| {"event": "hub", "repo": "AdaptiveChunking/ali_mem-split_s2", "keep_steps": [0, 1, 2, 4, 6, 8, 16, 32, 64, 128, 256, 512, 1000, 1024, 2048, 3000, 4096, 6000], "t": 2.1} |
| {"event": "start", "run": "ali_mem-split_s2", "arch": "ali", "objective": "ratio", "condition": "A", "params": 22785088, "langs": ["en", "tr", "fi"], "premium": {"en": 1.0, "tr": 1.12, "fi": 1.104}, "schedule": {"tau1": 0.4, "tau2": 0.7, "p1_objective": "global", "p2_objective": "global", "p3_objective": "global", "freeze_in_p3": false, "freeze_scope": "router"}, "corpus": "/local/scratch/sas245/beetle-hnet/scratch/data/tier1_corpus", "premium_kind": "flores_per_sent", "doc_filtered": {}, "weights": {"en": 0.74, "tr": 0.1536, "fi": 0.1064}, "device": "cuda", "ali_params": 268096, "ali_config": {"enabled": true, "budget_B": 65536, "staging_capacity": 262144, "policy": "frequency", "value_mode": "learned", "candidate_types": ["split"], "gate": "vector", "retrieval": true, "memory_scope": "persistent", "read_dropout": 0.1, "write_rule": "mean", "eta_min": 0.0, "consolidate_every": 50, "probe_seqs": 2, "probe_reps": 8, "horizon": 32, "predictor": "mlp", "lambda_util": 0.1, "max_span": 64, "byte_classes_path": "/local/scratch/sas245/beetle-hnet/scratch/data/byte_classes.json", "gate_hidden": 0, "type_dim": 16, "gate_bias_init": -2.0}, "t": 2.1} |
| {"event": "probe_schedule", "n": 45, "steps": [0, 1, 2, 4, 6, 8, 16, 32, 64, 128, 200, 256, 400, 512, 600, 800, 1000, 1024, 1200, 1400, 1600, 1800, 2000, 2048, 2200, 2400, 2600, 2800, 3000, 3200, 3400, 3600, 3800, 4000, 4096, 4200, 4400, 4600, 4800, 5000, 5200, 5400, 5600, 5800, 6000], "t": 2.1} |
| {"event": "probe", "step": 0, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 10933.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.628225, "cps": 85.414062, "mean_p": 0.513249, "boundary_entropy": 0.686295, "bpb": 8.12598}, "tr": {"chunks": 8187.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.452245, "cps": 63.960938, "mean_p": 0.498002, "boundary_entropy": 0.686826, "bpb": 8.097536}, "fi": {"chunks": 13540.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.719792, "cps": 105.78125, "mean_p": 0.517169, "boundary_entropy": 0.686661, "bpb": 8.065117}}, "bpb": {"en": 8.12598, "tr": 8.09754, "fi": 8.06512}, "gini_cpc": 0.099077, "gini_cps": 0.109267, "bpic": 1.737048, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.7198781817505053, "fi": 1.1457544508645294}, "inventory": {"size": 0, "admitted": 0, "staged": 0, "budget_B": 65536, "staging_capacity": 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": 4.0} |
| {"event": "train", "step": 0, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 5.62735, "bpb": 8.11854, "chunk_loss": 0.062086, "F": 0.59833, "G": 0.50836, "chunks": 612.69, "bytes_per_chunk": 1.671, "boundary_entropy": 0.69105, "router_noise": 0.0, "grad_norm": 11.4936, "lr": 1e-06, "router_trainable": true, "sec_per_step": 0.015, "probe_sec": 0.0159, "probe_extra_seqs": 0, "n_rows": 0, "extra_fwd_seqs": 0, "wall_s": 0.0, "overhead_frac_est": 0.0, "chunk_mismatch": 0, "n_loo": 0, "mean_delta_loo": NaN, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "read_hit_rate": {"split": 0.0}, "mean_gate": {"split": null}, "n_reads": 0, "n_cand": 15066, "read_suppressed_frac": 0.0, "dropout_reads_logged": 0, "inventory": {"size": 1198, "admitted": 1198, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 1198, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 12.575959205627441, "n_util_obs": 0, "util_count_spearman": NaN}, "probe_overhead_sec_total": 0.016, "probe_overhead_seq_frac": 0.0, "util_loss": -0.0, "alpha": 0.03, "t": 5.0} |
| {"event": "probe", "step": 1, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 10907.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.626731, "cps": 85.210938, "mean_p": 0.513143, "boundary_entropy": 0.686304, "bpb": 8.104946}, "tr": {"chunks": 8206.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.453295, "cps": 64.109375, "mean_p": 0.498082, "boundary_entropy": 0.686828, "bpb": 8.08811}, "fi": {"chunks": 13506.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.717984, "cps": 105.515625, "mean_p": 0.517039, "boundary_entropy": 0.686672, "bpb": 8.051641}}, "bpb": {"en": 8.10495, "tr": 8.08811, "fi": 8.05164}, "gini_cpc": 0.098141, "gini_cps": 0.108321, "bpic": 1.739232, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.7232688618519676, "fi": 1.1456017516500347}, "inventory": {"size": 1198, "admitted": 1198, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 1198, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 12.575959205627441, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 6.2} |
| {"event": "probe", "step": 2, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 10881.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.625237, "cps": 85.007812, "mean_p": 0.51299, "boundary_entropy": 0.686317, "bpb": 8.042515}, "tr": {"chunks": 8249.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.45567, "cps": 64.445312, "mean_p": 0.498257, "boundary_entropy": 0.686832, "bpb": 8.044143}, "fi": {"chunks": 13469.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.716017, "cps": 105.226562, "mean_p": 0.516808, "boundary_entropy": 0.686691, "bpb": 7.99291}}, "bpb": {"en": 8.04251, "tr": 8.04414, "fi": 7.99291}, "gini_cpc": 0.09659, "gini_cps": 0.106752, "bpic": 1.740299, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.7287961371547589, "fi": 1.1451932495779524}, "inventory": {"size": 2214, "admitted": 2214, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 2214, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 13.202800750732422, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 8.0} |
| {"event": "probe", "step": 4, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 10720.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.615986, "cps": 83.75, "mean_p": 0.512126, "boundary_entropy": 0.686378, "bpb": 7.831973}, "tr": {"chunks": 8393.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.463625, "cps": 65.570312, "mean_p": 0.498697, "boundary_entropy": 0.686846, "bpb": 7.890018}, "fi": {"chunks": 13214.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.702461, "cps": 103.234375, "mean_p": 0.515687, "boundary_entropy": 0.686774, "bpb": 7.79202}}, "bpb": {"en": 7.83197, "tr": 7.89002, "fi": 7.79202}, "gini_cpc": 0.089348, "gini_cps": 0.099422, "bpic": 1.754942, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.7526550961290328, "fi": 1.1403856766087166}, "inventory": {"size": 3772, "admitted": 3772, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 3772, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 15.497614860534668, "n_util_obs": 0, "util_count_spearman": NaN}, "t": 10.1} |
| {"event": "probe", "step": 6, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 10329.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.593518, "cps": 80.695312, "mean_p": 0.510407, "boundary_entropy": 0.686481, "bpb": 7.553233}, "tr": {"chunks": 8517.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.470475, "cps": 66.539062, "mean_p": 0.499227, "boundary_entropy": 0.686868, "bpb": 7.681682}, "fi": {"chunks": 12728.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.676625, "cps": 99.4375, "mean_p": 0.513619, "boundary_entropy": 0.686908, "bpb": 7.535099}}, "bpb": {"en": 7.55323, "tr": 7.68168, "fi": 7.5351}, "gini_cpc": 0.078957, "gini_cps": 0.088913, "bpic": 1.796795, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.7926873700366488, "fi": 1.1400243456961001}, "inventory": {"size": 4762, "admitted": 4762, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 4762, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 18.730995178222656, "n_util_obs": 1, "util_count_spearman": NaN}, "t": 12.8} |
| {"event": "probe", "step": 8, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 9666.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.555421, "cps": 75.515625, "mean_p": 0.507386, "boundary_entropy": 0.686612, "bpb": 7.298781}, "tr": {"chunks": 8645.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.477545, "cps": 67.539062, "mean_p": 0.499629, "boundary_entropy": 0.6869, "bpb": 7.479541}, "fi": {"chunks": 11845.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.629685, "cps": 92.539062, "mean_p": 0.510227, "boundary_entropy": 0.687082, "bpb": 7.306999}}, "bpb": {"en": 7.29878, "tr": 7.47954, "fi": 7.307}, "gini_cpc": 0.061003, "gini_cps": 0.070743, "bpic": 1.881284, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.8597887843179451, "fi": 1.1337061720922763}, "inventory": {"size": 6051, "admitted": 6051, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 6051, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 19.833911895751953, "n_util_obs": 2, "util_count_spearman": NaN}, "t": 15.3} |
| {"event": "probe", "step": 16, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3787.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.217606, "cps": 29.585938, "mean_p": 0.484704, "boundary_entropy": 0.686187, "bpb": 6.725295}, "tr": {"chunks": 7338.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.405347, "cps": 57.328125, "mean_p": 0.494671, "boundary_entropy": 0.686998, "bpb": 7.013836}, "fi": {"chunks": 4538.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.241242, "cps": 35.453125, "mean_p": 0.486334, "boundary_entropy": 0.686885, "bpb": 6.786461}}, "bpb": {"en": 6.7253, "tr": 7.01384, "fi": 6.78646}, "gini_cpc": 0.144829, "gini_cps": 0.151142, "bpic": 3.622039, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.8627560005234256, "fi": 1.1086167172326693}, "inventory": {"size": 17584, "admitted": 17584, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 17584, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 12.563864707946777, "n_util_obs": 7, "util_count_spearman": 0.14285714285714285}, "t": 18.7} |
| {"event": "probe", "step": 32, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 2584.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.14848, "cps": 20.1875, "mean_p": 0.486707, "boundary_entropy": 0.686948, "bpb": 6.305491}, "tr": {"chunks": 4613.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.25482, "cps": 36.039062, "mean_p": 0.490015, "boundary_entropy": 0.687549, "bpb": 6.652166}, "fi": {"chunks": 1350.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.071767, "cps": 10.546875, "mean_p": 0.478935, "boundary_entropy": 0.686925, "bpb": 6.421523}}, "bpb": {"en": 6.30549, "tr": 6.65217, "fi": 6.42152}, "gini_cpc": 0.256881, "gini_cps": 0.254514, "bpic": 6.637651, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.7161866291768004, "fi": 0.48334084381495385}, "inventory": {"size": 43723, "admitted": 43723, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 43723, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 7.918944358825684, "n_util_obs": 9, "util_count_spearman": 0.18333333333333332}, "t": 24.3} |
| {"event": "consolidate", "step": 50, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 29.5} |
| {"event": "train", "step": 50, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 4.21299, "bpb": 6.07805, "chunk_loss": 0.034245, "F": 0.22629, "G": 0.48861, "chunks": 231.72, "bytes_per_chunk": 4.419, "boundary_entropy": 0.69161, "router_noise": 0.0, "grad_norm": 2.4801, "lr": 5.1e-05, "router_trainable": true, "sec_per_step": 0.497, "probe_sec": 0.1118, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10419, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.00539, "n_shadow": 0, "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": 0.12273, "trunc_abs_frac": 1.6521, "mean_delta_full": -0.00895, "read_hit_rate": {"split": 0.8015}, "mean_gate": {"split": 0.11917}, "n_reads": 7393, "n_cand": 9224, "read_suppressed_frac": 0.1017, "dropout_reads_logged": 7393, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 7.1774444580078125, "n_util_obs": 14, "util_count_spearman": -0.12087912087912088}, "probe_overhead_sec_total": 3.441, "probe_overhead_seq_frac": 0.489, "util_loss": 0.083851, "alpha": 0.03, "t": 29.9} |
| {"event": "probe", "step": 64, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 4862.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.279377, "cps": 37.984375, "mean_p": 0.485571, "boundary_entropy": 0.685544, "bpb": 5.552157}, "tr": {"chunks": 5495.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.303541, "cps": 42.929688, "mean_p": 0.48695, "boundary_entropy": 0.686512, "bpb": 5.955099}, "fi": {"chunks": 4723.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.251076, "cps": 36.898438, "mean_p": 0.483984, "boundary_entropy": 0.68632, "bpb": 5.717752}}, "bpb": {"en": 5.55216, "tr": 5.9551, "fi": 5.71775}, "gini_cpc": 0.041938, "gini_cps": 0.034129, "bpic": 3.762069, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.086491444938696, "fi": 0.8987010059864803}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 8.47003173828125, "n_util_obs": 16, "util_count_spearman": -0.3088235294117647}, "t": 34.2} |
| {"event": "consolidate", "step": 100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 46.1} |
| {"event": "train", "step": 100, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 3.25118, "bpb": 4.69046, "chunk_loss": 0.029492, "F": 0.15826, "G": 0.45296, "chunks": 162.06, "bytes_per_chunk": 6.319, "boundary_entropy": 0.68221, "router_noise": 0.0, "grad_norm": 1.5429, "lr": 0.000101, "router_trainable": true, "sec_per_step": 0.332, "probe_sec": 0.0574, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04625, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.00796, "n_shadow": 0, "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": 0.08945, "trunc_abs_frac": 1.00778, "mean_delta_full": 0.10903, "read_hit_rate": {"split": 0.77625}, "mean_gate": {"split": 0.11934}, "n_reads": 6765, "n_cand": 8715, "read_suppressed_frac": 0.1013, "dropout_reads_logged": 6765, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 12.928787231445312, "n_util_obs": 22, "util_count_spearman": -0.25352907961603616}, "probe_overhead_sec_total": 6.434, "probe_overhead_seq_frac": 0.4944, "util_loss": 0.001188, "alpha": 0.03, "t": 46.5} |
| {"event": "probe", "step": 128, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3830.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.220077, "cps": 29.921875, "mean_p": 0.418279, "boundary_entropy": 0.646359, "bpb": 4.155607}, "tr": {"chunks": 4904.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.270894, "cps": 38.3125, "mean_p": 0.441968, "boundary_entropy": 0.661423, "bpb": 4.403069}, "fi": {"chunks": 3904.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.207538, "cps": 30.5, "mean_p": 0.431096, "boundary_entropy": 0.661179, "bpb": 4.346486}}, "bpb": {"en": 4.15561, "tr": 4.40307, "fi": 4.34649}, "gini_cpc": 0.060468, "gini_cps": 0.056655, "bpic": 4.489001, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.2309070420796344, "fi": 0.9430251423636495}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 16.755950927734375, "n_util_obs": 30, "util_count_spearman": -0.3032258064516129}, "t": 54.7} |
| {"event": "consolidate", "step": 150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 61.9} |
| {"event": "train", "step": 150, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.79474, "bpb": 4.03196, "chunk_loss": 0.031243, "F": 0.19266, "G": 0.38136, "chunks": 197.28, "bytes_per_chunk": 5.191, "boundary_entropy": 0.61786, "router_noise": 0.0, "grad_norm": 1.3952, "lr": 0.000151, "router_trainable": true, "sec_per_step": 0.314, "probe_sec": 0.0495, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0395, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.00809, "n_shadow": 0, "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": 0.07805, "trunc_abs_frac": 0.85046, "mean_delta_full": -0.07336, "read_hit_rate": {"split": 0.8351}, "mean_gate": {"split": 0.11977}, "n_reads": 7870, "n_cand": 9424, "read_suppressed_frac": 0.0997, "dropout_reads_logged": 7870, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 19.73956298828125, "n_util_obs": 36, "util_count_spearman": -0.2205920205920206}, "probe_overhead_sec_total": 9.715, "probe_overhead_seq_frac": 0.4963, "util_loss": 0.000376, "alpha": 0.03, "t": 62.2} |
| {"event": "consolidate", "step": 200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 75.2} |
| {"event": "probe", "step": 200, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3621.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.208068, "cps": 28.289062, "mean_p": 0.379678, "boundary_entropy": 0.585487, "bpb": 3.625383}, "tr": {"chunks": 4403.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.243219, "cps": 34.398438, "mean_p": 0.366743, "boundary_entropy": 0.583837, "bpb": 3.82766}, "fi": {"chunks": 3718.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.19765, "cps": 29.046875, "mean_p": 0.386026, "boundary_entropy": 0.619749, "bpb": 3.858102}}, "bpb": {"en": 3.62538, "tr": 3.82766, "fi": 3.8581}, "gini_cpc": 0.046814, "gini_cps": 0.044399, "bpic": 4.831545, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.1689440626524432, "fi": 0.9499332676411327}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 25.369842529296875, "n_util_obs": 48, "util_count_spearman": 0.041901867129830656}, "t": 76.1} |
| {"event": "eval", "step": 200, "phase": 1, "bpb": {"en": 3.63218, "tr": 3.8387, "fi": 3.85561}, "bytes_per_chunk": {"en": 5.294, "tr": 5.09, "fi": 5.856}, "bpb_mean": 3.7755, "frozen_inventory": true, "t": 77.6} |
| {"event": "train", "step": 200, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.52976, "bpb": 3.64967, "chunk_loss": 0.030859, "F": 0.18594, "G": 0.36671, "chunks": 190.41, "bytes_per_chunk": 5.378, "boundary_entropy": 0.58484, "router_noise": 0.0, "grad_norm": 1.2817, "lr": 0.000201, "router_trainable": true, "sec_per_step": 0.326, "probe_sec": 0.0395, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03006, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.02231, "n_shadow": 0, "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": 0.76522, "trunc_abs_frac": 0.7376, "mean_delta_full": 0.02539, "read_hit_rate": {"split": 0.82392}, "mean_gate": {"split": 0.12053}, "n_reads": 7735, "n_cand": 9388, "read_suppressed_frac": 0.0959, "dropout_reads_logged": 7735, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 25.4888916015625, "n_util_obs": 49, "util_count_spearman": 0.07428571428571429}, "probe_overhead_sec_total": 12.729, "probe_overhead_seq_frac": 0.4972, "util_loss": 0.000322, "alpha": 0.03, "t": 78.5} |
| {"event": "consolidate", "step": 250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 93.3} |
| {"event": "train", "step": 250, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.3903, "bpb": 3.44848, "chunk_loss": 0.031983, "F": 0.21777, "G": 0.34532, "chunks": 223.0, "bytes_per_chunk": 4.592, "boundary_entropy": 0.53212, "router_noise": 0.0, "grad_norm": 2.4167, "lr": 0.000251, "router_trainable": true, "sec_per_step": 0.302, "probe_sec": 0.0401, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02855, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.01544, "n_shadow": 0, "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": -0.0373, "trunc_abs_frac": 0.0373, "mean_delta_full": 0.03027, "read_hit_rate": {"split": 0.86877}, "mean_gate": {"split": 0.12354}, "n_reads": 9765, "n_cand": 11240, "read_suppressed_frac": 0.0954, "dropout_reads_logged": 9765, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 32.1575927734375, "n_util_obs": 59, "util_count_spearman": 0.06370543541788427}, "probe_overhead_sec_total": 16.858, "probe_overhead_seq_frac": 0.4978, "util_loss": 0.000363, "alpha": 0.03, "t": 93.6} |
| {"event": "probe", "step": 256, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3378.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.194104, "cps": 26.390625, "mean_p": 0.339773, "boundary_entropy": 0.526299, "bpb": 3.36143}, "tr": {"chunks": 4868.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.268906, "cps": 38.03125, "mean_p": 0.31675, "boundary_entropy": 0.50096, "bpb": 3.509813}, "fi": {"chunks": 4720.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.250917, "cps": 36.875, "mean_p": 0.356496, "boundary_entropy": 0.56831, "bpb": 3.659727}}, "bpb": {"en": 3.36143, "tr": 3.50981, "fi": 3.65973}, "gini_cpc": 0.06985, "gini_cps": 0.076611, "bpic": 4.375443, "worst_lang": "tr", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.3853658986484385, "fi": 1.2926905981563073}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 32.88102722167969, "n_util_obs": 59, "util_count_spearman": 0.04915254237288136}, "t": 96.3} |
| {"event": "consolidate", "step": 300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 110.3} |
| {"event": "train", "step": 300, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.27953, "bpb": 3.28867, "chunk_loss": 0.030706, "F": 0.1889, "G": 0.30094, "chunks": 193.44, "bytes_per_chunk": 5.294, "boundary_entropy": 0.46207, "router_noise": 0.0, "grad_norm": 1.4962, "lr": 0.0003, "router_trainable": true, "sec_per_step": 0.342, "probe_sec": 0.0433, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03173, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.02846, "n_shadow": 0, "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": 0.15096, "trunc_abs_frac": 1.29849, "mean_delta_full": 0.10864, "read_hit_rate": {"split": 0.83929}, "mean_gate": {"split": 0.12872}, "n_reads": 8789, "n_cand": 10472, "read_suppressed_frac": 0.1002, "dropout_reads_logged": 8789, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 38.89750671386719, "n_util_obs": 65, "util_count_spearman": 0.058173076923076925}, "probe_overhead_sec_total": 20.546, "probe_overhead_seq_frac": 0.4981, "util_loss": 0.00062, "alpha": 0.03, "t": 110.7} |
| {"event": "consolidate", "step": 350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 124.1} |
| {"event": "train", "step": 350, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.16995, "bpb": 3.13057, "chunk_loss": 0.031121, "F": 0.20908, "G": 0.27986, "chunks": 214.09, "bytes_per_chunk": 4.783, "boundary_entropy": 0.39831, "router_noise": 0.0, "grad_norm": 1.6439, "lr": 0.0002999, "router_trainable": true, "sec_per_step": 0.276, "probe_sec": 0.0681, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05533, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.01634, "n_shadow": 0, "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": -0.23375, "trunc_abs_frac": 0.2943, "mean_delta_full": -0.20758, "read_hit_rate": {"split": 0.87233}, "mean_gate": {"split": 0.14124}, "n_reads": 10147, "n_cand": 11632, "read_suppressed_frac": 0.1003, "dropout_reads_logged": 10147, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 46.27012634277344, "n_util_obs": 77, "util_count_spearman": -0.019112466480887533}, "probe_overhead_sec_total": 23.726, "probe_overhead_seq_frac": 0.4984, "util_loss": 0.000812, "alpha": 0.03, "t": 124.4} |
| {"event": "consolidate", "step": 400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 138.3} |
| {"event": "probe", "step": 400, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3199.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.183819, "cps": 24.992188, "mean_p": 0.260522, "boundary_entropy": 0.359137, "bpb": 2.907454}, "tr": {"chunks": 3262.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.180191, "cps": 25.484375, "mean_p": 0.262445, "boundary_entropy": 0.399347, "bpb": 3.17758}, "fi": {"chunks": 4196.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.223061, "cps": 32.78125, "mean_p": 0.27774, "boundary_entropy": 0.422909, "bpb": 3.385928}}, "bpb": {"en": 2.90745, "tr": 3.17758, "fi": 3.38593}, "gini_cpc": 0.048682, "gini_cps": 0.062369, "bpic": 5.323449, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9802645232949168, "fi": 1.2134823842620401}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 53.69563293457031, "n_util_obs": 84, "util_count_spearman": -0.023549660828186696}, "t": 139.3} |
| {"event": "eval", "step": 400, "phase": 1, "bpb": {"en": 2.91919, "tr": 3.18634, "fi": 3.37756}, "bytes_per_chunk": {"en": 5.125, "tr": 5.819, "fi": 4.846}, "bpb_mean": 3.16103, "frozen_inventory": true, "t": 140.8} |
| {"event": "train", "step": 400, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 2.08887, "bpb": 3.0136, "chunk_loss": 0.030551, "F": 0.18994, "G": 0.26424, "chunks": 194.5, "bytes_per_chunk": 5.265, "boundary_entropy": 0.36008, "router_noise": 0.0, "grad_norm": 1.4891, "lr": 0.0002998, "router_trainable": true, "sec_per_step": 0.35, "probe_sec": 0.0491, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04013, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.05508, "n_shadow": 0, "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": 0.74216, "trunc_abs_frac": 0.74216, "mean_delta_full": -0.13232, "read_hit_rate": {"split": 0.84752}, "mean_gate": {"split": 0.15174}, "n_reads": 9004, "n_cand": 10624, "read_suppressed_frac": 0.1018, "dropout_reads_logged": 9004, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 53.83033752441406, "n_util_obs": 84, "util_count_spearman": -0.03474756883495386}, "probe_overhead_sec_total": 26.964, "probe_overhead_seq_frac": 0.4986, "util_loss": 0.001093, "alpha": 0.03, "t": 141.9} |
| {"event": "consolidate", "step": 450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 155.7} |
| {"event": "train", "step": 450, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.99648, "bpb": 2.88031, "chunk_loss": 0.030746, "F": 0.20007, "G": 0.26282, "chunks": 204.88, "bytes_per_chunk": 4.998, "boundary_entropy": 0.33191, "router_noise": 0.0, "grad_norm": 1.1408, "lr": 0.0002995, "router_trainable": true, "sec_per_step": 0.284, "probe_sec": 0.1191, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10427, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.00159, "n_shadow": 0, "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": -0.60292, "trunc_abs_frac": 1.98237, "mean_delta_full": 0.12118, "read_hit_rate": {"split": 0.87677}, "mean_gate": {"split": 0.16064}, "n_reads": 9925, "n_cand": 11320, "read_suppressed_frac": 0.0979, "dropout_reads_logged": 9925, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 61.40771484375, "n_util_obs": 103, "util_count_spearman": -0.05502350305320037}, "probe_overhead_sec_total": 30.76, "probe_overhead_seq_frac": 0.4988, "util_loss": 0.001279, "alpha": 0.03, "t": 156.1} |
| {"event": "consolidate", "step": 500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 169.6} |
| {"event": "train", "step": 500, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.90805, "bpb": 2.75273, "chunk_loss": 0.031, "F": 0.21011, "G": 0.26509, "chunks": 215.16, "bytes_per_chunk": 4.759, "boundary_entropy": 0.30798, "router_noise": 0.0, "grad_norm": 1.3689, "lr": 0.0002992, "router_trainable": true, "sec_per_step": 0.278, "probe_sec": 0.039, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02501, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.05792, "n_shadow": 0, "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": 0.21589, "trunc_abs_frac": 0.24956, "mean_delta_full": -0.17283, "read_hit_rate": {"split": 0.89234}, "mean_gate": {"split": 0.16766}, "n_reads": 10792, "n_cand": 12094, "read_suppressed_frac": 0.0993, "dropout_reads_logged": 10792, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 69.16044616699219, "n_util_obs": 106, "util_count_spearman": 0.016913116860216137}, "probe_overhead_sec_total": 34.144, "probe_overhead_seq_frac": 0.4989, "util_loss": 0.002075, "alpha": 0.03, "t": 170.0} |
| {"event": "probe", "step": 512, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3490.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.20054, "cps": 27.265625, "mean_p": 0.256123, "boundary_entropy": 0.30516, "bpb": 2.703047}, "tr": {"chunks": 3441.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.190079, "cps": 26.882812, "mean_p": 0.262124, "boundary_entropy": 0.358324, "bpb": 3.068351}, "fi": {"chunks": 4154.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.220828, "cps": 32.453125, "mean_p": 0.271715, "boundary_entropy": 0.36599, "bpb": 3.221882}}, "bpb": {"en": 2.70305, "tr": 3.06835, "fi": 3.22188}, "gini_cpc": 0.033526, "gini_cps": 0.042881, "bpic": 5.117907, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9478351591110213, "fi": 1.1011672893336963}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 70.85360717773438, "n_util_obs": 107, "util_count_spearman": 0.006955465428397892}, "t": 173.7} |
| {"event": "consolidate", "step": 550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 185.2} |
| {"event": "train", "step": 550, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.90427, "bpb": 2.74728, "chunk_loss": 0.030812, "F": 0.20328, "G": 0.26107, "chunks": 208.16, "bytes_per_chunk": 4.919, "boundary_entropy": 0.30365, "router_noise": 0.0, "grad_norm": 1.469, "lr": 0.0002987, "router_trainable": true, "sec_per_step": 0.307, "probe_sec": 0.0428, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03425, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.00205, "n_shadow": 0, "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": -0.391, "trunc_abs_frac": 1.62393, "mean_delta_full": -0.02646, "read_hit_rate": {"split": 0.8909}, "mean_gate": {"split": 0.16714}, "n_reads": 10436, "n_cand": 11714, "read_suppressed_frac": 0.1007, "dropout_reads_logged": 10436, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 76.96479797363281, "n_util_obs": 115, "util_count_spearman": 0.05080091533180778}, "probe_overhead_sec_total": 37.706, "probe_overhead_seq_frac": 0.499, "util_loss": 0.002478, "alpha": 0.03, "t": 185.4} |
| {"event": "consolidate", "step": 600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 198.9} |
| {"event": "probe", "step": 600, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3346.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.192266, "cps": 26.140625, "mean_p": 0.250359, "boundary_entropy": 0.291579, "bpb": 2.599228}, "tr": {"chunks": 3620.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.199967, "cps": 28.28125, "mean_p": 0.262186, "boundary_entropy": 0.333975, "bpb": 2.923426}, "fi": {"chunks": 4178.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.222104, "cps": 32.640625, "mean_p": 0.273907, "boundary_entropy": 0.346139, "bpb": 3.138156}}, "bpb": {"en": 2.59923, "tr": 2.92343, "fi": 3.13816}, "gini_cpc": 0.03238, "gini_cps": 0.049773, "bpic": 5.090811, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0400547521143129, "fi": 1.155193497922503}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 84.77732849121094, "n_util_obs": 127, "util_count_spearman": -0.0035854893138357706}, "t": 199.8} |
| {"event": "eval", "step": 600, "phase": 1, "bpb": {"en": 2.61199, "tr": 2.93132, "fi": 3.13905}, "bytes_per_chunk": {"en": 4.994, "tr": 5.634, "fi": 4.622}, "bpb_mean": 2.89412, "frozen_inventory": true, "t": 201.3} |
| {"event": "train", "step": 600, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.89225, "bpb": 2.72993, "chunk_loss": 0.030748, "F": 0.2016, "G": 0.25666, "chunks": 206.44, "bytes_per_chunk": 4.96, "boundary_entropy": 0.29369, "router_noise": 0.0, "grad_norm": 1.2389, "lr": 0.0002982, "router_trainable": true, "sec_per_step": 0.341, "probe_sec": 0.0825, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07447, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.10464, "n_shadow": 0, "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": 4.78447, "trunc_abs_frac": 0.73844, "mean_delta_full": 0.01225, "read_hit_rate": {"split": 0.88442}, "mean_gate": {"split": 0.17232}, "n_reads": 10215, "n_cand": 11550, "read_suppressed_frac": 0.0995, "dropout_reads_logged": 10215, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 84.93251037597656, "n_util_obs": 127, "util_count_spearman": -0.005718035245594301}, "probe_overhead_sec_total": 41.113, "probe_overhead_seq_frac": 0.4991, "util_loss": 0.002784, "alpha": 0.03, "t": 202.4} |
| {"event": "consolidate", "step": 650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 216.3} |
| {"event": "train", "step": 650, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.838, "bpb": 2.65168, "chunk_loss": 0.030682, "F": 0.19806, "G": 0.25668, "chunks": 202.81, "bytes_per_chunk": 5.049, "boundary_entropy": 0.28571, "router_noise": 0.0, "grad_norm": 1.1359, "lr": 0.0002975, "router_trainable": true, "sec_per_step": 0.286, "probe_sec": 0.0481, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03801, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.03102, "n_shadow": 0, "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": 0.01796, "trunc_abs_frac": 0.30654, "mean_delta_full": 0.10916, "read_hit_rate": {"split": 0.88987}, "mean_gate": {"split": 0.17303}, "n_reads": 10205, "n_cand": 11468, "read_suppressed_frac": 0.0994, "dropout_reads_logged": 10205, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 92.91929626464844, "n_util_obs": 133, "util_count_spearman": 0.001617000438681507}, "probe_overhead_sec_total": 45.052, "probe_overhead_seq_frac": 0.4991, "util_loss": 0.003423, "alpha": 0.03, "t": 216.7} |
| {"event": "consolidate", "step": 700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 229.6} |
| {"event": "train", "step": 700, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.98954, "bpb": 2.8703, "chunk_loss": 0.031236, "F": 0.2131, "G": 0.26647, "chunks": 218.22, "bytes_per_chunk": 4.693, "boundary_entropy": 0.29076, "router_noise": 0.0, "grad_norm": 1.2354, "lr": 0.0002967, "router_trainable": true, "sec_per_step": 0.264, "probe_sec": 0.0953, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.083, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.12524, "n_shadow": 0, "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": -0.53592, "trunc_abs_frac": 2.48523, "mean_delta_full": -0.11808, "read_hit_rate": {"split": 0.8952}, "mean_gate": {"split": 0.17786}, "n_reads": 10635, "n_cand": 11880, "read_suppressed_frac": 0.099, "dropout_reads_logged": 10635, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 100.96186828613281, "n_util_obs": 140, "util_count_spearman": -0.13779274452778204}, "probe_overhead_sec_total": 48.115, "probe_overhead_seq_frac": 0.4992, "util_loss": 0.003563, "alpha": 0.03, "t": 230.0} |
| {"event": "consolidate", "step": 750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 243.1} |
| {"event": "train", "step": 750, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.78137, "bpb": 2.56998, "chunk_loss": 0.030952, "F": 0.21054, "G": 0.26287, "chunks": 215.59, "bytes_per_chunk": 4.75, "boundary_entropy": 0.27056, "router_noise": 0.0, "grad_norm": 1.0818, "lr": 0.0002959, "router_trainable": true, "sec_per_step": 0.269, "probe_sec": 0.1249, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.11055, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.02649, "n_shadow": 0, "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": 0.48346, "trunc_abs_frac": 1.61406, "mean_delta_full": -0.03033, "read_hit_rate": {"split": 0.90779}, "mean_gate": {"split": 0.17856}, "n_reads": 10937, "n_cand": 12048, "read_suppressed_frac": 0.1007, "dropout_reads_logged": 10937, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 109.00100708007812, "n_util_obs": 151, "util_count_spearman": -0.1307354478912513}, "probe_overhead_sec_total": 51.451, "probe_overhead_seq_frac": 0.4993, "util_loss": 0.003884, "alpha": 0.03, "t": 243.4} |
| {"event": "consolidate", "step": 800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 256.7} |
| {"event": "probe", "step": 800, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3424.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.196748, "cps": 26.75, "mean_p": 0.245429, "boundary_entropy": 0.263319, "bpb": 2.435192}, "tr": {"chunks": 3541.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.195603, "cps": 27.664062, "mean_p": 0.254624, "boundary_entropy": 0.310016, "bpb": 2.765111}, "fi": {"chunks": 4122.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.219127, "cps": 32.203125, "mean_p": 0.260659, "boundary_entropy": 0.299692, "bpb": 2.978237}}, "bpb": {"en": 2.43519, "tr": 2.76511, "fi": 2.97824}, "gini_cpc": 0.025647, "gini_cps": 0.041971, "bpic": 5.116984, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9941816421969613, "fi": 1.113746797335224}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 116.94253540039062, "n_util_obs": 163, "util_count_spearman": -0.14166977248778082}, "t": 257.6} |
| {"event": "eval", "step": 800, "phase": 1, "bpb": {"en": 2.44559, "tr": 2.77351, "fi": 2.98005}, "bytes_per_chunk": {"en": 4.89, "tr": 5.636, "fi": 4.722}, "bpb_mean": 2.73305, "frozen_inventory": true, "t": 259.2} |
| {"event": "train", "step": 800, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.85437, "bpb": 2.67529, "chunk_loss": 0.030972, "F": 0.20816, "G": 0.26054, "chunks": 213.16, "bytes_per_chunk": 4.804, "boundary_entropy": 0.27187, "router_noise": 0.0, "grad_norm": 1.2273, "lr": 0.0002949, "router_trainable": true, "sec_per_step": 0.34, "probe_sec": 0.0643, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05422, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.03526, "n_shadow": 0, "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": 1.08929, "trunc_abs_frac": 0.66586, "mean_delta_full": -0.06485, "read_hit_rate": {"split": 0.89361}, "mean_gate": {"split": 0.17743}, "n_reads": 10264, "n_cand": 11486, "read_suppressed_frac": 0.0978, "dropout_reads_logged": 10264, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 117.10017395019531, "n_util_obs": 163, "util_count_spearman": -0.14244416626779363}, "probe_overhead_sec_total": 54.642, "probe_overhead_seq_frac": 0.4993, "util_loss": 0.004037, "alpha": 0.03, "t": 260.4} |
| {"event": "consolidate", "step": 850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 277.1} |
| {"event": "train", "step": 850, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.80424, "bpb": 2.60297, "chunk_loss": 0.030853, "F": 0.20914, "G": 0.25591, "chunks": 214.16, "bytes_per_chunk": 4.782, "boundary_entropy": 0.26131, "router_noise": 0.0, "grad_norm": 1.0227, "lr": 0.0002938, "router_trainable": true, "sec_per_step": 0.337, "probe_sec": 0.0659, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05704, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.09194, "n_shadow": 0, "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": -0.28097, "trunc_abs_frac": 0.94355, "mean_delta_full": -0.0069, "read_hit_rate": {"split": 0.90086}, "mean_gate": {"split": 0.17943}, "n_reads": 10940, "n_cand": 12144, "read_suppressed_frac": 0.0985, "dropout_reads_logged": 10940, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 125.18403625488281, "n_util_obs": 172, "util_count_spearman": -0.13865993118297829}, "probe_overhead_sec_total": 58.679, "probe_overhead_seq_frac": 0.4993, "util_loss": 0.004217, "alpha": 0.03, "t": 277.3} |
| {"event": "consolidate", "step": 900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 290.5} |
| {"event": "train", "step": 900, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.60458, "bpb": 2.31492, "chunk_loss": 0.030774, "F": 0.20529, "G": 0.25215, "chunks": 210.22, "bytes_per_chunk": 4.871, "boundary_entropy": 0.25174, "router_noise": 0.0, "grad_norm": 1.0367, "lr": 0.0002927, "router_trainable": true, "sec_per_step": 0.267, "probe_sec": 0.0456, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0369, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.02311, "n_shadow": 0, "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": 0.34297, "trunc_abs_frac": 0.34965, "mean_delta_full": 0.249, "read_hit_rate": {"split": 0.91021}, "mean_gate": {"split": 0.18383}, "n_reads": 10715, "n_cand": 11772, "read_suppressed_frac": 0.0963, "dropout_reads_logged": 10715, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 133.258544921875, "n_util_obs": 183, "util_count_spearman": -0.09514541054099145}, "probe_overhead_sec_total": 61.773, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.004554, "alpha": 0.03, "t": 290.7} |
| {"event": "consolidate", "step": 950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 304.6} |
| {"event": "train", "step": 950, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.68214, "bpb": 2.42682, "chunk_loss": 0.030778, "F": 0.2052, "G": 0.24994, "chunks": 210.12, "bytes_per_chunk": 4.873, "boundary_entropy": 0.24739, "router_noise": 0.0, "grad_norm": 1.0425, "lr": 0.0002914, "router_trainable": true, "sec_per_step": 0.283, "probe_sec": 0.0411, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03289, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.01985, "n_shadow": 0, "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": 0.23306, "trunc_abs_frac": 0.8083, "mean_delta_full": -0.17389, "read_hit_rate": {"split": 0.90827}, "mean_gate": {"split": 0.18002}, "n_reads": 10694, "n_cand": 11774, "read_suppressed_frac": 0.0976, "dropout_reads_logged": 10694, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 141.34283447265625, "n_util_obs": 193, "util_count_spearman": -0.04649858447732493}, "probe_overhead_sec_total": 64.631, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.00478, "alpha": 0.03, "t": 304.8} |
| {"event": "consolidate", "step": 1000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 319.8} |
| {"event": "probe", "step": 1000, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3506.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.20146, "cps": 27.390625, "mean_p": 0.240579, "boundary_entropy": 0.242805, "bpb": 2.328317}, "tr": {"chunks": 3799.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.209855, "cps": 29.679688, "mean_p": 0.24683, "boundary_entropy": 0.283651, "bpb": 2.633965}, "fi": {"chunks": 4233.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.225028, "cps": 33.070312, "mean_p": 0.258504, "boundary_entropy": 0.276529, "bpb": 2.850997}}, "bpb": {"en": 2.32832, "tr": 2.63396, "fi": 2.851}, "gini_cpc": 0.024692, "gini_cps": 0.042006, "bpic": 4.91697, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0416719041219384, "fi": 1.1169882212902156}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 149.25917053222656, "n_util_obs": 206, "util_count_spearman": -0.059369155109871063}, "t": 320.7} |
| {"event": "eval", "step": 1000, "phase": 1, "bpb": {"en": 2.34038, "tr": 2.64118, "fi": 2.8566}, "bytes_per_chunk": {"en": 4.81, "tr": 5.304, "fi": 4.616}, "bpb_mean": 2.61272, "frozen_inventory": true, "t": 322.2} |
| {"event": "train", "step": 1000, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.73853, "bpb": 2.50817, "chunk_loss": 0.030787, "F": 0.20758, "G": 0.25053, "chunks": 212.56, "bytes_per_chunk": 4.817, "boundary_entropy": 0.2492, "router_noise": 0.0, "grad_norm": 0.9287, "lr": 0.0002901, "router_trainable": true, "sec_per_step": 0.377, "probe_sec": 0.1122, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09594, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.0088, "n_shadow": 0, "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": 0.06603, "trunc_abs_frac": 0.64376, "mean_delta_full": 0.12849, "read_hit_rate": {"split": 0.90604}, "mean_gate": {"split": 0.17774}, "n_reads": 10838, "n_cand": 11962, "read_suppressed_frac": 0.0986, "dropout_reads_logged": 10838, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 149.4256591796875, "n_util_obs": 206, "util_count_spearman": -0.05991688030008202}, "probe_overhead_sec_total": 68.302, "probe_overhead_seq_frac": 0.4994, "util_loss": 0.005142, "alpha": 0.03, "t": 323.6} |
| {"event": "probe", "step": 1024, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3469.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.199333, "cps": 27.101562, "mean_p": 0.2391, "boundary_entropy": 0.243449, "bpb": 2.325329}, "tr": {"chunks": 3503.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.193504, "cps": 27.367188, "mean_p": 0.242584, "boundary_entropy": 0.284657, "bpb": 2.634541}, "fi": {"chunks": 4266.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.226782, "cps": 33.328125, "mean_p": 0.256776, "boundary_entropy": 0.265758, "bpb": 2.821453}}, "bpb": {"en": 2.32533, "tr": 2.63454, "fi": 2.82145}, "gini_cpc": 0.035805, "gini_cps": 0.04728, "bpic": 5.048229, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9707544861920212, "fi": 1.1377026981033083}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 153.13169860839844, "n_util_obs": 212, "util_count_spearman": -0.09370611579699065}, "t": 330.9} |
| {"event": "consolidate", "step": 1050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 339.6} |
| {"event": "train", "step": 1050, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.63076, "bpb": 2.3527, "chunk_loss": 0.03097, "F": 0.21143, "G": 0.24736, "chunks": 216.5, "bytes_per_chunk": 4.73, "boundary_entropy": 0.23588, "router_noise": 0.0, "grad_norm": 0.9781, "lr": 0.0002886, "router_trainable": true, "sec_per_step": 0.326, "probe_sec": 0.122, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10975, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.02165, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -8.13125, "trunc_abs_frac": 2.11063, "mean_delta_full": -0.01302, "read_hit_rate": {"split": 0.91043}, "mean_gate": {"split": 0.17459}, "n_reads": 10947, "n_cand": 12024, "read_suppressed_frac": 0.0985, "dropout_reads_logged": 10947, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 157.5243682861328, "n_util_obs": 215, "util_count_spearman": -0.06413098601753242}, "probe_overhead_sec_total": 71.8, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.005646, "alpha": 0.03, "t": 339.9} |
| {"event": "consolidate", "step": 1100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 356.0} |
| {"event": "train", "step": 1100, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.74205, "bpb": 2.51325, "chunk_loss": 0.030812, "F": 0.20718, "G": 0.24778, "chunks": 212.16, "bytes_per_chunk": 4.827, "boundary_entropy": 0.24422, "router_noise": 0.0, "grad_norm": 1.0508, "lr": 0.0002871, "router_trainable": true, "sec_per_step": 0.328, "probe_sec": 0.0713, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05779, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.07079, "n_shadow": 0, "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": 1.25147, "trunc_abs_frac": 0.60345, "mean_delta_full": -0.04916, "read_hit_rate": {"split": 0.89615}, "mean_gate": {"split": 0.17888}, "n_reads": 10510, "n_cand": 11728, "read_suppressed_frac": 0.0984, "dropout_reads_logged": 10510, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 165.60552978515625, "n_util_obs": 226, "util_count_spearman": -0.045247878575234234}, "probe_overhead_sec_total": 75.427, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.0056, "alpha": 0.03, "t": 356.3} |
| {"event": "consolidate", "step": 1150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 379.7} |
| {"event": "train", "step": 1150, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.64203, "bpb": 2.36894, "chunk_loss": 0.030695, "F": 0.20505, "G": 0.2429, "chunks": 209.97, "bytes_per_chunk": 4.877, "boundary_entropy": 0.2258, "router_noise": 0.0, "grad_norm": 0.8649, "lr": 0.0002855, "router_trainable": true, "sec_per_step": 0.474, "probe_sec": 0.0602, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04796, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.1661, "n_shadow": 0, "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": 0.79516, "trunc_abs_frac": 0.7662, "mean_delta_full": -0.16895, "read_hit_rate": {"split": 0.90766}, "mean_gate": {"split": 0.17581}, "n_reads": 10812, "n_cand": 11912, "read_suppressed_frac": 0.0972, "dropout_reads_logged": 10812, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 173.71487426757812, "n_util_obs": 234, "util_count_spearman": -0.029016909594232816}, "probe_overhead_sec_total": 79.561, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.005974, "alpha": 0.03, "t": 380.0} |
| {"event": "consolidate", "step": 1200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 405.9} |
| {"event": "probe", "step": 1200, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3421.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.196575, "cps": 26.726562, "mean_p": 0.229746, "boundary_entropy": 0.215776, "bpb": 2.264526}, "tr": {"chunks": 3444.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.190245, "cps": 26.90625, "mean_p": 0.230104, "boundary_entropy": 0.255891, "bpb": 2.556107}, "fi": {"chunks": 4374.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.232524, "cps": 34.171875, "mean_p": 0.255349, "boundary_entropy": 0.239447, "bpb": 2.720896}}, "bpb": {"en": 2.26453, "tr": 2.55611, "fi": 2.7209}, "gini_cpc": 0.045509, "gini_cps": 0.056529, "bpic": 5.04778, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9677955867948005, "fi": 1.182872516502135}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 181.70037841796875, "n_util_obs": 248, "util_count_spearman": -0.01915248823424821}, "t": 406.7} |
| {"event": "eval", "step": 1200, "phase": 1, "bpb": {"en": 2.27444, "tr": 2.56411, "fi": 2.72929}, "bytes_per_chunk": {"en": 4.878, "tr": 5.835, "fi": 4.446}, "bpb_mean": 2.52261, "frozen_inventory": true, "t": 408.3} |
| {"event": "train", "step": 1200, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.61718, "bpb": 2.33309, "chunk_loss": 0.030575, "F": 0.20007, "G": 0.23959, "chunks": 204.88, "bytes_per_chunk": 4.998, "boundary_entropy": 0.22128, "router_noise": 0.0, "grad_norm": 0.9077, "lr": 0.0002837, "router_trainable": true, "sec_per_step": 0.596, "probe_sec": 0.0659, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0519, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.09177, "n_shadow": 0, "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": 1.86369, "trunc_abs_frac": 0.50315, "mean_delta_full": 0.1059, "read_hit_rate": {"split": 0.90506}, "mean_gate": {"split": 0.17534}, "n_reads": 10562, "n_cand": 11670, "read_suppressed_frac": 0.102, "dropout_reads_logged": 10562, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 181.8608856201172, "n_util_obs": 249, "util_count_spearman": -0.024450058297706957}, "probe_overhead_sec_total": 83.595, "probe_overhead_seq_frac": 0.4995, "util_loss": 0.005938, "alpha": 0.03, "t": 409.8} |
| {"event": "consolidate", "step": 1250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 425.5} |
| {"event": "train", "step": 1250, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.62943, "bpb": 2.35077, "chunk_loss": 0.030624, "F": 0.20093, "G": 0.23996, "chunks": 205.75, "bytes_per_chunk": 4.977, "boundary_entropy": 0.22215, "router_noise": 0.0, "grad_norm": 0.8573, "lr": 0.0002819, "router_trainable": true, "sec_per_step": 0.318, "probe_sec": 0.0551, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04436, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.01507, "n_shadow": 0, "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": 1.34795, "trunc_abs_frac": 3.42162, "mean_delta_full": -0.0756, "read_hit_rate": {"split": 0.90024}, "mean_gate": {"split": 0.1753}, "n_reads": 10378, "n_cand": 11528, "read_suppressed_frac": 0.1001, "dropout_reads_logged": 10378, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 189.9832763671875, "n_util_obs": 255, "util_count_spearman": -0.05624605009882253}, "probe_overhead_sec_total": 87.249, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.005925, "alpha": 0.03, "t": 425.7} |
| {"event": "consolidate", "step": 1300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 444.4} |
| {"event": "train", "step": 1300, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.67692, "bpb": 2.41929, "chunk_loss": 0.030884, "F": 0.21246, "G": 0.24637, "chunks": 217.56, "bytes_per_chunk": 4.707, "boundary_entropy": 0.21337, "router_noise": 0.0, "grad_norm": 0.953, "lr": 0.00028, "router_trainable": true, "sec_per_step": 0.381, "probe_sec": 0.039, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02552, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.03478, "n_shadow": 0, "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": 0.31583, "trunc_abs_frac": 0.26396, "mean_delta_full": -0.29932, "read_hit_rate": {"split": 0.90857}, "mean_gate": {"split": 0.17134}, "n_reads": 11010, "n_cand": 12118, "read_suppressed_frac": 0.1045, "dropout_reads_logged": 11010, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 198.1029052734375, "n_util_obs": 259, "util_count_spearman": -0.06893551948651114}, "probe_overhead_sec_total": 91.158, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.006444, "alpha": 0.03, "t": 444.7} |
| {"event": "consolidate", "step": 1350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 464.9} |
| {"event": "train", "step": 1350, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.59681, "bpb": 2.3037, "chunk_loss": 0.030708, "F": 0.20694, "G": 0.23976, "chunks": 211.91, "bytes_per_chunk": 4.832, "boundary_entropy": 0.2152, "router_noise": 0.0, "grad_norm": 0.8284, "lr": 0.000278, "router_trainable": true, "sec_per_step": 0.411, "probe_sec": 0.053, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0352, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.08403, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 0, "read_hit_rate": {"split": 0.90059}, "mean_gate": {"split": 0.16891}, "n_reads": 10735, "n_cand": 11920, "read_suppressed_frac": 0.1011, "dropout_reads_logged": 10735, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 206.19639587402344, "n_util_obs": 264, "util_count_spearman": -0.07265357499233674}, "probe_overhead_sec_total": 94.413, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.006452, "alpha": 0.03, "t": 465.3} |
| {"event": "consolidate", "step": 1400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 478.6} |
| {"event": "probe", "step": 1400, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3439.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.19761, "cps": 26.867188, "mean_p": 0.226824, "boundary_entropy": 0.196358, "bpb": 2.216309}, "tr": {"chunks": 3648.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.201514, "cps": 28.5, "mean_p": 0.224388, "boundary_entropy": 0.234137, "bpb": 2.489449}, "fi": {"chunks": 4270.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.226995, "cps": 33.359375, "mean_p": 0.248043, "boundary_entropy": 0.225256, "bpb": 2.645561}}, "bpb": {"en": 2.21631, "tr": 2.48945, "fi": 2.64556}, "gini_cpc": 0.031288, "gini_cps": 0.04878, "bpic": 4.995333, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0197558904036845, "fi": 1.1487034778809904}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 214.17837524414062, "n_util_obs": 272, "util_count_spearman": -0.05963483358381387}, "t": 479.4} |
| {"event": "eval", "step": 1400, "phase": 1, "bpb": {"en": 2.22285, "tr": 2.49632, "fi": 2.65104}, "bytes_per_chunk": {"en": 4.852, "tr": 5.45, "fi": 4.649}, "bpb_mean": 2.45674, "frozen_inventory": true, "t": 480.8} |
| {"event": "train", "step": 1400, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.59664, "bpb": 2.30346, "chunk_loss": 0.030638, "F": 0.20355, "G": 0.23459, "chunks": 208.44, "bytes_per_chunk": 4.913, "boundary_entropy": 0.20516, "router_noise": 0.0, "grad_norm": 0.8874, "lr": 0.0002759, "router_trainable": true, "sec_per_step": 0.336, "probe_sec": 0.0896, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0812, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.02351, "n_shadow": 0, "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": 1.38347, "trunc_abs_frac": 0.8886, "mean_delta_full": 0.0144, "read_hit_rate": {"split": 0.9006}, "mean_gate": {"split": 0.1733}, "n_reads": 10429, "n_cand": 11580, "read_suppressed_frac": 0.102, "dropout_reads_logged": 10429, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 214.33790588378906, "n_util_obs": 272, "util_count_spearman": -0.05916940465199414}, "probe_overhead_sec_total": 98.048, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.006936, "alpha": 0.03, "t": 482.0} |
| {"event": "consolidate", "step": 1450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 497.1} |
| {"event": "train", "step": 1450, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.65036, "bpb": 2.38097, "chunk_loss": 0.03067, "F": 0.20609, "G": 0.23707, "chunks": 211.03, "bytes_per_chunk": 4.852, "boundary_entropy": 0.20045, "router_noise": 0.0, "grad_norm": 0.8097, "lr": 0.0002738, "router_trainable": true, "sec_per_step": 0.311, "probe_sec": 0.0737, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05825, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.00881, "n_shadow": 0, "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": 0.56352, "trunc_abs_frac": 0.62689, "mean_delta_full": 0.30722, "read_hit_rate": {"split": 0.89775}, "mean_gate": {"split": 0.1655}, "n_reads": 10387, "n_cand": 11570, "read_suppressed_frac": 0.0971, "dropout_reads_logged": 10387, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 222.5381622314453, "n_util_obs": 285, "util_count_spearman": -0.07612529303308045}, "probe_overhead_sec_total": 101.56, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.006598, "alpha": 0.03, "t": 497.6} |
| {"event": "consolidate", "step": 1500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 510.0} |
| {"event": "train", "step": 1500, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.66908, "bpb": 2.40798, "chunk_loss": 0.030849, "F": 0.2106, "G": 0.2432, "chunks": 215.66, "bytes_per_chunk": 4.748, "boundary_entropy": 0.20102, "router_noise": 0.0, "grad_norm": 0.8606, "lr": 0.0002715, "router_trainable": true, "sec_per_step": 0.254, "probe_sec": 0.0457, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03574, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.00211, "n_shadow": 0, "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": 0.93264, "trunc_abs_frac": 2.51662, "mean_delta_full": 0.41321, "read_hit_rate": {"split": 0.90332}, "mean_gate": {"split": 0.16662}, "n_reads": 10764, "n_cand": 11916, "read_suppressed_frac": 0.1002, "dropout_reads_logged": 10764, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 230.70211791992188, "n_util_obs": 293, "util_count_spearman": -0.04285461948954093}, "probe_overhead_sec_total": 105.246, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.006968, "alpha": 0.03, "t": 510.3} |
| {"event": "consolidate", "step": 1550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 523.4} |
| {"event": "train", "step": 1550, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.58675, "bpb": 2.2892, "chunk_loss": 0.030674, "F": 0.20306, "G": 0.2355, "chunks": 207.94, "bytes_per_chunk": 4.925, "boundary_entropy": 0.1943, "router_noise": 0.0, "grad_norm": 0.8995, "lr": 0.0002692, "router_trainable": true, "sec_per_step": 0.266, "probe_sec": 0.0405, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03124, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.01709, "n_shadow": 0, "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": 0.9706, "trunc_abs_frac": 0.33846, "mean_delta_full": -0.08443, "read_hit_rate": {"split": 0.90235}, "mean_gate": {"split": 0.16372}, "n_reads": 10543, "n_cand": 11684, "read_suppressed_frac": 0.0962, "dropout_reads_logged": 10543, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 238.8048553466797, "n_util_obs": 299, "util_count_spearman": -0.059765212902067294}, "probe_overhead_sec_total": 108.598, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.006926, "alpha": 0.03, "t": 523.6} |
| {"event": "consolidate", "step": 1600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 537.6} |
| {"event": "probe", "step": 1600, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3319.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.190714, "cps": 25.929688, "mean_p": 0.220369, "boundary_entropy": 0.187757, "bpb": 2.17578}, "tr": {"chunks": 3582.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.197868, "cps": 27.984375, "mean_p": 0.218396, "boundary_entropy": 0.22008, "bpb": 2.420577}, "fi": {"chunks": 4207.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.223646, "cps": 32.867188, "mean_p": 0.242443, "boundary_entropy": 0.205794, "bpb": 2.581832}}, "bpb": {"en": 2.17578, "tr": 2.42058, "fi": 2.58183}, "gini_cpc": 0.03586, "gini_cps": 0.053295, "bpic": 5.10731, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0375090600458623, "fi": 1.172674540287203}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 246.7036590576172, "n_util_obs": 306, "util_count_spearman": -0.06813588184740894}, "t": 538.4} |
| {"event": "eval", "step": 1600, "phase": 1, "bpb": {"en": 2.18667, "tr": 2.43031, "fi": 2.58889}, "bytes_per_chunk": {"en": 4.995, "tr": 5.512, "fi": 4.634}, "bpb_mean": 2.40196, "frozen_inventory": true, "t": 539.9} |
| {"event": "train", "step": 1600, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.51587, "bpb": 2.18694, "chunk_loss": 0.030646, "F": 0.20636, "G": 0.23341, "chunks": 211.31, "bytes_per_chunk": 4.846, "boundary_entropy": 0.18804, "router_noise": 0.0, "grad_norm": 0.9255, "lr": 0.0002668, "router_trainable": true, "sec_per_step": 0.349, "probe_sec": 0.1595, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.14467, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.13957, "n_shadow": 0, "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": 1.17292, "trunc_abs_frac": 0.81908, "mean_delta_full": 0.17419, "read_hit_rate": {"split": 0.90656}, "mean_gate": {"split": 0.1636}, "n_reads": 10886, "n_cand": 12008, "read_suppressed_frac": 0.1033, "dropout_reads_logged": 10886, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 246.87110900878906, "n_util_obs": 306, "util_count_spearman": -0.06680362596980448}, "probe_overhead_sec_total": 111.964, "probe_overhead_seq_frac": 0.4996, "util_loss": 0.007351, "alpha": 0.03, "t": 541.0} |
| {"event": "consolidate", "step": 1650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 554.0} |
| {"event": "train", "step": 1650, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.58016, "bpb": 2.27969, "chunk_loss": 0.030573, "F": 0.20456, "G": 0.23184, "chunks": 209.47, "bytes_per_chunk": 4.889, "boundary_entropy": 0.18245, "router_noise": 0.0, "grad_norm": 0.8156, "lr": 0.0002643, "router_trainable": true, "sec_per_step": 0.264, "probe_sec": 0.0534, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04292, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.01892, "n_shadow": 0, "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": 0.44509, "trunc_abs_frac": 3.35891, "mean_delta_full": -0.29456, "read_hit_rate": {"split": 0.90408}, "mean_gate": {"split": 0.16076}, "n_reads": 10802, "n_cand": 11948, "read_suppressed_frac": 0.0999, "dropout_reads_logged": 10802, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 255.02035522460938, "n_util_obs": 313, "util_count_spearman": -0.028325284965445424}, "probe_overhead_sec_total": 115.794, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.007145, "alpha": 0.03, "t": 554.3} |
| {"event": "consolidate", "step": 1700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 567.0} |
| {"event": "train", "step": 1700, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.53916, "bpb": 2.22053, "chunk_loss": 0.030502, "F": 0.19971, "G": 0.22623, "chunks": 204.5, "bytes_per_chunk": 5.007, "boundary_entropy": 0.18463, "router_noise": 0.0, "grad_norm": 0.7556, "lr": 0.0002618, "router_trainable": true, "sec_per_step": 0.264, "probe_sec": 0.1419, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.12874, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.09006, "n_shadow": 0, "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": 0.59107, "trunc_abs_frac": 1.76697, "mean_delta_full": -0.19368, "read_hit_rate": {"split": 0.89918}, "mean_gate": {"split": 0.15743}, "n_reads": 10328, "n_cand": 11486, "read_suppressed_frac": 0.0968, "dropout_reads_logged": 10328, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 263.1134338378906, "n_util_obs": 320, "util_count_spearman": -0.029605516436212558}, "probe_overhead_sec_total": 119.356, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.00781, "alpha": 0.03, "t": 567.4} |
| {"event": "consolidate", "step": 1750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 579.8} |
| {"event": "train", "step": 1750, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.57034, "bpb": 2.26553, "chunk_loss": 0.030521, "F": 0.20071, "G": 0.22926, "chunks": 205.53, "bytes_per_chunk": 4.982, "boundary_entropy": 0.18363, "router_noise": 0.0, "grad_norm": 0.8057, "lr": 0.0002591, "router_trainable": true, "sec_per_step": 0.253, "probe_sec": 0.0631, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05414, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.01797, "n_shadow": 0, "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": 0.19543, "trunc_abs_frac": 0.41124, "mean_delta_full": -0.07253, "read_hit_rate": {"split": 0.89704}, "mean_gate": {"split": 0.1569}, "n_reads": 10429, "n_cand": 11626, "read_suppressed_frac": 0.1026, "dropout_reads_logged": 10429, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 271.16412353515625, "n_util_obs": 334, "util_count_spearman": -0.01599898484579608}, "probe_overhead_sec_total": 123.043, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.008568, "alpha": 0.03, "t": 580.1} |
| {"event": "consolidate", "step": 1800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 593.0} |
| {"event": "probe", "step": 1800, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3406.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.195713, "cps": 26.609375, "mean_p": 0.218406, "boundary_entropy": 0.170905, "bpb": 2.134848}, "tr": {"chunks": 3563.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.196818, "cps": 27.835938, "mean_p": 0.21611, "boundary_entropy": 0.204445, "bpb": 2.36231}, "fi": {"chunks": 4316.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.22944, "cps": 33.71875, "mean_p": 0.242929, "boundary_entropy": 0.1873, "bpb": 2.520843}}, "bpb": {"en": 2.13485, "tr": 2.36231, "fi": 2.52084}, "gini_cpc": 0.03615, "gini_cps": 0.053759, "bpic": 5.027204, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0056451130801762, "fi": 1.1723277065838935}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 279.13592529296875, "n_util_obs": 342, "util_count_spearman": -0.02506566593996648}, "t": 593.7} |
| {"event": "eval", "step": 1800, "phase": 1, "bpb": {"en": 2.14713, "tr": 2.37269, "fi": 2.5256}, "bytes_per_chunk": {"en": 4.889, "tr": 5.545, "fi": 4.585}, "bpb_mean": 2.34847, "frozen_inventory": true, "t": 595.2} |
| {"event": "train", "step": 1800, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.53854, "bpb": 2.21965, "chunk_loss": 0.030555, "F": 0.2023, "G": 0.22661, "chunks": 207.16, "bytes_per_chunk": 4.943, "boundary_entropy": 0.17678, "router_noise": 0.0, "grad_norm": 0.7722, "lr": 0.0002564, "router_trainable": true, "sec_per_step": 0.324, "probe_sec": 0.0659, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05647, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.02621, "n_shadow": 0, "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": 0.23381, "trunc_abs_frac": 1.53673, "mean_delta_full": 0.07562, "read_hit_rate": {"split": 0.90175}, "mean_gate": {"split": 0.15866}, "n_reads": 10426, "n_cand": 11562, "read_suppressed_frac": 0.101, "dropout_reads_logged": 10426, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 279.2955322265625, "n_util_obs": 342, "util_count_spearman": -0.025373605693271198}, "probe_overhead_sec_total": 126.536, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.007818, "alpha": 0.03, "t": 596.3} |
| {"event": "consolidate", "step": 1850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 610.5} |
| {"event": "train", "step": 1850, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.55478, "bpb": 2.24307, "chunk_loss": 0.030671, "F": 0.20755, "G": 0.2316, "chunks": 212.53, "bytes_per_chunk": 4.818, "boundary_entropy": 0.16886, "router_noise": 0.0, "grad_norm": 0.8021, "lr": 0.0002537, "router_trainable": true, "sec_per_step": 0.288, "probe_sec": 0.0452, "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": 0.05875, "n_shadow": 0, "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": 0.39104, "trunc_abs_frac": 0.39104, "mean_delta_full": 0.22742, "read_hit_rate": {"split": 0.9055}, "mean_gate": {"split": 0.15419}, "n_reads": 10857, "n_cand": 11990, "read_suppressed_frac": 0.0973, "dropout_reads_logged": 10857, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 287.40728759765625, "n_util_obs": 347, "util_count_spearman": -0.04322250217021524}, "probe_overhead_sec_total": 130.132, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.008019, "alpha": 0.03, "t": 610.7} |
| {"event": "consolidate", "step": 1900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 624.6} |
| {"event": "train", "step": 1900, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.53662, "bpb": 2.21687, "chunk_loss": 0.030703, "F": 0.20648, "G": 0.23003, "chunks": 211.44, "bytes_per_chunk": 4.843, "boundary_entropy": 0.17038, "router_noise": 0.0, "grad_norm": 0.8518, "lr": 0.0002508, "router_trainable": true, "sec_per_step": 0.284, "probe_sec": 0.0606, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04358, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.07186, "n_shadow": 0, "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": 0.29987, "trunc_abs_frac": 8.39865, "mean_delta_full": 0.1897, "read_hit_rate": {"split": 0.90729}, "mean_gate": {"split": 0.1577}, "n_reads": 10599, "n_cand": 11682, "read_suppressed_frac": 0.0983, "dropout_reads_logged": 10599, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 295.5125732421875, "n_util_obs": 351, "util_count_spearman": -0.042298269036039095}, "probe_overhead_sec_total": 133.668, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.008466, "alpha": 0.03, "t": 624.9} |
| {"event": "consolidate", "step": 1950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 639.2} |
| {"event": "train", "step": 1950, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.48676, "bpb": 2.14494, "chunk_loss": 0.030613, "F": 0.20627, "G": 0.22604, "chunks": 211.22, "bytes_per_chunk": 4.848, "boundary_entropy": 0.16369, "router_noise": 0.0, "grad_norm": 0.7741, "lr": 0.0002479, "router_trainable": true, "sec_per_step": 0.29, "probe_sec": 0.0374, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02859, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.03952, "n_shadow": 0, "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": 0.53894, "trunc_abs_frac": 2.47182, "mean_delta_full": -0.2865, "read_hit_rate": {"split": 0.90458}, "mean_gate": {"split": 0.15907}, "n_reads": 10589, "n_cand": 11706, "read_suppressed_frac": 0.0993, "dropout_reads_logged": 10589, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 303.59771728515625, "n_util_obs": 360, "util_count_spearman": -0.0291769301658529}, "probe_overhead_sec_total": 136.934, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.008308, "alpha": 0.03, "t": 639.4} |
| {"event": "consolidate", "step": 2000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 652.9} |
| {"event": "probe", "step": 2000, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3425.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.196805, "cps": 26.757812, "mean_p": 0.216535, "boundary_entropy": 0.152925, "bpb": 2.105419}, "tr": {"chunks": 3542.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.195658, "cps": 27.671875, "mean_p": 0.209364, "boundary_entropy": 0.185549, "bpb": 2.308958}, "fi": {"chunks": 4252.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.226038, "cps": 33.21875, "mean_p": 0.237329, "boundary_entropy": 0.170267, "bpb": 2.471907}}, "bpb": {"en": 2.10542, "tr": 2.30896, "fi": 2.47191}, "gini_cpc": 0.032746, "gini_cps": 0.049143, "bpic": 5.056779, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9941720511702903, "fi": 1.148536805029826}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 311.5474853515625, "n_util_obs": 368, "util_count_spearman": -0.023885066391618803}, "t": 653.9} |
| {"event": "eval", "step": 2000, "phase": 1, "bpb": {"en": 2.11483, "tr": 2.32122, "fi": 2.4796}, "bytes_per_chunk": {"en": 4.879, "tr": 5.635, "fi": 4.672}, "bpb_mean": 2.30522, "frozen_inventory": true, "t": 655.4} |
| {"event": "train", "step": 2000, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.64572, "bpb": 2.37427, "chunk_loss": 0.030752, "F": 0.21042, "G": 0.23248, "chunks": 215.47, "bytes_per_chunk": 4.752, "boundary_entropy": 0.16443, "router_noise": 0.0, "grad_norm": 0.7721, "lr": 0.000245, "router_trainable": true, "sec_per_step": 0.343, "probe_sec": 0.1005, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.08846, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.03825, "n_shadow": 0, "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": 0.51928, "trunc_abs_frac": 2.16793, "mean_delta_full": -0.29647, "read_hit_rate": {"split": 0.89592}, "mean_gate": {"split": 0.15433}, "n_reads": 10631, "n_cand": 11866, "read_suppressed_frac": 0.1017, "dropout_reads_logged": 10631, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 311.70751953125, "n_util_obs": 368, "util_count_spearman": -0.022924430309986265}, "probe_overhead_sec_total": 140.716, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.008688, "alpha": 0.03, "t": 656.6} |
| {"event": "probe", "step": 2048, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3447.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.198069, "cps": 26.929688, "mean_p": 0.218416, "boundary_entropy": 0.150688, "bpb": 2.099123}, "tr": {"chunks": 3590.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.19831, "cps": 28.046875, "mean_p": 0.212259, "boundary_entropy": 0.187974, "bpb": 2.305282}, "fi": {"chunks": 4376.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.23263, "cps": 34.1875, "mean_p": 0.244015, "boundary_entropy": 0.172075, "bpb": 2.462212}}, "bpb": {"en": 2.09912, "tr": 2.30528, "fi": 2.46221}, "gini_cpc": 0.03663, "gini_cps": 0.054266, "bpic": 4.970823, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0012135855233568, "fi": 1.1744871422425234}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 319.35235595703125, "n_util_obs": 378, "util_count_spearman": 0.009158342848741216}, "t": 670.8} |
| {"event": "consolidate", "step": 2050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 672.7} |
| {"event": "train", "step": 2050, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.48312, "bpb": 2.13969, "chunk_loss": 0.030549, "F": 0.20126, "G": 0.22101, "chunks": 206.09, "bytes_per_chunk": 4.969, "boundary_entropy": 0.15713, "router_noise": 0.0, "grad_norm": 0.7599, "lr": 0.0002419, "router_trainable": true, "sec_per_step": 0.326, "probe_sec": 0.0621, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04987, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.00924, "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": 3, "trunc_ratio_sum": 9.47368, "trunc_abs_frac": 0.69875, "mean_delta_full": 0.01392, "read_hit_rate": {"split": 0.90302}, "mean_gate": {"split": 0.15796}, "n_reads": 10475, "n_cand": 11600, "read_suppressed_frac": 0.0952, "dropout_reads_logged": 10475, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 319.84490966796875, "n_util_obs": 379, "util_count_spearman": 0.0027036590323466485}, "probe_overhead_sec_total": 143.951, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.008387, "alpha": 0.03, "t": 672.9} |
| {"event": "consolidate", "step": 2100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 686.1} |
| {"event": "train", "step": 2100, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.54129, "bpb": 2.22362, "chunk_loss": 0.030446, "F": 0.19891, "G": 0.21906, "chunks": 203.69, "bytes_per_chunk": 5.027, "boundary_entropy": 0.15382, "router_noise": 0.0, "grad_norm": 0.7974, "lr": 0.0002388, "router_trainable": true, "sec_per_step": 0.271, "probe_sec": 0.1144, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10551, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.06011, "n_shadow": 0, "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": 1.40754, "trunc_abs_frac": 9.13291, "mean_delta_full": 0.05017, "read_hit_rate": {"split": 0.89232}, "mean_gate": {"split": 0.15089}, "n_reads": 10242, "n_cand": 11478, "read_suppressed_frac": 0.0989, "dropout_reads_logged": 10242, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 327.9566650390625, "n_util_obs": 383, "util_count_spearman": -0.03302133029404126}, "probe_overhead_sec_total": 147.572, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.00881, "alpha": 0.03, "t": 686.4} |
| {"event": "consolidate", "step": 2150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 701.3} |
| {"event": "train", "step": 2150, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.49709, "bpb": 2.15985, "chunk_loss": 0.030505, "F": 0.20282, "G": 0.22302, "chunks": 207.69, "bytes_per_chunk": 4.93, "boundary_entropy": 0.15156, "router_noise": 0.0, "grad_norm": 0.7146, "lr": 0.0002357, "router_trainable": true, "sec_per_step": 0.304, "probe_sec": 0.0645, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04769, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.06236, "n_shadow": 0, "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": 0.74678, "trunc_abs_frac": 1.59365, "mean_delta_full": 0.21228, "read_hit_rate": {"split": 0.90127}, "mean_gate": {"split": 0.15096}, "n_reads": 10653, "n_cand": 11820, "read_suppressed_frac": 0.1059, "dropout_reads_logged": 10653, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 336.1172790527344, "n_util_obs": 397, "util_count_spearman": -0.0023934654939901066}, "probe_overhead_sec_total": 151.612, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.008458, "alpha": 0.03, "t": 701.6} |
| {"event": "consolidate", "step": 2200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 715.4} |
| {"event": "probe", "step": 2200, "phase": 1, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3410.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.195943, "cps": 26.640625, "mean_p": 0.213819, "boundary_entropy": 0.145877, "bpb": 2.081936}, "tr": {"chunks": 3899.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.215379, "cps": 30.460938, "mean_p": 0.21801, "boundary_entropy": 0.178527, "bpb": 2.285717}, "fi": {"chunks": 4487.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.238531, "cps": 35.054688, "mean_p": 0.241287, "boundary_entropy": 0.168251, "bpb": 2.437276}}, "bpb": {"en": 2.08194, "tr": 2.28572, "fi": 2.43728}, "gini_cpc": 0.043689, "gini_cps": 0.060868, "bpic": 4.809427, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0991891300400138, "fi": 1.217345703541838}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 344.1043395996094, "n_util_obs": 406, "util_count_spearman": -0.019417003949853993}, "t": 716.3} |
| {"event": "eval", "step": 2200, "phase": 1, "bpb": {"en": 2.09008, "tr": 2.29422, "fi": 2.44484}, "bytes_per_chunk": {"en": 4.886, "tr": 4.987, "fi": 4.245}, "bpb_mean": 2.27638, "frozen_inventory": true, "t": 717.8} |
| {"event": "train", "step": 2200, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.47448, "bpb": 2.12723, "chunk_loss": 0.030488, "F": 0.20383, "G": 0.22215, "chunks": 208.72, "bytes_per_chunk": 4.906, "boundary_entropy": 0.14979, "router_noise": 0.0, "grad_norm": 0.6836, "lr": 0.0002325, "router_trainable": true, "sec_per_step": 0.347, "probe_sec": 0.0416, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03171, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.03368, "n_shadow": 0, "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": 0.33161, "trunc_abs_frac": 0.52545, "mean_delta_full": -0.2327, "read_hit_rate": {"split": 0.91055}, "mean_gate": {"split": 0.14698}, "n_reads": 10759, "n_cand": 11816, "read_suppressed_frac": 0.0983, "dropout_reads_logged": 10759, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 344.2690124511719, "n_util_obs": 406, "util_count_spearman": -0.019993395992212146}, "probe_overhead_sec_total": 155.07, "probe_overhead_seq_frac": 0.4997, "util_loss": 0.007861, "alpha": 0.03, "t": 719.0} |
| {"event": "consolidate", "step": 2250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 731.7} |
| {"event": "train", "step": 2250, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.4992, "bpb": 2.16289, "chunk_loss": 0.030533, "F": 0.20395, "G": 0.22131, "chunks": 208.84, "bytes_per_chunk": 4.903, "boundary_entropy": 0.14877, "router_noise": 0.0, "grad_norm": 0.7165, "lr": 0.0002293, "router_trainable": true, "sec_per_step": 0.259, "probe_sec": 0.0493, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03884, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.0191, "n_shadow": 0, "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": 0.45789, "trunc_abs_frac": 14.25577, "mean_delta_full": 0.12103, "read_hit_rate": {"split": 0.90329}, "mean_gate": {"split": 0.15163}, "n_reads": 10498, "n_cand": 11622, "read_suppressed_frac": 0.0987, "dropout_reads_logged": 10498, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 352.3584289550781, "n_util_obs": 413, "util_count_spearman": -0.005814188210271274}, "probe_overhead_sec_total": 158.804, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.008379, "alpha": 0.03, "t": 731.9} |
| {"event": "consolidate", "step": 2300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 745.4} |
| {"event": "train", "step": 2300, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.51402, "bpb": 2.18427, "chunk_loss": 0.030488, "F": 0.20392, "G": 0.2203, "chunks": 208.81, "bytes_per_chunk": 4.904, "boundary_entropy": 0.14175, "router_noise": 0.0, "grad_norm": 0.7568, "lr": 0.000226, "router_trainable": true, "sec_per_step": 0.276, "probe_sec": 0.0762, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06049, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.02773, "n_shadow": 0, "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": 0.06438, "trunc_abs_frac": 1.53576, "mean_delta_full": 0.16654, "read_hit_rate": {"split": 0.90178}, "mean_gate": {"split": 0.14995}, "n_reads": 10659, "n_cand": 11820, "read_suppressed_frac": 0.0957, "dropout_reads_logged": 10659, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 360.43548583984375, "n_util_obs": 423, "util_count_spearman": 0.02393276615974866}, "probe_overhead_sec_total": 162.64, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.008662, "alpha": 0.03, "t": 745.7} |
| {"event": "consolidate", "step": 2350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 758.6} |
| {"event": "train", "step": 2350, "arch": "ali", "phase": 1, "objective": "global", "lm_loss": 1.56364, "bpb": 2.25586, "chunk_loss": 0.030515, "F": 0.20636, "G": 0.22225, "chunks": 211.31, "bytes_per_chunk": 4.846, "boundary_entropy": 0.14084, "router_noise": 0.0, "grad_norm": 0.7685, "lr": 0.0002226, "router_trainable": true, "sec_per_step": 0.263, "probe_sec": 0.1385, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.12465, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.03205, "n_shadow": 0, "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": 0.97678, "trunc_abs_frac": 0.97918, "mean_delta_full": -0.07458, "read_hit_rate": {"split": 0.89919}, "mean_gate": {"split": 0.14317}, "n_reads": 10641, "n_cand": 11834, "read_suppressed_frac": 0.1022, "dropout_reads_logged": 10641, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 368.533447265625, "n_util_obs": 431, "util_count_spearman": 0.020487783931009935}, "probe_overhead_sec_total": 166.19, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.008567, "alpha": 0.03, "t": 758.9} |
| {"event": "consolidate", "step": 2400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 772.1} |
| {"event": "probe", "step": 2400, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3467.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.199219, "cps": 27.085938, "mean_p": 0.212543, "boundary_entropy": 0.126488, "bpb": 2.058728}, "tr": {"chunks": 3556.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.196432, "cps": 27.78125, "mean_p": 0.205018, "boundary_entropy": 0.162431, "bpb": 2.253941}, "fi": {"chunks": 4300.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.22859, "cps": 33.59375, "mean_p": 0.234432, "boundary_entropy": 0.144408, "bpb": 2.389031}}, "bpb": {"en": 2.05873, "tr": 2.25394, "fi": 2.38903}, "gini_cpc": 0.034344, "gini_cps": 0.049045, "bpic": 5.010333, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9860103629997504, "fi": 1.1474317178469409}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 376.52813720703125, "n_util_obs": 439, "util_count_spearman": 0.005350806156214193}, "t": 773.0} |
| {"event": "eval", "step": 2400, "phase": 1, "bpb": {"en": 2.07048, "tr": 2.26139, "fi": 2.39887}, "bytes_per_chunk": {"en": 4.811, "tr": 5.626, "fi": 4.625}, "bpb_mean": 2.24358, "frozen_inventory": true, "t": 774.5} |
| {"event": "phase_transition", "step": 2400, "phase": 2, "objective": "global", "router_trainable": true, "encoder_trainable": true, "frozen_router_params": 0, "t": 775.4} |
| {"event": "train", "step": 2400, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.49892, "bpb": 2.16249, "chunk_loss": 0.03061, "F": 0.20972, "G": 0.22487, "chunks": 214.75, "bytes_per_chunk": 4.768, "boundary_entropy": 0.1342, "router_noise": 0.0, "grad_norm": 0.7153, "lr": 0.0002192, "router_trainable": true, "sec_per_step": 0.336, "probe_sec": 0.084, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07339, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.0953, "n_shadow": 0, "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": 0.95681, "trunc_abs_frac": 1.07405, "mean_delta_full": 0.17571, "read_hit_rate": {"split": 0.91338}, "mean_gate": {"split": 0.14412}, "n_reads": 11125, "n_cand": 12180, "read_suppressed_frac": 0.0965, "dropout_reads_logged": 11125, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 376.69879150390625, "n_util_obs": 439, "util_count_spearman": 0.005667742253359437}, "probe_overhead_sec_total": 169.366, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.008444, "alpha": 0.03, "t": 775.7} |
| {"event": "consolidate", "step": 2450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 788.4} |
| {"event": "train", "step": 2450, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.46199, "bpb": 2.1092, "chunk_loss": 0.030493, "F": 0.2023, "G": 0.22073, "chunks": 207.16, "bytes_per_chunk": 4.943, "boundary_entropy": 0.14174, "router_noise": 0.0, "grad_norm": 0.705, "lr": 0.0002158, "router_trainable": true, "sec_per_step": 0.259, "probe_sec": 0.044, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03398, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.05029, "n_shadow": 0, "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": -0.21863, "trunc_abs_frac": 0.36927, "mean_delta_full": 0.1474, "read_hit_rate": {"split": 0.90462}, "mean_gate": {"split": 0.14281}, "n_reads": 10604, "n_cand": 11722, "read_suppressed_frac": 0.1046, "dropout_reads_logged": 10604, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 384.847412109375, "n_util_obs": 443, "util_count_spearman": 0.010821799835420521}, "probe_overhead_sec_total": 172.78, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.008899, "alpha": 0.03, "t": 788.6} |
| {"event": "consolidate", "step": 2500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 802.2} |
| {"event": "train", "step": 2500, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.43509, "bpb": 2.07039, "chunk_loss": 0.030584, "F": 0.21017, "G": 0.22396, "chunks": 215.22, "bytes_per_chunk": 4.758, "boundary_entropy": 0.12901, "router_noise": 0.0, "grad_norm": 0.744, "lr": 0.0002123, "router_trainable": true, "sec_per_step": 0.275, "probe_sec": 0.0505, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04026, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.06756, "n_shadow": 0, "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": 0.24991, "trunc_abs_frac": 4.87653, "mean_delta_full": 0.08511, "read_hit_rate": {"split": 0.91844}, "mean_gate": {"split": 0.13708}, "n_reads": 11194, "n_cand": 12188, "read_suppressed_frac": 0.0967, "dropout_reads_logged": 11194, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 393.0103759765625, "n_util_obs": 450, "util_count_spearman": 0.03442809337792552}, "probe_overhead_sec_total": 175.812, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.008897, "alpha": 0.03, "t": 802.4} |
| {"event": "consolidate", "step": 2550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 814.9} |
| {"event": "train", "step": 2550, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.46327, "bpb": 2.11106, "chunk_loss": 0.030453, "F": 0.20172, "G": 0.21782, "chunks": 206.56, "bytes_per_chunk": 4.957, "boundary_entropy": 0.13194, "router_noise": 0.0, "grad_norm": 0.6981, "lr": 0.0002088, "router_trainable": true, "sec_per_step": 0.26, "probe_sec": 0.0508, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03632, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.09909, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 5, "trunc_ratio_sum": 1.06424, "trunc_abs_frac": 0.56221, "mean_delta_full": 0.18726, "read_hit_rate": {"split": 0.89985}, "mean_gate": {"split": 0.13762}, "n_reads": 10575, "n_cand": 11752, "read_suppressed_frac": 0.1012, "dropout_reads_logged": 10575, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 401.11773681640625, "n_util_obs": 454, "util_count_spearman": 0.02688721931206296}, "probe_overhead_sec_total": 179.06, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.008061, "alpha": 0.03, "t": 815.4} |
| {"event": "consolidate", "step": 2600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 832.0} |
| {"event": "probe", "step": 2600, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3453.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.198414, "cps": 26.976562, "mean_p": 0.211117, "boundary_entropy": 0.120776, "bpb": 2.037465}, "tr": {"chunks": 3602.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.198973, "cps": 28.140625, "mean_p": 0.204912, "boundary_entropy": 0.151645, "bpb": 2.237073}, "fi": {"chunks": 4334.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.230397, "cps": 33.859375, "mean_p": 0.234579, "boundary_entropy": 0.139357, "bpb": 2.364262}}, "bpb": {"en": 2.03746, "tr": 2.23707, "fi": 2.36426}, "gini_cpc": 0.033964, "gini_cps": 0.05157, "bpic": 4.981298, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0028147170023756, "fi": 1.1611934178483383}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 409.15545654296875, "n_util_obs": 458, "util_count_spearman": 0.02076723278495201}, "t": 832.8} |
| {"event": "eval", "step": 2600, "phase": 2, "bpb": {"en": 2.05016, "tr": 2.24816, "fi": 2.37132}, "bytes_per_chunk": {"en": 4.788, "tr": 5.626, "fi": 4.564}, "bpb_mean": 2.22321, "frozen_inventory": true, "t": 834.2} |
| {"event": "train", "step": 2600, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.42021, "bpb": 2.04893, "chunk_loss": 0.030487, "F": 0.20493, "G": 0.21799, "chunks": 209.84, "bytes_per_chunk": 4.88, "boundary_entropy": 0.12502, "router_noise": 0.0, "grad_norm": 0.7365, "lr": 0.0002053, "router_trainable": true, "sec_per_step": 0.4, "probe_sec": 0.1057, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09158, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.00337, "n_shadow": 0, "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": 1.06568, "trunc_abs_frac": 1.05416, "mean_delta_full": 0.07666, "read_hit_rate": {"split": 0.91202}, "mean_gate": {"split": 0.13752}, "n_reads": 10729, "n_cand": 11764, "read_suppressed_frac": 0.1014, "dropout_reads_logged": 10729, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 409.31988525390625, "n_util_obs": 458, "util_count_spearman": 0.020408560291235747}, "probe_overhead_sec_total": 183.084, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.00868, "alpha": 0.03, "t": 835.4} |
| {"event": "consolidate", "step": 2650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 849.4} |
| {"event": "train", "step": 2650, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.46566, "bpb": 2.11449, "chunk_loss": 0.030387, "F": 0.19861, "G": 0.21322, "chunks": 203.38, "bytes_per_chunk": 5.035, "boundary_entropy": 0.13291, "router_noise": 0.0, "grad_norm": 0.6668, "lr": 0.0002017, "router_trainable": true, "sec_per_step": 0.284, "probe_sec": 0.0397, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03049, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.04959, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": 0.69699, "trunc_abs_frac": 0.91523, "mean_delta_full": 0.22319, "read_hit_rate": {"split": 0.89993}, "mean_gate": {"split": 0.14384}, "n_reads": 10360, "n_cand": 11512, "read_suppressed_frac": 0.1016, "dropout_reads_logged": 10360, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 417.4539794921875, "n_util_obs": 464, "util_count_spearman": 0.024520229674481843}, "probe_overhead_sec_total": 186.513, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.008497, "alpha": 0.03, "t": 849.6} |
| {"event": "consolidate", "step": 2700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 862.8} |
| {"event": "train", "step": 2700, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.44706, "bpb": 2.08766, "chunk_loss": 0.030472, "F": 0.20444, "G": 0.21676, "chunks": 209.34, "bytes_per_chunk": 4.891, "boundary_entropy": 0.1279, "router_noise": 0.0, "grad_norm": 0.6738, "lr": 0.0001981, "router_trainable": true, "sec_per_step": 0.273, "probe_sec": 0.159, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.14624, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.00539, "n_shadow": 0, "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": -0.35857, "trunc_abs_frac": 0.63598, "mean_delta_full": 0.05889, "read_hit_rate": {"split": 0.90248}, "mean_gate": {"split": 0.13694}, "n_reads": 10689, "n_cand": 11844, "read_suppressed_frac": 0.0976, "dropout_reads_logged": 10689, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 425.57861328125, "n_util_obs": 473, "util_count_spearman": 0.038112583448883304}, "probe_overhead_sec_total": 189.73, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.008386, "alpha": 0.03, "t": 863.2} |
| {"event": "consolidate", "step": 2750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 878.6} |
| {"event": "train", "step": 2750, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.45251, "bpb": 2.09553, "chunk_loss": 0.030575, "F": 0.20673, "G": 0.22126, "chunks": 211.69, "bytes_per_chunk": 4.837, "boundary_entropy": 0.12713, "router_noise": 0.0, "grad_norm": 0.6742, "lr": 0.0001945, "router_trainable": true, "sec_per_step": 0.317, "probe_sec": 0.0561, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0319, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.00341, "n_shadow": 0, "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": 0.15422, "trunc_abs_frac": 0.57148, "mean_delta_full": 0.21475, "read_hit_rate": {"split": 0.91143}, "mean_gate": {"split": 0.13896}, "n_reads": 10826, "n_cand": 11878, "read_suppressed_frac": 0.0984, "dropout_reads_logged": 10826, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 433.7032470703125, "n_util_obs": 479, "util_count_spearman": 0.014367342000784275}, "probe_overhead_sec_total": 193.553, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.009201, "alpha": 0.03, "t": 879.1} |
| {"event": "consolidate", "step": 2800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 897.5} |
| {"event": "probe", "step": 2800, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3383.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.194392, "cps": 26.429688, "mean_p": 0.205954, "boundary_entropy": 0.119577, "bpb": 2.017669}, "tr": {"chunks": 3534.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.195216, "cps": 27.609375, "mean_p": 0.200539, "boundary_entropy": 0.145501, "bpb": 2.196678}, "fi": {"chunks": 4293.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.228218, "cps": 33.539062, "mean_p": 0.232727, "boundary_entropy": 0.132534, "bpb": 2.322866}}, "bpb": {"en": 2.01767, "tr": 2.19668, "fi": 2.32287}, "gini_cpc": 0.0365, "gini_cps": 0.054118, "bpic": 5.060839, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0042413881913834, "fi": 1.1740081922934478}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 441.61334228515625, "n_util_obs": 484, "util_count_spearman": 0.0066895216104850734}, "t": 898.4} |
| {"event": "eval", "step": 2800, "phase": 2, "bpb": {"en": 2.02903, "tr": 2.20756, "fi": 2.3302}, "bytes_per_chunk": {"en": 4.88, "tr": 5.675, "fi": 4.554}, "bpb_mean": 2.18893, "frozen_inventory": true, "t": 899.8} |
| {"event": "train", "step": 2800, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.40988, "bpb": 2.03403, "chunk_loss": 0.030433, "F": 0.20282, "G": 0.21527, "chunks": 207.69, "bytes_per_chunk": 4.93, "boundary_entropy": 0.12169, "router_noise": 0.0, "grad_norm": 0.6929, "lr": 0.0001909, "router_trainable": true, "sec_per_step": 0.469, "probe_sec": 0.0764, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03716, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.00953, "n_shadow": 0, "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": 0.63064, "trunc_abs_frac": 0.52348, "mean_delta_full": -0.14868, "read_hit_rate": {"split": 0.91222}, "mean_gate": {"split": 0.13905}, "n_reads": 10693, "n_cand": 11722, "read_suppressed_frac": 0.0968, "dropout_reads_logged": 10693, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 441.77728271484375, "n_util_obs": 484, "util_count_spearman": 0.006581194713177835}, "probe_overhead_sec_total": 197.339, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.009574, "alpha": 0.03, "t": 902.5} |
| {"event": "consolidate", "step": 2850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 926.8} |
| {"event": "train", "step": 2850, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.46159, "bpb": 2.10863, "chunk_loss": 0.030634, "F": 0.2131, "G": 0.22396, "chunks": 218.22, "bytes_per_chunk": 4.693, "boundary_entropy": 0.11827, "router_noise": 0.0, "grad_norm": 0.7038, "lr": 0.0001872, "router_trainable": true, "sec_per_step": 0.489, "probe_sec": 0.0406, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0291, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.05927, "n_shadow": 0, "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": 2.00038, "trunc_abs_frac": 3.47831, "mean_delta_full": 0.10681, "read_hit_rate": {"split": 0.90925}, "mean_gate": {"split": 0.12971}, "n_reads": 11191, "n_cand": 12308, "read_suppressed_frac": 0.0977, "dropout_reads_logged": 11191, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 449.9207458496094, "n_util_obs": 492, "util_count_spearman": 0.006600176884015029}, "probe_overhead_sec_total": 201.005, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.009666, "alpha": 0.03, "t": 927.0} |
| {"event": "consolidate", "step": 2900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 941.0} |
| {"event": "train", "step": 2900, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.39061, "bpb": 2.00623, "chunk_loss": 0.030361, "F": 0.19849, "G": 0.21028, "chunks": 203.25, "bytes_per_chunk": 5.038, "boundary_entropy": 0.12155, "router_noise": 0.0, "grad_norm": 0.6548, "lr": 0.0001835, "router_trainable": true, "sec_per_step": 0.286, "probe_sec": 0.074, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06542, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.02212, "n_shadow": 0, "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": -0.14954, "trunc_abs_frac": 0.38114, "mean_delta_full": -0.23136, "read_hit_rate": {"split": 0.90044}, "mean_gate": {"split": 0.13264}, "n_reads": 10346, "n_cand": 11490, "read_suppressed_frac": 0.0959, "dropout_reads_logged": 10346, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 458.05841064453125, "n_util_obs": 501, "util_count_spearman": -0.006777155823721589}, "probe_overhead_sec_total": 204.057, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.009121, "alpha": 0.03, "t": 941.2} |
| {"event": "consolidate", "step": 2950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 956.2} |
| {"event": "train", "step": 2950, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.5395, "bpb": 2.22103, "chunk_loss": 0.03063, "F": 0.2067, "G": 0.21964, "chunks": 211.66, "bytes_per_chunk": 4.838, "boundary_entropy": 0.12296, "router_noise": 0.0, "grad_norm": 0.6964, "lr": 0.0001799, "router_trainable": true, "sec_per_step": 0.315, "probe_sec": 0.0772, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03778, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.09301, "n_shadow": 0, "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": -0.11008, "trunc_abs_frac": 0.71293, "mean_delta_full": 0.17013, "read_hit_rate": {"split": 0.89243}, "mean_gate": {"split": 0.13065}, "n_reads": 10470, "n_cand": 11732, "read_suppressed_frac": 0.0995, "dropout_reads_logged": 10470, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 466.14447021484375, "n_util_obs": 506, "util_count_spearman": -0.006229030992229982}, "probe_overhead_sec_total": 207.316, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.009581, "alpha": 0.03, "t": 957.0} |
| {"event": "consolidate", "step": 3000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 978.4} |
| {"event": "probe", "step": 3000, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3350.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.192496, "cps": 26.171875, "mean_p": 0.204239, "boundary_entropy": 0.116097, "bpb": 1.999433}, "tr": {"chunks": 3566.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.196984, "cps": 27.859375, "mean_p": 0.198054, "boundary_entropy": 0.140022, "bpb": 2.175209}, "fi": {"chunks": 4269.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.226942, "cps": 33.351562, "mean_p": 0.2314, "boundary_entropy": 0.131951, "bpb": 2.302407}}, "bpb": {"en": 1.99943, "tr": 2.17521, "fi": 2.30241}, "gini_cpc": 0.037254, "gini_cps": 0.054776, "bpic": 5.07215, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0233167917249635, "fi": 1.1789451075386979}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 474.1201477050781, "n_util_obs": 510, "util_count_spearman": -0.0017094850980134979}, "t": 979.3} |
| {"event": "eval", "step": 3000, "phase": 2, "bpb": {"en": 2.01292, "tr": 2.18525, "fi": 2.30714}, "bytes_per_chunk": {"en": 4.898, "tr": 5.649, "fi": 4.57}, "bpb_mean": 2.16844, "frozen_inventory": true, "t": 980.5} |
| {"event": "train", "step": 3000, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.48787, "bpb": 2.14654, "chunk_loss": 0.030434, "F": 0.20087, "G": 0.21205, "chunks": 205.69, "bytes_per_chunk": 4.978, "boundary_entropy": 0.12, "router_noise": 0.0, "grad_norm": 0.6983, "lr": 0.0001761, "router_trainable": true, "sec_per_step": 0.508, "probe_sec": 0.0628, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04981, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.09078, "n_shadow": 0, "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": -0.53314, "trunc_abs_frac": 1.33894, "mean_delta_full": 0.09668, "read_hit_rate": {"split": 0.89429}, "mean_gate": {"split": 0.12681}, "n_reads": 10313, "n_cand": 11532, "read_suppressed_frac": 0.0929, "dropout_reads_logged": 10313, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 474.27618408203125, "n_util_obs": 511, "util_count_spearman": 0.00031323752294141604}, "probe_overhead_sec_total": 211.123, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.008671, "alpha": 0.03, "t": 982.4} |
| {"event": "consolidate", "step": 3050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1005.0} |
| {"event": "train", "step": 3050, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.43709, "bpb": 2.07329, "chunk_loss": 0.030489, "F": 0.20294, "G": 0.21384, "chunks": 207.81, "bytes_per_chunk": 4.928, "boundary_entropy": 0.11918, "router_noise": 0.0, "grad_norm": 0.6819, "lr": 0.0001724, "router_trainable": true, "sec_per_step": 0.457, "probe_sec": 0.0741, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06508, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.03053, "n_shadow": 0, "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": -0.11603, "trunc_abs_frac": 0.71331, "mean_delta_full": -0.07991, "read_hit_rate": {"split": 0.90486}, "mean_gate": {"split": 0.13226}, "n_reads": 10690, "n_cand": 11814, "read_suppressed_frac": 0.0973, "dropout_reads_logged": 10690, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 482.33612060546875, "n_util_obs": 518, "util_count_spearman": -0.0025669090487612785}, "probe_overhead_sec_total": 214.653, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.010487, "alpha": 0.03, "t": 1005.2} |
| {"event": "consolidate", "step": 3100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1019.4} |
| {"event": "train", "step": 3100, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.4885, "bpb": 2.14745, "chunk_loss": 0.030361, "F": 0.1973, "G": 0.20896, "chunks": 202.03, "bytes_per_chunk": 5.069, "boundary_entropy": 0.11467, "router_noise": 0.0, "grad_norm": 0.6749, "lr": 0.0001687, "router_trainable": true, "sec_per_step": 0.288, "probe_sec": 0.0496, "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": -0.0094, "n_shadow": 0, "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": -11.37684, "trunc_abs_frac": 0.70962, "mean_delta_full": -0.00415, "read_hit_rate": {"split": 0.89549}, "mean_gate": {"split": 0.1292}, "n_reads": 10162, "n_cand": 11348, "read_suppressed_frac": 0.0968, "dropout_reads_logged": 10162, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 490.41265869140625, "n_util_obs": 524, "util_count_spearman": -0.014843441371098576}, "probe_overhead_sec_total": 218.162, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.010196, "alpha": 0.03, "t": 1019.6} |
| {"event": "consolidate", "step": 3150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1033.0} |
| {"event": "train", "step": 3150, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.44715, "bpb": 2.0878, "chunk_loss": 0.030446, "F": 0.20267, "G": 0.2117, "chunks": 207.53, "bytes_per_chunk": 4.934, "boundary_entropy": 0.11562, "router_noise": 0.0, "grad_norm": 0.6918, "lr": 0.000165, "router_trainable": true, "sec_per_step": 0.274, "probe_sec": 0.0369, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02735, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.03539, "n_shadow": 0, "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": -0.97375, "trunc_abs_frac": 0.7782, "mean_delta_full": -0.10725, "read_hit_rate": {"split": 0.89983}, "mean_gate": {"split": 0.12894}, "n_reads": 10537, "n_cand": 11710, "read_suppressed_frac": 0.0953, "dropout_reads_logged": 10537, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 498.5524597167969, "n_util_obs": 528, "util_count_spearman": -0.01293140915552086}, "probe_overhead_sec_total": 221.84, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.009682, "alpha": 0.03, "t": 1033.3} |
| {"event": "consolidate", "step": 3200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1046.2} |
| {"event": "probe", "step": 3200, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3398.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.195254, "cps": 26.546875, "mean_p": 0.203942, "boundary_entropy": 0.10574, "bpb": 1.982368}, "tr": {"chunks": 3580.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.197757, "cps": 27.96875, "mean_p": 0.196087, "boundary_entropy": 0.126604, "bpb": 2.151401}, "fi": {"chunks": 4356.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.231567, "cps": 34.03125, "mean_p": 0.2332, "boundary_entropy": 0.123912, "bpb": 2.281467}}, "bpb": {"en": 1.98237, "tr": 2.1514, "fi": 2.28147}, "gini_cpc": 0.03876, "gini_cps": 0.05635, "bpic": 5.00547, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.012822220582848, "fi": 1.1859782742048948}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 506.51654052734375, "n_util_obs": 539, "util_count_spearman": -0.00790145055980056}, "t": 1047.1} |
| {"event": "eval", "step": 3200, "phase": 2, "bpb": {"en": 1.99533, "tr": 2.15857, "fi": 2.28784}, "bytes_per_chunk": {"en": 4.854, "tr": 5.627, "fi": 4.531}, "bpb_mean": 2.14725, "frozen_inventory": true, "t": 1048.6} |
| {"event": "train", "step": 3200, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.42487, "bpb": 2.05565, "chunk_loss": 0.030471, "F": 0.20636, "G": 0.21483, "chunks": 211.31, "bytes_per_chunk": 4.846, "boundary_entropy": 0.10621, "router_noise": 0.0, "grad_norm": 0.6258, "lr": 0.0001613, "router_trainable": true, "sec_per_step": 0.33, "probe_sec": 0.0406, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03123, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.09987, "n_shadow": 0, "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": 1.02326, "trunc_abs_frac": 0.95055, "mean_delta_full": 0.21844, "read_hit_rate": {"split": 0.90281}, "mean_gate": {"split": 0.1273}, "n_reads": 10599, "n_cand": 11740, "read_suppressed_frac": 0.1041, "dropout_reads_logged": 10599, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 506.67889404296875, "n_util_obs": 539, "util_count_spearman": -0.007827498838529363}, "probe_overhead_sec_total": 224.846, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.009301, "alpha": 0.03, "t": 1049.8} |
| {"event": "consolidate", "step": 3250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1063.4} |
| {"event": "train", "step": 3250, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.43166, "bpb": 2.06545, "chunk_loss": 0.03043, "F": 0.20047, "G": 0.21061, "chunks": 205.28, "bytes_per_chunk": 4.988, "boundary_entropy": 0.11202, "router_noise": 0.0, "grad_norm": 0.7207, "lr": 0.0001576, "router_trainable": true, "sec_per_step": 0.277, "probe_sec": 0.0718, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05516, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.00914, "n_shadow": 0, "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": 0.60216, "trunc_abs_frac": 2.89469, "mean_delta_full": -0.16084, "read_hit_rate": {"split": 0.89998}, "mean_gate": {"split": 0.12462}, "n_reads": 10411, "n_cand": 11568, "read_suppressed_frac": 0.0988, "dropout_reads_logged": 10411, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 514.79052734375, "n_util_obs": 545, "util_count_spearman": -0.02302800035426686}, "probe_overhead_sec_total": 228.289, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.010136, "alpha": 0.03, "t": 1063.7} |
| {"event": "consolidate", "step": 3300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1076.8} |
| {"event": "train", "step": 3300, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.40895, "bpb": 2.03269, "chunk_loss": 0.030558, "F": 0.20816, "G": 0.21674, "chunks": 213.16, "bytes_per_chunk": 4.804, "boundary_entropy": 0.1086, "router_noise": 0.0, "grad_norm": 0.6502, "lr": 0.0001539, "router_trainable": true, "sec_per_step": 0.266, "probe_sec": 0.0528, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04061, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.0049, "n_shadow": 0, "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": 0.10574, "trunc_abs_frac": 1.20484, "mean_delta_full": 0.18918, "read_hit_rate": {"split": 0.90661}, "mean_gate": {"split": 0.12279}, "n_reads": 10843, "n_cand": 11960, "read_suppressed_frac": 0.1007, "dropout_reads_logged": 10843, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 522.9075317382812, "n_util_obs": 554, "util_count_spearman": -0.038124199553585904}, "probe_overhead_sec_total": 231.714, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.008751, "alpha": 0.03, "t": 1077.0} |
| {"event": "consolidate", "step": 3350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1090.2} |
| {"event": "train", "step": 3350, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.35541, "bpb": 1.95545, "chunk_loss": 0.030483, "F": 0.20676, "G": 0.21462, "chunks": 211.72, "bytes_per_chunk": 4.837, "boundary_entropy": 0.10391, "router_noise": 0.0, "grad_norm": 0.7006, "lr": 0.0001501, "router_trainable": true, "sec_per_step": 0.269, "probe_sec": 0.0827, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07149, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.01342, "n_shadow": 0, "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": -1.87024, "trunc_abs_frac": 0.95473, "mean_delta_full": 0.01497, "read_hit_rate": {"split": 0.91306}, "mean_gate": {"split": 0.12325}, "n_reads": 10880, "n_cand": 11916, "read_suppressed_frac": 0.1007, "dropout_reads_logged": 10880, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 531.0582275390625, "n_util_obs": 562, "util_count_spearman": -0.047566591582052835}, "probe_overhead_sec_total": 235.358, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.009865, "alpha": 0.03, "t": 1090.5} |
| {"event": "consolidate", "step": 3400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1103.6} |
| {"event": "probe", "step": 3400, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3408.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.195828, "cps": 26.625, "mean_p": 0.204022, "boundary_entropy": 0.098046, "bpb": 1.970779}, "tr": {"chunks": 3526.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.194774, "cps": 27.546875, "mean_p": 0.194041, "boundary_entropy": 0.124936, "bpb": 2.148804}, "fi": {"chunks": 4344.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.230929, "cps": 33.9375, "mean_p": 0.23109, "boundary_entropy": 0.116409, "bpb": 2.257693}}, "bpb": {"en": 1.97078, "tr": 2.1488, "fi": 2.25769}, "gini_cpc": 0.03878, "gini_cps": 0.055329, "bpic": 5.030325, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9946179452008966, "fi": 1.1792407199563335}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 538.9718017578125, "n_util_obs": 569, "util_count_spearman": -0.04025660029315805}, "t": 1104.6} |
| {"event": "eval", "step": 3400, "phase": 2, "bpb": {"en": 1.98316, "tr": 2.156, "fi": 2.2639}, "bytes_per_chunk": {"en": 4.826, "tr": 5.723, "fi": 4.56}, "bpb_mean": 2.13435, "frozen_inventory": true, "t": 1106.1} |
| {"event": "train", "step": 3400, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.41704, "bpb": 2.04436, "chunk_loss": 0.030411, "F": 0.20117, "G": 0.20968, "chunks": 206.0, "bytes_per_chunk": 4.971, "boundary_entropy": 0.10422, "router_noise": 0.0, "grad_norm": 0.6944, "lr": 0.0001465, "router_trainable": true, "sec_per_step": 0.341, "probe_sec": 0.3948, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03094, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.02315, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 4, "trunc_ratio_sum": 2.3132, "trunc_abs_frac": 1.3891, "mean_delta_full": 0.03006, "read_hit_rate": {"split": 0.89561}, "mean_gate": {"split": 0.11936}, "n_reads": 10416, "n_cand": 11630, "read_suppressed_frac": 0.1004, "dropout_reads_logged": 10416, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 539.1309814453125, "n_util_obs": 569, "util_count_spearman": -0.040074341347135674}, "probe_overhead_sec_total": 238.923, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.01087, "alpha": 0.03, "t": 1107.5} |
| {"event": "consolidate", "step": 3450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1120.6} |
| {"event": "train", "step": 3450, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.38578, "bpb": 1.99925, "chunk_loss": 0.030401, "F": 0.20102, "G": 0.20939, "chunks": 205.84, "bytes_per_chunk": 4.975, "boundary_entropy": 0.10378, "router_noise": 0.0, "grad_norm": 0.7629, "lr": 0.0001428, "router_trainable": true, "sec_per_step": 0.27, "probe_sec": 0.0864, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06853, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.00665, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -10.50773, "trunc_abs_frac": 37.78771, "mean_delta_full": -0.00395, "read_hit_rate": {"split": 0.90182}, "mean_gate": {"split": 0.11873}, "n_reads": 10425, "n_cand": 11560, "read_suppressed_frac": 0.1003, "dropout_reads_logged": 10425, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 547.2059326171875, "n_util_obs": 574, "util_count_spearman": -0.029118198281440726}, "probe_overhead_sec_total": 242.282, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.00962, "alpha": 0.03, "t": 1121.0} |
| {"event": "consolidate", "step": 3500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1133.9} |
| {"event": "train", "step": 3500, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.36349, "bpb": 1.9671, "chunk_loss": 0.030367, "F": 0.19904, "G": 0.20665, "chunks": 203.81, "bytes_per_chunk": 5.024, "boundary_entropy": 0.09857, "router_noise": 0.0, "grad_norm": 0.6818, "lr": 0.0001391, "router_trainable": true, "sec_per_step": 0.265, "probe_sec": 0.0461, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03686, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.01705, "n_shadow": 0, "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": 0.41199, "trunc_abs_frac": 0.39651, "mean_delta_full": 0.37402, "read_hit_rate": {"split": 0.90474}, "mean_gate": {"split": 0.12431}, "n_reads": 10419, "n_cand": 11516, "read_suppressed_frac": 0.0943, "dropout_reads_logged": 10419, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 555.3169555664062, "n_util_obs": 579, "util_count_spearman": -0.043080420711058796}, "probe_overhead_sec_total": 245.66, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.009151, "alpha": 0.03, "t": 1134.3} |
| {"event": "consolidate", "step": 3550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1147.3} |
| {"event": "train", "step": 3550, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.40418, "bpb": 2.0258, "chunk_loss": 0.030453, "F": 0.20291, "G": 0.2117, "chunks": 207.78, "bytes_per_chunk": 4.928, "boundary_entropy": 0.10642, "router_noise": 0.0, "grad_norm": 0.672, "lr": 0.0001355, "router_trainable": true, "sec_per_step": 0.264, "probe_sec": 0.0456, "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": -0.00665, "n_shadow": 0, "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": -2.28706, "trunc_abs_frac": 0.53317, "mean_delta_full": -0.04199, "read_hit_rate": {"split": 0.89959}, "mean_gate": {"split": 0.11982}, "n_reads": 10455, "n_cand": 11622, "read_suppressed_frac": 0.0949, "dropout_reads_logged": 10455, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 563.4141845703125, "n_util_obs": 581, "util_count_spearman": -0.04716986226123304}, "probe_overhead_sec_total": 249.29, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.009255, "alpha": 0.03, "t": 1147.5} |
| {"event": "consolidate", "step": 3600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1160.6} |
| {"event": "probe", "step": 3600, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3432.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.197207, "cps": 26.8125, "mean_p": 0.205428, "boundary_entropy": 0.103578, "bpb": 1.957742}, "tr": {"chunks": 3597.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.198696, "cps": 28.101562, "mean_p": 0.194223, "boundary_entropy": 0.118208, "bpb": 2.112455}, "fi": {"chunks": 4372.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.232417, "cps": 34.15625, "mean_p": 0.231454, "boundary_entropy": 0.116232, "bpb": 2.229589}}, "bpb": {"en": 1.95774, "tr": 2.11245, "fi": 2.22959}, "gini_cpc": 0.037359, "gini_cps": 0.054966, "bpic": 4.976055, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0075502785343695, "fi": 1.1785421266310108}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 571.4066772460938, "n_util_obs": 591, "util_count_spearman": -0.028419191208902222}, "t": 1161.6} |
| {"event": "eval", "step": 3600, "phase": 2, "bpb": {"en": 1.96983, "tr": 2.11969, "fi": 2.23654}, "bytes_per_chunk": {"en": 4.845, "tr": 5.605, "fi": 4.565}, "bpb_mean": 2.10869, "frozen_inventory": true, "t": 1162.9} |
| {"event": "train", "step": 3600, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.40119, "bpb": 2.0215, "chunk_loss": 0.030487, "F": 0.20596, "G": 0.21425, "chunks": 210.91, "bytes_per_chunk": 4.855, "boundary_entropy": 0.10558, "router_noise": 0.0, "grad_norm": 0.6304, "lr": 0.0001319, "router_trainable": true, "sec_per_step": 0.334, "probe_sec": 0.0666, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05692, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.01379, "n_shadow": 0, "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": -0.68343, "trunc_abs_frac": 0.24934, "mean_delta_full": -0.05058, "read_hit_rate": {"split": 0.90176}, "mean_gate": {"split": 0.11687}, "n_reads": 10731, "n_cand": 11900, "read_suppressed_frac": 0.1014, "dropout_reads_logged": 10731, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 571.570556640625, "n_util_obs": 591, "util_count_spearman": -0.02846499683947361}, "probe_overhead_sec_total": 252.873, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.009094, "alpha": 0.03, "t": 1164.2} |
| {"event": "consolidate", "step": 3650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1177.1} |
| {"event": "train", "step": 3650, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.43456, "bpb": 2.06964, "chunk_loss": 0.030486, "F": 0.20508, "G": 0.2137, "chunks": 210.0, "bytes_per_chunk": 4.876, "boundary_entropy": 0.10169, "router_noise": 0.0, "grad_norm": 0.6591, "lr": 0.0001283, "router_trainable": true, "sec_per_step": 0.264, "probe_sec": 0.0538, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04157, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.02782, "n_shadow": 0, "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": 0.09423, "trunc_abs_frac": 0.68313, "mean_delta_full": -0.25857, "read_hit_rate": {"split": 0.89988}, "mean_gate": {"split": 0.11373}, "n_reads": 10696, "n_cand": 11886, "read_suppressed_frac": 0.1003, "dropout_reads_logged": 10696, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 579.708984375, "n_util_obs": 597, "util_count_spearman": -0.02946837447931761}, "probe_overhead_sec_total": 255.769, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.00902, "alpha": 0.03, "t": 1177.4} |
| {"event": "consolidate", "step": 3700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1190.0} |
| {"event": "train", "step": 3700, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.38902, "bpb": 2.00393, "chunk_loss": 0.030455, "F": 0.20587, "G": 0.2137, "chunks": 210.81, "bytes_per_chunk": 4.857, "boundary_entropy": 0.1033, "router_noise": 0.0, "grad_norm": 0.6366, "lr": 0.0001247, "router_trainable": true, "sec_per_step": 0.257, "probe_sec": 0.0894, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07558, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.04691, "n_shadow": 0, "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": 1.46665, "trunc_abs_frac": 6.75853, "mean_delta_full": 0.09151, "read_hit_rate": {"split": 0.90701}, "mean_gate": {"split": 0.12506}, "n_reads": 10603, "n_cand": 11690, "read_suppressed_frac": 0.1063, "dropout_reads_logged": 10603, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 587.8734130859375, "n_util_obs": 603, "util_count_spearman": -0.017582902038791585}, "probe_overhead_sec_total": 259.344, "probe_overhead_seq_frac": 0.4998, "util_loss": 0.011354, "alpha": 0.03, "t": 1190.3} |
| {"event": "consolidate", "step": 3750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1203.0} |
| {"event": "train", "step": 3750, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.37772, "bpb": 1.98763, "chunk_loss": 0.03032, "F": 0.19794, "G": 0.20612, "chunks": 202.69, "bytes_per_chunk": 5.052, "boundary_entropy": 0.10127, "router_noise": 0.0, "grad_norm": 0.6745, "lr": 0.0001212, "router_trainable": true, "sec_per_step": 0.259, "probe_sec": 0.0588, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04696, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.04025, "n_shadow": 0, "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": 8.08163, "trunc_abs_frac": 8.08163, "mean_delta_full": 0.00598, "read_hit_rate": {"split": 0.89713}, "mean_gate": {"split": 0.11811}, "n_reads": 10308, "n_cand": 11490, "read_suppressed_frac": 0.1006, "dropout_reads_logged": 10308, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 595.9449462890625, "n_util_obs": 611, "util_count_spearman": -0.016973903669038615}, "probe_overhead_sec_total": 263.22, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009435, "alpha": 0.03, "t": 1203.2} |
| {"event": "consolidate", "step": 3800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1216.9} |
| {"event": "probe", "step": 3800, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3370.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.193645, "cps": 26.328125, "mean_p": 0.201148, "boundary_entropy": 0.09564, "bpb": 1.946893}, "tr": {"chunks": 3568.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.197094, "cps": 27.875, "mean_p": 0.19438, "boundary_entropy": 0.118103, "bpb": 2.104757}, "fi": {"chunks": 4370.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.232311, "cps": 34.140625, "mean_p": 0.23052, "boundary_entropy": 0.106637, "bpb": 2.209115}}, "bpb": {"en": 1.94689, "tr": 2.10476, "fi": 2.20911}, "gini_cpc": 0.041373, "gini_cps": 0.058955, "bpic": 5.016979, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.017814218703361, "fi": 1.199675453484111}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 603.857177734375, "n_util_obs": 616, "util_count_spearman": -0.019131909705622986}, "t": 1217.7} |
| {"event": "eval", "step": 3800, "phase": 2, "bpb": {"en": 1.95966, "tr": 2.11252, "fi": 2.21734}, "bytes_per_chunk": {"en": 4.897, "tr": 5.631, "fi": 4.473}, "bpb_mean": 2.09651, "frozen_inventory": true, "t": 1219.2} |
| {"event": "train", "step": 3800, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.44794, "bpb": 2.08894, "chunk_loss": 0.030461, "F": 0.2052, "G": 0.21243, "chunks": 210.12, "bytes_per_chunk": 4.873, "boundary_entropy": 0.09933, "router_noise": 0.0, "grad_norm": 0.6791, "lr": 0.0001177, "router_trainable": true, "sec_per_step": 0.342, "probe_sec": 0.3055, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02849, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.12797, "n_shadow": 0, "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": 0.60767, "trunc_abs_frac": 0.70354, "mean_delta_full": 0.25911, "read_hit_rate": {"split": 0.90291}, "mean_gate": {"split": 0.113}, "n_reads": 10797, "n_cand": 11958, "read_suppressed_frac": 0.1018, "dropout_reads_logged": 10797, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 604.02197265625, "n_util_obs": 616, "util_count_spearman": -0.01863598236303256}, "probe_overhead_sec_total": 266.88, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009737, "alpha": 0.03, "t": 1220.3} |
| {"event": "consolidate", "step": 3850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1233.0} |
| {"event": "train", "step": 3850, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.54105, "bpb": 2.22327, "chunk_loss": 0.030652, "F": 0.20068, "G": 0.21072, "chunks": 205.5, "bytes_per_chunk": 4.983, "boundary_entropy": 0.10365, "router_noise": 0.0, "grad_norm": 0.7326, "lr": 0.0001142, "router_trainable": true, "sec_per_step": 0.262, "probe_sec": 0.1095, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09978, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.12559, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 2, "trunc_ratio_sum": 2.21717, "trunc_abs_frac": 1.13639, "mean_delta_full": -0.18372, "read_hit_rate": {"split": 0.88354}, "mean_gate": {"split": 0.11526}, "n_reads": 9961, "n_cand": 11274, "read_suppressed_frac": 0.1029, "dropout_reads_logged": 9961, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 612.126708984375, "n_util_obs": 624, "util_count_spearman": -0.016901149124640056}, "probe_overhead_sec_total": 270.476, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.010249, "alpha": 0.03, "t": 1233.4} |
| {"event": "consolidate", "step": 3900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1245.3} |
| {"event": "train", "step": 3900, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.4306, "bpb": 2.06391, "chunk_loss": 0.030503, "F": 0.20523, "G": 0.21181, "chunks": 210.16, "bytes_per_chunk": 4.873, "boundary_entropy": 0.09922, "router_noise": 0.0, "grad_norm": 0.6951, "lr": 0.0001108, "router_trainable": true, "sec_per_step": 0.243, "probe_sec": 0.0787, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06612, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.0289, "n_shadow": 0, "mean_delta_shadow": NaN, "n_split": 0, "mean_delta_split": NaN, "n_merge": 0, "mean_delta_merge": NaN, "n_dropped_unpaired": 0, "n_full": 3, "trunc_ratio_sum": -5.24964, "trunc_abs_frac": 0.53076, "mean_delta_full": -0.02787, "read_hit_rate": {"split": 0.90657}, "mean_gate": {"split": 0.10996}, "n_reads": 10770, "n_cand": 11880, "read_suppressed_frac": 0.0976, "dropout_reads_logged": 10770, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 620.19677734375, "n_util_obs": 632, "util_count_spearman": -0.009807510606915452}, "probe_overhead_sec_total": 273.6, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.00973, "alpha": 0.03, "t": 1245.5} |
| {"event": "consolidate", "step": 3950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1258.5} |
| {"event": "train", "step": 3950, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.48289, "bpb": 2.13936, "chunk_loss": 0.030409, "F": 0.20053, "G": 0.20745, "chunks": 205.34, "bytes_per_chunk": 4.987, "boundary_entropy": 0.09653, "router_noise": 0.0, "grad_norm": 0.7894, "lr": 0.0001074, "router_trainable": true, "sec_per_step": 0.263, "probe_sec": 0.0458, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03473, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.04007, "n_shadow": 0, "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": -0.00464, "read_hit_rate": {"split": 0.89988}, "mean_gate": {"split": 0.11471}, "n_reads": 10444, "n_cand": 11606, "read_suppressed_frac": 0.101, "dropout_reads_logged": 10444, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 628.3475341796875, "n_util_obs": 636, "util_count_spearman": -0.013258135438935992}, "probe_overhead_sec_total": 277.686, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009327, "alpha": 0.03, "t": 1258.7} |
| {"event": "consolidate", "step": 4000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1271.3} |
| {"event": "probe", "step": 4000, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3372.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.19376, "cps": 26.34375, "mean_p": 0.200423, "boundary_entropy": 0.094821, "bpb": 1.932804}, "tr": {"chunks": 3500.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.193338, "cps": 27.34375, "mean_p": 0.193404, "boundary_entropy": 0.116643, "bpb": 2.088473}, "fi": {"chunks": 4337.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.230557, "cps": 33.882812, "mean_p": 0.230337, "boundary_entropy": 0.106416, "bpb": 2.19326}}, "bpb": {"en": 1.9328, "tr": 2.08847, "fi": 2.19326}, "gini_cpc": 0.040172, "gini_cps": 0.057394, "bpic": 5.06129, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9978242335327917, "fi": 1.1899099413892569}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 636.2701416015625, "n_util_obs": 641, "util_count_spearman": -0.007383027718196504}, "t": 1272.1} |
| {"event": "eval", "step": 4000, "phase": 2, "bpb": {"en": 1.94507, "tr": 2.09629, "fi": 2.19892}, "bytes_per_chunk": {"en": 4.897, "tr": 5.698, "fi": 4.555}, "bpb_mean": 2.08009, "frozen_inventory": true, "t": 1273.6} |
| {"event": "train", "step": 4000, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.40829, "bpb": 2.03174, "chunk_loss": 0.030473, "F": 0.20282, "G": 0.21031, "chunks": 207.69, "bytes_per_chunk": 4.93, "boundary_entropy": 0.0985, "router_noise": 0.0, "grad_norm": 0.665, "lr": 0.000104, "router_trainable": true, "sec_per_step": 0.325, "probe_sec": 0.1184, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10744, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.07577, "n_shadow": 0, "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": 0.29145, "trunc_abs_frac": 0.34299, "mean_delta_full": -0.41962, "read_hit_rate": {"split": 0.90043}, "mean_gate": {"split": 0.11418}, "n_reads": 10517, "n_cand": 11680, "read_suppressed_frac": 0.102, "dropout_reads_logged": 10517, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 636.429931640625, "n_util_obs": 641, "util_count_spearman": -0.007142389150129532}, "probe_overhead_sec_total": 281.462, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008627, "alpha": 0.03, "t": 1274.9} |
| {"event": "consolidate", "step": 4050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1287.4} |
| {"event": "train", "step": 4050, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.42616, "bpb": 2.05751, "chunk_loss": 0.030465, "F": 0.20386, "G": 0.21248, "chunks": 208.75, "bytes_per_chunk": 4.905, "boundary_entropy": 0.0966, "router_noise": 0.0, "grad_norm": 0.6938, "lr": 0.0001007, "router_trainable": true, "sec_per_step": 0.255, "probe_sec": 0.0877, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07864, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.01209, "n_shadow": 0, "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": 0.20435, "trunc_abs_frac": 0.3952, "mean_delta_full": 0.16528, "read_hit_rate": {"split": 0.89622}, "mean_gate": {"split": 0.11646}, "n_reads": 10398, "n_cand": 11602, "read_suppressed_frac": 0.1074, "dropout_reads_logged": 10398, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 644.5401000976562, "n_util_obs": 647, "util_count_spearman": -0.014050254370886939}, "probe_overhead_sec_total": 284.915, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009707, "alpha": 0.03, "t": 1287.7} |
| {"event": "probe", "step": 4096, "phase": 2, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3405.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.195656, "cps": 26.601562, "mean_p": 0.201609, "boundary_entropy": 0.091339, "bpb": 1.927314}, "tr": {"chunks": 3589.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.198254, "cps": 28.039062, "mean_p": 0.194855, "boundary_entropy": 0.109602, "bpb": 2.078967}, "fi": {"chunks": 4327.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.230025, "cps": 33.804688, "mean_p": 0.228917, "boundary_entropy": 0.09949, "bpb": 2.19071}}, "bpb": {"en": 1.92731, "tr": 2.07897, "fi": 2.19071}, "gini_cpc": 0.036723, "gini_cps": 0.054294, "bpic": 5.011218, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.0132810269965233, "fi": 1.1756607402595305}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 651.8314819335938, "n_util_obs": 656, "util_count_spearman": -0.032148595078934965}, "t": 1300.6} |
| {"event": "consolidate", "step": 4100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1303.8} |
| {"event": "train", "step": 4100, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.35636, "bpb": 1.95681, "chunk_loss": 0.030439, "F": 0.20468, "G": 0.21059, "chunks": 209.59, "bytes_per_chunk": 4.886, "boundary_entropy": 0.09088, "router_noise": 0.0, "grad_norm": 0.6912, "lr": 9.75e-05, "router_trainable": true, "sec_per_step": 0.329, "probe_sec": 0.0623, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04682, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.01299, "n_shadow": 0, "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": 0.27723, "trunc_abs_frac": 0.84614, "mean_delta_full": -0.17501, "read_hit_rate": {"split": 0.90787}, "mean_gate": {"split": 0.11426}, "n_reads": 10751, "n_cand": 11842, "read_suppressed_frac": 0.1006, "dropout_reads_logged": 10751, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 652.64501953125, "n_util_obs": 656, "util_count_spearman": -0.03361850115226475}, "probe_overhead_sec_total": 288.623, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009518, "alpha": 0.03, "t": 1304.2} |
| {"event": "consolidate", "step": 4150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1317.7} |
| {"event": "train", "step": 4150, "arch": "ali", "phase": 2, "objective": "global", "lm_loss": 1.4131, "bpb": 2.03868, "chunk_loss": 0.030606, "F": 0.21008, "G": 0.21589, "chunks": 215.12, "bytes_per_chunk": 4.76, "boundary_entropy": 0.09524, "router_noise": 0.0, "grad_norm": 0.7431, "lr": 9.43e-05, "router_trainable": true, "sec_per_step": 0.276, "probe_sec": 0.1226, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.11052, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.07658, "n_shadow": 0, "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": -0.70833, "trunc_abs_frac": 0.89582, "mean_delta_full": -0.08533, "read_hit_rate": {"split": 0.89965}, "mean_gate": {"split": 0.10796}, "n_reads": 10830, "n_cand": 12038, "read_suppressed_frac": 0.0966, "dropout_reads_logged": 10830, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 660.7463989257812, "n_util_obs": 667, "util_count_spearman": -0.02631195863833535}, "probe_overhead_sec_total": 292.699, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009426, "alpha": 0.03, "t": 1318.0} |
| {"event": "consolidate", "step": 4200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1331.7} |
| {"event": "probe", "step": 4200, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3403.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.195541, "cps": 26.585938, "mean_p": 0.200435, "boundary_entropy": 0.089221, "bpb": 1.924208}, "tr": {"chunks": 3509.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.193835, "cps": 27.414062, "mean_p": 0.192044, "boundary_entropy": 0.109484, "bpb": 2.070781}, "fi": {"chunks": 4342.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.230822, "cps": 33.921875, "mean_p": 0.229222, "boundary_entropy": 0.103801, "bpb": 2.18221}}, "bpb": {"en": 1.92421, "tr": 2.07078, "fi": 2.18221}, "gini_cpc": 0.039758, "gini_cps": 0.055625, "bpic": 5.041052, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9912769047473456, "fi": 1.1804296424493126}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 668.6604614257812, "n_util_obs": 676, "util_count_spearman": -0.02589090137828656}, "t": 1332.6} |
| {"event": "eval", "step": 4200, "phase": 2, "bpb": {"en": 1.93599, "tr": 2.07936, "fi": 2.18971}, "bytes_per_chunk": {"en": 4.874, "tr": 5.69, "fi": 4.578}, "bpb_mean": 2.06835, "frozen_inventory": true, "t": 1334.0} |
| {"event": "phase_transition", "step": 4200, "phase": 3, "objective": "global", "router_trainable": true, "encoder_trainable": true, "frozen_router_params": 0, "t": 1335.1} |
| {"event": "train", "step": 4200, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.34778, "bpb": 1.94443, "chunk_loss": 0.030299, "F": 0.19608, "G": 0.20286, "chunks": 200.78, "bytes_per_chunk": 5.1, "boundary_entropy": 0.09006, "router_noise": 0.0, "grad_norm": 0.6935, "lr": 9.12e-05, "router_trainable": true, "sec_per_step": 0.349, "probe_sec": 0.1465, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.13407, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.0221, "n_shadow": 0, "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": 1.62573, "trunc_abs_frac": 1.48021, "mean_delta_full": 0.02087, "read_hit_rate": {"split": 0.8971}, "mean_gate": {"split": 0.11629}, "n_reads": 10270, "n_cand": 11448, "read_suppressed_frac": 0.0993, "dropout_reads_logged": 10270, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 668.8171997070312, "n_util_obs": 677, "util_count_spearman": -0.022650534572165506}, "probe_overhead_sec_total": 296.266, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008423, "alpha": 0.03, "t": 1335.4} |
| {"event": "consolidate", "step": 4250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1350.2} |
| {"event": "train", "step": 4250, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38003, "bpb": 1.99096, "chunk_loss": 0.030466, "F": 0.20459, "G": 0.21032, "chunks": 209.5, "bytes_per_chunk": 4.888, "boundary_entropy": 0.09128, "router_noise": 0.0, "grad_norm": 0.658, "lr": 8.81e-05, "router_trainable": true, "sec_per_step": 0.311, "probe_sec": 0.0571, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03274, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.0761, "n_shadow": 0, "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": -0.83961, "trunc_abs_frac": 0.66215, "mean_delta_full": 0.04053, "read_hit_rate": {"split": 0.90585}, "mean_gate": {"split": 0.11567}, "n_reads": 10689, "n_cand": 11800, "read_suppressed_frac": 0.1047, "dropout_reads_logged": 10689, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 676.908447265625, "n_util_obs": 682, "util_count_spearman": -0.018344538802027563}, "probe_overhead_sec_total": 299.6, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009444, "alpha": 0.03, "t": 1350.9} |
| {"event": "consolidate", "step": 4300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1366.4} |
| {"event": "train", "step": 4300, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38591, "bpb": 1.99944, "chunk_loss": 0.030316, "F": 0.1972, "G": 0.20339, "chunks": 201.94, "bytes_per_chunk": 5.071, "boundary_entropy": 0.09223, "router_noise": 0.0, "grad_norm": 0.7281, "lr": 8.5e-05, "router_trainable": true, "sec_per_step": 0.313, "probe_sec": 0.0491, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03758, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.05418, "n_shadow": 0, "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": 0.77904, "trunc_abs_frac": 0.64059, "mean_delta_full": -0.10141, "read_hit_rate": {"split": 0.89112}, "mean_gate": {"split": 0.10941}, "n_reads": 10157, "n_cand": 11398, "read_suppressed_frac": 0.0997, "dropout_reads_logged": 10157, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 685.0122680664062, "n_util_obs": 688, "util_count_spearman": -0.028020920188689232}, "probe_overhead_sec_total": 303.196, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009173, "alpha": 0.03, "t": 1366.6} |
| {"event": "consolidate", "step": 4350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1380.2} |
| {"event": "train", "step": 4350, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.40948, "bpb": 2.03345, "chunk_loss": 0.030421, "F": 0.20355, "G": 0.2096, "chunks": 208.44, "bytes_per_chunk": 4.913, "boundary_entropy": 0.08902, "router_noise": 0.0, "grad_norm": 0.6913, "lr": 8.21e-05, "router_trainable": true, "sec_per_step": 0.278, "probe_sec": 0.0518, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04167, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.00926, "n_shadow": 0, "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": -0.05145, "trunc_abs_frac": 13.0408, "mean_delta_full": -0.24592, "read_hit_rate": {"split": 0.89919}, "mean_gate": {"split": 0.1075}, "n_reads": 10596, "n_cand": 11784, "read_suppressed_frac": 0.0947, "dropout_reads_logged": 10596, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 693.059326171875, "n_util_obs": 692, "util_count_spearman": -0.03122197117796957}, "probe_overhead_sec_total": 306.918, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009717, "alpha": 0.03, "t": 1380.5} |
| {"event": "consolidate", "step": 4400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1394.7} |
| {"event": "probe", "step": 4400, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3352.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.19261, "cps": 26.1875, "mean_p": 0.197221, "boundary_entropy": 0.088081, "bpb": 1.913266}, "tr": {"chunks": 3500.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.193338, "cps": 27.34375, "mean_p": 0.190117, "boundary_entropy": 0.105557, "bpb": 2.057266}, "fi": {"chunks": 4379.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.232789, "cps": 34.210938, "mean_p": 0.231608, "boundary_entropy": 0.104213, "bpb": 2.161578}}, "bpb": {"en": 1.91327, "tr": 2.05727, "fi": 2.16158}, "gini_cpc": 0.043291, "gini_cps": 0.060962, "bpic": 5.051376, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 1.003777838744801, "fi": 1.2086016198819332}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 700.9389038085938, "n_util_obs": 694, "util_count_spearman": -0.029883263794436375}, "t": 1395.7} |
| {"event": "eval", "step": 4400, "phase": 3, "bpb": {"en": 1.92568, "tr": 2.06521, "fi": 2.16969}, "bytes_per_chunk": {"en": 4.924, "tr": 5.735, "fi": 4.572}, "bpb_mean": 2.05353, "frozen_inventory": true, "t": 1397.1} |
| {"event": "train", "step": 4400, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35969, "bpb": 1.96162, "chunk_loss": 0.030326, "F": 0.20029, "G": 0.20517, "chunks": 205.09, "bytes_per_chunk": 4.993, "boundary_entropy": 0.09008, "router_noise": 0.0, "grad_norm": 0.6253, "lr": 7.92e-05, "router_trainable": true, "sec_per_step": 0.362, "probe_sec": 0.1022, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.09244, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.06305, "n_shadow": 0, "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": -9.33104, "trunc_abs_frac": 1.18066, "mean_delta_full": 0.00889, "read_hit_rate": {"split": 0.90295}, "mean_gate": {"split": 0.10912}, "n_reads": 10606, "n_cand": 11746, "read_suppressed_frac": 0.1002, "dropout_reads_logged": 10606, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 701.10107421875, "n_util_obs": 694, "util_count_spearman": -0.03023205638452883}, "probe_overhead_sec_total": 310.121, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.010146, "alpha": 0.03, "t": 1398.6} |
| {"event": "consolidate", "step": 4450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1416.1} |
| {"event": "train", "step": 4450, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35406, "bpb": 1.95349, "chunk_loss": 0.030281, "F": 0.19443, "G": 0.20001, "chunks": 199.09, "bytes_per_chunk": 5.143, "boundary_entropy": 0.08836, "router_noise": 0.0, "grad_norm": 0.6868, "lr": 7.63e-05, "router_trainable": true, "sec_per_step": 0.356, "probe_sec": 0.1173, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10455, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.04733, "n_shadow": 0, "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": 0.12276, "trunc_abs_frac": 0.23852, "mean_delta_full": -0.43905, "read_hit_rate": {"split": 0.89701}, "mean_gate": {"split": 0.10995}, "n_reads": 10077, "n_cand": 11234, "read_suppressed_frac": 0.0984, "dropout_reads_logged": 10077, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 709.1805419921875, "n_util_obs": 698, "util_count_spearman": -0.03375328828306408}, "probe_overhead_sec_total": 314.049, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009192, "alpha": 0.03, "t": 1416.4} |
| {"event": "consolidate", "step": 4500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1432.8} |
| {"event": "train", "step": 4500, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32904, "bpb": 1.9174, "chunk_loss": 0.030403, "F": 0.20404, "G": 0.20907, "chunks": 208.94, "bytes_per_chunk": 4.901, "boundary_entropy": 0.08926, "router_noise": 0.0, "grad_norm": 0.6475, "lr": 7.36e-05, "router_trainable": true, "sec_per_step": 0.331, "probe_sec": 0.0594, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04549, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.09684, "n_shadow": 0, "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": 0.07638, "trunc_abs_frac": 1.16422, "mean_delta_full": 0.21431, "read_hit_rate": {"split": 0.90686}, "mean_gate": {"split": 0.1074}, "n_reads": 10837, "n_cand": 11950, "read_suppressed_frac": 0.0952, "dropout_reads_logged": 10837, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 717.2337036132812, "n_util_obs": 705, "util_count_spearman": -0.035600320457164686}, "probe_overhead_sec_total": 317.715, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009339, "alpha": 0.03, "t": 1433.0} |
| {"event": "consolidate", "step": 4550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1461.2} |
| {"event": "train", "step": 4550, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35724, "bpb": 1.95808, "chunk_loss": 0.030352, "F": 0.19931, "G": 0.20551, "chunks": 204.09, "bytes_per_chunk": 5.017, "boundary_entropy": 0.08902, "router_noise": 0.0, "grad_norm": 0.6725, "lr": 7.09e-05, "router_trainable": true, "sec_per_step": 0.579, "probe_sec": 0.0444, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02903, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.01621, "n_shadow": 0, "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": 0.5495, "trunc_abs_frac": 0.46347, "mean_delta_full": -0.1829, "read_hit_rate": {"split": 0.89632}, "mean_gate": {"split": 0.10652}, "n_reads": 10383, "n_cand": 11584, "read_suppressed_frac": 0.0989, "dropout_reads_logged": 10383, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 725.3174438476562, "n_util_obs": 712, "util_count_spearman": -0.033609741498825556}, "probe_overhead_sec_total": 321.476, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009791, "alpha": 0.03, "t": 1461.9} |
| {"event": "consolidate", "step": 4600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1478.0} |
| {"event": "probe", "step": 4600, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3399.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.195311, "cps": 26.554688, "mean_p": 0.198818, "boundary_entropy": 0.085557, "bpb": 1.902994}, "tr": {"chunks": 3529.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.19494, "cps": 27.570312, "mean_p": 0.191499, "boundary_entropy": 0.100998, "bpb": 2.048653}, "fi": {"chunks": 4360.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.231779, "cps": 34.0625, "mean_p": 0.230085, "boundary_entropy": 0.099542, "bpb": 2.153638}}, "bpb": {"en": 1.90299, "tr": 2.04865, "fi": 2.15364}, "gini_cpc": 0.039483, "gini_cps": 0.056756, "bpic": 5.025868, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9981000159965294, "fi": 1.1867180866315332}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 733.2472534179688, "n_util_obs": 713, "util_count_spearman": -0.03636019569330197}, "t": 1478.8} |
| {"event": "eval", "step": 4600, "phase": 3, "bpb": {"en": 1.91605, "tr": 2.05638, "fi": 2.15984}, "bytes_per_chunk": {"en": 4.871, "tr": 5.696, "fi": 4.577}, "bpb_mean": 2.04409, "frozen_inventory": true, "t": 1480.4} |
| {"event": "train", "step": 4600, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.36083, "bpb": 1.96326, "chunk_loss": 0.030508, "F": 0.2095, "G": 0.21413, "chunks": 214.53, "bytes_per_chunk": 4.773, "boundary_entropy": 0.09016, "router_noise": 0.0, "grad_norm": 0.6847, "lr": 6.82e-05, "router_trainable": true, "sec_per_step": 0.403, "probe_sec": 0.1243, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.10832, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.01331, "n_shadow": 0, "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": 0.17264, "trunc_abs_frac": 0.27952, "mean_delta_full": 0.13373, "read_hit_rate": {"split": 0.90575}, "mean_gate": {"split": 0.10666}, "n_reads": 10898, "n_cand": 12032, "read_suppressed_frac": 0.1012, "dropout_reads_logged": 10898, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 733.4141845703125, "n_util_obs": 713, "util_count_spearman": -0.036429256497071974}, "probe_overhead_sec_total": 325.271, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.010177, "alpha": 0.03, "t": 1482.1} |
| {"event": "consolidate", "step": 4650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1501.4} |
| {"event": "train", "step": 4650, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.3917, "bpb": 2.00779, "chunk_loss": 0.030401, "F": 0.20129, "G": 0.20775, "chunks": 206.12, "bytes_per_chunk": 4.968, "boundary_entropy": 0.08926, "router_noise": 0.0, "grad_norm": 0.6599, "lr": 6.57e-05, "router_trainable": true, "sec_per_step": 0.394, "probe_sec": 0.0462, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03413, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.11968, "n_shadow": 0, "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": 0.82088, "trunc_abs_frac": 0.82088, "mean_delta_full": 0.78235, "read_hit_rate": {"split": 0.89484}, "mean_gate": {"split": 0.10795}, "n_reads": 10339, "n_cand": 11554, "read_suppressed_frac": 0.0959, "dropout_reads_logged": 10339, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 741.4727783203125, "n_util_obs": 717, "util_count_spearman": -0.04203661675731648}, "probe_overhead_sec_total": 328.858, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009653, "alpha": 0.03, "t": 1501.8} |
| {"event": "consolidate", "step": 4700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1525.2} |
| {"event": "train", "step": 4700, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.41694, "bpb": 2.04421, "chunk_loss": 0.030532, "F": 0.20923, "G": 0.21287, "chunks": 214.25, "bytes_per_chunk": 4.779, "boundary_entropy": 0.08508, "router_noise": 0.0, "grad_norm": 0.644, "lr": 6.32e-05, "router_trainable": true, "sec_per_step": 0.496, "probe_sec": 0.0783, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04204, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.06533, "n_shadow": 0, "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": -4.45022, "trunc_abs_frac": 0.34766, "mean_delta_full": -0.0094, "read_hit_rate": {"split": 0.9061}, "mean_gate": {"split": 0.10341}, "n_reads": 10962, "n_cand": 12098, "read_suppressed_frac": 0.0979, "dropout_reads_logged": 10962, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 749.5487060546875, "n_util_obs": 724, "util_count_spearman": -0.029149533816327105}, "probe_overhead_sec_total": 332.075, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.010349, "alpha": 0.03, "t": 1526.5} |
| {"event": "consolidate", "step": 4750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1545.7} |
| {"event": "train", "step": 4750, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.39984, "bpb": 2.01954, "chunk_loss": 0.030436, "F": 0.20438, "G": 0.20963, "chunks": 209.28, "bytes_per_chunk": 4.893, "boundary_entropy": 0.08626, "router_noise": 0.0, "grad_norm": 0.6646, "lr": 6.08e-05, "router_trainable": true, "sec_per_step": 0.388, "probe_sec": 0.0351, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02632, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.06338, "n_shadow": 0, "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": 0.68632, "trunc_abs_frac": 0.93714, "mean_delta_full": 0.29091, "read_hit_rate": {"split": 0.89804}, "mean_gate": {"split": 0.10527}, "n_reads": 10552, "n_cand": 11750, "read_suppressed_frac": 0.1033, "dropout_reads_logged": 10552, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 757.5510864257812, "n_util_obs": 733, "util_count_spearman": -0.04153199061181606}, "probe_overhead_sec_total": 335.373, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008582, "alpha": 0.03, "t": 1545.9} |
| {"event": "consolidate", "step": 4800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1561.7} |
| {"event": "probe", "step": 4800, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3369.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.193587, "cps": 26.320312, "mean_p": 0.198027, "boundary_entropy": 0.085436, "bpb": 1.89794}, "tr": {"chunks": 3499.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.193283, "cps": 27.335938, "mean_p": 0.189344, "boundary_entropy": 0.097924, "bpb": 2.034702}, "fi": {"chunks": 4314.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.229334, "cps": 33.703125, "mean_p": 0.227384, "boundary_entropy": 0.094113, "bpb": 2.144969}}, "bpb": {"en": 1.89794, "tr": 2.0347, "fi": 2.14497}, "gini_cpc": 0.039003, "gini_cps": 0.056341, "bpic": 5.073511, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9984274215187666, "fi": 1.184653567310349}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 765.45556640625, "n_util_obs": 740, "util_count_spearman": -0.04023608793890459}, "t": 1562.5} |
| {"event": "eval", "step": 4800, "phase": 3, "bpb": {"en": 1.9102, "tr": 2.04291, "fi": 2.15038}, "bytes_per_chunk": {"en": 4.909, "tr": 5.717, "fi": 4.58}, "bpb_mean": 2.0345, "frozen_inventory": true, "t": 1564.1} |
| {"event": "train", "step": 4800, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.37638, "bpb": 1.98569, "chunk_loss": 0.030465, "F": 0.20807, "G": 0.21225, "chunks": 213.06, "bytes_per_chunk": 4.806, "boundary_entropy": 0.08648, "router_noise": 0.0, "grad_norm": 0.6483, "lr": 5.85e-05, "router_trainable": true, "sec_per_step": 0.393, "probe_sec": 0.0496, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03335, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.02667, "n_shadow": 0, "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": -0.81818, "trunc_abs_frac": 0.71031, "mean_delta_full": -0.0141, "read_hit_rate": {"split": 0.90564}, "mean_gate": {"split": 0.10501}, "n_reads": 10788, "n_cand": 11912, "read_suppressed_frac": 0.1078, "dropout_reads_logged": 10788, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 765.6201782226562, "n_util_obs": 740, "util_count_spearman": -0.04031888704191524}, "probe_overhead_sec_total": 339.115, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009364, "alpha": 0.03, "t": 1565.6} |
| {"event": "consolidate", "step": 4850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1579.6} |
| {"event": "train", "step": 4850, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.34802, "bpb": 1.94479, "chunk_loss": 0.030421, "F": 0.20184, "G": 0.20733, "chunks": 206.69, "bytes_per_chunk": 4.954, "boundary_entropy": 0.08721, "router_noise": 0.0, "grad_norm": 0.6671, "lr": 5.62e-05, "router_trainable": true, "sec_per_step": 0.285, "probe_sec": 0.0387, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02882, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.01493, "n_shadow": 0, "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": -0.67405, "trunc_abs_frac": 0.48333, "mean_delta_full": 0.06268, "read_hit_rate": {"split": 0.89918}, "mean_gate": {"split": 0.10715}, "n_reads": 10488, "n_cand": 11664, "read_suppressed_frac": 0.1021, "dropout_reads_logged": 10488, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 773.7313232421875, "n_util_obs": 748, "util_count_spearman": -0.0377082053176289}, "probe_overhead_sec_total": 342.209, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009496, "alpha": 0.03, "t": 1579.8} |
| {"event": "consolidate", "step": 4900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1592.8} |
| {"event": "train", "step": 4900, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38109, "bpb": 1.99249, "chunk_loss": 0.030428, "F": 0.19781, "G": 0.20272, "chunks": 202.56, "bytes_per_chunk": 5.055, "boundary_entropy": 0.08563, "router_noise": 0.0, "grad_norm": 0.6585, "lr": 5.41e-05, "router_trainable": true, "sec_per_step": 0.266, "probe_sec": 0.1734, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.15616, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.00693, "n_shadow": 0, "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": -1.78068, "trunc_abs_frac": 2.73369, "mean_delta_full": 0.07332, "read_hit_rate": {"split": 0.8909}, "mean_gate": {"split": 0.09883}, "n_reads": 10142, "n_cand": 11384, "read_suppressed_frac": 0.1008, "dropout_reads_logged": 10142, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 781.8142700195312, "n_util_obs": 753, "util_count_spearman": -0.041838645348510055}, "probe_overhead_sec_total": 346.089, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009395, "alpha": 0.03, "t": 1593.1} |
| {"event": "consolidate", "step": 4950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1605.7} |
| {"event": "train", "step": 4950, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.33484, "bpb": 1.92577, "chunk_loss": 0.030414, "F": 0.20209, "G": 0.2068, "chunks": 206.94, "bytes_per_chunk": 4.948, "boundary_entropy": 0.08575, "router_noise": 0.0, "grad_norm": 0.6536, "lr": 5.2e-05, "router_trainable": true, "sec_per_step": 0.256, "probe_sec": 0.042, "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": 0.04559, "n_shadow": 0, "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": -1.38122, "trunc_abs_frac": 0.7688, "mean_delta_full": -0.09806, "read_hit_rate": {"split": 0.8986}, "mean_gate": {"split": 0.10798}, "n_reads": 10368, "n_cand": 11538, "read_suppressed_frac": 0.0948, "dropout_reads_logged": 10368, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 789.8638916015625, "n_util_obs": 760, "util_count_spearman": -0.02918260068180947}, "probe_overhead_sec_total": 349.162, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009547, "alpha": 0.03, "t": 1605.9} |
| {"event": "consolidate", "step": 5000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1620.2} |
| {"event": "probe", "step": 5000, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3376.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.19399, "cps": 26.375, "mean_p": 0.197629, "boundary_entropy": 0.082465, "bpb": 1.889318}, "tr": {"chunks": 3501.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.193393, "cps": 27.351562, "mean_p": 0.188916, "boundary_entropy": 0.096864, "bpb": 2.026794}, "fi": {"chunks": 4318.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.229547, "cps": 33.734375, "mean_p": 0.226866, "boundary_entropy": 0.092533, "bpb": 2.127657}}, "bpb": {"en": 1.88932, "tr": 2.02679, "fi": 2.12766}, "gini_cpc": 0.039068, "gini_cps": 0.056096, "bpic": 5.067619, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9969267321825896, "fi": 1.1832933853616454}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 797.7796020507812, "n_util_obs": 768, "util_count_spearman": -0.03683186767681979}, "t": 1621.0} |
| {"event": "eval", "step": 5000, "phase": 3, "bpb": {"en": 1.90235, "tr": 2.0356, "fi": 2.13477}, "bytes_per_chunk": {"en": 4.895, "tr": 5.717, "fi": 4.585}, "bpb_mean": 2.02424, "frozen_inventory": true, "t": 1622.5} |
| {"event": "train", "step": 5000, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32616, "bpb": 1.91325, "chunk_loss": 0.030418, "F": 0.20499, "G": 0.20809, "chunks": 209.91, "bytes_per_chunk": 4.878, "boundary_entropy": 0.08175, "router_noise": 0.0, "grad_norm": 0.6192, "lr": 5e-05, "router_trainable": true, "sec_per_step": 0.352, "probe_sec": 0.0565, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.04656, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.10328, "n_shadow": 0, "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": 0.31824, "trunc_abs_frac": 7.75835, "mean_delta_full": 0.24799, "read_hit_rate": {"split": 0.90957}, "mean_gate": {"split": 0.10132}, "n_reads": 10642, "n_cand": 11700, "read_suppressed_frac": 0.1001, "dropout_reads_logged": 10642, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 797.9425048828125, "n_util_obs": 768, "util_count_spearman": -0.03586272046140968}, "probe_overhead_sec_total": 353.58, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009259, "alpha": 0.03, "t": 1623.5} |
| {"event": "consolidate", "step": 5050, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1636.5} |
| {"event": "train", "step": 5050, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.34682, "bpb": 1.94306, "chunk_loss": 0.030487, "F": 0.20749, "G": 0.21062, "chunks": 212.47, "bytes_per_chunk": 4.82, "boundary_entropy": 0.08279, "router_noise": 0.0, "grad_norm": 0.6882, "lr": 4.81e-05, "router_trainable": true, "sec_per_step": 0.266, "probe_sec": 0.0762, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06381, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.0294, "n_shadow": 0, "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": 0.011, "trunc_abs_frac": 0.52426, "mean_delta_full": 0.14097, "read_hit_rate": {"split": 0.91212}, "mean_gate": {"split": 0.10252}, "n_reads": 11064, "n_cand": 12130, "read_suppressed_frac": 0.0954, "dropout_reads_logged": 11064, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 806.0313720703125, "n_util_obs": 776, "util_count_spearman": -0.04453699640540143}, "probe_overhead_sec_total": 357.315, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008875, "alpha": 0.03, "t": 1636.8} |
| {"event": "consolidate", "step": 5100, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1649.7} |
| {"event": "train", "step": 5100, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.3742, "bpb": 1.98255, "chunk_loss": 0.030524, "F": 0.20465, "G": 0.21016, "chunks": 209.56, "bytes_per_chunk": 4.886, "boundary_entropy": 0.0844, "router_noise": 0.0, "grad_norm": 0.7178, "lr": 4.63e-05, "router_trainable": true, "sec_per_step": 0.267, "probe_sec": 0.0709, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05261, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.00362, "n_shadow": 0, "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": -0.07356, "trunc_abs_frac": 0.07356, "mean_delta_full": 0.46716, "read_hit_rate": {"split": 0.89696}, "mean_gate": {"split": 0.09986}, "n_reads": 10437, "n_cand": 11636, "read_suppressed_frac": 0.1014, "dropout_reads_logged": 10437, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 814.10498046875, "n_util_obs": 785, "util_count_spearman": -0.04063719948613702}, "probe_overhead_sec_total": 360.46, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009727, "alpha": 0.03, "t": 1650.2} |
| {"event": "consolidate", "step": 5150, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1663.5} |
| {"event": "train", "step": 5150, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.36675, "bpb": 1.9718, "chunk_loss": 0.030515, "F": 0.20883, "G": 0.21228, "chunks": 213.84, "bytes_per_chunk": 4.789, "boundary_entropy": 0.08631, "router_noise": 0.0, "grad_norm": 0.6845, "lr": 4.45e-05, "router_trainable": true, "sec_per_step": 0.275, "probe_sec": 0.045, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03132, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.00644, "n_shadow": 0, "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": -0.65439, "trunc_abs_frac": 0.5937, "mean_delta_full": -0.06982, "read_hit_rate": {"split": 0.90029}, "mean_gate": {"split": 0.10358}, "n_reads": 10636, "n_cand": 11814, "read_suppressed_frac": 0.095, "dropout_reads_logged": 10636, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 822.1798095703125, "n_util_obs": 789, "util_count_spearman": -0.04073305833101696}, "probe_overhead_sec_total": 363.474, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009348, "alpha": 0.03, "t": 1663.9} |
| {"event": "consolidate", "step": 5200, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1677.5} |
| {"event": "probe", "step": 5200, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3359.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.193013, "cps": 26.242188, "mean_p": 0.197962, "boundary_entropy": 0.081856, "bpb": 1.882832}, "tr": {"chunks": 3493.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.192951, "cps": 27.289062, "mean_p": 0.187873, "boundary_entropy": 0.095804, "bpb": 2.022172}, "fi": {"chunks": 4320.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.229653, "cps": 33.75, "mean_p": 0.227207, "boundary_entropy": 0.092121, "bpb": 2.117919}}, "bpb": {"en": 1.88283, "tr": 2.02217, "fi": 2.11792}, "gini_cpc": 0.039745, "gini_cps": 0.057346, "bpic": 5.078052, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.999682640322009, "fi": 1.1898329173376272}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 830.0846557617188, "n_util_obs": 799, "util_count_spearman": -0.04479431763358374}, "t": 1678.4} |
| {"event": "eval", "step": 5200, "phase": 3, "bpb": {"en": 1.89674, "tr": 2.03079, "fi": 2.12562}, "bytes_per_chunk": {"en": 4.903, "tr": 5.743, "fi": 4.578}, "bpb_mean": 2.01771, "frozen_inventory": true, "t": 1679.9} |
| {"event": "train", "step": 5200, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.31889, "bpb": 1.90275, "chunk_loss": 0.030335, "F": 0.20026, "G": 0.20491, "chunks": 205.06, "bytes_per_chunk": 4.994, "boundary_entropy": 0.08317, "router_noise": 0.0, "grad_norm": 0.6535, "lr": 4.29e-05, "router_trainable": true, "sec_per_step": 0.351, "probe_sec": 0.0876, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07701, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.00565, "n_shadow": 0, "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": -0.35769, "trunc_abs_frac": 1.24047, "mean_delta_full": -0.12148, "read_hit_rate": {"split": 0.89956}, "mean_gate": {"split": 0.09605}, "n_reads": 10523, "n_cand": 11698, "read_suppressed_frac": 0.101, "dropout_reads_logged": 10523, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 830.2457275390625, "n_util_obs": 799, "util_count_spearman": -0.04481452642956702}, "probe_overhead_sec_total": 367.159, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009197, "alpha": 0.03, "t": 1681.5} |
| {"event": "consolidate", "step": 5250, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1693.1} |
| {"event": "train", "step": 5250, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.31886, "bpb": 1.90271, "chunk_loss": 0.030376, "F": 0.20059, "G": 0.20511, "chunks": 205.41, "bytes_per_chunk": 4.985, "boundary_entropy": 0.0811, "router_noise": 0.0, "grad_norm": 0.6423, "lr": 4.14e-05, "router_trainable": true, "sec_per_step": 0.236, "probe_sec": 0.0446, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03258, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.02515, "n_shadow": 0, "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": 0.2362, "trunc_abs_frac": 0.73196, "mean_delta_full": 0.34058, "read_hit_rate": {"split": 0.90236}, "mean_gate": {"split": 0.09966}, "n_reads": 10462, "n_cand": 11594, "read_suppressed_frac": 0.1, "dropout_reads_logged": 10462, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 838.3134155273438, "n_util_obs": 805, "util_count_spearman": -0.043376634067285466}, "probe_overhead_sec_total": 370.229, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009463, "alpha": 0.03, "t": 1693.3} |
| {"event": "consolidate", "step": 5300, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1702.4} |
| {"event": "train", "step": 5300, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.37649, "bpb": 1.98585, "chunk_loss": 0.030418, "F": 0.2038, "G": 0.20807, "chunks": 208.69, "bytes_per_chunk": 4.907, "boundary_entropy": 0.08203, "router_noise": 0.0, "grad_norm": 0.6879, "lr": 3.99e-05, "router_trainable": true, "sec_per_step": 0.186, "probe_sec": 0.0368, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02609, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.02553, "n_shadow": 0, "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": 0.17447, "trunc_abs_frac": 0.37364, "mean_delta_full": 0.28909, "read_hit_rate": {"split": 0.89981}, "mean_gate": {"split": 0.09393}, "n_reads": 10607, "n_cand": 11788, "read_suppressed_frac": 0.1036, "dropout_reads_logged": 10607, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 846.368408203125, "n_util_obs": 809, "util_count_spearman": -0.04366351386013481}, "probe_overhead_sec_total": 372.535, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009598, "alpha": 0.03, "t": 1702.6} |
| {"event": "consolidate", "step": 5350, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1711.9} |
| {"event": "train", "step": 5350, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.30759, "bpb": 1.88645, "chunk_loss": 0.030294, "F": 0.19839, "G": 0.20252, "chunks": 203.16, "bytes_per_chunk": 5.04, "boundary_entropy": 0.08244, "router_noise": 0.0, "grad_norm": 0.7165, "lr": 3.86e-05, "router_trainable": true, "sec_per_step": 0.191, "probe_sec": 0.0406, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03236, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.02728, "n_shadow": 0, "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": 0.14566, "trunc_abs_frac": 0.1456, "mean_delta_full": 0.60364, "read_hit_rate": {"split": 0.90638}, "mean_gate": {"split": 0.09538}, "n_reads": 10572, "n_cand": 11664, "read_suppressed_frac": 0.1053, "dropout_reads_logged": 10572, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 854.4921264648438, "n_util_obs": 816, "util_count_spearman": -0.036674024577656776}, "probe_overhead_sec_total": 374.84, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.010074, "alpha": 0.03, "t": 1712.1} |
| {"event": "consolidate", "step": 5400, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1724.9} |
| {"event": "probe", "step": 5400, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3380.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.194219, "cps": 26.40625, "mean_p": 0.197827, "boundary_entropy": 0.081223, "bpb": 1.877082}, "tr": {"chunks": 3512.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.194001, "cps": 27.4375, "mean_p": 0.18926, "boundary_entropy": 0.09404, "bpb": 2.013821}, "fi": {"chunks": 4341.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.230769, "cps": 33.914062, "mean_p": 0.228141, "boundary_entropy": 0.088578, "bpb": 2.112839}}, "bpb": {"en": 1.87708, "tr": 2.01382, "fi": 2.11284}, "gini_cpc": 0.0396, "gini_cps": 0.057034, "bpic": 5.050476, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9988755337227117, "fi": 1.188188438780155}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 862.4102783203125, "n_util_obs": 828, "util_count_spearman": -0.027181014646701484}, "t": 1725.6} |
| {"event": "eval", "step": 5400, "phase": 3, "bpb": {"en": 1.89087, "tr": 2.02354, "fi": 2.12019}, "bytes_per_chunk": {"en": 4.873, "tr": 5.716, "fi": 4.576}, "bpb_mean": 2.01153, "frozen_inventory": true, "t": 1727.3} |
| {"event": "train", "step": 5400, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38279, "bpb": 1.99495, "chunk_loss": 0.030615, "F": 0.20905, "G": 0.21295, "chunks": 214.06, "bytes_per_chunk": 4.784, "boundary_entropy": 0.08329, "router_noise": 0.0, "grad_norm": 0.7111, "lr": 3.73e-05, "router_trainable": true, "sec_per_step": 0.332, "probe_sec": 0.0692, "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": 0.06069, "n_shadow": 0, "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": 0.22131, "trunc_abs_frac": 0.33005, "mean_delta_full": -0.18585, "read_hit_rate": {"split": 0.90612}, "mean_gate": {"split": 0.09493}, "n_reads": 10830, "n_cand": 11952, "read_suppressed_frac": 0.1015, "dropout_reads_logged": 10830, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 862.5755615234375, "n_util_obs": 828, "util_count_spearman": -0.027201720892390292}, "probe_overhead_sec_total": 377.797, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009197, "alpha": 0.03, "t": 1728.7} |
| {"event": "consolidate", "step": 5450, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1742.5} |
| {"event": "train", "step": 5450, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.31591, "bpb": 1.89846, "chunk_loss": 0.030524, "F": 0.20831, "G": 0.21199, "chunks": 213.31, "bytes_per_chunk": 4.8, "boundary_entropy": 0.08122, "router_noise": 0.0, "grad_norm": 0.6871, "lr": 3.62e-05, "router_trainable": true, "sec_per_step": 0.283, "probe_sec": 0.1495, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.13948, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.10443, "n_shadow": 0, "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": -0.38904, "trunc_abs_frac": 0.17012, "mean_delta_full": -0.07117, "read_hit_rate": {"split": 0.90977}, "mean_gate": {"split": 0.09791}, "n_reads": 10879, "n_cand": 11958, "read_suppressed_frac": 0.0983, "dropout_reads_logged": 10879, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 870.626953125, "n_util_obs": 837, "util_count_spearman": -0.034142865782617667}, "probe_overhead_sec_total": 382.304, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008858, "alpha": 0.03, "t": 1742.9} |
| {"event": "consolidate", "step": 5500, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1756.4} |
| {"event": "train", "step": 5500, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.30503, "bpb": 1.88277, "chunk_loss": 0.030372, "F": 0.20261, "G": 0.20681, "chunks": 207.47, "bytes_per_chunk": 4.936, "boundary_entropy": 0.07991, "router_noise": 0.0, "grad_norm": 0.6673, "lr": 3.51e-05, "router_trainable": true, "sec_per_step": 0.275, "probe_sec": 0.0661, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05645, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.01322, "n_shadow": 0, "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": -0.74771, "trunc_abs_frac": 0.95273, "mean_delta_full": 0.07869, "read_hit_rate": {"split": 0.90141}, "mean_gate": {"split": 0.0968}, "n_reads": 10579, "n_cand": 11736, "read_suppressed_frac": 0.0943, "dropout_reads_logged": 10579, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 878.70849609375, "n_util_obs": 842, "util_count_spearman": -0.044878552141928214}, "probe_overhead_sec_total": 386.808, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009792, "alpha": 0.03, "t": 1756.6} |
| {"event": "consolidate", "step": 5550, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1768.8} |
| {"event": "train", "step": 5550, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.35937, "bpb": 1.96115, "chunk_loss": 0.030352, "F": 0.19846, "G": 0.20267, "chunks": 203.22, "bytes_per_chunk": 5.039, "boundary_entropy": 0.08401, "router_noise": 0.0, "grad_norm": 0.7305, "lr": 3.41e-05, "router_trainable": true, "sec_per_step": 0.251, "probe_sec": 0.0693, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.05726, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.00183, "n_shadow": 0, "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": 1.75176, "trunc_abs_frac": 0.60765, "mean_delta_full": -0.15605, "read_hit_rate": {"split": 0.89637}, "mean_gate": {"split": 0.09999}, "n_reads": 10163, "n_cand": 11338, "read_suppressed_frac": 0.1061, "dropout_reads_logged": 10163, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 886.7894897460938, "n_util_obs": 851, "util_count_spearman": -0.03486921871939069}, "probe_overhead_sec_total": 390.268, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.010083, "alpha": 0.03, "t": 1769.2} |
| {"event": "consolidate", "step": 5600, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1781.1} |
| {"event": "probe", "step": 5600, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3388.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.194679, "cps": 26.46875, "mean_p": 0.197509, "boundary_entropy": 0.08089, "bpb": 1.872107}, "tr": {"chunks": 3504.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.193559, "cps": 27.375, "mean_p": 0.188441, "boundary_entropy": 0.09307, "bpb": 2.007611}, "fi": {"chunks": 4332.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.230291, "cps": 33.84375, "mean_p": 0.226886, "boundary_entropy": 0.088808, "bpb": 2.104795}}, "bpb": {"en": 1.87211, "tr": 2.00761, "fi": 2.1048}, "gini_cpc": 0.03959, "gini_cps": 0.05607, "bpic": 5.054526, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9942469436174648, "fi": 1.1829251981918942}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 894.67529296875, "n_util_obs": 857, "util_count_spearman": -0.034859139127015354}, "t": 1781.9} |
| {"event": "eval", "step": 5600, "phase": 3, "bpb": {"en": 1.88607, "tr": 2.01634, "fi": 2.11245}, "bytes_per_chunk": {"en": 4.893, "tr": 5.713, "fi": 4.601}, "bpb_mean": 2.00495, "frozen_inventory": true, "t": 1783.3} |
| {"event": "train", "step": 5600, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.34625, "bpb": 1.94223, "chunk_loss": 0.030414, "F": 0.20285, "G": 0.20714, "chunks": 207.72, "bytes_per_chunk": 4.93, "boundary_entropy": 0.08429, "router_noise": 0.0, "grad_norm": 0.7463, "lr": 3.33e-05, "router_trainable": true, "sec_per_step": 0.312, "probe_sec": 0.0583, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03964, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.02661, "n_shadow": 0, "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": -1.28923, "trunc_abs_frac": 1.30424, "mean_delta_full": -0.1091, "read_hit_rate": {"split": 0.89933}, "mean_gate": {"split": 0.09347}, "n_reads": 10506, "n_cand": 11682, "read_suppressed_frac": 0.0958, "dropout_reads_logged": 10506, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 894.8355712890625, "n_util_obs": 857, "util_count_spearman": -0.034807892766362335}, "probe_overhead_sec_total": 393.832, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.010054, "alpha": 0.03, "t": 1784.8} |
| {"event": "consolidate", "step": 5650, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1798.0} |
| {"event": "train", "step": 5650, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.32877, "bpb": 1.91702, "chunk_loss": 0.030511, "F": 0.20908, "G": 0.21166, "chunks": 214.09, "bytes_per_chunk": 4.783, "boundary_entropy": 0.0806, "router_noise": 0.0, "grad_norm": 0.6897, "lr": 3.25e-05, "router_trainable": true, "sec_per_step": 0.27, "probe_sec": 0.0426, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0329, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.06939, "n_shadow": 0, "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": 69.70689, "trunc_abs_frac": 0.46751, "mean_delta_full": 0.00177, "read_hit_rate": {"split": 0.90622}, "mean_gate": {"split": 0.09352}, "n_reads": 10891, "n_cand": 12018, "read_suppressed_frac": 0.0982, "dropout_reads_logged": 10891, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 902.9114379882812, "n_util_obs": 865, "util_count_spearman": -0.036929745499014295}, "probe_overhead_sec_total": 398.158, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008929, "alpha": 0.03, "t": 1798.3} |
| {"event": "consolidate", "step": 5700, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1810.4} |
| {"event": "train", "step": 5700, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.3405, "bpb": 1.93394, "chunk_loss": 0.030325, "F": 0.1991, "G": 0.20321, "chunks": 203.88, "bytes_per_chunk": 5.023, "boundary_entropy": 0.07934, "router_noise": 0.0, "grad_norm": 0.7231, "lr": 3.18e-05, "router_trainable": true, "sec_per_step": 0.251, "probe_sec": 0.2171, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.2079, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.02087, "n_shadow": 0, "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": 0.39533, "trunc_abs_frac": 4.60794, "mean_delta_full": -0.1691, "read_hit_rate": {"split": 0.89536}, "mean_gate": {"split": 0.09318}, "n_reads": 10413, "n_cand": 11630, "read_suppressed_frac": 0.0963, "dropout_reads_logged": 10413, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 910.9782104492188, "n_util_obs": 869, "util_count_spearman": -0.03769630695997239}, "probe_overhead_sec_total": 402.081, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009326, "alpha": 0.03, "t": 1810.8} |
| {"event": "consolidate", "step": 5750, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1822.8} |
| {"event": "train", "step": 5750, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.38715, "bpb": 2.00124, "chunk_loss": 0.030623, "F": 0.21188, "G": 0.21459, "chunks": 216.97, "bytes_per_chunk": 4.72, "boundary_entropy": 0.08036, "router_noise": 0.0, "grad_norm": 0.7115, "lr": 3.13e-05, "router_trainable": true, "sec_per_step": 0.245, "probe_sec": 0.0899, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07614, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.00503, "n_shadow": 0, "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": 0.78217, "trunc_abs_frac": 0.62808, "mean_delta_full": -0.10586, "read_hit_rate": {"split": 0.90303}, "mean_gate": {"split": 0.09323}, "n_reads": 10802, "n_cand": 11962, "read_suppressed_frac": 0.1026, "dropout_reads_logged": 10802, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 919.0294189453125, "n_util_obs": 874, "util_count_spearman": -0.039483039974488335}, "probe_overhead_sec_total": 405.507, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009879, "alpha": 0.03, "t": 1823.1} |
| {"event": "consolidate", "step": 5800, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1835.9} |
| {"event": "probe", "step": 5800, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3406.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.195713, "cps": 26.609375, "mean_p": 0.198421, "boundary_entropy": 0.078163, "bpb": 1.868772}, "tr": {"chunks": 3507.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.193725, "cps": 27.398438, "mean_p": 0.188119, "boundary_entropy": 0.091627, "bpb": 2.00393}, "fi": {"chunks": 4341.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.230769, "cps": 33.914062, "mean_p": 0.227272, "boundary_entropy": 0.086462, "bpb": 2.101075}}, "bpb": {"en": 1.86877, "tr": 2.00393, "fi": 2.10108}, "gini_cpc": 0.039819, "gini_cps": 0.055388, "bpic": 5.041052, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.9898392959787196, "fi": 1.179118298026108}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 926.9151611328125, "n_util_obs": 883, "util_count_spearman": -0.03815593814581086}, "t": 1836.7} |
| {"event": "eval", "step": 5800, "phase": 3, "bpb": {"en": 1.88263, "tr": 2.01287, "fi": 2.10814}, "bytes_per_chunk": {"en": 4.874, "tr": 5.724, "fi": 4.592}, "bpb_mean": 2.00121, "frozen_inventory": true, "t": 1838.2} |
| {"event": "train", "step": 5800, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.31993, "bpb": 1.90426, "chunk_loss": 0.030459, "F": 0.2038, "G": 0.20801, "chunks": 208.69, "bytes_per_chunk": 4.907, "boundary_entropy": 0.08253, "router_noise": 0.0, "grad_norm": 0.7178, "lr": 3.08e-05, "router_trainable": true, "sec_per_step": 0.323, "probe_sec": 0.0756, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.06364, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.00685, "n_shadow": 0, "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": 0.28706, "trunc_abs_frac": 1.12362, "mean_delta_full": 0.69049, "read_hit_rate": {"split": 0.90685}, "mean_gate": {"split": 0.09619}, "n_reads": 10690, "n_cand": 11788, "read_suppressed_frac": 0.1, "dropout_reads_logged": 10690, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 927.079345703125, "n_util_obs": 883, "util_count_spearman": -0.03804531756195774}, "probe_overhead_sec_total": 409.052, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.010732, "alpha": 0.03, "t": 1839.2} |
| {"event": "consolidate", "step": 5850, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1852.2} |
| {"event": "train", "step": 5850, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.33535, "bpb": 1.92651, "chunk_loss": 0.030435, "F": 0.20657, "G": 0.21017, "chunks": 211.53, "bytes_per_chunk": 4.841, "boundary_entropy": 0.08037, "router_noise": 0.0, "grad_norm": 0.713, "lr": 3.05e-05, "router_trainable": true, "sec_per_step": 0.264, "probe_sec": 0.0514, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.0419, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.11768, "n_shadow": 0, "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": 0.31975, "trunc_abs_frac": 0.29188, "mean_delta_full": 0.37125, "read_hit_rate": {"split": 0.90433}, "mean_gate": {"split": 0.08823}, "n_reads": 10937, "n_cand": 12094, "read_suppressed_frac": 0.097, "dropout_reads_logged": 10937, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 935.1304931640625, "n_util_obs": 889, "util_count_spearman": -0.03635085138547602}, "probe_overhead_sec_total": 412.526, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.010212, "alpha": 0.03, "t": 1852.4} |
| {"event": "consolidate", "step": 5900, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1864.8} |
| {"event": "train", "step": 5900, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.27516, "bpb": 1.83966, "chunk_loss": 0.030339, "F": 0.20007, "G": 0.20399, "chunks": 204.88, "bytes_per_chunk": 4.998, "boundary_entropy": 0.07954, "router_noise": 0.0, "grad_norm": 0.7102, "lr": 3.02e-05, "router_trainable": true, "sec_per_step": 0.253, "probe_sec": 0.0882, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.07688, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": 0.0536, "n_shadow": 0, "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": 0.43635, "trunc_abs_frac": 0.39531, "mean_delta_full": 0.32414, "read_hit_rate": {"split": 0.90285}, "mean_gate": {"split": 0.09206}, "n_reads": 10529, "n_cand": 11662, "read_suppressed_frac": 0.0985, "dropout_reads_logged": 10529, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 943.2474365234375, "n_util_obs": 897, "util_count_spearman": -0.04613674687001544}, "probe_overhead_sec_total": 415.843, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008574, "alpha": 0.03, "t": 1865.1} |
| {"event": "consolidate", "step": 5950, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1878.4} |
| {"event": "train", "step": 5950, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.37669, "bpb": 1.98615, "chunk_loss": 0.030467, "F": 0.20554, "G": 0.20841, "chunks": 210.47, "bytes_per_chunk": 4.865, "boundary_entropy": 0.0808, "router_noise": 0.0, "grad_norm": 0.7236, "lr": 3.01e-05, "router_trainable": true, "sec_per_step": 0.269, "probe_sec": 0.0358, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.02585, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.04016, "n_shadow": 0, "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": 0.45809, "trunc_abs_frac": 1.19614, "mean_delta_full": -0.10535, "read_hit_rate": {"split": 0.90596}, "mean_gate": {"split": 0.09191}, "n_reads": 10886, "n_cand": 12016, "read_suppressed_frac": 0.1005, "dropout_reads_logged": 10886, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 951.3178100585938, "n_util_obs": 905, "util_count_spearman": -0.042494706568239546}, "probe_overhead_sec_total": 419.325, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.009456, "alpha": 0.03, "t": 1878.6} |
| {"event": "consolidate", "step": 6000, "retired": 0, "split_retired": 0, "merge_admitted": 0, "admitted": 0, "swapped": 0, "uncertain_staged": 0, "t": 1892.1} |
| {"event": "train", "step": 5999, "arch": "ali", "phase": 3, "objective": "global", "lm_loss": 1.37408, "bpb": 1.98238, "chunk_loss": 0.030501, "F": 0.20764, "G": 0.21043, "chunks": 212.62, "bytes_per_chunk": 4.816, "boundary_entropy": 0.0787, "router_noise": 0.0, "grad_norm": 0.685, "lr": 3e-05, "router_trainable": true, "sec_per_step": 0.27, "probe_sec": 0.0418, "probe_extra_seqs": 16, "n_rows": 2, "extra_fwd_seqs": 16, "wall_s": 0.03175, "overhead_frac_est": 0.16667, "chunk_mismatch": 0, "n_loo": 14, "mean_delta_loo": -0.00787, "n_shadow": 0, "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": 0.27926, "trunc_abs_frac": 3.74076, "mean_delta_full": 0.19964, "read_hit_rate": {"split": 0.90108}, "mean_gate": {"split": 0.09425}, "n_reads": 10557, "n_cand": 11716, "read_suppressed_frac": 0.0988, "dropout_reads_logged": 10557, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 959.2438354492188, "n_util_obs": 913, "util_count_spearman": -0.04965040731562304}, "probe_overhead_sec_total": 423.739, "probe_overhead_seq_frac": 0.4999, "util_loss": 0.008625, "alpha": 0.03, "t": 1892.1} |
| {"event": "probe", "step": 6000, "phase": 3, "objective": "global", "router_trainable": true, "per_lang": {"en": {"chunks": 3374.0, "chars": 17403.0, "bytes": 17418.0, "cpc": 0.193875, "cps": 26.359375, "mean_p": 0.197165, "boundary_entropy": 0.077506, "bpb": 1.866558}, "tr": {"chunks": 3502.0, "chars": 18103.0, "bytes": 19742.0, "cpc": 0.193449, "cps": 27.359375, "mean_p": 0.187544, "boundary_entropy": 0.090001, "bpb": 2.000974}, "fi": {"chunks": 4335.0, "chars": 18811.0, "bytes": 19572.0, "cpc": 0.23045, "cps": 33.867188, "mean_p": 0.227321, "boundary_entropy": 0.086311, "bpb": 2.097158}}, "bpb": {"en": 1.86656, "tr": 2.00097, "fi": 2.09716}, "gini_cpc": 0.03993, "gini_cps": 0.057146, "bpic": 5.060387, "worst_lang": "fi", "cpc_parity_vs_en": {"en": 1.0, "tr": 0.99780260231899, "fi": 1.1886562009509187}, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 959.2438354492188, "n_util_obs": 913, "util_count_spearman": -0.04965040731562304}, "t": 1892.9} |
| {"event": "eval", "step": 6000, "phase": 3, "bpb": {"en": 1.88061, "tr": 2.00889, "fi": 2.10391}, "bytes_per_chunk": {"en": 4.896, "tr": 5.727, "fi": 4.601}, "bpb_mean": 1.99781, "frozen_inventory": true, "t": 1894.3} |
| {"event": "reps_dumped", "dirs": ["/local/scratch/sas245/beetle-hnet/runs/ali_mem-split_s2/probe/reps/th0.30", "/local/scratch/sas245/beetle-hnet/runs/ali_mem-split_s2/probe/reps/th0.40", "/local/scratch/sas245/beetle-hnet/runs/ali_mem-split_s2/probe/reps/th0.50", "/local/scratch/sas245/beetle-hnet/runs/ali_mem-split_s2/probe/reps/th0.60", "/local/scratch/sas245/beetle-hnet/runs/ali_mem-split_s2/probe/reps/th0.70"], "t": 1897.3} |
| {"event": "final", "run": "ali_mem-split_s2", "arch": "ali", "objective": "ratio", "condition": "A", "step": 6000, "params": 22785088, "seed": 2, "per_lang": {"en": {"cpc": 0.1938746193185083, "cps": 26.359375, "bpb": 1.8871370572997246, "bytes_per_chunk": 4.841831127693174, "reads_per_chunk": 1.5941929942834978}, "tr": {"cpc": 0.19344859967961112, "cps": 27.359375, "bpb": 2.0003227739356775, "bytes_per_chunk": 5.760646947655254, "reads_per_chunk": 1.5141849426449259}, "fi": {"cpc": 0.23045026845994365, "cps": 33.8671875, "bpb": 2.1228573277366287, "bytes_per_chunk": 4.577415355602508, "reads_per_chunk": 1.6137089874102917}}, "bpb_mean": 2.003439052990677, "gini_cpc": 0.03993013570133663, "gini_cps": 0.057146255166057314, "bpic": 5.06038711979306, "mean_boundary_entropy": 0.07869629561901093, "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": 423.73862648010254, "probe_overhead_seq_frac": 0.49990625, "wall_sec": 1900.4829647541046, "inventory": {"size": 65536, "admitted": 65536, "staged": 0, "budget_B": 65536, "staging_capacity": 0, "collisions": 0, "dropped": 0, "type_rejected": 0, "config_keys": 0, "admitted_surface": 0, "staged_surface": 0, "admitted_split": 65536, "staged_split": 0, "admitted_merge": 0, "staged_merge": 0, "admitted_skel": 0, "staged_skel": 0, "mean_count": 959.2438354492188, "n_util_obs": 913, "util_count_spearman": -0.04965040731562304}, "ali_params": 268096, "ali_config": {"enabled": true, "budget_B": 65536, "staging_capacity": 262144, "policy": "frequency", "value_mode": "learned", "candidate_types": ["split"], "gate": "vector", "retrieval": true, "memory_scope": "persistent", "read_dropout": 0.1, "write_rule": "mean", "eta_min": 0.0, "consolidate_every": 50, "probe_seqs": 2, "probe_reps": 8, "horizon": 32, "predictor": "mlp", "lambda_util": 0.1, "max_span": 64, "byte_classes_path": "/local/scratch/sas245/beetle-hnet/scratch/data/byte_classes.json", "gate_hidden": 0, "type_dim": 16, "gate_bias_init": -2.0}, "util_vs_frequency_spearman": -0.04965040731562304, "paired_vs_dropout": {"n": 1788, "pearson": -0.0030876698017904647, "spearman": 0.011169164198972185, "mean_abs_discrepancy": 0.6378756761550903, "mean_paired": 0.023726264014840126, "mean_dropout": 0.01686469279229641}, "predictor_calibration": {"n": 4068, "all": {"n": 4068, "pearson": 0.0846433533068819, "spearman": 0.06590938152129017, "mae": 0.08614727854728699, "reliability": [{"pred": -0.0006287364521995187, "obs": 0.004920631647109985, "n": 407}, {"pred": 0.008317101746797562, "obs": 0.006895688362419605, "n": 407}, {"pred": 0.011855889111757278, "obs": 0.01249079778790474, "n": 407}, {"pred": 0.014004413969814777, "obs": 0.024222392588853836, "n": 406}, {"pred": 0.01570117473602295, "obs": 0.032149795442819595, "n": 407}, {"pred": 0.017101813107728958, "obs": 0.01511900033801794, "n": 407}, {"pred": 0.018620165064930916, "obs": 0.029573889449238777, "n": 406}, {"pred": 0.02080720290541649, "obs": 0.0296308696269989, "n": 407}, {"pred": 0.023795345798134804, "obs": 0.031094778329133987, "n": 407}, {"pred": 0.03035973384976387, "obs": 0.0438702292740345, "n": 407}]}, "admitted": {"n": 4068, "pearson": 0.0846433533068819, "spearman": 0.06590938152129017, "mae": 0.08614727854728699, "reliability": [{"pred": -0.0006287364521995187, "obs": 0.004920631647109985, "n": 407}, {"pred": 0.008317101746797562, "obs": 0.006895688362419605, "n": 407}, {"pred": 0.011855889111757278, "obs": 0.01249079778790474, "n": 407}, {"pred": 0.014004413969814777, "obs": 0.024222392588853836, "n": 406}, {"pred": 0.01570117473602295, "obs": 0.032149795442819595, "n": 407}, {"pred": 0.017101813107728958, "obs": 0.01511900033801794, "n": 407}, {"pred": 0.018620165064930916, "obs": 0.029573889449238777, "n": 406}, {"pred": 0.02080720290541649, "obs": 0.0296308696269989, "n": 407}, {"pred": 0.023795345798134804, "obs": 0.031094778329133987, "n": 407}, {"pred": 0.03035973384976387, "obs": 0.0438702292740345, "n": 407}]}, "staged": {"n": 0, "pearson": NaN, "spearman": NaN, "mae": NaN}}, "t": 1900.6} |
| {"event": "hub_done", "repo": "AdaptiveChunking/ali_mem-split_s2", "pushed": [0, 1, 2, 4, 6, 8, 16, 32, 64, 128, 256, 512, 1000, 1024, 2048, 3000, 4096, 6000], "failed": [], "ok": true, "t": 1917.1} |
|
|
