gate": {"skel": 0.25186}, "n_reads": 4046, "n_cand": 5319, "read_suppressed_frac": 0.1013, "dropout_reads_logged": 4046, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 53.56695556640625, "n_util_obs": 197, "util_count_spearman": 0.259570058126893}, "probe_overhead_sec_total": 53.158, "probe_overhead_seq_frac": 0.4991, "util_loss": 0.009805, "alpha": 0.03, "t": 360.2} |
| {"event": "consolidate", "step": 1000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 376.9} |
| {"event": "probe", "step": 1000, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3657.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.210136, "cps": 28.570312, "mean_p": 0.238785, "boundary_entropy": 0.224064, "bpb": 2.338989}, "tr": {"chunks": 3645.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.201348, "cps": 28.476562, "mean_p": 0.249441, "boundary_entropy": 0.276991, "bpb": 2.639607}, "fi": {"chunks": 4666.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.248046, "cps": 36.453125, "mean_p": 0.27638, "boundary_entropy": 0.236917, "bpb": 2.821294}}, "bpb": {"en": 2.33899, "tr": 2.63961, "fi": 2.82129}, "gini_cpc": 0.047204, "gini_cps": 0.056874, "bpic": 4.740307, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9581778807685871, "fi": 1.1804076376689343}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 56.617584228515625, "n_util_obs": 208, "util_count_spearman": 0.2656370954998747}, "t": 377.7} |
| {"event": "eval", "step": 1000, "phase": 1, "bpb": {"en": 2.34879, "tr": 2.64826, "fi": 2.82833}, "bytes_per_chunk": {"en": 4.518, "tr": 5.506, "fi": 4.314}, "bpb_mean": 2.60846, "frozen_inventory": true, "t": 379.0} |
| {"event": "train", "step": 1000, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.71776, "bpb": 2.4782, "chunk_loss": 0.031381, "F": 0.22864, "G": 0.25751, "chunks": 234.12, "bytes_per_chunk": 4.374, "boundary_entropy": 0.22711, "router_noise": 0.0, "grad_norm": 1.0016, "lr": 0.0002901, "router_trainable": true, "sec_per_step": 0.409, "probe_sec": 0.039, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02892, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.19994, "n_shadow": 0, "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": 0.92026, "trunc_abs_frac": 0.81241, "mean_delta_full": 0.58496, "read_hit_rate": {"skel": 0.77552}, "mean_gate": {"skel": 0.24147}, "n_reads": 4201, "n_cand": 5417, "read_suppressed_frac": 0.0909, "dropout_reads_logged": 4201, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 56.682403564453125, "n_util_obs": 208, "util_count_spearman": 0.2662413604112831}, "probe_overhead_sec_total": 55.773, "probe_overhead_seq_frac": 0.4992, "util_loss": 0.010407, "alpha": 0.03, "t": 380.7} |
| {"event": "probe", "step": 1024, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3678.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.211343, "cps": 28.734375, "mean_p": 0.237761, "boundary_entropy": 0.218615, "bpb": 2.325852}, "tr": {"chunks": 3626.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.200298, "cps": 28.328125, "mean_p": 0.247677, "boundary_entropy": 0.280477, "bpb": 2.647069}, "fi": {"chunks": 4698.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.249747, "cps": 36.703125, "mean_p": 0.278735, "boundary_entropy": 0.232461, "bpb": 2.82212}}, "bpb": {"en": 2.32585, "tr": 2.64707, "fi": 2.82212}, "gini_cpc": 0.049844, "gini_cps": 0.059546, "bpic": 4.726879, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.947740944760627, "fi": 1.1817171116514642}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 58.095367431640625, "n_util_obs": 213, "util_count_spearman": 0.2970930963562156}, "t": 388.6} |
| {"event": "consolidate", "step": 1050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 403.6} |
| {"event": "train", "step": 1050, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.74215, "bpb": 2.5134, "chunk_loss": 0.030913, "F": 0.21442, "G": 0.25103, "chunks": 219.56, "bytes_per_chunk": 4.664, "boundary_entropy": 0.22292, "router_noise": 0.0, "grad_norm": 0.9039, "lr": 0.0002886, "router_trainable": true, "sec_per_step": 0.465, "probe_sec": 0.04, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03122, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.05584, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 1.35552, "trunc_abs_frac": 1.35552, "mean_delta_full": -0.44568, "read_hit_rate": {"skel": 0.7284}, "mean_gate": {"skel": 0.25531}, "n_reads": 3835, "n_cand": 5265, "read_suppressed_frac": 0.1012, "dropout_reads_logged": 3835, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 59.75718688964844, "n_util_obs": 219, "util_count_spearman": 0.26834036164899394}, "probe_overhead_sec_total": 58.556, "probe_overhead_seq_frac": 0.4992, "util_loss": 0.009701, "alpha": 0.03, "t": 403.9} |
| {"event": "consolidate", "step": 1100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 430.9} |
| {"event": "train", "step": 1100, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.72518, "bpb": 2.48891, "chunk_loss": 0.030959, "F": 0.21588, "G": 0.24929, "chunks": 221.06, "bytes_per_chunk": 4.632, "boundary_entropy": 0.21653, "router_noise": 0.0, "grad_norm": 0.9496, "lr": 0.0002871, "router_trainable": true, "sec_per_step": 0.543, "probe_sec": 0.0284, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.023, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.07489, "n_shadow": 0, "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": -0.3208, "trunc_abs_frac": 0.96779, "mean_delta_full": 0.18448, "read_hit_rate": {"skel": 0.75812}, "mean_gate": {"skel": 0.25425}, "n_reads": 4156, "n_cand": 5482, "read_suppressed_frac": 0.0936, "dropout_reads_logged": 4156, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 62.805938720703125, "n_util_obs": 228, "util_count_spearman": 0.24906080077842443}, "probe_overhead_sec_total": 61.566, "probe_overhead_seq_frac": 0.4992, "util_loss": 0.011947, "alpha": 0.03, "t": 431.1} |
| {"event": "consolidate", "step": 1150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 456.0} |
| {"event": "train", "step": 1150, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.59034, "bpb": 2.29438, "chunk_loss": 0.03078, "F": 0.2114, "G": 0.24047, "chunks": 216.47, "bytes_per_chunk": 4.73, "boundary_entropy": 0.20906, "router_noise": 0.0, "grad_norm": 0.8785, "lr": 0.0002855, "router_trainable": true, "sec_per_step": 0.511, "probe_sec": 0.0387, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02649, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.10651, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 8, "trunc_ratio_sum": 0.75667, "trunc_abs_frac": 0.88728, "mean_delta_full": 0.31332, "read_hit_rate": {"skel": 0.75118}, "mean_gate": {"skel": 0.25182}, "n_reads": 3979, "n_cand": 5297, "read_suppressed_frac": 0.1038, "dropout_reads_logged": 3979, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 65.79635620117188, "n_util_obs": 238, "util_count_spearman": 0.2767445353747828}, "probe_overhead_sec_total": 64.313, "probe_overhead_seq_frac": 0.4993, "util_loss": 0.01243, "alpha": 0.03, "t": 456.6} |
| {"event": "consolidate", "step": 1200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 480.1} |
| {"event": "probe", "step": 1200, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3462.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.198931, "cps": 27.046875, "mean_p": 0.23016, "boundary_entropy": 0.205348, "bpb": 2.274438}, "tr": {"chunks": 3793.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.209523, "cps": 29.632812, "mean_p": 0.242747, "boundary_entropy": 0.248953, "bpb": 2.545492}, "fi": {"chunks": 4766.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.253362, "cps": 37.234375, "mean_p": 0.276965, "boundary_entropy": 0.212174, "bpb": 2.748375}}, "bpb": {"en": 2.27444, "tr": 2.54549, "fi": 2.74837}, "gini_cpc": 0.05483, "gini_cps": 0.072318, "bpic": 4.719408, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0532448589244428, "fi": 1.2736180672339747}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 68.67938232421875, "n_util_obs": 251, "util_count_spearman": 0.2637740070806766}, "t": 481.0} |
| {"event": "eval", "step": 1200, "phase": 1, "bpb": {"en": 2.28428, "tr": 2.55524, "fi": 2.75533}, "bytes_per_chunk": {"en": 4.743, "tr": 5.298, "fi": 4.262}, "bpb_mean": 2.53162, "frozen_inventory": true, "t": 482.1} |
| {"event": "train", "step": 1200, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.63088, "bpb": 2.35286, "chunk_loss": 0.030843, "F": 0.21393, "G": 0.2431, "chunks": 219.06, "bytes_per_chunk": 4.674, "boundary_entropy": 0.20689, "router_noise": 0.0, "grad_norm": 0.8097, "lr": 0.0002837, "router_trainable": true, "sec_per_step": 0.534, "probe_sec": 0.0248, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01952, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.07312, "n_shadow": 0, "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": 2.60437, "trunc_abs_frac": 1.57171, "mean_delta_full": 0.08135, "read_hit_rate": {"skel": 0.74957}, "mean_gate": {"skel": 0.25761}, "n_reads": 3921, "n_cand": 5231, "read_suppressed_frac": 0.1069, "dropout_reads_logged": 3921, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 68.73971557617188, "n_util_obs": 251, "util_count_spearman": 0.2626921386249599}, "probe_overhead_sec_total": 66.846, "probe_overhead_seq_frac": 0.4993, "util_loss": 0.014783, "alpha": 0.03, "t": 483.3} |
| {"event": "consolidate", "step": 1250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 502.2} |
| {"event": "train", "step": 1250, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.68241, "bpb": 2.42721, "chunk_loss": 0.030757, "F": 0.21103, "G": 0.2383, "chunks": 216.09, "bytes_per_chunk": 4.739, "boundary_entropy": 0.20102, "router_noise": 0.0, "grad_norm": 0.9521, "lr": 0.0002819, "router_trainable": true, "sec_per_step": 0.383, "probe_sec": 0.0443, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0372, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.0756, "n_shadow": 0, "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": 0.14042, "trunc_abs_frac": 0.28729, "mean_delta_full": -0.51593, "read_hit_rate": {"skel": 0.74361}, "mean_gate": {"skel": 0.25425}, "n_reads": 3959, "n_cand": 5324, "read_suppressed_frac": 0.094, "dropout_reads_logged": 3959, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 71.73234558105469, "n_util_obs": 261, "util_count_spearman": 0.28121507704711446}, "probe_overhead_sec_total": 69.301, "probe_overhead_seq_frac": 0.4993, "util_loss": 0.014011, "alpha": 0.03, "t": 502.4} |
| {"event": "consolidate", "step": 1300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 514.9} |
| {"event": "train", "step": 1300, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.62623, "bpb": 2.34615, "chunk_loss": 0.030685, "F": 0.20752, "G": 0.23534, "chunks": 212.5, "bytes_per_chunk": 4.819, "boundary_entropy": 0.19579, "router_noise": 0.0, "grad_norm": 0.8946, "lr": 0.00028, "router_trainable": true, "sec_per_step": 0.253, "probe_sec": 0.0275, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02116, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.44393, "n_shadow": 0, "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": -2.11118, "trunc_abs_frac": 0.82312, "mean_delta_full": -0.03202, "read_hit_rate": {"skel": 0.74816}, "mean_gate": {"skel": 0.25307}, "n_reads": 3957, "n_cand": 5289, "read_suppressed_frac": 0.1029, "dropout_reads_logged": 3957, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 74.7601318359375, "n_util_obs": 270, "util_count_spearman": 0.28301108597588875}, "probe_overhead_sec_total": 72.255, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.015006, "alpha": 0.03, "t": 515.1} |
| {"event": "consolidate", "step": 1350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 527.7} |
| {"event": "train", "step": 1350, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.61464, "bpb": 2.32943, "chunk_loss": 0.030904, "F": 0.2168, "G": 0.23807, "chunks": 222.0, "bytes_per_chunk": 4.613, "boundary_entropy": 0.18514, "router_noise": 0.0, "grad_norm": 0.9332, "lr": 0.000278, "router_trainable": true, "sec_per_step": 0.255, "probe_sec": 0.0277, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02122, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.11645, "n_shadow": 0, "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": 98.49808, "trunc_abs_frac": 2.23194, "mean_delta_full": -0.00159, "read_hit_rate": {"skel": 0.76791}, "mean_gate": {"skel": 0.25232}, "n_reads": 4106, "n_cand": 5347, "read_suppressed_frac": 0.0955, "dropout_reads_logged": 4106, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 77.76284790039062, "n_util_obs": 282, "util_count_spearman": 0.26363737790567643}, "probe_overhead_sec_total": 74.847, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.020961, "alpha": 0.03, "t": 527.8} |
| {"event": "consolidate", "step": 1400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 540.3} |
| {"event": "probe", "step": 1400, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3497.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.200942, "cps": 27.320312, "mean_p": 0.221128, "boundary_entropy": 0.171484, "bpb": 2.224773}, "tr": {"chunks": 3794.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.209579, "cps": 29.640625, "mean_p": 0.229588, "boundary_entropy": 0.216156, "bpb": 2.490141}, "fi": {"chunks": 4826.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.256552, "cps": 37.703125, "mean_p": 0.272139, "boundary_entropy": 0.187846, "bpb": 2.646081}}, "bpb": {"en": 2.22477, "tr": 2.49014, "fi": 2.64608}, "gini_cpc": 0.055576, "gini_cps": 0.07312, "bpic": 4.682017, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0429782768007634, "fi": 1.2767442835142289}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 80.719482421875, "n_util_obs": 296, "util_count_spearman": 0.24040898231363936}, "t": 541.1} |
| {"event": "eval", "step": 1400, "phase": 1, "bpb": {"en": 2.23411, "tr": 2.50258, "fi": 2.65488}, "bytes_per_chunk": {"en": 4.727, "tr": 5.321, "fi": 4.22}, "bpb_mean": 2.46386, "frozen_inventory": true, "t": 542.3} |
| {"event": "train", "step": 1400, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.69254, "bpb": 2.44181, "chunk_loss": 0.030755, "F": 0.21213, "G": 0.23484, "chunks": 217.22, "bytes_per_chunk": 4.714, "boundary_entropy": 0.18208, "router_noise": 0.0, "grad_norm": 0.8743, "lr": 0.0002759, "router_trainable": true, "sec_per_step": 0.308, "probe_sec": 0.044, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03812, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.80625, "n_shadow": 0, "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": 1.13522, "trunc_abs_frac": 0.80981, "mean_delta_full": 0.77883, "read_hit_rate": {"skel": 0.73834}, "mean_gate": {"skel": 0.26257}, "n_reads": 3801, "n_cand": 5148, "read_suppressed_frac": 0.1039, "dropout_reads_logged": 3801, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 80.77784729003906, "n_util_obs": 296, "util_count_spearman": 0.24145355575461355}, "probe_overhead_sec_total": 77.261, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.020802, "alpha": 0.03, "t": 543.2} |
| {"event": "consolidate", "step": 1450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 555.4} |
| {"event": "train", "step": 1450, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.64781, "bpb": 2.37729, "chunk_loss": 0.030809, "F": 0.21582, "G": 0.23602, "chunks": 221.0, "bytes_per_chunk": 4.633, "boundary_entropy": 0.18138, "router_noise": 0.0, "grad_norm": 0.7823, "lr": 0.0002738, "router_trainable": true, "sec_per_step": 0.248, "probe_sec": 0.0259, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02115, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.06945, "n_shadow": 0, "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": 0.54091, "trunc_abs_frac": 0.80041, "mean_delta_full": -0.32074, "read_hit_rate": {"skel": 0.74785}, "mean_gate": {"skel": 0.24946}, "n_reads": 3912, "n_cand": 5231, "read_suppressed_frac": 0.102, "dropout_reads_logged": 3912, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 83.80001831054688, "n_util_obs": 309, "util_count_spearman": 0.24444411453795611}, "probe_overhead_sec_total": 79.382, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.014674, "alpha": 0.03, "t": 555.6} |
| {"event": "consolidate", "step": 1500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 566.9} |
| {"event": "train", "step": 1500, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.60595, "bpb": 2.31689, "chunk_loss": 0.030652, "F": 0.20938, "G": 0.22877, "chunks": 214.41, "bytes_per_chunk": 4.776, "boundary_entropy": 0.17074, "router_noise": 0.0, "grad_norm": 0.8161, "lr": 0.0002715, "router_trainable": true, "sec_per_step": 0.23, "probe_sec": 0.0352, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02605, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.04579, "n_shadow": 0, "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": 5.01245, "trunc_abs_frac": 0.93571, "mean_delta_full": 0.02415, "read_hit_rate": {"skel": 0.74718}, "mean_gate": {"skel": 0.25305}, "n_reads": 3910, "n_cand": 5233, "read_suppressed_frac": 0.0967, "dropout_reads_logged": 3910, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 86.85275268554688, "n_util_obs": 327, "util_count_spearman": 0.1967106892774678}, "probe_overhead_sec_total": 82.018, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.015928, "alpha": 0.03, "t": 567.2} |
| {"event": "consolidate", "step": 1550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 579.5} |
| {"event": "train", "step": 1550, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.56955, "bpb": 2.26438, "chunk_loss": 0.030694, "F": 0.2121, "G": 0.22672, "chunks": 217.19, "bytes_per_chunk": 4.715, "boundary_entropy": 0.15996, "router_noise": 0.0, "grad_norm": 0.7259, "lr": 0.0002692, "router_trainable": true, "sec_per_step": 0.252, "probe_sec": 0.047, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03326, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.17004, "n_shadow": 0, "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": 0.30356, "trunc_abs_frac": 0.78059, "mean_delta_full": 0.16511, "read_hit_rate": {"skel": 0.75043}, "mean_gate": {"skel": 0.25228}, "n_reads": 3912, "n_cand": 5213, "read_suppressed_frac": 0.0938, "dropout_reads_logged": 3912, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 89.8592529296875, "n_util_obs": 337, "util_count_spearman": 0.18104163813765908}, "probe_overhead_sec_total": 84.579, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.014929, "alpha": 0.03, "t": 579.8} |
| {"event": "consolidate", "step": 1600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 591.6} |
| {"event": "probe", "step": 1600, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3590.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.206286, "cps": 28.046875, "mean_p": 0.218234, "boundary_entropy": 0.153572, "bpb": 2.182849}, "tr": {"chunks": 3827.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.211401, "cps": 29.898438, "mean_p": 0.221308, "boundary_entropy": 0.194635, "bpb": 2.430428}, "fi": {"chunks": 4852.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.257934, "cps": 37.90625, "mean_p": 0.266728, "boundary_entropy": 0.166004, "bpb": 2.575097}}, "bpb": {"en": 2.18285, "tr": 2.43043, "fi": 2.5751}, "gini_cpc": 0.050963, "gini_cps": 0.068574, "bpic": 4.624012, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0247963794917263, "fi": 1.2503701014149102}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 92.81910705566406, "n_util_obs": 348, "util_count_spearman": 0.18901989822140777}, "t": 592.4} |
| {"event": "eval", "step": 1600, "phase": 1, "bpb": {"en": 2.19339, "tr": 2.44229, "fi": 2.58268}, "bytes_per_chunk": {"en": 4.609, "tr": 5.291, "fi": 4.15}, "bpb_mean": 2.40612, "frozen_inventory": true, "t": 593.8} |
| {"event": "train", "step": 1600, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.57424, "bpb": 2.27115, "chunk_loss": 0.03079, "F": 0.21631, "G": 0.23019, "chunks": 221.5, "bytes_per_chunk": 4.623, "boundary_entropy": 0.15725, "router_noise": 0.0, "grad_norm": 0.8371, "lr": 0.0002668, "router_trainable": true, "sec_per_step": 0.304, "probe_sec": 0.0247, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.01994, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.04267, "n_shadow": 0, "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": -0.25625, "trunc_abs_frac": 10.3075, "mean_delta_full": 0.18304, "read_hit_rate": {"skel": 0.75626}, "mean_gate": {"skel": 0.25413}, "n_reads": 4015, "n_cand": 5309, "read_suppressed_frac": 0.1049, "dropout_reads_logged": 4015, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 92.88111877441406, "n_util_obs": 349, "util_count_spearman": 0.18991328704061836}, "probe_overhead_sec_total": 87.591, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.014926, "alpha": 0.03, "t": 595.0} |
| {"event": "consolidate", "step": 1650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 606.5} |
| {"event": "train", "step": 1650, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.59514, "bpb": 2.3013, "chunk_loss": 0.030731, "F": 0.21402, "G": 0.22872, "chunks": 219.16, "bytes_per_chunk": 4.672, "boundary_entropy": 0.15523, "router_noise": 0.0, "grad_norm": 0.8048, "lr": 0.0002643, "router_trainable": true, "sec_per_step": 0.234, "probe_sec": 0.0331, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02656, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.23635, "n_shadow": 0, "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": 1.67737, "trunc_abs_frac": 1.72168, "mean_delta_full": 0.15588, "read_hit_rate": {"skel": 0.75212}, "mean_gate": {"skel": 0.24446}, "n_reads": 3908, "n_cand": 5196, "read_suppressed_frac": 0.1013, "dropout_reads_logged": 3908, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 95.89637756347656, "n_util_obs": 355, "util_count_spearman": 0.20781375290573548}, "probe_overhead_sec_total": 89.879, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.015861, "alpha": 0.03, "t": 606.7} |
| {"event": "consolidate", "step": 1700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 618.6} |
| {"event": "train", "step": 1700, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.61347, "bpb": 2.32774, "chunk_loss": 0.030688, "F": 0.21176, "G": 0.2279, "chunks": 216.84, "bytes_per_chunk": 4.722, "boundary_entropy": 0.15859, "router_noise": 0.0, "grad_norm": 0.8129, "lr": 0.0002618, "router_trainable": true, "sec_per_step": 0.243, "probe_sec": 0.0322, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02545, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.07261, "n_shadow": 0, "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": {"skel": 0.72943}, "mean_gate": {"skel": 0.25035}, "n_reads": 3812, "n_cand": 5226, "read_suppressed_frac": 0.1065, "dropout_reads_logged": 3812, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 98.91572570800781, "n_util_obs": 368, "util_count_spearman": 0.19119595080125892}, "probe_overhead_sec_total": 92.228, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.016097, "alpha": 0.03, "t": 618.8} |
| {"event": "consolidate", "step": 1750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 630.8} |
| {"event": "train", "step": 1750, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.55942, "bpb": 2.24977, "chunk_loss": 0.030599, "F": 0.21014, "G": 0.22411, "chunks": 215.19, "bytes_per_chunk": 4.759, "boundary_entropy": 0.15526, "router_noise": 0.0, "grad_norm": 0.7449, "lr": 0.0002591, "router_trainable": true, "sec_per_step": 0.245, "probe_sec": 0.0556, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04074, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.03477, "n_shadow": 0, "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": 0.93941, "trunc_abs_frac": 0.66583, "mean_delta_full": -0.43835, "read_hit_rate": {"skel": 0.74392}, "mean_gate": {"skel": 0.24222}, "n_reads": 3884, "n_cand": 5221, "read_suppressed_frac": 0.103, "dropout_reads_logged": 3884, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 101.9117431640625, "n_util_obs": 376, "util_count_spearman": 0.19550687657787197}, "probe_overhead_sec_total": 94.457, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.018538, "alpha": 0.03, "t": 631.1} |
| {"event": "consolidate", "step": 1800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 644.2} |
| {"event": "probe", "step": 1800, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3491.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.200598, "cps": 27.273438, "mean_p": 0.214586, "boundary_entropy": 0.151055, "bpb": 2.14572}, "tr": {"chunks": 3757.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.207535, "cps": 29.351562, "mean_p": 0.220827, "boundary_entropy": 0.1873, "bpb": 2.388948}, "fi": {"chunks": 4857.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.2582, "cps": 37.945312, "mean_p": 0.26505, "boundary_entropy": 0.153136, "bpb": 2.51446}}, "bpb": {"en": 2.14572, "tr": 2.38895, "fi": 2.51446}, "gini_cpc": 0.057631, "gini_cps": 0.075231, "bpic": 4.686658, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0345819925711388, "fi": 1.2871539429876726}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 104.89265441894531, "n_util_obs": 387, "util_count_spearman": 0.22210567096954806}, "t": 645.1} |
| {"event": "eval", "step": 1800, "phase": 1, "bpb": {"en": 2.15842, "tr": 2.39936, "fi": 2.52463}, "bytes_per_chunk": {"en": 4.711, "tr": 5.33, "fi": 4.161}, "bpb_mean": 2.36081, "frozen_inventory": true, "t": 646.4} |
| {"event": "train", "step": 1800, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.53305, "bpb": 2.21172, "chunk_loss": 0.030574, "F": 0.20874, "G": 0.22171, "chunks": 213.75, "bytes_per_chunk": 4.791, "boundary_entropy": 0.15211, "router_noise": 0.0, "grad_norm": 0.7997, "lr": 0.0002564, "router_trainable": true, "sec_per_step": 0.333, "probe_sec": 0.0253, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0196, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.00775, "n_shadow": 0, "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": 0.93894, "trunc_abs_frac": 0.76167, "mean_delta_full": 0.3407, "read_hit_rate": {"skel": 0.74496}, "mean_gate": {"skel": 0.25024}, "n_reads": 3882, "n_cand": 5211, "read_suppressed_frac": 0.1108, "dropout_reads_logged": 3882, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 104.95263671875, "n_util_obs": 387, "util_count_spearman": 0.2211915147481902}, "probe_overhead_sec_total": 96.77, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.018638, "alpha": 0.03, "t": 647.7} |
| {"event": "consolidate", "step": 1850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 664.9} |
| {"event": "train", "step": 1850, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.56086, "bpb": 2.25184, "chunk_loss": 0.030699, "F": 0.21463, "G": 0.22701, "chunks": 219.78, "bytes_per_chunk": 4.659, "boundary_entropy": 0.14545, "router_noise": 0.0, "grad_norm": 0.8167, "lr": 0.0002537, "router_trainable": true, "sec_per_step": 0.348, "probe_sec": 0.05, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03995, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.24468, "n_shadow": 0, "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": -24.30556, "trunc_abs_frac": 0.64089, "mean_delta_full": 0.00293, "read_hit_rate": {"skel": 0.75649}, "mean_gate": {"skel": 0.2344}, "n_reads": 4051, "n_cand": 5355, "read_suppressed_frac": 0.1049, "dropout_reads_logged": 4051, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 107.97938537597656, "n_util_obs": 394, "util_count_spearman": 0.21756604040121277}, "probe_overhead_sec_total": 99.473, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.018967, "alpha": 0.03, "t": 665.2} |
| {"event": "consolidate", "step": 1900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 680.9} |
| {"event": "train", "step": 1900, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.49752, "bpb": 2.16047, "chunk_loss": 0.030672, "F": 0.21252, "G": 0.22567, "chunks": 217.62, "bytes_per_chunk": 4.705, "boundary_entropy": 0.14561, "router_noise": 0.0, "grad_norm": 0.7048, "lr": 0.0002508, "router_trainable": true, "sec_per_step": 0.319, "probe_sec": 0.0377, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.031, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.18738, "n_shadow": 0, "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": -0.15951, "trunc_abs_frac": 1.70382, "mean_delta_full": -0.50806, "read_hit_rate": {"skel": 0.76641}, "mean_gate": {"skel": 0.23703}, "n_reads": 4098, "n_cand": 5347, "read_suppressed_frac": 0.0947, "dropout_reads_logged": 4098, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 110.99777221679688, "n_util_obs": 408, "util_count_spearman": 0.2020632305463864}, "probe_overhead_sec_total": 101.869, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.016659, "alpha": 0.03, "t": 681.1} |
| {"event": "consolidate", "step": 1950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 702.0} |
| {"event": "train", "step": 1950, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.5725, "bpb": 2.26864, "chunk_loss": 0.030851, "F": 0.21826, "G": 0.23227, "chunks": 223.5, "bytes_per_chunk": 4.582, "boundary_entropy": 0.14526, "router_noise": 0.0, "grad_norm": 0.7797, "lr": 0.0002479, "router_trainable": true, "sec_per_step": 0.429, "probe_sec": 0.0469, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02549, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.18632, "n_shadow": 0, "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": 0.83801, "trunc_abs_frac": 6.37683, "mean_delta_full": 0.73358, "read_hit_rate": {"skel": 0.75776}, "mean_gate": {"skel": 0.24154}, "n_reads": 4176, "n_cand": 5511, "read_suppressed_frac": 0.1001, "dropout_reads_logged": 4176, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 114.00807189941406, "n_util_obs": 423, "util_count_spearman": 0.19789426670340893}, "probe_overhead_sec_total": 104.623, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.017409, "alpha": 0.03, "t": 702.6} |
| {"event": "consolidate", "step": 2000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 729.5} |
| {"event": "probe", "step": 2000, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3532.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.202954, "cps": 27.59375, "mean_p": 0.215194, "boundary_entropy": 0.134823, "bpb": 2.119923}, "tr": {"chunks": 3751.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.207203, "cps": 29.304688, "mean_p": 0.215619, "boundary_entropy": 0.172817, "bpb": 2.34534}, "fi": {"chunks": 4886.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.259742, "cps": 38.171875, "mean_p": 0.266682, "boundary_entropy": 0.14764, "bpb": 2.477183}}, "bpb": {"en": 2.11992, "tr": 2.34534, "fi": 2.47718}, "gini_cpc": 0.056514, "gini_cps": 0.074178, "bpic": 4.66201, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0209393379985137, "fi": 1.2798085419406842}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 116.9539794921875, "n_util_obs": 435, "util_count_spearman": 0.2025668039983656}, "t": 730.4} |
| {"event": "eval", "step": 2000, "phase": 1, "bpb": {"en": 2.1298, "tr": 2.35664, "fi": 2.48705}, "bytes_per_chunk": {"en": 4.693, "tr": 5.435, "fi": 4.137}, "bpb_mean": 2.3245, "frozen_inventory": true, "t": 731.8} |
| {"event": "train", "step": 2000, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.53928, "bpb": 2.22072, "chunk_loss": 0.030803, "F": 0.22055, "G": 0.2303, "chunks": 225.84, "bytes_per_chunk": 4.534, "boundary_entropy": 0.13723, "router_noise": 0.0, "grad_norm": 0.8124, "lr": 0.000245, "router_trainable": true, "sec_per_step": 0.63, "probe_sec": 0.1226, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10661, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.05825, "n_shadow": 0, "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": 0.19665, "trunc_abs_frac": 0.43691, "mean_delta_full": 0.22156, "read_hit_rate": {"skel": 0.75937}, "mean_gate": {"skel": 0.23615}, "n_reads": 4071, "n_cand": 5361, "read_suppressed_frac": 0.1032, "dropout_reads_logged": 4071, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 117.01667785644531, "n_util_obs": 435, "util_count_spearman": 0.20286476111447369}, "probe_overhead_sec_total": 107.573, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.02038, "alpha": 0.03, "t": 734.1} |
| {"event": "probe", "step": 2048, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3500.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.201115, "cps": 27.34375, "mean_p": 0.214371, "boundary_entropy": 0.140526, "bpb": 2.106149}, "tr": {"chunks": 3747.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.206982, "cps": 29.273438, "mean_p": 0.214667, "boundary_entropy": 0.168079, "bpb": 2.330951}, "fi": {"chunks": 4935.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.262346, "cps": 38.554688, "mean_p": 0.269687, "boundary_entropy": 0.148047, "bpb": 2.457169}}, "bpb": {"en": 2.10615, "tr": 2.33095, "fi": 2.45717}, "gini_cpc": 0.060887, "gini_cps": 0.078531, "bpic": 4.657035, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.029174974944958, "fi": 1.304461751103078}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 119.86799621582031, "n_util_obs": 449, "util_count_spearman": 0.21142239670761076}, "t": 756.3} |
| {"event": "consolidate", "step": 2050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 760.0} |
| {"event": "train", "step": 2050, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.55242, "bpb": 2.23967, "chunk_loss": 0.030447, "F": 0.19977, "G": 0.21744, "chunks": 204.56, "bytes_per_chunk": 5.006, "boundary_entropy": 0.14479, "router_noise": 0.0, "grad_norm": 0.805, "lr": 0.0002419, "router_trainable": true, "sec_per_step": 0.527, "probe_sec": 0.0311, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02083, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.21074, "n_shadow": 0, "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": 1.11469, "trunc_abs_frac": 3.58704, "mean_delta_full": 0.23553, "read_hit_rate": {"skel": 0.72462}, "mean_gate": {"skel": 0.24743}, "n_reads": 3697, "n_cand": 5102, "read_suppressed_frac": 0.0974, "dropout_reads_logged": 3697, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 120.04356384277344, "n_util_obs": 450, "util_count_spearman": 0.210679569363328}, "probe_overhead_sec_total": 110.323, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.019154, "alpha": 0.03, "t": 760.4} |
| {"event": "consolidate", "step": 2100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 786.8} |
| {"event": "train", "step": 2100, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.47563, "bpb": 2.12889, "chunk_loss": 0.030528, "F": 0.20728, "G": 0.21778, "chunks": 212.25, "bytes_per_chunk": 4.824, "boundary_entropy": 0.13416, "router_noise": 0.0, "grad_norm": 0.7086, "lr": 0.0002388, "router_trainable": true, "sec_per_step": 0.534, "probe_sec": 0.121, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.11221, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.21987, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 6, "trunc_ratio_sum": 2.02985, "trunc_abs_frac": 1.0369, "mean_delta_full": 0.0627, "read_hit_rate": {"skel": 0.74685}, "mean_gate": {"skel": 0.24404}, "n_reads": 3909, "n_cand": 5234, "read_suppressed_frac": 0.0985, "dropout_reads_logged": 3909, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 123.08473205566406, "n_util_obs": 462, "util_count_spearman": 0.22304243802999654}, "probe_overhead_sec_total": 113.164, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.016818, "alpha": 0.03, "t": 787.1} |
| {"event": "consolidate", "step": 2150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 799.5} |
| {"event": "train", "step": 2150, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.48447, "bpb": 2.14163, "chunk_loss": 0.030751, "F": 0.21649, "G": 0.22654, "chunks": 221.69, "bytes_per_chunk": 4.619, "boundary_entropy": 0.13389, "router_noise": 0.0, "grad_norm": 0.7085, "lr": 0.0002357, "router_trainable": true, "sec_per_step": 0.251, "probe_sec": 0.0267, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0193, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.11265, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": -8.01, "trunc_abs_frac": 2.13126, "mean_delta_full": -0.00763, "read_hit_rate": {"skel": 0.76111}, "mean_gate": {"skel": 0.22671}, "n_reads": 4094, "n_cand": 5379, "read_suppressed_frac": 0.1055, "dropout_reads_logged": 4094, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 126.1234130859375, "n_util_obs": 474, "util_count_spearman": 0.20533752339105557}, "probe_overhead_sec_total": 116.565, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.019254, "alpha": 0.03, "t": 799.7} |
| {"event": "consolidate", "step": 2200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 811.5} |
| {"event": "probe", "step": 2200, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3491.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.200598, "cps": 27.273438, "mean_p": 0.211427, "boundary_entropy": 0.126398, "bpb": 2.090126}, "tr": {"chunks": 3717.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.205325, "cps": 29.039062, "mean_p": 0.210134, "boundary_entropy": 0.159047, "bpb": 2.303346}, "fi": {"chunks": 4873.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.259051, "cps": 38.070312, "mean_p": 0.263791, "boundary_entropy": 0.134338, "bpb": 2.449276}}, "bpb": {"en": 2.09013, "tr": 2.30335, "fi": 2.44928}, "gini_cpc": 0.058602, "gini_cps": 0.076263, "bpic": 4.695969, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0235670126129686, "fi": 1.291394104216374}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 129.0868377685547, "n_util_obs": 483, "util_count_spearman": 0.19637847881041398}, "t": 812.3} |
| {"event": "eval", "step": 2200, "phase": 1, "bpb": {"en": 2.10156, "tr": 2.31197, "fi": 2.46236}, "bytes_per_chunk": {"en": 4.738, "tr": 5.461, "fi": 4.137}, "bpb_mean": 2.29196, "frozen_inventory": true, "t": 813.6} |
| {"event": "train", "step": 2200, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.5273, "bpb": 2.20343, "chunk_loss": 0.030755, "F": 0.21381, "G": 0.22473, "chunks": 218.94, "bytes_per_chunk": 4.677, "boundary_entropy": 0.12868, "router_noise": 0.0, "grad_norm": 0.7986, "lr": 0.0002325, "router_trainable": true, "sec_per_step": 0.311, "probe_sec": 0.062, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05221, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.17829, "n_shadow": 0, "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": 0.59201, "trunc_abs_frac": 0.43537, "mean_delta_full": 0.4693, "read_hit_rate": {"skel": 0.7571}, "mean_gate": {"skel": 0.2274}, "n_reads": 4077, "n_cand": 5385, "read_suppressed_frac": 0.1018, "dropout_reads_logged": 4077, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 129.1497039794922, "n_util_obs": 484, "util_count_spearman": 0.19124996434537764}, "probe_overhead_sec_total": 118.734, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.019928, "alpha": 0.03, "t": 815.2} |
| {"event": "consolidate", "step": 2250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 826.8} |
| {"event": "train", "step": 2250, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.49681, "bpb": 2.15944, "chunk_loss": 0.030562, "F": 0.20947, "G": 0.2183, "chunks": 214.5, "bytes_per_chunk": 4.774, "boundary_entropy": 0.12383, "router_noise": 0.0, "grad_norm": 0.7206, "lr": 0.0002293, "router_trainable": true, "sec_per_step": 0.237, "probe_sec": 0.0859, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07472, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.06381, "n_shadow": 0, "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": 0.73597, "trunc_abs_frac": 0.73597, "mean_delta_full": -0.56335, "read_hit_rate": {"skel": 0.75213}, "mean_gate": {"skel": 0.23004}, "n_reads": 3893, "n_cand": 5176, "read_suppressed_frac": 0.0986, "dropout_reads_logged": 3893, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 132.14739990234375, "n_util_obs": 493, "util_count_spearman": 0.197358370341168}, "probe_overhead_sec_total": 121.41, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.018505, "alpha": 0.03, "t": 827.1} |
| {"event": "consolidate", "step": 2300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 838.9} |
| {"event": "train", "step": 2300, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.54128, "bpb": 2.2236, "chunk_loss": 0.030623, "F": 0.21133, "G": 0.21978, "chunks": 216.41, "bytes_per_chunk": 4.732, "boundary_entropy": 0.1271, "router_noise": 0.0, "grad_norm": 0.6908, "lr": 0.000226, "router_trainable": true, "sec_per_step": 0.242, "probe_sec": 0.0769, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0676, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.01744, "n_shadow": 0, "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": 0.65314, "trunc_abs_frac": 0.65314, "mean_delta_full": -0.65881, "read_hit_rate": {"skel": 0.73922}, "mean_gate": {"skel": 0.2355}, "n_reads": 3790, "n_cand": 5127, "read_suppressed_frac": 0.0997, "dropout_reads_logged": 3790, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 135.1708984375, "n_util_obs": 502, "util_count_spearman": 0.19794850619747442}, "probe_overhead_sec_total": 124.0, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.018804, "alpha": 0.03, "t": 839.2} |
| {"event": "consolidate", "step": 2350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 851.1} |
| {"event": "train", "step": 2350, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.46129, "bpb": 2.1082, "chunk_loss": 0.030612, "F": 0.21115, "G": 0.22079, "chunks": 216.22, "bytes_per_chunk": 4.736, "boundary_entropy": 0.12583, "router_noise": 0.0, "grad_norm": 0.7571, "lr": 0.0002226, "router_trainable": true, "sec_per_step": 0.242, "probe_sec": 0.0343, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02685, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.07099, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 7, "trunc_ratio_sum": 0.62603, "trunc_abs_frac": 1.82005, "mean_delta_full": -0.32547, "read_hit_rate": {"skel": 0.7645}, "mean_gate": {"skel": 0.22261}, "n_reads": 4100, "n_cand": 5363, "read_suppressed_frac": 0.1017, "dropout_reads_logged": 4100, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 138.17822265625, "n_util_obs": 518, "util_count_spearman": 0.20395086035816798}, "probe_overhead_sec_total": 126.622, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.018286, "alpha": 0.03, "t": 851.3} |
| {"event": "consolidate", "step": 2400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 863.7} |
| {"event": "probe", "step": 2400, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3499.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.201057, "cps": 27.335938, "mean_p": 0.210467, "boundary_entropy": 0.11978, "bpb": 2.067939}, "tr": {"chunks": 3704.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.204607, "cps": 28.9375, "mean_p": 0.210354, "boundary_entropy": 0.157308, "bpb": 2.269566}, "fi": {"chunks": 4853.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.257987, "cps": 37.914062, "mean_p": 0.262387, "boundary_entropy": 0.133199, "bpb": 2.392966}}, "bpb": {"en": 2.06794, "tr": 2.26957, "fi": 2.39297}, "gini_cpc": 0.057189, "gini_cps": 0.074873, "bpic": 4.705707, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0176550786355623, "fi": 1.283153419336272}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 141.10765075683594, "n_util_obs": 532, "util_count_spearman": 0.19356143316241411}, "t": 864.6} |
| {"event": "eval", "step": 2400, "phase": 1, "bpb": {"en": 2.08036, "tr": 2.27575, "fi": 2.40437}, "bytes_per_chunk": {"en": 4.761, "tr": 5.434, "fi": 4.165}, "bpb_mean": 2.25349, "frozen_inventory": true, "t": 865.9} |
| {"event": "phase_transition", "step": 2400, "phase": 2, "objective": "global", "router_trainable": true, "encoder_trainable": true, "frozen_router_params": 0, "t": 866.8} |
| {"event": "train", "step": 2400, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.4588, "bpb": 2.10461, "chunk_loss": 0.030517, "F": 0.20667, "G": 0.2179, "chunks": 211.62, "bytes_per_chunk": 4.839, "boundary_entropy": 0.12519, "router_noise": 0.0, "grad_norm": 0.6725, "lr": 0.0002192, "router_trainable": true, "sec_per_step": 0.315, "probe_sec": 0.03, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02373, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.12315, "n_shadow": 0, "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": 0.36624, "trunc_abs_frac": 0.31239, "mean_delta_full": 0.28485, "read_hit_rate": {"skel": 0.74909}, "mean_gate": {"skel": 0.23238}, "n_reads": 3917, "n_cand": 5229, "read_suppressed_frac": 0.0965, "dropout_reads_logged": 3917, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 141.1669158935547, "n_util_obs": 532, "util_count_spearman": 0.1928804551273056}, "probe_overhead_sec_total": 128.906, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.020985, "alpha": 0.03, "t": 867.0} |
| {"event": "consolidate", "step": 2450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 879.8} |
| {"event": "train", "step": 2450, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.47386, "bpb": 2.12633, "chunk_loss": 0.0308, "F": 0.21924, "G": 0.22938, "chunks": 224.5, "bytes_per_chunk": 4.561, "boundary_entropy": 0.12418, "router_noise": 0.0, "grad_norm": 0.7752, "lr": 0.0002158, "router_trainable": true, "sec_per_step": 0.261, "probe_sec": 0.0526, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04512, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.13284, "n_shadow": 0, "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": 1.28418, "trunc_abs_frac": 5.40404, "mean_delta_full": 0.21606, "read_hit_rate": {"skel": 0.77193}, "mean_gate": {"skel": 0.22899}, "n_reads": 4197, "n_cand": 5437, "read_suppressed_frac": 0.0972, "dropout_reads_logged": 4197, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 144.19520568847656, "n_util_obs": 541, "util_count_spearman": 0.1968916332734726}, "probe_overhead_sec_total": 131.735, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.019099, "alpha": 0.03, "t": 880.1} |
| {"event": "consolidate", "step": 2500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 892.2} |
| {"event": "train", "step": 2500, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.56957, "bpb": 2.26441, "chunk_loss": 0.030695, "F": 0.21625, "G": 0.2237, "chunks": 221.44, "bytes_per_chunk": 4.624, "boundary_entropy": 0.12383, "router_noise": 0.0, "grad_norm": 0.7151, "lr": 0.0002123, "router_trainable": true, "sec_per_step": 0.246, "probe_sec": 0.0292, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02284, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.20375, "n_shadow": 0, "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": -0.89476, "trunc_abs_frac": 0.89476, "mean_delta_full": -0.25024, "read_hit_rate": {"skel": 0.73792}, "mean_gate": {"skel": 0.2253}, "n_reads": 3832, "n_cand": 5193, "read_suppressed_frac": 0.0953, "dropout_reads_logged": 3832, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 147.22515869140625, "n_util_obs": 552, "util_count_spearman": 0.20371859340760085}, "probe_overhead_sec_total": 134.13, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.018944, "alpha": 0.03, "t": 892.4} |
| {"event": "consolidate", "step": 2550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 904.9} |
| {"event": "train", "step": 2550, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.49783, "bpb": 2.16092, "chunk_loss": 0.030535, "F": 0.20981, "G": 0.21927, "chunks": 214.84, "bytes_per_chunk": 4.766, "boundary_entropy": 0.12288, "router_noise": 0.0, "grad_norm": 0.7291, "lr": 0.0002088, "router_trainable": true, "sec_per_step": 0.255, "probe_sec": 0.0308, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02545, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.28482, "n_shadow": 0, "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": 0.74501, "trunc_abs_frac": 0.53837, "mean_delta_full": 0.17686, "read_hit_rate": {"skel": 0.74977}, "mean_gate": {"skel": 0.23232}, "n_reads": 3994, "n_cand": 5327, "read_suppressed_frac": 0.0974, "dropout_reads_logged": 3994, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 150.26214599609375, "n_util_obs": 562, "util_count_spearman": 0.19684286607748713}, "probe_overhead_sec_total": 136.7, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.019926, "alpha": 0.03, "t": 905.1} |
| {"event": "consolidate", "step": 2600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 917.8} |
| {"event": "probe", "step": 2600, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3525.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.202551, "cps": 27.539062, "mean_p": 0.208943, "boundary_entropy": 0.116299, "bpb": 2.050943}, "tr": {"chunks": 3737.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.20643, "cps": 29.195312, "mean_p": 0.208573, "boundary_entropy": 0.146407, "bpb": 2.251696}, "fi": {"chunks": 4859.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.258306, "cps": 37.960938, "mean_p": 0.261766, "boundary_entropy": 0.121783, "bpb": 2.357143}}, "bpb": {"en": 2.05094, "tr": 2.2517, "fi": 2.35714}, "gini_cpc": 0.055703, "gini_cps": 0.073371, "bpic": 4.680472, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0191486776025132, "fi": 1.275263749028692}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 153.23709106445312, "n_util_obs": 571, "util_count_spearman": 0.1809317251947244}, "t": 918.6} |
| {"event": "eval", "step": 2600, "phase": 2, "bpb": {"en": 2.06046, "tr": 2.25965, "fi": 2.3679}, "bytes_per_chunk": {"en": 4.731, "tr": 5.381, "fi": 4.142}, "bpb_mean": 2.22934, "frozen_inventory": true, "t": 919.9} |
| {"event": "train", "step": 2600, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.5047, "bpb": 2.17082, "chunk_loss": 0.030746, "F": 0.21643, "G": 0.22546, "chunks": 221.62, "bytes_per_chunk": 4.62, "boundary_entropy": 0.11597, "router_noise": 0.0, "grad_norm": 0.6941, "lr": 0.0002053, "router_trainable": true, "sec_per_step": 0.319, "probe_sec": 0.0451, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03557, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.05953, "n_shadow": 0, "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": -0.86686, "trunc_abs_frac": 1.74524, "mean_delta_full": -0.02155, "read_hit_rate": {"skel": 0.76308}, "mean_gate": {"skel": 0.22017}, "n_reads": 4171, "n_cand": 5466, "read_suppressed_frac": 0.1002, "dropout_reads_logged": 4171, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 153.3005828857422, "n_util_obs": 571, "util_count_spearman": 0.18096340029252692}, "probe_overhead_sec_total": 139.818, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.019711, "alpha": 0.03, "t": 921.1} |
| {"event": "consolidate", "step": 2650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 934.8} |
| {"event": "train", "step": 2650, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.43152, "bpb": 2.06525, "chunk_loss": 0.030529, "F": 0.2095, "G": 0.21563, "chunks": 214.53, "bytes_per_chunk": 4.773, "boundary_entropy": 0.11979, "router_noise": 0.0, "grad_norm": 0.6476, "lr": 0.0002017, "router_trainable": true, "sec_per_step": 0.278, "probe_sec": 0.0447, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03764, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.08553, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 1.0, "trunc_abs_frac": 1.0, "mean_delta_full": -0.02576, "read_hit_rate": {"skel": 0.75124}, "mean_gate": {"skel": 0.21299}, "n_reads": 3935, "n_cand": 5238, "read_suppressed_frac": 0.0925, "dropout_reads_logged": 3935, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 156.3041534423828, "n_util_obs": 580, "util_count_spearman": 0.17196775447770624}, "probe_overhead_sec_total": 143.055, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.019715, "alpha": 0.03, "t": 935.0} |
| {"event": "consolidate", "step": 2700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 949.2} |
| {"event": "train", "step": 2700, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.51883, "bpb": 2.19121, "chunk_loss": 0.030596, "F": 0.20978, "G": 0.21867, "chunks": 214.81, "bytes_per_chunk": 4.767, "boundary_entropy": 0.12322, "router_noise": 0.0, "grad_norm": 0.7707, "lr": 0.0001981, "router_trainable": true, "sec_per_step": 0.29, "probe_sec": 0.1041, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08793, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.14402, "n_shadow": 0, "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": 0.17557, "trunc_abs_frac": 0.52502, "mean_delta_full": 0.25517, "read_hit_rate": {"skel": 0.73928}, "mean_gate": {"skel": 0.22494}, "n_reads": 3913, "n_cand": 5293, "read_suppressed_frac": 0.0966, "dropout_reads_logged": 3913, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 159.3068084716797, "n_util_obs": 595, "util_count_spearman": 0.17679611564204123}, "probe_overhead_sec_total": 146.059, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.020169, "alpha": 0.03, "t": 949.4} |
| {"event": "consolidate", "step": 2750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 964.2} |
| {"event": "train", "step": 2750, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.45753, "bpb": 2.10277, "chunk_loss": 0.030558, "F": 0.21051, "G": 0.21908, "chunks": 215.56, "bytes_per_chunk": 4.75, "boundary_entropy": 0.11873, "router_noise": 0.0, "grad_norm": 0.7267, "lr": 0.0001945, "router_trainable": true, "sec_per_step": 0.299, "probe_sec": 0.028, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02276, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.12118, "n_shadow": 0, "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": 0.50122, "trunc_abs_frac": 0.97796, "mean_delta_full": 0.38407, "read_hit_rate": {"skel": 0.75597}, "mean_gate": {"skel": 0.22044}, "n_reads": 3987, "n_cand": 5274, "read_suppressed_frac": 0.1026, "dropout_reads_logged": 3987, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 162.35198974609375, "n_util_obs": 606, "util_count_spearman": 0.17996636355402498}, "probe_overhead_sec_total": 148.649, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.018661, "alpha": 0.03, "t": 964.4} |
| {"event": "consolidate", "step": 2800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 986.5} |
| {"event": "probe", "step": 2800, "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": 0.207306, "boundary_entropy": 0.113578, "bpb": 2.027111}, "tr": {"chunks": 3743.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.206761, "cps": 29.242188, "mean_p": 0.205055, "boundary_entropy": 0.139214, "bpb": 2.231436}, "fi": {"chunks": 4849.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.257775, "cps": 37.882812, "mean_p": 0.260135, "boundary_entropy": 0.1187, "bpb": 2.329845}}, "bpb": {"en": 2.02711, "tr": 2.23144, "fi": 2.32985}, "gini_cpc": 0.054878, "gini_cps": 0.07254, "bpic": 4.680086, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0193391169416304, "fi": 1.2708366044037582}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 165.3271942138672, "n_util_obs": 619, "util_count_spearman": 0.18400676884855005}, "t": 987.4} |
| {"event": "eval", "step": 2800, "phase": 2, "bpb": {"en": 2.03931, "tr": 2.23788, "fi": 2.34127}, "bytes_per_chunk": {"en": 4.738, "tr": 5.398, "fi": 4.161}, "bpb_mean": 2.20615, "frozen_inventory": true, "t": 988.7} |
| {"event": "train", "step": 2800, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.45053, "bpb": 2.09267, "chunk_loss": 0.030601, "F": 0.21411, "G": 0.2192, "chunks": 219.25, "bytes_per_chunk": 4.67, "boundary_entropy": 0.11297, "router_noise": 0.0, "grad_norm": 0.6847, "lr": 0.0001909, "router_trainable": true, "sec_per_step": 0.526, "probe_sec": 0.0424, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03665, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.16438, "n_shadow": 0, "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": 0.49852, "trunc_abs_frac": 0.33292, "mean_delta_full": -0.24992, "read_hit_rate": {"skel": 0.76098}, "mean_gate": {"skel": 0.21341}, "n_reads": 3986, "n_cand": 5238, "read_suppressed_frac": 0.0971, "dropout_reads_logged": 3986, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 165.38851928710938, "n_util_obs": 619, "util_count_spearman": 0.1839022019234932}, "probe_overhead_sec_total": 151.578, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.018031, "alpha": 0.03, "t": 990.7} |
| {"event": "consolidate", "step": 2850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1015.4} |
| {"event": "train", "step": 2850, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.54749, "bpb": 2.23256, "chunk_loss": 0.030773, "F": 0.21597, "G": 0.22427, "chunks": 221.16, "bytes_per_chunk": 4.63, "boundary_entropy": 0.1138, "router_noise": 0.0, "grad_norm": 0.688, "lr": 0.0001872, "router_trainable": true, "sec_per_step": 0.513, "probe_sec": 0.0706, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05721, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.05118, "n_shadow": 0, "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": 1.73922, "trunc_abs_frac": 1.39977, "mean_delta_full": 0.08496, "read_hit_rate": {"skel": 0.74686}, "mean_gate": {"skel": 0.21305}, "n_reads": 4045, "n_cand": 5416, "read_suppressed_frac": 0.109, "dropout_reads_logged": 4045, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 168.410888671875, "n_util_obs": 626, "util_count_spearman": 0.16711533099252368}, "probe_overhead_sec_total": 154.443, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.018322, "alpha": 0.03, "t": 1016.3} |
| {"event": "consolidate", "step": 2900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1040.9} |
| {"event": "train", "step": 2900, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.49305, "bpb": 2.15401, "chunk_loss": 0.030828, "F": 0.22116, "G": 0.2274, "chunks": 226.47, "bytes_per_chunk": 4.522, "boundary_entropy": 0.11472, "router_noise": 0.0, "grad_norm": 0.67, "lr": 0.0001835, "router_trainable": true, "sec_per_step": 0.499, "probe_sec": 0.0427, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03654, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.12937, "n_shadow": 0, "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": -0.91166, "trunc_abs_frac": 1.07094, "mean_delta_full": -0.08939, "read_hit_rate": {"skel": 0.76181}, "mean_gate": {"skel": 0.22391}, "n_reads": 4097, "n_cand": 5378, "read_suppressed_frac": 0.0993, "dropout_reads_logged": 4097, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 171.44073486328125, "n_util_obs": 639, "util_count_spearman": 0.17141478507553923}, "probe_overhead_sec_total": 157.202, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.020336, "alpha": 0.03, "t": 1041.2} |
| {"event": "consolidate", "step": 2950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1066.4} |
| {"event": "train", "step": 2950, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.43354, "bpb": 2.06816, "chunk_loss": 0.030512, "F": 0.20624, "G": 0.21462, "chunks": 211.19, "bytes_per_chunk": 4.849, "boundary_entropy": 0.11545, "router_noise": 0.0, "grad_norm": 0.682, "lr": 0.0001799, "router_trainable": true, "sec_per_step": 0.515, "probe_sec": 0.0611, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04627, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.0391, "n_shadow": 0, "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": 0.2125, "trunc_abs_frac": 0.27373, "mean_delta_full": 0.31335, "read_hit_rate": {"skel": 0.74567}, "mean_gate": {"skel": 0.22229}, "n_reads": 3873, "n_cand": 5194, "read_suppressed_frac": 0.0891, "dropout_reads_logged": 3873, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 174.46070861816406, "n_util_obs": 649, "util_count_spearman": 0.17913757435463645}, "probe_overhead_sec_total": 159.888, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.018917, "alpha": 0.03, "t": 1067.0} |
| {"event": "consolidate", "step": 3000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1087.4} |
| {"event": "probe", "step": 3000, "phase": 2, "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": 0.208289, "boundary_entropy": 0.10712, "bpb": 2.009458}, "tr": {"chunks": 3732.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.206154, "cps": 29.15625, "mean_p": 0.203145, "boundary_entropy": 0.137867, "bpb": 2.197162}, "fi": {"chunks": 4881.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.259476, "cps": 38.132812, "mean_p": 0.261174, "boundary_entropy": 0.117139, "bpb": 2.304884}}, "bpb": {"en": 2.00946, "tr": 2.19716, "fi": 2.30488}, "gini_cpc": 0.058042, "gini_cps": 0.0757, "bpic": 4.681631, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0235927037573789, "fi": 1.288347509056793}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 177.4366912841797, "n_util_obs": 658, "util_count_spearman": 0.17899580015158076}, "t": 1088.4} |
| {"event": "eval", "step": 3000, "phase": 2, "bpb": {"en": 2.022, "tr": 2.20459, "fi": 2.31444}, "bytes_per_chunk": {"en": 4.743, "tr": 5.384, "fi": 4.162}, "bpb_mean": 2.18034, "frozen_inventory": true, "t": 1089.7} |
| {"event": "train", "step": 3000, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.45482, "bpb": 2.09886, "chunk_loss": 0.030546, "F": 0.21072, "G": 0.21731, "chunks": 215.78, "bytes_per_chunk": 4.746, "boundary_entropy": 0.10878, "router_noise": 0.0, "grad_norm": 0.7003, "lr": 0.0001761, "router_trainable": true, "sec_per_step": 0.484, "probe_sec": 0.0355, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02559, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.01187, "n_shadow": 0, "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": 0.8762, "trunc_abs_frac": 1.68281, "mean_delta_full": -0.2498, "read_hit_rate": {"skel": 0.75132}, "mean_gate": {"skel": 0.21718}, "n_reads": 3997, "n_cand": 5320, "read_suppressed_frac": 0.1023, "dropout_reads_logged": 3997, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 177.49778747558594, "n_util_obs": 658, "util_count_spearman": 0.17899216638615095}, "probe_overhead_sec_total": 162.828, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.016963, "alpha": 0.03, "t": 1091.1} |
| {"event": "consolidate", "step": 3050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1103.4} |
| {"event": "train", "step": 3050, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.44594, "bpb": 2.08605, "chunk_loss": 0.030639, "F": 0.21536, "G": 0.22165, "chunks": 220.53, "bytes_per_chunk": 4.643, "boundary_entropy": 0.10807, "router_noise": 0.0, "grad_norm": 0.6903, "lr": 0.0001724, "router_trainable": true, "sec_per_step": 0.249, "probe_sec": 0.0311, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02339, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.22408, "n_shadow": 0, "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": -0.14628, "trunc_abs_frac": 1.80626, "mean_delta_full": 0.05862, "read_hit_rate": {"skel": 0.75596}, "mean_gate": {"skel": 0.22416}, "n_reads": 4058, "n_cand": 5368, "read_suppressed_frac": 0.0998, "dropout_reads_logged": 4058, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 180.52603149414062, "n_util_obs": 668, "util_count_spearman": 0.1929898823417713}, "probe_overhead_sec_total": 165.703, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.019559, "alpha": 0.03, "t": 1103.6} |
| {"event": "consolidate", "step": 3100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1117.2} |
| {"event": "train", "step": 3100, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.54174, "bpb": 2.22427, "chunk_loss": 0.0306, "F": 0.20862, "G": 0.21692, "chunks": 213.62, "bytes_per_chunk": 4.793, "boundary_entropy": 0.11513, "router_noise": 0.0, "grad_norm": 0.7205, "lr": 0.0001687, "router_trainable": true, "sec_per_step": 0.277, "probe_sec": 0.0426, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03611, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.08689, "n_shadow": 0, "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": 14.18907, "trunc_abs_frac": 1.12266, "mean_delta_full": 0.02139, "read_hit_rate": {"skel": 0.73107}, "mean_gate": {"skel": 0.2239}, "n_reads": 3795, "n_cand": 5191, "read_suppressed_frac": 0.0893, "dropout_reads_logged": 3795, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 183.54493713378906, "n_util_obs": 677, "util_count_spearman": 0.19096131488265802}, "probe_overhead_sec_total": 168.436, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.018742, "alpha": 0.03, "t": 1117.4} |
| {"event": "consolidate", "step": 3150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1131.3} |
| {"event": "train", "step": 3150, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.46137, "bpb": 2.10832, "chunk_loss": 0.030583, "F": 0.21188, "G": 0.21771, "chunks": 216.97, "bytes_per_chunk": 4.72, "boundary_entropy": 0.10835, "router_noise": 0.0, "grad_norm": 0.6715, "lr": 0.000165, "router_trainable": true, "sec_per_step": 0.282, "probe_sec": 0.0456, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03768, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.15245, "n_shadow": 0, "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": 0.25611, "trunc_abs_frac": 0.7819, "mean_delta_full": 0.2348, "read_hit_rate": {"skel": 0.7441}, "mean_gate": {"skel": 0.21978}, "n_reads": 3943, "n_cand": 5299, "read_suppressed_frac": 0.0992, "dropout_reads_logged": 3943, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 186.59603881835938, "n_util_obs": 682, "util_count_spearman": 0.18471884708022682}, "probe_overhead_sec_total": 171.209, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.019774, "alpha": 0.03, "t": 1131.5} |
| {"event": "consolidate", "step": 3200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1143.6} |
| {"event": "probe", "step": 3200, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3511.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.201747, "cps": 27.429688, "mean_p": 0.207091, "boundary_entropy": 0.1042, "bpb": 1.989667}, "tr": {"chunks": 3736.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.206375, "cps": 29.1875, "mean_p": 0.199691, "boundary_entropy": 0.136771, "bpb": 2.174459}, "fi": {"chunks": 4851.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.257881, "cps": 37.898438, "mean_p": 0.258952, "boundary_entropy": 0.110928, "bpb": 2.27533}}, "bpb": {"en": 1.98967, "tr": 2.17446, "fi": 2.27533}, "gini_cpc": 0.05619, "gini_cps": 0.073841, "bpic": 4.68937, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0229386944383396, "fi": 1.2782408185421363}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 189.5762176513672, "n_util_obs": 693, "util_count_spearman": 0.18726236793662165}, "t": 1144.5} |
| {"event": "eval", "step": 3200, "phase": 2, "bpb": {"en": 2.00213, "tr": 2.18467, "fi": 2.28573}, "bytes_per_chunk": {"en": 4.737, "tr": 5.354, "fi": 4.16}, "bpb_mean": 2.15751, "frozen_inventory": true, "t": 1145.9} |
| {"event": "train", "step": 3200, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.42414, "bpb": 2.0546, "chunk_loss": 0.030726, "F": 0.21753, "G": 0.22248, "chunks": 222.75, "bytes_per_chunk": 4.597, "boundary_entropy": 0.1085, "router_noise": 0.0, "grad_norm": 0.6709, "lr": 0.0001613, "router_trainable": true, "sec_per_step": 0.318, "probe_sec": 0.075, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06961, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.15162, "n_shadow": 0, "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": 0.79695, "trunc_abs_frac": 0.60582, "mean_delta_full": 0.44888, "read_hit_rate": {"skel": 0.76693}, "mean_gate": {"skel": 0.21431}, "n_reads": 4156, "n_cand": 5419, "read_suppressed_frac": 0.0994, "dropout_reads_logged": 4156, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 189.63973999023438, "n_util_obs": 693, "util_count_spearman": 0.18457840044003837}, "probe_overhead_sec_total": 173.612, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.021618, "alpha": 0.03, "t": 1147.4} |
| {"event": "consolidate", "step": 3250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1159.3} |
| {"event": "train", "step": 3250, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.40663, "bpb": 2.02934, "chunk_loss": 0.03043, "F": 0.20599, "G": 0.21005, "chunks": 210.94, "bytes_per_chunk": 4.855, "boundary_entropy": 0.10589, "router_noise": 0.0, "grad_norm": 0.673, "lr": 0.0001576, "router_trainable": true, "sec_per_step": 0.245, "probe_sec": 0.0511, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03713, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.06026, "n_shadow": 0, "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": 0.56626, "trunc_abs_frac": 0.54733, "mean_delta_full": 0.08858, "read_hit_rate": {"skel": 0.74383}, "mean_gate": {"skel": 0.21342}, "n_reads": 3859, "n_cand": 5188, "read_suppressed_frac": 0.1003, "dropout_reads_logged": 3859, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 192.6752166748047, "n_util_obs": 702, "util_count_spearman": 0.18493595930052695}, "probe_overhead_sec_total": 176.172, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.019005, "alpha": 0.03, "t": 1159.7} |
| {"event": "consolidate", "step": 3300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1172.2} |
| {"event": "train", "step": 3300, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.44826, "bpb": 2.0894, "chunk_loss": 0.030606, "F": 0.21289, "G": 0.2173, "chunks": 218.0, "bytes_per_chunk": 4.697, "boundary_entropy": 0.10998, "router_noise": 0.0, "grad_norm": 0.6518, "lr": 0.0001539, "router_trainable": true, "sec_per_step": 0.253, "probe_sec": 0.0275, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02131, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.21988, "n_shadow": 0, "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": 0.55054, "trunc_abs_frac": 18.43089, "mean_delta_full": 0.15157, "read_hit_rate": {"skel": 0.75055}, "mean_gate": {"skel": 0.21058}, "n_reads": 4059, "n_cand": 5408, "read_suppressed_frac": 0.099, "dropout_reads_logged": 4059, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 195.6963348388672, "n_util_obs": 711, "util_count_spearman": 0.17969543625129591}, "probe_overhead_sec_total": 178.957, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.018236, "alpha": 0.03, "t": 1172.4} |
| {"event": "consolidate", "step": 3350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1185.2} |
| {"event": "train", "step": 3350, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.48125, "bpb": 2.13699, "chunk_loss": 0.030777, "F": 0.2197, "G": 0.22316, "chunks": 224.97, "bytes_per_chunk": 4.552, "boundary_entropy": 0.10917, "router_noise": 0.0, "grad_norm": 0.7495, "lr": 0.0001501, "router_trainable": true, "sec_per_step": 0.261, "probe_sec": 0.0501, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0428, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.08958, "n_shadow": 0, "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": -0.49804, "trunc_abs_frac": 0.64219, "mean_delta_full": -0.23934, "read_hit_rate": {"skel": 0.75144}, "mean_gate": {"skel": 0.2179}, "n_reads": 4054, "n_cand": 5395, "read_suppressed_frac": 0.0997, "dropout_reads_logged": 4054, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 198.7206268310547, "n_util_obs": 721, "util_count_spearman": 0.1686379982193028}, "probe_overhead_sec_total": 181.374, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.019367, "alpha": 0.03, "t": 1185.4} |
| {"event": "consolidate", "step": 3400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1197.5} |
| {"event": "probe", "step": 3400, "phase": 2, "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": 0.208194, "boundary_entropy": 0.10068, "bpb": 1.975996}, "tr": {"chunks": 3759.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.207645, "cps": 29.367188, "mean_p": 0.200082, "boundary_entropy": 0.13341, "bpb": 2.146783}, "fi": {"chunks": 4858.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.258253, "cps": 37.953125, "mean_p": 0.258607, "boundary_entropy": 0.107677, "bpb": 2.257238}}, "bpb": {"en": 1.976, "tr": 2.14678, "fi": 2.25724}, "gini_cpc": 0.054995, "gini_cps": 0.072641, "bpic": 4.668916, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0225377478261302, "fi": 1.2717542629168228}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 201.70616149902344, "n_util_obs": 729, "util_count_spearman": 0.16210952031240083}, "t": 1198.3} |
| {"event": "eval", "step": 3400, "phase": 2, "bpb": {"en": 1.98811, "tr": 2.157, "fi": 2.26528}, "bytes_per_chunk": {"en": 4.729, "tr": 5.367, "fi": 4.152}, "bpb_mean": 2.1368, "frozen_inventory": true, "t": 1199.8} |
| {"event": "train", "step": 3400, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.39479, "bpb": 2.01225, "chunk_loss": 0.030526, "F": 0.20908, "G": 0.21327, "chunks": 214.09, "bytes_per_chunk": 4.783, "boundary_entropy": 0.10063, "router_noise": 0.0, "grad_norm": 0.7121, "lr": 0.0001465, "router_trainable": true, "sec_per_step": 0.317, "probe_sec": 0.1466, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.14068, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.04425, "n_shadow": 0, "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": 0.09684, "trunc_abs_frac": 0.09684, "mean_delta_full": -0.25085, "read_hit_rate": {"skel": 0.75467}, "mean_gate": {"skel": 0.209}, "n_reads": 3959, "n_cand": 5246, "read_suppressed_frac": 0.0998, "dropout_reads_logged": 3959, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 201.76719665527344, "n_util_obs": 729, "util_count_spearman": 0.16193247493087198}, "probe_overhead_sec_total": 184.902, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.019103, "alpha": 0.03, "t": 1201.3} |
| {"event": "consolidate", "step": 3450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1213.3} |
| {"event": "train", "step": 3450, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.39115, "bpb": 2.00701, "chunk_loss": 0.030452, "F": 0.20819, "G": 0.21158, "chunks": 213.19, "bytes_per_chunk": 4.803, "boundary_entropy": 0.09668, "router_noise": 0.0, "grad_norm": 0.6532, "lr": 0.0001428, "router_trainable": true, "sec_per_step": 0.245, "probe_sec": 0.034, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02791, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.22574, "n_shadow": 0, "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": 1.00579, "trunc_abs_frac": 6.61401, "mean_delta_full": 0.39899, "read_hit_rate": {"skel": 0.75281}, "mean_gate": {"skel": 0.20684}, "n_reads": 3953, "n_cand": 5251, "read_suppressed_frac": 0.0878, "dropout_reads_logged": 3953, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 204.81338500976562, "n_util_obs": 743, "util_count_spearman": 0.15084729617339468}, "probe_overhead_sec_total": 187.381, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.017049, "alpha": 0.03, "t": 1213.5} |
| {"event": "consolidate", "step": 3500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1229.7} |
| {"event": "train", "step": 3500, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.47183, "bpb": 2.1234, "chunk_loss": 0.030681, "F": 0.21286, "G": 0.21755, "chunks": 217.97, "bytes_per_chunk": 4.698, "boundary_entropy": 0.10434, "router_noise": 0.0, "grad_norm": 0.6947, "lr": 0.0001391, "router_trainable": true, "sec_per_step": 0.336, "probe_sec": 0.0388, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02594, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.04021, "n_shadow": 0, "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": -0.29203, "trunc_abs_frac": 3.42913, "mean_delta_full": 0.08423, "read_hit_rate": {"skel": 0.75704}, "mean_gate": {"skel": 0.20212}, "n_reads": 4085, "n_cand": 5396, "read_suppressed_frac": 0.1033, "dropout_reads_logged": 4085, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 207.8695526123047, "n_util_obs": 753, "util_count_spearman": 0.14310980167065016}, "probe_overhead_sec_total": 190.515, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.01959, "alpha": 0.03, "t": 1230.3} |
| {"event": "consolidate", "step": 3550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1248.0} |
| {"event": "train", "step": 3550, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.4767, "bpb": 2.13043, "chunk_loss": 0.030819, "F": 0.21997, "G": 0.22607, "chunks": 225.25, "bytes_per_chunk": 4.546, "boundary_entropy": 0.10835, "router_noise": 0.0, "grad_norm": 0.7216, "lr": 0.0001355, "router_trainable": true, "sec_per_step": 0.357, "probe_sec": 0.0475, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03647, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.20293, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 7, "trunc_ratio_sum": 0.70329, "trunc_abs_frac": 0.88283, "mean_delta_full": 0.42496, "read_hit_rate": {"skel": 0.76214}, "mean_gate": {"skel": 0.20116}, "n_reads": 4268, "n_cand": 5600, "read_suppressed_frac": 0.1057, "dropout_reads_logged": 4268, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 210.92242431640625, "n_util_obs": 764, "util_count_spearman": 0.15178597238652583}, "probe_overhead_sec_total": 193.157, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.020672, "alpha": 0.03, "t": 1248.2} |
| {"event": "consolidate", "step": 3600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1264.4} |
| {"event": "probe", "step": 3600, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3502.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.20123, "cps": 27.359375, "mean_p": 0.204893, "boundary_entropy": 0.100027, "bpb": 1.962795}, "tr": {"chunks": 3775.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.208529, "cps": 29.492188, "mean_p": 0.201147, "boundary_entropy": 0.129505, "bpb": 2.133783}, "fi": {"chunks": 4852.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.257934, "cps": 37.90625, "mean_p": 0.257383, "boundary_entropy": 0.105297, "bpb": 2.235351}}, "bpb": {"en": 1.9628, "tr": 2.13378, "fi": 2.23535}, "gini_cpc": 0.056617, "gini_cps": 0.074202, "bpic": 4.677385, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.036273477678793, "fi": 1.2817900240090028}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 213.8690185546875, "n_util_obs": 773, "util_count_spearman": 0.15439389577563606}, "t": 1265.3} |
| {"event": "eval", "step": 3600, "phase": 2, "bpb": {"en": 1.97545, "tr": 2.14156, "fi": 2.24595}, "bytes_per_chunk": {"en": 4.745, "tr": 5.34, "fi": 4.165}, "bpb_mean": 2.12099, "frozen_inventory": true, "t": 1266.6} |
| {"event": "train", "step": 3600, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.4592, "bpb": 2.10518, "chunk_loss": 0.030513, "F": 0.21027, "G": 0.21426, "chunks": 215.31, "bytes_per_chunk": 4.756, "boundary_entropy": 0.10376, "router_noise": 0.0, "grad_norm": 0.6898, "lr": 0.0001319, "router_trainable": true, "sec_per_step": 0.403, "probe_sec": 0.0345, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02269, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.07248, "n_shadow": 0, "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": 0.12011, "trunc_abs_frac": 10.6864, "mean_delta_full": -0.16162, "read_hit_rate": {"skel": 0.73157}, "mean_gate": {"skel": 0.20282}, "n_reads": 3712, "n_cand": 5074, "read_suppressed_frac": 0.1021, "dropout_reads_logged": 3712, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 213.92625427246094, "n_util_obs": 773, "util_count_spearman": 0.15450930011328945}, "probe_overhead_sec_total": 196.604, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.018433, "alpha": 0.03, "t": 1268.4} |
| {"event": "consolidate", "step": 3650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1298.0} |
| {"event": "train", "step": 3650, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.47464, "bpb": 2.12746, "chunk_loss": 0.030696, "F": 0.21634, "G": 0.22119, "chunks": 221.53, "bytes_per_chunk": 4.622, "boundary_entropy": 0.10688, "router_noise": 0.0, "grad_norm": 0.6851, "lr": 0.0001283, "router_trainable": true, "sec_per_step": 0.599, "probe_sec": 0.0789, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07047, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.06528, "n_shadow": 0, "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": 0.93852, "trunc_abs_frac": 0.43954, "mean_delta_full": 0.25056, "read_hit_rate": {"skel": 0.75957}, "mean_gate": {"skel": 0.21911}, "n_reads": 4069, "n_cand": 5357, "read_suppressed_frac": 0.099, "dropout_reads_logged": 4069, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 216.94801330566406, "n_util_obs": 788, "util_count_spearman": 0.14172621601945443}, "probe_overhead_sec_total": 199.766, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.018186, "alpha": 0.03, "t": 1298.3} |
| {"event": "consolidate", "step": 3700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1325.9} |
| {"event": "train", "step": 3700, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.37435, "bpb": 1.98277, "chunk_loss": 0.030473, "F": 0.20746, "G": 0.21135, "chunks": 212.44, "bytes_per_chunk": 4.82, "boundary_entropy": 0.10284, "router_noise": 0.0, "grad_norm": 0.6984, "lr": 0.0001247, "router_trainable": true, "sec_per_step": 0.568, "probe_sec": 0.0456, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02806, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.16338, "n_shadow": 0, "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": 2.89071, "trunc_abs_frac": 1.00178, "mean_delta_full": -0.00894, "read_hit_rate": {"skel": 0.75019}, "mean_gate": {"skel": 0.21148}, "n_reads": 3916, "n_cand": 5220, "read_suppressed_frac": 0.1011, "dropout_reads_logged": 3916, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 219.98471069335938, "n_util_obs": 803, "util_count_spearman": 0.1423783400893989}, "probe_overhead_sec_total": 202.563, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.018917, "alpha": 0.03, "t": 1326.7} |
| {"event": "consolidate", "step": 3750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1352.5} |
| {"event": "train", "step": 3750, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.39739, "bpb": 2.01601, "chunk_loss": 0.030732, "F": 0.21902, "G": 0.22154, "chunks": 224.28, "bytes_per_chunk": 4.566, "boundary_entropy": 0.09447, "router_noise": 0.0, "grad_norm": 0.6318, "lr": 0.0001212, "router_trainable": true, "sec_per_step": 0.527, "probe_sec": 0.0358, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02256, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.1436, "n_shadow": 0, "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": 1.01445, "trunc_abs_frac": 2.07858, "mean_delta_full": -0.20378, "read_hit_rate": {"skel": 0.77315}, "mean_gate": {"skel": 0.19416}, "n_reads": 4199, "n_cand": 5431, "read_suppressed_frac": 0.1048, "dropout_reads_logged": 4199, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 223.03656005859375, "n_util_obs": 814, "util_count_spearman": 0.14361451809639666}, "probe_overhead_sec_total": 205.068, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.01829, "alpha": 0.03, "t": 1353.0} |
| {"event": "consolidate", "step": 3800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1376.4} |
| {"event": "probe", "step": 3800, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3507.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.201517, "cps": 27.398438, "mean_p": 0.205316, "boundary_entropy": 0.096256, "bpb": 1.947649}, "tr": {"chunks": 3779.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.20875, "cps": 29.523438, "mean_p": 0.199093, "boundary_entropy": 0.128907, "bpb": 2.112306}, "fi": {"chunks": 4849.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.257775, "cps": 37.882812, "mean_p": 0.257587, "boundary_entropy": 0.10532, "bpb": 2.214601}}, "bpb": {"en": 1.94765, "tr": 2.11231, "fi": 2.2146}, "gini_cpc": 0.056142, "gini_cps": 0.073726, "bpic": 4.675072, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0358925144951612, "fi": 1.2791711472897822}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 226.02735900878906, "n_util_obs": 821, "util_count_spearman": 0.13592827085022438}, "t": 1377.3} |
| {"event": "eval", "step": 3800, "phase": 2, "bpb": {"en": 1.96073, "tr": 2.12053, "fi": 2.22342}, "bytes_per_chunk": {"en": 4.723, "tr": 5.35, "fi": 4.154}, "bpb_mean": 2.10156, "frozen_inventory": true, "t": 1378.5} |
| {"event": "train", "step": 3800, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.37797, "bpb": 1.98799, "chunk_loss": 0.030567, "F": 0.20926, "G": 0.21154, "chunks": 214.28, "bytes_per_chunk": 4.779, "boundary_entropy": 0.09994, "router_noise": 0.0, "grad_norm": 0.6464, "lr": 0.0001177, "router_trainable": true, "sec_per_step": 0.534, "probe_sec": 0.1085, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.1013, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.32549, "n_shadow": 0, "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": 0.35821, "trunc_abs_frac": 1.0203, "mean_delta_full": 0.19489, "read_hit_rate": {"skel": 0.74774}, "mean_gate": {"skel": 0.20931}, "n_reads": 3880, "n_cand": 5189, "read_suppressed_frac": 0.0977, "dropout_reads_logged": 3880, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 226.08660888671875, "n_util_obs": 821, "util_count_spearman": 0.1363073922854956}, "probe_overhead_sec_total": 208.209, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.01688, "alpha": 0.03, "t": 1379.8} |
| {"event": "consolidate", "step": 3850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1392.0} |
| {"event": "train", "step": 3850, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.36637, "bpb": 1.97126, "chunk_loss": 0.030378, "F": 0.20255, "G": 0.20669, "chunks": 207.41, "bytes_per_chunk": 4.937, "boundary_entropy": 0.10135, "router_noise": 0.0, "grad_norm": 0.7366, "lr": 0.0001142, "router_trainable": true, "sec_per_step": 0.25, "probe_sec": 0.0302, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02194, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.00168, "n_shadow": 0, "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": -1.38185, "trunc_abs_frac": 0.71296, "mean_delta_full": 0.07629, "read_hit_rate": {"skel": 0.74056}, "mean_gate": {"skel": 0.21496}, "n_reads": 3805, "n_cand": 5138, "read_suppressed_frac": 0.1014, "dropout_reads_logged": 3805, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 229.1429443359375, "n_util_obs": 831, "util_count_spearman": 0.1369016206198308}, "probe_overhead_sec_total": 211.812, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.017336, "alpha": 0.03, "t": 1392.3} |
| {"event": "consolidate", "step": 3900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1404.1} |
| {"event": "train", "step": 3900, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.39987, "bpb": 2.01959, "chunk_loss": 0.030546, "F": 0.20996, "G": 0.21367, "chunks": 215.0, "bytes_per_chunk": 4.763, "boundary_entropy": 0.10158, "router_noise": 0.0, "grad_norm": 0.6666, "lr": 0.0001108, "router_trainable": true, "sec_per_step": 0.243, "probe_sec": 0.0477, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04058, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.04736, "n_shadow": 0, "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": 0.98566, "trunc_abs_frac": 0.81154, "mean_delta_full": -0.08514, "read_hit_rate": {"skel": 0.73758}, "mean_gate": {"skel": 0.21094}, "n_reads": 3876, "n_cand": 5255, "read_suppressed_frac": 0.1001, "dropout_reads_logged": 3876, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 232.19810485839844, "n_util_obs": 844, "util_count_spearman": 0.13009564991004943}, "probe_overhead_sec_total": 214.288, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.019593, "alpha": 0.03, "t": 1404.4} |
| {"event": "consolidate", "step": 3950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1416.3} |
| {"event": "train", "step": 3950, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.42713, "bpb": 2.05892, "chunk_loss": 0.030642, "F": 0.21603, "G": 0.2186, "chunks": 221.22, "bytes_per_chunk": 4.629, "boundary_entropy": 0.09617, "router_noise": 0.0, "grad_norm": 0.6876, "lr": 0.0001074, "router_trainable": true, "sec_per_step": 0.242, "probe_sec": 0.0498, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03883, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.11167, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 1.22573, "trunc_abs_frac": 1.22573, "mean_delta_full": -0.40613, "read_hit_rate": {"skel": 0.75252}, "mean_gate": {"skel": 0.20113}, "n_reads": 3877, "n_cand": 5152, "read_suppressed_frac": 0.091, "dropout_reads_logged": 3877, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 235.2335968017578, "n_util_obs": 856, "util_count_spearman": 0.12896571384259228}, "probe_overhead_sec_total": 216.692, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.017618, "alpha": 0.03, "t": 1416.5} |
| {"event": "consolidate", "step": 4000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1429.6} |
| {"event": "probe", "step": 4000, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3519.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.202207, "cps": 27.492188, "mean_p": 0.204641, "boundary_entropy": 0.094697, "bpb": 1.933711}, "tr": {"chunks": 3744.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.206817, "cps": 29.25, "mean_p": 0.19697, "boundary_entropy": 0.131267, "bpb": 2.097994}, "fi": {"chunks": 4859.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.258306, "cps": 37.960938, "mean_p": 0.256716, "boundary_entropy": 0.100849, "bpb": 2.199373}}, "bpb": {"en": 1.93371, "tr": 2.09799, "fi": 2.19937}, "gini_cpc": 0.056044, "gini_cps": 0.073695, "bpic": 4.680086, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0227986402897995, "fi": 1.2774381117721338}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 238.2176055908203, "n_util_obs": 866, "util_count_spearman": 0.1139658772805771}, "t": 1430.4} |
| {"event": "eval", "step": 4000, "phase": 2, "bpb": {"en": 1.9488, "tr": 2.10581, "fi": 2.20828}, "bytes_per_chunk": {"en": 4.723, "tr": 5.349, "fi": 4.142}, "bpb_mean": 2.08763, "frozen_inventory": true, "t": 1431.7} |
| {"event": "train", "step": 4000, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.4495, "bpb": 2.09119, "chunk_loss": 0.030659, "F": 0.21799, "G": 0.21966, "chunks": 223.22, "bytes_per_chunk": 4.587, "boundary_entropy": 0.09992, "router_noise": 0.0, "grad_norm": 0.7445, "lr": 0.000104, "router_trainable": true, "sec_per_step": 0.332, "probe_sec": 0.1042, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.095, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.07543, "n_shadow": 0, "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": 1.27137, "trunc_abs_frac": 1.03002, "mean_delta_full": 0.2099, "read_hit_rate": {"skel": 0.75173}, "mean_gate": {"skel": 0.20166}, "n_reads": 3900, "n_cand": 5188, "read_suppressed_frac": 0.1005, "dropout_reads_logged": 3900, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 238.276123046875, "n_util_obs": 866, "util_count_spearman": 0.11413317392728246}, "probe_overhead_sec_total": 220.414, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.018073, "alpha": 0.03, "t": 1433.2} |
| {"event": "consolidate", "step": 4050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1445.3} |
| {"event": "train", "step": 4050, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.36762, "bpb": 1.97306, "chunk_loss": 0.030565, "F": 0.21106, "G": 0.21407, "chunks": 216.12, "bytes_per_chunk": 4.738, "boundary_entropy": 0.09034, "router_noise": 0.0, "grad_norm": 0.6766, "lr": 0.0001007, "router_trainable": true, "sec_per_step": 0.248, "probe_sec": 0.057, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04748, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.0398, "n_shadow": 0, "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": 0.19537, "trunc_abs_frac": 0.19537, "mean_delta_full": 0.95422, "read_hit_rate": {"skel": 0.74642}, "mean_gate": {"skel": 0.19956}, "n_reads": 3853, "n_cand": 5162, "read_suppressed_frac": 0.0971, "dropout_reads_logged": 3853, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 241.32310485839844, "n_util_obs": 880, "util_count_spearman": 0.115119382870183}, "probe_overhead_sec_total": 223.225, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.019114, "alpha": 0.03, "t": 1445.5} |
| {"event": "probe", "step": 4096, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3508.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.201574, "cps": 27.40625, "mean_p": 0.205047, "boundary_entropy": 0.091696, "bpb": 1.926225}, "tr": {"chunks": 3738.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.206485, "cps": 29.203125, "mean_p": 0.196796, "boundary_entropy": 0.130696, "bpb": 2.087981}, "fi": {"chunks": 4824.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.256446, "cps": 37.6875, "mean_p": 0.2573, "boundary_entropy": 0.104528, "bpb": 2.1836}}, "bpb": {"en": 1.92623, "tr": 2.08798, "fi": 2.1836}, "gini_cpc": 0.05505, "gini_cps": 0.072687, "bpic": 4.700249, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0243615795110343, "fi": 1.2722133577802406}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 244.0853271484375, "n_util_obs": 887, "util_count_spearman": 0.11753239264337033}, "t": 1457.2} |
| {"event": "consolidate", "step": 4100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1459.4} |
| {"event": "train", "step": 4100, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.44156, "bpb": 2.07973, "chunk_loss": 0.030723, "F": 0.21844, "G": 0.22196, "chunks": 223.69, "bytes_per_chunk": 4.578, "boundary_entropy": 0.09866, "router_noise": 0.0, "grad_norm": 0.6894, "lr": 9.75e-05, "router_trainable": true, "sec_per_step": 0.281, "probe_sec": 0.0438, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.034, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.1933, "n_shadow": 0, "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": 0.81332, "trunc_abs_frac": 0.91305, "mean_delta_full": 0.28723, "read_hit_rate": {"skel": 0.77156}, "mean_gate": {"skel": 0.1911}, "n_reads": 4205, "n_cand": 5450, "read_suppressed_frac": 0.0982, "dropout_reads_logged": 4205, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 244.39471435546875, "n_util_obs": 888, "util_count_spearman": 0.11880079333983205}, "probe_overhead_sec_total": 226.044, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.019572, "alpha": 0.03, "t": 1459.6} |
| {"event": "consolidate", "step": 4150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1471.5} |
| {"event": "train", "step": 4150, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.3888, "bpb": 2.00361, "chunk_loss": 0.030822, "F": 0.22318, "G": 0.22434, "chunks": 228.53, "bytes_per_chunk": 4.481, "boundary_entropy": 0.09199, "router_noise": 0.0, "grad_norm": 0.6708, "lr": 9.43e-05, "router_trainable": true, "sec_per_step": 0.243, "probe_sec": 0.0509, "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": -0.00024, "n_shadow": 0, "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": -0.5943, "trunc_abs_frac": 0.87083, "mean_delta_full": -0.31331, "read_hit_rate": {"skel": 0.77022}, "mean_gate": {"skel": 0.19007}, "n_reads": 4324, "n_cand": 5614, "read_suppressed_frac": 0.1161, "dropout_reads_logged": 4324, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 247.47508239746094, "n_util_obs": 901, "util_count_spearman": 0.11596421798840151}, "probe_overhead_sec_total": 228.631, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.018382, "alpha": 0.03, "t": 1471.8} |
| {"event": "consolidate", "step": 4200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1483.6} |
| {"event": "probe", "step": 4200, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3525.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.202551, "cps": 27.539062, "mean_p": 0.20521, "boundary_entropy": 0.091485, "bpb": 1.923306}, "tr": {"chunks": 3754.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.207369, "cps": 29.328125, "mean_p": 0.195917, "boundary_entropy": 0.124094, "bpb": 2.077831}, "fi": {"chunks": 4851.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.257881, "cps": 37.898438, "mean_p": 0.257266, "boundary_entropy": 0.09863, "bpb": 2.174432}}, "bpb": {"en": 1.92331, "tr": 2.07783, "fi": 2.17443}, "gini_cpc": 0.055236, "gini_cps": 0.072877, "bpic": 4.676999, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0237848904789497, "fi": 1.2731641174188484}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 250.47938537597656, "n_util_obs": 908, "util_count_spearman": 0.12319032958855518}, "t": 1484.5} |
| {"event": "eval", "step": 4200, "phase": 2, "bpb": {"en": 1.93637, "tr": 2.0853, "fi": 2.18311}, "bytes_per_chunk": {"en": 4.732, "tr": 5.335, "fi": 4.155}, "bpb_mean": 2.06826, "frozen_inventory": true, "t": 1485.8} |
| {"event": "phase_transition", "step": 4200, "phase": 3, "objective": "global", "router_trainable": true, "encoder_trainable": true, "frozen_router_params": 0, "t": 1487.1} |
| {"event": "train", "step": 4200, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38048, "bpb": 1.99162, "chunk_loss": 0.03047, "F": 0.20941, "G": 0.21119, "chunks": 214.44, "bytes_per_chunk": 4.775, "boundary_entropy": 0.09197, "router_noise": 0.0, "grad_norm": 0.6546, "lr": 9.12e-05, "router_trainable": true, "sec_per_step": 0.311, "probe_sec": 0.0331, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02277, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.12389, "n_shadow": 0, "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": {"skel": 0.74828}, "mean_gate": {"skel": 0.19662}, "n_reads": 3915, "n_cand": 5232, "read_suppressed_frac": 0.0976, "dropout_reads_logged": 3915, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 250.53773498535156, "n_util_obs": 908, "util_count_spearman": 0.12378463822096329}, "probe_overhead_sec_total": 230.96, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.017839, "alpha": 0.03, "t": 1487.3} |
| {"event": "consolidate", "step": 4250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1499.8} |
| {"event": "train", "step": 4250, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38136, "bpb": 1.99288, "chunk_loss": 0.030439, "F": 0.20303, "G": 0.20703, "chunks": 207.91, "bytes_per_chunk": 4.925, "boundary_entropy": 0.09864, "router_noise": 0.0, "grad_norm": 0.6892, "lr": 8.81e-05, "router_trainable": true, "sec_per_step": 0.254, "probe_sec": 0.0373, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0311, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.31873, "n_shadow": 0, "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": 0.6233, "trunc_abs_frac": 0.6686, "mean_delta_full": -0.0342, "read_hit_rate": {"skel": 0.74053}, "mean_gate": {"skel": 0.20669}, "n_reads": 3833, "n_cand": 5176, "read_suppressed_frac": 0.0924, "dropout_reads_logged": 3833, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 253.60598754882812, "n_util_obs": 919, "util_count_spearman": 0.12915001466062603}, "probe_overhead_sec_total": 233.442, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.017311, "alpha": 0.03, "t": 1500.0} |
| {"event": "consolidate", "step": 4300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1511.9} |
| {"event": "train", "step": 4300, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.46361, "bpb": 2.11155, "chunk_loss": 0.030585, "F": 0.21225, "G": 0.2153, "chunks": 217.34, "bytes_per_chunk": 4.711, "boundary_entropy": 0.10263, "router_noise": 0.0, "grad_norm": 0.7478, "lr": 8.5e-05, "router_trainable": true, "sec_per_step": 0.245, "probe_sec": 0.1302, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.11802, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.1538, "n_shadow": 0, "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": -0.70726, "trunc_abs_frac": 3.53937, "mean_delta_full": 0.15262, "read_hit_rate": {"skel": 0.74662}, "mean_gate": {"skel": 0.20777}, "n_reads": 3981, "n_cand": 5332, "read_suppressed_frac": 0.096, "dropout_reads_logged": 3981, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 256.6544189453125, "n_util_obs": 922, "util_count_spearman": 0.12463490941719434}, "probe_overhead_sec_total": 235.834, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.018655, "alpha": 0.03, "t": 1512.3} |
| {"event": "consolidate", "step": 4350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1529.5} |
| {"event": "train", "step": 4350, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32342, "bpb": 1.9093, "chunk_loss": 0.030578, "F": 0.21191, "G": 0.2132, "chunks": 217.0, "bytes_per_chunk": 4.719, "boundary_entropy": 0.0904, "router_noise": 0.0, "grad_norm": 0.7108, "lr": 8.21e-05, "router_trainable": true, "sec_per_step": 0.35, "probe_sec": 0.0817, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07351, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.2478, "n_shadow": 0, "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": 0.49797, "trunc_abs_frac": 0.7525, "mean_delta_full": -0.396, "read_hit_rate": {"skel": 0.76028}, "mean_gate": {"skel": 0.19347}, "n_reads": 4050, "n_cand": 5327, "read_suppressed_frac": 0.0956, "dropout_reads_logged": 4050, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 259.70599365234375, "n_util_obs": 928, "util_count_spearman": 0.12067051347079842}, "probe_overhead_sec_total": 238.56, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.019057, "alpha": 0.03, "t": 1529.7} |
| {"event": "consolidate", "step": 4400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1544.6} |
| {"event": "probe", "step": 4400, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3537.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.203241, "cps": 27.632812, "mean_p": 0.204848, "boundary_entropy": 0.089535, "bpb": 1.910738}, "tr": {"chunks": 3747.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.206982, "cps": 29.273438, "mean_p": 0.195418, "boundary_entropy": 0.124319, "bpb": 2.061918}, "fi": {"chunks": 4875.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.259157, "cps": 38.085938, "mean_p": 0.257252, "boundary_entropy": 0.096869, "bpb": 2.159918}}, "bpb": {"en": 1.91074, "tr": 2.06192, "fi": 2.15992}, "gini_cpc": 0.055689, "gini_cps": 0.073361, "bpic": 4.665844, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0184089376045666, "fi": 1.2751221708158145}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 262.7008972167969, "n_util_obs": 940, "util_count_spearman": 0.12119446854930471}, "t": 1545.6} |
| {"event": "eval", "step": 4400, "phase": 3, "bpb": {"en": 1.92507, "tr": 2.06996, "fi": 2.16766}, "bytes_per_chunk": {"en": 4.724, "tr": 5.354, "fi": 4.125}, "bpb_mean": 2.05423, "frozen_inventory": true, "t": 1546.9} |
| {"event": "train", "step": 4400, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.43716, "bpb": 2.07339, "chunk_loss": 0.03078, "F": 0.2178, "G": 0.22056, "chunks": 223.03, "bytes_per_chunk": 4.591, "boundary_entropy": 0.09341, "router_noise": 0.0, "grad_norm": 0.71, "lr": 7.92e-05, "router_trainable": true, "sec_per_step": 0.376, "probe_sec": 0.0414, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02571, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.03226, "n_shadow": 0, "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": 0.44822, "trunc_abs_frac": 0.75124, "mean_delta_full": 0.31946, "read_hit_rate": {"skel": 0.7596}, "mean_gate": {"skel": 0.19147}, "n_reads": 4073, "n_cand": 5362, "read_suppressed_frac": 0.095, "dropout_reads_logged": 4073, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 262.76312255859375, "n_util_obs": 940, "util_count_spearman": 0.1212082733120844}, "probe_overhead_sec_total": 241.003, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.020186, "alpha": 0.03, "t": 1548.6} |
| {"event": "consolidate", "step": 4450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1560.5} |
| {"event": "train", "step": 4450, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.37549, "bpb": 1.98441, "chunk_loss": 0.030602, "F": 0.21335, "G": 0.2154, "chunks": 218.47, "bytes_per_chunk": 4.687, "boundary_entropy": 0.09398, "router_noise": 0.0, "grad_norm": 0.6677, "lr": 7.63e-05, "router_trainable": true, "sec_per_step": 0.253, "probe_sec": 0.0696, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04477, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.1585, "n_shadow": 0, "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": 0.68281, "trunc_abs_frac": 0.6749, "mean_delta_full": 0.60905, "read_hit_rate": {"skel": 0.7547}, "mean_gate": {"skel": 0.19456}, "n_reads": 3975, "n_cand": 5267, "read_suppressed_frac": 0.1029, "dropout_reads_logged": 3975, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 265.803466796875, "n_util_obs": 953, "util_count_spearman": 0.12511961826803414}, "probe_overhead_sec_total": 244.534, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.017546, "alpha": 0.03, "t": 1561.2} |
| {"event": "consolidate", "step": 4500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1591.8} |
| {"event": "train", "step": 4500, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38837, "bpb": 2.00299, "chunk_loss": 0.030699, "F": 0.21729, "G": 0.21899, "chunks": 222.5, "bytes_per_chunk": 4.602, "boundary_entropy": 0.09375, "router_noise": 0.0, "grad_norm": 0.7313, "lr": 7.36e-05, "router_trainable": true, "sec_per_step": 0.617, "probe_sec": 0.1396, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.12735, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.11388, "n_shadow": 0, "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": 1.06038, "trunc_abs_frac": 1.06277, "mean_delta_full": 0.19861, "read_hit_rate": {"skel": 0.77202}, "mean_gate": {"skel": 0.19122}, "n_reads": 4260, "n_cand": 5518, "read_suppressed_frac": 0.0937, "dropout_reads_logged": 4260, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 268.85107421875, "n_util_obs": 957, "util_count_spearman": 0.11917819165217548}, "probe_overhead_sec_total": 247.514, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.017849, "alpha": 0.03, "t": 1592.1} |
| {"event": "consolidate", "step": 4550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1616.9} |
| {"event": "train", "step": 4550, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.36182, "bpb": 1.96469, "chunk_loss": 0.03064, "F": 0.21588, "G": 0.21734, "chunks": 221.06, "bytes_per_chunk": 4.632, "boundary_entropy": 0.09235, "router_noise": 0.0, "grad_norm": 0.6808, "lr": 7.09e-05, "router_trainable": true, "sec_per_step": 0.501, "probe_sec": 0.0673, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05413, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.35618, "n_shadow": 0, "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": -0.17979, "trunc_abs_frac": 0.28641, "mean_delta_full": -0.62604, "read_hit_rate": {"skel": 0.7732}, "mean_gate": {"skel": 0.19383}, "n_reads": 4183, "n_cand": 5410, "read_suppressed_frac": 0.1014, "dropout_reads_logged": 4183, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 271.8758544921875, "n_util_obs": 965, "util_count_spearman": 0.12081673635127824}, "probe_overhead_sec_total": 250.33, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.018567, "alpha": 0.03, "t": 1617.1} |
| {"event": "consolidate", "step": 4600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1643.8} |
| {"event": "probe", "step": 4600, "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": 0.205332, "boundary_entropy": 0.086394, "bpb": 1.90001}, "tr": {"chunks": 3707.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.204773, "cps": 28.960938, "mean_p": 0.19186, "boundary_entropy": 0.129341, "bpb": 2.051136}, "fi": {"chunks": 4870.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.258891, "cps": 38.046875, "mean_p": 0.257545, "boundary_entropy": 0.09613, "bpb": 2.150381}}, "bpb": {"en": 1.90001, "tr": 2.05114, "fi": 2.15038}, "gini_cpc": 0.055693, "gini_cps": 0.07342, "bpic": 4.683563, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.007822148550164, "fi": 1.2741745948204464}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 274.8499755859375, "n_util_obs": 969, "util_count_spearman": 0.11735980216432053}, "t": 1644.8} |
| {"event": "eval", "step": 4600, "phase": 3, "bpb": {"en": 1.91379, "tr": 2.05922, "fi": 2.15835}, "bytes_per_chunk": {"en": 4.722, "tr": 5.442, "fi": 4.143}, "bpb_mean": 2.04379, "frozen_inventory": true, "t": 1646.1} |
| {"event": "train", "step": 4600, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.42334, "bpb": 2.05345, "chunk_loss": 0.030507, "F": 0.20621, "G": 0.20978, "chunks": 211.16, "bytes_per_chunk": 4.849, "boundary_entropy": 0.0952, "router_noise": 0.0, "grad_norm": 0.7338, "lr": 6.82e-05, "router_trainable": true, "sec_per_step": 0.625, "probe_sec": 0.0713, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04375, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.2887, "n_shadow": 0, "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": 0.802, "trunc_abs_frac": 0.55045, "mean_delta_full": 1.06271, "read_hit_rate": {"skel": 0.74476}, "mean_gate": {"skel": 0.20196}, "n_reads": 3907, "n_cand": 5246, "read_suppressed_frac": 0.0996, "dropout_reads_logged": 3907, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 274.9095458984375, "n_util_obs": 969, "util_count_spearman": 0.11704109879126046}, "probe_overhead_sec_total": 253.265, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.016362, "alpha": 0.03, "t": 1648.4} |
| {"event": "consolidate", "step": 4650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1671.2} |
| {"event": "train", "step": 4650, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.33833, "bpb": 1.9308, "chunk_loss": 0.030499, "F": 0.2088, "G": 0.21077, "chunks": 213.81, "bytes_per_chunk": 4.789, "boundary_entropy": 0.09637, "router_noise": 0.0, "grad_norm": 0.6479, "lr": 6.57e-05, "router_trainable": true, "sec_per_step": 0.462, "probe_sec": 0.0439, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03834, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.2331, "n_shadow": 0, "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": 1.33737, "trunc_abs_frac": 98.71268, "mean_delta_full": 0.42108, "read_hit_rate": {"skel": 0.75881}, "mean_gate": {"skel": 0.19629}, "n_reads": 4046, "n_cand": 5332, "read_suppressed_frac": 0.1003, "dropout_reads_logged": 4046, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 277.94769287109375, "n_util_obs": 978, "util_count_spearman": 0.1221775937159921}, "probe_overhead_sec_total": 256.001, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.017356, "alpha": 0.03, "t": 1671.5} |
| {"event": "consolidate", "step": 4700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1684.8} |
| {"event": "train", "step": 4700, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35864, "bpb": 1.96011, "chunk_loss": 0.030544, "F": 0.20963, "G": 0.21106, "chunks": 214.66, "bytes_per_chunk": 4.77, "boundary_entropy": 0.094, "router_noise": 0.0, "grad_norm": 0.6687, "lr": 6.32e-05, "router_trainable": true, "sec_per_step": 0.27, "probe_sec": 0.0306, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02282, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.09229, "n_shadow": 0, "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": -2.62812, "trunc_abs_frac": 2.62812, "mean_delta_full": 0.03906, "read_hit_rate": {"skel": 0.74581}, "mean_gate": {"skel": 0.19201}, "n_reads": 3920, "n_cand": 5256, "read_suppressed_frac": 0.0995, "dropout_reads_logged": 3920, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 280.9847106933594, "n_util_obs": 991, "util_count_spearman": 0.11706596785446148}, "probe_overhead_sec_total": 258.38, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.019231, "alpha": 0.03, "t": 1685.0} |
| {"event": "consolidate", "step": 4750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1698.4} |
| {"event": "train", "step": 4750, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.31882, "bpb": 1.90266, "chunk_loss": 0.030601, "F": 0.21375, "G": 0.21487, "chunks": 218.88, "bytes_per_chunk": 4.678, "boundary_entropy": 0.09117, "router_noise": 0.0, "grad_norm": 0.6751, "lr": 6.08e-05, "router_trainable": true, "sec_per_step": 0.272, "probe_sec": 0.0392, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03005, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.01426, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 0.37178, "trunc_abs_frac": 0.48768, "mean_delta_full": 0.18231, "read_hit_rate": {"skel": 0.77053}, "mean_gate": {"skel": 0.19318}, "n_reads": 4100, "n_cand": 5321, "read_suppressed_frac": 0.0998, "dropout_reads_logged": 4100, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 284.0487060546875, "n_util_obs": 997, "util_count_spearman": 0.12438244748741496}, "probe_overhead_sec_total": 261.093, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.019493, "alpha": 0.03, "t": 1698.6} |
| {"event": "consolidate", "step": 4800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1712.4} |
| {"event": "probe", "step": 4800, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3533.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.203011, "cps": 27.601562, "mean_p": 0.205301, "boundary_entropy": 0.08427, "bpb": 1.889087}, "tr": {"chunks": 3742.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.206706, "cps": 29.234375, "mean_p": 0.193141, "boundary_entropy": 0.126763, "bpb": 2.038505}, "fi": {"chunks": 4865.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.258625, "cps": 38.007812, "mean_p": 0.25675, "boundary_entropy": 0.094652, "bpb": 2.134734}}, "bpb": {"en": 1.88909, "tr": 2.03851, "fi": 2.13473}, "gini_cpc": 0.055475, "gini_cps": 0.073147, "bpic": 4.673147, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0182014578058607, "fi": 1.2739472449461258}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 287.04034423828125, "n_util_obs": 1005, "util_count_spearman": 0.1278337911842115}, "t": 1713.4} |
| {"event": "eval", "step": 4800, "phase": 3, "bpb": {"en": 1.90359, "tr": 2.04611, "fi": 2.14443}, "bytes_per_chunk": {"en": 4.724, "tr": 5.394, "fi": 4.15}, "bpb_mean": 2.03138, "frozen_inventory": true, "t": 1714.9} |
| {"event": "train", "step": 4800, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.39373, "bpb": 2.01073, "chunk_loss": 0.030771, "F": 0.21872, "G": 0.22351, "chunks": 223.97, "bytes_per_chunk": 4.572, "boundary_entropy": 0.09307, "router_noise": 0.0, "grad_norm": 0.7064, "lr": 5.85e-05, "router_trainable": true, "sec_per_step": 0.369, "probe_sec": 0.0904, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07767, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.02357, "n_shadow": 0, "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": -24.01923, "trunc_abs_frac": 1.14761, "mean_delta_full": 0.00068, "read_hit_rate": {"skel": 0.77002}, "mean_gate": {"skel": 0.18913}, "n_reads": 4202, "n_cand": 5457, "read_suppressed_frac": 0.098, "dropout_reads_logged": 4202, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 287.1054992675781, "n_util_obs": 1005, "util_count_spearman": 0.12790604526205712}, "probe_overhead_sec_total": 264.514, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.020579, "alpha": 0.03, "t": 1717.1} |
| {"event": "consolidate", "step": 4850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1730.3} |
| {"event": "train", "step": 4850, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.34695, "bpb": 1.94324, "chunk_loss": 0.030701, "F": 0.21771, "G": 0.21992, "chunks": 222.94, "bytes_per_chunk": 4.593, "boundary_entropy": 0.09347, "router_noise": 0.0, "grad_norm": 0.705, "lr": 5.62e-05, "router_trainable": true, "sec_per_step": 0.273, "probe_sec": 0.0992, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08288, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.03212, "n_shadow": 0, "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": 0.13151, "trunc_abs_frac": 1.40673, "mean_delta_full": -0.34615, "read_hit_rate": {"skel": 0.7736}, "mean_gate": {"skel": 0.18733}, "n_reads": 4097, "n_cand": 5296, "read_suppressed_frac": 0.105, "dropout_reads_logged": 4097, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 290.13507080078125, "n_util_obs": 1023, "util_count_spearman": 0.14103486495077636}, "probe_overhead_sec_total": 267.325, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.021849, "alpha": 0.03, "t": 1730.7} |
| {"event": "consolidate", "step": 4900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1743.7} |
| {"event": "train", "step": 4900, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35264, "bpb": 1.95144, "chunk_loss": 0.030467, "F": 0.20776, "G": 0.20955, "chunks": 212.75, "bytes_per_chunk": 4.813, "boundary_entropy": 0.089, "router_noise": 0.0, "grad_norm": 0.711, "lr": 5.41e-05, "router_trainable": true, "sec_per_step": 0.266, "probe_sec": 0.1188, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.11142, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.20107, "n_shadow": 0, "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": 0.63909, "trunc_abs_frac": 0.47629, "mean_delta_full": 0.94611, "read_hit_rate": {"skel": 0.74693}, "mean_gate": {"skel": 0.18338}, "n_reads": 3893, "n_cand": 5212, "read_suppressed_frac": 0.1007, "dropout_reads_logged": 3893, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 293.19378662109375, "n_util_obs": 1027, "util_count_spearman": 0.14714964448962425}, "probe_overhead_sec_total": 269.929, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.019226, "alpha": 0.03, "t": 1744.0} |
| {"event": "consolidate", "step": 4950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1756.7} |
| {"event": "train", "step": 4950, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.33516, "bpb": 1.92623, "chunk_loss": 0.030671, "F": 0.21704, "G": 0.2182, "chunks": 222.25, "bytes_per_chunk": 4.607, "boundary_entropy": 0.08519, "router_noise": 0.0, "grad_norm": 0.7022, "lr": 5.2e-05, "router_trainable": true, "sec_per_step": 0.259, "probe_sec": 0.0452, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03609, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.22591, "n_shadow": 0, "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": 0.66763, "trunc_abs_frac": 1.30106, "mean_delta_full": 0.37761, "read_hit_rate": {"skel": 0.76014}, "mean_gate": {"skel": 0.17197}, "n_reads": 4031, "n_cand": 5303, "read_suppressed_frac": 0.101, "dropout_reads_logged": 4031, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 296.24249267578125, "n_util_obs": 1033, "util_count_spearman": 0.14608015861660806}, "probe_overhead_sec_total": 272.896, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.0175, "alpha": 0.03, "t": 1756.9} |
| {"event": "consolidate", "step": 5000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1770.3} |
| {"event": "probe", "step": 5000, "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": 0.203888, "boundary_entropy": 0.084849, "bpb": 1.880279}, "tr": {"chunks": 3725.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.205767, "cps": 29.101562, "mean_p": 0.191835, "boundary_entropy": 0.124774, "bpb": 2.025217}, "fi": {"chunks": 4844.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.257509, "cps": 37.84375, "mean_p": 0.256124, "boundary_entropy": 0.095225, "bpb": 2.122986}}, "bpb": {"en": 1.88028, "tr": 2.02522, "fi": 2.12299}, "gini_cpc": 0.055588, "gini_cps": 0.073259, "bpic": 4.694415, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0184764220815323, "fi": 1.274581189606038}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 299.2225036621094, "n_util_obs": 1043, "util_count_spearman": 0.15212882322978707}, "t": 1771.2} |
| {"event": "eval", "step": 5000, "phase": 3, "bpb": {"en": 1.89534, "tr": 2.03418, "fi": 2.13182}, "bytes_per_chunk": {"en": 4.733, "tr": 5.405, "fi": 4.159}, "bpb_mean": 2.02045, "frozen_inventory": true, "t": 1772.7} |
| {"event": "train", "step": 5000, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.33417, "bpb": 1.9248, "chunk_loss": 0.030587, "F": 0.21295, "G": 0.21386, "chunks": 218.06, "bytes_per_chunk": 4.696, "boundary_entropy": 0.09032, "router_noise": 0.0, "grad_norm": 0.7103, "lr": 5e-05, "router_trainable": true, "sec_per_step": 0.357, "probe_sec": 0.1191, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10989, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.00308, "n_shadow": 0, "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": 0.53064, "trunc_abs_frac": 0.50777, "mean_delta_full": 0.14142, "read_hit_rate": {"skel": 0.75038}, "mean_gate": {"skel": 0.18957}, "n_reads": 3944, "n_cand": 5256, "read_suppressed_frac": 0.0996, "dropout_reads_logged": 3944, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 299.28314208984375, "n_util_obs": 1043, "util_count_spearman": 0.15187445916891143}, "probe_overhead_sec_total": 275.51, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.017855, "alpha": 0.03, "t": 1774.8} |
| {"event": "consolidate", "step": 5050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1788.3} |
| {"event": "train", "step": 5050, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.3514, "bpb": 1.94965, "chunk_loss": 0.030557, "F": 0.21271, "G": 0.21436, "chunks": 217.81, "bytes_per_chunk": 4.701, "boundary_entropy": 0.08905, "router_noise": 0.0, "grad_norm": 0.6332, "lr": 4.81e-05, "router_trainable": true, "sec_per_step": 0.276, "probe_sec": 0.0464, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0389, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.03065, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 1, "trunc_ratio_sum": 1.0, "trunc_abs_frac": 1.0, "mean_delta_full": 0.25732, "read_hit_rate": {"skel": 0.75}, "mean_gate": {"skel": 0.18618}, "n_reads": 3939, "n_cand": 5252, "read_suppressed_frac": 0.0957, "dropout_reads_logged": 3939, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 302.345947265625, "n_util_obs": 1052, "util_count_spearman": 0.14554361900115054}, "probe_overhead_sec_total": 278.547, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.017528, "alpha": 0.03, "t": 1788.6} |
| {"event": "consolidate", "step": 5100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1801.7} |
| {"event": "train", "step": 5100, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.31842, "bpb": 1.90207, "chunk_loss": 0.030396, "F": 0.20642, "G": 0.20637, "chunks": 211.38, "bytes_per_chunk": 4.844, "boundary_entropy": 0.08616, "router_noise": 0.0, "grad_norm": 0.6714, "lr": 4.63e-05, "router_trainable": true, "sec_per_step": 0.267, "probe_sec": 0.0414, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0338, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.06379, "n_shadow": 0, "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": 0.89477, "trunc_abs_frac": 27.71682, "mean_delta_full": -0.32916, "read_hit_rate": {"skel": 0.75481}, "mean_gate": {"skel": 0.18087}, "n_reads": 3848, "n_cand": 5098, "read_suppressed_frac": 0.1014, "dropout_reads_logged": 3848, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 305.39361572265625, "n_util_obs": 1060, "util_count_spearman": 0.14431418596949985}, "probe_overhead_sec_total": 281.151, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.016625, "alpha": 0.03, "t": 1801.9} |
| {"event": "consolidate", "step": 5150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1817.1} |
| {"event": "train", "step": 5150, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.30752, "bpb": 1.88635, "chunk_loss": 0.030611, "F": 0.21429, "G": 0.21475, "chunks": 219.44, "bytes_per_chunk": 4.666, "boundary_entropy": 0.08817, "router_noise": 0.0, "grad_norm": 0.6913, "lr": 4.45e-05, "router_trainable": true, "sec_per_step": 0.31, "probe_sec": 0.1068, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09264, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.17534, "n_shadow": 0, "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": -1.11643, "trunc_abs_frac": 0.86545, "mean_delta_full": -0.11454, "read_hit_rate": {"skel": 0.77568}, "mean_gate": {"skel": 0.17879}, "n_reads": 4184, "n_cand": 5394, "read_suppressed_frac": 0.1004, "dropout_reads_logged": 4184, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 308.43756103515625, "n_util_obs": 1073, "util_count_spearman": 0.15086862131547296}, "probe_overhead_sec_total": 283.679, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.018839, "alpha": 0.03, "t": 1817.4} |
| {"event": "consolidate", "step": 5200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1833.7} |
| {"event": "probe", "step": 5200, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3526.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.202609, "cps": 27.546875, "mean_p": 0.203496, "boundary_entropy": 0.083595, "bpb": 1.872591}, "tr": {"chunks": 3756.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.207479, "cps": 29.34375, "mean_p": 0.19283, "boundary_entropy": 0.123419, "bpb": 2.016912}, "fi": {"chunks": 4878.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.259316, "cps": 38.109375, "mean_p": 0.256807, "boundary_entropy": 0.094531, "bpb": 2.1138}}, "bpb": {"en": 1.87259, "tr": 2.01691, "fi": 2.1138}, "gini_cpc": 0.056476, "gini_cps": 0.074123, "bpic": 4.665461, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0240398195377187, "fi": 1.279887285510436}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 311.4102478027344, "n_util_obs": 1080, "util_count_spearman": 0.15047919503002713}, "t": 1834.6} |
| {"event": "eval", "step": 5200, "phase": 3, "bpb": {"en": 1.8875, "tr": 2.02468, "fi": 2.12233}, "bytes_per_chunk": {"en": 4.725, "tr": 5.367, "fi": 4.138}, "bpb_mean": 2.0115, "frozen_inventory": true, "t": 1836.1} |
| {"event": "train", "step": 5200, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35235, "bpb": 1.95102, "chunk_loss": 0.030795, "F": 0.22055, "G": 0.22224, "chunks": 225.84, "bytes_per_chunk": 4.534, "boundary_entropy": 0.08994, "router_noise": 0.0, "grad_norm": 0.6819, "lr": 4.29e-05, "router_trainable": true, "sec_per_step": 0.423, "probe_sec": 0.0728, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06013, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.0987, "n_shadow": 0, "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": 0.938, "trunc_abs_frac": 0.67606, "mean_delta_full": -0.04618, "read_hit_rate": {"skel": 0.77086}, "mean_gate": {"skel": 0.18376}, "n_reads": 4232, "n_cand": 5490, "read_suppressed_frac": 0.0992, "dropout_reads_logged": 4232, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 311.4753723144531, "n_util_obs": 1080, "util_count_spearman": 0.15044692868732157}, "probe_overhead_sec_total": 286.742, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.019394, "alpha": 0.03, "t": 1838.6} |
| {"event": "consolidate", "step": 5250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1853.7} |
| {"event": "train", "step": 5250, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38084, "bpb": 1.99213, "chunk_loss": 0.03066, "F": 0.21851, "G": 0.21883, "chunks": 223.75, "bytes_per_chunk": 4.577, "boundary_entropy": 0.08836, "router_noise": 0.0, "grad_norm": 0.7039, "lr": 4.14e-05, "router_trainable": true, "sec_per_step": 0.317, "probe_sec": 0.0552, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03546, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.14776, "n_shadow": 0, "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": 0.92284, "trunc_abs_frac": 0.86126, "mean_delta_full": 0.10046, "read_hit_rate": {"skel": 0.75543}, "mean_gate": {"skel": 0.1866}, "n_reads": 4034, "n_cand": 5340, "read_suppressed_frac": 0.1016, "dropout_reads_logged": 4034, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 314.54364013671875, "n_util_obs": 1086, "util_count_spearman": 0.16038582499711956}, "probe_overhead_sec_total": 289.401, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.017442, "alpha": 0.03, "t": 1854.4} |
| {"event": "consolidate", "step": 5300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1884.4} |
| {"event": "train", "step": 5300, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.37062, "bpb": 1.97739, "chunk_loss": 0.030491, "F": 0.20862, "G": 0.21083, "chunks": 213.62, "bytes_per_chunk": 4.793, "boundary_entropy": 0.08895, "router_noise": 0.0, "grad_norm": 0.6791, "lr": 3.99e-05, "router_trainable": true, "sec_per_step": 0.61, "probe_sec": 0.0599, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03934, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.10867, "n_shadow": 0, "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": 1.08474, "trunc_abs_frac": 2.93795, "mean_delta_full": 0.04106, "read_hit_rate": {"skel": 0.75569}, "mean_gate": {"skel": 0.181}, "n_reads": 3981, "n_cand": 5268, "read_suppressed_frac": 0.1035, "dropout_reads_logged": 3981, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 317.60784912109375, "n_util_obs": 1101, "util_count_spearman": 0.15947869514109936}, "probe_overhead_sec_total": 292.682, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.018081, "alpha": 0.03, "t": 1884.9} |
| {"event": "consolidate", "step": 5350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1917.0} |
| {"event": "train", "step": 5350, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.33173, "bpb": 1.92127, "chunk_loss": 0.030788, "F": 0.21597, "G": 0.21681, "chunks": 221.16, "bytes_per_chunk": 4.63, "boundary_entropy": 0.09038, "router_noise": 0.0, "grad_norm": 0.7282, "lr": 3.86e-05, "router_trainable": true, "sec_per_step": 0.648, "probe_sec": 0.0353, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02597, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.28857, "n_shadow": 0, "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": 0.41325, "trunc_abs_frac": 0.71428, "mean_delta_full": 0.02561, "read_hit_rate": {"skel": 0.74594}, "mean_gate": {"skel": 0.19033}, "n_reads": 3949, "n_cand": 5294, "read_suppressed_frac": 0.0993, "dropout_reads_logged": 3949, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 320.6544494628906, "n_util_obs": 1109, "util_count_spearman": 0.1625475485490112}, "probe_overhead_sec_total": 295.942, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.020537, "alpha": 0.03, "t": 1917.3} |
| {"event": "consolidate", "step": 5400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1947.6} |
| {"event": "probe", "step": 5400, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3471.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.199448, "cps": 27.117188, "mean_p": 0.202656, "boundary_entropy": 0.083172, "bpb": 1.869673}, "tr": {"chunks": 3734.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.206264, "cps": 29.171875, "mean_p": 0.192071, "boundary_entropy": 0.121385, "bpb": 2.010538}, "fi": {"chunks": 4869.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.258838, "cps": 38.039062, "mean_p": 0.256269, "boundary_entropy": 0.09316, "bpb": 2.104122}}, "bpb": {"en": 1.86967, "tr": 2.01054, "fi": 2.10412}, "gini_cpc": 0.059579, "gini_cps": 0.077191, "bpic": 4.698691, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0341731752870387, "fi": 1.2977690052166557}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 323.6392822265625, "n_util_obs": 1117, "util_count_spearman": 0.1648970283791394}, "t": 1948.6} |
| {"event": "eval", "step": 5400, "phase": 3, "bpb": {"en": 1.88376, "tr": 2.01858, "fi": 2.11299}, "bytes_per_chunk": {"en": 4.729, "tr": 5.363, "fi": 4.145}, "bpb_mean": 2.00511, "frozen_inventory": true, "t": 1949.8} |
| {"event": "train", "step": 5400, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.25708, "bpb": 1.81359, "chunk_loss": 0.030492, "F": 0.20657, "G": 0.20707, "chunks": 211.53, "bytes_per_chunk": 4.841, "boundary_entropy": 0.08595, "router_noise": 0.0, "grad_norm": 0.6559, "lr": 3.73e-05, "router_trainable": true, "sec_per_step": 0.691, "probe_sec": 0.0772, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04848, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.06129, "n_shadow": 0, "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": 0.65043, "trunc_abs_frac": 0.57426, "mean_delta_full": 0.27657, "read_hit_rate": {"skel": 0.76159}, "mean_gate": {"skel": 0.18703}, "n_reads": 3942, "n_cand": 5176, "read_suppressed_frac": 0.0982, "dropout_reads_logged": 3942, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 323.6998291015625, "n_util_obs": 1117, "util_count_spearman": 0.16459269465106555}, "probe_overhead_sec_total": 299.036, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.018534, "alpha": 0.03, "t": 1951.9} |
| {"event": "consolidate", "step": 5450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1973.9} |
| {"event": "train", "step": 5450, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.30229, "bpb": 1.87881, "chunk_loss": 0.030579, "F": 0.21423, "G": 0.21592, "chunks": 219.38, "bytes_per_chunk": 4.668, "boundary_entropy": 0.08289, "router_noise": 0.0, "grad_norm": 0.6514, "lr": 3.62e-05, "router_trainable": true, "sec_per_step": 0.447, "probe_sec": 0.102, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09416, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.11463, "n_shadow": 0, "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": 0.49158, "trunc_abs_frac": 2.40455, "mean_delta_full": 0.10384, "read_hit_rate": {"skel": 0.7711}, "mean_gate": {"skel": 0.17666}, "n_reads": 4194, "n_cand": 5439, "read_suppressed_frac": 0.0973, "dropout_reads_logged": 4194, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 326.77117919921875, "n_util_obs": 1124, "util_count_spearman": 0.17156065392228811}, "probe_overhead_sec_total": 302.737, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.018719, "alpha": 0.03, "t": 1974.2} |
| {"event": "consolidate", "step": 5500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1987.7} |
| {"event": "train", "step": 5500, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.3207, "bpb": 1.90536, "chunk_loss": 0.030579, "F": 0.21393, "G": 0.21408, "chunks": 219.06, "bytes_per_chunk": 4.674, "boundary_entropy": 0.08542, "router_noise": 0.0, "grad_norm": 0.6778, "lr": 3.51e-05, "router_trainable": true, "sec_per_step": 0.274, "probe_sec": 0.0382, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02915, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.10592, "n_shadow": 0, "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": 0.64601, "trunc_abs_frac": 4.98579, "mean_delta_full": 0.04938, "read_hit_rate": {"skel": 0.75791}, "mean_gate": {"skel": 0.18462}, "n_reads": 3998, "n_cand": 5275, "read_suppressed_frac": 0.0888, "dropout_reads_logged": 3998, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 329.84576416015625, "n_util_obs": 1128, "util_count_spearman": 0.17599792164117917}, "probe_overhead_sec_total": 305.941, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.021547, "alpha": 0.03, "t": 1987.9} |
| {"event": "consolidate", "step": 5550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2000.9} |
| {"event": "train", "step": 5550, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.30184, "bpb": 1.87815, "chunk_loss": 0.030659, "F": 0.21976, "G": 0.21972, "chunks": 225.03, "bytes_per_chunk": 4.55, "boundary_entropy": 0.07999, "router_noise": 0.0, "grad_norm": 0.6574, "lr": 3.41e-05, "router_trainable": true, "sec_per_step": 0.268, "probe_sec": 0.0347, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02706, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.07379, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -0.07451, "trunc_abs_frac": 0.61752, "mean_delta_full": -0.1521, "read_hit_rate": {"skel": 0.77006}, "mean_gate": {"skel": 0.1733}, "n_reads": 4193, "n_cand": 5445, "read_suppressed_frac": 0.1064, "dropout_reads_logged": 4193, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 332.90216064453125, "n_util_obs": 1137, "util_count_spearman": 0.18729435587684198}, "probe_overhead_sec_total": 308.65, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.018871, "alpha": 0.03, "t": 2001.3} |
| {"event": "consolidate", "step": 5600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2014.2} |
| {"event": "probe", "step": 5600, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3477.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.199793, "cps": 27.164062, "mean_p": 0.202757, "boundary_entropy": 0.082543, "bpb": 1.862936}, "tr": {"chunks": 3714.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.205159, "cps": 29.015625, "mean_p": 0.191124, "boundary_entropy": 0.12259, "bpb": 2.003743}, "fi": {"chunks": 4864.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.258572, "cps": 38.0, "mean_p": 0.256056, "boundary_entropy": 0.092957, "bpb": 2.095869}}, "bpb": {"en": 1.86294, "tr": 2.00374, "fi": 2.09587}, "gini_cpc": 0.059057, "gini_cps": 0.076704, "bpic": 4.706097, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.026858913978811, "fi": 1.2941991562313995}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 335.89971923828125, "n_util_obs": 1150, "util_count_spearman": 0.18265578842443003}, "t": 2015.2} |
| {"event": "eval", "step": 5600, "phase": 3, "bpb": {"en": 1.87834, "tr": 2.01166, "fi": 2.10548}, "bytes_per_chunk": {"en": 4.724, "tr": 5.401, "fi": 4.148}, "bpb_mean": 1.99849, "frozen_inventory": true, "t": 2016.4} |
| {"event": "train", "step": 5600, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32856, "bpb": 1.91671, "chunk_loss": 0.030643, "F": 0.21619, "G": 0.21706, "chunks": 221.38, "bytes_per_chunk": 4.626, "boundary_entropy": 0.08951, "router_noise": 0.0, "grad_norm": 0.6794, "lr": 3.33e-05, "router_trainable": true, "sec_per_step": 0.341, "probe_sec": 0.0403, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02108, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.28047, "n_shadow": 0, "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": 1.04392, "trunc_abs_frac": 1.50389, "mean_delta_full": 0.42407, "read_hit_rate": {"skel": 0.76086}, "mean_gate": {"skel": 0.18785}, "n_reads": 4028, "n_cand": 5294, "read_suppressed_frac": 0.0961, "dropout_reads_logged": 4028, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 335.9613342285156, "n_util_obs": 1150, "util_count_spearman": 0.18286985926977212}, "probe_overhead_sec_total": 311.412, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.019497, "alpha": 0.03, "t": 2018.3} |
| {"event": "consolidate", "step": 5650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2031.2} |
| {"event": "train", "step": 5650, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.31584, "bpb": 1.89836, "chunk_loss": 0.03055, "F": 0.21228, "G": 0.21368, "chunks": 217.38, "bytes_per_chunk": 4.711, "boundary_entropy": 0.0849, "router_noise": 0.0, "grad_norm": 0.6634, "lr": 3.25e-05, "router_trainable": true, "sec_per_step": 0.263, "probe_sec": 0.0638, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0524, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.12381, "n_shadow": 0, "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": 0.86194, "trunc_abs_frac": 0.5509, "mean_delta_full": 0.45544, "read_hit_rate": {"skel": 0.76058}, "mean_gate": {"skel": 0.17954}, "n_reads": 4009, "n_cand": 5271, "read_suppressed_frac": 0.1, "dropout_reads_logged": 4009, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 338.9828186035156, "n_util_obs": 1161, "util_count_spearman": 0.1904729648742437}, "probe_overhead_sec_total": 314.365, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.020839, "alpha": 0.03, "t": 2031.5} |
| {"event": "consolidate", "step": 5700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2044.6} |
| {"event": "train", "step": 5700, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.39334, "bpb": 2.01017, "chunk_loss": 0.030669, "F": 0.21762, "G": 0.21846, "chunks": 222.84, "bytes_per_chunk": 4.595, "boundary_entropy": 0.08707, "router_noise": 0.0, "grad_norm": 0.7025, "lr": 3.18e-05, "router_trainable": true, "sec_per_step": 0.269, "probe_sec": 0.0793, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07152, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.3822, "n_shadow": 0, "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": 1.0453, "trunc_abs_frac": 2.17205, "mean_delta_full": 0.6864, "read_hit_rate": {"skel": 0.76055}, "mean_gate": {"skel": 0.18475}, "n_reads": 4126, "n_cand": 5425, "read_suppressed_frac": 0.1028, "dropout_reads_logged": 4126, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 342.0687255859375, "n_util_obs": 1177, "util_count_spearman": 0.193366296585965}, "probe_overhead_sec_total": 316.773, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.019661, "alpha": 0.03, "t": 2044.9} |
| {"event": "consolidate", "step": 5750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2057.4} |
| {"event": "train", "step": 5750, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.37599, "bpb": 1.98513, "chunk_loss": 0.030656, "F": 0.21512, "G": 0.21561, "chunks": 220.28, "bytes_per_chunk": 4.649, "boundary_entropy": 0.08504, "router_noise": 0.0, "grad_norm": 0.7051, "lr": 3.13e-05, "router_trainable": true, "sec_per_step": 0.253, "probe_sec": 0.0347, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02388, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.05776, "n_shadow": 0, "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": 0.69317, "trunc_abs_frac": 1.48841, "mean_delta_full": 0.37954, "read_hit_rate": {"skel": 0.75831}, "mean_gate": {"skel": 0.18423}, "n_reads": 4013, "n_cand": 5292, "read_suppressed_frac": 0.0959, "dropout_reads_logged": 4013, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 345.13720703125, "n_util_obs": 1184, "util_count_spearman": 0.198934445133521}, "probe_overhead_sec_total": 319.513, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.017371, "alpha": 0.03, "t": 2057.6} |
| {"event": "consolidate", "step": 5800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2070.3} |
| {"event": "probe", "step": 5800, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3518.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.202149, "cps": 27.484375, "mean_p": 0.203516, "boundary_entropy": 0.082414, "bpb": 1.858035}, "tr": {"chunks": 3745.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.206872, "cps": 29.257812, "mean_p": 0.192497, "boundary_entropy": 0.119936, "bpb": 1.998594}, "fi": {"chunks": 4876.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.25921, "cps": 38.09375, "mean_p": 0.25629, "boundary_entropy": 0.093024, "bpb": 2.089587}}, "bpb": {"en": 1.85804, "tr": 1.99859, "fi": 2.08959}, "gini_cpc": 0.056927, "gini_cps": 0.074581, "bpic": 4.673532, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0233626343250863, "fi": 1.2822718215900009}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 348.14410400390625, "n_util_obs": 1195, "util_count_spearman": 0.1988773834631165}, "t": 2071.2} |
| {"event": "eval", "step": 5800, "phase": 3, "bpb": {"en": 1.87355, "tr": 2.0061, "fi": 2.09892}, "bytes_per_chunk": {"en": 4.725, "tr": 5.363, "fi": 4.135}, "bpb_mean": 1.99286, "frozen_inventory": true, "t": 2072.4} |
| {"event": "train", "step": 5800, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.29096, "bpb": 1.86246, "chunk_loss": 0.030503, "F": 0.21054, "G": 0.21125, "chunks": 215.59, "bytes_per_chunk": 4.75, "boundary_entropy": 0.08465, "router_noise": 0.0, "grad_norm": 0.6817, "lr": 3.08e-05, "router_trainable": true, "sec_per_step": 0.335, "probe_sec": 0.1005, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08975, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.30088, "n_shadow": 0, "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": 0.68572, "trunc_abs_frac": 1.32439, "mean_delta_full": 0.13199, "read_hit_rate": {"skel": 0.77576}, "mean_gate": {"skel": 0.1788}, "n_reads": 4141, "n_cand": 5338, "read_suppressed_frac": 0.1055, "dropout_reads_logged": 4141, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 348.2080078125, "n_util_obs": 1196, "util_count_spearman": 0.1984137238314424}, "probe_overhead_sec_total": 321.977, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.019653, "alpha": 0.03, "t": 2074.4} |
| {"event": "consolidate", "step": 5850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2087.3} |
| {"event": "train", "step": 5850, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.34692, "bpb": 1.94319, "chunk_loss": 0.030447, "F": 0.20596, "G": 0.20741, "chunks": 210.91, "bytes_per_chunk": 4.855, "boundary_entropy": 0.08419, "router_noise": 0.0, "grad_norm": 0.7041, "lr": 3.05e-05, "router_trainable": true, "sec_per_step": 0.264, "probe_sec": 0.0376, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02865, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.00404, "n_shadow": 0, "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": 0.94673, "trunc_abs_frac": 1.00278, "mean_delta_full": 0.07524, "read_hit_rate": {"skel": 0.74268}, "mean_gate": {"skel": 0.17671}, "n_reads": 3853, "n_cand": 5188, "read_suppressed_frac": 0.103, "dropout_reads_logged": 3853, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 351.26129150390625, "n_util_obs": 1207, "util_count_spearman": 0.20092341904282973}, "probe_overhead_sec_total": 324.655, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.020845, "alpha": 0.03, "t": 2087.6} |
| {"event": "consolidate", "step": 5900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2099.7} |
| {"event": "train", "step": 5900, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.37684, "bpb": 1.98636, "chunk_loss": 0.030639, "F": 0.21695, "G": 0.21824, "chunks": 222.16, "bytes_per_chunk": 4.609, "boundary_entropy": 0.08719, "router_noise": 0.0, "grad_norm": 0.7298, "lr": 3.02e-05, "router_trainable": true, "sec_per_step": 0.25, "probe_sec": 0.11, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09241, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.21133, "n_shadow": 0, "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": -0.51624, "trunc_abs_frac": 2.84242, "mean_delta_full": -0.16534, "read_hit_rate": {"skel": 0.76057}, "mean_gate": {"skel": 0.17957}, "n_reads": 4120, "n_cand": 5417, "read_suppressed_frac": 0.0961, "dropout_reads_logged": 4120, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 354.32427978515625, "n_util_obs": 1220, "util_count_spearman": 0.2043022882105897}, "probe_overhead_sec_total": 327.069, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.020039, "alpha": 0.03, "t": 2100.1} |
| {"event": "consolidate", "step": 5950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2112.0} |
| {"event": "train", "step": 5950, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.30742, "bpb": 1.88621, "chunk_loss": 0.030525, "F": 0.20953, "G": 0.21082, "chunks": 214.56, "bytes_per_chunk": 4.773, "boundary_entropy": 0.08839, "router_noise": 0.0, "grad_norm": 0.7566, "lr": 3.01e-05, "router_trainable": true, "sec_per_step": 0.245, "probe_sec": 0.0304, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02112, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.08579, "n_shadow": 0, "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": 0.51995, "trunc_abs_frac": 0.92696, "mean_delta_full": -0.21517, "read_hit_rate": {"skel": 0.7547}, "mean_gate": {"skel": 0.18308}, "n_reads": 3978, "n_cand": 5271, "read_suppressed_frac": 0.0882, "dropout_reads_logged": 3978, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 357.38775634765625, "n_util_obs": 1228, "util_count_spearman": 0.2060909805277281}, "probe_overhead_sec_total": 330.162, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.020447, "alpha": 0.03, "t": 2112.3} |
| {"event": "consolidate", "step": 6000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 2126.9} |
| {"event": "train", "step": 5999, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32619, "bpb": 1.91329, "chunk_loss": 0.030714, "F": 0.21826, "G": 0.21908, "chunks": 223.5, "bytes_per_chunk": 4.582, "boundary_entropy": 0.08855, "router_noise": 0.0, "grad_norm": 0.7263, "lr": 3e-05, "router_trainable": true, "sec_per_step": 0.291, "probe_sec": 0.0848, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07823, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.25374, "n_shadow": 0, "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": 0.98889, "trunc_abs_frac": 0.60532, "mean_delta_full": 0.28625, "read_hit_rate": {"skel": 0.78175}, "mean_gate": {"skel": 0.18254}, "n_reads": 4309, "n_cand": 5512, "read_suppressed_frac": 0.0956, "dropout_reads_logged": 4309, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 360.4039306640625, "n_util_obs": 1234, "util_count_spearman": 0.2093735972320026}, "probe_overhead_sec_total": 332.778, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.020778, "alpha": 0.03, "t": 2126.9} |
| {"event": "probe", "step": 6000, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3524.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.202494, "cps": 27.53125, "mean_p": 0.203231, "boundary_entropy": 0.081521, "bpb": 1.855682}, "tr": {"chunks": 3732.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.206154, "cps": 29.15625, "mean_p": 0.191252, "boundary_entropy": 0.121463, "bpb": 1.991533}, "fi": {"chunks": 4861.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.258413, "cps": 37.976562, "mean_p": 0.255892, "boundary_entropy": 0.091662, "bpb": 2.084612}}, "bpb": {"en": 1.85568, "tr": 1.99153, "fi": 2.08461}, "gini_cpc": 0.055886, "gini_cps": 0.073561, "bpic": 4.682017, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0180739008710593, "fi": 1.2761506854944136}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 360.4039306640625, "n_util_obs": 1234, "util_count_spearman": 0.2093735972320026}, "t": 2127.8} |
| {"event": "eval", "step": 6000, "phase": 3, "bpb": {"en": 1.87103, "tr": 2.00003, "fi": 2.0933}, "bytes_per_chunk": {"en": 4.724, "tr": 5.378, "fi": 4.133}, "bpb_mean": 1.98812, "frozen_inventory": true, "t": 2128.9} |
| {"event": "reps_dumped", "dirs": ["/local/scratch/sas245/beetle-hnet/runs/ali_mem-skel_s0/probe/reps/th0.30", "/local/scratch/sas245/beetle-hnet/runs/ali_mem-skel_s0/probe/reps/th0.40", "/local/scratch/sas245/beetle-hnet/runs/ali_mem-skel_s0/probe/reps/th0.50", "/local/scratch/sas245/beetle-hnet/runs/ali_mem-skel_s0/probe/reps/th0.60", "/local/scratch/sas245/beetle-hnet/runs/ali_mem-skel_s0/probe/reps/th0.70"], "t": 2131.4} |
| {"event": "final", "run": "ali_mem-skel_s0", "arch": "ali", "objective": "ratio", "condition": "A", "step": 6000, "params": 22785088, "seed": 0, "per_lang": {"en": {"cpc": 0.20249382290409698, "cps": 27.53125, "bpb": 1.8757852886849464, "bytes_per_chunk": 4.684656349404911, "reads_per_chunk": 0.5621358876300082}, "tr": {"cpc": 0.20615367618626748, "cps": 29.15625, "bpb": 1.9921305415955628, "bytes_per_chunk": 5.3693276665471865, "reads_per_chunk": 0.5658763889600082}, "fi": {"cpc": 0.25841263090744776, "cps": 37.9765625, "bpb": 2.1117318775414704, "bytes_per_chunk": 4.117358798768612, "reads_per_chunk": 0.6686011811270968}}, "bpb_mean": 1.9932159026073266, "gini_cpc": 0.05588582446936141, "gini_cps": 0.07356056229539765, "bpic": 4.682017000907815, "mean_boundary_entropy": 0.08854883164167404, "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": 332.7783329486847, "probe_overhead_seq_frac": 0.499859375, "wall_sec": 2134.0191237926483, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 0, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 65536, "staged_skel": 0, "mean_count": 360.4039306640625, "n_util_obs": 1234, "util_count_spearman": 0.2093735972320026}, "ali_params": 268096, "ali_config": {"enabled": true, "budget_B": 65536, "staging_capacity": 262144, "policy": "frequency", "value_mode": "learned", "candidate_types": ["skel"], "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": 0.2093735972320026, "paired_vs_dropout": {"n": 2664, "pearson": 0.06890196174492981, "spearman": 0.08033049316911257, "mean_abs_discrepancy": 1.0446057319641113, "mean_paired": 0.07614783197641373, "mean_dropout": 0.05991087481379509}, "predictor_calibration": {"n": 6359, "all": {"n": 6359, "pearson": 0.13304607699358897, "spearman": 0.156916545929357, "mae": 0.13441164791584015, "reliability": [{"pred": 0.0023132613860070705, "obs": 0.03271646797657013, "n": 636}, {"pred": 0.014444655738770962, "obs": 0.024504762142896652, "n": 636}, {"pred": 0.020898422226309776, "obs": 0.04482361674308777, "n": 636}, {"pred": 0.02605515345931053, "obs": 0.037269365042448044, "n": 636}, {"pred": 0.030572518706321716, "obs": 0.04951855167746544, "n": 635}, {"pred": 0.03494088724255562, "obs": 0.05362846702337265, "n": 636}, {"pred": 0.03953668102622032, "obs": 0.054467253386974335, "n": 636}, {"pred": 0.04513830691576004, "obs": 0.08542212843894958, "n": 636}, {"pred": 0.053437087684869766, "obs": 0.0719672292470932, "n": 636}, {"pred": 0.08540016412734985, "obs": 0.13967818021774292, "n": 636}]}, "admitted": {"n": 6359, "pearson": 0.13304607699358897, "spearman": 0.156916545929357, "mae": 0.13441164791584015, "reliability": [{"pred": 0.0023132613860070705, "obs": 0.03271646797657013, "n": 636}, {"pred": 0.014444655738770962, "obs": 0.024504762142896652, "n": 636}, {"pred": 0.020898422226309776, "obs": 0.04482361674308777, "n": 636}, {"pred": 0.02605515345931053, "obs": 0.037269365042448044, "n": 636}, {"pred": 0.030572518706321716, "obs": 0.04951855167746544, "n": 635}, {"pred": 0.03494088724255562, "obs": 0.05362846702337265, "n": 636}, {"pred": 0.03953668102622032, "obs": 0.054467253386974335, "n": 636}, {"pred": 0.04513830691576004, "obs": 0.08542212843894958, "n": 636}, {"pred": 0.053437087684869766, "obs": 0.0719672292470932, "n": 636}, {"pred": 0.08540016412734985, "obs": 0.13967818021774292, "n": 636}]}, "staged": {"n": 0, "pearson": NaN, "spearman": NaN, "mae": NaN}}, "t": 2134.1} |
| {"event": "hub_done", "repo": "AdaptiveChunking/ali_mem-skel_s0", "pushed": [0, 1, 2, 4, 6, 8, 16, 32, 64, 128, 256, 512, 1000, 1024, 2048, 3000, 4096, 6000], "failed": [], "ok": true, "t": 2148.0} |
|
|
